Natural capital indicators for OECD countries FINAL REPORT Natural capital indicators for OECD countries Prepared by WCMC on behalf of RIVM 1. INTRODUCTION AND METHODS The particular purpose of this study is to investigate whether certain elements of a Natural Capital Index framework (NCI), as proposed by the liaison group established by the Subsidiary Body on Scientific, Technical and Technological Advice (SBSTTA) to the Convention on Biological Diversity (CBD) (see UNEP (1997a,b), can be implemented in practice with reference to the OECD countries. The study has been carried out in support of the regional programme for environmental assessment within the OECD, and is a contribution to the development of practical assessment tools designed to indicate the changing status of biological diversity. The study has addressed change in the amount and quality of natural ecosystems. “Natural ecosystems’ are broadly defined as self-regenerating systems which are natural or semi-natural ecosystems, irrespective of ecological quality, larger than 100 ha, such as nature areas, extensively used areas (such as areas with nomadic livestock), all forest (except for planted forest with exotic species), rangelands of native pastures, all freshwater (except for man-made waters) and marine areas (UNEP, 1997a,b). Urban and anthropogenic agricultural landscapes are not discussed here. ‘Amount’ is assessed by measures of area, and 'quality' by measures of population change in characteristic species. These approaches are intended to complement one another, and in some cases to provide information where data on one or other variable are difficult to obtain. In the case of grassland, we have also tested the feasibility of using information on soil degradation as a measure of quality. The study builds on procedures developed by WCMC for the WWF Living Planet Report (Loh et al., 1998), and which generated the first time-series index designed to represent global trends in biological diversity. This report outlines possible approaches to: 1) setting baseline’ values for ecosystem area within the OECD, 2) assessing current ecosystem area and condition#and 3) establishing long term population trend indicators for species within these ecosystems. It should be noted that the approach sketched here is designed primarily to use the variety of data that are already available, and to function at global or continental scale. The general method appears applicable at regional or national level, where it would benefit from the high quality data available for a few areas. Some countries are in a position to undertake new field research on land cover and species diversity; eg. a planned biodiversity monitoring scheme in Switzerland will include repeat diversity sampling on plots countrywide down to 10m” in size (Hintermann and Weber, 1999). Data of this quality are highly desirable for many purposes, and could be used to support global analysis, but for the foreseeable future it will be necessary to make use of existing data, however disparate, incomplete, or coarse in scale. A)N 22S WW r Ss MAC Natural capital indicators for OECD countries The area of the five main terrestrial ecosystem types has been mapped, where possible, for two time periods, and change in area of each calculated using GIS. The first period may be taken as the baseline for comparison with later periods. In this study, estimates of original or potential natural vegetation have provided the area estimates for ecosystem area baselines. Development of a harmonised land cover data set is fundamental to this process, and provided that spatial data can be collated according to an integrated system, estimates of future change can be calculated whenever new data become available. A number of additional sources may be taken into account both to refine the process of setting a baseline, and to make realistic estimates of pre-current cover for which spatial data have not been collated or are unavailable (ie. intermediate points in the series). Available data on species population trends reported in the literature have been collated and processed in order to indicate possible change in aspects of ecosystem quality. It is exceptional to obtain population time-series starting earlier than the 1970s. For the species illustrated, the index value is set to unity at the earliest period for which actual time-series data are available, and for many purposes this may be taken as the baseline. This baseline will continue to provide a standard for comparison as future population data are added to the series, but earlier baselines, corresponding in time to those set for land cover, must be set using other information. This ‘other information’ might consist of anecdotal reports from hunters or colonists (eg. in the case of American bison), or, perhaps, extrapolations based on biomass and carrying capacity. We have not in the present report explored the setting of 'original' species population baselines where time-series data are absent, but have no reason to doubt that it is possible, provided that the basis for such estimation is explicit. Historical ecology may provide information on this in the coming years. The issue of defining 'baselines' is important because the point at which the baseline is set determines the way in which subsequent trends are interpreted. To take a simplistic example, postwar forest area in North America and Europe shows a stable or increasing trend, but this must be seen against the background of very extensive clearance by human communities over past centuries. Within the time constraints of this study it has not been possible to provide estimations on change in wetland areas, area of the major habitat types at intermediate time points, and more natural baseline figures on species abundance. Although natural baseline figures are important as general calibration points, intermediate figures for especially the last decades provide politically significant information on the character and extent of the loss and sometimes gains of biodiversity. Although a large number of conceptual and other difficulties remains, the work to date appears to show that it is possible to bring together these existing data and theoretical frameworks in a practical system capable of generating a working index of biodiversity change. Given data limitations imposed by the short span of the project, the actual results obtained are illustrative rather than definitive ‘ aU wy ae (bubba. 