P<
if

UNITED NATIONS & é .
“J J vv t ~ \
5 FOUN DATION acoatl Ni sll

ay

a ™,
R 4 %
4 —~Z =

‘RY FROM THE AMERICAN PEOPLE
Sat, ot

Mesoamerican Reef Alliance, ICRAN-MAR Project

Land use change modelling for
three scenarios for the MAR region

Technical Report

Technical report on the collection of geographic data, the regression
analysis of explanatory factors of land use patterns, the development of a
set of three alternative scenarios, and the modelling of land use changes
using the CLUE-S model. This work was carried out as part of the
ICRAN-MAR project’s sub-result 1.2, “Trends in land use integrated with
spatial, hydrological and oceanographic models for use in modelling”.

Joep Luijten, Lera Miles and Emil Cherrington
UNEP World Conservation Monitoring Centre

5 October 2006

@) @

UNEP WCMC

This report was made possible through support provided by the office of Guatemala-
Central American Programs, Latin America and Caribbean Bureau, U.S. Agency for
International Development, under the terms of Grant No.596-G-00-03-00215-00. The
opinions expressed herein are those of the author(s) and not necessarily reflect the
views of the U.S. Agency for International Development or of UNEP.

INUIT XO OZ

Table of contents

Bistiof tables Aictirecccscecctcestncsesesccceesctsxsssscecerertreeravestce-coreererevraecessaceetttrertesssossecernetse: Vv
ISH OF FIG UNOS occ: sii sseccovintccevs sassuedccactecerucsccecsucazacerecstecteeioutueterseucecressectaisereecescessses Vii
SU LOIIINE LN caccerececeacecece cas 2600 EC con 00700000000 1012000 0C000 9000000000 6c OcoCEASS opoo gas cS DaoC TB USG Soo 1
1 Data collection and preparation ................cccessseseeeeeseseeesseseeeeeeeeesesnesecseeeeeeeseess 2
1.1 Outline of methodology and preparation Steps .......... eee eeeeeeeeeeeeeeeeeteeesetreeeeenes 2
1.1.1 Stage 1: Creation of ASCII grids with identical number of value cells................ 2
elle Stage 2: Conversion to a text file for use by SPSS regression module.............. 3

de2) (Grid’extentiandignidinesoluitiompsercsttecceeesee se toneeceeee ese cere eraeeeseseeeeeeecoaees arc eeceeeaseser 3
1.2.1 Creation of watershed boundaries shapefile .................0..cccccsccccccceeessseeeeeeeeeaes 3
te2-2 ConversionitomastemimaskS reece cscs crete career ee ee 4

d23 sEandiuse/land/covenclassificatiom een seccaes sector cao 5
1.3.1 Reduced number of land cover ClASSES..............ccccccccccceeeseteceeeeeeeestteceeeeessesaaeees 5
123425 SOUrCESION land (COVE Catal erneeee cee terttacceseaceees eter eceee sean nse ane ene 6
(23:3 » (Outputvextentrandicellisize see teeter eee rete eee cect caeecsent esc stneeauneeep cess 7
1.3.4 Reclassification methodology using ArCGIS ...............ccccccesccceceessssseeseeeceesssseeees 7
(E325 Calculation of area by country and land Cover type...............:::ccceccecesseeeeeseeeeeees 9
1.3.6 FandicovermfonuseliniN-sRE GM ccsecscecec cr csre ee eoteeese eee ae 10

1.4 Explanatory factors of land use patterns..............cceeccceeceseeceeteeeeteeeeeneeeesseeeeeaes 10
1.4.1 Topographic factors - elevation and SlOPeC.................:cccccccceeessseceeeeesestseeeeeeeenees 10
1.4.2 Demographic factors — population density...................cccceccssccceceeesssteeeeeeeeesstaes 13
1.4.3 Demographic factors — location of settlements ................cccccccceseeeeetteeeestseeeeeaes 13
14:4 “Soiltandi geology factors) sees see eee Bree cee reer oot mate 14
1.4.5 Climate factors - precipitation and length of dry season..................c..cccceseceeees 15
1.4.6 Contextual factors — protected Areas ........... cc ccccccccecstcccceeccessssteeceeesstsseeeeeeeens 17
1.4.7 Contextual factors — access to roads and markets. ..............00::ccccccccessseeeesseeeees 18
1.4.8 Contextual factors — tourist hotspots and areas of coastal development ......... 21

2 Analysis of drivers of land use Chang..............:ccsseccccesssceceesseeeeeseseeeeeessneeees 22
2.1. Land Use Change Adjacent to the Mesoamerican Reef................:::ccccccesseesereeseees 22
2.1.1 ES {eXo Tic aasceceanectodtecédonascderac descshAans bacestecbenc an che caHaaBetne chica craceadaaceectHestsdadasHaadasas 22
2.1.2 BGlIZC i ists cecsse:ancanctsneauite pecsccAeeee Pee ee ee ee LN. Or aN 23

7 sea (ire elim © (UF) (0 | eee Areaeerticr onc esonbacco seco penne ec ence acosccaroncodedecs scbarencuemrceracrecancensececnne 25
2.1.4 PROM (U Le (Seas ceendeeeegannsee iccronccodhedadnete cick acd ches iaeina doen tne eRnerer ene emer iced oa 26

2A De ee REGIONAL SVMUNS SIS ir. «ees sete ee on eect cat eect Sasa coe sac Re aa hot A 2 26

2.2 Statistical analysis of explanatory factors for land use patterns ................:ccceee 27
2.2.1 Method log yee erste reece re ee aN een rere ane eee eee ane 27
2.2.2 Evaluating statistical significance and goodness Of fit ................ceeceeeeeeeeteees 28
2.2.3 REgressioniresultse ys eee ek AE See ies 8 ee a Cee 29

3 GEO-4 scenarios and the ICRAN MAR project..............:::csssssccsssseceeeesseeeeeees 34
3. HO ESCENANIOVSUIMIMANISS Eee crtececaceae secre eee ace eee aaa cae een eed ces douaaes avetuevaaee 34
Sala AbouititherG EOz4i:Scemanlosy sre pease cseettcsstccee sees ee seems ieee sence eee eee es < 34
3.1.2 MarketsiFinst:s-eess3-025 fee et oe ate baie oe ates ee ei cere a le oy te eS 35
3.1.3 POlicyaR iste Fe ee Re RI Ie TEA SNS DI Re tah ve ented 37

Sah STS ETI OTT? [TLS scopae ce eacooeeo0eoascosoocaccecoseaossencarseanadadeascdee conoboapo00 san daGecCacoT Ieper 38

3.2 Population and land cover change: comparisons between scenarios .................-++- 39
3.3. Land cover quantification: the IFs and IMAGE models.................::c:cseceeteteeeeees 44
3.3.1 ModellbaCkQnounndcexncccen ene eects eters eee encaslsoisseeenenessicsensexsenaneermeen asso 44
3.3.2 Bringing the models together: Methods ...............ccce cece ceeeeeeeeeeee tte eeteeeeneeeeenes 46

3.4 Future changes in protected Areas 0.0... ccc cee eee ereeeeeneeeetteeeetneeeeesraeereees 48

4 Modelling land use changes for scenarios using CLUE-CS............::::::eeceeee 52
4.1 Important aspects of model development .............. ec cccee ee ceeeeeeeteeeeteteetereeserenenes 52
4.1.1 (EFT AC LAUISEY CEVE occocescedecaadcon oe abacuoneonesecoucaqq000sdeedeonddaauBa6ae Ansa DB DSSBee Bie sbeBaaaceDoTasHaaadsC 52
4.1.2 ProbabilityiSunfacesmtsstseesecrsstrceectetce ree eee cctenes tat ovocsers cecsescecerersertensst 52
4i1.3\ Neighbourhood Settings... cease ccte cece esac nnncscesansessancesntescnesenteesncetrerrssncnes 53

4e2 ae I MPUted ata PNe pW alatiOMpecrcccsenesteteceeere tee eset eee eel eiscctateacetls.a7.2csccceceuareeeoe once ose=n 53
4.2.1 Files used) bylC EW Bes ie reece eceeteer tie erence eens aches ssutsgasassussedeecetarentacesl 53
4.2.2 Land use requirements for different SCENAMIOS ...............ecceeeceeeeeeeteeeee eet ttttteeees 54
4.2.3 Mainimodeliparametens teensex cre cestecstee tates eee rect nas -oo.cneeoarteerearacesste= 56
4.2.4 ReEgnreSSiom|panaime@tenSte: ts seseecces cee sstae eects sctesscscassesssecsusceueesenacouneeeaees 57
4.2.5 Conversionielasticnticspeeseserrctrenccererersce eee teria nateasess ss. es.sssoenaacnacessennnemea 59
CNP > (GOLETRSTOM ENO) scecoscocc eco docoedcsu0cdsnassd6bantivaSdeadoo neaaAarcanSac seco oeBEASERE badosnaAbodocsaod 60
4.2.7 Dynamic location factor grids: Protected Areas .................:ccceeeeeeteet eterno 62
4.2.8 Dealing with absence of pine forest in Me@XICO................cccceeeeeeceeeeeeeeeteeeeeeees 62

CMs STIMULI Men RESIS eicesenececseeso cdc sacesocbocteoiatiadcoosebeohodbe bared sae Hee aaa eoopecEreBBEEE eee oaseBasoaScouD 63
4.3.1 Simulated changes in land use and in forest COVED ..............::ccsceeeeeeeeeeeeeeeeeee 63
4.3.2 Average and maximum deviation Of SOIUtION «2.2.0.0... cece c ccc ceteeeeeteeeeenenees 63

5 Workshop, conclusions and recommendations ..............::ccccssesecssssseeeessseeees 72
Hedi p MeCHMicall| WONKSIMNOPicecascercccce ee ceet eee eee eects coea Sac cce ws duteee ce vocetsMesvex ses auetnceusecusales 72
5.2 Conclusions and Recommendations ................::cccceceeeeeeeeesreeeeeeeeesneeeeenseeeesnneeeee 72
5.2.1 Application of CLUE-S model to the MAR region ...............ccceeeeeeeeeeereeeee 72
5i2e2ee VVOTKShOpLandhthalmim Giiecetmentte tee ecetesestisce cere: --cocsssseercceresess act tevin .ssesacnaseett 73
GIMIRETERCM COS secccccrreecceerertrrrtt recente detttat enna sceccecertstassascacriststostsiccresstsucossccoateersnss 74
MMA DDCIMGI COS secrcecscceceerc eceecen-cceceascactnseccecmesnacetersscccveressacscavecucsddvenetesssccererssacacers 78
Appendix 1. Avenue script for NoData filling and filtering ..............0ce eee eeeeeeeeeneeeeeees 78
Appendix 2. Avenue script for creating dynamic protected areas gridS......0....0.. eee 82
Appendix 3: Complete list of available spatial data... eee ceeeceeeeteeeeeteeeeeteeeees 85
Appendix 4: Ecosystem Map land cover classification ........00...0ecceecceeseeeseeesseetseeecseeeseees 93
Appendix 5. Land use requirements for future SCEMAMIOS «00.02... eee ceteeeeeeteeeeetseeeeees 94
Appendix 6. CLUE-S Training Package (EXerciSes)............0..ccccceeesceeeseeesseeeesesseeeesseeeees 110

List of tables

Table 1-1: Spatial extents for the raster datasets, by country. The coordinates are based on Universal
Transverse Mercator (UTM) projection for zone 16 with the NAD 1927 Central American datum. 4

Table 1-2: Land use classes used for the land use change modelling. The original Ecosystem Map

dataset had a more detailed classification that WaS TEAUCEO, ............cccccccscscccscsescsesesssesessseevevevecees 5
Table 1-3: Total area (km*) of each land cover type in the reclassified and rasterized Ecosystem map

data (final version 4 created on 6" Febriary’ 2006) tlt.n2) semeetcoterrete | meee tmens S) ne nse sh kT meats 6
Table 1-4: Reclassification table for the 2003 Ecosystem map using field DESCRIPTIO...........00..000.... 8
Table 1-5: Reclassification table for the 2004 Belize Ecosystem map using field ECOSYSTEM .......... 8

Table 1-6: Potential explanatory factors that will be included in the regression analysis and simulated
of land use changes. The number (#) has been used for numbering of the CLUE-S regression
results parameter files and therefore starts at 0. Cost of access to roads was eventually left out
from the analysis because it is strongly correlated to cost of access to markets. There weare no

categoricaliexplanatonysfactorsm tee ee ee: 12
Table 1-7: Grid reclassification (resampling) scheme for the number of dry MonthS................0.000002-+- 16
Table 1-8: Prevailing designation types of WDPA areas in Mexico, Guatemala, Honduras and Belize.

SODA sth sccesetodies sas vacteas Doves seats Sune stee ts Sethe coer T eT TR oe Ee RE nn ae 18

Table 1-9: Friction values for land cover with 250 m grid cells. On land cover, average walking speed
was estimated at 4km/hr, but reduced to 3 km/hr in forest and increase to 5 km/hr in urban areas.
zal sain saieaaiss:htadecanessansereansr sic ctccecee reer tet res Cotes a eer 20

Table 1-10: Friction values for different road type with 250 m Tid COIS. ............ccccccccsesececseseeseseeeeeeseees 20

Table 1-11: Friction multipliers for slope. There is no accounting for slope direction; it is assumed that
travelling both up-slope and down-slope incurs a reduction in travel SPC@O...........0..0ccccccsececseeeeee 21

Table 2-1: Summary of the logic regression analysis for Belize. For each dynamic land use, the
regression coefficients for all statistically significant explanatory location factors are listed, with
the four most significant ones in bold. Note that the absolute value of a regression coefficient is
no indicator of its level of significance, so even relatively small values may be in the top four. ... 30

Table 2-2: Summary of the logic regression analysis for Mexico. For each dynamic land use, the
regression coefficients for all statistically significant explanatory location factors are listed, with
the four most significant ones in bold. Note that the absolute value of a regression coefficient is
no indicator of its level of significance, so even relatively small values may be in the top four. ... 31

Table 2-3: Summary of the logic regression analysis for Guatemala. For each dynamic land use, the
regression coefficients for all statistically significant explanatory location factors are listed, with
the four most significant ones in bold. Note that the absolute value of a regression coefficient is
no indicator of its level of significance, so even relatively small values may be in the top four. ... 32

Table 2-4: Summary of the logic regression analysis for Honduras. For each dynamic land use, the
regression coefficients for all statistically significant explanatory location factors are listed, with
the four most significant ones in bold. Note that the absolute value of a regression coefficient is
no indicator of its level of significance, so even relatively small values may be in the top four.... 33

Table 3-1: Forest cover, Change DY SCONAMO)............2..c-ssc20cccecceerecctssecsescoensocsssansosscesonesazeesssvastasessieesiss 40
Table 3-2: Mapping of land cover types between IFs, IMAGE and CLUEGCS ....0.........ccccccccccceseccseceeeeeee 46

Table 4-1: Input files used by CLUE-S. The “created” column indicates which software is used to
create the files and the “mandatory” column indicates whether the file is a minimum input data
requirement. Files created using CLUE-S are plain text files and may also be edited in a text

OGM OFF so sxe sso Secnszca coasters coos See toes AE eT EE LE OTE OI 53
Table 4-2: Belize: Distribution of present land use and land demand for the scenarios. Blue coloured
land use types were kept fixed at present values and not allowed to change over time. ............. 54

Table 4-3: Mexico: Distribution of present land use and land demand for the scenarios. Blue coloured
land use types were kept fixed at present values and not allowed to change over time. Note that
there is no pine forest in Mexico; this required some adjustments to the model. ......................-.- 54

Table 4-4: Guatemala: Distribution of present land use and land demand for the scenarios. Blue
coloured land use types were kept fixed at present values and not allowed to change over time.
The area savanna is very small but not exactly zero (the distinction is significant)...................... 55

Table 4-5: Honduras: Distribution of present land use and land demand for the scenarios.
Blue=forced fixed at initial area, not allowed to Changed OVEF tM. ........2..:c:ccccccecceereceeeteeeeenees 55

Table 4-6: Present land use distribution with red coloured values for those types for which the demand
was kept constant over time because the demand changes were smaller than the iteration
tolerance of CLUE-S, or so small that the model was prevented from reaching a solution. The
required change in land was added to another land use type, as indicated within parenthesis. .. 56

Table 4-7a: Main model parameters as used for the SiMUIALIONS. 0.2.2.0... cccccce cette ette etre et tteeeeeteeeees 56
Table 4-8: Default conversion elasticities for the land USE tYP@S. .........ccccccccccceteeteeeneteneeeeeserseeneeennes 60
Table 4-9: default conversion matrix. Note that some adjustments had to be made for all countries to
allow for sufficient change options, as indicated in blue in the next four tables. ...........-...:0.000 6
Table 4-10: Modified conversion matrix for Belize conversion. The medium grey coloured rows and/or
columns are associated with land use types that were kept constant and did not change........... 61
Table 4-11: modified conversion matrix for Mexico. The medium grey coloured rows and/or columns
are associated with land use types that were kept constant and did not change. ................:0006 61
Table 4-12: modified conversion matrix for Honduras. The medium grey coloured rows and/or
columns are associated with land use types that were kept constant and did not change........... 61
Table 4-13: modified conversion matrix for Guatemala. The medium grey coloured rows and/or
columns are associated with land use types that were kept constant and did not change........... 62

Table 4-15: mean and maximum deviation between demand and allocated land use, in percentage of
absolute area, for land use in the final simulated year, 2025 ). These statistics are calculated for
every simulated year but presented here only for the final year). The maximums (2”7 and gia
columns) are specified in the main parameter file and are slight adjustments from the default
settings in CLUE-S, respectively, 0.35% and 3.0%. In almost all cases the highest deviation
applies to land use that occupies the least area and is not kept constant, which almost always is
Ug OEY eR el aaa net ect i a A i a Pe colar C SSSA SERRE EERE PEER CEE EEE ee RE cB nccodied 63

vi

List of figures

Figure 1-1: Spatial extents and data area for raster datasets for the counties within the MAR region... 4
Figure 1-2: Rasterization of the Ecosystem map vector data on linked field NUM for the 2003

Ecosystem Map data (left) and the 2004 Belize Ecosystem Map (right) ...........:cccccccccecscetecsseeees 9
Figure 1-3: SRTM tile numbers that were downloaded. Tile 20_10 was included as the earlier versions

of the watersheds boundaries indicated that it extended more to the east and southeast. .......... 11
Figure 3-1: Role of three models used to simulate land COVEr CHANGE .............c:ccccecseesseeeeteteeeteeeesees 40
Figure 3-2: National human population at 2005 and 2025 by SCeNariOS (IFS) ...........ccccccceceteeeteeteeetees 41
Figure 3-3: Percentage change in national populations, 2000 to 2025, by scenario (IFS).................... 42
Figure 3-4: Change in land cover, 2005 to 2025, all countrieS COMDINGC..............00..0cccccccccsceseeseesseceees 42
Figure 3-5: Percentage change in land cover, 2005 to 2025, all countries combined.....................00++- 43
Figure 3-6: Land cover for watershed area at 2005; all COUNtrIeES COMBINE .................6cccccccctseeeeeeeeees 43
Figure 3-7: Land cover for watershed area at 2025 by scenarios; all countries combined.................-. 44

Figure 3-8: Land cover at baseline year (2004 for Belize, 2000 for Guatemala, Honduras and eri

Figure 3-9: Change in land cover for Belize, 2005 to 2025, scenarios

Figure 3-10: Change in land cover for Guatemala, 2005 to 2025, SC@NAMIOS ...........ccccccseseeeeeesteeseee 49
Figure 3-11: Change in land cover for Honduras, 2005 to 2025, SC@NALMIOS ............:cccccccceeeeeeteeeeeeeees 50
Figure 3-12: Change in land cover for Mexico, 2005 to 2025, SC@ENAMIOS.............0ccccceeeeetteeteeeeeeteetsees 50
Figure 4-1: Present land cover and simulated land cover for the three scenarios in 2026................... 64
Figure 4-2: Baseline (2000/2004) land USC .........cccccccccccsececsccccescceescecesseesessescssesesscecsssesensseeesseessseeusses 65
Figure 4-3: Simulated land cover for scenario 1, Markets FirSt, iN 2025..........0..cccccccccccsssceessesesseeeeeses 66
Figure 4-4: Simulated land cover for scenario 2, Policy First, if 2025 ............ccccccccscccsscssseesseesseesseesseees 67
Figure 4-5: Simulated land cover for scenario 4, Sustainability First, 19 2025 .............cccccccccessessseeesees 68
Figure 4-6: Simulated areas of change with 2025 land cover for scenario 1, Markets First................. 69
Figure 4-7: Simulated area of change with 2025 land cover for scenario 2, Policy FirSt............:..:000 70
Figure 4-8: Simulated areas of change with 2025 land cover for scenario 4, Sustainability First......... Tal

Figure 1: Spatial extents and mask for raster datasets for the four MAR counties. Coordinates are in
UTM zone 16 with NAD 1927 Central American Gatum. ............::ccccceceeccceeseeetceesseeeseeeenseeenes 134

Vil

Summary

Mesoamerica — the region in which the Mesoamerican Barrier Reef Systems fall — is
recognized internationally for its biodiversity. For example, Conservation International has
identified the area as a biodiversity hotspot, with a high proportion of endemic species
(Myers et a/. 2000). The area’s natural ecosystems are also recognized to be threatened.
The World Bank-funded Central America Ecosystems Mapping Project, which concluded in
2002, estimated that 49% of Central American land had been converted to agriculture
(Vreugdenhil et a/. 2002).

With a focus on the Mesoamerican Reef, the International Coral Reef Action Network’s
Mesoamerican Reef Alliance (ICRAN-MAR) project is focusing its attention on how changing
land use affects the health of the region’s reef ecosystems. The project region includes
southern Mexico, and all of Belize, Guatemala, and Honduras.

This report details the steps undertaken to map current and potential future land cover for
this ICRAN MAR region. Geographic data was collated, three alternative land cover
scenarios for 2005 to 2025 were developed, a regression analysis was undertaken to identify
the strength of different factors affecting land use patterns and land use changes under these
scenarios were modelled.

The land cover maps for the present day and for 2025 were used as a key input to a
hydrological model of watersheds discharging adjacent to the Mesoamerican Reef, prepared
by the World Resources Institute (WRI). A hydrologic modelling report is also available on
this CD.

A workshop was held in August 2006 to disseminate project results and to provide training in
the use of the models. A preliminary version of this report was distributed to workshop
participants.

1 Data collection and preparation

1.1 Outline of methodology and preparation steps

To identify drivers of deforestation, a regression analysis was undertaken in SPSS. The
method involves a comparison of land use with the explanatory factors on a cell-by-cell basis
within a raster map. Consequently, it is important that all raster data associated with the
explanatory factors are prepared consistently: all raster maps must have exactly the same
extent, same cell size, and the same numbers of grid cells that are not Null (NoData). A
difference of just one cell will cause an offset in the order in which the statistical analysis are
carried out and results will be meaningless.

To assure consistency across the raster inputs, the same preparation and conversion
procedure was applied to every dataset. The data preparation involves two stages as follows.

1.1.1 Stage 1: Creation of ASCII grids with identical number of value cells

Stage 1 involves the creation of the raster datasets in Arc/Info ASCII format so that they can
be used (i) by the CLUE-S model and (ii) for the subsequent Stage 2 processing steps.

1. Identify and acquire the best available and most suitable data, in vector or raster
formats. Different data from different sources will be used.

2. Review the quality of the dataset, and edit the dataset to resolve any data errors or
other problems (areas with missing data; non-adjacent polygons; misclassification of
data). If necessary reclassify the data into a more appropriate system.

3. Create any derived datasets, if applicable. An example of this is the creation of a
dataset for the number of dry months from monthly precipitation data.

4. Convert vector data or resample raster data to the same raster grid resolution and
spatial extent (see Section 1.2).

5. Apply a focal mean filter (continuous data) or a focal majority filter (categorical data)
to fill any occasional Null cells' and “add a few grid cells width” of data on the edges
of the maps. This critical step ensures that when data are clipped in the next step,
there are absolutely no Null cells within the watershed boundaries. An Avenue script
was developed for use in ArcView 3.3 (Appendix 1).

6. Clip all rasters to the MAR extent, and then clip them further to the individual extents
of the countries (Table 1-1). This step can be carried out using the Raster Calculator
in ArcMap.

7. Export all data from GRID to ASCII text format. This can be carried out using the
conversion tools in ArcToolBox (Conversion Tools > From Raster > Raster to ASCII?)

' |The conversion of vector data to raster data sometimes results in Null cells wnere they would not be expected.
This reason for this appears to be non-adjacency of polygons in the vector data. Grid cells are assigned as Null
when their centre points fall in the empty area between the two polygons.

? Step 7 — 10 required several Gigabytes of disk space because the ASCIl files were quite large and there were
many of them.

1.1.2 Stage 2: Conversion to a text file for use by SPSS regression module

Stage 2 involves the further processing of the output datasets from stage 2 into a number of
different formats to obtain plain text files that can be imported by SPSS. The CLUE-S user
manual and exercises (Verburg 2004, Verburg et a/. 2004), offer a more detailed explanation.

8. Separate grids must be created for every land cover type because binary logistic
regression analysis is used. This can be carried out using the Raster Calculator. For
the ICRAN MAR region, there were 4 countries * 10 land use types = 40 different
grids. Each grid is then converted to ASCII format, as in step 7.

9. Using the File Converter program that is supplied with CLUE-S, convert the ASCII
grids to text files in which all raster values are listed in a single column, with no
header. This must be undertaken for all land use types and all explanatory factors,
creating a large number of files. For example, for 10 land use types and 15
explanatory factors, there are 4*(10+15) = 100 single-column files. A consistent file
naming convention should be used to avoid confusion and mistakes.

10. Copy the contents of the single-column files into an overall file that can be loaded in
SPSS (this file is called stats.txt by the CLUE-S File Converter). The total number of
columns in this file must equal the sum of the number of land use types and the total
number of explanatory factors. This file was created using the TextPad text editor (the
option to create this file using the CLUE-S File Converter resulted in a runtime error,
possibly as a result of the large grid size. Record the order of the data columns.

1.2 Grid extent and grid resolution

1.2.1 Creation of watershed boundaries shapefile

WRI provided a base watershed boundaries shapefile. This illustrates that not all the
watersheds in the four MAR countries drain to and have a direct impact on the
Mesoamerican Reef’, and is a vital component in analysing the impacts of land cover change
on the reef system. A version delineated from the 90 m DEM was completed on 4 August
2005 and a version based on the 250 m DEM on 24 January 2006. Neither shapefile was
readily usable in this exercise because WRI had removed watersheds less then 80 ha in
size. This had resulted in an erratic boundary that did not correctly represent the water/land
boundary. Furthermore, to retain flexibility in the final resolution used for modelling, it was
considered undesirable to restrict the boundaries to a particular DEM extent.

Several edits were carried out to create an improved and more flexible boundaries shapefile
for preparation of data for the regression analysis and the land use modelling. The overall
area of WRI’s 90 m and 250 m shapefiles (for inland boundaries) was combined it with the
best land/water/country boundary shapefile (/and_country_2Ojuly05.shp, used for the mask’s
coastline). Next, the combined shapefile was improved in January 2006 by extensively
editing the coastline of Mexico and Honduras so that it better matched the coastline from the
Ecosystem map and the Landsat TM colour composites. The final MAR watershed shapefile
MAR_BASIN 3B RECLASSMASK_5FEBO6.SHP was Created.

The shapefile was converted to a raster at 250 m resolution as BASIN250. This raster has
NoData values outside the catchment area and has four different grid values: 1 for Mexico, 2

"GIS analysis (using the WRI watershed delineations and the administrative boundaries provided by CCAD)
reveals that all of Belize’s six districts, fourteen of Guatemala’s twenty-two departments, sixteen of Honduras’
eighteen departments and three of Mexico's thirty-two states possess lands in the hundred or so watersheds
draining to the reef.

for Belize, 3 for Guatemala and 4 for Honduras. These values are used later on in the
modelling process.

1.2.2 Conversion to raster masks

As mentioned above, it is critically important that all input data associated with the land use
and the explanatory factors are prepared consistently, meaning that all grids must have the
same extent, cell size and NoData area. A difference of just a single cell will render the
results of the statistical analysis meaningless.

In cooperation with the hydrological modeller, a grid cell size of 250 m was chosen’. The
extent of the grids for the MAR watershed and every country is given in Figure 1-1.

Figure 1-1: Spatial extents and data area for raster datasets for the counties within the MAR region.

The regression analysis was carried out at 250 m. It should be noted that explanatory factors
of land use changes can be scale dependent. That is, certain spatial relationships that may
be observed (i.e., are statistically significant) at a certain scale, but may be less or not all
significant at other scales. However, Kok and Veldkamp (2001) and Kok (2004) have
concluded that changing the spatial resolution does not lead to major changes in the set of
variables composing the equation that explain land use patterns in Central America.

Table 1-1: Spatial extents for the raster datasets, by country. The coordinates are based on Universal
Transverse Mercator (UTM) projection for zone 16 with the NAD 1927 Central American datum.

AllofMAR | 40000 | 794000 | 1519000 | 2 390 000

|
Belize 261 500 | 412500 | 1 757 500 | 2 045 250 604
41 250 | 368 750
260 250 | 793.000 | 1521 000

# Columns

3 046 407

349 762

542 309
1 267 903

‘A minimum polygon size of about 150 ha was applied during the creation of the 2003 Central American
Ecosystem map, and a minimum of 10 ha was used for the more detailed 2004 Ecosystem Map for Belize. A
resolution of 250 m (6.25 ha grid cell) is thus small enough to preserve the data resolution.

From the raster BASIN250, a separate raster mask was created for each of the four countries.
This involved three steps:
1. Set the appropriate analysis extent and cell size in the Spatial Analyst options menu.
Use the values as specified in Table 1-1.

2. Use the raster calculator and the expressions below:

omnes

For BZ, Con (
For GT, ia aes 250] )
For HN, Con([basin250] == 4, 0,

3. Save the output of the raster calculator permanently, using the names: MASK_MX_250,
MASK _BZ 250, MASK_GT_250 and MASK_HN_ 250. Each of these grids only | has zero
values and can be used as analysis mask for further data preparation.

1.3 Land usel/land cover classification

1.3.1 Reduced number of land cover classes

A reduced land cover classification with ten classes (Table 1-2) was developed for use by the
CLUE-S land use model and the scenario analysis. The need for such a classification was
outlined early in the project and a proposed classification in principle agreed upon during a
conference call on 16 September 2005. The dataset was derived from the 2003 Ecosystem
Map dataset for Central America and the 2004 update for Belize. Appendix 4 gives the
legend used for the original and reduced classifications.

Table 1-2: Land use classes used for the land use change modelling. The original Ecosystem Map
dataset had a more detailed classification that was reduced.

[ 0 — Other/Unknown 5 — Savanna |
| 1 — Broad-leaved forest 6 — Wetland/Swamp |
2 — Pine forest 7 — Mangroves |
eee 8 — Urbanized |
— Scrub 9 — Water |

The ten land use classes represent different production systems that are distinctly different in
terms of (i) natural and spectral characteristics, (ii) relevant national policies and key drivers
of land use change in the past and future, and (iii) management practices and possible
changes in those practices as they relate to the overall objective of the project.

The proposed classification changed over time, during a total of four revisions:

e In September, a seven-class system was proposed: Forest, Pasture, Scrub,
Cropland/Agriculture, Wetland, Savanna, and Other (includes urban, water bodies).

e During the 16 Sept 2005 conference call we agreed that mangroves should be added as
a separate class and that forest should be split in two forest types (broad-leaved and pine
forest). This brought the total to nine classes.

' The “Other/Unknown” land use class includes any land cover types that cannot be reclassified as any other
types. For the scenario simulations it is assumed that “Other/Unknown” remains constant over time (i.e., neither
the total area, nor the spatial distribution changes over time. The Other and Water classes were not included in
the statistical analysis of land use factors and the associated areas were not changed by the CLUE-S model.

e During the creation of the first reclassified raster, it was noticed that neither the 2003
Ecosystem Map nor the 2004 Belize Ecosystem map contained pasture as a separate
category. Pasture may be included within the agricultural land class. Pasture was
therefore dropped from the classification, resulting in a total of eight classes.

e Having reviewed the first reclassified dataset, Lauretta Burke suggested that urban and
water should be included as two separate classes rather than be grouped in the “other”
class. The final version is therefore composed of ten classes.

Table 1-3: Total area (km’) of each land cover type in the reclassified and rasterized Ecosystem map
data (final version 4 created on 6"” February 2006)

| Mexico Belize Guatemala Honduras
| 0. Other/Unknown 251.7 13.5 8.3 232.9
| 1. Broad-leaved forest 31 760.9 12 684.2 17 322.6 | 20 555.3 |
| 2. Pine forest livia 0.0" | 771.8 840.4 | 12 198.9
3. Agriculture/pasture 3 398.0 4235.1 10 505.9 | 43 720.4
4. Scrub 14 990.6 274.8 151.6
5. Savanna 62.3 1 886.4
6. Wetland/Swamp 1921.0 931.8
7. Mangroves 2316.8 | 720.0
8. Urban 145.4 189.4
9. Water 153.2
54 402.06 21 860.13 33 894.31 79 243.94

1.3.2 Sources of land cover data

For Mexico, Honduras and Guatemala, data were derived from the revised 2003 Ecosystem
Map. For Belize: data were derived from the revised 2004 Ecosystem map for Belize.

Both datasets contain a mangrove class. Emil Cherrington shared a separate mangrove
dataset for Belize that is arguably more up-to-date. While this dataset appears more detailed
(there are many more smaller polygons), it does not include all the mangrove areas within
the 2004 Ecosystem map. As substituting the mangroves from the 2004 Ecosystem map with
the improved mangrove data would result in data gaps, for which the land cover is unknown,
this has not been undertaken.

The 2003 Ecosystem map had various data quality problems, in particular non-adjacent
polygons along the Belize/Mexico and Belize/Guatemala border and in locations where rivers
form national boundaries. For example, an area of about 75 km long and just 400 m wide
along the straight border was not classified. This resulted in some visible reclassification
errors and gaps in the first version of the reclassified land cover raster.

The 2003 Ecosystem map was extensively edited to fix these errors and improve polygon
adjacency with the 2004 Belize ecosystem data (which was not edited). After review, it
appeared that the second reclassified land dataset still contained errors, mostly in the form of
single NoData cells. The original 2003 Ecosystem map was extensively edited (half a day) to
fix these remaining problems via: better edge-matching of polygons along rivers; addition of
numerous missing water bodies, particularly along the Mexican coastline; development of a
script in Avenue to iteratively apply a 3x3 neighbourhood majority filter (Appendix 1). This
script was applied to the ecosystem raster dataset, prior to clipping.

' Total absence of a particular land use type, here pine forest in Mexico, is a special case that requires some
tweaks/workarounds in the CLUE-S model to avoid runtime errors. See section 4.2.6 for details.

The known unresolved data quality issues with the land cover map are as follows:

1. In the 2003 ecosystem map, a very large part of Honduras has been classified as
‘Sistemas agropecuarios’, and in the 2004 Belize dataset there is a class ‘Agricultural
uses’. As this is likely to be a mixture of cropland and pasture, this class has been named
“Agriculture/Pasture” to avoid confusion.

2. The errors that could be observed near the Belize/Mexico border in the first reclassified
raster have been fixed. However, some other abnormalities in the Mexican Yucatan —the
sudden land use changes at the 19th and 20th parallel and the 90th meridian— have not
been resolved as these are problems with the source data, not the reclassification.

1.3.3 Output extent and cell size

A cell size of 250 m was chosen. Earlier in the project, it had been assumed that the entire
country of Honduras would be included. WARI’s_ latest watershed shapefile
(mar_basin_3b.shp, 4 August 2005) showed that not all of this country would be included, so
the raster analysis extent was adjusted to avoid unnecessarily large grids. The final analysis
extent is:

West: 40 000 (no changes)
794 900 (was 920 000 but eastern part of Honduras now excluded)
Nf 2 390 000 (was 2 400 000)
South: 1 519 000 (was 1 430 000 but southern part of Honduras now excluded)

1.3.4 Reclassification methodology using ArcGIS

1.3.4.1 Step 1: Creation of clip/mask shapefile and grid

An overall MAR watershed shapefile MAR BASIN 3B RECLASSMASK_4FEB06.SHP was
created, based on the WRI version. It includes both sets of watershed boundaries that WRI
delineated from the 90 m and 250 m DEM, which were completed on respectively 4 August
2005 and 24 January 2006, and also all smaller watersheds excluded by WRI. The coastline
has been extensively edited to better match the coastline from the Ecosystem map data and
the Landsat colour composites.

The shapefile was converted to a raster at 250 m resolution: BASIN250, as described in
section 1.2.1. This raster has NoData values outside the catchment area, and values inside
the catchment area according to country: 1 for Mexico, 2 for Belize, 3 for Guatemala and 4
for Honduras.

1.3.4.2 Step 2: Creation of land cover reclassification tables

Two reclassification tables (dbf files) were created for the ecosystem datasets:
ECOMAP2003_ RECLASS.DBF (Table 1-4) and ECOMAP2004BZ_RECLASS.DBF (Table 1-5).

1.3.4.3 Step 3: Reclassification & rasterization of the Ecosystem Map data

The two reclassification tables were linked to their corresponding vector datasets in ArcMap.
Next, the Feature to Raster tool was used to rasterize the ecosystem data (see Fig. 1-2)."

1.3.4.4 Step 4: Combining the rasterized 2003 and 2004 Ecosystem datasets

The next step was the combination of the two grids created in the previous step in such a
way that the EcomBz04 v3 values takes priority over ECOMAPO3 v3. This was carried out
using the following Raster Calculator expression, and the result saved as grid COMBRAW_V3.

IsNull ([ecombz04 v3]),[ecomap03_ v3], [ecombz04 v3])

Table 1-4: Reclassification table for the 2003 Ecosystem map using field DESCRIPTIO

Arbustales de coniferas 4

Arbustales de latifoliadas 4

Arbustales mixtos

Arbustales xeromorficos subdeserticos
Areas con escasa vegetacion Other

Arrecifes coralinos Other

Bosques deciduos de latifoliadas Broad-leaved forest

Bosques manglares Mangroves
Bosques siempreverdes de coniferas Pine forest

Bosques siempreverdes y semisiempreverdes de latifoliadas
Bosques siempreverdes y semisiempreverdes mixtos
Cuerpos de agua

Otros

Pantanos y humedales

Plantaciones forestales

Broad-leaved forest
Broad-leaved forest

-=|3]p

Wetland/Swamp
Broad-leaved forest

Paramos
Sabanas
Sin datos

Agriculture/Pasture

Sistemas agropecuarios
Sistemas productivos acuaticos (camaroneras, salineras)
[ Urbano

| Urbanized

Table 1-5: Reclassification table for the 2004 Belize Ecosystem map using field ECOSYSTEM

NEWCLASS [NUM |

Lowland pine forest

Agricultural uses Agriculture/Pasture 3
Coral ree Water 9
Lowland broad-leaved dry forest
Lowland broad-leaved moist forest Broad-leaved forest
Lowland broad-leaved wet forest | Broad-leaved forest

| Pine forest

Savanna
Mangroves
Water
Water

Lowland savanna
Mangrove and littoral forest

Open sea
Seagrass

' It should be noted that these conversions could not be successfully completed in ArcGIS 9 (it hung the
application). The reason for this is unknown. ArcView 3.3 was used instead.

O}]N} Or} Ph

.

? Coral reef, sea grass and open sea are included in the original source data and were reclassified as Water, but
these ecosystem types are not relevant to the land cover change analysis.

Shrubland

Sparse Algae
Submontane broad-leaved moist forest

Submontane broad-leaved wet forest

Other
Broad-leaved forest
Broad-leaved forest

Submontane pine forest Pine forest

Urban Urban

Water Water a
Wetland Wetland/Swamp 6

Features to Raster

Input features: [2003 Ecosystem Map x] S| Input features: [Ecosystem Belize 2004 7] |
Field: | ecomap2003_reclass.NUM * j Field: | ecomap2004bz_reclass. NUM v |

Output cell size: 250 Output cell size: 250

Output raster: [ecomap03_vd B| Output raster: ecomb204_v2 Taal |
cont _| co

Figure 1-2: Rasterization of the Ecosystem map vector data on linked field NUM for the 2003
Ecosystem Map data (left) and the 2004 Belize Ecosystem Map (right)

2 x] Features to Raster

1.3.4.5 Step 5: Application of a hole-filling majority filter

The raster ECOMRAW_v3 had some imperfections. First, some apparently randomly located
grid cells were Null where they would not be expected to be Null. This was traced back as
the result of non-matching polygons in the original vector data, where the centre point of the
grid cells fell exactly in the empty area between the two polygons. Even a grid cell that had
>95% of its area covered by the vector data could still become NoData in this way. Second,
the coastline of the Ecosystem Map dataset did not exactly match the coastline of the
clip/mask shapefile. For the statistical analysis it is crucial that all datasets contain exactly
the same number of value grid cells (not NoData cells).

A hole-filling Avenue script (GRIDTOOLS, Fill NoData Holes in Grid) was used. This
iteratively applies a majority filter. This script not only fills any single NoData cells, but also
buffers the raster as described in section 1.1.1. A 1-cell thick buffer is added in each iteration
and a total of five iterations were carried out. The resulting dataset was ECOMFILTER_V3.

1.3.4.6 Step 6: Clip to the watershed extent and coastline

Lastly, the ECOMFILTER_V3 was clipped to the extent of the watershed using the mask grid
BASIN250 and the result saved as ECOMAPFINAL_V4. This is the final land cover grid.

Con (IsNull ([basin250]), SetNull([ecomfilter v3)) , [ecomfilter v3])

1.3.5 Calculation of area by country and land cover type

This was easily carried out in the Raster Calculator using BASIN250 and ECOMFINAL v4.
Recall that BASIN250 has four different values for each country values (1=MX, 2=BZ, 3=GT
and 4=HN) and the final land cover grid has values from 0 to 9. The expression below
produces a grid that has unique values for each land cover type in each country.

({basin250] * 10) + [ecomfinal v4]

The resulting grid was saved as cLS4cntTRY_V4. This grid’s attribute table lists the number of
cells per land cover per country (Table 1-1). The grid values range from 10 (Other/Unknown
in Mexico) to 49 (Water in Honduras). As each grid cell is 250 m, the area in km? was
calculated by dividing the Count value by 16.

