Full text of "Response of an atmospheric prediction model to time-dependent sea-surface temperatures."

See other formats

NAVAL POSTGRADUATE SCHOOL

Monterey, California

THESIS

RESPONSE OF AN ATMOSPHERIC PREDICTION MODEL
TO TIME- DEPENDENT SEA-SURFACE TEMPERATURES

Peter

Henry Ranelli

March 19 84

Thesis Advisor:

L. Elsberry

Approved for public release; distribution unlimited,

121567k

SECURITY CLASSIFICATION OF THIS PACE (Whan Data

Entarad)

REPORT DOCUMENTATION PAGE

READ INSTRUCTIONS
BEFORE COMPLETING FORM

1. REPORT NUMBER

2. GOVT ACCESSION NO.

3. RECIPIENT'S CATALOG NUMBER

4. TITLE (and Submit)

Response of an Atmospheric Prediction
Model to Time- Dependent Sea-Surface
Temperatures

5. TYPE OF REPORT & PERIOD COVERED

Master's Thesis;
March, 19 84

6. PERFORMING ORG. REPORT NUMBER

7. AUTHORS

Peter Henry Ranelli

8. CONTRACT OR GRANT NUMSERC*,)

• ■ PERFORMING ORGANIZATION NAME AND ADDRESS

Naval Postgraduate School
Monterey, California 9 3943

tO. PROGRAM ELEMENT. PROJECT, TASK
AREA 4 WORK UNIT NUMBERS

II. CONTROLLING OFFICE NAME ANO ADDRESS

Naval Postgraduate School
Monterey, California 93943

12. REPORT DATE

March 19 84

'3. NUMBER OF PAGES

112

M. MONITORING AGENCY NAME « AOORESSfU dlftarant from Controlling OHIca)

IS. SECURITY CLASS, (ol thta report)

Unclassified

15*. DECLASSIFICATION/ DOWNGRADING
SCHEDULE

'» DISTRIBUTION STATEMENT (ot thla Raport)

Approved for public release; distribution unlimited.

17. DISTRIBUTION STATEMENT (ot tha abatract antarad In Block 20. II dlltarant trom Raport)

IB. SUPPLEMENTARY NOTES

U KEY WOROS (Conllnuo on ra<raraa alda II nacaaaary and Idantlty by block numbar)

Atmosphere-Ocean Coupled Models T-EOTS
Sea Surface Temperature
NO GAPS

20. ABSTRACT (Contlnua on rararaa alda II nacaaaary and Idantlty by block numbar)

DO , '*:*» 1473

EDITION OF 1 NOV «9 IS OBSOLETE
S/N 0102- LF- 014- 6601 1

UNCLASSIFIED

SECURITY CLASSIFICATION OF THIS PAGE (Whan Data Mntarac

#20 - ABSTRACT - (CONTINUED)

force the system. The 10-day predictions are compared to
determine if a coupled atmosphere-ocean model would improve
or deteriorate the atmospheric predictions. The case
analyzed occurred after the oceanic spring transition so
that only small increases in SST occurred. Use of time-
dependent SST resulted in only small changes in latent,
sensible and total heat fluxes, and in storm tracks and
intensities. Thus, further case studies of the atmospheric
response are necessary to indicate whether coupled
atmosphere-ocean models are required on 10-day time scales.

SN0,02-LF-0U-6601 „ UNCLASSIFIED

SECURITY CLASSIFICATION OF THIS PAGEfWh»n Dmtm Enfrmd)

Approved for public release; distribution unlimited,

Response of an Atmospheric Prediction Model
to Time-Dependent Sea-Surface Temperatures

Peter Henry Ranelli
Lieutenant Commander, United States Navy
B.S., Rensselaer Polytechnic Institute, 1975

Submitted in partial fulfillment of the
requirements for the degree of

MASTER OF SCIENCE IN METEOROLOGY AND OCEANOGRAPHY

from the
NAVAL POSTGRADUATE SCHOOL
March 19 84

ABSTRACT

The purpose of this research is to explore the need for
time -de pen dent sea-surface temperatures in atmospheric model
predictions up to 10 days. The Navy Operational Global
Atmospheric Prediction System is used in this study. First,
a control run is made in which the sea-surface temperature
(SST) is fixed in time. In the test case, the observed SST
analyzed each 12 hours by the Fleet Numerical Oceanography
Center are used to force the system. The 10-day predictions
are compared to determine if a coupled atmosphere-ocean model
would improve or deteriorate the atmospheric predictions.
The case analyzed occurred after the oceanic spring transi-
tion so that only small increases in SST occurred. Use of
time-dependent SST resulted in only small changes in latent,
sensible and total heat fluxes, and in storm tracks and
intensities. Thus, further case studies of the atmospheric
response are necessary to indicate whether coupled atmosphere-
ocean models are required on 10-day time scales.

TABLE OF CONTENTS

I. INTRODUCTION H

II. BACKGROUND 13

A. THE AIR-SEA INTERACTION PROBLEM 14

B. U.S. NAVY ROLE/REQUIREMENTS 16

C. COUPLED ATMOSPHERE -OCEAN MODELS 17

D. COUPLING SCHEMES 18

III. PROCEDURE 21

A. NAVY OBSERVATIONAL GLOBAL ATMOSPHERE
PREDICTION SYSTEM (NOGAPS) 21

B. SEA SURFACE TEMPERATURES 22

C. INITIAL CONDITIONS 24

D. EXPERIMENTAL PROCEDURES 26

IV. HEAT FLUX ANALYSIS 27

A. SEA SURFACE TEMPERATURE CHANGES 2 7

B. SURFACE HEAT FLUX CHANGES 32

1. Sensible Heat Flux 33

2. Latent Heat Flux 35

3. Total Heat Flux 37

C. SUMMARY 39

V. CYCLOGENESIS PREDICTIONS 41

A. INTRODUCTION 41

B. STORM TRACKS 42

C. FORECAST DIFFERENCES 48

D. SUMMARY 50

VI. CONCLUSIONS 54

APPENDIX A: NAVY OPERATIONAL GLOBAL ATMOSPHERIC

PREDICTION SYSTEM (NOGAPS) 56

APPENDIX B: THERMAL OCEAN PREDICTION SYSTEM-
EXPANDED OCEAN THERMAL STRUCTURE
(TOPS-EOTS) 61

APPENDIX C: SYSTEMATIC ERROR IDENTIFICATION

SYSTEM (SEIS) 64

APPENDIX D: FIGURES 67

LIST OF REFERENCES 108

INITIAL DISTRIBUTION LIST 111

LIST OF FIGURES

1. Methods for coupling atmospheric and oceanic models

(a) minimal feedback, (b) non-synchronous and

2. Schematic of experiment design for (a) control run

(b) SST run. SST fields input every twelve hours.
History files output every six hours. 6 3