48 1 “ a1) etdiel iid Me) » wahoo seat eiklctin | ef eT a = eT Hf i oa Rt lesen z pom iy A soles 1VA mingtike aml WF SIT ihren nrc i ) oe) sy soita! wg? 7. BT } a ca wie wy lal vay rs di fi iy ty x sh wth: oN wr a lite ae Ls pangs ii yt: ’ 4} 54d 7 ‘ eae ay cma Bai \ } veh fan lial ion 0 ae y a TC sieesin Bild yi i 1A ‘eth ail F Natural capital indicators for OECD countries 2. ECOSYSTEM QUANTITY The intention here is to derive numerical estimates of change in area of natural habitats (ie. of self- regenerating or largely unconverted habitat) at the level of OECD regions using an explicit and consistent methodology. We have aimed to collate a standardised set of spatial data on the occurrence of natural habitats within the OECD in order that estimates of change in area can be calculated by a GIS. Data on forest extent were available at WCMC. Data on past and present grassland and other terrestrial ecosystems were derived from outside sources as noted below (and see Annex I for data sources). Some datasets were in digital form, some were digitized at WCMC from hard copy maps. Because at most scales of analysis, ecosystems tend to grade into one another, there are fundamental logical difficulties in demarcating ecosystem boundaries and in classifying ecosystem types. Four highly generalised terrestrial habitat types are recognised for the purposes of this preliminary investigation, and wetlands are addressed briefly. The five basic ecosystem types specified (forest, grassland, desert including semi-desert, tundra, wetlands) and four OECD continental regions (North America/Mexico, Europe, Japan/Korea, Australia/NZ), produce a matrix with 20 cells. However, not all the habitats of interest occur to significant extent in all OECD regions, leaving 16 habitat/region combinations for which data are required. Void combinations are shaded in Table 1. Table 1. Habitat and species data coverage. forest grassland desert and tundra wetland semi-desert spp. | hab. | spp. N America Europe Japan/Korea Australia/NZ Note: * indicates data available, empty cells indicate no data located, shaded cells are void (although some natural grasslands exists in Japan/Korea, none is taken into account in this analysis). 2.1 Forest WCMC, working with WWF and others, has developed in digital format a world map of both current forest cover and approximate past forest cover. The ‘current forest’ coverage represents a compilation of spatially referenced data from a large number of sources, some derived from 1990s remote sensing imagery and some derived from printed maps often a decade or two older. This dataset is revised incrementally as new information becomes available. The “past forest’ coverage is an informed estimate of the general global maximum distribution of forest cover after the last glacial Digitized by the Internet Archive in 2010 with funding from UNEP-WCMC, Cambridge http://www.archive.org/details/finalreportnatur99wcmc Natural capital indicators for OECD countries period but before significant clearance by early human communities. Put in other terms, these two datasets approximate to ‘actual’ and ‘potential’ forest vegetation, respectively. Only one change has been made to the data for this study: the past forest cover in New Zealand has been revised to reflect the approximate extent of forest at around 1300AD, prior to the main impact of Polynesian peoples. The change in area of forest cover can be derived from these data using a GIS. See Figure 1 and Table 2. The scale and quality of source data mean that such estimates will be relatively robust at global, OECD and continent level, but often unreliable at the level of individual countries. Figure 1. Approximate change in forest area in OECD regions. Note: 1400 AD represents approximate original area. change in forest area 9,000,000 8,000,000 7,000,000 6,000,000 5,000,000 —~— Australia/New Zealand — Japan/South Korea —North America —*— Europe sq km 4,000,000 3,000,000 2,000,000 1,000,000 1400 1985 period Because globally consistent time-series spatial data are not readily available, it has been impossible to make rigorous estimates of forest area at a sequence of fixed intermediate time-points. It would be possible to use the variety of other information available (eg. national sources, FAO land cover and forest statistics) to make reasonable estimation of forest area at intermediate points during the present century. For north temperate areas this would be expected to steepen the first sector of the graph between ‘original’ and the middle of this century, and flatten the second sector from around 1950 onward, or produce an upward trend, reflecting the spread of plantation forests. This procedure is unable to yield information on ccentemporary rates of change or on forest condition. The former requires focused local or country-level studies of change in forest area, several of which have been reported. The latter may be estimated by change in population of indicator species (see section 3, below). It may also be assessed in terms of fragmentation, ie. the size and connectedness of forest patches, or wilderness value, or percentage of old growth forests. The 7 ; on Ce ee nah ah pilin ia semana Lk cau) ltecaail bas hld ds Wea iho’ (ay a ‘eos Mee Ve aly ‘ba! tii ig, thay is otra owt UU nr ane peat | = gener ‘Wael PO Tee: Woe dats a ihaae ool jared: ee age 1G a Giles, ih 1 ie awe pe) diet it seniogeyh su A duped ft) f Wy, roel: tly) weit i Natural capital indicators for OECD countries percentage of old growth forest (forest more than 200 years old) as a first approximation of the forest quality might be determined for most countries, eg. about 1% in Western Europe, about 15% in the USA, 52% in Canada and 25% in New Zealand (Dudley, 1992). The other aspects could not be investigated within the context of the present study. 