1.3.6 Land cover for use in N-SPECT

The combined & reclassified land cover grid ECOMFINAL_v4 was developed for use as the
“current” land cover by both the CLUE-S model and the N-SPECT model. It is important that
the same land cover grid is used in by both models to allow accurate evaluation of the
impacts of the land cover change simulated by CLUE-S on the results of the N-SPECT
model. For N-SPECT, it is necessary to reclassify/remap the 10 different land cover types to
10 corresponding ones from the set of 22 land cover types supported and hard-coded into
the N-SPECT model.

1.4 Explanatory factors of land use patterns

A set of potential explanatory factors was compiled on the basis of a literature review and
other knowledge about the dominant factors that have affected the directions of land use
changed in the past and/or affect the prevailing land use patterns. CLUE-S operates by
extrapolating the current land use pattern and driving forces of change to the future (Kok &
Veldkamp 2001, Wassenaar et a/. 2005, Kok & Winograd 2002, Kok 2004, Cherrington
2005). Table 1-7 lists the factors that have been identified and for which data are available at
this time.

Each location factor is represented in the form of a grid that is clipped to the boundaries of
the country based on the extents listed in Table 2. There is a separate grid for each country
because the regression analysis and CLUE-S model runs are performed on a country basis.
The main categories of explanatory factors are described below. It has been assumed that
only factors on this list have to be accounted for; on the other hand, some of these factors
may not be significant.

1.4.1 Topographic factors - elevation and slope

1.4.1.1 Data source

The Shuttle Rader Topography Mission (SRTM) data provided the most consistent and
highest resolution elevation data for Central America. CIAT has processed the original 90 m
resolution STRM data to fill any NoData holes using digitized contours from topographic
maps and other elevation products. These processed data were used in this project.

1.4.1.2 Data processing

CIAT data were available in 5 x 5 degree tiles. A total of eight tiles covering 10-25N and
80-95 W (Fig. 1-3) were required to cover the entire MAR catchment .These were merged
into a seamless mosaic, SRTMFULL, in geographic coordinates and WGS-1984 datum. This
DEM was projected to UTM zone 17 using a modified Raster Project tool to a 250 m DEM,
SRTM250 BL cc. This name reflects the discovery that the bilinear and cubic convolution
resampling methods both gave the same result as the grid resolution was increased from

10

0.0008333° (approximately 90 m) to 250 m. The factor grids for elevation and slope
(degrees) were computed from this DEM.

In the early stages of the project, a comprehensive accuracy assessment of the SRTM data
was conducted along with a review of relevant geographic transformations and tools for
projecting raster datasets in ArcGIS. The out-of-the-box Raster Project tool in ArcToolbox is
that it cannot perform geographic transformation of raster datasets. This is a known issue
with ArcGIS 9.0/9.1. Consequently, a modified, functional version of that tool was developed
by Joep Luijten.

Figure 1-3: SRTM tile numbers that were downloaded. Tile 20_10 was included as the earlier
versions of the watersheds boundaries indicated that it extended more to the east and southeast.

11

(a!

(4/0) wl

cuVdOM Iq!ssOd

L00¢ dwerypleA g XO ‘GOOZ seeuasse/\

uooaj}o1d jemed) eale pajoajoig

LL

LuvdGM A\qissod (1/0) wt L00Z dwexpjeA g 40M ‘SOOZ Jeeusssen (90 uer) WdGM (uonoajoud |Inj) Seaue pajoajold | OL
SIMNOL Ajax (1/0) yl 0 BABWEATAS MMM//-}U ‘GQOz uo}BuleYD spedxea 4M ‘eAeyw eales sjodsjoy wsuno} ‘je}seoD pea
(4 OL < uonejndod
INWSOV A\qissod yeoj4 L00Z dweypjaA 8 40M ‘LVIO BOEYNS UONOI4 | YJIM SAI}IO) S}oyxJeLW 0} SSBD0e JO SOD vp
s9dd1S ON yeo|4 L00Z dweryplaA 9 40 :GO0Z JeeUasse/n Wd WLYS ado|s 9
1VWAs13 ON yu] L007 duerxpleA 40 W3d WLYS apnyniy S
Aydesbodo,
(ules WLU 99 > YIM Sy}UOW
NOWANG ON (2-0) ql L007 dweypla ¥ 40M 'SOOZ Jeeuassen WIJDPUON\ WO panueq | aaynoesuoo) poved Alp yybueq v
YANIVY ON jul L00z dweypla/ 9 40M 'GOOZ JeeUasseN aseqejep WIIQPHOM LVIO |Jeyules Jenuue ueayy €
ayewd
|
NIVYGS ON 1-0 ‘yeO|4 L00z dweyplen 8 40m “GOO Jeeuassenn OV1 YSLOS ABeUlesp [10S z
H1d3qS ON yeo|4 L00Z duexpljaA 8 40M ‘GOOZ JeeUasse/N OV1 YALOS yidap |10S L
| BojoaB nos ||
N3ddOd 124!7 ye0|4 GO0z Jeeuassen '|00Z AwexPIa/ 9 40 OVI LVIO '€A MdD Susp UO!}eiNdod 0
o1ydeabowaq
|
xijaud
al4 | olueudg adh Aq paysebinsjasuaiajoy aounos e}eq 4oyoey Asoyeurjdxg as

‘suojoey Auojeuejdxe jeouobejeo Ou aseaM asaY | ‘SJOYJEUI O} SST JO JSOD O} pajejalod AjBuosjs si j! asneoaq
sishjeue ay} WoL Jno Ya] AJenjuare SEM speOs 0] SS9IIe JO JSOD ‘CE Je SUe}S asojasay) pue Salij saJaW/eIed s}jnsai uoIssaiBal S-4N7O ay} Jo Buuaquinu
J0J pasn uaaq sey (#) saquinu ey, ‘sabueyo asn puey jo payeynuys pue sisAjeue uolssai6a ay) Ul Papnjou! aq jj/M Jey] SIOJOeY AUOJeUe/AX~a JeUBJOd :9-L A/QeL

1.4.2 Demographic factors — population density

1.4.2.1 Data sources

There was a choice of two population density datasets, as follows: (i) CIAT and colleagues
completed the third version of the Latin American and Caribbean (LAC) Population Database
in March 2005 (CIAT et al. 2005). It contains vector population maps (population per
administrative unit) and raster surfaces created with an accessibility model; (ii) CIESEN
released the latest Gridded Population of the World (GPW) database v3, together with the
Global-Rural Urban Mapping Project (GRUMP) data (Balk et a/. 2004) in December 2005. A
third dataset, Landscan 2004, was produced by the Oak Ridge National Laboratory, USA,
but the project team was not able to obtain this dataset.

When overlaying on a Landsat image, the LAC dataset is visibly less accurate than GPW v3.
GPW’s actual population density for 1990, 1995 and 2000, estimated density for 2005, 2010
and 2015, and population density grid appears more accurate. This may be because CIAT
used comparatively coarse road maps and urban areas data for population modelling.
However, the GPW data do not show much spatial variation in population density in Belize,
and to a lesser extent in the Mexican Yucatan. Belize City does not stand out at all in the
GPW dataset, probably because population densities are averaged across administrative
areas, of which there are only six in Belize (compared to 3 696 in Honduras). The LAC
dataset is slightly better for Belize, although it still does not look accurate in the vicinity of
Belize City.

1.4.2.2 Data processing

The LAC dataset was selected for Belize and Mexico, and the GPW v3 dataset for
Guatemala and Honduras. The original data at 1-km resolution were resampled to 250 m. No
other processing was carried out. For actual (1990-2000) population data, the GPW3 “ac”
grids (adjusted population density to match UN totals) were used.

1.4.3 Demographic factors — location of settlements

1.4.3.1 Data source

As there was no available consistent urban point dataset with associated population
information, a dataset was pieced together from four different sources data. Many of these
were identified by Emil Cherrington. For Honduras, a dataset (HN_SETTLEMENTS IGN-
CCAD.SHP) from CCAD was used. For Belize a dataset (BZ SETTLEMENTS BTFS.SHP) from
the Belize Tropical Forest Studies was used. For Mexico and Guatemala the settlements
from the Selva Maya CD (see Table 1-6) were used (Bz-GT-
MX SETTLEMENTS SELVA_C2000.SHP). El Salvador was included because it closely borders
the MAR catchment; the GRUMP v1 settlement data were used here, even though they were
not very accurate.

1.4.3.2 Data processing

Each dataset was projected to UTM16 and the fields in the attribute table that contained the
city/settlement name and population size were renamed, respectively, NAME and POPSIZE.
The datasets were then merged. The Selva Maya data appeared quite inaccurate and so,
where possible, alternative data were used. Guatemala City was missing and was added
manually, with its location based on the ESRI and GRUMP settlements and ESRI world
cities. The resulting shapefile MAR SETTLEMENTS POP.SHP has a field SOURCE that
indicates were the point features originated from. It should be noted that the GRUMP dataset

13

is very small scale, but was useful for comparison to indicate whether any key cities are
missing, rather than for precise pinpointing of locations.

1.4.4 Soil and geology factors

1.4.4.1 Data source

The highest quality consistent soil dataset for Central America is the ‘Soil and Terrain
database for Latin America and the Caribbean’ (SOTERLAC) and its associated SOTER-
based soil parameters estimates (version 1). Both datasets were downloaded from
http://www. isric.org/UK/About+Soils/Soil+data/Geographict+data/Regional/. The soil
parameters dataset contained everything that was needed - shapefiles, legend files and a
large MS Access database that contained all parameters for every soil profile ID (PR/D) and
unique SOTER unit (VEWSUID).

1.4.4.2 Data processing

Essentially, the entire area has been characterized using 1585 unique SOTER units,
corresponding with 5855 polygons, and the soils described using 1660 profiles. Each SOTER
unit is associated with one or more profiles, each given a relative weight and totalling 100%.
Each profile and its soil parameters are specified by up to five different layers (D1 = 0-20 cm;
D2 = 20-40 cm; D3 = 40-60 cm; D4 =60-80 cm and D5 = 80-100 cm), but the deepest layers
can be less than 20 cm thick. The soil parameters vary between soil profiles. The attribute
data of the shapefiles only contained the soil parameters for the top layer (D7).

To calculate the average soil depth and average drainage (over all soil layers), an
aggregation had to be made across soil profiles and soil layers, as follows:

For soil depth and drainage:

© Opened sSOTWIS_SOTERLAC_1.MDB and exported table SOTERsummaryFile to a
SOTER_SUMMARY_FILE.DBF.

© Edit SOTER_SUMMARY FILE.DBF and add a field ProfDepth (Number, 2 decimal
places) to store the effective depth of that profile within a SOTER unit.

© Calculated the ProfDepth, incm, as: 0.01 * ([BotDep] - [TopDep] * [Prop].

© Added a field ProfDrainage (Number, 4 decimal places) to store the effective
drainage rates of that profile within a SOTER unit.

® Calculated the ProfDrainage, incm, as: 0.01 * [Drain2] * [Prop].

® Summary on the Newsuid, taking the Sum of ProfDepth (which will be in between 0
and 100 cm) and the Average of ProfDrainage (which will be in between 0 and 1).
The file was saved as CUMULATIVE_BY_NEWGUID .DBF.

® Linked CUMULATIVE BY NEWGUID.DBF tO SOTERLAC2 SOTWIS.SHP and created a
legend based on the cumulative soil depth (Sum_Cum_Depth).

® Convert feature to raster on the cumulative depth field and drainage field. The
resulting grids were saved as SDEPTH and SDRAIN.

© Applied the majority filter 20 times to each grid (so many times to fill major gaps near
the islands) and saved grid as SDEPTHFT20 and SDRAINFT20.

14

© Mask and clipped a total of 8 grids using the raster calculator:

Saved as MXSDEPTH
Saved as MXSDRAIN
0|) Saved as BZSDEPTH
T20]) Saved as BZSDRAIN
°20]) Saved as GTSDEPTH

}) Saved as GTSDRAIN
Saved as HNSDEPTH
Saved as HNSDRAIN

1.4.5 Climate factors - precipitation and length of dry season

1.4.5.1 Data source

CIAT’s WorldClim database (http://biogeo.berkeley.edu/worldclim/worldclim.htm) was used.
The database contains grids of monthly mean temperatures and precipitation in several
resolutions (30 degree-seconds and 2.5, 5 and 10 degree-minutes). The finest resolution, 30
degree-seconds (about 1-km x 1-km), was considered more than sufficient for land use
modelling.

1.4.5.2 Data processing

WorldClim data at 30 degree-seconds resolution were downloaded for tiles #22 and #23 (the
MAR catchment covers a small part of each tile). These monthly grids were mosaicked and
stored aS PREC_1, PREC_2, .., PREC_12. Each grid was then projected to UTM 16 NAD
1927 and clipped to the extent of the MAR. A bilinear interpolation was used. The resulting
grids are stored aS PRECIP1, PRECIP2, .., PRECIP12. A grid of annual precipitation
ANNUALRAIN was computed by adding the 12 grids.

The calculation of dry season length was more complex. Based on existing literature, a
month with less than 60 mm of precipitation is considered a dry month. An inspection of the
range of values of the monthly grids showed that November through May are generally the
drier months. While the minimum value in both July and August is also below 60 mm, the
much shorter period and very few grid cells with a value < 60 makes this insignificant
compared to the other seven months.

A short Avenue script was written as follows, to calculate a grid that indicates whether each
of these seven months is a dry month. Note that the script is hard-coded to use the
precipitation grids from months 11, 12, 1, 2, 3, 4.and 5.

ol
)

Q

oO

' oR
)

D

»}

+

7. FindTheme ("
7.Finc iTheme ( ig
.FindThe

eave out other months.

15

).asgrid) .Cor

O.asgrid) .Con(

The output grid has 8-digit numbers only. The first digit is always 2 and has no meaning: it
exists solely to make sure that the first digit is not a 0 (resulting in a number less than 8 digits
long). The 2nd through 8th digit indicate whether, in exactly the following order, the month of
November, December, January, February, March, April, and May is a dry month (value=1) or
not (value=0). A reclassification table was manually created (Table 1-6).

Note that the number Dry_Months is the number of consecutive dry months. For example, a
value of 20101010 is reclassified as 1 because there is a maximum of just one consecutive
dry month (albeit it occurs three times), not three consecutive dry months. The resulting grid

was Saved aS DRYMONTHS (Table 1-7).

Table 1-7: Grid reclassification (resampling) scheme for the number of dry months.

DryMonths

Value Count
iii

20000011 {104
20000100 /18129

20000110 /42598

20000000 [59136
20000010 /10397

20000101 {142

= ip l= lo

20000111 |187
20001000 |2164

20001010 |4224

20001100 {5290

20001110 |67622
20010110 |25

20011000 |14
120011100 |1320

W [Nh [Nh [@ [hy [= [= Jw il

Value Count

DryMonths

20011110 |34593

Fi
2

20101110 |4406 3
20111000 3

20111100 |6604
20111110 |61067

20111111 [29

21111100 |10403
114659
6930

Next, the majority filter was applied five times and the grids saved as ANNUALRAINFT and
DRYMONTHSFT. Lastly, the grids were clipped using the raster calculator

Con ( [BZ
Con (
Con ([
Con (
Con (

Con (

16

Saved as MXDRYMON
Saved as MXRAINYR
Saved as BZDRYMON
Saved as BZRAINYR
Saved as GTDRYMON
Saved aS GTRAINYR
Saved as HNDRYMON
Saved aS HNRAINYR

1.4.6 Contextual factors — protected areas

1.4.6.1 Data source

The World Dataset of Protected Areas (WDPA) that is maintained by UNEP-WCMC has
been used. Initially the 18-May-2005 WDPA version was made available. A comparison with
similar data from both CCAD and MesoStor showed major discrepancies, in particular for
Honduras. There were obviously missing data in the WDPA dataset. Revisions were started
during Emil Cherrington’s visit in December 2005 and an improved dataset for MX, GT, BZ
and HN was made available in January 2006. A further revision was completed in May 2006,
along with a hypothetical future protected area dataset for the scenarios. The only key
difference between the Jan and May versions was the inclusion in May of a large area in
Belize (Gallon Jug Estate). Table 1-8 lists the prevailing WDPA types in the four countries
along with the number of polygons of each type.

1.4.6.2 Data processing

The WDPA dataset contains protected areas with different types of designation (national
parks, biological reserves, etc.). For the regression, the degree of protection from land use
change is important. The IUCN category (IUCN 1994) was used to generate an estimate of
protection level.

The following assumptions were made.

© The areas are legally protected from land use change if they are in IUCN Categories | to
IV. There are some exceptions for category III (Natural Monument), but this general rule
will be used in the CLUE-S model.

© The areas may be subject to some level of change (but certainly not complete change) if
they fall in IUCN categories V and VI.

© Any area that does not have a category assigned (115 areas for the MAR countries), was
treated as if it was fully protected from change.

Data processing steps:
© Two new fields named PROTECTED1 and PROTECTED2 were added to the WDPA
shapefile WOPA_MAR_SUBSET_UTM16.SHP. The field were of type integer.

© All polygons of the categories | to IV and the “unset” ones were given a value of 1 for
PROTECTED1 (full protection). All polygons in categories V and VI were given a value
of 1 for PROTECTED2 (partial protection).

© The shapefile was rasterized on both fields and the resulted grids saved under the
same name as the fields, PROTECTED1 and PROTECTED2. Note that these grids have
values only for the WDPA area, not for the entire country.

® Lastly, the following equations were used to created the final clipped grids:

on(isNull([protectedl - protectedi]),0,[protectedl - protectedl)), saved
as WDPAR1
Con(isNull([protected2 -— protected2]),0,[protected2 - protected2]), saved
as WDPAR2

17

Table 1-8: Prevailing designation types of WDPA areas in Mexico, Guatemala, Honduras and Belize.

Designation Type Areas ' Designation Type CIE
Anthropological Reserve ;

1 (1) Multiple Use Reserve 1(1
Archaeological Reserve 12 (12) | National Park 121 (40)
Archaeological Site os 2 (1) National Park - Buffer Zone ARG)
Area de Proteccion Especial ASKS) Natural Monument 10 (3)
Biological Reserve | 12 (8) Natural Resources Protection Area
Biosphere Reserve 47 (15) Nature Reserve
Biosphere Reserve Core Zone 89 (8) Private Natural Reserve
Bird Sanctuary 7 (3) _ Private Reserve
Crocodile Reserve 1 (0) Protected Biotope
Cultural Monument 7 (3) Regional Park
Fisheries No Take Zone | 11 (1) Reserva de Manantial 2 (2) |
Flora and Fauna Protection Area 402 (4) Sanctuary
Forest Reserve 18 (17) Wildlife Refuge
Mangrove Reserve 1 (0) Wildlife Sanctuary 9 (8)
Marine National Park 1 (0) Zona de Amortiguamiento
Marine Reserve 29 (4 Zona de Veda Definitiva
Multiple Use Area 10 (5) Area Productora de Agua

1.4.7 Contextual factors — access to roads and markets

The accessibility of transportation links and markets are important explanatory factors of land
use patterns and how land use changes. Accessibility is more than a measure of distance. It
has been described as the ease with which a location may be reached from another location.
The concept of accessibility has been used in rural development policy as an indicator or
rural deprivation and as a variable in location analysis.

Farrow and Nelson (2001) and Nelson (2000) developed a raster GIS-based methodology for
calculating accessibility grids using cost-distance functions. The same methodology was
used here for calculating accessibility of roads and markets.

1.4.7.1 Data source and data processing - roads

Numerous roads datasets of varying quality and ground year were identified (Appendix 2).
The best quality regional dataset was the one from MesoStor (RED_VIAL_LINE.SHP). In
addition, other datasets, thought to be more accurate, were available for Belize. A national
map was created by Jan Meerman, as an update of the Land Information Centre’s (LIC)
roads dataset, using 2000-03 Landsat Imagery. Furthermore, Emil Cherrington made
available a 2005 road dataset for southern Belize.

The processing of the road data was cumbersome for several reasons. First, while all three
datasets included a road classification, these classifications were different and needed to be
reconciled. Second, overlaying the data for Belize showed that they were all different, though
no single dataset seemed superior to the others. Some existing roads were missing in the
MesoStor data included in the more recent Meerman data, but the opposite was true for
other roads. The Belizean datasets were also most detailed, including many tracks.

1 Total number of WDPA areas in Mexico, Guatemala, Honduras and Belize. The number between parenthesis is
the number of areas that are wholly or partially within the MAR catchment boundaries.

18

Thus, reconciling these differences and the creation of a single combined dataset was the
first processing step. QuickBird Satellite imagery, viewed through Google Earth, was used to
resolve discrepancies about the existence or precise location of roads. The combined
dataset MAR_ROADS MODELLING. SHP has the fields TYPE and SOURCE. The former field is
used for symbology. For consistency between countries, tracks were omitted. The SOURCE
field shows from shapefile each road originated.

1.4.7.2 Data source and data processing - markets

It was assumed that markets exist in the larger cities, so a dataset of settlement locations
with population data was needed. No such single dataset for the entire MAR or Central
America existed, however, several other datasets that covered a country where available.
The settlements data from the Selva Maya CD provided good coverage in Mexico, Belize
and all but the southern part of Guatemala. Better data for Belize were available from the
Belize Tropical Forest Studies project (http://www.green-hills.net/btfs/). Several datasets
were available for Honduras and the one from IGN/CCAD was the most complete. The
datasets were merged and reviewed, resulting in the combined dataset
MAR_SETTLEMENTS POP. SHP.

1.4.7.3 Data processing - accessibility

The methodology described by Farrow and Nelson (2000) was followed, although their
accessibility wizard (an Arcview GIS 3.2 extension) was not used, to retain control over all
processes. Several new grids were prepared as an input to the cost-distance functions.

© To avoid edge effects that may be caused by the exclusion of roads that are just
outside the MAR boundary, the catchment extent was buffered at 50 km and
rasterized. This raster MASK50K was used as a temporary analysis extent

(Xmin=-10 000; Xmax=844 000; Ymin=1 469 000, Ymax=2 440 000).

© The land cover raster was recreated to include the 50 km buffer zone. The resulting
grid had the same 10 classes (Table 1-9) and was saved as ECOMAP50K. A value of 0
(unknown) was assigned to those areas that fall in the buffer zone and that do not
have land. Note that the precise value doesn’t matter.

© Slope affects travel time. Slope in degrees was calculated from the DEM. Any areas
not covered by the DEM and oceans were assigned a slope of 0. Again, the precise
value of the additional areas in the buffer zone doesn’t matter. The resulting grid was
saved aS SLOPE5OK.

© Roads were rasterized using four classes: 1=paved roads; 2=major roads; 3=major
roads dry season only; 4=other roads (Table 1-10). Tracks were omitted. The source
grid SRC-ROADS was reclassified to contain only zeros, SRC_ROADS_0.

© Settlements with a population of at least 5 000 and 10000 were selected from
MAR_POP_BUF_75KM.SHP and rasterized to SRC_POP5K and SRC_POP10K. The only grid
value is 0.

© Next the friction surface was created. As the cell size is 250 m, friction values were
expressed in seconds. This was a two-step process. First, three input grids (slope,
land cover and roads) were reclassified to their friction values (see Tables 1-9 to 1-
11), resulting in FRIC_ROADS, FRIC_SLOPE and FRIC_LAND.

© Next, the three reclassified “semi-friction” grids were combined into a single surface
using the following expression. The output was saved as FRICTION, and had friction

values from 8 to 2 700 seconds, indicating difficulty of passing through a 250 m grid
cell.

19

The same friction surface, but expressed per map unit passed through:

Accessibility in terms of travel time, in hours, was calculated using ArcView 3, as:

ACCESS ROADS = [Sre_ roads 0) .costdistance( [friction] ,ni He sqistl yw alatall
ACCESS POP5K [Sre_pop5k]} . costdistance( [friction] ,ni IL jolak IL taal AL
ACCESS POP10K = /Sre_pop10k) .costdistance( [friction] ,nil,nil,nil

© Because of the strong interdependence between access to roads and access to
market, only one of these factors (ACCESS _POP10) was ultimately included in the
regression analysis. Lastly, the ACCESS-POP10K grid was masked and clipped to
create four country-scale grids in the final format, using the raster calculator:

1, [ACCESS POP10K]), Saved as MZACSMRK
[2 )P10K]), Saved as BZACSMRK
K]), Saved as GTACSMRK

)K]), Saved as HNACSMRK

Table 1-9: Friction values for land cover with 250 m grid cells. On land cover, average walking speed
was estimated at 4km/hr, but reduced to 3 km/hr in forest and increase to 5 km/hr in urban areas.

Land cover type Speed

Friction value
sec per 250 m

0. Other/Unknown
1. Broad-leaved forest
2. Pine forest
3. Agriculture/pasture
4. Scrub
5. Savanna
6. Wetland/Swamp
7. Mangroves
8. Urban
[ 9. Water

=rah

Table 1-10: Friction values for different road type with 250 m grid cells.

Road type Speed Friction value

| 1. Paved road
2. Major road
3. Major road (dry season only) _|
4. Other road |

20

Table 1-11: Friction multipliers for slope. There is no accounting for slope direction; it is assumed that
travelling both up-slope and down-slope incurs a reduction in travel speed.

Slope Friction value |
multiplier

|

0 — 5 degrees ; ei|fat 1 |

5 — 10 degrees | Pal
10 — 20 degrees 3
> 20 degrees all 5

1.4.8 Contextual factors — tourist hotspots and areas of coastal development

1.4.8.1 Data source

The most relevant dataset was the tourism threat layer from the Selva Maya data CD,
covering the Yucatan, Belize and the Peten region of Guatemala (the northern half). It is
composed of hexagonal polygons of 100 ha, with an attribute “Qualification” (Calificacia) that
indicates what part of the polygon is under threat. Nearly all of the areas under threat are
predominantly mangroves. In addition, WWF (email from Melanie McField) supplied the
approximate location of tourism hotspots, drawn on maps in a PowerPoint file. This
confirmed the accuracy of the Selva Maya dataset, though Honduras was not covered by
either. The two main tourist areas on mainland Honduras are the cities of La Ceiba and
Trujillo. These cities were added to the Selva Maya dataset.

While there is general consensus that urban development near tourist hotspots is a major
threat to the land in those areas, its use as an explanatory factor in the statistical analysis
difficult, because the impact of tourism is highly localized, whereas the statistical analysis
and subsequent modelling is carried out at a national level.

The problem can be split in two. First, coastal development can never be an explanatory
factor for (urban) developed land that is not near the coast, particularly in Honduras, which
has major urban areas inland. Second, the available data for tourism hotspots point out the
areas that are under the greatest pressure, rather than actual areas of tourism-induced urban
development. Overlaying the Selva Maya tourism threat layer with the ecosystem map land
cover data shows that nearly all of the areas under threat are mangroves.

Consequently, it is likely that a regression analysis between the land cover data and tourism
hotspots will not show a significant relationship. However, the areas under threat from
coastal development were included in the statistical analysis in order to confirm this
suspicion.

1.4.8.2 Data processing

® A field named RECLASS was added to the Selva Maya shapefile and all polygons
with a qualification > 100 (out of 1000) were given a value of 1. All other polygons
were given a value 0. The shapefile was rasterized on the field and the resulted grid
saved as COASTD.

® The following equation was used to created the final clipped grid:

Yon (IsNull ([coastd]),0, [coastd]), saved aS TOURIS.

21

2 Analysis of drivers of land use change

2.1 Land Use Change Adjacent to the Mesoamerican Reef

This section reviews literature on land use changes in the project region over the past twenty
or so years. It was originally released as a working document entitled “Drivers of Land Use
Change Adjacent to the Mesoamerican Reef: A Preliminary Review, by Emil Cherrington,
Coastal Zone Management Institute, Belize City, in August 2005.

The individual Annotated Bibliographies compiled for the FAO’s 2000 Forest Resource
Assessment for México, Belize, Guatemala and Honduras provide a great deal of additional
insight into country-level environmental landscape changes in the respective countries (FAO
1999, FAO 2000a, FAO 2000b, FAO 2000c).

2.1.1 Mexico

The states of Campeche, Quintana Roo and the Yucatan fall in the MAR project area. . While
national-level statistics are readily available on land cover change, GIS analysis is required
to quantify changes within the project area. Considerable work on the drivers of land use
change has been carried out for part of this area by the Southern Yucatan Peninsular Region
(SYRP) project, a joint initiative between Mexico’s ECOSUR and the USA’s Harvard Forest
(Harvard University) and Clark University.

Land use change can be summarized over the past thirty years as the result of an expansion
of agricultural activities and rapid increase in population. Change seems to have reduced in
light of the Mexican government's promotion of the regional Mundo Maya archaeo-
ecotourism initiative, which has also seen the designation of a number of protected areas in
the project region since the late 1980s. It is acknowledged that even ecotourism will continue
to affect the local environment.

2.1.1.1 Historical Land Use Change

Following a forestry (selective logging)-dominated period for the first half of the twentieth
century which went bust by the late 1960s due to international market conditions, the
Mexican government sought to use its southern frontier “as a release valve for land stress
elsewhere in Mexico.” Peasant farmers were drawn to the area due to readily accessible land
in the form of communally-owned ejidos, “the primary form of land tenure in Mexico,” created
by Article 27 of the 1917 Constitution (Merrill 1996).

Infrastructural development, such as the completion of Highway 186 in 1970, which
connected the capitals of Campeche and Quintana Roo to the rest of the nation, also
encouraged land use change. Emphasizing agriculture, Mexican governments of the 1970s
and early 1980s sought to “[reshape the] forest frontier into a rice and cattle producing area.”
Seasonal wetlands known as bajos were converted to large-scale rice paddies, but poor
practices led to failure of this venture. The land has since been used for pastureland. Other
agricultural activities include cattle ranching, fruit orchards, and the cultivation of chilli
peppers, corn and beans.

Trade liberalization in the 1990s accompanied land reforms in which farmers received formal

title to ejidos, allowing them to sell and lease plots (if this is agreed to by their communities).
Subsidies and price controls were eliminated, as was further distribution of land.

22.

Mexican participation in the regional Mundo Maya initiative is currently being promoted by
the government, which is seeking to capitalize on the region's rich history. A number of
protected areas have hence been designated since the late 1980s. According to WRI (2004),
coastal development is a major issue due to resort developments, particularly on the
Caribbean coast of Mexico.

2.1.1.2 Explicit / Implicit Drivers

As indicated above, in the recent past, government agricultural policies in the form of
subsidies, price controls and ready distribution of land encouraged deforestation in southern
Mexico. These have been discontinued with trade liberalization and a new emphasis on
tourism. However, even with nature-based tourism, in the archaeologically-rich inland and in
coastal areas, a demand is placed on land resources.

It remains to be seen how poverty and population growth also affect land use change in
southern Mexico, although discontinued distribution of lands may lead communities to
encroach on protected areas. The effort at making ejidos transferable by sale and lease is
aimed at improving the economic situation of peasant farmers by proving them with access
to credit. The elimination of subsidies for export crops should not impact demand in local
markets for food, especially given steady population growth.

Regional influences, such as Plan Puebla-Panama, are explored in section 2.1.5.

2.1.2 Belize

Land use change over the past twenty plus years of Belize’s history can be summarized as
the continuous expansion of agriculture (including aquaculture), and infrastructural expansion
driven by population growth (including immigration) and tourism. These changes have
occurred after Belize’s attainment of independence from Great Britain in 1981. Despite a
rapidly changing natural environment, deforestation was not acknowledged as an issue until
resource assessments of the mid- to late-1990s which indicated that deforestation was
occurring at rates of roughly 24 280 ha a year in the early 1990s (FAO 2000a). Whereas in
the 1980s, Belize boasted 97% forest cover, the most recent (2004) assessment indicates
that forest cover is closer to 63%, down from 72% in the beginning of the 1990s (DiFiore
2002, Fairweather & Gray 1994, Meerman 2005).

2.1.2.1 Historical Land Use Change

For most of the past three hundred and fifty years of Belize’s history, forestry was the
mainstay of the territory's economy. Colonial masters intentionally suppressed agriculture to
maintain forest resources, even as already-independent neighbouring republics had begun
their phase of agricultural development. The 20" Century saw a gradual decline of forestry
due to depressed prices on the world market, and the rise of a national economy founded on
the export of agricultural and marine products. Passage of the Land Reform Ordinance in
1962 further shifted emphasis to agriculture, and between 1971 and 1982, 212 465 ha of
land were transferred farmers. Plummeting prices for Belize’s agricultural exports starting in
the late 1970s, even further spurred agricultural expansion and made once-independent
subsistence farmers even more dependent on international market forces.

23

The mid-1960s also saw the gradual establishment of a tourist industry based largely on the
territory's offshore attractions though the industry, did not really take off until the post-
Independent 1980s, following the creation of a Ministry of Tourism & the Environment whose
efforts centred on marketing the nation as a Caribbean tourist destination (McMinn & Cater
1998). By the late 1990s, tourism began to displace agriculture as the major engine of
economic growth, averaging 20.2% of GDP per year between 1997 and 2001 (GOB 2002).
Although tourism relies on Belize’s natural assets, the industry has exerted its own impact on
the national landscape, particularly in coastal areas, where the most rapid changes are
believed to be occurring.

2.1.2.2 Explicit / Implicit Land Use Policies

The Belizean Government continues to be the largest landowner in Belize, and almost 37%
of the country’s land is vested in protected areas. Only a few of these are privately-owned.
The government encourages smail and large-scale enterprises in tourism or agriculture, in
the face of ever-mounting foreign debt and continuing trade deficits. The implicit government
policy has been support for the agricultural, aquacultural, and tourism sectors (over say
forestry) because of the revenues and contribution to GDP generated.

The main export crops include citrus, bananas, and sugarcane, while locally-consumed crops
include beans, rice and corn. With regard to the export crops, sugarcane is cultivated mostly
in the north of Belize, while citrus and bananas are cultivated in the centre and south of the
country. In the 1970s, for instance, revenue from sugar exports accounted for roughly 70% of
export revenue (Merrill 1992). While there was no formal agriculture policy until 2003,
agriculture was and still is widely promoted, though there are questions as to the impact of
trade liberalization. Traditionally there have been price guarantees for Belizean crops in the
American and European markets, even though such support is now waning.

Boles (2005) cites poverty as a significant driver of land use change, indicating that it has
driven deforestation in southern Belize via slash and burn milpa agriculture. Some speculate
that integration of Belize into the Caribbean Single Market & Economy (CSME) initiative may
mean increased immigration from the Caribbean and hence greater demand for land.

Belize has one of the most extensive protected areas systems in the world, and almost one
protected area has been added to the national list each year. Nevertheless, there have been
de-gazettements of protected areas and sections thereof in recent years. The ongoing
National Protected Areas Policy & System Plan (NPAPSP) project seeks to define a national
policy on protected areas, and to rationalize their future existence.

The lack of an overarching, explicit land use policy and plan has resulted and continues to
result in haphazard development. The National Lands Act encourages prospective
landowners to ‘develop’ the land, whereby development is defined as modification of the
land’s original cover. There is a continued outlook in some quarters that natural habitat as
‘useless’ land to be ‘developed,’ irrespective of its biophysical potential. Due to the continued
importance of coastal areas to tourism, a continuous, largely unregulated development in
coastal areas (on the coastal mainland and on offshore islands) led to the establishment both
of a national Coastal Zone Management Authority and, more recently, guidelines for
development in coastal areas. Some institutional weakening of the Coastal Zone
Management Authority has, however, occurred since its initial sponsorship through the
UNDP-GEF and EU ran out in mid-2004.

A project in the pipeline through the UN Convention: to Combat Desertification, includes the
preparation of a national land use plan to guide future development efforts. An ongoing land
titling initiative is occurring through the Land Management Programme, which is conducting
cadastral surveys in the northern half of Belize. The LMP is intended to stimulate economic

24

growth through secure land tenure. It remains to be seen if the end result will be further
emphasis on productive enterprises such as agriculture.

In light of the above analysis, it seems that population growth, migration, coastal
development, and agricultural / aquacultural expansion will be the main factors driving land
use change in Belize in the near future.

2.1.3 Guatemala

While eight of Guatemala’s southern Pacific states’ fall outside in the MAR project area, most
of the information available covers the whole country. The National Institute of Forestry
(INAB) reports that in the 1980s, deforestation occurred at a rate of roughly 60 000 ha per
year, while in the 1990s, the rate was roughly 90 000 ha per year (FAO 1999). This change
seems to have driven jointly by agricultural expansion and human population dynamics,
including migration to the largely forested eastern highlands of the Petén in northern
Guatemala (FAO 1999). FAO (1999) further states that forest policy had changed three times
over the twenty-year period, and that there has been competition between the forest and
agricultural sectors, though since the 1990s, forestry has played a larger role in the
economy.

2.1.3.1 Historic Land Use Change

Large areas of land were converted to agriculture from the early 20" century onwards.
Around the middle of that century, Guatemalan governments promoted agriculture as the
major avenue of economic growth. Government policy was that the wide expanses of forest
were essentially “useless” and should be converted to “productive” uses. In reality, some of
the areas where such land conversion occurred, such as the Petén, are infertile. Land was
openly distributed to peasant farmers, and promotion of agricultural activities took the form of
subsidies, price guarantees and laws encouraging development via land conversion.

Commercial, export-oriented agriculture has been practiced mostly in southern Guatemalan
states (most of which fall outside of the project area), while shifting cultivation, cattle ranching
(and illegal logging) have predominated in the Petén (FAO 1999). Shriar (2002) cites the
Petén as being 70-80% forested in 1970, but only 50% forested by the late 1990s.

The late 1980s through the mid-1990s saw the establishment of various protected areas
such as the Maya Biosphere Reserve in the Petén, and institutional changes empowering the
national Commission on Protected Areas (CONAP) and the INAB. The role of forests in the
national economy has likewise changed, with the introduction of market incentives to prevent
deforestation, including concessions, and exploration of carbon sequestration as options for
revenue generation.

FAO (1999) recognizes the increasing role of managed forests in the Guatemalan economy,
but notes that there is a national debate as to whether agriculture has stabilized or will
continue to expand, and on the effectiveness of protected areas in maintaining forest
resources. CONAP has delegated management duties of various parks to NGOs.

2.1.3.2 Explicit / Implicit Drivers

Shriar (2002) points to a growing population in areas such as the Petén, while FAO (1999)
discusses the significance of “migration, colonization and dependency” on change. FAO

! These are Escuintla, Huehuetenango, Jutiapa, Quezaltenango, Retalhuleu, San Marcos, Santa Rosa, and
Suchitepequez. According to FAO (2001), these areas produce sugarcane, cotton and cattle for export.

25

(1999) also points to the emergence of forestry as a major player in the Guatemalan
economy as being able to drive sustainable use of forests, particularly because of economic
incentives coming from the government. Other factors mentioned by both Shriar (2002) and
FAO (1999) are the availability of land (even despite protected area designations), and the
incidence of rural poverty, which limits communities’ options economically.

2.1.4 Honduras’

While the other nations of the project area are acknowledged to be underdeveloped,
Honduras is one of the few Highly Indebted Poor Countries in Latin America (Jansen et al.
2005). The nation has a more diverse topography than the rest of the region, with a large
mountainous area and largely infertile soil (Merrill 1993). As with the other nations,
agriculture is a major contributor to GDP. The World Bank figures cites the nation’s
population growth at 2.6% per annum (World Bank 2004c).

2.1.4.1 Historical Land Use Change

Martinez et al. (1999) indicate that almost half of the forests that existed in 1965 had been
converted to other uses by 1992. FAO (2000b) and Merrill (1993) also indicate that large
areas of forest land were converted to agriculture in the latter half of the twentieth century,
continuing into the late 1980s. Farmers focused on the production of livestock, and the
cultivation of coffee, bananas, sugar, and basic grains. Despite the poor soil of the nation’s
mountainous landscape, agriculture has mainly expanded, rather than intensified, and
resulted in the erosion of an estimated 2.3 million ha (FAO 2000b). In the 1990s, following
trade liberalization, commercial agriculture declined.

Other factors contributing to continuous land conversion have been population growth, and
the incidence of natural disasters. Hurricane Mitch in 1998 had substantial impacts on both
natural forests and human-dominated landscapes (FAO 2000b).

2.1.4.2 Explicit / Implicit Drivers

The major cited drivers of deforestation have included expansion of agriculture & cattle-
ranching, population growth & colonization, land tenure, energy production needs,
competition between forestry and agricultural policies, forest fires, crop disease and natural
disasters such as hurricanes (FAO 2000b).

According to Jansen et al. (2005: 18), trade and market liberalization in the 1990s saw the
discontinuation of “land distribution and rural credit provision,” agricultural extension services,
consumer subsidies and guaranteed prices. In theory, this should have discouraged the
expansion of export-based agriculture. The authors suggest that increased emphasis should
be placed on intensification of existing agriculture as a means of poverty alleviation. They
also recommend putting measures in place to limit population growth.

Bonta (2005: 95) states that “by 2000, Honduras alone possessed over 100 protected

areas...including 37...‘cloud forests’ that had been set aside by presidential decree in 1987”.
He suggests that many Honduran protected areas are protected merely on paper.

2.1.5 Regional Synthesis

Certain cross-cutting themes seem to emerge from the four countries, including:

' Only the southern departments of Choluteca and Valle are excluded from the project region.

26

(i) A strong emphasis on agricultural activities in the last few decades, at the
expense of forest land. In the case of both Belize and Honduras, it appears that
agriculture is expanding rather than intensifying

A former emphasis on forest management (excluding Honduras), faltering in the mid-20"
Century due to the international market

(ii) An expansion of road networks and settlements driven by population dynamics of

both growth and migration

2.1.5.1 Future Land Use Change

A number of factors operating at national and regional scales can be expected to influence
future land use changes. For one, each of the countries of the region are the signatories to
some form of trade liberalization agreement, whether it be the Free Trade Agreement of the
Americas, the Central America Free Trade Agreement or the Caribbean Single Market and
Economy. The conventional wisdom is that these will discourage agricultural expansion by
removing subsidies and price guarantees

Other sources indicate that such liberalization will instead encourage agricultural expansion,
because countries will have to export more products to maintain previous levels of revenue.
Plan Puebla-Panama can be expected to open up previously inaccessible areas to
development. The regional fisheries & aquaculture policy advocated by the PREPAC project
may in turn lead to increased aquacultural activities in coastal areas.

Population growth and migrations will themselves exert pressures on national land
resources. Such migrations may be within individual countries, or between nations in the
region, such as expected to impact Belize through the CSME initiative. Tourism is expected
to continue to grow, with a proportionate increase in demand for land in coastal areas and
offshore islands. The influence of climate change on land suitability will also become
increasingly important in the future.

While the list of possible future causes of land use change can only go on, with regard to
spatially explicit causes, infrastructural development and expansion of both settlements and
roads, and expansion (rather than intensification) of agriculture and aquaculture seem like
the most plausible factors.