3. Sea-surface temperature fields used in the initial
conditions in the model run. Contour interval is 2°C. 59

4 . The difference between the SST field input at six hours
of the model run and the initial SST field. Contour
interval is 0.5°C. Thin solid lines are higher SST,
thick solid is no change and the dashed lines are lower
temperatures. 7Q

5. As in Fig. 4, except for 30 h. 71

6. As in Fig. 4, except for 54 h. 72

7. As in Fig. 4, except for 78 h. „ 73

8. As in Fig. 4, except for 102 h. 74

9. As in Fig. 4, except for 126 h. 75

10. As in Fig. 4, except for 150 h. 76

11. The cumulative change in the SST field summed over

24 h intervals. Contour interval is 2°C. 77

12. (a) Mean sensible heat flux for the Pacific Ocean, SST
run. Contour interval is 1.0 gm-cal/cm^-h. (b) Differences
in the sensible heat flux, control run minus SST run.
Contour interval is 0.5 gm-cal/cm^-h . 78

13. As in Fig. 12, except for the Atlantic Ocean. 79

14. (a) The standard deviations of the sensible heat flux for
the Pacific Ocean in the SST run. Contour interval is
1.0 gm-cal/cm^-h . (b) The differences in the standard
deviations of the two model runs, control run minus SST
run. Contour interval is 0.5 gm-cal/cnr-h . 80

15. As in Fig. 14, except for the Atlantic Ocean. 81

16. As in Fig. 12, except for the latent heat flux.

(a) Contour interval is 5 gm-cal/cm2-h. (b) Contour
interval is 1 gm-cal/cm2-h. 82

17. As in Fig. 12 , except for the latent heat flux in the
Atlantic Ocean. (a) Contour interval is 5 gm-cal/cm2-h.

(b) Contour interval is 1 gm-cal/cm2-h . 83

18. As in Figs. 14, except for the latent heat flux.

(a) Contour interval is 3 gra-cal/cm2-h . (b) Contour
interval is 1 gm-cal/cm2-h. 84

19. As in Fig. 14, except for the latent heat flux in the
Atlantic Ocean. (a) Contour interval is 3 gm-cal/cm2-h.

(b) Contour interval is 1 gm-cal/cm2-h. 85

20. As in Fig. 12, except for the total heat flux.

(a) Contour interval is 5 gm-cal/cm2-h . (b) Contour
interval is 1 gm-cal/cm2-h. 86

21. As in Fig. 12, except for the total heat flux in the
Atlantic Ocean. (a) Contour interval is 5 gm-cal/cm2-h.

(b) Contour interval is 1 gm-cal/cm2-h. 87

22. As in Fig. 14, except for the total heat flux.

(a) Contour interval is 5 gm-cal/cm2-h . (b) Contour
interval is 0.5 gm-cal/cm2-h. 88

23. As in Fig. 14, except for the total heat flux in the
Atlantic Ocean. (a) Contour interval is 5 gm-cal/cm2-h.

(b) Contour interval is 1 gm-cal/cm2-h . 89

24. Tracks for storm P4 . Solid is analysis, dashed is
control run and dotted is SST run. "x" indicates
position at 126 h. "o" indicates position at 138 h. 90

25. Storm parameters for the storm P4 . (a) SST at storm
center in °C. (b) Heat flux at storm center in gm-
cal/cm2-h. (c) Radius of storm in km. (d) Central
pressure in mb . Solid is analysis, dashed is control

run and dotted is SST run. 91

26. As in Fig. 24 except for storm A2 . "x" indicates
position at 126 h. "o" indicates position at 162 h. 92

27. As in Fig. 25 except for storm A2 . 93

28. As in Fig. 24 except for storm A3. "x" indicates
position at 138 h. 94

29. As in Fig. 25 except for storm A3. 95

30. As in Fig. 24 except for storm Pi. "x" indicates
position at 6 h. 9 6

31. As in Fig. 25 except for storm PI. 9 7

32. As in Fig. 24 except for storm P2. 98

33. As in Fig. 25 except for storm P2. 99

34. As in Fig. 24 except for storm Al . 100

35. As in Fig. 25 except for storm Al . 101

36. As in Fig. 24 except for storm P3. 102

37. As in Fig. 25 except for storm P3. 103

38. Horizontal distribution of model large-scale prognostic
variables. 104

39. Vertical distribution of model large-scale prognostic
variables. 105

40. The twenty-six TOPS-EOTS ocean thermal structure
parameters. 106

41. SEIS derived location errors measures. 107

ACKNOWLEDGEMENTS

I would like to express my sincere gratitude to Professor
R. L. Elsberry for assistance and guidance during this
research. The time and effort that Dr. T. E. Rosmond always
had available to assist me with running NOGAPS has been
greatly appreciated. Dr. T. Tsui and Mr. P. Harr provided
considerable aid, expertise and support which made my work
more pleasurable. Finally, I would like to thank LCDR S. A.
Sandgathe, who provided the initial inspiration and continuing
motivation during this project.

I. INTRODUCTION

It is well known that important interactions between the
atmosphere and the ocean exist on time scales of a month or
longer. The interaction between the atmosphere and the
ocean on shorter time scales is less well understood. How-
ever, heat fluxes from the ocean into the atmosphere are
believed to play an important role in many atmospheric
circulations .

Rapid advancements in the last three decades have greatly
improved both the quality and the speed of numerical weather
prediction models. Presently, the accuracy of the model
forecast decays as the forecast time increases and the fore-
casts are, in general, no better than climatology after five
or six days. To improve model forecasts beyond this present
limitation, some type of feedback between the atmosphere and
the ocean most likely will be required.