2.2 Grassland, semi-desert and desert In terms of vegetation structure and composition these systems, particularly those labelled ‘grassland’ and ‘semi-desert’, tend to grade imperceptibly into one another, and so exemplify in acute form the demarcation problems mentioned above. We attempted to sidestep this problem by using climatic data to distinguish between them. The humidity/aridity dataset used in Middleton and Thomas (1997) was retrieved from the UNEP/GRID website. This surface was mapped to those terrestrial parts of the world not already assigned to forest or tundra. We then proposed that regions classified as hyperarid in this dataset correspond to ‘desert’; arid and semi-arid regions equate to ‘semi-desert’; and dry subhumid and humid regions (beyond the limits of forest) represent potential grassland systems. This proposed classification was then plotted and examined with reference to available vegetation or land cover maps. A significant mismatch was evident between the potential grassland areas distinguished on climatic criteria and the grassland represented on other map sources. Assuming that national sources are likely to be most representative of conditions at national or continental scale, we incorporated grassland cover from other map sources (see Annex I). The climatically defined hyperarid regions were accepted as ‘desert’; the ‘semi-desert’ category thus includes all the non-forest and non-tundra regions within the arid and semi-arid zone that are not mapped as grassland. For presentation purposes in this study, the semi-desert and desert categories have been lumped together. Estimates of current grassland area were also derived from map sources, typically from different sources than the original (potential) grassland. In some cases, the sets of data on original and current grassland appeared reasonably consistent, ie. where the area of current grass is smaller than original, this fitted with evidence of decline from other sources. In one case (Mexico) grassland area according to the 'present' cover source is much larger than originally, doubtless because of classification differences. Other than Hungary and Sweden, no country level data could be located for original natural grassland in Europe within the project period; this is probably because many authorities accept that little or no strictly natural (non-alpine) grassland exists in the region. Where such data are necessary for calculation, eg. of the percentage of natural grassland remaining, we have used the current grass estimates (but therefore inflated the percent values). Estimates of past and present grassland area for New Zealand have been derived from text and a data table, not a map (see Annex I). Estimates of original occurrence and current extent of natural grassland have not yet been collated at global level. In the absence of a complete and consistent set of grassland area estimates, data from two independent sources were investigated for suitability as indicators of change in self-regenerating area of grassland: the GLASOD project assessment of soil degradation severity, and attributes from the global land cover map produced by Moscow State University. The latter was not used, largely because the methodology behind the classification was not apparent from the supporting literature available. ais \ mays aii: NT aM, yas a. rales Wort 7 sins: sf ‘ 7 y ar ~ . he = . Natural capital indicators for OECD countries The GLASOD project aimed to assess human-induced reduction in capacity of soil to support human life, essentially through loss of productivity following soil erosion or damage to soil structure. We make the assumption that this can also be interpreted as reduction in self-regenerating area of grassland and other dryland systems. According to the NCI-framework it could also be used as a short term substitute for ecosystem quality information on biological variables The GLASOD attribute data referring to soil erosion or deterioration relate to physiographic map units. These units are not grid squares, but complex polygons that vary greatly in size and shape, thus limiting the precision with which the GLASOD data can be applied. This is because the type and severity of degradation can in principle vary between low and very high in different parts of any given map unit. However, the GLASOD dataset remains the best global evaluation of soil degradation. For the purpose of this study, we accept that areas identified by the above process as grassland, that are also ranked as non-degraded or low to medium severity degradation by GLASOD, represent ‘current self-regenerating grassland' cover. Similarly, areas ranked as high to very high severity are unlikely to represent good quality self-regenerating natural grassland. See Figure 2 and Table 2. This procedure can also be applied to semi-desert and desert systems, but these steps are not illustrated in this report. Figure 2. Approximate change in grassland area in OECD regions. Note: 1400 AD represents original area, 1900 approximate modern area of grassland, and 1985 approximate current extent of self- regenerating natural grassland. 2.3 Tundra change in grassland area and condition 2,500,000 2,000,000 1,500,000 —— Australia/New Zealand E —& Japan/South Korea > —S—North America —4— Europe 1,000,000 500,000 x ST xK & eee 0 C}-— = : 1400 1900 1985 period Tundra is defined as the treeless zone in the northern hemisphere located at higher latitudes than coniferous forests, but at lower latitudes than regions of permanent ice cover. Some tundra-like habitats occur in montane locations elsewhere; these will not be assessed in this project. Although it ic ‘bot pred Be: rary vA eed iva TUN K sescek. cat yar: Gh Live ial “tthonuke: oe ‘Hupalion | “agate nits as alan ; an ied eget { Aad oR wget t wap, pe SA at « Ayre gait paterrwarennd, gt ae Me STL IE Ipslinionalrings vctioh d . si itl amas Es “fil ate ga ait} sae Date hata a hens Hen 4 Ae aes te p aes fed one, vad ‘eres: axe rap a wii se Gone a's aps Nana a Par a her. funrseh tits im Reape Cate ; iy Ye mmm Kine 3 bol ody en nak si mie! rey WE 3 eieatir ta anit Peeves] an * win BORAGE a spiseni! 4 ts rss, iG ; i =" ny ig noel pana vane ’ tv cody F Gina §, Natural capital indicators for OECD countries is straightforward to locate the tundra region on a world map, data on change in area (in effect, change in the perimeter of forest areas or ice cover) are not available at a scale suitable for global- level analysis. However, there is no information to suggest that change in tundra area at global scale has been significant in the same way that change in forest cover over the past two millennia has been. Change in species that may serve as indicators of tundra habitat quality are discussed below. 2.4 Wetlands No comprehensive and reliable global map of wetland area, present or past, exists. Rivers, flood plains, and lakes may vary in area to some extent both naturally and through drainage, where a net loss of aquatic habitat is involved. At the same time ecosystem quality is liable to decline because of water quality impacts, over-exploitation, water-use, disturbance and habitat destruction. Unfortunately, changes in area are rarely monitored as they occur, and historical sources are in most cases inadequate to reconstruct past drainage. Good national level text information and numerical data relating to drainage are available for a small minority of countries, including some OECD members, but most information is too anecdotal or too local in scale to have been applied in the present project. 