2.2 Statistical analysis of explanatory factors for land use patterns

2.2.1 Methodology

One set of parameters for the CLUE-S land use change model is derived from regression
equations that describe the relationship between each individual land use type and a
relatively small but diverse number of “explanatory factors” or “location factors”. The
regressions attempt to quantify the relationships between the location of all land cover types
(dependent variables) and a set of explanatory factors (independent variables).

These regression equations are used to compute the relative suitability of a particular
location for each of the possible land use types during a simulated future scenario. The
regression coefficients are then input as model parameters. The regression analysis is one of
the most critical and comprehensive tasks during the preparation of the CLUE-S model.

The regression analyses were completed using the statistical program SPSS v 11.5. A
binomial (binary) logistic regression was used, as is appropriate when the dependent
variable is a dichotomy (i.e. 0/1 values for each land cover class). Unlike OLS (ordinary least
squares) regression, logistic regression does not assume linearity of the relationship

27

between the independent variables and the dependent, does not require normally distributed
variables, does not assume homoscedasticity', and in general has less_ stringent
requirements. It does, however, require that observations are independent and that the logit
(effect) of the independent variable is linearly related to the dependent.

The spatial relationships between land use and the selected set of variables were quantified
in a two-step procedure using binary logistic multiple regression analysis. Independence
between variables is a prerequisite for this method. The use of a stepwise regression
procedure solves multi-collinearity problems. In step one, significantly contributing variables
were selected with a stepwise forward regression, using the 0.05 significance criterion. In
step two, this set of variables was used to construct multiple regression equations.

The regression analysis was performed separately for every land use type and stratified by
dividing the study region into the four countries (or parts thereof).

The CLUE-S user’s manual (Verburg 2004) and the associated exercise 4, “How to do the
Statistical analysis” (Verburg et a/. 2004) explain how to conduct these analyses in SPSS.
The guidelines provided in these documents provided the basis for the analysis, though
additional online information proved useful.”

2.2.2 Evaluating statistical significance and goodness of fit

The output of a logistic regression in SPSS includes various statistics on the significance of
the individual regression coefficient and the overall fit of the regression equation. These are
found in the “Variables in the Equation” section of the output. The final regression model is
the last step model for which adding another variable would not improve the model
significantly.

2.2.2.1 Regression coefficients

The standard regression coefficients (standardized betas) are used to indicate the relative
importance of individual variables in a given equation. Note that you cannot compare the
various coefficients for the partial factor across rows. That is, the absolute value of a
regression coefficient is meaningless if it is not considered within the context of the total
number of significant factors and their respective importance.

2.2.2.2 Wald test

The Wald test is used to test the statistical significance of individual logistic regression
coefficients (8 coefficients) for each independent variable, i.e., to test the null hypothesis that
a particular logit (effect) is zero. A Wald test calculates a Z statistic, which is B / SE. Values
greater than zero indicate that their effect is not significant, and these independent variables
may well be dropped from the model.

Initially, all explanatory factors were included in the regression. When the results indicated
that a factor(s) was not statistically significant, the insignificant factor(s) was specifically
removed (i.e., not selected as an independent variable) and the regression analysis was
repeated. This process was iterated until all Wald values were zero or near-zero.

1 Homoscedasticity = constancy of the variance of a measure over the levels of the factor under study.

2 Other useful sources included http://www2.chass.ncsu.edu/garson/PA765/logistic.htm and
http://www.ats.ucla.edu/stat/spss/topics/logistic_ regression htm

28

2.2.2.3 R-squared

The adjusted coefficient of determination (R?), reported in the SPSS regression output,
serves as a measure for the amount of variation in the dependent variable that is explained
uniquely or jointly by the independents. However, note that it is a pseudo-R? that is not
equivalent to the R* found in Ordinary Least Squares (OLS) regression. Hence, this R?
statistic should be interpreted with great caution.

2.2.2.4 Relative Operating Characteristic (ROC)

The ROC characteristic is a measure of the goodness of fit of a logistic regression model,
similar to the R? statistic in Ordinary Least Squares regression. A completely random model
gives a ROC value of 0.5; a perfect fit results in a ROC of 1.0. The ROC was calculated only
for Belize and Guatemala as these datasets are relatively small. Attempts to calculate the
ROC for Mexico and Honduras resulted in the computer being locked for hours on end. The
ROC values should also be interpreted with care. For example, the equation for savanna in
Guatemala has an ROC of 1.0, which seems excellent, but is meaningless because the area
of Savanna is extremely small, so the regression equation and ROC are not significant.

2.2.3. Regression results

The statistically significant regression coefficients along with the total number of significant
location factors (NF) and the ROC statistics are listed in Table 2-1 to 2-4. The default number
of decimal values in the SPSS output was increased from 3 to 4 because some coefficients —
in particular for elevation and annual precipitation, which are relatively large numbers--- are
significant only at the third or fourth decimal.

There are no results for Water, because the regression analysis was not conducted for this
land cover type. Also, note that the results for mangroves and savanna in Guatemala are not
significant because the area of these land cover types is very small (respectively 0.8 and 0.3
km?). These data should be ignored, but are included here because the CLUE-S model
requires regression coefficients for all land use type present.

29

oe

(oNsUayoeseyo
Bunesedo
0c6'0 6r6'0 vv6'0 218°0 9b'0 9S8'0 vl60 062'0 986'0 anne|es) OOY
(s10}Oe}
uol}e90}
Jaquinu
jueoisiuBis)
8 6 OL LL 6 OL 6 LL LZ jeul4

80/9'0- - pbSl 0+ S9EL'0- 9826 '0+ eroco+ | (zeae | (e)9evst+ | LL19 0+ lea cae eae |
9106°0+ | (p) sogs"L- - (p) 9298°0+ SLLz‘O+ o6L0+ | (e) 9ppt'z- Lolei+ — | (z) 0994 b+ Eee Ss LSWdM ‘Ol
(¢) p900°%+ (z) 9eS¢°7+ 8872'0+ G80S'0- L691 L- - SZOv'L- ss 8bzr'0- (p) €6z0"S- SINNOL 6
Sausov 8
(L) zzso'z- | (L) oveze- | (z) ogge'o+ | (z) ggezo+ | (L)sospo- | (L)e66z‘0- | (L) vegso- | (z)Sez670- | (1) seae'o+ eS LMIWSOW ‘Z

eke He
LZ€Z 0+ v6S0'0- S260 0+ | OvZr 0+ (vy) yOZL"0- | (Z) 6z80°0 . OvL0'0+ (€) LL80°0+ 08@r 0+ s9dd1$ 9
(py) 85z0°0- = (L) LZ6L°0- (L) 9Z91°0- (¢) 98L0°0- | (pv) ZL00°0- €000'0- (L) b800°0+ Z000'0- (€) S6L2'0- 1IWAd14 ‘S
> €LZL‘O+ (py) 08S9°0+ - _|__ £660'0+ (€) 6Z6'0- Ory 0+ 2 8ES10- SbS6'0- NOWANG ‘¥
(Z) ZL00°0+ (€) »Z00°0+ (€) SLOO"0+ | (¢) g000°0- | L00'0+ © S000 ‘0+ 0L00'0+ (p) 7000'0- (L) 8£L0'0- YANIVY '€
Pees 9Z66'L- sosoe | otro | (2) vosrtz- €£000'0+ golgz+ | (p) eps'pz- Se90'e+ 9/80'E1- Nivuds ‘z
2910'0- : 900'°0+ vZ10°0- LL0'0+ 6810°0+ 60€0 0+ ZZv0'0+ ZLLO'O+ 5 H1daads ‘1
0L¥0'0- 0900°0+ = z9Z0 '0- Lyz0'0+ 01Z0'0- =] (Z) Z820°0- 00r0'0- 8200°0+ (Z) LZzL0- NaGdOd 0
| | BIC
LLZL'€- o9zP's- 9eve'S- pLLe + 8680't- L90L°S- OL8S z- zS€9'S- L2Sp'L- 690€ Zz yue}suog
Ja}e\\ ueqin sonoibueyy puejai euuevaes qnios ainjseqiBy }S9JO} BUld yso04 1g UuMOUyUL)
6 8 Z 9 S v € Zz 1 0

‘QOUBDIIUBIS JO [BAB] S}i JO JOJEDIPU! OU SI JUBIDIYJa09 UOISSa/Ba/ e JO ANjeA ajnjosqe ay} Jey] aJON ‘pjog
Arojeuejdxa jueoyiubis Ayeoysieys [je 10 S}UaI94ja09 uolssaJBad ay) ‘asn pue] WWeUAp Yyoed O04 ‘8Z1]Eg JJ SIS;

“inoy do} ay) ul aq Aew sanjed jews Ajaaije/as UAB OS
ul sauo jJueOYIUBIS JSOW JNOJ BY} YIM ‘pajsi] Be S1OJIeJ UOHEIO|

‘Ajeue uolssaibas oI6o] ay} Jo AIEWWWNS ‘1-2 e/GeL

Le

(o1NSUa}yOeseYyo

Bunesado

969'0 SSZ0 Z1L°0 anne|as) 904

1 (ss0}0B}

uoNed0|

ie) Jaquinu

yueoyiuBis)

Ol 9 OL OL L OL LL jeul4
L6CL t+ - - 96h 0+ ee (p)eatrt- | (e) szbz- ZSvdGM ‘}1
ZZ0S'1+ . 180S'\+ : econ - Z096'€- LSWdOM ‘OL
1L9€°0+ - = L06E'0+ 9SS0'L- - 9ssb't- Bees'0- SINNOL ‘6
sausov 8
evoe'o- | (4) zee8't- gssoo+ _| (z) 28910+ geco'o- | (t)vergo- | (Heszeo- | = ——_| (z) 80LF-0+ ZE8h'0- Lawsov ‘2 |
SE66°0+ OLzb'0- 1600+ Lzge0+ —_| (b) bzee'o+ 1811'0- 1981 0- 9L€0'0+ SP0E'0- s9ad1s 9
Z8910- (¢) vELO'O+ =| (L) Z06z°0- (L) ZSLL'0- (z) b900°0- (¢) Z800°0+ 7000'0+ INSS3aYd | (¢)ze00'0+ | (L) $¢0z'0- iWAd14 ¢
LZ610- (y) gzer'o- | (€) 8925°0- [@ LLS€'0- (¢) 99¢¢°0- (Z) S96p'0+ GLE0'0+ LON (v) 0Z02'0- (p) €Z20°L- NOWANG ‘7
9Z00'0- - 9Z00'0- 2000'°0+ | (») szoo'0- gooo'o+ _| (z) 6vo0'0- Pee | goo0'0+ _| (e) 9600°0- MANIVY '€
Goce |) | Iw ezeez+ | (e) zevere- 18bS"L- 1Z9L'0+ geuvi+ | | ozazo+ | (e) ogaree Nivuds ‘z
SEL0‘0- egioo- _| (z) s9vo'0+ IE 09€0'0- oa 1120'0+ | (v) 89¢0°0+ DEG L£60'0+ Hidads ‘1
ie oes | te) ee00'0+ 9000°0- 0000+ 0800'0- g£00'°0+ yoo: | ee] Le00°0- 1800'0- NaddOd ‘0
zS96'¢+ 61ZS'0- 6SOr' z+ 66ZE0- Ober I- LLY? O6rE'e+ SZLL'0+ L198°9+ jue|su0

JayeM ueain sanoibueyy puenai BUUeARS qnios ainseg/iBy | S810} Ould ‘\sa104 1g umouxun)
6 8 Z 9 G v 5 Z b 0

‘QOUBDIJUBIS JO [@A9| S}I JO JOJeIIPU! OU SI JUaIDYe0d UOISsa/Bal e JO ANjeA ajnjos
Arojeuejdxa juesiyiubis Ayeansieys [je 10f SJuaIIJa09 UOIssa/Bad ay) ‘asn pue) aJWeUAp Ydea JO “ODIXB/fy 40} SIS:

“Inoy do} ay] ul aq Aew! sanjen jjews Ajaaiejas UaAa OS

ge ay] Jey) aJON ‘pjog ul SauO jueDyIUBIS JSOW 1NOJ 94} YIM ‘pajsi) ase S4OJIEJ UOHEIO}

‘Ajeue uoissaibas a16o) ay} Jo ANBWIWINS -Z-2 a/GeL

ce

“(wy g'0 pue €'0 Ajannoadsas) jIu si ejewayend ul sad} Jen09 pue| asay) Jo eaJe ay) asneoaq aigeljaiun pue jueoyluBls jou ase senoiBuew pue euuenes Jo} s}insey |

(dijslUa}yOeseYyo
Buneiado
€96'0 71460 SSl0 8¢20 906°0 2820 0001 anne|a1) DON

(s40}0e}

uolje90|

Jo Jequinu

queoiyiuBbis)

OL OL v “AN Jeul4

it

V+ (€) S80L'L- (Z) 8ESbL- Or 0+ (¢) L96€°b+ 2 ZSVddM ‘11
QLL0'2+ L@zze- (€) s8Se"L- e (Z) SOL6"L+ LSVdGM ‘OL
SduSOV 8
| (2) orzgro+ | - glso'0- | (L) zsve"0- 61g0'0- | (L) 66ze"0+ - LINSOW ‘LZ
[ee ce ee ee ee | () Levoro- ZL10'0- 6L00'0- 9100'0- - s9ad1$ °9
6vE0'0- l()stooro- | - ~—| ( egtore- | (1) o¢o0'0+ | (1) 6000°0- z000'0- | (Lt) $z00‘0+ - (L) 826L"0- LvAga1a S
| veceo- | ezzeo+ | - (z) ogse"0+ LLOV0- L0ZZ'0- 90S0’0- (») vegz'0+ NOWANG ‘+
LL00'0- eset eee 9000'0+ Loo0'0- ‘| (z) 9L00°0- yooo'0- | (e) szeo'o- YANIVY “€
oe LL2S' b- 8L1Z'0- ozee'l- | (p) esee"L+ NIVUGS Z
00€0'0- 7010+ @8€0'0+ 28Z0'0+ (py) ebzo"0+ =| (bv) 8660°0+ 2€00°0- Hidaqs ‘1
$9S0'0- (L) €L00'0+ | (€) 0Z8S"0- (€) €900°0+ 6100°0- 7000 '0- (€) 9L00°0- 7000'0+ (2) osso'0+ | N3GdOd *
€886°6+ Gees LI- OLbp'es- pe | ee 916b'S- 1€96'0- S@vz'Z1- 1120'I- 126€ Lb yuejsuo9
{ag |e | —pueema —| — acre gnios aimseduBy | yep auld | yses0j Te | uMoUxuN |
ee a ee Se ee Oe g b € Z L 0

“unoy do} ay} ul aq Aew sanjea jews AJaAyejes UA OS
‘QOUBDIJIUBIS JO [AA9| S}I JO JOJESIPU! OU SI! JUBIIIYI09 UOISsa/Bas e JO AN/EA ajnjosge ay} Jey] BJO ‘pjOg Ul SaUO JUBDYIUBIS JSOW INOJ BY} YJIM ‘pa}Si] ae $10JOeJ UO!}EIO0]
Arojeuejdxa JuesyiuBis Ajeoaysyeys |j€ 10) Sjuaioya0o uolssei/6as ay) ‘asn pue] JweUAp yoea JO4 eyeurajzenyd oJ sisAjeue uojssai6ad o/60) ay} Jo AIBUIUINS °€-Z a/GeL

(OnsUajoeseyo

Bunesado

SAI}E[9J) 90uN

(su0}oe}

uolje00|

jo Jaquinu

yueoyluBis)

6 g Z 6 8 6 8 6 “4'N jeuly
100S°0+ ee oa (giz) 6019+ | OZLY'L- zsoe'o | (z)ezbe'z+ | (p) bL09"0- | (e) e9vsL- peesl+ | (e) poor i+ LSvdOM ‘OL
=e

eee - - - Opbr'|- - - - SINNOL 6
SGUHSOV 8

| (z) vszz"o- | (z) vors'e- | (y)sezz'o- | (v) sz90r0+ | (1) ogtz'o- 601¥'0- 6L0Z'0- gozo'o- | (b) 9ezz'o+ | (b) LLOLO- LIWSOV ‘LZ
€8610- , . vr6L'0- (py) p8gs"0- (¥) pbz"0- 2 8400 0+ 9/70 ‘0+ G089'c- soadis ‘9
(py) 7900"0- (¢) szoo'o+ | (€/z) S9LL'O- | (L) Zee0'0- 2Z00'0+ 6000 ’0- (L) bL00"0- (Z) 0L00°0+ (e) ZL00°0+ | (Z) z6b0'0- LWAd14
(e) L9SS*L+ @ZLZ 0- 0668 '0- (€) pL28°0- €S02'0- (¢) €862'L- (Z) Z0vZ°0+ vvv0'0+ (Z) 8Z02°0- LESL'0- NOWANG ‘7
(L) 9900°0+ 9100'0+ (pv) 0200°0+ (z) b€00'0- (e) 6£00°0+ | (L) SOL0"0- (¢) 2000°0+ | (L) 9200°0- (vy) 0L00'0- 6000°0+ YANIWY '€
©869°0+ Wiewer | 2 8) p1S0'L- 698e'Z+ Z0S9'|- 686L'0- pLZS'L+ Leos‘o- | () zpse"b+ NIVuGS ‘Z
$6Z0'0- S910'0- - - €150'0+ | 6800'0- 6S00°0- LLL0°0+ LLS0°0+ Hidads ‘|
6S00'0- (L) 9200°0+ 2 (Z) pSLL'O- vZ00'0- © (pv) $600°0- © 2000'0- NaddOd ‘0
ZS28'vh- 6868'S- 0zP0'9- LL6z°S+ LyS9'rL- 0662'S1+ €69b'L+ 9169 L+ G6Lb c- Z9€7 Ol- jue}suo9

JO}VeAA uedin senoibuey puepay =| euueres qnios ainjseq/By JSO10} BUI ysq04 “1g uMOUUl)
6 8 Z 9 g v € | Z L 0

‘QOUBDIJIUBIS JO [9A] S}! JO JOJEIIPU! OU SI} JUBIDIJa0I UOISSa/BaJ e JO AN/EA ajnjosge ay] JEY] 9JON |

“Inoy do} ay] ul ag Aew! sanje jjews Ajaaiejas UaAa OS
pjog ul Sauo JUeDYIUBIS JSOW INOJ AY} YIM ‘pa}si) ase S1OJOeJ UOH}EDO}

Arojeuejdxa JueIWIUBIS A/EDSHE}S |/e Jf SJUBINYJ20I UOISsaiBad ay] ‘asn puey BWeUAp yea JO4 ‘SeANPUOH 4/0) sisfjeue uoissaibal a16o] ay} Jo AIEWIWWNS p-Z a/QeL

3 GEO-4 scenarios and the ICRAN MAR project

Acknowledgements

With thanks to the Latin American scenarios working group for GEO-4, Barry Hughes, Bas
Eickhout, Joep Luijten, Lauretta Burke, Emil Cherrington, Jan Meerman, Gladys Hernandez,
Igor Lysenko and the GEO-4 Outlook team.

This chapter was originally released as: Miles, L., 2006. GEO-4 scenarios and the ICRAN MAR
project. UNEP World Conservation Monitoring Centre. Project report. 14 August 2006.

3.1 Scenario summaries

Four scenarios for the ‘Latin America and the Caribbean’ UNEP region have been drafted in
preparation for the publication of Global Environment Outlook 4 (GEO-4) in 2007. These
narratives were developed by the LAC scenarios working group for GEO-4. The scenarios
envisage differing social, political and economic trajectories, emphasising outcomes for the
environment and human well-being. Three of the four draft scenarios have been selected for
exploration of possible futures for land cover change within the ICRAN MAR project; the
fourth scenario (Security First) is not presented here.

The GEO scenarios consider the period from 2007 through to 2050 for the whole of Latin
America and the Caribbean, and encompass the overall interaction between human
development and the environment. The ICRAN MAR project considers the period up to 2025,
for the watersheds draining directly onto the reef and focuses on the impact of land cover
change on coral reefs. This chapter summarises and adapts the GEO scenarios with a focus
on this topic and timescale. The scenarios published in GEO 4 and in the forthcoming GEO-
LAC will therefore differ in many respects from those presented here.

It is assumed that climate change and variability is not susceptible to further human influence
up to 2025 — the change will occur has already been set in train. Changes in climate are
therefore identical throughout the scenarios; what varies is the resilience and response of
societies within each of the scenarios. For example, coral bleaching events can be expected
to increase in frequency in every scenario; but the approach to and coordination in tackling
the issue varies.

A comparison of modelled population and land cover changes up to 2025 for the scenaries
follows the narrative description; the methods are described in Sections 3.2 and 3.3.

3.1.1 About the GEO-4 scenarios

UNEP is working on the fourth Global Environment Outlook (GEO-4), for release in 2007, 10
years after the first GEO, and 20 years after the Brundtland report (WCSD, 1987). The
Global Environment Outlook process was initiated by UNEP for global environmental
assessment and reporting process, in response to several Decisions of the UNEP Governing
Council. The aim is to ensure that environmental problems and emerging issues of wide
international significance receive appropriate, adequate and timely consideration by
governments and other stakeholders. Projects are undertaken under the GEO programme at
global, regional and local scales.

There are seven GEO regions, each divided into subregions for finer scale analysis and
reporting. The Latin America and Caribbean (LAC or ALC) region is composed of the
Caribbean, Meso-America and South America regions. The Meso-America subregion is the
one relevant to the ICRAN MAR project, being composed of Belize, Costa Rica, El Salvador,

34

Guatemala, Honduras, Mexico, Nicaragua and Panama. Many GEO processes have been
undertaken in the LAC region. The most relevant for the ICRAN MAR project are GEO LAC
2000', GEO LAC 2003’, Caribbean Environmental Outlook (1999, 2005)? (includes Belize),
GEO Centroamerica 20047, GEO Biodiversidad (Centroamenica) 2003°, GEO Guatemala
2003°, GEO Honduras 2005’ and GEO México 2004°.

GEO-3 presented a set of divergent global scenarios running from 2002 to 2032: Markets
First, Policy First, Security First and Sustainability First (UNEP, 2002). These scenarios are
being updated and extended to 2050 for GEO-4, with the global narratives being based on
the work of seven regional working groups. Each regional scenario focuses on regional
priorities defined by contributors to the GEO process. The LAC group met first as part of the
global scenarios meeting in Bangkok, September 2005, and then in a follow-up meeting in
Trinidad & Tobago, in February 2006. Each meeting included representatives from
throughout the LAC region. In addition, feedback has been sought from a broader group
including the regional team working on the state and trends section of GEO-4.

The narratives will be represented in the GEO-4 report alongside a set of quantitative
outcomes. A process of reconciliation of the assumptions made in the different regional
scenarios and by the modelling team is currently underway, with the first order draft of
GEO-4 being circulated for review in May 2006.

It is anticipated that the adaptation of the GEO scenarios for the ICRAN MAR project will
render the project outcomes more immediately accessible to policy makers who have already
encountered the GEO work through UNEP’s outreach efforts. It also allows the ICRAN MAR
project to benefit from the substantial amount of work undertaken through GEO, including the
modelling of regional scale land-cover change within an integrated modelling framework. The
MAR project has discarded the Security First scenario, which results in a level of land cover
change in between those of Markets and Policy First.

In the following sections, the global overview of each scenario is presented as described in
GEO-3, and is followed by a regional summary based on the draft for GEO-4.

3.1.2 Markets First

3.1.2.1 GEO-3 scenario overview

“Most of the world adopts the values and expectations prevailing in today’s industrialized
countries. The wealth of nations and the optimal play of market forces dominate social and
political agendas. Trust is placed in further globalization and liberalization to enhance
corporate wealth, create new enterprises and livelihoods, and so help people and
communities to afford to insure against — or pay to fix — social and environmental problems.
Ethical investors, together with citizen and consumer groups, try to exercise growing
corrective influence but are undermined by economic imperatives. The powers of state

" http:/Avww.unep.org/geo/reqreports.htm
? http:/Awww.unep.org/geo/regreports. htm
http://www.unep.org/geo/regreports.htm
4 draft pdf obtained

° pdf obtained

3

8 http://www.pnuma.org/dewalac_ingles/quatemala03 _i.htm

y http://www.serna.gob.hn/documentos/GEO Honduras 2005.pdf

http://www. ine.gob.mx/ueajei/publicaciones/consultaPublicacion.html?id_ pub=448

35

Officials, planners and lawmakers to regulate society, economy and the environment continue
to be overwhelmed by expanding demands.” (UNEP, 2002).

3.1.2.2. MAR region summary (based on GEO-4 draft)

Economy and governance

Public policy is geared towards supporting commercial interests and promoting the
open exchange of goods and services. Social and environmental policies receive little
attention or financial support; it is assumed that economic growth is in itself a sufficient
route to progress.

Remittances (funds sent home by migrant workers) are more important than foreign
investment or aid; this is especially valuable for Mexico’s economy.

New industrial parks are built to entice national and foreign investment.

Tourist visits to the MAR region increase until around 2025. With limited regulation, the
impact of tourism on coastal ecosystems also increases. Visits then start to drop off as a
result of deteriorating habitats and increasing pollution.

Population and standard of living

Populations increase, but the growth rate slows with falling birth rates.

For all MAR countries, the highest rates of urbanisation are seen under this scenario,
with 80% of the regional population living in urban areas by 2025. Most development is
unplanned, and built on the coast or around the industrial parks.

Social services are reduced, and inequity in resource distribution increases.

Emigration increases, with people from all countries of Central America moving
northwards. This is especially relevant for Mexico, which after 2010 sees a lower rate of
national population growth within this scenario than in any other. Migration also occurs
within the country, with agriculturalists moving from the dry central region to the south,
including the Yucatan penisula.

Environmental impacts

Although sustainable development is much discussed, this scenario sees the greatest
rate of agricultural expansion. Rates of habitat loss, fragmentation and soil erosion
increase. Comparing the MAR countries, the rate of agricultural expansion is greatest in
Mexico, Belize and then Guatemala. However, Honduras sees the highest rates of
decrease in natural habitats, because the area remaining is already substantially
reduced’.

Agrochemical pollution increases, despite the influence of emissions standards.

The terrestrial protected area network expands slightly by 2025, to encompass 10% of
all biomes. For 20% of the new sites, natural ecosystems are successfully protected from
change over the scenario period. 60% are partially protected from change, and 20% fail
to be protected (see Section 3-4).

Water quality decreases and abstraction for tourism and agriculture increases, as a result
of limited interest in promoting good watershed management practices.

Both agricultural and natural ecosystems are vulnerable to an increasing frequency of
climate extremes. Fire frequency increases, especially in the dry forests of Honduras
and Guatemala.

1 56% of Honduras was already dedicated to agriculture by 2000, as opposed to 31% in Guatemala, in 19% in
Belize and only 6% in Mexico.

36

3.1.3. Policy First

3.1.3.1 GEO-3 scenario overview

“Decisive initiatives are taken by governments in an attempt to reach specific social and
environmental goals. A coordinated pro-environment and anti-poverty drive balances the
momentum for economic development at any cost. Environmental and social costs and gains
are factored into policy measures, regulatory frameworks and planning processes. All these
are reinforced by fiscal levers or incentives such as carbon taxes and tax breaks.
International ‘soft law’ treaties and binding instruments affecting environment and
development are integrated into unified blueprints and their status in law is upgraded, though
fresh provision is made for open consultation processes to allow for regional and local
variants. ” (UNEP 2002).

3.1.3.2. MAR region summary (based on GEO-4 draft)

Economy and governance

e Whilst many policies are more reactive than strategic, governments take a close interest
in social and environmental problems.

e Exports of primary goods continue to form a crucial part of the region’s economy, and
the tourism sector grows significantly with public support.

e By 2025, this is the scenario with the highest GDP per capita growth rates for Guatemala,
Belize and Honduras. For Mexico, Markets First has a slightly higher growth rate, partly
as a result of increased remittances from North America.

Population and standard of living

e Equity increases, with progress towards the Millennium Development Goals on
education, income and health. Emigration decreases as quality of life improves.

e Over the MAR region, population growth continues, but the rate of increase slows more
rapidly than in Markets First, especially in Honduras and Guatemala.

e Urbanisation continues, but is subject to stronger planning constraints.

Environmental impacts

e Land use becomes better regulated, especially around riverine corridors. Implementation
is patchy, but the rate of deforestation decreases. Over the MAR region, deforestation
continues to result in erosion and land degradation, but at a lesser rate than in the
Markets First scenario. In Mexico, forest cover decreases only until 2010, when an
ambitious national forestry plan reverses the trend. Mexican forest area surpasses 2000
levels by 2025.

e By 2015, cooperation on the management of transboundary watersheds develops in
the MAR region. Water quality increases as a result.

e Certification schemes for timber, agriculture and fisheries are encouraged.

e The terrestrial protected area network expands by 2025 to encompass 10% of all
biomes and all single-site endemic species by 2025. For 65% of the new sites, natural
ecosystems are completely protected from change over the period. 25% are partially
protected from change (allowing sustainable use), and 10% fail to be protected (see
Section 3-4). The marine protected area network also grows, with a focus on enhancing
resilience to coral bleaching’.

4 through reserve network design to optimise larval dispersal opportunities and to include more resilient reef types
(Schuttenberg 2001)

37

e Research is undertaken into adaptation measures to cope with the changing climate. By
2025, more diverse agricultural systems are being encouraged with the aim of resilience
to climate change impacts.

e Policies are adopted to assign economic values to coastal ecosystems such as
mangroves that provide protection from sea surges. However, coastal developments
continue to expand, and coastal degradation continues.

3.1.4 Sustainability First

3.1.4.1 GEO-3 scenario overview

“A new environment and development paradigm emerges in response to the challenge of
sustainability, supported by new, more equitable values and institutions. A more visionary
state of affairs prevails, where radical shifts in the way people interact with one another and
with the world around them stimulate and support sustainable policy measures and
accountable corporate behaviour. There is much fuller collaboration between governments,
citizens and other stakeholder groups in decision-making on issues of close common
concern. A consensus is reached on what needs to be done to satisfy basic needs and
realize personal goals without beggaring others or spoiling the outlook for posterity. ” (UNEP
2002).

3.1.4.2 MAR region summary (based on GEO-4 draft)

Economy and governance

Economic cooperation between the MAR countries increases.
Governments make a strong commitment to sustainable development. Efficiency in the
use of energy, land and material resources is promoted. There are efforts to adopt an
ecosystem approach to land use planning, with particular attention to watershed
protection. Awareness campaigns are directed both at industry and the general public,
and help to change consumption patterns.

e The tourist industry continues to grow, but smaller packages become more popular, so
that there are fewer large developments.

e For Belize, Guatemala and Honduras, GDP per capita growth rates are greater than
those for Markets First, but are slightly smaller than for Policy First. Most other quality of
life indicators are strongest under this scenario.

Population and standard of living

e Considerable resources are directed to poverty alleviation as the scenario progresses.
Many of the Millennium Development Goals are achieved by 2015, and further progress
is made by 2025.

e For Guatemala and Honduras in particular, this is the scenario with the lowest rate of
population increase. The rate of population growth in this scenario for Mexico is
therefore higher than in Markets First, partly because fewer people feel the need to
migrate to find work. Overall, population growth rates decrease.

e There is less growth in urban area within this scenario than any other; most urban
development is concentrated in medium and small cities.

38

Environmental impacts

e A shared environmental agenda arises in the region. National regulation and incentives
develop further to control pollution and generate local payments for local environmental
services such as water.

e Atnational to local scale, Agenda 21 gains strength, promoting involvement of community
and business groups in areas such as integrated land management. The rate of loss and
fragmentation of key habitats decreases.

e The move towards organic agriculture and the use of biological controls is unexpectedly
assisted by rising oil prices, which increase the cost of agrochemical use. Extension
services for these more sustainable practices develop. Food yields improve. The
combined impact of increased efficiency of natural resource use and ecosystem
restoration means that by 2025, agricultural area begins to decrease slightly in all MAR
countries.

e Several large Clean Development Mechanism projects are implemented, with forest
landscape restoration initiatives being particularly successful in Honduras.

e The terrestrial protected area system expands to represent all key regional ecosystems
and species, including more transboundary reserves. It includes at least 10% of all
biomes and all single-site endemic species by 2025. For 30% of the new sites, natural
ecosystems are completely protected from change over the period. 65% are partially
protected from change (allowing sustainable use), and 5% fail to be protected despite the
best intentions (see Section 3-4). The marine protected area network also grows, with no-
catch zones being established by local agreement to conserve fisheries.

3.2 Population and land cover change: comparisons between scenarios

This section summarises the population and land cover changes across the scenarios. Land
cover change was modelled using a combination of three models (for methods, see next
section). Figure 3-1 summarises the questions addressed using the different models. For
Mexico, the annual rate of agricultural expansion within Markets First was multiplied by 1.5,
to represent internal migration by farmers from the dry central parts of the country.

The rate of land cover change under the different scenarios was estimated for the whole of
the four countries based on results from IFs and IMAGE. The changes in land cover were
then applied to the watershed area, assuming the rate of land cover change within the MAR
model region would match that within the remainder of the countries. CLUE-S was used to
allocate land cover within the region.

Differences between the scenarios can more easily be seen by comparing the changes in
human population or land cover (Figures 3-2, 3-3, 3-4 and 3-4) than by comparing the total
population and area values (Figures 3-2, 3-5 and 3-6). Greater detail for land cover change is
available in the Section 3.3.

The population of all four countries continues to grow under all scenarios (Figures 3-2 and
3-3). The population figures shown here represent the whole countries, not just the MAR
region. The highest growth rates are consistently found in Guatemala and Honduras, but
there is high variation between scenarios. All except Mexico experience the smallest
increase under Sustainability First; for Mexico, Markets First is smallest. Variation in growth
rate between scenarios from 2005 to 2025 is smaller for Mexico than for other countries, with
Markets First at 17% and the other three scenarios from 21 to 22%. For Belize, rates vary
from 24% to 39%, for Guatemala from 44% to 68%, and for Honduras from 37% to 58%.

39

IMAGE

IFs
[Provides

socioeconomic

What is the rate
of change in the

drivers to IMAGE region? Baseline
What proportion ecosystem
What proportion of of change map

occurs in each
ecosystem type?

regional ___ change
occurs _in__ each

country?

CLUE-S
Where does land
cover occur
within the MAR
watershed?

Rate of change
per habitat
type

Figure 3-1: Role of three models used to simulate land cover change

The greatest increases in urban and agricultural land are seen under Markets First, followed
by Policy First. Where there is an increase in wildland under Sustainability or Policy First, it is
usually scrubland, which may regenerate to forest in time. For Mexico, forest area increases
under Policy First.

To illustrate the variation between scenarios and countries in detail, change in forest cover
can be examined. When considering total change in all forest classes, all countries lose most
forest in Markets First (Table 3-1). However, there are differences between the response of
the different countries to the different scenarios. Belize, Guatemala and Honduras all lose
least forest in Sustainability First (in the case of Honduras, there is an increase in forest
area), whilst Mexico still loses a substantial amount of forest to agriculture in that scenario.
Whilst there is a gradual decrease in area devoted to agriculture (including pastureland), the
major increase by 2025 is in the area of scrubland, rather than of forest (Figure 3-5). This is
especially true in Mexico.

Whilst these scenarios provide a range of outcomes, more radical changes are also be
possible. For GEO Honduras, the Polestar model was used to quantify the scenarios. !t
simulated a decrease in forest area of ~20% by 2020 for the Markets-First equivalent, and an
increase of 15% under the Sustainability-First equivalent.

Table 3-1: Forest cover change by scenario

Markets First Policy First Sustainability First

|
IMAGE
Central -12.5 -5.1 +1.6
America
Belize -6.2 -2.2 -0.2
Guatemala -9.2 -3.9 -1.3
Honduras -14.1 -7.0 +0.8
Mexico -3.5 arnt -2.1

Within any country, the MAR modelling framework allocates an equal percentage change to
broad-leaved forest, pine forest and mangroves between 2005 and 2025. However, the

Land unit Percentage change, 2005 to 2025
|

40

percentage change in the area of each forest type over the whole of the four countries
between 2005 and 2025 differs (Figures 3-9 through 3-12) because there is variation
between countries in the baseline forest area belonging to the three categories (Figure 3-8)
and in the percentage change allocated to that country.

Country: Honduras Country: Mexico
r 100
50
2
Qa
°
o
= 0
5 Country: Belize Country: Guatemala
=
100 7
504
Reha His, _ ot aiialinli
A oar Sai
3 s =i “ai & =
=< xe) re} x x) a
© a © G a ©
= € = = Gm 2005
a7) a7) Gam «2025
=] é 3
a Scenario 79)

Figure 3-2: National human population at 2005 and 2025 by scenarios (IFs)'

‘Belize population is modelled as 0.26 million at 2005, and at 2025 varies little, from 0.33 million (Policy First,
Sustainabilty First) to 0.34 million (Markets First).

41

Scenario: Sustainability 1st

80 -

40

Scenario: Policy 1st

| 80

; 40

Scenario: Markets 1st

Increase in population, 2005 to 2025 (%)

Honduras |

au}
o
iS
i)
2
o
3
oO

Country

Figure 3-3: Percentage change in national populations, 2000 to 2025, by scenario (IFs)

Scenario: Sustainability 1st
6000

2000 |

-2000 7
-6000 + ———— ee
Scenario: Policy ist
an iL
= i 6000
SS L
= 2000
2
< -2000
-6000

Scenario: Markets ist

— ~ ~ o a oO Q n c

no n o
ENS Oa Mre) oaks RE Gs boo ae maees
£ 2 2
Oo <2 Cees . ome mete ee Pe
So = o a + a Q = ee a

ne} £ > > ao o

3 & < Oe oth =

ao N oO s ~

= oO

Country

Figure 3-4: Change in land cover, 2005 to 2025, all countries combined

42

Scenario: Sustainability 1st
80 7
60 7
40 |
w 20 |
NX ecarrras)
a OF
2 Scenario: Policy ist ca
So
Ss r 80
x F 60
2
© t 40
£ ees)
s fee Scenario: Markets ist
=X go 4
60 7
40 7
205]
0
2-2 BR Po as eee Seta, aS
S oT o Q + w D 2 oo a
32) S To n 5 o
oO ac
rS) a. < ‘D oD =
oO N oa = ~N
i Land cover ©

Figure 3-5: Percentage change in land cover, 2005 to 20285, all countries combined

Area, 2005, km2

0.Other
1.Broad-l forest ~
2.Pine forest
3.Ag/pasture ~
4.Scrub
§.Savanna
6.Wetld/Swamp
7.Mangroves
8.Urban
9.Water

Land cover

Figure 3-6: Land cover for watershed area at 2005; all countries combined

43

Scenario: Sustainability 1st |

2025 area, km2

80000 7
40000 |
0- —- =
o 7) 7 o 2 © a o c ro
AE BAM IGAY SS WPR Pye)! Re Rane Etc mE:
Oi ye Sin Eh a Bide ase, hier Meiees wei ans
i) or o Q + & (2) = a o
3 £ oS n cs] o
Cee es aes De or r=
Cie PE aa a
a Landcover ©

Figure 3-7: Land cover for watershed area at 2025 by scenarios, all countries combined
3.3 Land cover quantification: the IFs and IMAGE models

3.3.1 Model background

Two of the models used in support of GEO are relevant to the MAR project. IMAGE-2' is a
gridded integrated assessment model, operating at the global scale. It is able to simulate
issues like the impact of global climate change on crop production. International Futures
(IFs)* is a ‘macro-agent’ based model, also operating at the global scale, but at the resolution
of countries rather than on a spatial grid. It represents major agent classes (households,
governments, firms), simulating relationships in a variety of global structures (demographic,
economic, social, and environmental) (Hughes 2004). It is available online for use in scenario
exploration and teaching. IFs provides the driver variables for the GEO-4 scenario. IMAGE
projects land use based on these drivers and other interrelated factors, using a half-degree
grid, but its outputs are intended to be interpreted on a regional scale.

Within the ICRAN MAR project, the Conversion of Land Use and its Effects (CLUE-S) model
has been selected for land cover change modelling on a 250-m grid. The GEO models are
used to obtain percentage change in land cover types (rather than area of change) through
time, to drive the CLUE-S model.

Each model has been configured independently, so uses its own land cover classification.
The land cover classes have been mapped onto one another to give a minimum set as
shown in Table 3-2. The major assumptions are that (i) despite inconsistencies between land
cover definitions, the ratio of change in land cover between countries and the Meso-America

i http://www. mnp.nl/image/
2 http://ifsmodel_ora/; http://www. ifs.du.edu/

44

region within the IFs model is still a good proxy for the ratio within the IMAGE model; (ii) the
relative change between land cover types within the IMAGE model is a reasonable indicator
of the change between equivalent types within the CLUE-S model. The ‘other’ class within
CLUE-S and the ‘other class within IFs represent rather different concepts, and do not map
onto one another. The ‘other’ class within CLUE-S represents only 0.25% of the land area for
the four countries at the baseline year, and is not allowed to change in area. In IFs, the
‘other’ class represents 18% and is subject to change. Here, the IFs class is used to assist in
calculating change in the scrub, savanna, wetland and swamp categories within CLUE-S.

The baseline year for Belize is 2004; for Mexico, Guatemala and Honduras it is 2000 (Figure
3-8). These are the years for which the latest Ecosystem Map land cover data was available
(Meerman & Sabido 2001, Vreugdenhill et a/. 2002).

IMAGE simulates a historical 8.7% loss over the whole of Central America from 1990 to
2005. Looking into the future, IMAGE simulates a 12.5% loss in forest cover from 2005 to
2025 under Markets First, a 5.1% loss under Policy First and a 1.6% loss under
Sustainability First.

In IFs, forest area is initiated using FAO data, with simulated changes being dependent upon
the rate of conversion to cropland and grazing area. This rate is driven by agricultural supply
and demand, based upon factors such as human population and land development costs.
Urban area expands into all other land cover classes equally. The IMAGE model, conversely,
uses a terrestrial vegetation model factoring in impacts of climate and soils. As IMAGE does
not model urban area changes, IFs values have been used for change to urban land.

Country: Belize Country: Guatemala

37500
30000
22500
15000
g 5 7500
x ee
g 2
< 45000

7500
0 aa ee]
o¢e¢eseee & 8 SR Pease SS Aes
Ss Se Se es SS isu suns! ustee «Oe 2) 0s
Od 8 Bow S Sub se Ss © & 28 O88 pS Ss
GO & fF ES i @ Fe 8 Sy Sr Gy ey Se ay FS
Ht = HQ os $ a n = $
so < o Fs = Ry GE ce o Ze
mi) nN ied) =~ N oO
a Ss Landcover © 2 Ss
= © = o

Figure 3-8: Land cover at baseline year (2004 for Belize, 2000 for Guatemala, Honduras and Mexico)

45

3.3.2 Bringing the models together: methods

The land cover values for the future scenarios as applied to CLUE-S are derived by
allocating the percentage change as seen in IMAGE, distributed between countries
according to the proportionate national changes in IFs. The land cover types differ between
the three models, and are mapped onto one another as shown in Table 3-2. The resulting
values are used to drive the CLUE-S land allocation routine as described in detail below.