This thesis is the first in a series of case studies
designed to study the necessity and feasibility of coupling
an atmospheric model and an oceanic model. The techniques
for running an atmospheric model with a time-dependent sea-
surface temperature (SST) were developed and then used to
make two atmospheric model prediction runs. One was a con-
trol run in which the SST was held constat, as is presently
done in operational models. In the second model run, a

time -de pen dent SST was used to simulate a coupled air-ocean
model. Actual SST analyses were used in this "perfect-prog"
approach. The changes in surface heat fluxes were then
analyzed.

The following chapters describe the experiment and the
changes in the model response between the two model runs.
Chapter II is background on the scientific considerations for
this experiment. Chapter III is an explanation of the experi-
ment design including a brief description of the atmospheric
model used. Chapter IV is the analysis of the changes in the
heat flux. Chapter V analyzes the changes in this cyclo-
genesis as forecast by the two model runs and compared to
the actual storm development. Chapter VI contains the
conclusions reached in this study and recommendations for
further study.

II. BACKGROUND

The theoretical limit of predictability of numerical
weather prediction is on the order of 15 days [Rosmond et
al., 1983]. To improve from present forecast capability of
perhaps three to five days to the theoretical limit will
require a large amount of effort and research. Many problems
stand in the way of the researcher attempting to reach this
goal. A nearly perfect numerical model will be required
before a 15-day forecast can be attained. More complete
understanding of many atmospheric processes and many improve-
ments to available models will be required.

During the short time scales, atmospheric changes on a
synoptic scale are mainly a result of dynamical forcing.
The barotropic and baroclinic processes of the atmosphere
which are the main contributors to the dynamical forcing
are well represented in numerical models. The forecast
problem on the short time scale is then one of correctly
initializing the numerical model and then integrating the
initial conditions forward in time. For the medium range
(5-15 day) time scales, external forcing of the atmosphere,
such as diabatic processes become increasingly important in
determining the atmospheric response. In the 15-30 day range,
the forecast problem becomes less of an initial value problem
and more a problem in which the external forcing and/or the

diabatic processes begin to dominate the predicted circula-
tions. Diabatic processes are in turn forced by external
factors such as the oceanic heat source.

A. THE AIR-SEA INTERACTION PROBLEM

There are many sources and sinks of heat in the atmosphere,
including latent heat release, solar radiation, longwave
radiation and the heat fluxes across the air-sea interface.
Air-sea interaction becomes an important physical process
that has to be modelled on 15-day time scales. Sandgathe
(19 81) has concluded that numerical model forecasts of mari-
time cyclogenesis require an accurate specification of the
air-sea fluxes. All fluxes across the air-sea interface can
be modelled as a function of the sea-surface temperature
(SST) . It is the hypothesis of this study that a sophisti-
cated model capable of forecasting on a 5-15 day time scale
will require a time-dependent SST. Without this time-dependent
SST, representation of the interface fluxes may be inaccurate
and this will deteriorate the forecast.

There are many oceanic processes that can cause signifi-
cant changes in the SST on 10-day time scales and result in
a response in the atmospheric model. Large changes in SST
can occur during the spring and autumn transition periods in
the ocean. During this period, there are warm and shallow
ocean mixed layers that may deepen and cool rapidly in
response to atmospheric forcing [Camp and Elsberry, 19 78;
Elsberry and Camp, 19 78; Elsberry and Raney, 19 7 8J . During

winter, the ocean mixed layer is deeper and the changes in
SST due to winter storms are smaller. Summer storms are
less intense than winter storms and do not force a large
change in the SST.

Western ocean boundary current regions, such as the
Kuroshio and the Gulf Stream, are prime areas of cyclogene-
sis [Sanders and Gyakum, 19 80J. These ocean current features
meander eastward after departing from the east coast of
continents. Locations of the associated SST gradients will
be important in determining the atmospheric response to these
features .

Equatorial regions may also experience significant
changes in SST, which alter the surface heat flux and the
amount of deep convection in this region. This response is
felt in less than one day in the lower troposphere and, in
special circumstances, can affect the long waves of the
mid-latitudes within seven days.

An oceanic forecast model would be required to provide
the time-dependent SST to force an operational atmospheric
model. While the oceanic model might be run independently
to generate the SST, it is assumed that the atmospheric
model might provide important feedback to the oceanic model.
Therefore, a coupling of the two models might be required to
provide the best forecasts both in the atmosphere and the
ocean.

B. U.S. NAVY ROLE/REQUIREMENTS

Environmental factors both in the air and the ocean will
be an important consideration in many decisions made by
operational commanders. On a short time scale, immediate
operations such as flight operations from an aircraft
carrier or an underway replenishment may be affected. The
ability of anti-submarine warfare operators to find a sub-
marine or the submarine's ability to avoid detection is
always very dependent on knowledge of the thermal structure
of the ocean and the resulting acoustical propagation paths.
Long range plans, such as an ocean crossing, planning for
an amphibious landing, or a major fleet exercise could be
closely linked to environmental considerations.

Modern weapon systems are becoming more sophisticated
and complex. Environmental conditions are an important
factor in the development and operational application of
these systems. During the weapon development, climatological
variations are often used for design purposes. When these
weapon systems are deployed, their effective utilization
requires an accurate and complete forecast of the actual
atmospheric variations to be encountered.

The Fleet Numerical Oceanography Center (FNOC) and the
Naval Environmental Prediction Research Facility (NEPRF)
recognized the need to improve medium-range forecasts and
began a study of the feasibility of developing a coupled
atmospheric-oceanic model in early 19 82. The goal was to

eventually provide an accurate forecast of both the ocean
and the atmosphere for a 15-day period. The studies by
Elsberry et al., (19 82) and Rosmond et al., (19 83) concluded
that additional research is required before the FNOC atmos-
pheric forecast is extended beyond five days, or before an
atmospheric model and an oceanic model are coupled. The
question explored here is whether a viable forecast beyond
five days requires a time- dependent SST, which might be pro-
vided by a coupled air-ocean model. Before a coupled model
becomes operationally useful, the sensitivity of the atmos-
pheric model to the air-sea interaction must be understood.

C. COUPLED ATMOSPHERE-OCEAN MODELS

Previous studies regarding coupled models have mainly
focused on the climatological effects of a changing SST on
the atmosphere. These studies are mainly concerned with the
atmospheric response over months or years. A comprehensive
review of these studies was done by Elsberry et al. , (1982).