2.5 Ecosystem area summary Figure 3. Approximate change in forest and grassland area in OECD overall. Note: 1400 AD represents approximate original area, 1985 approximate current extent. Original grassland extent may be underestimated These procedures outlined above result in a generalised map of original and current extent of forest and grassland, and current extent of semi-desert, desert and tundra ecosystems in OECD regions. In a GIS, the size of the areas concerned, and the percentage that current cover is of former cover can be calculated. Soil degradation data are used to estimate current area of good quality grassland and semi-arid systems. In the OECD overall, present forest area (including plantation) is about 65% of change in ecosystem area 18000000 16000000 14000000 12000000 10000000 —& forest >< grass sq km 8000000 6000000 2000000 1400 1985 period past area of which only a part is old growth forst;present grassland is at most 40% of former and a third of this is badly degraded. See Figures 1, 2 and 3 and Table 2. aul nee Tah Le Tv VVPITP J a | gaa fet ae bwail My yor Dot ruc Be AWeeshanny marrmaeyin |! yee: , ab ANH ut 78S'L0'7 192'9S0'6 be 890'2ZL'L 029'282'2 s9 bL0'SZ0'0L OL2'e0r'SL (a4U9) GD30 O9L'c6e"L Z99"L0S"E 9c 78E'BCS Lbe900'% G8 €pe'68e"9 029°F60'8 Bovey YON 0 0 0 (0) 0 ce cLE' PL LEE OZP Baloy yynos/jueder 0 [Ser Zeos Le OPS pec olPT — [BET TST puelesZ MENTETEeNSNVY] 0 SNOID3Y G90 [80SezT Toe cer Keyin | [0 {0 [6PE'c [Store SpUeHOUTEN SU) 0 0 OOr [4a (44 Ze €89'cl 96L PE PueHeZIMS] L08°8¢ 0 4 eb 8SS GZ €Z8'S0€ 6c OLP UapemMs T Pe66e «eB Be [ee [bsrerr [STS COP UjedS} (0 T 66Le {sere TE gach) eonyod (0 [OOF /62T T [6c — [OSe6s TSTOLE puelod} 60868 0 iy cOEELT 1O0'6Ez | AEMION 0 0 001 S S 0 882 L197? | Binoquiexny PAA | PI Zel Vt oreo SBE cbc ATEN} [0 {0 /Oor So0e + (S00 Z LOS? 896°09 ~__ pueyely] [ele le {0 6ce'L P98" pueyed]] (0 Z jeseT [SO9'6L as [SPL Z 6F8"69 ATe6UnH| 0 ara OOL c66 LL c66LE ve 602'SP cescel 9088/9] (0) 0 (oor cee"L cee't 0€ 020°POL 909°6 PE AuewieS, (0 T TZZET TITEL eae LY 0c —Stst~‘“‘i«‘é«d SGOT = Br8.7¢ TES souesy Fi TO/ATO# pueyurs} (0 |0 JOOr [9ec Pole GLP er rewusC) (0 |” {oor td [c0e'Pe Z09°8L onandey yoe25| 0 0 001 6 6 74 189 SP0'6c wnibjeg 0 0 OOL GZS'S GZS oP €E9'9E c8c 6L enysny a a | (866019 =[eercorz Zé 6 T8P ZT [99 (Sle beC TPT LES sens ee [0 [9SP'EZe 8 [Sse Ze VT $9 [egererz [ESP STE T T COV LOST 8cL'G2e P Gee'é G86'6ec 16 SOL'261'S lepl6e'9 epeued 0 0 0 0 0 g 80°SL 626 P6 BSsJOy YyNo (0 0 — 0 0 9 [S87 ee eercle uede 0 0 {002 000°0E [000°ST 0 nsaeag [eeocle puejeeZz MeN] 0) S8L Leo's c6 8c 98P Sle eS (42) ec eer OOL Fle TS | eyeasny SAIYLNNOD G930 (Aypruny eed eaJe juasaid| ease sed es! ease juasaid ease jsed Aq pauyep]) % se juasaid % se jueseid LuY¥sssd Vauv VauV Vauv vauanni | * cane usayog |YF8V489H04) | cauos *S90.NOS JUAJAJJIP WO. svaie juasaid pue ysed ‘eyep OU = pu :a0N ‘(,UPY) UOISAL TAO oy} Ul vase wla}sXsoda Jo sajeumyse paseq-dey] *Z aqeL, a = Natural capital indicators for OECD countries ECOSYSTEM QUALITY As noted above, there is a total of 16 habitat/region combinations for which species data are needed. A sample of relevant species population time-series data is already available; some of these data were used in the WWF Living Planet Report (Loh et al. 1998). We have expanded this set so far as possible in order to provide optimum coverage of OECD ecosystems and continental regions within the limits of this trial project. Adequate data are available to calculate species indices over time for six of the 18 habitat/region combinations using the procedure developed by WCMC for Loh et al. (1998). These are shown as line graphs in Figures 4-9. As noted below, we are confident that some, perhaps most, gaps in these sets of data can be filled given a more substantial research phase and willingness to include more local populations. At this stage, the extent to which trends in species characteristic of a given habitat class can be taken as indicative of prevailing trends in that habitat, and whether this might be a property of all habitat types and communities of species, have been little explored. However, these cannot realistically be investigated until substantial sets of data on species and on habitat extent have been assembled, and one aim of this study is to contribute to that end. The fact that the species indices graphed here show major differences in pattern and trend, some of which can at face value be correlated with habitat area and condition, is encouraging, and suggests that the general methodology is both robust and sufficiently sensitive. 3.1 Forest The forest species data derive from long term monitoring of north temperate region bird populations in the UK (Europe) and USA (North America). These extracts make up the largest samples of species data in the study. Indices calculated from these data are graphed in Figure 4 and 5 respectively. The broad similarity of the Europe and North America graphs is remarkable; both curves are relatively flat and show a small net increase over the 25 year period represented. This could be interpreted as reflecting a phase of stability in forest condition in these areas over recent decades, although following from centuries of decline in forest area, and may be correlated with an increase in forest area in some regions through plantations. Information on the % area of old growth forest (see e.g. section 2.1), secondary forest and plantations, and other variables at the ecosystem level such as demographic structure could provide significant supplementary information on the quality of the forest. Indicators developed in the Helsinki process, Montreal process and Tarapoto proposal are most promising in this respect. No data have been collated at present on species in south temperate or tropical forests in the OECD. 3.2 Grassland, semi-desert and desert These habitats are represented by data on grassland birds in Europe and USA (North America), and by data on mammals (kangaroos) of semi-desert regions in Australia. These habitat Natural capital indicators for OECD countries categorisations are not rigid; eg. the grassland birds are liable to occur to some extent in semi- natural grassland, cultivated plains and dry farmland margins, as well as natural grassland. Indices calculated from these data are graphed in Figures 6-8. As with forest birds, there is distinct similarity between the Europe and North America graphs, although less marked in this instance. The North America curve is relatively flat and shows a net decline over the period represented, but the trend is very steeply declining and near linear in the case of the European grassland species. Both these species graphs parallel the widely acknowledged decline in area and quality of grasslands in the regions represented. The graph representing Australian semi-desert species is more uneven over time than those for species groups discussed above, but shows little net change by the end of the period shown. 