Table 3-2: Mapping of land cover types between IFs, IMAGE and CLUE-S
Land cover type | Land cover | Land cover types (IMAGE) Assumptions

(CLUE-S') type (IFs) for CLUE-S
application
Nia N/a CLUE-S requires

0. Other/Unknown no change to this
NO CHANGE pi class
1. Broad-leaved | Forest Carbon plantations Equal probability
forest Regrowth forest (abandoned) of change of
2. Pine forest Regrowth forest (timber) CLUE-S types

Warm mixed forest

Tropical woodland

Tropical forest

[On a global scale, this category
would include other forest types
not present in the Meso-America

7. Mangroves region
Food crops IFs types are
Biofuel crops subtypes of
pasture Grazing Grass and fodder CLUE-S type
4. Scrub Other Scrubland a ee
Other IMAGE savanna, desert, | Equal probability
5. Savanna grassland/steppe [on basis that it; of change of

will include wet grasslands] CLUE-S types

6. Wetland/swamp

Urban Excluded from IMAGE by|IFs_ increase in
reducing land area per cell| urban area is
accordingly; not modelled in| applied directly,

future. with the
expansion
reducing the
8. Urban ‘other’ category.

N/a N/a CLUE-S requires
9 Water (NO no change to this
CHANGE class

1 The 10 land cover classes used by the CLUE-S model are not dictated by CLUE-S. Instead, this is a reduced
classification of the land cover classes of the source ‘Ecosystem Map’ land cover dataset, which was developed
and agreed upon by the watershed partners.

46

For any given year, scenario and IFs land cover type:

f = area for country in IFs (mill ha)

F = area for region in IF s (mill ha) = 2f

m = area for country equivalent to that in IMAGE (mill ha) [derived in this exercise]

M = sum of area for appropriate categories from Table X for region in IMAGE (mill ha)
=2m

c = area for country in CLUE-S (km?) [derived from map data for year 1, and for later
years in this exercise]

C = area of country

i) Estimation of area in IMAGE for country in year n

Foe urban land (see Table 3-2)

Myn (urban) fyn

where fyn (urban) = fyn-1 (urban), SMOothing was carried out to ensure percentage change in cover
was not zero in alternate years (this was an issue for the very small amounts of urban land
cover in Belize).

For grazing, forest, crop land categories from IFs
Myn = Myn (fyn/F yn)

For scrub and savanna / wetland categories from CLUE-S and IMAGE, taking national
proportions from the Other category in IFs:
Myn = Myn (fyn/F yn)

ii) Percentage change assigned to CLUE-S for year n
= 100 (Myn - M yn-1)/ Myn-1

iii) Land cover assigned to equivalent CLUE-S categories for year n. Area per class is then
normalised to country area, based on the fraction represented by that land cover class in that
year for that country. The area belonging to ‘other’ and ‘water’ CLUE-S categories do not
change between years.

d yn = Cyn-1 + C yn-1 ((Myn -m yn-1)/ Myn-1)

Cyn= d yn + ((C - 2d yn all classes) ~ dyn) /C)

Figures 3-9 through 3-12 show the calculated change in land demand for each land use type

under all three scenarios. The land demand each year for each land use types is given in
Appendix 5.

47

3.4 Future changes in protected areas

A protected area scenario dataset was created based on the scenario assumptions. These
maps represent one hypothetical expansion of the network, rather than a recommended set
of designations. No distinction is made between managed and unmanaged forests within
CLUE-S, but a distinction can be made between the different protection categories. Existing
and new protected areas are allocated to the following categories within the scenarios:
- Sustainable-use (probability of conversion is reduced by designation; driven by the
logistic regression)

- No-use (no change after designation to natural ecosystems contained within the area)
- Failed (no protection from land cover change)

The protected areas are implemented within CLUE-S as follows:

1) all no-use areas are designated at the start of the scenario period. No land use change
occurs within the natural ecosystem cells inside these areas over the period. This category is
applied to IUCN categories I-IV and uncategorised protected areas.

2) New sustainable-use areas may be designated at any time. These areas are applied
within the model as a dynamic factor grid, which influences the probability of land cover
change, rather than via rule-based restrictions. This is therefore a ‘partially protected’ rather
than a strict ‘sustainable use’ designation. Applied to categories V-VI.

In summary, areas are assigned to the scenarios as follows:

Markets First — expansion of terrestrial network to 10% of all biomes/countries by 2025;
new sites allocated as 20% no-change, 60% sustainable-use, 20% failed

Policy First — expansion of terrestrial network to 10% of all biomes/countries + all single-
site endemic species by 2025 and 20% of all biomes/countries by 2050; new sites
allocated as 65% no-change, 25% sustainable-use, 10% failed

Sustainability First - expansion of terrestrial network to 10% of all biomes/countries + all
single-site endemic species by 2025 and 20% of all biomes/countries by 2050; new sites
allocated as 30% no-change, 65% sustainable-use, 5% failed

Failure indicates that there is no barrier to land use change in this protected area.

In Policy First and Sustainability First, the new protected areas are first allocated to priority
areas for biodiversity to attain at least 10 percent of each biome/country combination.
Additional areas are then allocated to cover single-site endemic species that have not
captured, based upon the Alliance for Zero Extinction point dataset. These additional areas
are circles of equivalent size to the area required for that species (or for the mean area
where this is not specified), thus giving an artificial appearance to the scenario data. The
coverage of some biomes is therefore expanded to greater than 10 per cent by 2025 within
these two scenarios. A number of protected areas were assigned outside the MAR region of
Guatemala, Honduras and Mexico.

48

Scenario: Sustainability 1st
1000 4
0+ eee See — ———

-1000 |
S -2000 -——
<3 Scenario: Policy 1st
°
5 - 1000
£ EEE es |
© Psst) TS i 0
2 + -1000
6 + 2000

Scenario: Markets ist
1000 4
-1000 | i

-2000 ~

Put ee bite occ LL ca) Bete e
£ oe Q 2 = c G ° 2 ©
(2) fe) 7) -® G S =
fe) = = @ n > Ss > =) s
fo} ere o Q + © 22) oO oO fo)
s) £ ° 2 3 Ss
8 o < o oD ne

9 :

i, MR Se S

Vr o
Land use

Figure 3-9: Change in land cover for Belize, 2005 to 2025, scenarios

Scenario: Sustainability ist

Change in area, km2

Scenario: Markets ist

1000 |
0 ee T T ——— T
-1000 4
-2000
as D on oO 2 is.) Qa oO c i
2M AB aiiehd pet teow Je enE ys tga!
= 4 = is) oO 2 2 oO
fo) 2 2 7 7) 7 2 a > Ss
o oP o Q + a (2) = Se oO
Z £ zr) n co] =
3 a << wo oD nN
° :
Cre hs a S
= 2
Landuse ©

Figure 3-10: Change in land cover for Guatemala, 2005 to 2025, scenarios

49

Scenario: Sustainability 1st

3500
1500
-500 Sant sanecih Godel howe = a
-2500
N
E
= |
g
% 3500
& 1500
o
2 -500
=
(S) -2500
Scenario: Markets ist
3500
1500
-500
-2500
Opps 2ewtical iO eS igie oa Eee
fo) =a ® = vt 3 g o co fo7)
hd a <x wo = =
e) 7
a N o = =
‘s Landuse ©
Figure 3-11: Change in land cover for Honduras, 2005 to 2025, scenarios
Scenario: Sustainability 1st
= -2000
as Scenario: Policy ist
3 |
F 1000
£ ie | aa pL Ag Ne Oe ll 0
Ss fb
£ r -1000
rs) — -2000

-2000

ny =
o 7 G 2 $s 2 = g 7 ®
ad ® o 5 = £ (= 3} i a
= 5 oO nD [s) o oO = _
OO al) Or a aa a eae ES
cS} z ® £ + a g G } ro)

Ss @ 2 epee

) - :

OF ee cg =

oO
aa Land use

Figure 3-12: Change in land cover for Mexico, 2005 to 2025, scenarios

50

4 Modelling land use changes for scenarios using CLUE-S

4.1 Important aspects of model development

4.1.1 Land use data

In working with CLUE-S, it became clear that the Ecosystem Map land cover data use were
not ideal for use by this model. The model had difficulty with the enormously skewed land
use distributions and with the large homogenous areas, particularly in Honduras, which is in
part a consequence of the relative large minimum mapping units that were used in the vector
Ecosystem Map dataset (10 ha for Belize; 150 ha for the other countries). It is likely that the
model would have had less trouble with a remotely sensed classification rather than raster
data converted from vector data. It is therefore strongly recommended to use different, more
detailed remotely sensed data if the study is to be undertaken or refined in the future.

4.1.2 Probability Surfaces

The single most important aspect of the model development process is to inspect and verify
the probability surfaces that CLUE-S uses to allocate land. These surfaces are constructed
from the regression parameters and they directly reflect the goodness of fit of the regression
equations. They are extremely helpful in assessing whether the way the model is going to
allocate land makes sense. If this is not the case, the regression equations should be
inspected for any possible typographical errors or, in case they are not statistically
significant, replaced by a better equation. Unfortunately, the Relative Operating Statistic
(ROC, introduced in Section 2.2.) could be calculated only for Belize, so we do not know the
goodness of fit for the equations for the other countries. Therefore, the manual inspection of
probability surfaces is essential. Adjustments in the model were made where necessary.

Not surprisingly, the probability surfaces for the Belize looked most plausible, whereas those
for Honduras were worst, with Guatemala and Mexico in between. The simulation of
Honduras posed a particular problem. The probability surfaces for all but Broadleaved
Forest, Pine Forest, and Agriculture\Pasture looked quite unlikely or clearly incorrect. The
cause is twofold. First, the three dominant land use types occupy over 95% of the total area,
the remaining seven types occupy < 5% of the area. The regression results are not
significant for those seven types. As a consequence, the model was not able to reach a
solution, i.e., it was unable to properly allocate the demanded land-use based on the
probability surfaces.

Two workarounds were applied:

e lf the difference in the calculated demand for a land use types hardly changed
throughout the simulation period (2000/2004 to 2025), then the land demand was
kept constant and the changes that should have occurred were added to another
(most comparable) land use type to ensure that the total area remained the same.
The regression equations were not changed, because by keeping the land demands
equal the equations were essentially bypassed and not relevant.

e The regression equations for urban land in Mexico, Guatemala and Honduras were
replaced by equations that made more sense. A generic regression equation based
on population density, distance to markets and the presence of tourism hotspots was
adopted. This is explained in more detail in Section 4.3.4.

52

4.1.3

Neighbourhood settings

The user manual and available parameters suggest that CLUE-S can allocate land use in a
user-specified (e.g, a 3x3 or 4x4) neighbourhood. Specifying such neighbourhoods (in
NEIGHMAT.TxT) and giving a weight of 1 to the neighbourhood suggests that the probability
surface are constructed in a way that reflects the neighbourhood settings, and that new land
use can only be allocated inside the that neighbourhood. This would be useful if you require
new grid cells of, for example, forest or urban, to be adjacent to existing cells. However,
Peter Verburg confirms that land allocation cannot be influenced in this way.

4.2 Input data preparation

4.2.1

Files used by CLUE-S

Table 4-1: Input files used by CLUE-S. The “created” column indicates which software is used to
create the files and the “mandatory” column indicates whether the file is a minimum input data
requirement. Files created using CLUE-S are plain text files and may also be edited in a text editor.

Description

Filename
MAIN.1

Created | Mandatory
Main parameters file. Listed on exactly 19 lines. Some | CLUE-S yes

parameters settings will dictate whether the optional files
must be specified or not.

ALLOC1.REG

ALLOC2 .REG

ALLOW. TXT

NEIGHMAT. TXT

REGI* .*

Regression parameters. Length of file depends on
number of land use types and location factors.

Neighbourhood results. These are additional
regression parameters based on the enrichment factor
equation. NOT USED

Change matrix. The number of rows and columns equal
the land cover types, here 10x10.

Neighbourhood settings. Defines the shape and size
(in the form of a small weight matrix) of the analysis
neighbourhood for every land use type.

Area restriction file. A grid that defines where land use
changes can and cannot occur. The * is a wildcard here;
it does not indicate the simulate year. All active cells
must have the value 0, restricted cells a value of -9998,
and all others cells -9999 (NoData).

| DEMAND IN

COV_ALL.0

SC1GR#. FIL

SC1GR#.FIL
(SAME
ABOVE)

AS

SC1GR#. *

es

ArcView

Land use requirements. Calculated at the aggregate
level and organized by rows (simulated years starting at
0) and columns (for every land use types). The * denotes
a unique number, not simulated year.

Initial land use. A grid of all land use types at the start
(year 0). Grid values must match the land use codes
listed in the main parameters file.

Excel

ArcView

yes

yes

Static location factor grid, where # is the number of the
location factor; or

Land change restrictions grid. Grid that indicates
whether a land use conversion can occur. The # does
not denote a location factor. Instead it is a unique values
that is also specified in the alloc1.reg.

Dynamic location factor grid, where # is the number of
a location factor. The * is the simulated year starting at 0,
not_a wildcard. Note that also the file srcigr#-fill is

ArcView

ArcView

ArcView

yes

53

needed and it is identical to src1gr#.0.

LOCSPEC#.FIL | | ocation specific preference addition. Grid used to | ArcView

increase the probability at specific locations. Must be
specified for every land use type. NOT USED.

4.2.2 Land use requirements for different scenarios

4.2.2.1 Requirements calculated using IMAGE model

Land use requirements were calculated by Lera Miles using outputs from the International
Futures and IMAGE and GEO-4 models, as part of the scenario development process
(Section 3.3).' Tables 4-2 through 4-5 give the percentage area of each land use type at the
present time along with future demands. The precise demand values for every land use type
in every simulated year are given in Appendix 5.

Table 4-2: Belize: Distribution of present land use and land demand for the scenarios. Blue coloured
land use types were kept fixed at present values and not allowed to change over time.

Present Markets Policy Security Sustain.
0. Other/Unknown 0.06% 0.06% 0.06% 0.06% 0.06%
1. Broad-leaved forest 58.02% 54.33% 56.69% 55.16% 57.78%
2. Pine forest 3.53% 3.31% 3.45% 3.36% 3.52%
3. Agriculture/pasture 19.37% 23.20% 20.64% 22.70% 18.85%
4. Scrub 1.26% 1.23% 1.13% 1.10% 1.37%
5. Savanna 8.63% 8.23% | 8.35% 8.11% 8.65%
6. Wetland/Swamp 4.26% 4.07% 4.12% 4.01% 4.27%
7. Mangroves 3.29% 3.08% 3.22% 3.13% 3.28%
8. Urban 0.87% 1.79% 1.64% 1.68% 1.51%
9. Water 0.70% 0.70% 0.70% 0.70% 0.70%

Table 4-3: Mexico: Distribution of present land use and land demand for the scenarios. Blue coloured
land use types were kept fixed at present values and not allowed to change over time. Note that there
is no pine forest in Mexico; this required some adjustments to the model.

2. Pine forest

3. Agriculture/pasture

4. Scrub

Present Security Sustain. |
0. Other/Unknown 0.45% 0.45% 0.45%
1. Broad-leaved forest 57.33% 54.43%

5. Savanna
6. Wetland/Swamp

7. Mangroves

0.26%

| 9. Water

1.00%

4.18% 4.02%

0.43%
1.00%

25.96%

30.84%

0.11%
3.48%

4.14%

0.11%
3.46%
3.97%

' CLUE-S does not dictate a particular method for calculating the land use requirements. The use of IMAGE and
International Futures was possible only because we had access to these models. Simpler methods are possible

and recommended. For example, land use requirements could be calculated using appropriate economic demand
models or, very simply, by setting hypothetical land use requirements for the scenarios end year and interpolating
the land requirements between start year and end year using a linear or exponential growth model.

54

Table 4-4: Guatemala: Distribution of present land use and land demand for the scenarios. Blue
coloured land use types were kept fixed at present values and not allowed to change over time. The
area savanna is very small but not exactly zero (the distinction is significant).

Markets Policy Security Sustain.

Present

0.02% 0.02% 0.02% 0.02% 0.02%
51.11% 45.76% 48.61% 46.63% 49.74%
2. Pine forest 2.36% 2.26% 2.41%
3. Agriculture/pasture 31.00% 36.27% 34.15% 36.64% 29.81%
4. Scrub | 12.66% 12.56% 11.31% 14.97%

| 5. Savanna | 0.001% 0.001% 0.001% 0.001% 0.001%
6. Wetland/Swamp 0.04% | 0.04% 0.04% 0.04%

7. Mangroves 0.00% 0.00% 0.00%
0.71% 0.76% 0.66%
2.34% 2.34% 2.34%

Table 4-5: Honduras: Distribution of present land use and land demand for the scenarios.
Blue=forced fixed at initial area, not allowed to changed over time.

[Present] Markets| Policy | Security| Sustain.

0. Other/Unknown 0.29% 0.29%

1. Broad-leaved forest 25.94% | 22.47% 24.42% 22.90% 26.64%

2. Pine forest 15.39% 13.34% 13.59% 15.81%

3. Agriculture/pasture 53.74%

4. Scrub 0.23%

5. Savanna 1.41% 1.27% 1.32% 1.24%

6. Wetland/Swamp 0.60% 0.54% 0.56% 0.52%

7. Mangroves
0.32% 0.32%

9. Water 0.74% 0.74% 0.74%

4.2.2.2 Adjustments in land demands for CLUE-S model

The initial plan was to use the calculated demands in Appendix 5 as the land demand files
for CLUE-S. From the initial model runs, it became clear that the model could not always
reach a solution. This was caused by some land use types that occupy only a small fraction
of the total area and hardly change over time. The annual changes may be close to within
the convergence criteria of the model. Examples include scrub and mangroves in Honduras
(Table 4-5). This adds complexity to the model without yielding any benefits - the input data
made it difficult for the model to reach a solution.

Some adjustments were made to the land demands that provided a workaround for this
issue. The adjustment was to keep the area of certain land use types constant, and adding
the hectares that should have changed to those of the most similar land use type (so that
total area remains constant). Note that the demand for Other and Water is always kept
constant. Table 4-6 indicates which land use change demands were kept constant. For
example, for Mexico, Savanna was fixed and the change in Savanna that should have
occurred was allocated to Scrub (that is what the +#5 means). Wetland was also kept
constant, with the change that should have occurred being added to mangroves.

Table 4-6 shows that generally, along with Water and Other, the land use types that occupy
< 0.5% of the total area were kept constant. Wetland in Mexico occupies a larger fraction
(3.47%), but was kept constant as it hardly changes across all scenarios (Table 4-3).

Note that Urban land was considered too important to keep constant, even though the area

can be quite small. Instead, problems with the allocation of Urban land were tackled by
modifying the regression equations.

55

Table 4-6: Present land use distribution with red coloured values for those types for which the demand
was kept constant over time because the demand changes were smaller than the iteration tolerance of
CLUE-S, or so small that the model was prevented from reaching a solution. The required change in
land was added to another land use type, as indicated within parenthesis.

| Belize Mexico
0. Other/Unknown 0.06% 0.45%
1. Broad-leaved forest | 58.02% 57.33%
|

| 2. Pine forest 3.53% 0.00%
| 3. Agriculture/pasture 19.37% 6.13%
| 4. Scrub 1.26% (+#5) 27.06%

Guatemala Honduras

0.02%

25.94%
15.39%

(+#5,6,7) 12.66%

5. Savanna 8.63% 0.11% | (+#4) 1.41%
6. Wetland/Swam 4.26% 3.47% 0.60%

7. Mangroves 3.29% (+#6) 4.18%
8. Urban 0.87%
9. Water | 0.70%

4.2.3. Main model parameters

The main parameter file has 19 lines with numbers. For further details, please see the User’s
Manual. Below are the main parameters used for Belize. Only lines 5, 6, 8 and 9 are different
for the four countries. These parameters can be read from the header of the ascii grid files.

Table 4-7a: Main model parameters as used for the simulations.

| Line | Parameters | Description

10 Number of land use
Number of regions
Max number of independent variables in eq
AP Total number of driving factors

Number of rows

Number of columns

Cell area in ha (250 m grid)

Xll coordinate of grids

Yll coordinate of grids

Number coding land use types

Codes for conversion elasticities

Iteration variables for land use types

Start and end year of simulation

Number and codes of dynamic explanatory factors
Output file choice

Region specific regression choice

Initialization of land use history

Neighbourhood calculation choice
Location specific preference addition.

country-dependent
country-dependent

6.25
8 country-dependent

9 country-dependent
10 0123456789

' The start year for Belize was 2004.

56

Table 4-7b: Country specific variables
Belize

Line 3. nvariables
Line 5. nrows

Guatemala

Honduras

Line 12. Iteration

Line 6. ncols 604 1262 1310 2131
Line 8. xllcorner 261500 213250 41250 260250
Line 9. yllcorner 1757500 1971250 1596500 1521000

Some variables proved more challenging:

4.2.4

Line 10: Number coding of land use types. In principle very straightforward, but note that
the numbering must be without any gaps. Mexico has no pine forest (land use type 2)
within the MAR, yet this land use must be included.

Line 11: Conversion elasticities. Too many values of, or close to, 1 will stabilize the
system and may prevent the model from reaching a solution. The elasticities had to be
relaxed.

Line 12: Iteration variables. Settings that are too relaxed could result in land use
allocations different from the expected one (as specified in the land use requirements
file), whilst settings that are too strict could significantly increase the simulation time --
because more iterations are needed each time set—or cause the model to fail to reach
the desired iteration’. The default settings (0.3 and 3.0) were a good starting point and
only some minor adjustments had to be made.

Line 18: Neighbourhood function. Was not used because there was _ insufficient
information available about the potential influence within the neighbourhood. See section
4.1.3 for further information about neighbourhood settings.

Line 19: Location-specific preference addition. A potentially powerful feature, which was
not used because insufficient spatial data was available to support it.

Regression parameters

The regression equation parameters (Tables 2-2 through 2-4) were reformatted to
ALLOC1.REG files for each country. Although this is, in principle, a straightforward task, initial
model runs indicated that for all countries but Belize, the model was unable to reach a
solution. The land use allocation module was not able to significantly change the land
allocation even after thousand of iterations.

' These iteration variables are the smallest settings with which the model was able to reach a solution. The
variables are slightly different among countries, which is a reflection of the different country sizes, land
distribution, and in particular smallest and largest (percentage wise) land allocation.

? The model execution stops if no solution has been reached after 20 000 iterations for any simulated year.

57

Numerous tests runs with adjusted model parameters were run to identify the cause’. It was
concluded that the problems were the combined effect of the following aspects of the model
and the MAR data:

Regression results for land use types with a small percentage of the area, such as
Urban, were inaccurate because the sample size that the regression is based on is
too small. One solution is to substitute a “common sense” equation.

The land use distribution of all four countries is extremely skewed, with the most
dominant land use type occupying over 50% of the area and the least dominant one
less than 1%, and Mexico, Guatemala and Honduras have one or more land use type
occupying less than 0.1% of the area (Table 4-6). The results of the regression
analysis are naturally insignificant for the least dominant land use types and the
probability maps that CLUE-S calculates for those land use types not significant
either and in fact very incorrect. They have predominantly zero values.

The CLUE-S allocation module seems restricted in that it may be unable to make
simultaneous adjustments in allocated areas across a wide range’. For example, the
model had problems making an adjustment of < 10 ha in one land use type and
5 000 ha in another land use type during the same iteration. Consequently, the model
is unable to arrive at very small demand figures and the maximum allowed deviation
of those land use types is almost always exceeded, causing the model to iterate until
the maximum of 20 000 iterations is met.

The underlying problem may by relatively small iteration variables. These variables
(the numbers between the “*” columns in the LoG.FIL) have only three decimal
places, and if the operating values are only significant at the third decimal, it is very
difficult for the model to make small adjustments. For example, if an iteration variable
is 0.002 then a change to 0.003 represents a 50% adjustment in the area.

As a workaround the following changes were attempted:

First, if initial model runs show that iteration variables are significant only at the third
decimal place, the relative SEENEMOE are reduced (this will increase the value of the
iteration variable because TPROF P-y = Pz: + BLAS, + ITER.)

Second, the regression equations for land use types occupying less then 0.5% of the
area were deemed inaccurate and insignificant, despite an apparently high ROC
value. The regression coefficients were removed from the ALLoc1.REG files and
replaced by a constant (probability) of 0.5. Whilst not a proper parameter setting, this
change is insignificant because few grid cells are affected, and it allows the model to
find a solution at last.

Land use demands that change very little between the initial year and 2025 —for
example, the area of Mangroves in Mexico, which are always around 0.09 — 0.11% of
the total area (Table 4-6), were adjusted to remain constant (at the initial value),
similar to the Water and Other/Unknown categories. The difference in area was
allocated to the most similar land use type so that the total area remained the same.

' Temporary adjustments in model parameters that were made include: (1) use of much larger iteration variables,
in both absolute and relative mode, thus allowing the model to reach a solution quicker and reducing the number
of iterations needed; (2) use of a change matrix composed oniy of 1s, thus providing the greatest flexibility by

allowing any land use type; (3) replacement of the least accurate regression equations by a constant equation of

0.5.

? This is not documented anywhere based on own experience using the model.

58

e For all countries, the regression coefficients for Urban were considered inaccurate
and were replaced by a subjectively chosen function with a dependency on
populetion density, access to markets, and tourist hotspots (-1 + 0.0005*LF, -
aa +: 90). ®.). The parameters used in this function were slightly adjusted
after me pectod of the probability surfaces and test simulation runs.

e For Honduras, the following regression constants were changed:

o For Savanna, the regression constant was changed from —14.65 to —13.0.
This did not expand the area with a non-zero probability, but it did increase
the probability of areas that already had a non-zero probability. It was
expected that this would made a conversion to Savanna more likely as its
probability would now exceed that of (most importantly) Broadleaved Forest,
and other types.

o For Other (#0), Scrub (#4), Wetland/Swamp (#6), Mangroves (#7) and Water
(#9), the regression equation was replaced by a constant of 1 (the value is
unimportant). These land use types were kept constant ae the simulation.

o A new regression equation for Urban was selected: + 0.005*LU, =
).3*LU;. This produced a much-improved srofseteite surface that is
dependent on population density and accessibility to markets.

The neighbourhood function will ensure that no pixellated areas can exist.

4.2.5 Conversion elasticities

This parameter relates to the reversibility of the land use change. Land use types with a high
capital investment (e.g., urban; permanent crops such as banana plantations) will not easily
be converted in other uses as long as there is sufficient demand. Other land use types easily
shift location when the location becomes more suitable for different land use types. Arable
land often makes place for urban development, while expansion of agricultural land occurs at
the forest frontier.

For each land use type, a value needs to be specified that represents the relative elasticity to
change, ranging from 0 (easy conversion) to 1 (irreversible change).

0: Means that all changes for that land use type are allowed, independent from the
current land use of a location. This means that a certain land use type can be
removed at one place and allocated at another place at the same time: for example,
in shifting cultivation.

1: means that grid cells with one land use type cannot be added and removed at the
same time. This is relevant for land use types that are difficult to convert, i.e., urban
settlements and primary forests. This stabilizes the system, for example, preventing
simultaneous deforestation and reforestation in different areas.

>0...<1: Means that changes are allowed, however, the higher the value the higher
the preference that will be given to locations that are already under this land use type.

After initial trial and error runs, it was clear that the key was to use elasticities that are as
high as possible, but that do not stabilise the system too much and would prevent the model
from reaching a solution. Values of 1 stabilize the system and from the initial model runs it
appeared that these values are too stringent, i.e., they can prevent CLUE-S from reaching a
solution even after thousands of iterations. Changing a value of 1.0 to 0.95 makes the model
significantly more flexible. Likewise, with values of 0 for more flexible land use types such as
agriculture, the model changes land use too much throughout the area. Higher values such
as 0.2 for agriculture and 0.5 for scrub gave model results that appeared more plausible, i.e.,

59

a less complete overhaul of the land use pattern. The suggested settings (Table 4-8) are
based on expert knowledge of actual past land use patterns and observed model behaviour.

Table 4-8: Default conversion elasticities for the land use types.

Land use type elasticity |
0. Other/Unknown 0.9
1. Broad-leaved forest 0.95 |
2. Pine forest 0.95
3. Agriculture/pasture =] 0.0
4. Scrub 0.2
5. Savanna 0.5
6. Wetland/Swamp oe]
7. Mangroves 0.8
8. Urban 0.9
9. Water wi 1.0

4.2.6 Conversion matrix

Table 4-9 indicates the default conversion settings, and Tables 4-10 through 4-13 the actual
values used for each country. Note that the conversion to and from Other/Unknown (#0) and
Water (#9) are not allowed. The ‘demand’ for these land use types is unlikely to change, and
the CLUE-S model operates better if conversions are prohibited. Per-country adjustments
were made for those land use types that were artificially kept constant. For example, for
Mexico, the rows and columns associated with Pine Forest (non existent) Savanna and
Wetland/Swamp were constant Os as well.

Care had to be taken that there is always at least 1 “from” land use types for every “to” land
use type (besides the “to” land use type itself), and vice versa; otherwise the model may be
unable to reach a solution because no conversion can be carried out. Most importantly, in
ALLOW.TXT the values in all rows and columns of all land use types kept constant had to be
set to 0, except for the value on the same row and column. This adjustment in ALLow. Txt is
critical to prevent the model from starting calculations with these cells.’

Table 4-9: default conversion matrix. Note that some adjustments had to be made for all countries to
allow for sufficient change options, as indicated in blue in the next four tables.

| 1. Broad-leaved forest |__|
0

Vavscnib.
5. Savanna

6. Wetland/Swamp
7. Mangroves

I.

ololo|o|w

Hl
—/—|/ 32/

i
LL
TEE
AEE

fej}
]

lo)

o|—

‘This was found out by trial and error, and is not a documented model feature

60

Table 4-10: Modified conversion matrix for Belize conversion. The medium grey coloured rows and/or
columns are associated with land use types that were kept constant and did not change.

1. Broad-leaved forest
2. Pine forest
3. Agriculture/pasture

| 4. Scrub
| 5. Savanna
6. Wetland/Swamp

PS: Waters eee

Table 4-11: modified conversion matrix for Mexico. The medium grey coloured rows and/or columns
are associated with land use types that were kept constant and did not change.

| 1. Broad-leaved forest |/
3. Agriculture/pasture |

| 6. Wetland/Swamp _|

Table 4-12: modified conversion matrix for Honduras. The medium grey coloured rows and/or
columns are associated with land use types that were kept constant and did not change.

6. Wetland/Swamp :

' Conversion from agriculture to broad-leaved and pine forest had to be allowed so that the model could reach a
solution, whereas conversion from savanna to either forest type was not allowed to prevent large shifts in the
location of savanna.

61

Table 4-13: modified conversion matrix for Guatemala. The medium grey coloured rows and/or
columns are associated with land use types that were kept constant and did not change.

| 0. Other/Unknown |
| 1. Broad-leaved forest _|
|2.Pineforest
3. Agriculture/pasture |
4. Scrub

|5.Savanna
|7.Mangroves |
pasleban |
eEwater rs |

4.2.7 Dynamic location factor grids: Protected Areas

From all location factors listed in Table 1-6, only the last two - fully protected areas and
partially protected areas - were used as dynamic location factor grids that are different in
every simulated year. While population density is often a dynamic location factor in CLUE-S
applications, the lack of spatially-explicit population scenarios made this impossible.

Two shapefiles (EXISTING _PA.SHP and SCENARIO124.SHP) that were used for creating all
necessary dynamic location grids. The fields USE_CLASS, SC1-USE, SC2-USE, SC3-USE
and SC4-USE indicated whether the polygon was fully protected (“NO_USE”), partially
protected area (“SUST-USE”), or net designated under that scenario (‘EXCLUDED’). For the
third scenario about 20% of the areas were identified as “FAILED”. There is no difference
from the land use model’s point of view between an protected area labelled FAILED or
EXCLUDED - in both cases the area is not considered protected, because under that future
scenario its protection failed, or it was not protected in the first place.

An Avenue script was developed for creating all location factor grids in a fully automatic way
and saved in ASCII grid format (Appendix 2). This script uses a specially created shapefile
SCEN COMBINED 4RASTER.SHP that has 9 additional fields: YEARINCL, S1PAS1, S1PA2,
S2PAS1, S2PA2, S3PAS1, S3PA2, S4PAS1, and S4PA2.

4.2.8 Dealing with absence of pine forest in Mexico

There is no pine forest on the map for the MAR region of Mexico. This is the only total
absence of any land use type in the four countries. It required special attention and some
adjustments in model parameters to avoid a runtime error (overflow error).

One solution might be to adjust the numbering of the land use types to fill the gap, i.e.,
numbers 0-1,3-9 (2 is the missing pine forest) would have to be changed to 0-8. As this
process is cumbersome and error-prone, the following tweaks were made instead:

e Added dummy regression coefficients for land use type 2 in file ALLoc1.REG because

this file must contain regression coefficients for all land use types. A dummy equation
with a constant of 0 and a regression coefficient of 0 for the first factor grid was used.

' Conversion from pine forest and scrub to urban was prohibited to force change from broad-leaved forest to
urban and generate more plausible urban expansion

62

e In cov1.ALL, changed the value of four grid cells from 1 (broadleaved forest) to 2
(pine forest), thus introducing artificial pine forests. Four cells in the bottom-left corner
of the grid were chosen solely because these cells are easy to identify. These cells
were returned to broadleaved forest in the simulated land use grid cov22.ALL. Note
that the change was made for 4 cells instead of 1 cell, so that the corresponding area
has no significant decimal value, and CLUE-S will not make a rounding error when
the demand figures are read in (the demand values in Loc. FIL suggest that they are
rounded to 1 decimal place, although that may a formatting matter).

e Inthe demand file DEMAND1.FIL, replaced the 0 hectare value in the third column with
25. That is the area of the additional cell in ha. The values in the second column
(broadleaved forest) are reduced by 25 hectares.

e In ALLOW.TXT, changed all values in the third row and third column to O except the
value at position [3,3]. Thus, pine forest cannot change to anything else.

4.3 Simulation results

4.3.1 Simulated changes in land use and in forest cover

Figures 4-1 to 4-5 show the present and simulated land use patterns for 2025. These land
cover raster data were shared with WRI on 20" July 2006 for use in the N-SPECT
hydrological simulations. Figures 4-6 to 4-8 show the areas of change only, making the new
areas of each land use class easier to identify.

A minor anomaly in the simulated land use pattern near San Pedro Sula, Honduras can be
seen, with some areas of forest “sandwiched” in between new urban land. The cause of this
was identified (probability surfaces and regressions) but could not easily be resolved. It is
merely a reflection of the probabilistic nature of the model and that the exact allocation by the
model of land use at a local level cannot easily be influenced.

4.3.2 Average and maximum deviation of solution

The iterative allocation module never achieves an allocation that fully matches the demand.
This is controlled by the iteration variables on line 12 of the main parameter file. A relative
iteration mode with an average deviation of 0.5% and a maximum individual deviation (for
any individual land use type) of 3.0% was used. Table 4-15 below gives the actual deviations
that were achieved, which are always equal to or lower than the maximum values.

Table 4-15: mean and maximum deviation between demand and allocated land use, in percentage of
absolute area, for land use in the final simulated year, 2025 ). These statistics are calculated for every
simulated year but presented here only for the final year). The maximums (2 and 3” columns) are
specified in the main parameter file and are slight adjustments from the default settings in CLUE-S,
respectively, 0.35% and 3.0%. In almost all cases the highest deviation applies to land use that
occupies the least area and is not kept constant, which almost always is Urban

Maximums 1. Market First 2. Policy First 4. Sustain. First
r —tL_____—

Mean | Max Mean Max Mean Max Mean Max
(Belize 0.3% 1.5% 0.29% 1.00% 0.30% 1.08% 0.27% | 0.80%

Mexico 0.5%

Guatemala

0.34%

3.0% 0.31% | 2.96%

3.0% 0.48% | 1.85%

2.27%

Honduras

63

a Baseline 2000/2004 ; Market First 2025

8 _—

=

: en

Figure 4-1: Present land cover and simulated land cover for the three scenarios in 2025.

64

| ME CtherUnknown
BR Broad-teaved Fores!

Bw Pine forest

|) Agriculture/Pasture

Figure 4-2: Baseline (2000/2004) land use

65

HE Cthrer/Uniknown
ee] Broad-leaved Forest
Be] Pine forest

Ea Agniculture/Pasture
4 Scrub

Savanna

[] Wetland/Swamp
BB Mangroves

ea Urban

be] Water

Figure 4-3: Simulated land cover for scenario 1, Markets First, in 2025

66

Figure 4-4: Simulated land cover for scenario 2, Policy First, in 2025

67

WE Cine Unknown
GE Broad-eaved Forest

ee Pine forest

Ea Agriculture /Pasture

Savanna

7 Wetland/Swame
eo Mangroves
ee Urban

Po VWWister

BE Ctr) rUnknown
oe Broadseaved Forest
ea Pine forest

k | Agriculture/Pasture

Figure 4-5: Simulated land cover for scenario 4, Sustainability First, in 2025

68

BRE Other Unknown
| (2 Broaddeaved Forest

Wy Pine forest

|] Agriculture /Pasture

7 Serub
Savanna

iE

iE | Wetland'Swame

| 8 Mangroves
BD Uren

Be veer

Figure 4-6: Simulated areas of change with 2025 land cover for scenario 1, Markets First

69

Figure 4-7: Simulated area of change with 2025 land cover for scenario 2, Policy First

70

BE Other'Unknown
a Broad4oaved Forest
eS Pine forest

(] Agricuture/ Pasture
4 Savanna

BE Urvan

ee Veter

BE Othen'Urknown
| (29 Broad-leaved Forest

oa Pine forest

7] Agriculture Pasture

3 Mangroves
Be Urean
HE voter

Figure 4-8: Simulated areas of change with 2025 land cover for scenario 4, Sustainability First

71

5 Workshop, conclusions and recommendations

5.1 Technical Workshop

A Workshop on Watershed Management, Land Cover Change Analysis, and Modeling of
Land-based Sources of Pollution and Sediment Discharge to the MAR was held 15-18
August 2006 at Galen University, Belize. The workshop consisted of a policy session (1 %
day) and a technical session (2 % days). Two presentations about the scenario development
and the land use change modelling were given during the policy session, alongside
presentations on the background of the land-use-change threats to the Mesoamerican Reef,
and the policy implications of the MAR project.

The last day of the workshop was dedicated to training in land use change modelling using
the CLUE-S model. The training programme, exercises and further supporting information
--all bundled in a 30 page training package—can be found in Appendix 6.

Proceedings from the workshop have been compiled separately, and summarise feedback
received from workshop participants.

5.2 Conclusions and Recommendations

The following conclusions and recommendations are compiled based upon Joep Luijten’s
experience with the application of CLUE-S to the MAR region, and on feedback received
during the workshop.

5.2.1 Application of CLUE-S model to the MAR region

Q The CLUE-S methodology has been successfully used to simulated land use changes in
the MAR region over the next 25 years. A separate model was developed for each
country. This was the correct approach, as it allows more accurate models that better
capture the relevant (and different) explanatory factors in each of the countries.

a Simulated land use changes in different directions under the three scenarios, with
substantial conversion from forest to agricultural land under the Markets First and Policy
First scenarios. Under the Sustainability First scenario, changes towards other land use
types could be observed too; most notably changes towards scrub and new forest areas.

a Whilst CLUE-S is a relatively easy model to use, the overall land use change modelling
component of the project is quite complex. CLUE-S is not a model that can be quickly
applied and run; preparation and implementation requires substantial time (months) and
many data conversions. A significant portion of the hours required was spent on data
collection and/or creation and quality assurance, data preparation for use in SPSS
(regression analysis) and CLUE-S, and the regression analysis. Once the model was
properly calibrated, the final simulation runs were a relatively straightforward task.

a CLUE-S does not dictate any particular method for calculating the land use requirements.
The use of IMAGE and International Futures was possible only because we had access
to these models. Simpler methods are possible and recommended, especially where
specific regional policies are to be applied. For example, land use requirements could be
calculated using appropriate economic demand models or simply by setting hypothetical
land use requirements for the final scenario year and interpolating the land requirements
between the start year and end year using a linear or exponential growth model.

72

a The regression analysis was a somewhat weak part of the study in that many
relationships were not very significant and had to be manually tweaked or replaced by
either a more logical regression equation (e.g., Urban), as detailed in Section 4.2.4. This
is a direct consequence of the characteristics of the land use data that were used. The
Ecosystem Map data were relatively coarse and polygon-based. It is believed that the
regression analysis and the way CLUE-S allocates land based on the probability surfaces
would give better results (and need fewer adjustments in model parameters and/or
workarounds for the model not being able to reach a solution) if an original remote
sensed raster dataset is used. Workshop participants knew of several recently released
new land cover datasets, in particular for Guatemala.

a Another potential approach for improving the regression analysis is to use a “balanced
sample” dataset instead of the full dataset for the region. A balanced sample is a dataset
for a region that clearly exhibits the relevant relationships between a particular land use
type and one or more location factors. Only full datasets were used for the MAR study.

a CLUE-S can be used if land use data from only a single year are available because the
model is parameterized: in principle, based on the results of the regression analysis of
the present land use pattern and a set of potential explanatory factors. However, it is
always better if land use data for two or even three years are available, and these
additional data can be used to improve the models. Having data for two years, y; and yo,
allows one to parameterize the models based on an regression analysis of the data for y,,
then run the model from y, to y2, compare the simulated land use pattern with the actual
land use pattern at y2, and adjust (calibrate) the improve the model fit. If a third year ys is
available, then a simulation run from y2 to y3 can be undertaken for model verification.

5.2.2 Workshop and training

a This overall work was covered during a 1-hour presentation during the policy part of the
workshop and a 1-day training day in CLUE-S. The response to the questionnaires
indicated that, in general, the participants found working with CLUE-S very useful.

a The CLUE-S training was quite intense. Any future training should dedicate at least 2 or 3
days to CLUE-S, as that will allow participants to spend more time on three important
aspects of the study: (i) in-depth understanding and hands-on working with the actual
data for Belize, Mexico, Guatemala and Honduras; (ii) how to use their own datasets; and
(iii) the regression analysis of location factors and methods for incorporating additional or
different location factors into the model. Any follow-up training should include these
aspects, as some attendees requested this in the questionnaires.