Arpe (19 81) has examined the sensitivity of an atmospheric
model to a different SST field on short to medium range time
scales. Arpe worked with the ECMWF model, which presently
uses the climatological SST to calculate surface fluxes.
Arpe substituted large scale temperature anomalies, some with
values as high as 3°C, into the model. The sea-surface
temperature anomalies were fixed in time. He showed that
forecasts beyond. six days were sensitive to large scale SST

anomalies. He also showed higher SST values in a region
resulted in a more rapid and intense development of cyclones.

This study is the first in a series of case studies de-
signed to evaluate what effects a coupled atmosphere-ocean
model may have on an atmospheric prediction in the medium-
range forecast periods. Rather than using an oceanic model
to provide the time-dependent SST, the actually observed SST
evolution will be specified. This "perfect prognosis" of
specifying the SST avoids any errors which might be introduced
by an ocean model. Studies of the atmospheric response and
feedback processes using "perfect-prog" SST forcing is essen-
tial to identify potential atmospheric model-dependent
problems before full interaction of the two models is
attempted. 3y avoiding errors that might result from biases
in the oceanic model, the analysis of the atmospheric model
response is made simpler.

D. COUPLING SCHEMES

Three methods have been described by Elsberry et al.,
(19 82) for coupling an atmospheric model to an oceanic model.
The methods are, in increasing sophistication, minimal feed-
back or weak coupling, non-synchronous coupling and synchronous
coupling (Fig. 1) . The coupling strategies must take into
account the inherent differences in the time steps of an
atmospheric model and an oceanic model. Atmospheric models
normally have an advective time step on the order of five
minutes, with the calculation of diabatic effects every 30

to 40 minutes. Oceanic models have a time step on the
order of one hour.

The simplest coupling scheme is the minimal feedback type
of coupling (Fig. la) . For this type of coupling, the
analyzed SST field at the initialization time of the atmos-
pheric model is used as input for the model initial conditions.
The SST then remains constant for the entire model run. This
type of coupling is used in present operational models. Beyond
five-day forecast periods, air-sea interaction plays an in-
creasing role in the atmospheric response and this type of
coupling may not prove accurate for medium range forecasts.

The second method is non-synchronous coupling (Fig. lb) .
In this case, a SST prediction for the entire forecast period
is assumed to be available and is used as input during the
appropriate times in the atmospheric model runs. The time-
dependent SST can be provided from either an ocean model
forecast that has been run independently of the atmospheric
model or from analyzed SST fields in a "perfect prog" hindcast.
This method provides a more realistic representation during
periods of changing physical conditions than does the weakly
coupled method. The non-synchronous method may be impracti-
cal for operational use if an independently-run ocean prediction
develops biases which would overwhelm a forecast by the atmos-
pheric model during a 10-15 day forecast. However, this type
of coupling is an excellent research tool when using the
"perfect-prog" SST fields.

The final and most sophisticated scheme is the fully
synchronous coupling (Fig. lc) . The two models are run
concurrently and provide feedback to each other at the
appropriate point in the model integration. In this way,
the SST used in the atmospheric model and the atmospheric
forcing (surface wind, surface heat fluxes, precipitation)
for the oceanic model are being continually updated. While
this is obviously the most complicated of the three schemes,
it should also provide the best forecast. Only a fully
synchronous coupled atmospheric-oceanic model should provide
a 15-day forecast capability [Rosmond et al . , 1983].

In summary, the goal of this research is to determine
the atmospheric forecast model response to a time-dependent
SST. The study was conducted using a non-synchronous coupling
with "perfect-prog" SST to isolate the atmospheric response.
By studying the response in this type of model, understanding
of the role of air-sea interaction can be improved. This
better understanding of the model air-sea interface processes
should then ultimately lead to a fully synchronous coupled
model .

III. PROCEDURE

An experiment was designed to examine the sensitivity of
an atmospheric prediction model to time-dependnet sea-surface
temperatures (SST) . The model chosen for this experiment
was the Navy Operational Global Atmospheric Prediction Sys-
tem (NOGAPS) . Time-dependent sea-surface temperatures were
obtained from the Fleet Numerical Oceanography Center (FNOC)
twice-daily analyses.

The experiment was straightforward in design. Two model
runs to ten days forecast time were made with NOGAPS. In
the first run, designated the control run, the SST was held
constant as captured in the initial conditions. In the second
run, designated the SST run, the SST were updated every twelve
hours. The model atmospheric response in each of the two
runs was then analyzed for changes. Specifically, the changes
in the intensity of cyclogenesis and storm tracks were
examined.

A. NAVY OPERATIONAL GLOBAL ATMOSPHERIC PREDICTION SYSTEM
(NOGAPS)

NOGAPS is the Navy's state-of-the-art atmospheric predic-
tion model. The model was made available by Dr. T. Rosmond
of NEPRF. The version of NOGAPS used in this experiment con-
tains all modifications made to the system through July 19 83.

The NOGAPS forecast model is a six-layer, sigma coor-
dinate, primitive equation model. It is based upon the UCLA

general circulation model (GCM) , described by Arakawa and
Lamb (19 77) . The diabatics of the model are of full GCM
sophistication. NOGAPS includes the parameterization of the
planetary boundary layer (PBL) after Randall (19 76) and
Deardorff (1972); cumulus convection using the Arakawa-Schubert
(19 74) scheme; and radiation as described by Katayama (19 72)
and Schlesinger (19 76) . It should be noted that NOGAPS
differs from the UCLA GCM in that the PBL is not allowed to
exceed the first sigma level. This effectively limits the
PBL to the bottom 200 mb of the atmosphere. A more complete
description of NOGAPS is contained in Appendix A.

The extremely complete package of diabatic processes
used in the model was felt to be an important consideration
in the selection of NOGAPS for this experiment. The full
parameterization of both the PBL and the cumulus convection
was necessary for the effects of the changing SST to be felt
in the rest of the model. Without full diabatics, the response
to the SST would have been diminished. Before a fully
synchronous coupled air-ocean model becomes operational,
the boundary layer physics in both the atmosphere and the
ocean, as well as the interaction between the two, will have
to be more fully understood. This experiment is just one
step toward that goal.