3.3 Tundra This ecosystem is represented in this study only by a small sample of North American mammals (caribou, snowshoe hare) and birds. The graph shows extreme variation over time, and the index ends the period at a substantially higher value than it started. This variation is not entirely unexpected because some tundra species are well known for showing high amplitude cyclical variation in population numbers. The mainly upward trend from the mid-1980s probably reflects the recent trend for higher mean annual temperature; this leads to increased vegetation production and this in turn supports increased caribou numbers. Many workers believe that this recent temperature rise is linked to global climate change phenomena. 3.4 Wetlands Data on wetland birds, reptiles and a small number of fishes are shown in graph form in Loh et al. (1998). These figures (Fig. 2b, Map 4) are not repeated here. As noted below, time-series data on populations of truly aquatic inland water species are very rare, and the indicators so far created that relate to species using freshwater habitats rely heavily on wetland birds. Future development of species level indicators of freshwater condition will need to explore the feasibility of making greater use of time-series of catch data (which do exist, mainly for species of economic importance, eg. salmonids). In addition, the potential for more ad hoc indicators, eg. the number of European rivers with salmon runs in 1900 and the number in 1990, will need to be explored. 1 ‘ 1 vt : ‘* ngs any wi ” ~ Pal skete d m - te) orcad 7 } ¥ i! i. : bem a a 4 Natural capital indicators for OECD countries Figure 4. Europe: forest species Note: based on data for 47 bird species. 0.8 06 +7 o4 0.2 1970 1975 1980 1985 1990 1995 Figure 5. North America: forest species Note: based on data for 123 bird species. 0.8 0.6 + 0.4 02 1970 1975 1980 1985 1990 1995 1997 (Mt Natural capital indicators for OECD countries Figure 6. Europe: grassland species Note: based on data for 4 bird species. 0.8 FT 06 0.4 02 1970 1975 1980 1985 1990 Figure 7. North America: grassland species Note: based on data for 25 bird species. Natural capital indicators for OECD countries Figure 8. Australia/New Zealand: semi-desert species Note: based on data for 3 mammal species. —— 0.2 1980 1985 1990 1995 1997 Figure 9. North America: tundra species Note: based on data for 4 birds and two mammals. w et 1970 1975 1980 1985 1990 1995 1997 Natural capital indicators for OECD countries 4. ISSUES 4.1 Availability, quality and classification of land cover data For the purposes of this project, and for many other applications, the preferred state would be to have available a digital dataset, global in scope but with fine-scale resolution, with labels and attribute data collated in consistent format, and forming a time-series extending for several decades. In fact, the data available tend to be patchy in coverage, with much variation in origin, date, resolution and quality, and linked to inconsistent ecosystem classifications. Even though the OECD is probably the largest political and economic grouping where generally good quality data are available for each national or regional component, each such component typically uses a different ecosystem or vegetation classification. One fundamental limitation of using an ecosystem approach to assess change in land cover is that at most scales of analysis the several parameters by which an ecosystem is conventionally described (eg. vegetation type and structure) do not as a rule show abruptly discontinuous variation. The demarcation of boundaries between ecosystems therefore relies on subjective judgement in many instances. This limitation is assumed to be of reduced significance in the context of the present project, using a very simple ecosystem classification and working at global or continental level. We have in some cases edited or aggregated classifications applied to spatial data in order to improve uniformity. In future, the feasibility of using bioclimatic data as a primary input to land cover assessment should be investigated. We have not attempted to reinterpret the existing dataset on current forest. This is potentially a source of imprecision, given the logical impossibility of labelling land cover unambiguously in the many tropical and temperate situations where wooded land with some grass cover grades imperceptibly into grassland with sporadic trees. Given the working definition of ‘self-regenerating’ natural ecosystems (see Introduction, and UNEP, 1997a,b), it is desirable to distinguish self-regenerating from man-made grassland. There is continuing discussion over the origin and maintenance of grasslands, although there is a consensus that natural and anthropogenic fire has a central role in promoting open grassland over wooded land in many areas. Much contemporary grassland exists where climax forest vegetation has been cleared by humans, and normal succession interrupted subsequently by fire, clearance or grazing pressure. Even where not consisting of sown grass species, such grassland may be regarded as more or less ‘natural’, depending on its age of establishment. The problem of defining and classifying grassland in terms of naturalness is particularly difficult, and not capable of unique solution unless valid criteria can be defined. Defining criteria at global level is problematic for at least two reasons. Firstly, there is an essentially unbroken continuum of management intervention, from zero to high intensity regimes (in terms of factors such as mowing, fertilizing, grazing, burning). Secondly, different stretches of post-forest grassland could in principle have originated at any point within the past four millennia, and have cycled through seral stages of