73

6 References

Ankersen, T.T., Arriola, L. 2001. At the Crossroads of Conservation & Development: the
Challenge of the Maya Forest. Report to the Coalicion de la Selva Maya and the Tropical
Ecosystem Directorate of the United States Man & the Biosphere Program. 13 pp.
http://conservation.law.ufl.edu/pdf/mayaforest.pdf. Accessed 14 September 2006.

Balk, D., Pozzi, F., Yetman, G., Deichmann, U., Nelson, A. 2004. The Distribution of People
and the Dimension of Place: Methodologies to Improve the Global Estimation of Urban
Extents (draft). Center for International Earth Science Information Network (CIESIN),
Columbia University; International Food Policy Research Institute (IFPRI); The World
Bank; and Centro Internacional de Agricultura Tropical (CIAT).
http://sedac.ciesin.columbia.edu/gpw/docs/UR paper webdraft1.pdf. Accessed 4
October 2006.

Batjes, N.H., 2005. /SRIC-WISE global data set of desired soil properties on a 0.5 by 0.5
degree grid. Report 2005-08. ISRIC — World Soil Information. December 2005.

http://www. isric.org/isric/webdocs/Docs/ISRIC_Report_2005 08 pdf. Accessed 14
September 2006.

Batjes, N.H., 2005. SOTER-based soil parameter estimates for Latin America and the
Caribbean (version 1.0). Report 2005/02. ISRIC — World Soil Information. December

2005. http://www. isric.org/isric/iwebDocs/Docs/ISRIC_ Report 2005 02.pdf. Accessed 14
September 2006.

Belize Tourism Board. No date. National Tourism Policy: Strategic Vision for Belize Tourism
in the New Millennium. Ministry of Tourism, Government of Belize. Belize City, Belize.

http://www.belizetourism.org/policy.html

Belize Tourism Board No date. Belize Cruise Tourism Policy. Ministry of Tourism,
Government of Belize. Belize City, Belize.
http://www.destinationsbelize.com/cruise_s.htm. Accessed 14 September 2006.

Boles, E. 2005. United Nations Convention on Biological Diversity: Thematic Assessment
Report of Belize. National Capacity for Self-Assessment project, Ministry of Natural
Resources. Belmopan, Belize. 58 pp.

Buck, M. 2002. Project Summary: Central American Project for Geographic Information:
Honduras. Report to PROCIG, the Central American Geographic Information Project.

http://www. procig.org/ing/procig-honduras.htm. Accessed 14 September 2006.

Central Statistical Office. 2001. Environmental Statistics for Belize: 2001. Environmental
Statistics Unit, Central Statistics Office, Ministry of Budget Management, Investment and
Public Utilities. Belmopan, Belize. 115 pp.

Central Statistical Office. 2005. Environmental Statistics for Belize: 2004. Environmental
Statistics Unit, Central Statistical Office, Ministry of National Development. Belmopan,
Belize. 219 pp.

Cherrington, E., 2005. Drivers of land use change adjacent to the Meso-american reef: a
preliminary review. Unpublished report to UNEP-WCMC. August 2005.

Chomitz, K.M., Gray, D.A. 1996. Roads, Land Use, and Deforestation: A Spatial Model
Applied to Belize. Poverty, Environment, and Growth Working Paper No. 5. International
Food Policy Research Institute, World Bank. Washington, DC. 26 pp.
http://www.worldbank.org/research/peq/wps03/roads.pdf. Accessed 14 September 2006.

CIAT, UNEP, CIESIN, Columbia University and the World Bank (2005) Latin American and
Caribbean Population Database. Version 3. Centro Internacional de Agricultura Tropical
(CIAT), United Nations Environment Program (UNEP), Center for International Earth

74

Science Information Network (CIESIN), Columbia University, and the World Bank.
http://gisweb.ciat.cgiar.org/population/report.htm. Accessed 4 October 2006.

Deininger, K.W., Minten, B. 1996a. Poverty, Policies, and Deforestation: the Case of Mexico.
Poverty, Environment, and Growth Working Paper No. 5. International Food Policy
Research Institute, World Bank. Washington, DC. 54 pp.

http://www.worldbank.org/research/peg/wps05/mexico.pdf. Accessed 14 September
2006.

Deininger, K.W., Minten, B. 1996b. Determinants of Forest Cover and the Economics of
Protection: An Application to Mexico. Poverty, Environment, and Growth Working Paper
No. 10. International Food Policy Research Institute, World Bank. Washington, DC. 22

pp. http://www.worldbank.org/research/peg/wps10/mxforest.pdf. Accessed 14
September 2006.

DiFiore, S. 2002. Remote Sensing and Exploratory Data Analysis as Tools to Rapidly
Evaluate Forest Cover Change and Set Conservation Priorities along the Belize River,
Belize. MS thesis, Department of Ecology, Evolution and Environmental Biology,
Columbia University. New York, NY. 68 pp.

Fairweather, P.N., Gray, D.A. 1994. The Land Use of Belize: 1989/92. Land Information
Centre, Ministry of Natural Resources. Belmopan, Belize. 15 pp.

Fairweather, P.N. 2002. Sistemas de Informacion Geografica en Belice [Geographic
Information Systems in Belize]. Report to PROCIG, the Central American Geographic

Information Project. http://www. procig.org/reunion-ctg/documentos/belice/belice-
resumen.pdf. Accessed 14 September 2006.

FAO, 1999. FRA 2000 Annotated Bibliography. Forest Cover Change: Guatemala. Working
Paper 13. Rome, Italy. 65 pp. ftp://ftp.fao.org/docrep/fao/005/ac631e/ac631e.pdf

FAO, 2000a. “FRA 2000 Annotated Bibliography. Forest Cover Change: Belize.” Working

Paper 40. Rome, Italy. 38 pp. ftp://ftp.fao.org/docrep/fao/007/ad677e/ad677e00.pdf.
Accessed 14 September 2006.

FAO, 2000b. FRA 2000 Bibliografia comentada. Cambios en la cobertura forestal: Honduras.
Working Paper 44. Rome, Italy. 40 pp.
ftp://ftp.fao.org/docrep/fao/007/ac768s/ac768s.pdf. Accessed 14 September 2006.

FAO, 2000c. FRA 2000 Bibliografia comentada. Cambios en la cobertura forestal: México.
Working Paper 35. Rome, Italy. 35 pp.
ftp://ftp.fao.org/docrep/fao/006/ad667s/ad667s00.pdf. Accessed 14 September 2006.

Farrow, A., Winograd, M. 2001. Land use modelling at the regional scale: an input to rural
sustainability indicators for Central America. Agriculture, Ecosystems. & Environment 85:
249-268.

Farrow, A., Nelson, A. 2001. Accessibility modelling in ArcView 3. An extension for
computing travel time and market catchment information. (User's manual). CIAT.
http://www.ciat.cgiar.org/access/index.htm. Accessed 14 September 2006.

Hughes, B.B. 2004. Introduction to International Futures (IFs). Graduate School of
International Studies, University of Denver.
http://Awww.du.edu/~bhughes/reports/Introduction%20to%20IFs%20V2 O.pdf. Accessed
1 June 2006.

Instituto Geografico Nacional, Guatemala. 2002. Guatemala Summary. Report to PROCIG,
the Central American Geographic Information Project. http://www.procig.org/reunion-

ctg/documentos/quatemala/resumen-actividades-ign-quatemala.pdf. Accessed 14
September 2006.

75

IUCN 1994. Guidelines for Protected Area Management Categories. CNPPA with the
assistance of VWCMC. IUCN, Gland, Switzerland and Cambridge, UK. 261pp.

http://www.unep-weme.org/protected_areas/categories/index.html. Accessed 14
December 2005.

lyo, J., Mendoza, P., Cardona, J., Cansino, A., Davis, R. 2003. An Overview of Land
Administration & Management in Belize. Workshop on Land Policy, Administration and
Management in the English — Speaking Caribbean, March 20003, Port of Spain,
Trinidad. 50 p

p.
http://Aweb.archive.org/web/20030520070825/http://www.mhtc.net/~terra/carib_workshop
/pdf/belizecfs.pdf. Accessed 14 September 2006.

Jansen, H.G.P., Siegel, P.B., Pichon, F. 2005. /dentifying the Drivers of Sustainable Rural
Growth & Poverty Reduction in Honduras. DSGD Discussion Paper No. 19.
Development Strategy & Governance Division, International Food Policy Research

Institute. Washington, DC. 136 pp. http://www. ifpri.org/divs/dsgd/dp/papers/dsqdp19.pdf.
Accessed 14 September 2006.

King, R.B., Baillie, |.C., Bissett, P.G., Grimble, R.J., Johnson, M.S., Silva, G.L. 1986. Land

Resource Survey of Toledo District, Belize. Land Resource Development Centre.
Tolworth, United Kingdom.

King, R.B., Baillie, |.C., Dunsmore, J.R., Grimble, R.J., Johnson, M.S., Wright, A.C.S. 1989.
Land Resource Survey of Stann Creek District, Belize. Bulletin 19. Natural Resources
Institute, Overseas Development Administration, Chatham, UK. 262 pp.

King, R.B., Baillie, 1.C., Abell, T.M.B., Dunsmore, Gray, D.A., Pratt, J.H., Versey, H.R.,
Wright, A.C.S., Zisman, S.A. 1992. Land Resource Assessment of Northern Belize.
Bulletin 43. 2 volumes. Natural Resources Institute, Overseas Development
Administration, Chatham, United Kingdom. 513 pp.

King, R.B., Pratt, J.H., Warner, M.P., Zisman, S.A. 1993. Agricultural Development Prospects
in Belize. Bulletin 48. Natural Resources Institute, Overseas Development
Administration, Chatham, United Kingdom. 159 pp.

Kok, K., Winograd, M. 2002. Modelling land-use change for Central America, with special
reference to the impact of Hurricane Mitch. Ecological Modelling 149: 53-69.

Kok, K., 2004. The role of population in understanding Honduran land use patterns. Journal
of Environmental Management 72: 73-89.

Kok, K., Veldkamp, A. 2001. Evaluating impact of spatial scales on land use pattern analysis
in Central America. Agriculture, Ecosystems & Environment 85: 205-221.

Loening, L.J., Markussen, M. 2003. Pobreza, Deforestacion y Pérdida de la Biodiversidad en
Guatemala. Discussion Paper No. 91. Ibero-America Institute for Economic Research.

44 pp. http://www.iai.wiwi.uni-goettingen.de/content/pdfliai-db91.pdf. Accessed 14
September 2006.

Meerman, J., Sabido, W. 2001. Central American Ecosystems: Belize. Programme for
Belize, Belize City. 2 volumes 50 + 88 pp.

Merrill, T (ed.). 1992. Belize: A Country Study. Call No. F2368.G893 1993. Federal Research
Division, U.S. Library of Congress. http://Icweb2.loc.gov/frd/cs/bztoc.html. Accessed 14
September 2006.

Merrill, T. (ed.). 1993. Honduras: A Country Study. Call No. F1503.H75 1995. Federal
Research Division, U.S. Library of Congress. http://Icweb2.loc.gov/frd/cs/hntoc.html.
Accessed 14 September 2006.

76

Merrill, T.L., Miré, R. (eds.) 1996. Mexico: A Country Study. F1208.M5828 1997. Federal
Research Division, U.S. Library of Congress. http://Icweb2.loc.gov/frd/cs/mxtoc. html.
Accessed 14 September 2006.

Ministry of Agriculture & Fisheries. 2003. The National Food & Agriculture Policy (2002-
2020). Government of Belize, Belmopan, Belize. 63 pp.

Myers, N., Mittermeier, R.A., Mittermeier, C.G., da Fonseca, G.A., Kent, J. 2000. Biodiversity
hotspots for conservation priorities. Nature 403(6772):853-858.

Nelson, A., 2000. Accessibility, transport and travel time information. \Internal document.
CIAT, Cali, Colombia.

Sader, S.A., Hayes, D.J., Irwin, D.E., Saatchi, S.S. 2001. Preliminary Forest Cover Change
Estimates for Central America (1990s), with Reference to the Proposed Mesoamerican
Biological Corridor. Proceedings of the Year 2001 Annual Conference of the American
Society of Photogrammetric Engineering and Remote Sensing. St. Louis, MO. 11 pp.

http:/wwwahcc.msfc.nasa.gov/corredor/mbr_asprs_paper.pdf. Accessed 14 September
2006.

Schuttenberg, H.Z. (ed). 2001. Coral Bleaching: Causes, Consequences and Response.
Selected Papers presented at the 9th International Coral Reef Symposium on Coral

Bleaching: Assessing and Linking Ecological and Socioeconomic Impacts, Future
Trends and Mitigation Planning. Coastal Resources Center, Univ. of Rhode Island.

Turner II, B. L., Cortina Villar, S., Foster, D., Geoghegan, J., Keys, E., Klepeis, P., Lawrence,
D., Macario Mendoza, P., Manson, S.M., Ogneva-Himmelberger, Y. 2001. Deforestation
in the southern Yucatan peninsular region: an integrative approach. Forest Ecology and
Management 154(3): 353-370.

UNEP 2002. Global Environment Outlook 3. United Nations Environment Programme,
Nairobi, Kenya. http://www.unep.org/qeo/geo3/. Accessed 14 September 2006.

Verburg, P.H., 2004. Manual for the CLUE-S model. September 2004.
http://www.cluemodelnl/. Accessed 14 September 2006.

Verburg, P.H., Veldkamp, T, Lessschen, J.P. 2004. Exercises for the CLUE-S model.
http://www.cluemodel.ni/Ex4 pdf. Accessed 14 September 2006.

Vreugdenhill, D., Meerman, J., Meyrat, A., Gomez, L.D., Graham, D.J., 2002. Map of the
ecosystems of Central America. Final report. World Bank, Washington, D.C.

Wassenaar, T., Gerber, P., Rosales, M., Ibrahim, M., Verburg, P.H., Steinfeld, H. 2005.
Projecting land use changes in the Neotropics: the geography of pasture expansion into
forest. Submitted.

WCSD, 1987. Our Common Future. World Commission for Sustainable Development /
Oxford Paperbacks. 398pp.

White, W.A., Raney, J., Tremblay, T.A., Crawford, M.M., Smith, S.S. 1996. Deforestation in
Belize 1989/92-1994/96. Bureau of Economic Geology, University of Texas. Austin, TX.
49 pp.

White, W.A., Raney, J., Tremblay, T.A., Crawford, M.M., Smith, S.S.. 1998. Remote Sensing
Analysis of Land Cover & Land Use, Central Belize. Bureau of Economic Geology,
University of Texas. Austin, TX. 39 pp.

77

7 Appendices

Appendix 1. Avenue script for NoData filling and filtering

The following Avenue script was created to fill NoData holes and to apply a majority filter to
rasters, as mentioned in Section 1.1.1. This script removes any minor imperfections in the
raster grids and adds a small buffer of value cells at the edge so that every raster being used
for the statistical analysis has exactly the same number of value cells. The script can handle
both mean filter and majority filters, and up to 10 iterations at once.

' SWBM.Grid.FillNodataGaps

July 14, 2004. Joep Luijten

This script was written fill the common Nodata cells in SRTM elevation

data. The NoData cells are typically areas with steep gradients, river

valleys, etc. The fill is done iteratively. See ESRI article 22853.

http: //support.esri.com/index.cfm?fa=knowledgebase.techarticles.articleShow&d=22853
The number of necessary iterations depends on how large the data gaps are.

4/1/06. Provide selection menu to choose focalstats type (MEAN ort MAJORITY) and
output type (floating or integer). This enables to use this script also to fill
gaps in classified data. The majority filter wasn't as straightforward, though.

The use of the MajorityFilter() does correctly fill any NoData holes inside a grid
providing that the second argument is set TRUE, however, it does not convert NoData
cells to value cells at the edge. On the other hand, a FocalStats() of type

GRID _STATYPE MAJORITY does create value cells at the edge (1 cell wide per
iteration). Hence, it was deemed necessary to implement a succession of both
methods at each iteration step to achieve the desired result.

theView = av.GetActiveDoc

theTheme = theView.GetActiveThemes.Get (0)
rawDem = theTheme.GetGrid

titmsg = "Fill NoData gaps in GRID"

' Check to proceed

if (MsgBox.YesNo("Fill NoData gaps in GRID theme" ++ theTheme.GetName + "?",titmsg, FALSE) =
FALSE) then

return nil

end

' Method
iMethodList = {"MEAN 3x3 filter, return floating grid",
"MEAN 3x3 filter, return integer grid",
"MAJORITY 3x3 filter, return integer grid"}
iMethod = Msgbox.ListAsString (iMethodList, "Select FocalStats method", titmsg)

if (iMethod = nil) then

return nil
else

iMethodIndex = iMethodList.FindByValue (iMethod)

if (iMethodIndex = 0) then
bMean = true
bFloat = true

elseif (iMethodIndex = 1) then
bMean = true

bFloat = false

elseif (iMethodIndex = 2) then

bMean = false
bFloat = false
end
end

' Prompt for number of iterations

errorMsg = "You must enter a number between 1 and 10"
while (true) é ae
Niter = MsgBox.Input ("Number of iterations (1-10]:",titmsg, "3

if (nIter = NIL) then return nil end

if (nIter.IsNumber.Not) then

MsgBox .Warning(errorMsg, titmsg)

else

nIter = nIter.AsNumber

if ((nIter < 1) or (NIter > 10)) then
MsgBox.Warning(errorMsg, titmsg)

else
break

Appendix 1 78

end
end
end

' Check for any values < 0 or >= 32768 (in SRTM data, 32768 is used for NoData).
gStats = rawDem.GetStatistics

setMin = false

if (gStats.Get(0) < 0) then

setMin = Msgbox.YesNo("The grid contains negative values (as low as"++
gStats.Get(0).AsString ++ "). These are unusual --though not impossible-- elevation"++
"values that you may want to set to zero. Do you want to do this?",titmsg, TRUE)

end

setMax = false

if (gStats.Get(1) >= 32768) then

setMax = Msgbox.YesNo("The grid contains very high values that are unlikely elevations."++
"Note that SRTM data often contain values 32768 (NoData) and 95xxx (incorrect) ."++
"You are strongly recommended to set these values to NoData. Okay?",titmsg, TRUE)

if (setMax) then

mxMsg = "You must enter a number between 0 and 100000"
while (true)
mxCut = MsgBox.Input ("Maximum cutoff value (excluded) :",titmsg,"32768")

if (mxCut = NIL) then return nil end
if (mxCut.IsNumber.Not) then
MsgBox. Warning (mxMsg, titmsg)
else
mxCut = mxCut.AsNumber
if ((mxCut < 0) or (mxCut > 100000)) then
MsgBox.Warning(mxMsg, titmsg)
else
break 'Value is OK
end
end
end
end
end

' Apply min and max values
if (setMin and setMax) then

gO = ((rawDem < 0.asgrid).Con(0.asgrid, (rawDem >= mxCut.asgrid) .SetNull (rawDem) ) )
elseif (setMin and setMax.Not) then

gO = ((rawDem < 0.asgrid) .Con(0.asgrid, rawDem) )

elseif (setMin.Not and setMax) then

gO = ((rawDem >= mxCut.asgrid) .SetNull(rawDem) )

else

gO = rawDem

end

' Perform iterative fill. Two methods in succession for the majority filter.

theNbrHood = NbrHood.Make ' Default 3x3 rectangular neighborhood

if (nIter >= 1) then
if (bmean = true) then
gl = (g0.IsNull) .Con((g0.FocalStats (#GRID_STATYPE MEAN, theNbrHood, FALSE) ),g0)
else
gltmp = ((g0.IsNull) .Con(g0.MajorityFilter(TRUE, TRUE) ,g0))
gl = (gltmp.IsNull) .Con((gltmp.FocalStats (#GRID_STATYPE MAJORITY, theNbrHood, FALSE) ),gltmp)
end
if (nIter = 1) then
gFinal = gl
end
end

if (nIter >= 2) then
if (bmean = true) then
g2 = (gl.IsNull) .Con((gl.FocalStats (#GRID STATYPE MEAN, theNbrHood, FALSE) ),gl)
else
g2tmp = ((gl.IsNull) .Con(gl.MajorityFilter(TRUE, TRUE),gl))
g2 = (g2tmp.IsNull) .Con( (g2tmp.FocalStats (#GRID_STATYPE MAJORITY, theNbrHood, FALSE) ),g2tmp)
end
if (nIter = 2) then
gFinal = g2
end
end

if (nIter >= 3) then
if (bmean = true) then
g3 = (g2.IsNull) .Con((g2.FocalStats (#GRID_STATYPE_MEAN, theNbrHood, FALSE) ),g2)
else
g3tmp = ((g2.IsNull) .Con(g2.MajorityFilter(TRUE, TRUE),g2) )
g3 = (g3tmp.IsNull) .Con((g3tmp.FocalStats (#GRID_STATYPE MAJORITY, theNbrHood, FALSE) ),g3tmp)

Appendix 1 79

end

if (nIter = 3) then
gFinal = g3

end

end

if (nIter >= 4) then
if (bmean = true) then

g4 = (g3.IsNull) .Con((g3.FocalStats (#GRID_STATYPE_MEAN, theNbrHood, FALSE) ),g3)
else

g4tmp = ((g3.IsNull) .Con(g3.MajorityFilter(TRUE, TRUE) ,g3)

g4 = (g4tmp.IsNull) .Con((g4tmp.FocalStats (#GRID_STATYPE_MAJORITY, theNbrHood, FALSE) ), g4tmp)
end
if (nIter = 4) then

gFinal = 94

end
end

if (nIter >= 5) then
if (bmean = true) then

g5 = (g4.IsNull) .Con((g4.FocalStats (#GRID_STATYPE_MEAN, theNbrHood, FALSE) ),g4)
else

gStmp = ((g4.IsNull) .Con(g4.MajorityFilter (TRUE, TRUE) ,g4) )
g5 = (gStmp.IsNull) .Con((gStmp.FocalStats (#GRID_STATYPE MAJORITY, theNbrHood, FALSE) ),g5tmp)
end

if (nIter = 5) then
gFinal = g5

end

end

if (nIter >= 6) then
if (bmean = true) then
g6 = (g5.IsNull) .Con((g5.FocalStats (#GRID_STATYPE_ MEAN, theNbrHood, FALSE) ),g5)
else
gé6tmp = ((g5.IsNull) .Con(g5.MajorityFilter(TRUE, TRUE) ,g5))
g6 = (gétmp.IsNull) .Con( (g6tmp.FocalStats (#GRID_STATYPE_MAJORITY, theNbrHood, FALSE) ),g6tmp)
end
if (nIter = 6) then
gFinal = g6é
end
end

if (nIter >= 7) then
if (bmean = true) then

g7 = (g6.IsNull) .Con( (g6.FocalStats (#GRID_STATYPE_MEAN, theNbrHood, FALSE) ) , g6)
else

g7tmp = ((g6.IsNull) .Con(g6.MajorityFilter(TRUE, TRUE) ,g6)

g7 = (g7tmp.IsNull) .Con((g7tmp.FocalStats (#GRID_STATYPE_MAJORITY, theNbrHood, FALSE) ),g7tmp)
end

é (niter ="7))) ‘then
gFinal = g7

end

end

if (nIter >= 8) then
if (bmean = true) then

g8 = (g7.IsNull) .Con((g7.FocalStats (#GRID_STATYPE_MEAN, theNbrHood, FALSE) ),g7)
else

g8tmp = ((g7.IsNull) .Con(g7.MajorityFilter (TRUE, TRUE) ,g7) )

g8 = (g8tmp.IsNull) .Con((g8tmp.FocalStats (#GRID_STATYPE MAJORITY, theNbrHood, FALSE) ) , g8tmp)
end
if (nIter = 8) then

gFinal = g8

end
end

if (nIter >= 9) then
if (bmean = true) then
g9 = (g8.IsNull) .Con((g8.FocalStats (#GRID_STATYPE_MEAN, theNbrHood, FALSE) ) , g8)
else
g9tmp = ((g8.IsNull) .Con(g8.MajorityFilter (TRUE, TRUE) ,g8) )
g9 = (g9tmp.IsNull) .Con((g9tmp.FocalStats (#GRID_STATYPE_MAJORITY, theNbrHood, FALSE) ), g9tmp)
end
if (nIter = 9) then
gFinal = g9
end
end

if (nIter >= 10) then
if (bmean = true) then
gl10 = (g9.IsNull) .Con((g9.FocalStats(#GRID_STATYPE MEAN, theNbrHood, FALSE) ), g9)

else
gl0tmp = ((g9.IsNull) .Con(g9.MajorityFilter (TRUE, TRUE) ,g9)

(gl0tmp. IsNull) .Con((gl0tmp.FocalStats (#GRID_STATYPE MAJORITY, theNbrHood, FALSE) ) , g10tmp)

Appendix 1 80

end

if (nIter = 10) then
gFinal = gl0

end

end

' Make final grid
if (bFloat = true) then
gFinal2 = gFinal.Float

sOper = "("+nIter.asstringt+"pass)"
else

gFinal2 = gFinal.Int

end

' Construct new title name

if (bMean = true) then

sname = theTheme.GetNamet++"("+nIter.asstring++"pass MEAN filter)"
else

sname = theTheme.GetNamet++"("+nIter.asstring++"pass MAJORITY filter)"
end

' Add filled grid theme to view
newGTheme = GTheme.Make (gFinal2)
newGTheme. SetName (sname)
theView. AddTheme (newGTheme)
theView. Invalidate

Appendix 1 81

Appendix 2. Avenue script for creating dynamic protected areas grids

The script below was used to create dynamic location grids for the protected area scenarios.
The script requires a view that contains five themes: the protected areas shapefile
(SCEN_COMBINED_4RASTER.SHP) and the four mask grids for the countries (MASK _Bz_ 250,
MASK_GT_250, MASK_MX_250, MASK_HN 250). The protected areas shapefile must have
eight additional fields S1PA7, S2PA2, A1PA1, .., S4PA2, each with 0 and 1 values, indicating
whether the polygon is a full or partially protected area. The scripts generates the grid
SRC1G11.FIL (static fully protected areas), REGION NO USE _S1MKT.FIL (also static fully
protected areas a value of —9998 for those areas) and SRC1G12.0, SRC1G12.1, ..., through
to SRC1G12.25 (dynamic partially protected areas).

' Create.Dynamic.ProtectedAreas.Grids

' Location factor numbers (as in CLUE-S regression files)
locFacNum_fullProt = 11

locFacNum partProt = 12

baseOutFolder = "D:\Work WCMC\CLUES\dyndata\"

' Get active view
theView = av.FindDoc("protected areas")
thePrj = TheView.Getprojection
if (theView.Is(View) .Not) then
msgbox.Info ("Active document must be a view","")
return nil
end

' Select country to process
country = MsgBox.ListAsString({"BZ","MX","GT","HN"}, "Select country","")
if (country = NIL) then
return nil
end
' select scenario to process
scenario = MsgBox.ListAsString (
{"1 Market First","2 Policy First","3 Security First","4 Sustainability First"),
"Select scenario","")
if (scenario = NIL) then
return nil
else
scenNo = scenario.Left(1).AsNumber
end

' Select protected areas shapefile (modified to with special fields added)
thmList = theView.GetThemes
if (thmList.Count > 0) then
wdpaThm = MsgBox.ListAsString(thmList, "Select the Protected Areas shapefile." +
"The attribute table must include fields S"+tscenNo.asstring+
"PAl and S"+scenNo.AsString+"PA2.","")
if (wdpaThm = NIL) then
return nil
else
theFTab = wdpaThm.GetFTab
fldList = theFtab.GetFields
end
else
return nil
end

' Output directory

outDir = Msgbox.Input ("Output folder","",baseOutFolder + country)
if (outDir = NIL) then return NIL end

if (scenNo = 1) then

subdir = "slmkt"
elseif (scenNo = 2) then
subdir = "s2pol"
elseif (scenNo = 3) then
subdir = "s3sec"
elseif (scenNo = 4) then
subdir = "s4sus"
end

outDir = outDir + "\" + subdir

' Get mask grid. If the hardcoded name not found the selection menu will be shown.
maskName = "mask_" + country + "_250"
maskThm= theView.FindTheme (maskName)

Appendix 2 82

if (maskThm <> NIL) then
maskGrid = maskThm.Getgrid
else
if (thmList.Count > 0) then
maskThm = MsgBox.ListAsString(thmList,
"Select the mask grid for " + country,"")
if (maskThm = NIL) then
return nil
else
maskGrid = maskThm.Getgrid
end
else
return nil
end
end

" Set analysis extent same to mask grid

aRect = maskgrid.getExtent

aCell = maskgrid.getCellSize

Grid.SetAnalysisExtent (#GRID_ENVTYPE_VALUE, aRect)
Grid.SetAnalysisCellsize(#GRID_ENVTYPE VALUE, aCell)

' Create grid with only zeros for country extent
zeroGrid = ((maskGrid.IsNull) .setnull(0.asgrid) )
' **** STATIC LOCATION FACTOR FOR FULLY PROTECTED AREAS ****

' Filename for static location factor grid
cluename = "Srclgr" + locFacNum_fullProt.asstring + ".fil"

' Query to select all WDPA to include
fldl = "S" +scenNo.asString + "pal"

expr = "(("+£ld1+"] = 1)"

theBitmap = theFTab.GetSelection
theFtab.Query(expr, theBitmap, #VTAB_SELTYPE_ NEW)
theFTab.UpdateSelection

' Convert shape to grid.

tmp1Grid = Grid.MakeFromFTab (theFTab, thePrj,nil,nil)
tmp2Grid = (tmplGrid.IsNull) .Con(0.AsGrid,tmplGrid) 'Grid with 0s and ls
finGrid = ((maskGrid.IsNull).setnull(tmp2grid)) 'Clip grid

' Save grid in ascii format

theFn = (outdir + "\" + cluename) .AsFileName

if (File.Exists(theFn)) then File.Delete(theFn) end
fingrid.SaveAsAscii (theFn)

' Also make a correspnding area restriction file. Active cells must have value
' of 0, restricted cells -9998, all other cells (NoData) -9999.

resGrid = (finGrid = 1.asgrid) .Con(-9998.asgrid, finGrid)

theFn = (outdir + "\region_no_use_" + subdir + ".fil") .AsFileName

if (File.Exists(theFn)) then File.Delete(theFn) end

resGrid.SaveAsAscii (theFn)

resGrid = nil

fingrid nil
tmp2grid = nil
tmplgrid = nil

' **** DYNAMIC LOCATION GRIDS FOR PARTIALLY PROTECTED AREAS ****

' Number of years for which to save dynamic grids

if (country = "BZ") then
nYears = 21 '2004 to 2025
else

nYears = 25 '2000 to 2025
end

' Save 2 WDPA location factor grid for each year
for each i in 0..nYears

' New grid name
cluename = "Srclgr" + locFacNum_partProt.asstring + "." + i.asstring

' Actual simulated year. First year will be 0 for grid naming convention.

if (country = "bz") then
year = 2004 +i

else

year = 2000 +i

end

' Select field name

fldl = "S" +scenNo.asString + "pa2"

Appendix 2 83

fld2 = "Yearincl"

" Query to select all WDPA to include

expr = "(["+fldl+"] = 1) and (["+f£1d2+"] <= "+year.asstring + ")"
theBitmap = theFTab.GetSelection

theFtab.Query(expr, theBitmap, #VTAB SELTYPE NEW)
theFTab.UpdateSelection Y =z

' Convert shape to grid.

tmplGrid = Grid.MakeFromFTab(theFTab, thePrj,nil,nil)

tmp2Grid = (tmplGrid.IsNull) .Con(0.AsGrid,tmplGrid) 'Grid with Os and 1s
finGrid = ((maskGrid.IsNull) .setnull (tmp2grid) ) "Clip grid

" Save grid in ascii format

theFn = (outdir + "\" + cluename) .AsFileName

if (File.Exists(theFn)) then File.Delete(theFn) end
fingrid.SaveAsAscii (theFn)

fingrid = nil
tmp2grid nil
tmplgrid nil

' Add theme to view
‘grdThm = GTheme.Make (finGrid)
'grdThm. SetName (cluename)
'theView. addTheme (grdThm)

end

zerogrid = nil

Appendix 2 84

"BJEP 2SU} SN 0} JOU JSabDNS ‘[Njasn Ajan aq AjayuN
SI BJep Siu} ‘aiojasayy “ejewayend pue ezijag wWejsee ul Avejnoped
‘uoiBad ay} jnouBnosy} (Smopeus pnojo pue) aBesanoo pnojo Ayoyed
jeuejsqns mous dsl Xepul/ipse/0808-Npe pun soeiuun ddejojby/-dyy
Wwod4 (ZZ pueg) sayisodwog

7Syonpold eyep/o| ob/WIOd JeAdd0eb MWMN//.d}Yy
‘aiqeyiene Mou ae eovaWy jesjUaD
JO} EYEP NOT JEU} GO/G/Z UO IID pajiewa (oo Tesypes@epsogi)
eplog pseyoly ‘payissejd sapun puey jesnjyjnowbe

s}sab6Bns sousuadxa YA, Jae] a}ep 19d QGOF aUe S}sOo pajyewsy

‘pajyeBysanul Bulag si aseajas jasejeg "sasayul yo Ajenred
due BjOW Basu} ‘edie yOaloid ay} UIYYIM AjJe}O} ase Sa}Is [O1}UOD
BaJYUL “BAOge ejep uOlNjOSa, aMO| BU} JO UOI}EIqIed JO} asf)

‘puaba] ul }SaJoj-uou / }Sas0} uado / sa10} AUD

‘ainpayos aseejau /ayepdn jxau aJ JajuaD e}eq SONA

pue AjIsiaAluA UO}SOg pa}oe}UON IID ‘papiAoid yonposd aBueyo
J@AOO Pue] e JO pue SyJUOW xis AuaAa pajepdn aq jij ‘UOIBad JOaloid
ul Says Bulules} ON “UOHEOYISSe|O pa}ewojne jua}sisuod Ajjeqo|g

‘yoalosd siy} Joy JUeHOdWI
SS9Q| ‘SIPOW] Jo Aouajsisuod jeqo|6 your; Aeyy ‘panjonul syadxe |e90|
se Aoeinooe jeuoiBai poo “e}ep NOILVLIOAA LOdsS uo paseg

s}ules}SUOD / S}UBWIWOD / SN}e}S

ée}ep 30H

€ xipuaddy

(Z00z
8 0002) 000z
pue (p69 | (SuolBas oj!
WOE | 06 ‘686L) O66L | IIs) jeqoiID

4809085
sous0s
000g | }espuey ALL
(yespuey) punose pue JO 4JOMJeU eyep
WOE SO6G6I ‘SO861 TEqo| | ays josjuod Was
sebeul
9661 -2661 UO
Wy | peseq — 0002 leqoip 0002 Vu
L00¢ ®O JOAOD
wy | 0} 0002 0 lego|p pue] SIGOW
(0002
99d 9} 6661
Wy | AON) 0002 leqo| 0002919
Ayjuno5 awiauL
uonnjosay siea, /yuayxg | / aweTN JasejeQg

SLASVLVG YSAO9 GNV1/ 3SN GNV1T

“sJapeal 0} BOUADIAI JO ASed U! B]qGe} ay} UI Ya] UGeq aAeY Ssa90e

pue Aqiqeyiene eyep uo sajou jewuoju| “jOaloid siy} JO} pajoa9jas sjeseyep ay} a}edIPul SMOJ papeys MOjjaA ay] ‘papeojumop UsEeq OS|e sAeY e}ep
asau} jo Aue ‘aseyd uolda||09 eyep ay) BuuNp payuap! osam yeu} e}ep feeds JayjO pue sjaseyep asn puel/eACd pug] |e 40 }SI| Ja|duod e SI MO|ag

eyep jeneds ajqejiene jo si] aJajdwo5 :¢ xipueddy

98

jpaz6
“6861 Ozlleg JO OS/) pue] as — ajqejiene ABojopoyjaw paylejaq

|__

“eyep yyoq
juawaduuoo sn }ouUed am Os jase}ep (S,}!WS) Sly} Ul papnjoul jou
ase yeu) SonosHuew se payissejo Seale SnoJOWNU Ule}UOD Jaye] Sy)
“‘dewoo 700z ay} ul senosBueyy ay} Woy JUaJayIP Alef”, “UMOUMUN
S| 12}SJ BOINOS AU} }NG Ja}Ses © WO PSLAUOd 9q 0} SIeadde

eyep au} ‘os|\y ‘suoBAjod jjewWs Jo s}o] Sule}UOD puke pajle}ep a10W

Ajjesauab si eyep ay ‘7 762 pOOT eAcuBUeW pueRjulew zq ‘914

vS | 0€ SX LOdS

26-06-6861

ysa}e|
shes |lwy

€ xipuaddy

(p66L) AeID

azieg 9 Joyyeanuie4

sonoibueyy
Ww / IWINZD

Ajuo azijag

senoiBuew

“9ZI[9g 10} UOI}do jseg JEyVEq 1401074
pawinssy ‘dewooy ¢00z au} 0} pasedwwos sayepdn Jouiw Ajanyejay MO|Eq 89S Aluo ezieg dewooz 7g
-aigeyiene si eyep ayeunooe SOW Puke JUBA BJOLW YOIUM sanoibuew
JO} ‘8ZIJag 10} }deoxea ‘aounos se asn 0} UONdo jsag ‘ebesaAOd |IN4 €00z dewoo3
Wau} BSN JOU 0} }Sebons
OS BJEP DINDM Jo Aoesnooe ul aouapyUod JO yOe7] “(UOISISA 700Z)
AZI|/Ag JO} JUaIaYIP ALBA JNQ (UOISIBA EQOZ) SEINPUOH JO} aes a4} OSIX4aIN
$1 e}eEp au} Jey} PEmMOYs Gewooy au} YM UOSHedWOD ‘payepdn pue sayepdn jou yng soanoiBuew
snoue, Buinujuo9

penosduui Ajuanbey (jj) Ajpesoddns si yng ‘saaouBuew smoys AjuO

“2002 Se] ul ejewiayend

ul doysiom e je SeujUNOO 94} UdaMjeqg sjusUee/Be WoL

peyjnsad yeu} Wa}sAs UO!eoyIssejo au} 0} SeBueYO SlWos sapnjou!
pue ‘edly B}S0D Jo UOISIAA pasiAal Ajajajdwioo e sapnjoul ‘dew ou}
UO SJOJB UMOU $}998JJO9 UOISJ8A PasiAel a4} JoAemoY ‘AueHeul! IL
JESPUET GE-C661 SwWes a4} UO pased j[I}s SI YOIUM ‘(1 00Z U! pasea|el
SEM UOISJAA }SJI} BU}) UOISIA EOOTZ Su} SI SIU{ “Seas UDI) 10} BuO
pue suoiejue|q JO} auo ‘swiajsAg jeunynoUBy Jo} ssejo suo sule}U0o
(uoneoyisselo OOSANN pey!pow e) puaba] wa}shsooe pay!ejeq

JO}SOSa\\|
yBnoiyy papeojumop aq Jouued eyep JAAD puke] JEU} SAAS }]

Npe len JSeO@SIIm pue Aob eSeUMUIMM jaluep = SjIe}UOD

‘ROB €SEU O}SSU IIAIOS//-dyy ye 10}Sosay) YBnosyy
ajqeyiene suake| eyep jeuoiBaJ JayjO “‘paynyysqns eyep SIGOW }ua0e1
— a|geyienk JOU SI G/~OOZ 40) EYep UeaW Jespue7 Uy SWaj|qoid
jeoluyse| Zauljaseq e se eyep WWYI1S pue yespue7 Q00Z ASN pinoD

Kdoo
e sey daor

puebe
ysiueds

UJIM SOSSEID ZZ
‘uoleoyisselo
OOSANN
Pey!pow

jesidoiy ued

OWOM-daNn

000 0SZ:1

‘ejep
G yegpue]
uo paseq
‘suoBhjod

(espueq)
wi og

yoyli-sod pue

-ald yjoq ‘66
-€661 ‘Ayjunoo
Aq souen

S00z pue 7007
‘€002 ‘0002

(oo1xaIy

Burpnjoxe
eouaLy UOISIAR £007
j24}U89) (avo9) dew
jeuoilBay suia}shsoog
jeuoibeay | YIAYSS VSVN

28

‘Jdd 001Ke
7eljaioul en/xurWieun yesboab! cApul//-d}y4

W 0 eu/zo0c/eyeAl

ao ehew

EAS U! UOISNOU! 10} ONL Aq passadoid-a1 pue QOOZ Ul dIIGnd apew
eyeq ‘sabewi WW yespue] e661 Buisn uoeoyissejo 1809 pue|
9861-8961 S,colxewW Jo ayepdn uy “~E6| JO AoJUSAU! }SaJ0} UBOIXEY\y

JO-EARBLWEAISS MMM//:d}),4

/UOHEOjUSA/SO}OUd jeWee YW pajyajduioo YyBnoy} ‘dew
wa}shso0a 984} JO} pasn jsapjo au} si AieBew! Wy ese jeu} S}ON

‘STIVLAG YOs NOILLOAS AWNOIDSSY/1VE019, JAS

‘dew wajsksooy

L00z leniu! ay} jo Wed UeJNpUOH ay} SI eVep siy} yUILY | “GOOZ IsNBNy
YyZ UO B}JEP UO UOH}EWWOJU! JaYyYNY J0j JaBUOWAL| UeSNS pajlewe jlWy

“eyep Jug09J BJOW BAeY JOU SOP 1 VWID

‘

“Aiysai0} pue Ayloeded ‘asn pue; papuawiwodal suleju0D

Wi YOWUW-See/ajeos SSOlo/bl0 Jeibo jelo qamsib//.dyy

W}YSWa}SASOO5/OjuI AjIsiaAip-jeoiBojoiqs//:dyjy
ipd'pooz ezq sAsooe

pue jpd'(o1qgnd) sisAjeuy ded dSdVdN 99S ‘Spurs! pue sjoas
au} Sepnjoul os|y “o}e ‘(yO uer) AreBew! WL |euoHIppe ‘Jompjay
ySnoiu; uodn paaoiduu! jng ‘Aiebew! WL 66-S661 Woy AjjeulbuoO

JPa"| [OA - 100% Opiges 9 UEWSS/Y
“2}8]0SGO MON ‘dew Wa}shsooF Jo UOISJAA (L00Z) JeNIU]

S002 euNnr 9}e| ul
0} papiAoid — ajqeyrene ABojopoujayy Oa!oid YYIWYYN JO s}insay