B. SEA-SURFACE TEMPERATURES

The time-dependent SST used in this experiment were
obtained from FNOC. FNOC performs SST analyses twice daily

at 0000 GMT and 1200 GMT. The analysis is an integral part
of the Thermal Ocean Prediction System-Expanded Ocean Thermal
Structure (TOPS-EOTS) . TOPS is a synoptic, mixed layer fore-
cast model. EOTS is an ocean thermal analysis procedure which
uses information blending techniques to blend XBT and sur-
face ship reports to a three dimensional grid. Satellite-
derived SST reports are not presently used in the analysis.
The combined TOPS-EOTS had only been in an operational status
a few months when the NOGAPS initial conditions and SST were
captured. However, in a four-month study the TOPS-EOTS
combination had less noise in the daily analysis than the
conventional EOTS [Clancy and Pollack, 19 83J. A more com-
plete description of the TOPS-EOTS system can be found in
Appendix B .

Observed sea-surface temperatures were used in the experi-
ment to substitute a "perfect-prog" of an ocean model. A
case study with non-synchronous coupling of the models, using
a "perfect-prog" forecast of the SST, is a test to isolate
atmospheric model dependent errors [Rosmond, et al . , 1983].
The effect of variations in model-predicted SST would be
enough to cause changes in the atmospheric response. By
using the observed SST in a "perfect-prog," these oceanic
model errors could be removed and attention focused in the
atmospheric model response to changing SST. It is felt that
if an atmospheric model does not respond to the SST changes
in a "oerfect" ocean model, then it is unlikely that coupled

atmospheric-ocean models would be required on 7-10 day time
scale .

C. INITIAL CONDITIONS

The initial conditions used to make the model runs for
this experiment were captured at the time of the FNOC opera-
tional forecasts. Four different sets of initial conditions
were captured in late April and May of 19 83. Initial condi-
tions had to be captured in real time at multiple time periods
for several reasons. First, to simulate an actual operational
forecast it was necessary to obtain the initial conditions
used by FNOC at the time of the actual forecast. A delay in
obtaining the initial conditions would have the advantage of
increasing the number of observations used to determine the
initial conditions . This would have improved the initial
conditions for the model and resulted in a better forecast.
However, it would have partially destroyed the objectives of
this experiment which were to determine the sensitivity of an
operational forecast model to time-dependent SST. Second,
the NOGAPS analysis, data assimilation and prediction model
require a several day period to stabilize after the model
is restarted. This "spin-up" period allows internal gravity
waves and other imbalances to filter out of the initial
fields used. After this "spin-up" period the model forecast
fields are much smoother.

The time period was chosen to coincide with the occur-
rence of the spring transition period in the ocean. During

the transition, large changes in the SST are possible due
to the rapid shallowing of the mixed layer. However, it was
desirable to select a period before the seasonal thermocline
had become very strong. The atmospheric analyses were moni-
tored for occurrences of cyclogenesis and a storm track across
a large portion of the Pacific Ocean. If this occurred, the
increased mixing due to the increased surface wind stress
could act to mix through the incipient seasonal thermocline
and rapidly deepen the mixed layer. The increased mixing
would reduce the SST due to the entrainment of cold water
into the mixed layer. In the atmospheric model runs with the
time-dependent SST, the reduced SST should act to impede the
cyclogenesis compared to the control run.

Given these constraints, the most favorable conditions
for a model run appeared in the Pacific Ocean in late May.
The initial conditions captured were for 1800 GMT 26 May
19 83. This was a NOGAPS 6-hour update and not an actual
forecast. Since this was a full initialization for a NOGAPS
forecast, this should not have caused a problem in the
experimental forecasts .

In capturing the SST fields for this ten-day period,
the TOPS-EOTS analysis was not available from the 0000 GMT
3 June analysis to the end of the 10-day forecast at 1800
GMT 5 June 19 83. The lack of a changing SST for the last
three days of the forecast period may cause some differences
in the overall final forecast. However, three days is too

short a period for a significant effect on the model. Thus,
the major goal of being able to compare the effect of time
dependent SST on two model runs could still be obtained.

D. EXPERIMENTAL PROCEDURES

In the control run from 1800 GMT 2 6 May 19 83, the SST
were held fixed at the initial values, as is presently done
in the operational forecasts. The model was integrated to
ten days (rather than five days as is the case of the opera-
tional forecasts) with no changes in any of the input initial
fields (Fig. 2a) . A complete history tape was written every
six hours during the model run for future analysis. These
fields include the winds, heights, humidities and temperatures
for several levels. Various PBL parameters were output as
well, including the total heat flux, moisture (latent heat)
flux, sensible heat flux and long and short wave radiative
heat fluxes. Precipitation fields associated with cumulus
convection and large scale lifting were also output.

The second model run, designated the SST run, was made
using the "perfect-prog" time-dependent SST. The new SST
were input every 12 h at 0000 and 1200 GMT during the fore-
cast (Fig. 2b) . No time interpolation of the SST fields
to smooth the effect of the change was performed. The last
of the changing SST was input at 162 h and held constant
for the remainder of the integration. A similar history
tape was generated from the SST run as for the control run.

IV. HEAT FLUX ANALYSIS

The surface fluxes of sensible and latent heat in NOGAPS
are parameterized according to Deardorff (19 72) based on
the sea-surface temperature and the values of T and q from
the dynamic portion of the model. The surface heat fluxes
are then used to force the remainder of the diabatic processes
in the model. The first changes in the model response to the
time-dependent SST will be seen in the surface heat fluxes.
This chapter analyzes the SST changes during the SST run and
the resulting changes in the surface heat fluxes. The next
chapter considers the effects of these changes on the overall
synoptic pressure patterns, specifically the changes in cyclo-
genesis and storm tracks.

A. SEA- SURFACE TEMPERATURE CHANGES

The initial SST fields (Fig. 3) are from the TOPS-EOTS
analysis at 1200 GMT 26 May 19 83. This field has a predominant
north- south gradient with very little structure, except along
the coastal regions. The warmest areas of 26 °C are found
in the southwestern corner of the ocean basins while the
coldest regions of -1°C are found in the northwestern corner
of the basins .

The Pacific Ocean SST field shows the Kuroshio current
as a strong gradient along the east coast of Japan. Along
the west coast of North America, a plume of warm water extends

northward into the Gulf of Alaska. South of this feature,
the southward flowing eastern boundary current has resulted
in lower SST along the coast. The gradient is not as strong
as along the western boundary.