vegetation development several times since their origin. The NCI-framework is aiming at a highly pragmatic approach. It defines self-regenerating grassland as “all grassland areas, irrespective of its quality and use, except for planted pasture for permanent vet We ee ee : ee ee oa eal SA acal; QoS ry lle ae atts apogee. tee oe Le Py wir Carts ai oo 7 - , ~~ oral avsiear|s Aer, iS =Hey oe Ware 1) j ¥ } ‘ S i se Rae iate ttl ws toe. is: Aisi é | re ae, - ane ‘a ie BaatMlaNe a0 ND yr jor i hy ‘tie! ia i Are a hyo Hal fee tipat 1 oo ee aka a Cir! oh teat » ee) ee As ee; See Call yee Keel) Lope em ik ees: Batioete hi): task Steetee Mares are " lows de i Dee ie salen | Sy pty sla ty i Te Cait! im Pai i Pay ine Svbaijeds, i aM (ay Yee alge ope) mary cane yey ; any vi Pa ies Natural capital indicators for OECD countries livestock”, the latter which is defined man-made and assessed differently. Depending on its use and other human interventions the quality of self-regenerating grasslands can vary from 100% quality (natural) to 0% (totally degraded). The quality of it is measured by comparing it with a postulated natural or low-impact baseline grassland ecosystem specific for the area. By looking at both the loss (gain) of area and quality within the remaining area an indication is given of the overall remaining “natural capital”. Within the limitations of this study we have adopted a simplistic approach, and where external map sources have been used to delimit grasslands, we have simply used the map units labelled as ‘natural grassland’ (eg. as in the CORINE classification of European vegetation) without further investigation. We have not treated any upland grasslands in the UK as natural (and coastal grass habitats are too restricted to enter analysis at this resolution), but we have included steppe grasslands in Hungary (labelled as ‘natural’ even though with varying degrees of naturalness). For some purposes, the relative biodiversity value of unsown grassland is a more important issue than age and relative naturalness, and some zero-input grass swards maintained by close grazing (eg. on the chalk downlands of southern Britain) are extremely species rich at the quadrat scale although essentially anthropogenic in origin. 4.2 Availability and quality of species data Where time-series land cover data are not available to provide information on change in extent of natural habitats, we have sought to provide information on change in quality by means of data on species populations. It was concluded in the course of an earlier study (WCMC, 1995) that "good time-series data for global populations of individual species are very rare". This remains a true statement. However, a relatively larger (though still absolutely small) pool of data becomes available if the conditions implicit in that statement are relaxed, ie. if populations below global species level are considered, and if multiple sets of short time series are processed. The method developed by WCMC for Loh er al. (1998) was designed to make use of the less-than-perfect data that are available, and so incorporated information on national or geographic populations, sometimes local, and in several cases made use of short and asynchronous time-series of population estimates. For this project we initially carried out a search for quantitative data on populations as near species level as possible. This met with moderate success, and confirms the conclusions drawn previously. We believe that much progress could be made in any future extension of this study by focusing from the outset on quantitative data relating to populations below species level, even including very local datasets. In some instances, eg. Japan, it is apparent that data of interest do exist, but it has not proved feasible to establish effective liaison with key agencies within the project period (in part because of the mismatch in language). Within the time available we have not been able to collate data on sample species for all 18 possible OECD habitat/region combinations (see Table 1). The previous study (WCMC, 1995) suggested that reasonable data are most likely to be found for species of widespread interest, either for economic reasons (game animals, fishes, cetacea) or from a natural history viewpoint (birds and possibly butterflies). Experience since that time has tended to confirm this prediction, with just a few qualifications. ingll ie 5 4 ae yf tell h ae } o i ' i D a : jw heey bel eat i Oe Lae i Cad is a Pani eee: Cat I ibe Ng ; by i pemeye ONS ty we aris he liad iM, py Le bmea Fel Pe tay aed dy, ‘| = ia) fl i " p ice ey ee Bihar iat Mp fe aa y Pies, 3 ar eites| aaa Huey NANI tty “) ' Dy » ‘inelja i sali al HPs oth Ob" ej ww arena i Pr eaiiet Laiagel ty Dt a a et ae a edi’ Ae Fish sillsges tu ake te adit es an | a (oie, ny a His Amt i a Bia" i ry a "tl i alee ara whee ji LS a allen — ng fie» Lhe esl ns Ayabiine ap Pe ba! PN ey i ed aba t at i i * Pe is pf ‘be ; 4 ral! \ vite i j pes j pein Wel f at fj , ates hay =) Al ts ey | lee + Per) 4 i ‘ DL ? j wir / fp! yt Ay af wai i ral ] Ci i A ofr i » i fas a 1 hy yt mf wager Ms be faved tay ‘ ey AN hie a iM if ia 1 i Ls uh Tay fie ay ; Natural capital indicators for OECD countries Globally, by far the greatest monitoring effort for any group of species is devoted to marine fishes of economic importance. By far the greatest volume of time-series data relate to stock estimates and catch levels in the marine fish populations targeted by industrialised fisheries of developed countries. Birds come second to marine fishery stocks. The bird species that are surveyed regularly by networks of mainly amateur ornithologists in developed countries are by far the best known large terrestrial group. In recent years considerable attention has been devoted to the monitoring of amphibian numbers, against a background of rising concern for the widespread decline and extirpation of local amphibian populations. Although many time-series are very local in scope, and mostly relate to North American or European species, a considerable volume of data is becoming available. Accepting the geopolitical bias toward populations in developed countries, and the taxonomic bias toward commercial marine fishes and north temperate birds, there is a sparse and irregular mosaic of time-series population data available, and these can be used to generate meaningful indices of biodiversity change. Freshwaters make up a class of habitats that are not yet well covered by approaches outlined in this study. The area