766) pue
9861 sey deor

Adoo
e sey deor

WOE /ASZL

paseq

‘ouBe aydiiny
"sesseio

gi sey 000 OLL:1
puabs] yneyaq WA
"96 JO [B}0.) yespuey

000 0SZ:1
yespue]

000 0Sz:1
zg yespueq

WL ‘WOE | P66) Pue 9861

€ xIpueddy

0002 OoIxal|

000Z Auojuanuy
jSas04 jeuoHeN

oolxal/]

sojoydjevee +
sabew! (jsem)
86°CLXL ‘PE/E

XZ '€6/2 XZ

sSeJNPUOH

(aBe1ano0o
Ayyunoo
|INJ you)

seinpuoyH

s90inos
Jayjyo Auew

Asoyuanu}
}SdJO4 JeuoeN
all BU9S,, IDANI

dew
wa}shsooF ¢00Z

ejep uonejaba)
JeBuowadly

ate)

SJOJCSIPU! LVID
“SENY YO}
BUEDINH LVIO

+ 70-66-86 azijeg jo dey,

WL yespue7] ezijag | weshsooz 700Z
(L002) opiges

86-96-€661 azi|}eg 9 UBLUda//\|
(S661) Mexo1g

661 9zI|/8q 9 JaBuoWal]

‘JOU SI}! Jey} SMOUS SUOBAjOd ay} Jo AejiaAo ue YBnoujye ‘dew
Wwaj}shsooa ay} se awes ay} SI SI4} }eU) JseBHns AYd ‘ OGTASWH
ul eyepejay ‘Juasaid ye paeys aq youUed ‘Waj}Ul SYgW Aq pauinboy

“pajajdwoo jsnf sem ejep ay} uaUM ‘eQ0z

judy ul deor 0} eyep siy} poueys sowey OBNH J0}01/ ‘Ua}aq AjUO
SJ9AOD 1NG dJeINIIE PasapISuOD pue Paliejap AlaA ‘ZOO YUeW
Woy EJP +15 LYSANV1 Pue 1OdS Buisn ‘ejewayend ‘saso|4
‘SOM Pue dYNOO-DAWAD Aq pajyeaio sem sjaseyep siu} ‘76

-9661 Ul BUIeYy JO AU BU} Aq UayYeLapUN YOM ay} Jo UO!}ENUI}UOD

L00Z ul deor ym e}ep siy} paseys uyns jaiueq 000 u! peysi|qnd
‘Alabew! 86-266) Woy jeuoeua}u] UOeAasUOD Aq pa}yeain

/Np|/Nps nyejo ywes//-dyy
$/yo1easal/Aydelboeb/s}jueuipledap/npa myejo MMWM//-dyY
“eyewayeny

ul uajaq Buiuiofpe ease ul uayepapun Apeasje wes6oid Buljapoyy

JwyYysd

(pooz) ye 18 seW

“OOIXO|Y OpNfour },UPIpP
eWOdZ 100 JeUuIDIO SU} yeu} JoAamoy “ajON ‘dew wa}shsooy
UBDAWY |€1}UBD OU} JO UOISIAS’ EQOT BY} Jo Wed se GYDD Aq
ajqejiene apew jas eyeq ‘AljUNOD BJOUM SJBAOD ‘papiAosd Aoeinooe
JO JUaluSsasse ‘pasn osje sojoUd jeay OdIXa/\J 10} UO!}EOYISsejO
Jano pue| aj9|dWOd e pajonpuod “LYNYVWWAS / IDANI/ WYNN

88 € xIpuaddy
umouyun
sueak abewi
BOINOS “L002

ve Wy 0SZ aya|dwiog eyewayend VOVIN

sieak om}

Asana saBueyo

Adoo }SO10j + JO}EM

e sey daor ‘ouBe ‘jsqu04
Adoo | ‘ou6e yorum jo
e sey door | ‘sassejo Z1

Adoo
e sey dsor gL WW OSZ

z00Z 9} dn

WL yespueq
86-2661

A\uo uajad

XN
awos pue

‘Za ‘19 ved

UWJa4ON

0002 ‘6661

S}E}S
ayoadweg
‘ueyeon,A
“S ‘OOIxaYy

7002-9861
saBueyo jso04

pue jsao4

i(fe))
dey uolyeye6an,
ekew

YNSOO4
/ PANE / He|D

(9nJ} a}iINb jou)

dewooz €00Z

68 € xipueddy

Jayjeq Ajjelauab wees sjasejyeq S[e}0}) NN Yoyew! Oo} Ajisuap uoNe|ndod pajsnipe :pajsajes
aiem spub ,6e, ay) ‘uoyeindod jenjoe ay} 104 “OW] S.LVID UeU} ayeinooe alow sueedde
puB Aysuap dod pue ‘g,oz pue OL0Z ‘S007 10} peyeusa pue ‘Q00Z PUE S661 ‘O66L pub wy | ‘sleek g jo Aayunoo

JO} Aysuap uoneindod jemoy “so 2eq Ul aiqellene EJeEP dINMU leuy pue E UOISIeA ysa}e7 | ‘Adoo e sey deor | sdajs ul 'g1oz— 0661 | Aq pue jeqojg, €A Md9

‘Md ue} ayesnooe ssa AjGISIA SI JaSeyep SIU}
jeu} weajo sawooeg }! abew! yespuey e uo Bulkeyano pue EA Md 0} pasedwiod usu
jepow Ajiqisseooe ue YM pajeald sadeyins Jayses pue (jun ullupe Jed dod) sdew
uoneindod s0joan sule}uog ‘7UOME|NdOd/H10 JeIDS Jel Gamsib//-dyY ‘ays qam 1VWIO woi4 | Adoo e sey daor | wy 1

LID aseqejyeq
‘dod ueaqqued
pue eouawy ule]

pub
‘0002 — O96

(OWT) leuolbay
Ayyunoo yore
JO} ay ajyeredas
ul — jeuoibay

SoalJUID
uojeindod GVvOO

IN3LX3 NV&YN / NOILVINdOd

‘payoeye eyep uonejndod ou yng ‘aweu UMO} puke JOHN}SIP UO Bep Sule}UOD | Adoo e sey daor

Adoo e sey deor jeuoiBay suod GVO90
Adoo e sey deor jeuolBay syodiiy QVOO
‘sdew o1ydes6odo} wos4 | Adoo e sey deor seinpuoyH speol 1VIO

‘uo}Buay [IW Aq pajeeso speod pasiney (GOOZ IWWZO SavOU-HiNos za) | Adooe sey deor S002 ezijeg Wwayjnos IWINZO

(‘9'g Woy UONOSUIP WayPLOU a10W UI BuloB suo au} 0} pasoddo se) uoloeuIp
Adoo e sey daor €0-000Z

ysemyou ul AjID ezIjag Wo’ peod Jofew ay) jo UOISNjOU! au} S! Jese}eEP QYOO/MO}SOseW,
Adoo e sey daor & SO66L

ay} UM abueyo ejqisiA solew y ‘Ueuaay) Aq ‘Aiabewy yespueq €0-000z Builsn ‘jaseyep
SPEOJ (D7) $,A}UaD UOeWJOJU] PpUe] Au} JO payepdn ue si sil} (G0: 99q) Mioyuaau| eyed
jeneds ezijeg ou} 0} Bulpioooe J wo paniesey (€0-000% SaLa Savou za)

(yes ejBoos wo) Asabew! ym Buiedwoo ueym speos Buissitu
pue saiouedaiosip snoiago aie aiay) ‘Joseyep ajeds Mo] Ajiiey e SuleWAd }! “JaAeMoyY ‘e}yep
ajqejiene seg ‘syoes/sujed pue ‘peod Jey} ‘(AjUo uoseas Ap) panedun-AemyBiH ‘(Je90h
j1e) pevedun-AemyBbiy ‘paned-AemuBiy :sassejo g ul payissejoas uaeq eAey ejeq jasyo
winjep e aney },Usa0p yng eyep GYD eu} se ewes au} ejdioulid Uy (NIT ‘TWIA am)

uo}Bulwayy *4

Ajuo azijeg | ‘ueweayy c

jeuoiBay Jo}Sosaly

‘eyep asay} asn 0} JOU paploap SEM }! OS ‘Wy | < 40 JaSyO

peq e uM jng ‘eyep ewes “(jlWy WOj) dHS*°NMONMNN-YA WATAS Sdwou XW

-19-za pue (qo ekem enjeag Woy) dHS° NOIOWDINOAWOD WSIA jeuoibay

speos eAey eNOS

Asyyunoo yoee
JO} ay a}yeredes

JW SEGEWI/SO}USUINDOP/SM peda MMM//-C}YG
‘(siuj wooly deyns = eyep = GvVOO-Oste) «= WsCZ-00L_ = ynoqe: = yO. JESYO

ue JaAeY 0} SlU9as Os} ‘SISAJeUe 4JOMJaU JO} POOH jou — sal}io yo abpa ye dos speoy | Adoo e sey door ulus —~—— jeuoiBay speoy GVO0
savouw

pajsanbay 4eak aounos
s}usWIWOZ | / paniaoay | / uonnjosay yuayxy yase}eq

SLASVLVG YSHLO

06 € xIpueddy

ujdap jlos yo puB sey
ose Ng QVINALOS 24} Se Salvadoud jios aes ay} Jo UeW Jo PB sulejUOD Jaseyep FSIM pub aesBbap s'0
TeyEQ+OYdeIH0S5 +105 /eyep+oneway |/e}ep+|}0S/S|OS+NOqy py /bu0 OUST MM] AyyY ‘g00z Jaquiaoaq jeqo1D CAASIM OIUSI

ZO) =pyuejuooy Wo Xepulju OUST eu GyY pue eolawiy ue)
INS] WO GO/E/EL UO 824 10} papeo|UMoP yng (Pr$) OW4 WO eseyoInd UeD | Adooe sey deor eH IN OV1 HALOS

silos

BP SID HALOS 8U} 0} poexul| eq ued yep esey] Sa}eu}se
ESVO?T%=PHUS}UOIE WY XSPUl/|U"OHIS!SUul]//;ayY Jayalueied yim
paulejgo osje sem sa}ewU}sa suajaweJed IOs Jo jaseyep Auepuooas (ueaqqued au;

“Buljjapow asn puej JoJ JUdIOYJNS UY} BIOW S| UONNJOSa/ Ly | “(UiLU OL pue
G ‘GZ ‘08S DE) SUO!Njosal jeJaAes U! UONe}IdIoeid pue seunjesedwia} UeoW AjyjUOW Se}
aSN 0} eyep jseq ou} palapilsuoy “Wj WOpPPOM/WTOpLOM/npe AsjeyJeq Oaboiq7;dyy | Adooe sey daor WIJDPHON LYWIO
wy Us EOZLNA/Z00Z/PS/D10 Oey MMM. TAY
‘(paysanbas sem Adod) INOY-GD uo spueMuo
ajqejiene (SeunpuoH U! COL ~ ‘ezI|9g Ul Z) SUOI}E}S DEWI}O Jo YLOMJaU Jego|H e Woy eyeq O96, Woy e}ep aWOS leqo|5 wid OVA
SLVIAIIS

(/Bio Jeibo so wySs//dRY “WQGO} (LVI) ‘adojs
wo. papeojumop) joo} yealolg Ja}sey paeziwojsno sdeaor Bbuisn 2z6L pajdwesei ‘(WwW 6) pue uoleAgja
GVN/9L WL 0} payaloid Ajayeunooy “eep O6 WLS Pally S.LVIO Woy deor Aq payealD oas-Bep OE jeuibio Jeuoibay | Jo} WAd WHS

3dO1S / NOILVAS 13
ap ekew enjas wo 19 s}UaWa}}as
azis uojeindod pue sjuawajes (4HS*000ZD WATSS SLNINSILLES xW-LO-za) | 4dooe sey deor 0002 punouy | jo ped pue Zq ‘XW | eAeW BABS

‘lw WOd pariaoal 'saipnys }Se104

jeoidosy azijag Woy azis uoneindod + sjuewemes (aHS ‘SALA SLINAWATLLS za) | dooe sey deor

umouyun azijeg | sjuawajes S41g

uo}Huiayy |i Woy panisooy

“@ZIS UOHejndod papnjou! jou pip GYOD Puke |INANI Woy Jesej}ep JayJO ‘SeiNDUOH, s}UsWa|}}es
40} eyep yulod uone|ndod payiejap jsow ay) (4HS*CWOO-NOI SLNAWATLLAS NH) | 4409 e sey deor Whee a) senpuer QvS9/NSI
‘Kayunoo yey} OJ ajqeyiene eyep aaneusaye Ou se jaseyep . pce

s}uawayyes e ajidwioo 0} pasn aJam sopenye WOJj sjulod sjuawalyas a
JUSWA}9S YVIN e all 1p FA EIR IO MES say ele cama saree |eiar 7 stasis

‘AiaBbeuw! yespue 7 wo ulebe suobAjod au} azi}/6ip 0} 19}39q 9q Aew }| ‘a}esnoOe AJ9A JON

PaAlaoal JA JOU }nq pa}sanbas ssacoy yood ueospue]

IEqo|

[= ‘9ZI|9g 10} }deoxa ‘Jaseyep Oy] ey} UeUY

L6 € xIpueddy

“seljuNoo
JOUJO BY} JO} eYEP SUSAN BU} YIM Jayeq s}y pue aiqisnejd aiow swaes siyy yBnouy
‘eyeP SIA GWOD au} Ul Ue} SJAA SYaMmay AjUeOyIUBIS sapnjou| “ezljag Ul SYaAIJ Joleyy way} sey dsor azijeg IVINZO

“eJYEP GVOD 3u} UeU} SIAN Aepuodas a0 SapnjoU] “sdew o1ydesbodo}
% OG:) woy pezyGip (sian Asepuooas pue Ayewud Aq pednoj6) suanu jeloiyoO

way) sey deor 40S: seinpuoy Lvid

“EYEP JOJOSA JAPjO BY} UCU} S}jesNdOe SJOW 9q Pinoys SJBAU
payeauljap WAC 24) [!e19A0 YBnoy} ‘sjaseyep (GYOd) Jaujo Woy asou} o} JeoHUAp! aq
IM SJOAU Ul pausing Aj]eE907 ‘WAG WLYS W 06 Wo payeauljap sjaseyep ayeunooe quo

“apnjoul 0} JUeM you AewW am yeu} SLUea}S
wes Asan Ajqeqoud aie asayj—azijag |esjuaO ul Ayisuap Weajs YBIy Ajaxiun ‘os\y
"eyep Sa/einyo SOU, S,1WID Se Sweays Alepuooas Auew se apnjoul },use0p seuinpuoH
Joy Jeseyep QYOO 9y} Woy Buissilu aie MOG Ul! papnjou! sweays Japso |.) [Jews AUeW
}nq ‘payieyap aiow aie eyep GyYOD Jey} sMoYs (aje9s WL:1) AAOG UW UOSHedWOD jaug
‘sdew o1ydes6odo} 4 OOL-4 0G: WOy pazi!Bip Ajsoyy ‘SeweU saA SUIeE}UDD JO}SOSAaL\
uo se ayisqam GyQD uO ejep owes © [UY Xepuljjo}SOSayy/AOD eseu ojSSu JiAles// dhyy

jeuoibay IMM

“Aqyunoo
‘yeuoiBay

way} sey daor 000 0Sz:1 | Aq wds GV99 / J0}Sose/y

syod
0} —Ayliqissaooy

speod ‘sjaysew
payeaio deor 0} Ayiiqissaooy

“eyep Sajid ‘speos Buisn ‘pseziny Ayiqisseooy S,1 VIO Buisn payeag

‘uoseal
Sly} JO} paysalay ‘QWINALOS 24} Woy jUsIaYyIP Osje ase suoNedo) payoid snouwen
}€ SOSSEIJO |1OS By} pue ejep jios eAeW) ay} WO JUBIAYIP Ajaynjosqe ase Aayy JaAemMoH

“eyep eAeW AAS du} O} |IleEJ}apP ajqesedwoo sey }! asnedaq Wy QGZ-00Z JO ajeos e je
Ul eq 0} sieadde eyep ay] ‘psooas UO dew jeuoiBas pajieyap e yONs aney jou op Aau} yeu} (dys ‘j!losoey)
Aes Kauy yng aoinos ejep ayy sem OVS Jeu} edIpul eyepejyay) GO Biequnig 9u} wWos4 é 0S2-002 OV4 / dVNOO

“sn pajzlul|
os ejyep saiadoid jios syoe7 (BIO eAeWEAASMMW//.dyU) Yom eAey) eAJaS au}
jo ed se uoleoyissejo OW 4 0} paziuoweY aJam yng AyjUNOD YyoRA U! JUBIaYIP AjjeuIBUO SeJNPUOH }OU
B1OM UOHEDISSe|O 94} SHUIY} [LU “YALOS OVA 24} UeY} pajiejep jeyeds iow Yyonyy SoA 40SZ'b | Inq ‘195 ‘La “XW ehelN BAJ2S

Jwyywy eulbed/eqozoz%}ausajujAsisp/xw'qob 1Baulyaniasdew//:dyy
eyepejau 3aS yJOjans Osnysooiseq soyep/dew/xurqob 1beul Jentesdeu/- dy
‘TT TEMAIA/OOIXOLUOYISGaM/XU Gob Ibeuroajeb/- ayy Janies dew slay} uo papnjouy éé WL -% 0GZ:1 oolxay IOSNI

(UY Zeeb/ON V/yoeasayje oe esell MMM//-d}U) (OWINALOS ‘'8'!) aymasja ajqeyiene
aq Aew saseqejep Jayeq yey} sajejs ayisqeyy ‘0002 BuluoZ jeo\Goj00Z-o1By jeqo|D ON pub wy OL leqoiQ | 000z ZAVO VSVII

c6 € xipueddy

"SaAoIDUEW AjjJUeUIWOpeld ale jealU] JOPUN Seale auj Jo |je Auean year)

Japun si ey OO} esau} Jo Wed Jeym sajeo|pul yey (BBDYYeD) ,uOHeo\IENo, anquye 19 Jo ele
ue pue yoea ey Ql JO SuOBAjod padeys-uobexay jo pasodwod s,}j see] SOD Wisiunoy | Adod e sey deor LO0¢ | Ue}eq pue Za XIN eke BAIaS
SLOdSLOH WSIHNNOL
‘Injasn Alaa JOU OS Bale JO SaWeU puke eyep anquye syoe] jeseyep GYDO
/ Jo\SOSayy Jo BSOu} se BWeS ay} Je SaUepUNoG UOBAjod ‘Janamoy ‘100Z UeU} JEpjO aq
pinoo ejeq *L00Z ul UOS|aN Apu Woy panisoas daor yeu) GO e}eP You! / SenPUCH UO Ados e sey deor L002 - 0002 seinpuo} LvId
“pasn Sauobajeo NNO! BlEP JOISOSaNW/GVOO ou} Yoyeu! pue ajajdu09
aq 0} sipadde uolsJan siy} pue gooz Auenuer ul OINOM Aq payepdn jeseyeq ‘seinpuo}Y __2@ pue NH ae (OWOM-daNn)
ul Sauepunog ajeinooeu! pue suOIssiWO JOfeyy “SeINPUO}] JO} JaSe}ep S,1 VID Se Jem pue FA AG} PeePe
JESe}eP WDD eu} 0} pasedwioo seoualayip jued\IUBIs pemoys jese}ep GOOZ-GO-8L SUL Adoo e sey daor goog uer payepdy | Ajuo yng = jeqci VddM
‘RJEP S,1VID Se aWeS ay} OS|e SEINPUOH 104 JO}SOSaYy WOy asoU} 7
se aules au} ave (jujy SedeU/SO}JUSLINOOP/SM POO MMWM//-d}Y) aysqam GYOO eu} Woy seale payoajoid
peojumop ejep ay, ‘euesvaweosay ugibal e/ ap sepibajold sajeinjey Sealy ap edey | Adooe sey daor e00z Aew jeuoibey | Gvod / J0}sosey\y

SvauV GaLOaLOdd

€6 p Xipueddy

SOJEP UIS
uoDejaban ese2s3 UD sealy
ouedin ji
(Se@JauIles ‘seJauJeWeD} soagende songonpoud sewaysis
enBe ap sodsn> gag

UeqIT) a SOUI|PJOD SAjDSy fy
JEN He S3}P}S3JOY SAUODEZUC|d Ry
sasn |einymouBy souensadoibe sewaysis
aebly asieds iy Sa|ePaWNyY A SOUEIUEd fee
ssesbeas SOWEIEd am
3991 2/0) Sa|eINJeu sajezqsed
235 UsdO fe SeLIEGES
PUERS\\ $029 JSSPQNs SOIYJOWOIAX Sa|e}sNquy
PURIQNIUS (ey SO}XILU SA[|EISNGUy
JEM a }SBJ0J |210}}!| PUe aACIOUS|y) py SPPEYOYQe] ap saeysnquy
ueq.n) il BUURARS PUE|MO7 SBIAJIUOD 3p Sa|BISNCUy gee
sanoJ0UeW fae :
1S3J0y JAM PSAP3a-PeO/g SUB]VOWANS gy saie|buew sanbsog
foe J }53J0} JSIOW Pares-PEOg SURJUOWGNS Hy Sepeliojge| ap sonppap sanbsog jay
gael: = 38340) BUId SUEUOLUGNS ggg S0}XIW SONpDapiwes sanbsog
aumseg/aumynouBy [—] }saJ0j auld PUE;MO7 Ry SEPe!0j92| 2p sonpDspiwas Sanbsog way
1S3J0) SUld BBB }59J0} JSM PACS POG PUE|MO] fay SO}XILU SAPJaAaidwaisiwss A sapJansdwais sanbsog
353104 panes-peosg Bey 53104 JSIOLU PaABar peojq PUE|MO] fag SEPEOYGe| ap sapsarsidwaisiwas A sapJanaidwais sanbsog gay
UMOLYUN/ JBL i 452810} AIP paAeat PROG PUe/MO] jm SPJajIUOD ap Saplanaidwais sanbsog pay
de; wayshsoo3 paanpay / paylssepay [A] $007 azg wayshsos3 [Al deja) Waysiso2rg co0z [A]
UOIJEDIJISSE|D paonpay / payissejoay azijag dew wajsksooz 700Z eoWauy jeijuay dew wia}shsooz 007

“suonejnuuis aBueys asn pue] YVI\j eu} 10} paydope sem yeu} UOHedyIssejo peonpas pue payissejoe) ay} pue (Sesse|o g})
azijag 10} dey wa}shsooz p00z eu} ‘(Sessejo ZZ) EovVeWY jeUaD JO} deyy Wa}shsooF E00Z JeuIBIO ay} JO} SUOHJEdyISSE}D BY} SMOYS MO|Eq 9/42} 9y

UOI}EDISSE]D J9AOD pue] dey wa}sAsoog :p xipueddy

v6 g xipusddy
91688 SLO08- 0€89c yooz0s_ | o9zzz | evgzelt [| ose [ Sz0z|
8EL68 y9v081 98992 9¢920S 6LS¢L Z882611
LZ€68 9E8081 ZEG9T 69086P C6SZELL
| 6lest | 20sec | reso | coses €OZL8L osesz | seer | series | zyecoct | ose. | zzoz |
| 6lest | lore | soseo | 8968 295181 geez | zo06er | 6zreL | 7S99021 | 202"
6LESL ZvELE 88069 81868 LPsLgl G6L9OZ y8c6lPp 7cLLZbel OSEL 6102
6LESL 9Z60€ 60%69 92268 QGZL8L Zoeoz | 166elp | LgZeeeL
92062 SS269 0LL68 evZLSL 6cv97 ve6s9or 6geeezi | osel | LL0c|
y808Z 02002 87268 659181 eesoz__| Ovzeor | 90152 | elrpeci | oser | 910)
6LESL LOLLZ? Z8EOL 28968 QZSL8L 87997 96S8SP LepSZ ELB6ETL OSEL GLO
6LESL g9L9z Z6SOL 61606 9sgzel LgZ97 GSeeSr g99SZ g09Erzl OSEL vLozc
6LESL OvzSz 00802 87606 SZLPsl 71692 Sr Lerr 6898S 6LZLb¢z2l OSEL €L0z¢
6LESL G6Sr~e 86602 O9SL6 g9Esel 6E0LZ 8L6cbr LOLOL G9Z0SZL OSEL ZLO
6LESL vS6EC vELLZ 99126 66S981 99LZzZ voller ZLEQL L@¢vSel OSEL LLOZ
6LESL 6922 B0rLZ 96226 OZ8Z8L 86cL2 9L9zEr OvS9L Z86LSZL OL0z
6LESL voce? 96rLZ 71626 GOL88L LLELZ vSOLEr GE9OL LpS6SZl OSEL 6002
6LESL 6L9L7Z v8SLi 620€6 Lregel 9Srl7 Oer6er 6cL9L G60L9cL OSEL 8002
gS80c Z89LZ LSLE6 Lo9gel 6ESLZ 6S92cr vesgl vL8zocl OSEL 2002
9¢00z 6LZLLL 98ze6 798881 7Z9LZ 2879¢P 6E692 vesr9cl OSEL | 9002 |
€0vE6 860681 LOLL7 8S9rcr veolZ 609971 oset | —_Sodz |
GLLE6 8E988L L8rZz eLSecr GLILLL 6LY89ZL OSEL | _-v00z |
GLLEG 3 l L8vle erect GLbL Z| OSel €002
186 l al | L OSEl 2002
| elgeat GLI | L002
( bri 00! 6 £9881 Y elsecr | Ss | 0002
YVSA
JB}JE/\\ ueqin) sonoibuel pues BUUBARS qnios JSed/joVy “JO} BUld }S9j0j 1g UMOUUl/)
6 8 Z | 9 S v € Zz L ty) {om

(L) }Sal4 SJoIeW — 8Z119q

“weak yoes jeoijuap! Ajasioaid aie sje}O} MOJ By} pue SJEWIDap 1NOJ aAeY SUaquuNu Uy} Saflj jndul S-4AN7D pue yeeuspeaids
BOINOS JU} Ul ‘J@ABMOY ‘J9}}9q Sajqe} asau} }y Ady} yey} OS SanjeA JaHa}U! 0} papuNnos aie eyeq ‘(awWeS By} }SOWMe JEU} BZI|9g 10} YBNouUye) saujuNoCD
Q}OYM ay} jou ‘uOoIBas YY OY} UIY}IM Je} }EY} SJUNOD oy} Jo Sed au} 0} Ajdde seaie payodas ay} yey} pulw Ul JeEag ‘sjapow ADVI] pue seinjn4
JEUOHeEUa}U] 94} WO’ Ss}ndjno Bulsn payejnojeo a1aMm s}usWasJINbes pue| asey| “(Sasodind jusWISsasse JO} payejnojeo sem jnq pajejnwis jou sem }siJI4

‘

Awundas

OWEUBIS PJIY} BY}) SAU}UNOD INOJ aU} PU SOUeUBIS INOJ JO SUO!JEUIQUUOD [|e JO} ‘ey Ul ‘S}UsWWaJINbe’ asn pue] ay} aAIK Mojaq saiqe} 91 aYL

SOMBUIDS 91N}Nj 10} S}JUBWAIINba’ asN pue7 “sg xipuaddy

G6 G xiIpuaddy

LeScsl | 808¢z €OCLSP S6ESZ 69L6ECL OSEl Sc0¢
QLLc8L LL8ve C8S6bV elSSZ 860171 OSel vc0c
6c0E8L 6962bb O€9SZ OcOEreL OSEL €20e
c8cESL CIEObY 9PZSZ CE6HPCL OSEL (4404
cesses LOLbby c9esZ sesorcl OSEL

O9ZE8L OLLEeby 896S2 vLS8bel

OPSE8l 6ESCrb creel
QLbE8L LLEcbYy ZLZ81S¢1
L8cesl S6ZL bP cOCESCL

860E81 O8cl by oeeg2 SCSPSCL
988csl LLLObY v89Ssel
9OZE8t LOL8EV OLLZSZL

ScSbsl SOSSEY

Ovessl 606CEP OS66SC1
vSLO8t 430)94 9P29L VIELOCL OSEL LLOc
ZE0Z8l SO6LCb 9892 O9CE9CL
OZEZ8L S8cler €L69Z 60LP9CL
629281 LL9OO¢Y SS69L LZ6Lb9CL

QE6SCr 26692 98pS9CL OSEL 200¢
TAS Ag 8eoZ2 vLL99CL OSEL 900¢

886281
86881

809881 98Srerv €9899c1 OSEL S00¢
8e98sl ELSECH 6LY89CL v00d
CLSECY £002

JO} OUld }sal0} 1g
G L

umouyun
0 oni

Jaye ueqin SOAOIDUEL puenen
6 8 Z 9

euueAeS
S

(Z) 3Su14 Adtjog — azaq

96

G xipueddy

989 7LGL8 C62LZL1 LCEL96V 99eel ZL8SOCL

94989 LSO88 €9c8ZL €80L6r €lgez 8986021

S0689 92S88 97C6LL 920987 8S8EeL S68ELCL

cELB9 86688 L8LO8L SOLL8r LOLyZ €68Z1L71 OSEL

LSE69 Z9v68 OLLSt OLLOLY cverl LO8LecL OSEL

cLG69 02668 8v078L 80Z2LLPr CLSPL OvOSCCL

GSZ69 G6868 966181 ZEL8OV OL88c7cL

€6669 Ov668 880c8l SevSor LSOCECL

v8lLOL 97668 090281 €vycgov 8ccGL CCVOETL

vLlEol €L668 ceOc8L COVESP CevSl GLL6ET?L

vSSOL 28868 L86L8L CEvasy 9c9SL CS6CrCL

88S0Z LSv06 ECLESL LOLYSY L99SZ QESEPCL

Lc90Z SLOL6 G9cPrel LO6LSY 269G2 LoLvycl

yS90Z O8S16 80vS8L Se96br CELGL 90LZbrCL

28902 vvlc6 LSS9O8L 89ELbL 89ZGZL L6cSPCL

6ELOL €€Zc6 €vlZZ8b OLCGbr €c8SZ LOZ9vCL

Zv60L 66226 9/8281 OPSLbP Lv09L 9L867C1 OSEL

GSLLLZ S98¢6 600881 9682Er 69292 LEGESCL OSEL

OLELL Cv6CE | 99L88L CCEVEV 00S9Z 62ELSTL OSEeL

Ge8SlZ 6LOE6 LZES8sl 99LOEr OELOL ZOLL9OCL OSEL

68212 €80€6 7SP88L vLLLov 6692 €OLP9ICL OSEeL
GLLE6 8e98sl ELSecv GLZLLL 6LV89CL OSEL
HLLE6 8F988 1 CLSE?P Al? Gl 1A OSE}

Jaye/\\ ueqin senojbuel| puejear euuernes jseq/Jby 10} OUld }SOJ0j Ta) umou yun
6 8 Es 9 S v & 4 L 0

(¢) ysul4y Ayunsag — ezijeg

26 G xIpueddy
ZOLLZ 861681 886z L661) ggg9L_ | LZbE9z) _—~|«OSEL Sz02
bwLIL | 226881 S066Z zeechp | Z689Z esgeszl [ose —Sss|—Ssébz0z
OSZLZ 9pZeel 92662 | rd 6E69L [spspozt | oser €202
6L8bZ 6LS8eL Lv662 Slézly | 18692 veesez) | Ose! aay
8S8LZ | [ 26z9e1 89662 e0eely ezozd __—*| Sz6s9zl __| OSEL zee
86812 590881 68662 LEgely S90ZZ zpeoezt | 0SE1 0z0z
ZS8bZ 6z1881 Sp96z 6LSSLY 910zL | zbesozt ‘| oseL 6102
90812 61881 O0e6z LDL L969Z yoosezt _| Ose! 8L0z
O9ZLZ eszeel S687 vZe6Ly Bl69L__—«| @6Lv9z _—*| OSEL (aa.
PLLLZ ezeeet s098z ovZlep 8989L gleeozl | OSEL 9102
LOQLL 68eeel SSz9ez Soler gigoz__—'| gsszozt _| oseL SL0Z
e69LZ 662881 vvS8z SoSeer 9pB9Z orocoz! | oSeL blz
SOLLZ [ecteal L28ez S8elzp 6S89Z veecoz! | OSL E102
LULUL 8bS681 OLL6z voclzr ZZ89L eeveozt [loser | —cloz
OfZLZ €z6681 v6E6Z €zg0zy | $889z ssgeozl | OSeL LL0z
69LLZ [rzZe06} 68962 0000z¢ L269L erergzl _| OSEL oL0z
S6LLZ 622061 e6Lp9zl | OSEL 6002
LLSLZ £90061 [z6osez! | Osel 8002
8ZBLZ 968681 ggesozl | OSEL 2002
Spell 822681 —[tsesozt [oser (| —‘s002
L98bZ 6SS681 elesezl | OSEL S002
000zZ gegee! 6lys9zl__| OSEL 002
00022 g29881 51 OSE! £002
PG 9002 Ssos | z002
veel | 0002/ SL1€6 ig 1002
= L186 0g 0002
lk __ [| uv3A
Jaye ueqin senoibuey| puenann euuenes qnios yseq/iby JO} Ul }S9J0} 7g umouyuny
6 [ 8 l 9 S Owe 6 Zao onee Oe kom |

(p) 3s414 Ayiqeureysng — azijag

86 G xIpueddy

OvSSS x4 vlyece Z6LLGL L0c9 O6rl9St | ZSLerr [0 esolloe [ozise | — Szoz|

Ovsss ree e6lezz | vosler [oezo | rerpsst |ooczer [0 | eszezoe |ozsz | 0c |

OvSss yc0E7 gl6ec~ LLYZ6L 0vz9 geeorst |esoler [0 —[9Leseoe | OZLSZ €z02

Ovsss Ltesce Rx 744 CLOC6L 09z9 ZOLBESL euoser _{9__ OSZZP0E

OvSSS L8Lez veESec 90961 6229 CEOTESL 6rb0cr ZLO6SOE

|orsss |6rzizc | veoge7~ Z61L¥61 Lesszest =| i96rlp =o 0S—~SC=*Y:«C 90 OLLSZ 0202

OvSss LOZLZ 6cc9c~ OS9E6l LLV8ZS1 poseeor {9 __ €LZL9OL0€E OLLSZ 6102

OvSss 18202 7ev9Cr €LOE6L 6CELESL 26000¢ ZESL8OE OLLSZ gL0c

povees _t vaca ___ 1 80ce veSZ6l Gplyest | 1oee6e [0 | O6Z980€ OLLSZ L102

|Ovsss | v6Z6l _—| 0089zz p8616L ozegest |esesse [0 | €z6L60¢ OLLSZ

OvSsSs vOE6L Gg69zz -esret 9S96ES1 7S68LE 0 €v6960E OLLSZ

OvSss €egel | S60L¢z LOzz6l 0627S) G8SLLE 0 CQOLOLE OLLSZ vloc

OvSSs 69E81 | €0cZ¢z Ly6Z61 987SrS1 ZOEPOE 0 Z80SOLE

Ovsss OL6ZL OLEL2z LLOC6L 9b08rS1 O8ZZSE 0 LOO6OLE

OvSSs BSrLl Glee ZEKE) LZZ0SS1 GEEOSE 0 GZ8ZLLE

Orsss ZLOLL 6LSLec €OLS6L ax) O9PESSL veSEre 0 vSSOLLE OL1LSZ OL0zc

OvSSs gS991 62SL22 9S6r6L [ezeo «| 9zevGsi —| Lzozve 0 O6LZLLE OLLSZ — Sone

OvSSs ZOE9L gesl7z 608r61 gLeg Z8L9SSL SZSObE 0 bzeztle =[ozrse si 8002

OvSss ZS6S1 LbSlez 79961 eLeg ZvSLSSl QE06EE lo |[€bpperre |ozrse | 2002 |

OvSss z9Gl 955/22 SLSP6L osessst |esszee [0 ZGO6LLE OLLSZ 9002

Ovsss 6251 v9SL77c OLEr6L Gez09Sl BLO9EE 0 OLOBLLE OLLSZ S002

OvSss | 996¢1 Z6E8TC €Z6E6L 8L08bS1 LESQee 0 LZOLELE OLLSZ 7002

Ovsss 6SZY1L LL@627% LLbe6L GZESESL BLSLEE 0 GZecrLe OLLSZ €00z

Ovsss QZSrL ZvOOETS 6LOE6L 6r9EcS1 GZESEe 0 OS9ESLE OLLSZ 2002

OvSSs LeSrL L980€z LOSZ6L Svz9 OLZELISL | ¢g06ee | 0 698H9LE OLLSZ L00z

Ovsss OrSrl osglez | OOLZ6L 0ez9 09066r1 OO86EE 0 O609ZLE OLLSZ 0002
YVSA

JOJe/\\ ueqir) sanoijbuey| | puee/\\ BuueAeS qnios JSed/iby JO} BUld }Sol0j “1g UMOUUl)
6 8 Z 9 S € z L 0 oni

(L) 3sa14 sjayseW — ODIKayy

66

G xipuaddy

Orsss GOSez 8brce7 v76c6l a) LLZeprl L8rSZe vLOOBLE OLLSZ Gz0zc
ees | GeScez 6CLEBL €9z9 GOESPrL QSSELe 0 66/881 OLLSZ | vz0e |
OvSSs e2Lee% CSeebl WEA) ieee ae Oa Ye | Oe aor]
|orsss 8S82E2 QESE6L LL29 GLEErPL (Aan a AS a ya ar]
pees OZLLZ v66CE7 | LEZE6L €879 LLberrl ZS8L9E 0 ZOLPELE OLLSZ

OvSss vllO@ 6CLEET SE6E6L 0629 Leverrl OL6S9E 0 ZS6S6LE OLLSZ
lorsss —s- | EL vOz 908ZEZ ZEEEEL 0229 O€s0Srl 7CEVOE 0 SZSIBLE OLLSZ

Orsss S002 G8rze7 8ZLZ6L 0Sz9 LOLESPL v89Z9E 0 LELZOLE OLLSZ

Ovsss L896L LOLZET O€LZ6L Lez9 €9PSOrl SSOL9E 0 OLLZSLE OLLSZ
|OvSSS _—_—s|s-LLE6L LG8LEz GESL6L ZLz9 |SLZ2Zbl | Leb6SE 0 gerezle [ozise | — gtoz |
|OvSSS _—*|- vEGRL LESLETZ rv606! | z6L9 | yc66Zbl | SEBZSE 0 celpzle [ozrse | —_—stoz |
Ovsss vosSsl L8L0€7% €v6061 Z6L9 GcvS6rL €6LESE 0 GSBCOLE OLLSZ vLO7c
O”sss 82081 Lv00EZ €v6061 Z6L9 BEZOLSL OO86rE 0 OOLESLE OLLSZ €L0c
OvSSS OS9ZL GLE6cC €v6061 7619 698SZSL QG8SPrE 0 ELOEVLE OL1SZ coca
|OvSsS | GEZZL Z6S877 €v606L Z61L9 ZL80¢S1 6S6LPE 0 LOLEELE OLLSZ OCA
Orsss LLS9L 828172 €r6061 Z6L9 68SSSSL SOLSEe 0 ZLEETLE OLLSZ 2
Ovsss L6r9L | €L08zz vELL6L al 6619 ZScrSSl C6SLEE 0 6LESZLE OLLSZ 6002
|OvSSS _—|- PZLOL LvL87z vZEL6l s0z9 vL6cSSl SLOLEE 0 BS9LZLE OLLSZ | __- 800 |
ores L8zg7z vISL6L LLc9 QLSLSSL CQG9CE 0 L6V6ZLE OLLSZ | 2002 |
Ovsss GLv8cc | COZL6L L129 zvc0SSl 6YO9EE 0 LZELELE OLLSZ | __ 9002 |
|OvssS _—s«|§ SEzSL 8vS872 | £68161 €@z9 8068S GESSee 0 GSLEELE OLLSZ G00
|orsss |sSeé6rl _—_—| SZzL6zez Ov6L6L Szz9 GE06ESL y6EQEE 0 E6LLPLE OLLSZ 7002
ose _ 6ELbL G086zz €86L6L 9729 90L6ZS1L SreLEE 0 Z6EOSLE

Ovsss 8SSrL vevOET 9z0z6L 87z9 SrL6LSL vOLSEEe 0 €O06SLE

Orsss ezSrl LSOLET 90261 |6cec9 —__—*«|s #ZL-GOSL €G68EE 0 LSGLOLE

Orsss | OvSrl OS9Lez OOLZ6L 0¢z9 09066F1 OO86EE 0 O609ZLE

UWaA

| JaJe/\\ ueqin SSAOIDUB// puejea\\ BUUeARS qniaS jseq/Jby “JO} BUld jsos0j “1g UMOUUl)

| 6 8 Z 9 S v € 4 L 0 ont

(z) ysat4 Ado — 001xeW

OOL G xIpueddy
| ores | BLEET |9Sr6ce _—|:~O€Sz6L vvc9 96z8Er1 666Ecr 0 S6SSrLE OLLSZ Gz0c
Ovsss €00€Z | Lyg6z~ —s|s Le ze6L 292@9 9Z96Er1 O6S6LP 0 |zerspre | ozise | — zoz|
|ovsss | rz97z_—s| eRGzz__—C| SBE 0629 br60rrl | O€ZSLP 0 |epgoste | Oztsc | E2007 |
|ovsss _—| oSzzz GO000EZ Ly9v6l ZLEg Srecrrl 6L60LP 0 |seresre |ozrse | zzz |
Ovsss LZ8L7z v8LOEZ OSeS6L Seca OPSErrl ss990¢ 0 Q8SSSLE OL1SZ Lz0z¢
Ovsss 96rL2 Z9EOET LLOO6L LS€9 LiSPPrl serc0r 0 OZO8SLE OLLSZ 0z0z
Ovsss GZLL7Z G8LOEZ OLPS6L Beg LOSLSPL ZVEBBE 0 v6SSSLE OL1SZ 6102
Ovsss 6rL02 0L00Ez LZ8v6L | 8Le9 LZL8SPl QLZ96E 0 LELESLE OLLSZ gL0z
Ovsss 69€02 LE86cz Svcrel 00€9 8Z9P9r1 THCEBE 0 6Z80SLE OLLSZ LL0e
OvSSs ©6661 9996722 S99E6L | 69LLZbL | BEZOGBE 0 G8rSrle OLLSZ 9102
OvSSs vO961 8676727 Z60€61 Lo9zZZpl | sgzzee —s [0 ELLOvLE OLLSZ GLOZ
Ovsss Z8L6L €09872 L80€61 LOSS6rl 9S6cge 0 6OBEELE OLLSZ vLozc
Ovsss 99281 OzZL7z LLOEBL GLLELSL LOZEZE 0 OOSLZLE OLLSZ €L02
8rs9cc L90E6L |19c9 ~~—«|| ezgoest —| OOSHZE 0 ZHB6OLE OL1SZ ZLOZ
986Szcc LSOE6L L9c9 vORZySL LSEOZEe 0 ZEO860E OLLSZ LLOz
O”sss OPSZL 9ELS7z SrOE6l 19z9 O68P9S1 €SZ99€ 0 ELE9BOE OLLSZ 0102
Ovsss 6SLLL L09S7@z 8Z8Z6L vSc9 LZGLOSL L6ZE9E 0 vE8ZG60E OLLSZ 6002
OvSSs pLZ9l yZ09zz Lloz6L | Zbz9 Z8L8SS1 oZE09€ 0 | 8€z660E OLLSZ | 8002 |
Ovsss €6E9L Legge L6EZ61 0vz9 v88rSsl OLPLSE 0 L8SSOLE OLLSZ 2002
OvSss GLOOL G669¢c v8Lc6l €€79 ZLOLSSL L6SPSE 0 pOSLLLEe OLLSZ 9002
OSrLzz €L6L6L 9zz9 QLE8PSL €SLLSE 0 L608LLE OLLSZ G002
vresl gLegz~ G00z6L L@z9 GOP8ESL OEe6re 0 B966CLE OLLSZ 700z
OvSss LO6BPL LLL 6c zE0c61 8729 €8r8cSl O€69re 0 689LPLE OLLSZ €002
Orvsss LOLPL 8L00EZ 6S0c6L 6cc9 ZEOBLSL EvSpre 0 ZLEESLE OLLSZ | 200z |
OvSSS 96S¢1 [yseocz | reoz6t [ecco [oreeost |roizre |o | toevore [oztse | +100z|
OvSSS OrStl osgrez | oolz6t ~—s foeza =~ OOGGHL [oosece [0 | o60d/1e | ozISzZ 0002
JO}E\\ ueqin sanoibuely puejan\ euuenes qnios jseq/iby “JO} SUld jsoj0j ‘1'q umouyun
6 8 Z 9 S v € Zz L 0 oni