In the Atlantic Ocean, the predominant gradient is
oriented NW-SE over most of the ocean north of 40 °N due to
the strong influence of the Gulf Stream. The Gulf Stream is
evident as a strong gradient extending from the east coast
of the U.S. to the northeast above Great Britain. The
structure of the Gulf Stream in this analysis begins at
35°N and not in the Florida straits as expected.

The differences between the SST fields used as input to
the SST run and the initial SST were computed to determine
the horizontal variations of the changes in the SST field.
These differences were analyzed at 24-h intervals to remove
any diurnal effects. The 0000 GMT analyses for each day were
used to observe the first change in SST. However, this
selection did introduce a diurnal effect between the time-
dependent SST fields and the initial SST at 1800 GMT 26 May
19 83. It is most evident along the edges of continents,
since the diurnal surface temperature change is much greater
over continents than over the ocean surface. The changes
over the land have been shaded out in the following figures,
but the land effect can be seen as a strong gradient near
the coastal boundaries.

The first change in the SST field was inserted at six
hours of the model run. Changes in SST (Fig. 4) are generally

less than 0.5°C. In the Pacific Ocean, the changes are in
north-south bands of alternating cooling and warming regions.
The largest temperature change of 2°C is found in the region
of the Kuroshio. In the Atlantic, there is a cooling along
the east coast of North America, with changes as large as
2°C. The area of cooling extends eastward into the middle
of the basin. Large areas of lower SST are also found in the
southern part of the basin and over most of the northern area.
Changes in these two areas are generally less than 0.5°C.
Warming occurs over most of the eastern Atlantic with small
regions extending to the area north of Cuba.

The SST input at 30 h had larger departures from the
initial field (Fig. 5). Large areas of 0.5°C temperature
change can be seen as well as some areas of 1.0 °C change in
the middle of the basin. A warming trend, especially in the
Atlantic Ocean, is evident as the area of positive temperature
changes increases.

The 54-h SST change (Fig. 6) continues the trend to higher
temperatures. The area of lower temperatures in the region
of the Kuroshio and the Gulf Stream has begun to shrink.
Most of the Pacific basin has temperature differences of less
than 0.5°C. The only exception is a large area of temperature
decreases exceeding -1.0 °C in the southeastern part of the
basin. The Atlantic is also warmer. The only region of
lower temperatures is in the center of this basin.

The SST changes for 78 h (Fig. 7) , 102 h (Fig. 8) , and
12 6 h (Fig. 9) continue the warming trend that has been

occurring over most of both ocean basins. The Kuroshio area
temperatures increase rapidly during this period and has
differences as large as +2.0°C. There is a second area of
maximum temperature increases in the middle of the Pacific
basin. The area of temperature decrease in the eastern
Pacific is still present with almost the same areal coverage
of previous times. However, the magnitude of the changes is
decreasing. In the Atlantic near the Gulf Stream, a center
of temperature increase replaces most of the previous area
of temperature decrease. Maximum increases of 2.0°C are
found in this area of the Gulf Stream and in the area to the
northeast of Cuba. The largest change is at 10 2 hours, when
a 2.5°C change is analyzed in the region to the north of Cuba.
The area of small temperature decreases in the central Atlantic
Ocean remains approximately constant. At 12 6 h, this area
begins to shrink but the central value is larger in magnitude.

The final SST field analyzed was input to the model at
150 h. At this time, changes in the SST field (Fig. 10) had
resulted in a much warmer ocean surface than the initial
SST field. Most of the western Pacific Ocean has a tempera-
ture increase of at least 0.5°C with large areas over 1.5°C.
The area of temperature decrease has become smaller than
24 h previously and is now mainly located in the southeastern
part of the basin. A separate area of small temperature
decreases is also present south of the Aleutian Islands. The
Gulf Stream area has warmed significantly. Most of the area

of temperature decrease along the east coast of the U.S. has
disappeared. The largest temperature increases of 2.0°C
are found in the northern region of the Atlantic basin. Most
of the Atlantic has warmed a minimum of 0.5°C with large
areas over 1.0 °C. The area of temperature decreases in the
center of the Atlantic has remained throughout this period.

The final SST field was input at 162 h of the model run.
Thus, the changes for 150 h are representative of the changes
for the rest of the model run.

Since an area may have both positive and negative departures
relative to the initial SST during the integration, it is
difficult to summarize the SST changes. One method of estab-
lishing a trend would be to take the 150 h SST departures and
divide by 7.25 days. The method used in this study was to
take the simple sum of the seven daily SST departures from the
initial value. The sum of the 24-hour SST changes (.Fig. 11)
shows the areas of total SST increases and decreases to the
model. In general, the cumulative temperature departures
over the Pacific Ocean have positive values of 2.5°C over
much of the basin. Lower temperatures are seen along the
coasts of both continents and in a large region of the eastern
Pacific. The Atlantic Ocean has a similar change. A large
area of temperature decrease is in the center part of the
Atlantic Ocean, but the magnitude of the total change is
less than 2.5°C. Along the east coast of the U.S. is a
second area of large temperature decrease. The rest of the

basin has a temperature increase with the largest magnitudes
in the southwestern corner and in the north of the basin.

In summary, the SST departures showed a general warming
trend over much of the two ocean basins. There were also
areas of lower SST ■ s in both oceans. The magnitude of the
positive SST changes was almost double that of the negative
temperature departures. Lower SST values occurred during
the first four days along western boundaries of both oceans.
However, temperature increases occurred in this area in the
last SST input, which was maintained for the last 78 h of the
integration.

B. SURFACE HEAT FLUX CHANGES

The surface heat fluxes (sensible, latent and total) are
responsive to changes in the SST. The surface fluxes are
analyzed here to determine the direct effect that a time-
dependent SST had on the model. Subsequent changes in the
atmospheric model will depend on how efficiently these fluxes
are transported from the surface via the PBL to the free
atmosphere. In this regard, a model with higher vertical
resolution than the six-layer version of NOGAPS used in this
study will probably improve the model response to a time-
dependent SST.