criterion has little validity for inland waters other than floodplains and other extensive superficial wetlands, and virtually no truly aquatic freshwater species has useful time- series population data. Most of the species using freshwaters and that have useful population data are birds; some of these are diving birds permanently residing on or around the water surface, others are birds of water margins. Data illustrating population trends in these species are certainly useful (eg. Loh et al., 1998) , but it would be preferable also to have available time-series data on a good sample of freshwater fishes and invertebrates. It may be that other methods will be needed to make use of the less quantified trend data that are available, and the validity of using catch data from inland water fisheries should be investigated. The possibility of deriving low-impact baselines from similar, relatively unaffected ecosystems should be investigated where historical data are absent. 16 ic 7 =i ag Pe MV . ! on) Ves) gh & bb Pa aaed yee tue? i - ‘ary asyeat me bia} ave (fen | aides. a hs i gl a c deflate. aan) ois hy vrabell te Thi aii ; ' —_" haa re, bla etn Nal, ol beg Pnate!.. nits a mo) Roe a er Le Mae lh pair Jeri 7 - 7: ae, an i= AFP OY 1" 7 7 rete ‘iy ‘sys § ei 7 Wad : why vinyl baud mp poke fpeeyine. edttt BeOtAT oe WIC ; i a , ; Bey sadn pee ry TS eee, ee ries ey, Dib AY « ~*~ ” = . ~ ei ae ’ ijapiyt re igs Hatt ele Fala pe (as Nites | . Fe ‘ } 4 pallid ial. + dal Ai CGIR if poy 0 Ms AL ay, es Bek PL Daigeet Maas ay me he se nn Tiley te eglt : i PMighe i side ai et a! ve a a whi DP i RN A 7 4 Mayantidy** iy eh Natural capital indicators for OECD countries 5. CONCLUSIONS AND RECOMMENDATIONS 5.1 Shed 5.3 5.4 The goal of this feasibility study has been to apply a consistent methodology to land cover and time-series species population data in order to generate meaningful indices of ecosystem area and quality, respectively, within the OECD region. Using readily available data sources, map-based estimates of approximate original and current forest and grassland have been generated. Current forest area, including plantation forest, in the OECD is about 65% of past area. Measures of degradation, fragmentation or wilderness can be applied to these data; this has been done for grassland. Alternatively, according to the NCI-framework these are not integrated in figures on the size of self- regenerating ecosystems but will be part of the information on ecosystem quality. The greatest area of grassland in the OECD occurs in North America (including Mexico); current area is about 25% of past area, and about half of this is badly degraded. A novel method to process available time-series population data has been further tested. Sample indices of change in species of forest, grassland, semi-desert and tundra have been produced. Points of similarity and difference among these indices suggest that these indices do contain significant information about change in species of different ecosystems. Ecosystem quality indicators at the ecosystem level could be very useful in covered and complex ecosystems such as forests (WCMC, 1995). Proposals on indicators in the Montreal process, Helsinki process and Tarapoto proposal are important in this respect. Although preliminary estimation of baseline figures on species abundance, intermediate time points and wetlands were not feasible within the limitations of this study, the goal of this study to investigate whether the NCI-framework could be a feasible, significant and universal assessment methodology have largely been met. It is recommended that a full-scale project under the OECD be undertaken, aiming to: 1) Further refine and harmonise the sets of digital land cover data used as the basis for assessing ecosystem area, and refine assessment of condition. 2) Collate a geographically and taxonomically broader set of time-series and baseline data on species populations, the better to reflect change in different ecosystems, and devoting particular effort to filling the information gap on freshwaters. 3) Develop effective liaison with in-country sources of data, and with personnel shaping OECD policy goals in relation to environmental planning. a De Aes iphere 1 pay BH) sapeiopary jae by. ¥ ean nay Latog | ae ied we “ “ » ae ie aioe ae Noirerdte ge ‘pti; ey vp as (fede ot ae, Si stint ty or ry oe ane a ene iF 4 Ly ie Mi: a . fh niga are i mi | idea i ra ia ta aM ol 7 “i i Natural capital indicators for OECD countries 6. REFERENCES Hintermann, U. and Weber, D. 1998. Biodiversity Monitoring in Switzerland. Report on the status of the project at the end of 1998. Hintermann & Weber Ltd., for Swiss Agency for the Environment, Forests and Landscape. Middleton, N. and Thomas, D. (Eds). 1997. World Atlas of Desertification. Second Edition. UNEP. Arnold, London. Loh, J., Randers, J., MacGillivray, A., Kapos, V., Jenkins, M., Groombridge, B. and Cox, N. 1998. Living Planet Report 1998. WWF International, Gland, Switzerland. UNEP. 1997a. Recommendations for a core set of indicators of biological diversity. Note by the Executive Secretary. UNEP/CBD/SBSTTA/3/9. (available from the CBD Secretariat website, http://www.biodiv.org/). UNEP. 1997b. Recommendations for a core set of indicators of biological diversity. Background paper prepared by the liaison group on indicators of biological diversity. UNEP/CBD/SBSTTA/3/Inf.13. (available from the CBD Secretariat website, http://www. biodiv.org/). WCMC. 1995. Biodiversity indicators for integrated environmental assessments at the regional and global level. Unpublished report, commissioned by RIVM. ~eia! iF ; De i a Nore at 7 th i Sh acieniy ehh aro | | hein 8 1 ith i) jay ye ee ee itt as . a - s ' a. - ert Aon hs gpl me Nema aal Lin Sy’ Aes . ag ie "ie a b\ip> Reba shart ne _) on. eae, Ta ah ¥ it a y oy oe hires sang otic ‘fi py vlia Tipe fi, af ae rf | i an meal Poy act th wu aha by A tin ey ‘ : nee llfp de ¢ ifs we) pee oy) nelle MAY ls iain Midi Tesh clcaka!. WANES aes all as r Natural capital indicators for OECD countries ANNEX I. Sources of map data. FOREST: CURRENT Primarily from CD-ROM: Iremonger, S., Ravilious, C. and Quinton, T. (Eds). 1997. A global overview of forest conservation. WCMC and CIFOR, Cambridge, UK. Includes documentation of multiple sources used. Some open forest cover from: U.S.G.S. EROS Data Center/GLCCD version 1.2. 