(€) 3s414 Aylundag — oo1xaW

LOL G xipuaddy

OvSSS €LO6L LS661e ‘| OLPLEL | 80¢9 vO880Z1 908862
|ovsss | geser _—i| p9661Lz | S9OL6L | 9619 E6880Z1 120662
Opes | 208! __ 216612 €12061 S8l9 €8680Z1 LEEGEC

OvSSS O9r81L 066612 LOEO6L vZt9 9Z060Z1 €09662
|orsss ss [zgzet | p000zZ __—| 010061 Z9L9 LLb6021_ | 698662
|OovssS _—-| 9gog1_ _—| gL00zz 859681 LSLO | s9z6021 —_| 9EL00E

LOESLOE OLLSZ TAA
GLYSLOE OLLSZ vc0c

vSOSLOE
9E8SLOE
cCO9LOE
ELCOLOE

DISO/OJOJOJSO/O/OJOJ/O/o|a@

OvSSs vg90zz ZOvO61 GLL9 6LLS691 SSLE0E QPESTZOE OL1SZ
OvSss 6SEL7Z GOLL6L 0029 GELZ8OL 9Lz90€ G6SPEOE OLLSZ
Ovsss | 8OZZ1 Lv07Zz ZEBLEL G@z9 | 6cesool | ore6oe | GS6EPOE
OvSSS | ZZSZL eele7z 60261 0Sz9 O9EPS9L ZvyeLe OEPESOE
OvSss ZEVETTS G6vE6L | S£z9 9ZZOPOL ZLOSLE €Z0E90E
|ovsss ss [ Lelzb | zzz 09ze6l 8929 a a

|orsss i eyaaaad 22061 0929 €SPrz991 LELZLE 0 ELSOVOE OLLSZ €1L02

OvSss ZE9L77 S6LZ6L €Sc9 69PELIL Q6POLE 0 LEESEOE pete 7102

OvSSs 6E0Le7c GOSZ6L Svz9 SL rPsoL LL8g0E 0 LLZOEOE OLLSZ LLOZ

Ovsss |s6est _—s| ospozz_—s| Zeez6L | gez9 c0es69l _| SELZ0€E 0 |zevecoe | ozvsz |_|

OvSSs |9LlZst _—| 9SzLzz 2897261 6¢z9 2970891 ZELOLE 0 E6LEEOE

Ovsss LL0ee7z OvOe6L 0979 ZZOS991 6SLELE 0 Q9EPrP0E OLLSZ 8002

Ovsss G687cc L6EC6L ZLe9 SLZ6p91 LLZOLE 0 ZS9SSOE OLLSZ 2002

OvSSs L7LE7 SSZE61L v8c9 | Z6lve9l | LOE6LE 0 Z90290€

OvSSS g9Srcc €ZL HEL 96c9 | LISSLOL —s |: Leweze 0 06S8Z0€

Ovsss 696S77C ZELEBL €8z9 8ZLS6SL LvBSce 0 v08Z60€

Ovsss | eesrl —s |: Leezez | €€€E61 68rLZS1 L6Z6cE 0 6YLZLLE OLLSZ £002

Ovsss 8orrl v088zzc | O€6Z6L OLOLZPSL Q9LZEE 0 GO99ELE

Ovsss 69bbl LEZOET | SLSZ6L €9PE7cSl g9Z9EE 0 LOSOSLE L00z

Ovsss OvSrl OS9LEz OOLZ6L 09066r1 OOS6EE 0 O609LLE 0002
YWSA

JOJe/\\ ueqin, sanoibuey| pues BUUEARS qnsaS seq/iby “JO} SUI }SoJOj “1g UMOUUl)
6 8 Z 9 S 4 € zZ L 0

(p) 3sa14 Ayiqeureysng — ooxay

ZOL G xiIpueddy
€l2L [oe |[velser |oceeecr | tees ZZ60SS1 ¢z8 Gz0z
08zl oe __| resver_ LeZOezL LELSL 6yLLOSL Sz8 yz0c
9gzl o¢ Geeecr LLLZLeL L@z9L 69ZLZS1 Sze €7z02
e621 0¢ cSLecr eeecoel ee vrELSSL Gzg 7202
6621 o¢ 0260Zr | 8SZS6LL | OO@LZZ OZELBSL G78 | _—-bzoz |
GOEL 0€ L6L6LP 1889811 vecLogl G78 | _Oz0z |
90El 0€ 6ELecP L6PSLL1 VLOOLOL Gzg | «6 LO |
LOEL O€ 6S¢bcr veCrOLlL zwle6lot =6| Sze (isd LO
g0EL 0€ LS19¢b O90ESLL ee
60EL 0€ Gc06cr C86lPLll | przseot _| Sze 9102
OLEL oe LLZLev CLOLELL a a Tora
LZeL Le [eeeeer —s | cezeiit §=[e9cos §~=—ss | posgsol =[ Sze bO |
Zeel Le LES9er 6092011 S808 __ LOZS99L Szg eee
Zpel Le €LZ6Er 0€09601 €SZL8 L98PZ91 S79 ZLOZ
SEL Le visiby | 6errsor [zeorg | zecgot = |sz7e | sib |
p9El ze Qrrry —[ tvoezol | oelze _—s| 9e6zeol
v9El ze 60ESPr L8E6901 78Cc8 1909691
GOEL ze OLZ9PP 8zZS901 74s) SLL6691
GOEL ze eo 99072901 78Sz8 L9ZZOLL
QE Ze G86LbP €6E8SOL 0€Zz8 80ESOLL
Q9El [43 €L88rPr OPZrSOL 62828 LLEBOLL
Z9EL ze vS6bbr Z96ESOL GLLES OrZeLZL
LSEL ze SZOLbY €vLesOL Brees LSOSLZL
eSel Le oer SZEzSOL L8S¢eg LS8Z2ZL Gz8
Brel Le GLLEEY L6vLSOL ZLBES GLOLZZZL Sze
vel Le |sez6ep | ggsosor if QSZZELL Sze

ie. eR

puejai\ euuenes qnios jseq/iby “JOJ DUI }SOJ0} “1g umouyuf)
9 S v € 4 L 0

(L) 3sul4 sjaysew — ejewajend

£OL G xiIpuaddy

ZOEL 0€ goge6e | 8ZSZS11 ze66L ~—s« SEQLHOI _—'| Sze Sz0z
GOEL 0¢€ GZe66E OSLZSLL 78108 LZ8zS9L Sze vz0z
60E1 0¢ 6Z866E 9089rLL Sev08 866591 Sz €z0z
ZLEL Le L800 SLPLPLL 58908 OrLEQ9L Sz aaiyd
SLEL Le €ZOLOPF ESLOELL Z€608 syzegol | S78 L202
SLEel Le €9910P OS80ELL SZLL8 €ZECZOL zg 0z0z
ZLEL Le SSbbrOr S809ZL1 LOELe ZPBSL9L $28 610
pe Le O€ZLOP OZELZLL bzple | €zeeZor _[sze gL0z
€Lel Le Z6660¥ SZSOLLL | beSL8 €6Z0891 Gz Miwa
LLEL 0€ veep 6PSLLLL 09918 egzesor ‘| Sze 910z
OLEL 0€ BLPS LP ZZLLOLL SZL18 GZ9S891 Sz GL0z
GLEL Le ZE9L ZY 7798601 PlL6L8 S8rsg9l Szg rloz
LZel Le LvLlev 610601 8r0z8 €SZl69L Sz €L0z
9zel Le Lescer SLZL8OL z81z8 900691 Sz ZL0Z
Zeel Le 8L86Er LLeezol _| €reze ZLL9691 S78 LL0z
ZEel Le 006S¢r S96r7901 OPrz8 €ry669l Gz o10z
6EEl Le LOSSPP 28¢z901 986z98 6ZECOLL Seg 6002
Zvel Le COLSbr ZL00901 9Z2z8 ELZSOLL Gzg g00z
pl veel Le pO9Srr SESZSOl 598ze p6080Z1 Sz 2002
vl Orel Le SOSSbP Z90SS01 s00es ZLZ60LLL Sze 9002
vl Brel Le LOvStP 26SZS01 PrLes SrBELZL Sz g00z
vl Lvel Le 98lzrr Leezsol 9zeEes 98SZLZ1 S28 700z
GZ Orel Le 8S68er Zygisol —_| 90S¢8 96Z1ZL1 Gz €00z
GZ Orel Le 8ZZSer ZOVLSOL 989e8 200221 zg Z00Z
GZ Shel Le L6vecr 0901S01 v98es 069821 Sz L002
GZ veel Le S@c6cr 88S0S01 8e0r8 9SezeZLl Sz 000z
UVSA
sanoibuel PUBHa\\ BUUBAES imc seq/jiby "JOJ OUI 38910} 7g umouyun)
Z fe) g 4 € Zz L 0 ont

(Z) ysa14 Adijog — eyewazeny

vOL G xlpueddy

6L26Z O€LSZ | 89 ez 62 LLZEVE GO8LbZl GL99L 06708S1 Sze | _-szoz |
|6lc6Z _—_—si|s: EE 6H 69 z9ZL 62 LZ6p8E | zrosect | SvOLZ OLLEssl Gze | __—-¥e07 |
|6lc6L | ehlee —s: | 69 LZZL 0¢ vSS98E | QLerec, | €69S6S1 €z0
-eizes_} sere 08zl 0€ LLL8E LLbSIZL SLLLL zece09l 4) 720
61262 €LS7c 68z1 0€ O6Z68E €629021 COLOLOL Szg LZ07
2621 og GSCl6E CVL86LL 006982 ZOL8LOL S78 | _—-Oz0z |

2621 0€ vyer6e OSrLé6Ll vOLEZ 6LEzcOL Sze 6102

L621 0€ v8ZL6E SS8Pr8ll 20682 GOS9Z9L Sze gLoz

L621 oe 66L00¢ | zszgztt | 80162 LEQOE9L Szg L102

9621 oe 960E0r | O69LZLL | epZvegl Sze | __ 9b 0z |

9621 O€ 228807 | GLLSOLL y78se9l Sze GLOZ

ZOEL 0€ 906Zl 688911 LOZOP91 Sze vLO7c

iy c/a 60€1 O€ 66Z6ly «(| GL9shll [| sec6e _—s | LeSzroL_ =| Sze €L0%
6LZ6L gZZ91 SLEL L€ oggezp | zeecorlt =| rzz6e =f ezeppol =| Sze] ZO |
6LZ6L v7c9l LZeL Le L6beer LOLZELL 2867 | 961 9r91 See __{ fig
6L26Z €Z9SL LZEL Le L6cObP CCOveLl LS66Z ZLOSPOL Sze O10
61262 LOLS SEL Le 8Sc0bP GLZOLLL veeos €z6SS9L Szg 6002
6LZ6L g09rl ZZ 62EL Le €€ZOPy vSS8OLL 21-208 €LSE9OL Sz8 8002
61262 GOLPL cl LEel Le €6LObY GO6001L1 S60L8 96S1Z91 Sze | __- 2002 |
61262 699EL eZ ZEEl Le LOLObY Z6ce601 LLyLs 6SE6L91L G78 9002
6L26Z QZLEL eZ eeel Le 6ZLOPY 9ZZS801 9818 vZ0Z891 G28 G00z
[ 6L262 Brel el Geel Le Greler €Z98Z01 €62c8 0629691 Sze v00c
61262 eerel bl LEEL Le 8SLSEPr LZ9LZOL GeZ7g 66ESOL1 Sze €002
61262 OzLZL bl Ove Le €BSeer 2O09P901 vLLEs SSPrLZL G78 Z002¢
6L26Z vO8ll GL Ze Le ClyLey vLOZSOL OL9E8 OPPecZl L002
6L76L LEStL GL vel Le G¢c6cr 88S0SOl ger QS7@ceLl Gzg 0002
UWaA

JO}e/\\ ueqin senoijbueyy Pues BUUBAES | qnsaS Se d/joV “JO} SUld }Sol0j “1g UMOU UN)
6 8 Z 9 S 4 € Zz l ) on

(€) ysul4 Ayiunoas — eyewajens

SOL

G xIpueddy

61262 L0czez €Z 68EL [ze [o0sz0s _—| zssolol 98218 _—«|: wSBSB8OL Sze gz0z
6LZ6L 8E9lZ eZ gel ze 9e¢Z0S El90l0l ‘| ELere geegsol _| Sze y20@
6LZ6L 6zL bz eZ sel ze y9EZ0S 8S90l0l | seale ol6gs91 _| Sze Ses Cae
6L26Z zz90 eZ Ogel Ze 262206 €OZ0L0l _ | zogie Zcpzeol _| Sze 720%
6LZ6Z 19002 eZ El ze 622Z0S y9Z0L0l _| 6gsle z96Z891 | Sze Lz0@
6L26L SSS6L eZ SZEL ze LS1-L0S 6080101 | E1619 ezyssol _| Sze 0z0z
6LZ6Z 80161 fe a QZEL ze y8S10s = Z8SSLOL LZ618 1996891 _| Sze 6102
6LZ6L v0981 eZ ZLEL ze v66S6P €6€0z0l _| o€0ze €ggo69l _| Sze gL0z
6LZ6L €SL8h eZ gZEl ze LZe06¢ oLzszol _| 680z8 zgoz69l _—| Sze L10¢ |
6LZ6L pSZL1 eZ 6ZEL ze pL Ler Leooeol | ovize 6sze6ol _| Sze 9102
6LZ6L 86eZ1 €Z L8eL ze 6E06Zb L68reol _| vOzze B9rr69l _| Sz8 SLoz
6LZ6L Z8Z9L eZ = Zeel ze ZeSeSh zSp0eol _| 9zzze g06r691 _| Sze yL0z
6LZ6Z LZe91 eZ sel ze 90088 Zo09zol —_ | Sbzze glesegl | Sze e107
61262 08S} eZ Q8el ze LLve6y o6sizol _| Z9%z8 SSZS691 [sze 710%
61262 6vESL eZ Sse ze 62696 ZOLZ10l ‘| 98728 egl9691 __| Sze LLOz
6LZ6L Z68b1 eZ 06€1 ze OZELOS SSZzLoL__| 90¢z8 6959691 __| Sze 0L0z
6LZ6L Z8brl 7 EET Zé ELSS6r 9629101 | 60vz8 e698691 | Sze 6002
6LZ6L LZOvL vl L8El ze ce968h esgozol _| €1Sz9 gzsoozl _| Sze 8002
61262 7 S8el ze 6eLesy gz6rz0l | ZL9z8 Z9620L1__ | Sz8 2002
6LZ6Z 7 peel ze S6LLLb Z668z0l _| 61278 Zg0soZl _| Sze 9002
6L26L rl a ze Op8lLp zOLeeol _| vzez9 SyeZ0Ll _| Sze $00z
6LZ6Z 7 SZEL ze 69EEO bz99c0l _| 690€8 ggzzlZt | Sze 7002
6LZ6L bl ZOEL ze 898rSp 6zLOvol _| zLeee B0EZLZL | Sze €002
6L26Z SZ 6SEL Ze LSEQby ry9erOl _| 9SceR8 6eezzl) | Sze 2002
6LZ6Z GZ zS€l Le OL8Zer 8rLZpol _| 008E8 ISEZ7Zl _| Sze 100z
6L26Z GZ prel Le Sec6cr 88S0SOl 860r8 9SezeZl Sze =| ae
| UVaA
JOJe/\\ ueqir) sanoibuel| puepjea\\ BuueAeS qniaS seq/Jby JO} SUld }sosOj 71g UuMOUUL)
6 8 i) 9 S v € z L 0 ont

(y) ysa14 Ajiqeureysng — eyeweyend

901

G xipueddy

€0Szr 6yE00l O8zrl pOvELer 19z9SOl | s90ezt | zezez |
Leger LOLLOL €SPeelp 20Z8901 SC666Z1
68 Ler OZ6LOL 6LELSLY 6c96Z01 ySL6LSL
|s96re si ZESEP 6ZZZ01 | O8L0cZy | ESOLG6OL PEPSERL L8ZE7 720
|66lye | ELL GL8er |6sscol | eecri _—| Seo6gor O8Prz0ll 2892581 g8zez LZ0z
ere OSSZ 8leZry 66Er0L eer | 882S9r SO6ELLL vy69Z81 gecec B02
299¢2 Strrr 6L6P0L Lepr 8SrS7Z9Or LL6SZLL 6912681 88ZEC 6102
8S9br LEvSOl €L6c6Sr 688ZELL ZSELLBL 88ZE7 gL0z
6L8bP 8S6S01 GLv09SP 9686r11 GSSZE6L L82E7 L102
860SP ZL901 SL6LcSP SZ8LOLL SLLiG6l__| Lezez 9L0¢
6LESr 866901 99rS6rr L68EZLL 9LOSZ6L L8ZE7 | SLog |
L98SP L6z801 [erst | eZ 99h SEcrBLL OSPS66L L8ZE7 | vb |
9L79r 88S60L €6cS/ S6L9EbP vLOPELL ve6cl0c eet te
L969¥ 888011 pbrSl 6EELObY 6LOSOZL L9v0E07 g8ze7c ZLOC
O@SLb G6LZLL G6SSL GO6BLLEY v8rSLcL oorsroe | secez | tod]
| €208¢ OOSELL ObZSL LLL8rer gl6SccL €89S902 L8ZE7 0102
LELSP LSOELL €6LSL Lpssrer 0069721 8EEL902 88cec 6002
LOzer ZOBELL OvgsL OLSerer €sglecl 7668902 L8ZE7 8002
Gozey eseell | 288i BLLL ber L988zcl LS90202 88zez 2002
8ZE8Pr | ZOLPLL €C6SL OSLEcer 0zg6zczl LS@2L02 88zEz 9002
Z6ESP eSerLl O86S1 LOPOSer GO80Eezl 916202 LEZET G00z
GSL8P €69EL1 OS9ErEY €cggecl LvZ0202 887E7c 700
6L6LP QELELL 6r9SL LZ80SEr LSp97ecl 0869902 887e7 €002
v89Lr | L8SZLL v8rS 1 v908SEr 98zrecl €€6c907 88Ze7 2002
Srrly GzOcLL OzESL LZLS9EP LOLeccl 0SZ6S02 88Ze7c L002
6928 Eley 69rL LL yrOcler v686LCl LESGsoz 88zez 0002
EDN
JB}E/\\ sanoibuel\ pues eBUUBAeES “JOJ BUld jso10j “1g uMOouUyul)
6 Z | 9 g Z L 0 on

(L) S414 S}oxseW — SPaNnpUuOH

ZOL

G xIpueddy

pEll | S@zbr | pLerOL GLOCOSP OLESPLL €LOSE6L L8ZE7 Sz0z

LegZ | Zerby GlL6v0L O68rbSr 9rzSSI1 66S9¢61 L8ZEZ vz02

2181 | LSObb OzrSOl Ovz9cSr ZOLZOLL ZS78S61 88zez | _—-€zo |

ye6L | y98br €z6SO0L pOLEL Leylosy 7906911 6286961 g8zezc 7202

LL6Z ZL0S¢ L@v90l QI8EL LZ988rr vl6SZI1 22S 861 g8zez LZ0z

8108 762Sb vE690L g98el 02869¢%b 867811 vyce66l ggzez 0z0z

0S08 LveSb OSOZOL L66E1 09rlSbr EE9Z8LL LZLL00z L8ZE2 6102

7808 O6ESP 99LZ0L 8Zl rl Sc0Sbbb LvEZ6LL GLL6002 ggzez 8102

ae) Leys G8zZol gSzrl ZLOZEbY OOLZ6LL veLLZLOZ~ B8zez L102

918 O6¢Sb ZOvZOL 68Erl 98 L0crb vEsLOcl 0016202

818 6ESSP 8ISZOL OzSrL €LOLOry 8259071 e60EE0z

Glzg | L68S¢ 6PEROL GeZrl vrle76er LLOZLZL 6rez0e

ata) prycor L8L60L zSOSL LOSOLEY ZOVLIEL pevLSoe

6828 96S9r vLOOLL }etest «| Segoger _| 006zzezL 9650902

9zes LS69r ZS8OLL 98SS1 ZOLSPEP p8esecl 2€86902

v9E8 LOELY Z69LLL GS8SL OvP6cer G88Eecl 2016202 0102

v9E8 96ELb ZOBLEL GS8SL 66E6CEr €00vEzl soe6Z0z

g9es v8rlp OLLZLL GSeSL LS@6cEv O60vETL €Sv6202 28787 8002

g9E8 elSlp OzEzIL |ssest «| 6ocecer _—| BOzrEzL LS96202 g8zez 2002

g9¢8 LO9Lb 8ZSTZLL GS8SL L906zer G6cvEecl 862602 B8ZE7 9002

29€8 LSZLb 6ELZLL 9S8Sl 0z06zEr ElPbecl 2666202 S002
ev9ly G8v7cLl GLZSL LeZleer ELSLEZL OllLSZ0z 7002
QESLY (4xaqa GLSSL Levorer 0z98zz1 GEzOL0% £002
6erly OS6LIL | 9€rSt —_—s«|s OBOSSEr velSecl €Z€S902 2002

6828 LZeLp GZLZLLL 9672S LEQe9er raya LZv0902

6978 ELZLY 69rL LL QSISL vrOcler y6861Z1 LESSSGOz

JOJEA\ sanoibueyy puelja\\ BUUBAeES qnios Jsed/iDV JO} BUld }S9j0j “1g UMOUUP)
6 Z 9 S v € Z L ) oni

(Z) ysa14 Adtjod — SeanpuoH

80L g xipueddy

Leyes 68bSC LoeZ veSly 6€086 LOZeL LZceS9olp O8lZZ01 8SOSIL8L L8CEC Sc0c
vSp6ely c8v980l cELOEsl 88cEC yc0c

OSOOOL S8beLLy LZZS601 esegrsl 88cES €c0e
8SOLOL 96S289b 660S0L1 LOLc98L L8CEC

990201 ZSSL99% ZlopLtt | pezZz2981 _—_—| egzez
LL9L LS9e¢ GLOEOL vErseor gezeclLt | 60Sse6el | eezez
L8rss 8rS Lz L992 BOLEr QEEEOL pleel BSL 8L9b S9ZOELL ZLSPO6L g8ze7z 6102
L87eS 62602 GOZL BLBEr Q6SE0L 60SEL €02009% QLLOELL SLPSL6L 18ZE7% 8L0z
L8bss | LLvoz OSZZ B86er GS8E0l GP9EL pecessr O6ZErLL eSP9z6l ggzez L102
L8r9s | €€002 v6LL L60¢¢ ZLLPOL OSZeL gegsosr LEZ6rL O6EZE6L ggcez 9L0z
L8veS | pOv6L Beez LOtbr ZLEVOL QL6EL 6EL8PSb GOE9SLL p8esr6l ggzez GLoz
L8~8s | 62681 2982 L6Sbb 08zSOl g0zrlL BPSSeSr €6SO9LL OL9SS6L ggzez vL0z
L87gs L6E8l 9682 LL6¢b 761901 LOSPL 668¢cS¢ LO6Y9LL 898z961 Z87EZ e102
L988 pros | S262 v9ESP GOLZOL G6LbL Z800LSb 006911 LLLOZ6L 88ZEz coc
L8¢8s ZOSLL GS6L vSLSv 720801 SLEL6rry OPSELLL GepzZ6l | gecec LL0z
L8P9s LSLZL vs6Z prov 9r6801 yeersry yESZZLL pelys6l | zecec | —_—oLod |
L8veS OSS9L 6108 897Z9r 8eZ601 L890 br veOestl 7cves6l VEXA4 | __- 6002
L8rss prlol vS08 06E9r 925601 9689Sbb brlestt €£02002 g8zez 8002
L8veS vbSSl 8808 ELSOPr 218601 vOrSl 6rcerby eLee6lL 9290102 ggcez 2002
L8rgs OPLSL | 218 GE99Pr vOLOLL 68rS1 L@G6cbr EQES6ll bsc6loc =| egzezc —s| (900 |
Less erSrl 2G18 LSL9¢ v6EOLL vLSSL O€6SL Pr LLveocl 6sgzzoc | escee | so0c|
L8res 9Zl PL 0818 0S89r ZL9011 epySL SPLLObPy L0890zL 6OPEEN~ g8ze7c | ___- #002
L8veS LEZEL 208 cr69b OSL LLESL L6ER6Er ZLLOLZL 2S06€02 ggzez €002
L8r9s L62E1 g7ze8 vEOly QPOLLL 66cS1 pe96eer 6LPELZL Lz79rv0c gece Z002
L8r8s ZLOEL Lvz8 belly 6SCLLL | 8zzSt 66808Er SL99LZL €LLOSOZ
L8~ss OSOEL 6928 €Lelpy 69rL LL QSISL prOcler p6861cl lesssoz ‘| eezez 0002
UWA
Ja}e/\\ ueqin) senoibuey| PUBS BUuUeAeS qnias seq/JbW “1O} SUld }s9jOj “1g UMOU Uf)
6 8 Z 9 S v € 4 L 0 | om

(€) }Sal4 Aylundag — seanpuoyH

601 g xipueddy
[teres | 8610S [ ZUS8LL Gerel | v6leser | ZLezszL LOOLLLZ 88zEz Gz0z
L8res v0S8 6210S | 9SESLL 986eSer LL9vSZL ZELPL LT g8zez yz0z
L8rss 4900S | S6LSLL ell6ver L@S9SZL 6SCLLLZ g8zez €z0z
| £666 veoslt SSSSber G8Egscl o6eoziz | eszec | zzz!
| S@66¢ ZZ8LLL |vZpet «| €€ELeer ‘| Spzo9gzL v7SETLZ L8Z€Z LZ02
9S86r LEZZLL erat _] 9OLLEtr 2012921 Z999ZLZ ggzez 0202
2086 Z9SZI1 9rZ81 GrOerer 9970971 O9SEZIZ g8zez 6102
Lv L6v ESrZll 80081 7668bcer €zpesel €Sv07Lz ggze7z gL0z
L696 LZEZLL 6OLLL vysrSer SPS9Scl O6ZLLLZ g8zez L102
LE96r COLLIE L OESZ1 60809¢r 969rSZ1 pLlyll] ggcez 9102
78S6r Z9OLLL |o6cZ _—s«|s egzgger —| SPezSzl PSOLLLZ L87E7 GLOz
pSL6Py OLPLLL O6rLL v9sogzr 690SSZL ZOSPLLTZ ggzezc vL0z
| @zs8 8@66r O88ZIL 0691 BcerSer OO0EZSZ1 ZOSGSLLTZ peacee _}_sh 02
L87gs 9SO8L 2€S8 0010S Z8Z8b1 06821 996L ber L0S6SZL L8cezl~ L8ZE2 ZLOZ
L8rgs LOSZI zSS8 GZZ0S 669811 | 60st —_—s |: O69L eb eszt9ct | rgo9ziz | zecec LLOZ
Less BELLI 8958 60S 6OL6LL |e6cat _—«| BEcSEzp =| EzeEgzL _—| 9086zIZ 88zez OL0z
L8¢gs LvS9L 6ES8 ZSZ0S gg98lt pete __] B9z8rcr 606SZ1 079ZZ1Z 88ZEz 6002
L87eS OSLOL O1ss 7900S LOze8tL vOLLL O€LL9cy ZLvSScl 68ESLLZ B8zez 8002
L8v9s 6SSS1 L8r8 786 980bZer LOLLSZL LZLZOLZ 18ZE? 2002
L8Pgs €9lSl 7S¢8 | 0896p —Ss | v6 CZLL €2698cr 218921 6660012 88zez 9002
L8res els aa) 68r6r eregll GSg866cr 8z97r7l 6E8EG0T g8ce7z g00z
L898 6bL HL Z6E8 O£06r 6SLSLL | 20991 —_—s«|§: LOSPLEr SSO8Ezl ZEL9807
L8~9s ezZel L9E8 }ezsep «|| beoplt _—si[ zen vl06cer €OSEEZL €9¢8202 L82EZ €002
L8v9s BOEEL Ofes B09E11 GLSErer pl68ecl Z€80L02 g8ze7 2002
L87eS €80€1 00€8 ZESZlb 8682SEr gerrecl g81e90z g8zez L002
L8rvgs OSOEL 6928 69rL LL | 9SISL |: brOZZEr v6861cl LESGsOz 88zEz 0002
Tae ETE
Ja}e/\\ ueqin S9AOJDUP//\ Puepea\\ BuUueAeS qnsiaS jsed/iby JO} BUld yS9j0j 1g uMmouyus)

(p) 3sa14 Ayiqeureysng — seanpuoy

Appendix 6. CLUE-S Training Package (Exercises)

ICRAN-MAR Watershed Management Workshop

Training Course
“Land cover change modelling using the CLUE-S model”

Friday 18 August

Dr. Joep Luijten
Consultant to the UNEP World Conservation Monitoring Centre

UNEP WCMC

Fi SUNIT 7 TA’
‘=, USAID SQrctxpation

* PROM THE AMLARKLAN PiDPLE

Appendix 6

Land cover change modeling using the CLUE-S model

Friday 18 August
Training schedule (revised)

09:00 Introduction to land use change modelling and the CLUE-S model
Different types of land use change models
History and applications of CLUE-S in the world

CLUE-S model structure and key input files
Separate regression analysis of driving factors in SPSS

e

e

@

e
10:00 Introduction to case study area (Sibuyan island, Philippines)
10:15 Break

10:30 Practical CLUE-S

e System requirements and installation. Demo vs full version
e Exercise 1: Learning to know the user-interface and displaying results.
e Overview of input data files and model parameters files
e Exercise 2: Parameter files and simulating alternative scenarios
12:00 Lunch

13:00 Practical CLUE-S (continued)

Regression equation parameters files and probability surfaces
Land use conversion matrix and conversion sequences

Creating land use requirement (demand) files

Spatial policies and area restriction files

Conversion elasticities and crop rotations

Exercise 3: Creating new area restriction and land requirement files

14:30 Background on the MAR land use change scenario simulations, and
CLUE-S data sets for Belize, Guatemala, Mexico and Honduras

e Separate data and simulation per country

e Calculation of the land demand for different scenarios

e Dynamic and static driving factors; protected areas data
14:45 Break

15:00 MAR simulations, continued

e Regression equations and probability surfaces
e Exercise 4: Working with actual scenario data for Belize

16:30 End

Appendix 6 111

More information about CLUE-S model
http://www.cluemodel.nl/

Software used

For the training we will use the latest version of CLUE-S, also named Dyna-CLUE. This
version was released in February 2006 is a further development of v2.4

For visualization we use ArcGIS 9.1 with Spatial Analyst extension (ArcView 3 with the
Spatial Analyst extension can also be used in combination with CLUE-S).

Further reading

Below is a list of selected further reading related to the land use change modeling,
technical documentation of CLUE-S and its applications, and the development and
application of scenarios. All papers are included in PDF format on the data CD. The
technical report that describes the MAR scenarios and land use modeling in detail is:

Luijten, J., L. Miles and E. Cherrington, 2006. Land use change modeling for
scenarios for the MAR region. Technical report. ICRAN-MAR Project, UNEP-
WCNC.

Land use change modelling (in general

e Verburg, P.H., P.P. Schot, M.J. Dijst and A. Veldkamp, 2004. Land use change modelling:

current practice and research priorities. GeoJournal 61: 309-324.

e Parker, D.C., S.M. Manson, M.A. Jansen, M.J. Hoffman and P. Deadman, 2003. Multi-
agent systems for the simulation of land-use and land-cover change: A review. Annuals of

the Association of American Geographers 93(2): 413-337.

e Verburg, P.H. and A. Veldkamp, 2005. Introduction to the Special Issue on spatial

modelling to explore land use dynamics. Intl. J. of Geog. Info. Science 19(2) 99-102.

e Briassoulis, H, 2004. Analysis of Land Use Change: Theoretical and Modeling

Approaches. In: The Web Bcok of regional Science.

Land use change modelling in Central America

e Farrow, A., M. Winograd, 2001. Land use modelling at the regional scale: an input to rural

sustainability indicators for Central America. Agric. Ecosyst. & Environ. 85: 249-268.

e Kok, K., M. Winograd, 2002. Modelling land-use change for Central America, with special

reference to the impact of Hurricane Mitch. Ecological Modelling 149: 53-69.

e Kok, K., A. Veldkamp, 2001. Evaluating impact of spatial scales on land use pattern

analysis in Central America. Agric. Ecosyst. & Environ. 85: 205-221.

e Kok, K., 2004. The role of population in understanding Honduran land use patterns.

J. of Environmental Management 72: 73-89.

e Wassenaar, T., P. Gerber, M. Rosales, M. Ibrahim, P.H. Verburg, H. Steinfield, 1996.
Projecting land use changes in the Neotropics: the geography of pasture expansion into

forest. Global Environmental Change (in press)

Appendix 6 112

CLUE-S and its applications

Verburg, 2004. Manual for the CLUE-S model. Wageningen University, the Netherlands.
Verburg, P.H., W. Soepboer, R.L.V. Espaldon, 2002. Modeling the spatial dynamics of
regional land use: The CLUE-S model. Environmental Management 30(3): 391-405.
Verburg, P.H., C.J.E. Schulp, N. Witte, A. Veldkamp, 2005. Downscaling of land use
scenarios to assess the dynamics of European landscapes. In: Special issue: Future land
use in Europe: Scenario based studies on land use and environmental impact.

Verburg, P.H., K.P. Overmars, M.G.A. Huigen, W.T. de Groot, and A. Veldkamp, 2006.
Analysis of the effects of land use change on protected areas in the Phillippines. Applied
Geog. 26: 153-173.

Verburg, P.H., A. Veldkamp, 2004. Projecting land use transitions at forest fringes in the
Philippines at two spatial scales. Landscape Ecology 19 (1): 77-98 (2004).

Statistical analysis of land use change and explanatory factors

Koning, G.H.J., A. veldkamp, L.O. Fresco, 1998. Land use in Ecuador: a statistical
analysis at different aggregation levels. Agric. Ecosyst. & Environ. 70: 231-247.

Verburg, P.H., 2004. CLUE exercise - How to do the statistical analysis. September 2004.
Downloaded from http://www.cluemodel.nl/

Lessschen, J.P., P.H. Verburg and S.J. Stall, 2005. Statistical methods for analysing the
spatial dimension of changes in land use and farming systems. LUCC report series No. 7.
ILRI and Wageningen University, Nairobi and Wageningen.

SPSS 2004. SPSS Regression models 13.0.

Developing future scenarios and empowering stakeholders

Miles, L., 2006. GEO-4 scenarios and the ICRAN MAR project. UNEP World Conservation
Monitoring Centre. Project report. 14 August 2006.

Verburg, P.H., C.J.E. Schulp, N. Witte, A. Veldkamp, 2006. Downscaling of land use
change scenarios to assess the dynamics of European landscapes. Agriculture,
Ecosystems and Environment 114: 39-56.

Rounsevell, M.D.A. et al, 2006. A coherent set of future land use change scenarios for
Europe. Agriculture, Ecosystems and Environment 114: 57-68.

Potting, J. and J. Bakkes, J. (eds), 2004. The GEO-3 scenarios 2002-2032: Quantification
and analysis of environmental impacts. UNEP-DEWA/RS.03-4 and RIVM 402001022.

Selected data for Central America

Balk, D., M. Brickman, B. Anderson, F. Pozzi and G. Yetman, 2005. A global distribution of
future population: Estimates to 2015. (GPW v3). CIESEN, Columbia University.

Balk, D. and G. Yetman, 2004. The global distribution of population: Evaluating the gains
in resolution refinement. (GPW v3). CIESEN, Columbia University.

CIAT, 2005. Latin America and the Carribean (LAC) population database. International
Center for Tropical Agriculture, Colombia.

Vreugdenhill, D., J. Meerman, A. Meyrat, L.D. Gomez and D.J. Graham, 2002. Map of the
ecosystems of Central America. Final report. World Bank, Washington, D.C.

Meerman, J. and W. Sabido. 2001. Central American Ecosystems: Belize. Programme for
Belize, Belize City. 2 volumes 50 + 88 pp. http://biological-diversity.info/Ecosystems.htm
Batjes, N.H., 2005. SOTER-based soil parameter estimates for Latin America and the
Caribbean (version 1.0). ISRIC — World Soil Information, Wageningen, the Netherlands.

PDF files for all readings ca be found in ..\ Training\Documentation\

Appendix 6 113

What is included on the CLUE-S training CD?

Everyone who participates in the training on Friday will receive a data CD that includes
the following files organized in several folders:

Directory on CD Description
Software
. \CLUE-S\Dyna_CLUE_full\ Installation package of the full Dyna-CLUE model (latest

version of CLUE-S released in January 2006). It includes
the sample data for Sibuyan Island.

.. \CLUE-S\MAR_executable\ A specially compiled version of the Dyna-CLUE main
executable for use in the ICRAN-MAR project. It has been
optimized for memory usage and execution speed.

..\Other\ Installers for supporting software that we will used during
the training: TextPad 4.7, WinZip 9, and Adobe Reader.
CLUES_BZ Complete Dyna-CLUE model (program files and all data

files) for Belize, which will be used during the training.

Please note: If you copy this folder from CD, all files will be
read-only. You must right-click the folder, select Properties,
uncheck “read-only”, and apply it to all subfolders and files.

Documentation

.. \MAR_Modeling\ Full technical report of the land use modeling for the MAR
region, along with the CLUE-S user manual.

.. \Readings\ Scientific publications and documents related to some
source data that serve as (optional) further reading.

Scenario_results Land cover grid for the base year and for the three
simulated scenarios in 2025. There are also three grids
that show the areas of change in land cover.

Data

.\CLUE-S & Location factor grids, dynamic factor grids, and land use

.. \CLUE-S\MAR\ grids for the entire MAR (in the subfolder), as separate
files for Belize, Mexico, Honduras and Guatemala. These
files are the actual input files for CLUE-S. Files have been
compressed in *.zip files. Please note, metadata have
been included with the combined MAR data, not the
individual country datasets.

..\Clipmask\ Raster datasets of the precise spatial extent and
resolution for the MAR and the four countries, as they
were used to prepare (clip) all CLUE-S input data.

.. \Basedata\ Vector and raster datasets that were used for creating the
location factor grid for the MAR. These include original,
third party data dataset and derived datasets. All data
have been loaded in MAR Data Master.MXD

.\Avenue Scripts\ ArcView Avenue scripts for (i) creating dynamic factor
grids for protected areas, (ii) calculating the length of the
dry season, and (iii) for filling NoData gaps in grids.

Maps PDF files of two large format (AQ) maps.

Appendix 6 114

Background of the demonstration case study area, Sibuyan
Island, the Philippines

For the first two exercises the case study of Sibuyan Island is used. This is the same
case study area for which data are included with the demo version of CLUE-S. The
datasets are relatively small and simulations execute quickly, so it is ideal to start with.

Sibuyan Island is located in the Romblon Province in the Philippines. The island
measures 28 km east to west at its widest point and 24 km north to south, with a land
area of approximately 456 km? surrounded by deep water. Steep mountain slopes
covered with forest canopy characterize the island. The land surrounding the high
mountains slopes gently to the sea and is mainly used for agricultural, mining and
residential activities.

The island was selected as a case study because of its very rich biodiversity. About 700
vascular plant species live on Sibuyan Island including 54 endemic to the island and 180
endemic to the Philippine archipelago. Fauna diversity is low, but endemism is high. This
makes the island a ‘hot spot' for nature conservation and relevant for a detailed study of
land use change. For this application a spatial resolution of 250 x 250 meter is used.

The Phil lpplnes

+ =
N

Fe ae
fo Magawang

FS et eon L
*. . wh Mirdanae ¢
/f gte S :

‘

un

“ *

100 © 100200 Kilometers

Location of Sibuyan Island

Appendix 6 115

Five different land use types are distinguished for the simulation (see table below).
Important: for CLUE-S the land use numbering must start at 0, not 1.

Land use types on Sibuyan Island.

Land use code Land use type

0 Forest

1 Coconut plantations

2 Grassland

3 Rice fields

4 Others (mangrove/beach/villages/etc)

Four different files with land requirements (demand) scenarios have been created for the
period from 1997 to 2011. The land requirements are not very realistic for this short time
period, but allow us to clearly analyze the differences between the scenarios. Figure 2
summarizes the land requirements defined in the four scenarios:

1. Slow growth scenario, in this scenario a continuation of the land transformation rates
of the past ten years is assumed, meaning deforestation and an increase in the area
of coconut plantations, grassland and rice-area.

2. Fast growth scenario, in this scenario a higher rate of land transformation is
assumed, leading to rapid conversions of forest to coconut, grassland and rice fields.