The analysis consisted of determining the mean and
standard deviation for each run for the sensible, latent
and total heat fluxes. The mean and standard deviation were
computed over the 40 six-hourly calculations. The difference

of the two model runs was then determined to illustrate the
change in model response. in the following figures upward
heat flux (from the ocean to the atmosphere) is positive.
A positive number in the difference fields indicates less
heat flux was available to the atmosphere in the SST run.
1. Sensible Heat Flux

The sensible heat flux is dependent on the air-sea
temperature difference rather than just the SST. A higher
SST may not result in an increased sensible heat flux if the
air temperature also increases and results in a smaller
air-sea temperature difference. Sensible heat flux is the
smallest of the three heat fluxes analyzed here.

The sensible heat fluxes for the Pacific Ocean (Fig.
12) and the Atlantic Ocean (Fig. 13) show upward sensible
heat flux over most of the ocean basins for both model runs.
Areas of downward heat flux (indicating the air is warmer than
the ocean surface) are found along the western edge of the
Pacific Ocean extending southeastward from Kamchatka as far
as 30 °N. This southeastward extension is slightly less in
the SST run. An area of negative fluxes is also found in the
northeastern Pacific Ocean. The magnitude of the flux is
much smaller in the SST run. The only area of net downward
heat flux in the Atlantic Ocean is in the northwestern portion
of the basin which is also the area of lowest SST. The effect
of the Gulf Stream is very evident as the maximum upward
flux extends from the east coast of the U.S. to the northeast

and into the center part of the Atlantic. Fluxes in the
Atlantic are larger than 6 gm-cal/cm2-hr as opposed to the
largest flux of 3 gm-cal/cm2-hr in the Pacific.

In the Pacific Ocean, the change in the flux very
clearly follows the change in SST (Fig. 11) . Areas of higher
SST have resulted in a positive change in the flux and large
changes in the SST correlate with large changes in flux. An
exception is the area of the Pacific centered at 30 °N, 180°.
This is the area of highest SST change but the corresponding
heat flux change is not large. This suggests the air tempera-
ture had also increased and resulted in a lower air-sea
temperature difference than expected. The Atlantic Ocean
sensible heat flux difference also closely follows the changes
in SST.

For both basins, the change in sensible heat flux

was small compared to the mean values for the forecast period.

2
Largest changes of 1.5 gm-cal/cm -h in the Pacific Ocean and

2
0.5 gm-cal/cm -h in the Atlantic Ocean occurred along the east

coast of continents. This is most likely due to the higher

temperatures associated with western boundary currents than

with the rest of the ocean.

The standard deviation of the sensible heat flux for

the Pacific Ocean (Fig. 14) and the Atlantic Ocean (Fig. 15)

are of the same magnitude as the mean values. The contours

generally follow the pattern of the mean field. Largest

variations are associated with the western boundary currents.

The large variation in flux is most likely due to the
passage of cyclones along the area of largest SST gradient.
In advance of a cyclone there is zero or downward heat flux
while there are large upward fluxes behind the cold front.
The fluctuations in heat flux can be associated with the
pattern of warm advection in advance and cold advection behind
the cold front.

The difference of the standard deviations for the two
model runs shows very little change for the central and
eastern part of the oceans. A large increase in the amount
of variation in the SST run near the western boundary cur-
rents can be seen. The changes in SST are consistent with a
higher standard deviation in the SST run.
2 . Latent Heat Flux

The transfer of water vapor across the air-sea inter-
face results in a transfer of energy due to the latent heat
of evaporation required to evaporate the water. The latent
heat flux is proportional to the difference between the
saturation vapor pressure at the ocean surface and the vapor
pressure just above the ocean surface. Increasing the SST
will increase the saturation vapor pressure and the latent
heat flux will increase as a result. The energy tapped during
the evaporation process is subsequently released to the atmos-
phere through condensation of the water vapor. Latent heat
flux is the largest of the various heat transfer processes
that occur at the air-sea interface and is expected to
dominate the total surface heat flux.

The mean fields of the latent flux (Figs. 16 and 17)
show a general trend of increasing flux from the north to the
south. This pattern corresponds to the general SST field.
The highest fluxes near the western boundary currents are on
the order of 15 gm-cal/cm -h in the Pacific and 20 gm-cal/cm2-h
in the Atlantic. The effect of the Gulf Stream can be seen
more than halfway across the Atlantic and as far as 50 °N.

The differences in the latent heat flux for the two
model runs show there was an increase in the heat flux into
the atmosphere over most of both oceans. The largest in-
creases occurred in the western part of the oceans. Decreases
in heat flux (control run compared to the SST run) were asso-
ciated with decreases in the SST. These decreases occurred
in the eastern and southeastern part of the Pacific and in

the center and along part of the western boundary of the

Atlantic Ocean. The Kuroshio had changes of -3 gm-cal/cm -hr

over a small region to the east of Japan. The decrease over
the Gulf Stream region was as large as -5 gm-cal/cm -hr and
occurred over a larger area than associated with the Kuroshio.
The standard deviations of the latent heat flux (Figs.
18 and 19) are of the same general pattern and magnitude as
the mean field. As for the sensible heat flux,. this is most
likely due to the effect of the storm passages along the
highest SST gradient. There is also a latitudinal dependence
resulting from the SST distribution. The diurnal variation
of SST may account for the larger values of standard deviation

in the SST run. The resulting variation in the air-sea
temperature difference would be larger in the SST run since
both the SST and the air temperature vary.
3. Total Surface Heat Flux

The total heat flux is the sum of the latent, sensible,
solar (shortwave) and back (longwave) radiation. Latent and
sensible heat fluxes have previously been discussed. The
total heat flux is expected to have a larger diurnal component
than the sensible and latent heat fluxes because of the strong
downward component during the day.

The longwave radiation is the heat energy loss by
the ocean to the atmosphere or space. The energy is propor-
tional to the fourth power of the temperature (Stefan-Boltzman
Law) . However, back radiation from the sea surface may be
absorbed by clouds or water vapor and reradiated. The effec-
tive back radiation is the net longwave radiation loss from
the sea surface. Since the SST is relatively constant, the
controlling factors are the amount of water vapor in the
atmosphere and the cloud amount.

The mean fields of the total heat flux (Figs. 20 and
21) show downward heat flux over most of the ocean. Exceptions
are found near the Kuroshio and the Gulf Stream and in a
region in the central Atlantic. This region may be an exten-
sion of the effect of the Gulf Stream. In these regions, the

upward surface heat fluxes exceed the solar flux. The largest

magnitudes of 10 gm-cal/cm -h occur near the Gulf Stream and

the central part of the two oceans.