1998. Land cover characteristics database. The data have 1- km nominal spatial resolution, derived from 1992-1993 monthly AVHRR images, analysed for NDVI. FOREST: ORIGINAL Mexico Map of Ecoregions of Latin America and the Caribbean Scale: |: 15,000,000. In: Dinerstein, E., D.M Olson, D.J. Graham, A.L. Webster, S.A. Primm, M.P. Bookbinder and G. Ledec. 1995. A Conservation Assessment of the Terrestrial Ecoregions of Latin America and the Caribbean. Published in association with The World Wildlife Fund. The World Bank, Washington, DC. Australia Camahan, J.A. 1989. Australia - Natural Vegetation. Australian Surveying and Land Information Group, Department of Administrative Services. Scale: 1:5,000,000. Shows the probable state of Australia's vegetation around 1788 when European settlement began, i.e. pre-settlement vegetation. The vegetation cover is defined in terms of its growth form, foliage cover and, in most cases, predominant plant genus North America (Canada, USA) Map of Ecoregions of the USA and Canada. Scale: 1:15,000,000. In: Ricketts, T.H., E. Dinerstein, D.M. Olson, C.J. Loucks, W.M. Eichbaum, D.A. DellaSala, K.-C. Kavanagh, P. Hedao, P.T. Hurley, KM. Camey, R.A. Abell, and S. Walters. 1997. 4 conservation assessment of the terrestrial ecoregions of North America. Volume I - The United States and Canada. Draft Report. World Wildlife Fund. Washington, DC. Europe Bohn, U. and Katenina, G.D. 1994. Map of Natural Vegetation (of Europe). Komarovy Botanical Institute, St Petersburg. Scale1:2,500,000. Japan Potential Natural Vegetation Map of Japan. Scale: 1: 8,000,000. Miyawaki and Okunda, 1975. New Zealand Map of forest cover before Polynesian settlement, after McGlone (1989) in The State of New Zealand's Environment. Chapter Eight. The State of our Land. Available in .pdf from http://www.mfe.govt.nz/about/publication/ser/ser.htm Russian Federation, South Korea Milanova, E.V. and Kushlin, A.V. (Eds). 1993. World Map of Present-Day Landscapes. Prepared by Moscow State University and the United Nations Environment Programme. GRASSLAND New Zealand Current and original grassland statistics from The State of New Zealand's Environment. Chapter Eight. The State of our Land. Available in .pdf from http://www.mfe.govt.nz/about/publication/ser/ser.htm Australia Source as for original forest. The following grassland groups have been included : Astrebla (Mitchell grass), Dichanthium (bluegrass), Graminoids, Other grasses, Triodia and/or Plectrachne. Other grasses were included in current cover only where they overlapped the original areas of “other grasses’. We inferred that new areas of “other grasses” were likely to be crops or regeneration other than natural grassland. North America / Mexico Original grassland from GLOBAL 200 Ecoregions. Dinerstein, E. et. al. WWF. 1998. Details of classes included available at WCMC. Current grassland from U.S.G.S. EROS Data Center/GLCCD version 1.2 (1998) North America land cover characteristics database. The data have 1-km nominal spatial resolution, derived from 1992-1993 monthly AVHRR images, analysed for NDVI. The following classes were taken from the “North America Seasonal Land Cover regions legend’: Grassland Grassland (short grass prairie) Grassland (short-mid grass prairie) Grassland (tall grass prairie) Grassland (warm season grasses) Europe Original grassland Source as for forest. The following grassland types were included: L— Forest steppes M - Steppes 19 ‘ vs he) ‘ mi 6 iobeals =e : nA | Natural capital indicators for OECD countries M4 — Desert steppes Current grassland Source: CORINE Land Cover Data (received at WCMC in 1996). “natural grassland” was the only class included. GLOBAL The World Map of Present-day Landscapes compiled by Moscow State University and UNEP, based on climate and soils (Milanova & Kushlin, 1993) was used to plot a global map of deserts/grasssland: GLASOD The “severity” class in the GLASOD soil degradation dataset was used to classify current grasslands. (For Turkey, where we had no current grassland data, the original grasslands were overlaid with GLASOD). Severity: (low) — 4 (very high). HUMIDITY INDEX (UNEP/GRID) Humidity index was used in areas not covered by current forest or current grassland or tundra. The following zones were used: Desert = hyperarid Desert = arid/semi-arid ?grassland = dry sub-humid / humid /cold TUNDRA The figures for tundra were taken from CAFF Habitat Report No. 5 Gaps in Habitat Protection in the circumpolar Arctic: A preliminary Analysis (February 1996), except for Norway, Sweden and Finland where the area north the original forest line was taken as tundra. ANNEX II. Species population trend indices. This study relies on procedures developed by WCMC for the WWF Living Planet Report (Loh et al., 1998). Each species is given equal weighting, regardless of taxonomic grouping, population size, and the number of regional populations for which data are available. For the OECD project the calculations refer to five year intervals from 1970 to 1995 and also a figure for 1997. The aim is to give equal weighting to small and large populations to calculate the index of change, therefore the arithmetic means of all different samples cannot be used. Nor can an averaged percentage change be used because of the asymmetry of these (eg. a change from 100 to 5 is a 95% decrease, a change from 5 to 100 is a 2000% increase — in any meaningful index these two should cancel each other out). The methodology is accordingly based on geometric means. For each sample year (ie. the 5-year intervals) for each species the data are logged and the difference between sample years is calculated. The mean change between sample years is then calculated by averaging the differences for all species and then anti-logging (ie. 10°). We then have the mean changes (trends) for each sample year interval. It is then necessary to normalise in order to connect all the trends into one line - ie. the 1970 population is set at 1, and the subsequent virtual populations are calculated from that point. Standard Error is calculated using the formula: Standard Error = Standard Deviation / Square Root (n). Error bars belong to latter of the two sample years for any particular interval. 20 i ] cD ia 1 Ar f tee i ii { 1 2 a tee a 7 i il : i i 4 =f i i i i ho J = i i i u 1 i, ; i ip Mae j \ ia ‘ : ny n i t I ~ J ’ = ’ ; at } i 1} 4 a j yim! i ts : 2 ae i a ee i i ‘ Ob PY ' ie BY Oty) ' 4 i iD I up ‘ i rr 0 ee pein: Faas Wy Pe are Pe ty a vee Os as ya HOH. Py Hip vie mit mpl Wibetel ry, f eat ie Hens Paes i ty ae hie Cone da Neeadlhy Aiea , ae ea ete pais dt’ Diy ar) ' Tih Ll ity i» ei: mi 5 Air j Ne P fy fieldivighe dt \ r : ‘os 4 ile a iV