3. Food-focus scenario, a high rate of land transformation is foreseen, however,
compared to the ‘fast growth scenario’ relatively more land is dedicated to rice
cultivation in order to supply food for the population of the island.

4. Export oriented scenario, the same high land conversion rate applies. However, it is
assumed that high copra prices make it profitable to dedicate most land to coconut
plantations and less land to food crops.

25000 = — acer mrovasiesineiseieaneee ees teme — eee eeeeroeretones
mw iao7
000 Dsicw
Ofes
25000 Diood
Oexpert
20000
15000 -
10000 +
of ruil
; ooh, om
coco mut grassland ne others

Demands for each land use type, for the base year (1997) and four scenarios. The
combined demand of all land use types is the same each year (45162.5 ha).

Appendix 6 116

Exercise 1: Learning the CLUE-S user-interface and displaying
results in ArcGIS

Objective: This exercise makes you familiar with the user-interface of CLUE-S and how
you can display the simulation results in ArcG/S/ArcMap. The precise definition of the
different parameters and input files is discussed Exercise 2 and in the user manual.

1.0 INSTALLING CLUE-S

CLUE-S (Dyna-CLUE) has been pre-installed on all computers in the training lab and the
data on the training CD have been copied to the folder C:\Training\.

If you are using your own laptop, or if want to install CLUE-S and the training data later in
your office, then you can install them as follows:

Open Windows Explorer and browse to the training CD.

Double-click Clues_Training.exe and extract all files to a location on the hard
disk. The default location is “C:” but you may specify another one.

a Double-click setup.exe from the Training\Software\CLUE-S\Dyna_CLUE_Full\
to install CLUE-S. Keep the default destination directory of “C:\CLUES”.

Q [MAR simulations only]. A ‘tailored’ main executable was compiled for use in
this project. Copy Training\Software\CLUE-S\MAR_executable\clues.exe to
the installation directory and overwrite the existing file.

1.1 START CLUE-S
CLUE-S can be started in two different ways:

1. Click Start | Programs | CLUE-S tools | CLUE-S
2. Open the directory where CLUE-S is installed with explorer and double-click ‘clues.exe’

The user-interface should appear on the screen (Figure 1-1).

The “Neighborhood Result” and “Neighborhood setting” buttons only appear after
checking the “Neighborhood variables” checkbox. These functions are not used in the
exercises. A description of the functions can be found in the CLUE-S manual.

Appendix 6 117

1.2 MAIN FUNCTIONS

The user interface makes it possible to edit the main input files through a built-in text
editor and allows the user to choose the scenario conditions. When all parameters are set
the simulation can start by clicking the ‘Run CLUE-S’ button. Simulation results will be
saved to output files that can be imported by a GIS for display and analysis (CLUE-S
does not have any built-in graphical capabilities).

te

© Modelling Framewo
ag ae
<a
ta

a

Figure 1-2. Explanation of the menu options of the CLUE-S interface
Most commonly used menu options are

Main parameters: Edit the main settings of the model (file: main. 1)
Regression results: Edit the regression equations (alloc1.reg)

Change matrix: Edit the land use conversion matrix (allow. txt)

Calculate probability maps: Create probability maps for every land use types
Neighborhood settings: Edit the neighborhood settings (neighmat. txt)
Neighborhood results: Edit the neighborhood results (alloc2. reg)

1.3 START THE SIMULATION

* Make sure that all input files are correctly defined (correct input files for Sibuyan Island
are supplied with both the demo version full version of CLUE-S)

« Select an ‘area restriction’ input file

The ‘area restriction’ file indicates which cells of a rectangular grid are part of the case-study area
and can also contain information on the locations that belong to an area with restrictions to land use
conversion, e.g. a natural park. You must always create and select an area restriction file, even if
there are no actual restrictions. In exercise 2 you will learn more about this file.

Select a ‘land requirements’ (demand) input file

The ‘land requirements’ file contains for each year that is simulated the required area of the different
land use types. These claims can be calculated in other models or can be based on trend
extrapolation and demographic projections. Different land requirements are possible for different
scenarios. The demand values must always be expressed in hectares.

° Click ‘Run CLUE-S’.

Appendix 6 118

The simulations will now start and the status bars show the progress (Fig. 1-3).

NOTE: The status-bar for the iterative procedure shows the average difference between the allocated area of
the different land use types and the required allocation of the different land use types. The simulation of one
year is finished if the allocated area deviates less than the specified maximum allowed deviation. Only when
one of the land use types exceeds the specified maximum deviation between allocation and requirements for
one of the land use types the iterations will continue and a special indicator will appear on the screen.

Progress if the deviation
between the allocated area |
and required allocation 2

Progress of iteration

Figure 1-3. Explanation of the CLUE-S model run and progress information

1.4 END OF THE SIMULATION

When all simulations are made successfully the model will display the message ‘finished’
and a button that gives access to the LOG-file will appear (Fig. 1-4). The log file contains
information on the input files and run-time information on the iterations and may be
consulted when errors occur or unexpected results are found.

fegper_cavtc ft

Figure 1-4. Interface after successfully finishing the simulation

Appendix 6 119

1.5 DISPLAY OF SIMULATION RESULTS & ASCII TO RASTER CONVERSION

All results of the simulation are saved in the installation directory. To display the
simulation results it is needed to use a GIS package. In this tutorial we will use
ArcGIS 9.x with the Spatial Analyst Extension.

CLUE-S saves the simulated land use data in ASCII GRID format. This can easily be
imported in both ArcView 3.x and ArcGIS 8.x/9.x. The simulation results are stored in files
called: cov_all.* where * indicates the year after the start of the simulation. Select the file
you want, e.g. cov_all.14 and click ‘OK’. One minor complication is that CLUE-S does not
save the files with an *.asc extension but with the year numbers as the extensions (.e.g,
cov_all.0, cov_all.1, cov_all.2, ..., cov_all.14, where year O is the start year and year
14 is the end year of the simulation). We have to manually add the extension .asc
otherwise the GIS import routine will not recognize the file:

> Follow the steps below to display a land use map generated by the CLUE-S model:

e Rename the simulation output file: Go to My Computer and browse to the CLUE-S
installation directory. Right-click the cov_all.* file that has the highest number and
add “.asc’”. For Sibuyan, you would rename cov_all.14 to cov_all.14.asc.

e Open ArcMap: Click Start | Programs | ArcGIS | ArcMap.

e Activate Spatial Analyst extension: Tools | Extensions | Check ‘Spatial Analyst’ |
and click OK.

e Open ArcToolBox (the red icon on the Standard toolbar) and import the simulated
land use grid: Conversion Tools | To Raster | ASCII to Raster. The menu shown in
Fig. 1-5 will now appear. Specify the following information:

a Input ASCII raster file: from the CLUE-S directory select a cov_all.*.asc file.
Set File of Types to “File (*.ASC)

Q Output raster: you may specify any name, but make sure you use a temporary
directory. It is important that you do not save the file in the CLUE-S directory
because if you do that many times the directory becomes cluttered with
temporary files and CLUE-S program files.

Output data type: keep the default setting INTEGER.
Click OK when all data have been entered.

ae
* ASCIE to Raster

inpwt ASCH rester file
[EMALUES Urarucere_ ail 14 ase

Ciusput raster

i w&

a\Tenginovt 4

i date optional)
fintecen >
OK | Cancel | Emvecreneets. | Show Helo >> |

Figure 1-5. Convert ASCII grid to raster using ArcToolBox

=i

The result of the simulation can now be seen and analysed using ArcMap (Fig. 1-6).

Appendix 6 120

It is now possible to change the graphical presentation by changing the colours of the
map into colors that are easily associated with the different land use type. For Sibuyan
Island the suggested colors is the table below can be used.

Table 1-1: Land use types and suggested colors for Sibuyan Island.

Land use code Land use type Color

0 Forest Dark green
1 Coconut plantations Orange

2 Grassland Light green
3 Rice fields Blue

4 Others (mangrove/beach/villages/etc) Red

* wetitied « Arcttap « ArcView

Coctegpatiry Teak
Convery Tacs

= Pircen Sueeher
= Te Bag
ze Te Toman Laramie

& To Raabe
PF ROC Ba
FP OEM tm Retr
# Peahse ha
oP Prat ts Rasta

# Suever Foe oe

‘a
a
Pel
dl
-
ae
le
y
o
A
&
y |

Figure 1-6. Simulation result displayed in ArcMap

> Repeat the above steps for the results for different years of the simulation (for
example, years 0, 5, 10 in addition to 14) with the Sibuyan data supplied with CLUE-S
and see how results change over time.

[End of exercise 1]

Appendix 6 121

Exercise 2: Parameter and input data files and simulating
alternative scenarios

Objective: In this exercise you will learn about the different parameter and region-specific
data files used by CLUE-S. You will run simulations using different scenario data files and
modify some parameters, and then compare the results in ArcMap.

2.1 PARAMETER AND OTHER DATA FILES USED BY CLUE-S

CLUE-S stores model parameters and region-specific data files in various files. The table
below gives an overview of all files that you may use. All files are plain text files that can
be edited using CLUE-S or a text editor such as Notepad or TextPad. All files are located
in the CLUE-S installation directory, C:\Clues.

P Please review the table below to get a general idea of the parameters being used.

Table 2-1. Input files used by CLUE-S. The “created” column indicates what software is
used to create the files and the “required” column indicates if the file is required. All files
created CLUE-S are plain text files and may also be edited in a text editor.

Filename Description Created | Required

Main.1 Main parameters file. Listed on exactly 19 lines. Some | CLUE-S_ | yes
parameters settings will dictate whether the optional files
must be specified or not.

Alloc1.reg Regression parameters. The length of file depends on | CLUE-S__| yes
number of land use types and location factors.

Neighbourhood’ _ results. Additional regression | CLUE-S_ | no
parameters based on the enrichment factor equation.

Change matrix. The number of rows and columns equal

the land cover types, here 10x10.

Neighmat.txt | Neighbourhood settings. Defines the shape and size | CLUE-S | no
(in the form of a small weight matrix) of the analysis
neighbourhood for every land use type.

Regi*.* Area restriction file. A grid that defines where land use | ArcView | yes
changes can and cannot occur. The * is a wildcard here;
it does not indicate the simulated year. All active cells

must have the value 0, restricted cells a value of -9998,
and all others cells -9999 (NoData). IL

Alloc2.reg

Allow.txt

Demand. in* Land use requirements. Calculated at the aggregate | Excel / yes
level and organized by rows (simulated years starting at Textpad
0) and columns (for every land use types). The *
denotes a unique number, not simulated year.

Cov_all.0 Initial land use. A grid of all land use types at the start | GIS yes

(year 0). Grid values must match the land use codes in
the main parameters file and numbering starts at 0.

Sc1gr# fil Static location factor grid, where # is the number of | GIS yes
the location factor;
Sc1gr#.* Dynamic location factor grid, where # is the number of | GIS no

a location factor. The * is the simulated year starting at
0, not a wildcard. Note that also the file src1gr#-fill is
needed and it is identical to src1ogr#.0.

Appendix 6 122

2.1 SCENARIO CONDITIONS

The CLUE-S model has a number of parameters that need to be specified before a
simulation can be made. The setting of these parameters is dependent on the
assumptions made for a particular scenario. In this exercise we will explore four different
scenario conditions, i.e., one or more of these settings will be different among scenarios.

1. Land requirements

2. Spatial policies (area restrictions)
3. Conversion elasticity

4. Land use conversion sequences

Different scenarios allow the comparison of different possible developments and give
insight in the functioning of the model. Such analysis is most easy by visual comparison
or through the calculation of the differences between the two scenarios in a GIS.

In this exercise you will first run the model with the baseline scenario: use the original
settings of the ‘main parameters’, select ‘region_nopark’ and ‘demand.in1’. |mport the
results (e.g. for the start and end of the simulation, year 0 and year 14). Next, run the
model again with four alternative settings as specified in the following sections (2.2 to
2.5). Compare the results in ArcView.

2012 Scenario 1

2012 Scenario 2

Figure 2-1. Simulation results for two different scenarios

Appendix 6 123

2.2 LAND REQUIREMENTS (DEMAND)

The land requirements are input to the model. For each year of the simulation these
requirements determine the total area of each land use type that needs to be allocated by
the model. The iterative procedure will ensure that the difference between allocated land
cover and the land requirements is minimized. Land requirements are calculated
independently from the CLUE-S model itself, which calculates the spatial allocation of
land use change only. The calculation of the land use requirements can be based on a
range of methods, depending on the case study and the scenario. The extrapolation of
trends of land use change of the recent past into the near future is a common technique
to calculate the land use requirements. When necessary, these trends can be corrected
for changes in population growth and/or diminishing land resources.

For policy analysis it is also possible to base the land use requirements on advanced
models of macro-economic changes, which can serve to provide scenario conditions that
relate policy targets to land use change requirements. For example, land demand for the
Mesoamerican Barrier Reef (MAR) region were calculated using the IMAGE model.

2.2.1 Simulating scenarios with different land requirements

Four different files with land requirements are provided with the model for the period from
1997 to 2011. The land requirements in these scenarios are not very realistic for this
short time period but allow us to clearly analyse the differences between the scenarios.
The scenarios are based on the following assumptions:

demand.in1: Slow growth scenario, in this scenario a continuation of the land
transformation rates of the past ten years is assumed, meaning
deforestation and an increase in the area of coconut plantations, grassland
and rice fields.

demand.in2: Fast growth scenario, in this scenario a higher rate of land transformation
is assumed, leading to rapid conversions of forest to coconut, grassland
and rice fields.

demand.in3: Food-focus scenario, a high rate of land transformation is foreseen,
however, compared to the ‘fast growth scenario’ relatively more land is
dedicated to rice cultivation in order to supply food for the population of the
island.

demand.in4: Export oriented scenario, the same high land conversion rate applies.
However, it is assumed that high copra prices make it profitable to
dedicate most land to coconut plantations and less land to food crops.

> Select one of the land requirement scenarios and run the model keeping all other
settings equal to the first run of the model. Analyze the results in ArcMap through
displaying the land use pattern at the start of the simulation and at the end of the
simulation. Repeat this for another scenario of land requirements and compare the
results.

NOTE: Each simulation, the model will overwrite the results of a previous simulation. If you want to save the
results, rename the output files or move the output files to another directory.

Appendix 6 124

2.3 SPATIAL POLICIES (AREA RESTRICTIONS)

This option indicates areas where land use changes are restricted through spatial (land
use) policies or tenure status. Maps that indicate the areas for which the spatial policy is
implemented must be supplied. Some spatial policies restrict all land use change in a
certain area, e.g., when in a forest reserve all logging is banned. Other land use policies
restrict a set of specific land use conversions, e.g., residential construction in designated
agricultural areas. In this exercise we will only address policies that restrict all land use
changes in designated areas.

With the DEMO version of the model we supply three area restriction files that can be
selected through the user-interface. Each file contains a map designating the areas
where land use change is restricted. The maps are shown in Figure 14 but can also be
imported in ArcView as ASCII Raster file similar to the procedure used to import the
results of the simulations. The files are located in the installation directory.

Area restriction files:

region_nopark.fil: | no spatial policies included

region_park1.fil: one large nature park following the boundaries of the Department of
Environment and Natural Resources of the Philippines

region_park2. fil: instead of one large nature park protection is proposed for small
areas which are assumed to face large land use change pressure.

Figure 2-2. Maps of restricted areas (in black)

> Run the CLUE-S model with the different area restriction files keeping all other settings
equal to the first run of the model. Compare the results with the initial situation (1997,
year 0) and compare the impact of the different area restrictions.

a Q: Is strict protection of the nature reserve needed for the developments until
2011 as simulated by the model?

Q Q: Do the protected areas in ‘park 2’ protect areas that would otherwise be
deforested? What is the consequence of strictly protecting these areas?

NOTE: Each simulation, the model will overwrite the results of a previous simulation. If you want to save the
results, rename the output files or move the output files to another directory.

Appendix 6 125

2.4 CONVERSION ELASTICITY

The conversion elasticity is one of the land use type specific settings that determine the
temporal dynamics of the simulation. The conversion elasticity is related to the
reversibility of land use changes. Land use types with high capital investment or
irreversible impact on the environment will not easily be converted in other uses as long
as there are land requirements for those land use types. Such land use types are
therefore more ‘static’ than other land use types. Examples of relatively static land use
types are residential areas, but also plantations with permanent crops (e.g., fruit trees).
Other land use types are more easily converted when the location becomes more
suitable for other land use types. Arable land often makes place for urban development
while expansion of agricultural land can occur at the same time at the forest frontier. An
extreme example is shifting cultivation: for this land use system the same location is
mostly not used for periods exceeding two seasons as a consequence of nutrient
depletion of the soil.

These differences in behavior towards conversion of the different land use types can be
approximated by the conversion costs. However, costs cannot represent all factors that
influence the decisions towards conversion such as nutrient depletion, esthetical value
etc. Therefore, in the model we have assigned each land use type a dimensionless factor
that represents the relative elasticity to conversion, ranging from 0 (easy conversion) to 1
(irreversible change). The user should specify this factor based on expert knowledge or
observed behaviour in the recent past. An extended explanation of the possible values of
the conversion elasticity and how behaviour changes when the land requirements
increase or decrease in time is given below.

0: Means that all changes for that land use type are allowed, independent
from the current land use of a location. This means that a certain land use
type can be removed at one place and allocated at another place at the
same time, e.g. shifting cultivation.

>0 and <1: Means that changes are allowed, however, the higher the value, the
higher the preference that will be given to locations that are already under
this land use type. This setting is relevant for land use types with high
conversion costs.

4Jé Means that grid cells with one land use type can never be added and
removed at the same time. This is relevant for land use types that are
difficult to convert, e.g., urban settlements and primary forests. A value of
one stabilizes the system and prevents that in case of deforestation other
areas are reforested at the same time.

The conversion elasticities of all land use types are specified in the ‘Main Parameters’
input file (main.1, line 11) that can be edited through the user interface (click the ‘Main
Parameters’ button). An explanation of all other parameters in this file can be found in the
user manual). The first conversion elasticity corresponds with land use type 0, the second
with land use type 1, etc.

Table 2-3. Current settings of the conversion elasticities

Land use code Land use type Conversion elasticity
0 Forest 1.0

1 Coconut plantations 0.8

2 Grassland 0.2

3 Rice fields 0.2

4 Others 1

Appendix 6 126

> Run the baseline scenario for Sibuyan island with the CLUE-S model with the current
settings and with alternative settings for the conversion elasticity. Change the conversion
elasticity by:

Click on the ‘Main Parameters’ button. The main parameters can now be edited.

Line 11 contains the conversion elasticity settings of the different land use types in
the same order as the land use type coding. Change these values to new values.

Click on ‘Save’.

Run the model after selecting the ‘Area restrictions file’ and the ‘Land
requirements’ file (similar to the first run of the model).

Display the results with ArcView.

Compare the differences in spatial pattern of land use change as result of the
changes in conversion elasticity.

® Dyne-C be

Figure 2-3. Conversion elasticities are listed on line 11 in the parameter file main.1 5

NOTE: Each simulation, the model will overwrite the results of a previous simulation. If you want to save the
results, rename the output files or move the output files to another directory.

Appendix 6 127

2.5 LAND USE CONVERSION SEQUENCES

Not all land use changes are possible and some land use changes are very unlikely (€.g.,
arable land cannot be converted into primary rain forest). Many land use conversions
follow a certain sequence or cycle, e.g. fallow land and forest regrowth often follow
shifting cultivation. Figure 2-4 indicates a number of possible land use trajectories
identified on Sibuyan island.

oramary fores!

detorestatias foggy repre

pnmary forest aoe SECONGATY forest

IN

shifting cultrvation

Figure 2-4. Possible land use trajectories on Sibuyan island.

The conversions that are possible and impossible are specified in a land use conversion
matrix. For each land use type it is indicated in what other land use types it can be
converted during the next time step. Figure 2-5 provides a simplified example of a land
use transition sequence. Forest can be converted in either agricultural land or grassland,
while it is impossible to obtain new (primary) forest through the conversion of agricultural
land or grassland directly. The figure also illustrates the translation of these conversion
sequences into a land use conversion matrix, which can be used by the model.
Depending on the definition of this conversion matrix and the time-steps chosen, complex
land use sequences are possible.

Land use charge sequerice Land ise conversion matrix

future oe wee
meng une 4g 3 =
na en 5 | z
praaert b 6&
barat sce Ww *. si

forest “—— > agriculture ee » grassland Forest + | +) tay

° + ~ ~ + ° +

met Agrcumure ~ Feet

Grasdand = +ig) +

* comarman pera ;

> CervErinn not parE
Figure 2-5. Land use transition sequence

The land use conversion matrix can be edited by clicking the ‘Change matrix’ button. It is
also possible to use a text editor (e.g. Notepad) to edit the file ‘allow.txt’ in the
installation directory. The rows of this matrix indicate the land use types during time step f
and the columns indicate the land use types in time step t+7. If the value of a cell is 1 the
conversion is allowed while a 0 indicates that the conversion is not possible. The rows
and columns follow the number code of the land use types.

Appendix 6 128

Example: in the matrix below all conversion are possible except the conversion from
coconut plantation into rice fields.

From / To-> Forest Coconut Grassland _ Ricefields Others
Forest 1 ll 0 1
Coconut 1 ls 1 1 1
Grassland 1 1 1 1 1
Rice fields 1 1 1 1 1
Others 1 1 1 1 1

> Run the baseline scenario for Sibuyan island with a different setting of the conversion
matrix (keeping all other settings equal) and analyse the differences in outcome with
ArcView. We suggest to compare a model run that allows all changes with a model run in
which the conversion of grassland into agricultural land (coconut plantation and rice
fields) is no longer possible due to soil degradation. Compare the results.

Note: Some land use conversion settings will have no effect because they are overruled by the conversion
elasticity and land requirement settings. In the baseline scenario we have assumed that the ‘others’ land use
type is not changing and forest cannot ‘re-grow’ from other land use types as long as its total land area is
decreasing. Consequently, changing the conversion settings for these land use types in the conversion matrix
will have no effect on the simulation results.

[End of exercise 2]

Appendix 6 129

Exercise 3: Defining spatial policies and creating new land
requirements

Objective: In this exercise you will learn how to prepare a new land requirements file and
also a new area restrictions file. The combination of these new files represents a new
scenario and you will then simulate your own scenario.

3.1 INTRODUCTION

For some scenarios it is interesting to define areas where land use changes are restricted
because of spatial policies, e.g. the conservation of nature. In the previous we have seen
that spatial policies should be defined in an ‘area restriction’ file. This file contains a map
of the study area indicating the extent of the case-study area and the zones of the case-
study area where spatial restrictions apply.

The ‘area restriction’ file is located in the installation directory and called ‘region’.fil’
where * can be defined by the user to indicate the conditions specified in the file. With the
demo version of CLUE-S three different area restriction files are supplied, one without
any spatial policy and two file indicating different extents of a nature reserve.

> Import these ‘area restriction’ files in ArcMap using the procedure as you used for the
land cover grid in Exercise 1.5. It is best to copy these files to a temporary directory and
then rename them there by adding “.asc’” to the file.

Q Question: What are the different grid values in the area restriction files? What
value is used for a restricted area? And what value for a non-restricted area?

3.2 PREPARATION OF A NEW AREA RESTRICTION FILE

In this exercise you will create a new ‘area restriction’ file to simulate a scenario of the
effects of a strict protection of all remaining lowland forest on Sibuyan island. Therefore
we assume that during the simulations it is not possible to convert any of the remaining
forest areas below an altitude of 100 meter.

To make the area restriction file we need to identify:

e The extent of the case study

e The locations below 100 meter altitude

e The locations with forest at the start of the simulations

Therefore it is needed to import the land use map of year O (the start of the simulation) in
ArcView. This land use map shows the extent of the study area (all grid-cells that are
designated to a land use type) and the locations with forest at the start of the simulations.
This land use map can be found in the installation directory (C:\Clues) and is called
‘cov_all.0’.

To identify the locations below 100 meter an altitude map is needed. Since altitude is one
of the location factors used in the simulations for Sibuyan island this map is already
present in the installation directory. For this case study altitude is location factor number
7, so the elevation dataset in file ‘sc1gr7.fil’.

> Import both files using the ASC// to Raster option in ArcToolBox.

Appendix 6 130

goon ge Ce
Fee Got ew peers Sexton Ino fiecee page

Spare asi © ape Parone

ca ©

[latvhuiod| #9 0 & Leeade] < 169

Figure 3-1. Map query with the Raster Calculator in ArcMap
In the ‘area restriction’ file the following coding should be used:

0 all grid cells that belong to the study area outside the ‘restricted
area’. These are the grid cells that are allowed to change.

-9998 all grid cells for which land use conversions are not allowed during
the simulation (the ‘restricted area’)

-9999 (No Data) all other grid cells (outside the simulation area)

> Prepare an ‘area restriction’ file to prevent any forest areas below an altitude of 100
meter from changing. You can follow the steps below or use your own procedure:

Q Select all locations with forest located below an altitude of 100 meter at the start of
the simulation by a ‘map query’ (Spatial Analyst | Raster Calculator) (Fig. 3-1).
This will result in a new temporary theme ‘Calculation’ indicating all selected
locations by a value of 1.

a Classify the results of the previous step to the coding system of the area
restriction file, as listed above (Spatial Analyst | Reclassify) (Fig. 3-2). This should
create new temporary layer ‘Reclass of Calculation’ with values of —9998 and 0.

Q Export the result of the previous step as an ASCII file ‘region5.asc’ in the
CLUE-S installation folder (ArcToolBox | Conversion Tools | From Raster | Raster
to ASCII). Note that you must specify either a .txt or .asc extension

a Using Windows Explorer browse to the installation directory and rename the file
region5.asc to region5.fil (you may use a different number but you must use the
region*.fil naming convention otherwise CLUE-S does not recognize it).

Appendix 6 131

: ate mes 3x4
Inge raster |Cabidste =| cs |

Fewbess haeled [Nake

Set vaiues to rnchasmty

Load | Save |

I Charge eeseng walues by HoDole

pe
Qutpes nester i Tempeniey>

Figure 3-2. Reclassifying a grid as an area restriction file.

> Restart the CLUE-S model and the new area restriction file should appear in the list of
area restriction files and can be selected for the simulation. Run the model with this file
and compare the result with a simulation without protection of forest resources.

> Prepare your own area restriction file based on a hypothetical spatial policy. You can
also prepare area restriction files by delineating areas in ArcView that need to be
converted to grid cells.

Note: If the area restrictions violate the land requirements specified in the ‘land requirements’ file the model
will not succeed in allocating land use changes and stop the simulation. This can occur when all forest is
assumed to be protected while at the same time a decrease in land requirements for forest is specified.

3.3 CREATING YOUR OWN LAND REQUIREMENTS FILE

> You will now start defining your own scenario by generating a new land requirements
input file for CLUE-S. Follow the steps and data guidelines below:

Open Microsoft Excel to facilitate the calculations.

Specify for each year (1997-2011) the land requirements of the different land
use types in a table following the specifications below:
Please note: Demand must always be expressed in hectares (10 000 m7’).

a Each row indicates a year; each column a land use type following the order of
the land use coding.

a Make sure to include also the land requirements for 1997 (year 0). These
should be similar to the land use map of 1997 (29518.75, 7237.5, 5243.75,
1400, 1762.5 ha for respectively forest, coconut, grassland, rice and others).

a The total land area required shouid equal the size of the island (45162.5 ha),
i.e., the sum of the values on each row should equal 45162.5 for each year.

Suggestion: you can temporarily add an extra column G or H and use a formula to
verify that the row totals are always equal (for example: G2 = SUM(A2:F2)).

Appendix 6 132

a We suggest not to change the land use requirements for the ‘others’ land use
class and to create logical scenarios without sharp increases or decreases.
This should prevent problems or very long run times during the simulation.

A) Ge je Ge pat Pyne Tee fee gee ce koe aifizi

Dees @AT PRM o- Oren Mots F

ra ei

not

1B: ba

Ae MA sheet tf Teeth See EE bseroe 1 at
niall ees epee nem

Figure 3-3. Entering land use requirements in a spreadsheet.

a When all values have been defined, select the values (without land use type
names and year numbers) and paste the contents into a text editor (e.g.
Notepad). Insert a line at the top of the file with the number of lines (years) for
which the land requirements are specified (15 in our example).

Bretied Meiriad

29508 PS $22.98 SP4a. FS Teee. 88 1/824
2OFhT OF PEW .11 SB56 8 VTE EFS
SU eth ay Pee. 71 Sake OF tae 1s 1S
FOTbe Po TRIQ IF SSPE. FB eRe TT AEF.
26509 29 7683.73 SASS 39 1518.27 1742.5
2EPG? . ie TIS SIFFS VERE TES
2aete. Te POO? 1 SPILT VSTT. SE 17625
a7 ite ae SOW. 7S Sart. ee 1200.00 182.8
21568, Bs 6130.28 £933.88 TH28.57 1742-5
FITE6. 7S EFN1 TS 478A 7H 147 1S ETS
21 ee4 28 B253 47 6955.26 1685.71 1762.5
265s, BeOS. 18 GAG ST 1716.79 147.5
POSE? 68 PSS. Pe GST? OB TRE. BA 174275
2675138 S692 37 6458.04 1771, 63 1762.5
2h oe. oe $600.06 6906.08 1820.08 1762.5

Eros pea aie nee ate Sher ee
Figure 3-3. Land use requirements copied to a text file demand.in*

a Save this file in the installation directory as ‘demand.in* where * can be
defined by the users, e.g. demand.in5.

Q Restart the CLUE-S model; it is now possible to select the new land
requirement file and simulate the land use changes.
Q Import and analyse the results in ArcMap.

[End of exercise 3]

Appendix 6 133

Background on the MAR Land Use Change Simulations

In the previous exercises you worked with data for Sibuyan Island. This is a very small
dataset and simulations ran very quickly, which made it very suitable for a relatively short
training day and allowed you to quickly inspect the changes in simulation outcomes after
you made adjustments in area restriction files, land demand and conversion elasticities.
Now you will start working with some actual data for that we used for the MAR.

The MAR catchment is approximately 190,400 km? large, with 41% of the area in
Honduras, 29% in Mexico, 18% in Guatemala and 12% in Belize. Those 12% represent
the entire country of Belize whereas only parts of the other countries are included.

Dataset prepared for each country

First of all, it is important to know for that the regression analysis and the CLUE-S
simulations were done separately for each country (or part of it). The reasons are:
Q Land use pattern and the drivers of land use change are different for the countries
because of different policies, biophysical conditions or other factors, so performing
a separate analysis allows a more accurate analysis.
Q Smaller data files by country facilitate easier data management.

Nevertheless, these size of the data even for an individual country is much larger that for
Sibuyan Island. For all four countries the smallest possible spatial extent was defined and
the country grid were clipped to these extents.

Size of the grid for Sibuyal Island and the MAR countries. Cell size is 250 m. Simulation
times are observed on a laptop with a 2GHz Pent. M processor and 2GB of RAM

Country # Rows # Columns # Data cells Average time for
in grid in grid (not Null) a simulation run

|
| Sibuyal 108 7,226 < 10 seconds
Belize 1151 604 | 349,762 % - 1 hour
Guatemala 1503 1310 | 542,309 2 - 3 hours
Mexico 1674 1262 886,433 3 - 4 hours
Honduras 1,267,903 4 - 5 hours
All of MAR 3,046,407 N/A

West (xmin) East (xmax)

Belize 261,500 412,500
Mexico 213,250 528,750
Guatemala 41,250 368,750
Honduras 260,250 793,000

South (ymin) North (ymax)

Belize 1,757,500 2,045,250
Mexico 1,971,250 2,389,750
Guatemala 1,596,500 1,972,250
Honduras 1,521,000 1,772,250

Figure 1: Spatial extents and mask for raster datasets for the four MAR counties. Coordinates
are in UTM zone 16 with NAD 1927 Central American datum.

Appendix 6 134

GEO-4 Scenarios

We adapted three of the four Global Environment Outlook 4 (GEO-4) scenarios Latin
America and the Caribbean for use within the ICRAN MAR project. The scenarios
envisage differing social, political and economic trajectories, emphasizing outcomes for
the environment and human well-being.

1. Markets First: Under this scenario economic growth is prioritized over social and
environmental objectives. Everything becomes merchandise, including natural
resources and basic goods such as water and culture. In general, regional
environmental degradation continues to worsen.

Zs Policy First: Environmental awareness develops within government more rapidly
than in the private sector or amongst the general public. The resource base is better
managed, with policies being developed to alleviate the more serious environmental
problems.

3. Sustainability First: In this world, economic, social and environmental
dimensions combine to shift the trajectory towards environmentally sustainable
development. International cooperation within the region increases, with policies
being directed to achievement of the Millennium Development Goals and sound
natural resource management.

More details about the scenarios can be found in Miles (2006), which is included in the
readings list and as a PDF file on the CD.

Land requirements

Land requirements for every scenario were calculated using the IMAGE model. In the

next exercise we will focus on Belize. The table below gives the distribution of land use at

present and the calculated land demand under the scenarios in 2025 for Belize. The total

area is 21860.13 km’. The area of land use types 0 and 9 is assumed to remain constant.
Land use distribution at present and for the scenarios in Belize.

Present | Markets First Policy First | Sustainability

(2004) 2025 2025 First 2025

. Mangroves 3.29% 3.08% 3.22% 3.28%
. Urban 0.87% 1.79% 1.64% 1.51%
Water 0.70% 0.70% 0.70% 0.70%

0. Other/Unknown 0.06% 0.06% 0.06% 0.06%
1. Broad-leaved forest 58.02% 54.33% 56.69% 57.78%
2. Pine forest 3.53% 3.52%
3. Agriculture/pasture 19.37% 23.20% 20.64% 18.85%
4. Scrub 1.26% 1.23% 1.13% 1.37%
5. Savanna 8.63% 8.23% 8.35% 8.65%
6. Wetland/Swamp 4.26%
7

8

9

Appendix 6 135

Exercise 4: Working with actual scenario data for Belize

Objective: In this exercise you will learn some of the actual data that were used for the
Belize simulations, and how dynamic location factors were used. You will also create and
review probability surfaces. At the end of the day you should have a sufficient knowledge
of the data to simulate and analyze the actual different scenarios yourself.

4.0 COPY THE CLUE-S MODEL AND DATA FOR BELIZE FROM CD

a Using Windows Explorer, browse to the folder C:\7raining\Data\CLUE-S\ on your
computer (or the CD). This folder has 4 zip files that contain all land use and location
factor grids in regular grid and ASCII format.

a Double-click BZ.ZIP and unzip (extract) the file in a temporary location on your
computer, e.g., c:\temp. Remember where you extracted the file.

a Also copy the entire CLUE_BZ folder from tne CD to a place on the harddisk. Then
right-click the folder, select Properties,uncheck “read-only”, and apply it to all
subfolders and files. CLUE-S will give an error if the folder is read-only!

4.1 REVIEW OF THE LAND USE DATA

The baseline land cover map was based on the 2004 version of the Belize Ecosystem
Map and the revised 2003 Ecosystem Map for Central America land use data. The
original land cover classification was reduced to 10 classes (Table 4-1) and the data was
converted from a vector to a raster format with a 250 m grid cell size.

Note: CLUE-S requires that the land use numbering to start at 0, not 1.

Table 4-1: Reduced land use classification used for the MAR

Value Land use type Value Land use type
0 Other/unknown 5 Savanna

1 Broad-leaved forest 6 Wetland/Swap
2 Pine forest Uf Mangroves

3 Agriculture/pasture 8 Urban

4 Scrub 9 Water

> Let’s now look at the reclassified 2004 land use data for Belize:

a Open ArcMap and load the layer file ‘Belize Present land Cover (2004).lyr that
is in the BZ data folder (the layer file source grid is ..\BZ\grid\bzecomap).

a Review the land cover data. Keep in mind that these data were based on the
2004 Ecosystem Map for Belize (Meerman & Sabido. 2001), reclassified and
converted from vector to raster data with a cell resolution of 250 m.

4.2 STATIC AND DYNAMIC LOCATION FACTORS

For every region you must specify a number of driving factors of land use change. These
‘location factors’ were determined by a statistical regression analysis. Table 4-2 lists all
location factors that were analyzed for the MAR. The numbering must starts at 0.

A location factor can be static of dynamic, as is indicated in the last column of the table.

a Static: the location factor is constant over the entire simulation grid. The grid is
saved in ASCII format with the following naming convention: SRxGR.FIL

Appendix 6 136

Q Dynamic: the location factor changed over time. Instead of a single ASCII grid we

have to prepared a grid for every year: SRxGRD.y

where
= the number of the location factor (0 to 11)
y = the simulated year started at 0 (for belize, 0 to 21)

For the MAR land use simulations only one location factor was dynamic: No. 11,
protected areas with partial protection. Of course, population density will also change
over time, and accessibility to markets and road may also change if no roads are built.
However, the scenario descriptions that were developed described the future changes for
the country as a whole and did not provide sufficient details about exactly what, where
and how changes might occur on a regional or local scale.

Table 4-2: Location factors (LF) used for the MAR land use simulations.

No __ Description [unit] CLUE-S Original Dynamic
file GRID file
0) Population density [# per km7] SC1GRO.FIL POPDEN No
1 Soil depth [meter] SC1GR1.FIL SDEPTH No
2 Soil drainage [0-1] SC1GR2.FIL SDRAIN No
3 Mean annual rainfall [mm] SC1GR3.FIL RAINYR No
4 Length dry period [consecutive months SC1GR4.FIL DRYMON No
with < 60 mm rain]
5 Elevation [meter] SC1GR5.FIL ELEVAT No
6 Slope [degrees] SC1GR6.FIL SLPDGS No
7 Accessibility to markets SC1GR7.FIL ACSMKT No
[travel time in hours]
8 Accessibility to roads SC1GR8.FIL ACSRDS No
[travel time in hours]
9 Coastal area / tourism hotspots [0/1] SC1GR9.FIL TOURIS No
10 Protected areas / full protection [0/1] SC1GR10.FIL WDPAR1 No
11 Protected area / partial protection [0/1] ‘SC1G6R11.FIL) WDPAR2 Yes
SC11GR.0
SC11GRD.21

> You will now review the location factor grids.

Appendix 6

Q Browse to the ‘BZ’ folder that you just extracted. You should see two sub-

folders, “asci’’ and “grid’. Both folders contain the same dataset but in
different formats. It is easier to work with the data in the “grid” folder because
these data can be readily loaded as a layer in ArcMap.

Open a new ArcMap document and load location factor 8, “Accessibility to
Markets” (bzACSMKT). The unit of this data is travel time in hours. It was
calculated using a methodology developed by researchers at CIAT, Colombia.
Do you think the travel times are fairly realistic?

Also load location factor grid 10 (6zWDPAR1) and 11 (bzWDPAR2). A grid
value of 1 means that the grid cell is a protected area, a value of 0 not. Are
you familiar with the protected areas in the country?

Load all other grids location factor (i.e., don’t load the “bzLUC_” grids — these
are grids for individual land use types). Review all location factor grids and
make sure that you understand these grids and their units.

137

4.3 REGRESSION EQUATIONS AND PROBABILITY SURFACES

Note: in this part of the exercise you will be learning about some of the more advanced
features of CLUE-S. Nonetheless, you will need to understand these features if you are
planning to use CLUE-S for actual simulation modeling of land use change.

The allocation of land across the region is done by CLUE-S based on probabilities, which,
in turn, are calculated using regression equations that account for the effect of one or
more location factors. For example, the regression analysis showed a significant
relationship with Urban land and three location factors, as follows:

Probability LUs = 0.5 + 0.01 LFo — 0.37 LF7 + 0.70 LFy

where LU, = Land use type 8, Urban
LF, = Population density
LF; = Accessibility to markets
LF, = Coastal area / tourism hotspot

Note the negative relationship for LF7. Thus, the farther away a grid cell is from a
market, the lower the probability that land use at that location changes to Urban.

> The regression equation above and similar equation for other land use types are
specified in the file alloc1.reg. You will now briefly review that file.

Q Browse to the folder CLUE_BZ that contains all files for Belize.

Q_ Right-click on the file alloc.1reg and open the file in a text editor such as
Notepad or Textpad. You can select NotePad by choosing “Open With” and
the selecting Notepad from the list of available programs.

Q Scroll to the end of the file. Do you recognize the parameters from the
equation above? Please refer to pages 22-23 of the CLUE-S user manual for
more information about the precise format of this file.

The regression equations are important because during run-time CLUE-S uses these
equations to create probability surface for every land use type. These surfaces show the
probability of changes towards that land use type across the entire study area.

> Let's look at some of these probability surfaces.

a Using Windows Explorer, browse to the ..\Training/CLUES_BZ\ directory and
double-click clues.exe to start the model.

Select “Calculate Probability Maps” from the Mode main menu.
Select one of the area restriction files and one of the demand files.

Press the “Run CLUE-S” button to start the model. The model will now only
calculate the probability maps. This will take about 1-2 minutes. Press the
“Calculations Finished” button but do not close the application.

a Go back to Windows Explorer in the CLUE_BZ folder and refresh the contents
of the folder view. You should now see the probability surfaces that were
created with the names prob1_0.1, prob1_1.1, ..., and prob1_9.1.

Rename the extension of all ten probability files from “.1” to “.asc”.

Open ArcMap and ArcToolBox. Use the “ASCII to Raster” tool to import the
files prob1_6.asc (probability for Wetland) and prob1_8.asc (probability
surface for Urban). Make sure to select “FLOAT” as the Output data type.

Appendix 6 138

a Review the output. It should look like the next figure (colors may be different).
The highest probability for Urban should be near Belize City, near some other
coastal areas, and close to the main highways. The probability for Wetland
should be highest in the northeastern part of the country.

PE High : 0.380

i Low : 0.02)

Probability surface for Urban Probability surface for Wetland

Note: No probabilities are calculated for restricted areas (“no change” areas). These are NoData
cells and are white coloured in the maps below. The probability surfaces are very useful for
verification of the validity (significance) of regression equations and the restricted areas: Areas
with the higher probabilities should correspond to where the land use type presently is.

4.3 Running a complete simulation run

> A full simulation run for Belize may take % to 1 hour, so it is unlikely that there is
enough time during the training day to do this. However, you can try it, of course.
In CLUE-S, unselect “Calculate Probability Maps” from the Mode main menu.

Select an area restriction files and a land demand files. Note that you must
combine files that have the same number (1 = Markets First; 2 = Policy First;
4 = Sustainability First). Then click the “Run CLUE-S button.

a When the simulation has completed, rename the file cov_all.21 to
cov_all_21.asc and import this file in ArcMap. Make sure to select
“INTEGER” as the Output data type.Review the land use pattern.

[End of exercise 4]

Appendix 6 139