The regions of downward heat flux are dominated by
the solar radiation term which results in a higher SST. The
start of the model runs is before sufficient warming of the
ocean has occurred for the flux of sensible and latent heat
to be large enough to balance the solar heating.

The total heat flux differences between the two model
runs are relatively small, which suggests the total heat flux
is a result of processes not strongly dependent on the SST.
Specifically, at this time of the year, the magnitude of the
solar heat flux is beginning to increase, especially in the
subtropics and lower mid-latitudes. The atmosphere is almost
transparent to the incoming solar radiation and may have a
net gain due to the decreased back radiation and sensible and
latent heat fluxes at the sea surface.

The standard deviations of the total heat flux (Figs.
22 and 23) show a very large variation about the mean heat
flux. This variation is of the same magnitude as the mean
fields, with progressive increases toward the south. The
largest values of standard deviation are found in the
Atlantic Ocean. These large values are due to the diurnal
cycle of the solar radiation, as well as the sensible and
the latent heat fluxes. The increasing variation from the
north to the south is consistent with solar radiation being
the dominant factor.

The difference between the two runs also shows the
dependence of the total heat flux on the solar radiation.
The difference is rather small , and does not appear to have

a strong latitudinal dependence as in the mean field. The
time- dependent SST near the western boundary currents
contributed to the variation in the sensible and the latent
heat fluxes, but did not have as large an effect on the
total heat flux. Thus, the mean and variation of the total
heat flux are more dependent on the solar radiation than on
a time- dependent SST.

C. SUMMARY

Three facets of the air-sea energy transfer have been
examined. The sensible, latent and the total heat flux are
a function of the SST. The surprising result is the rather
small difference between the mean total heat flux for the
SST and control model runs. The total heat flux was strongly
dependent on the solar radiation, and the changes in SST did
not significantly affect the total heat flux. The differences
between the two model runs for the sensible and the latent heat
flux fields are small, but could be attributed to the SST
change. The statistical significance of these changes can
not be determined from only one case. The differences of the
standard deviations for the sensible and the latent heat
fluxes showed the SST run had more variation in the area of
the western boundary currents. The differences between the
standard deviations of the total heat flux for the control
run and the SST run were small. This is consistent with the
diurnal variations in the solar radiation being more important
in determining the total heat flux than the SST changes. This

result is likely to be seasonally dependent and at different
locations the SST variations may become more important in the
determination of the total heat flux.

The net change in the energy available to the atmosphere
has not been determined. Both the sensible and latent heat
fluxes showed general increases with the higher SST in the
SST run. This should result in more energy in the atmosphere
although the total heat flux values did not appear to change
significantly. A question remains on the interrelations among
the various heat flux components and the resulting energy
available to the atmosphere. Further study into this question,
particularly in the tropical regions, will be necessary.

V. CYCLOGENESIS PREDICTIONS

A. INTRODUCTION

The atmospheric response of most importance to an opera-
tional forecaster is the forecast of the development and
subsequent movement of a cyclone. In a 10-15 day forecast,
the model's ability to correctly forecast cyclogenesis would
be one of the primary requirements . The two model runs were
analyzed as to skill in predicting cyclone development and
movement in relation to the actual storm.

An analysis of actual storm developments is thought to
be of greater interest than average scores such as root mean
square (RMS) height errors or Si scores. The analysis of
cyclogenesis prediction used the Systematic Error Identifica-
tion System (SEIS) presently under development at the Naval
Environmental Prediction Research Facility (NEPRF) . SEIS is
designed to track individual low storms in either the sea-
level pressure or the 500 mb height field. The SEIS program
uses the analyzed and the forecast fields to determine various
storm parameters, such as the position, the central pressure,
the shape and the radius of the storm. These parameters are
compared to determine forecast errors. SEIS was modified for
this research to intercompare the two model forecasts runs
as well as the model run with the analyzed storm parameters.
The analysis used the 0000 and 1200 GMT sea-level pressure

fields prepared by FNOC. A more complete description of SEIS
can be found in Appendix C.

In addition to the SEIS-derived storm parameters, the
average SST and total surface heat flux following the storm
center were examined to illustrate the differences caused by
changing SST. A simple average over nine grid points
centered on the storm was computed at 12 h intervals.

Seven storms were identified by the SEIS program during
the 10-day model run. Four of these storms were in the
Pacific Ocean and three in the Atlantic Ocean. These storms
are identified with a letter for the ocean in which they
developed and a number that indicates the sequence in which
they developed. For example, storm Al indicates the first
storm developed in the Atlantic Ocean. In the next section,
the tracks of the individual storms and the differences in
movement of the predicted storms and the analyzed storms are
described. The following section summarizes the forecast
errors and' the differences between the storms forecast by
the two model runs.

B . STORM TRACKS

The storms of most significance were the ones that
developed late in the forecast period, because these storms
will have been exposed to the time-dependent SST for the
longest time. Storms P4, A2 and A3 developed over the open
ocean after the 90 h forecast time. If the model response,
specifically in cyclogenesis, is going to be different as a

result of the time-dependent SST, it should be observed in
these three storms. The fact that NOGAPS was able to develop
and maintain storms beyond five days forecast time is signi-
ficant in itself, since such a capability is essential to
extend operational atmospheric model runs into the medium-
range time period.

Storm P4 developed at 90 h over Asia and continued until
the end of the forecast. The SST model run developed this
storm 12 h earlier. In all three cases the storm was very
weak and extended over considerable distance. The SEIS
program had difficulty fitting a regular pattern to this
feature which resulted in the large variability in the early
storm tracks (Fig. 24) . The storm began to organize and
develop between 126 h and 138 h when it first crossed from
the Asian continent into the Pacific Ocean. After 138 h,
the forecast tracks were to the south and lag the actual storm
track, which indicated the model was slow in the movement of
this storm. The actual storm traveled to the east and stalled
at 160°E, 42°N. The forecast storms moved on a northeasterly
track and never stalled as in the analysis. The track for
the SST run was marginally closer than the control run to
the actual track position over the ocean.

The storm parameters (Fig. 25) also show the large varia-
bility in the early part of the storm's life and the organizing
effect of the ocean surface heat flux at 126 h. The organiz-
ing effect resulted in the storm being much better defined,

the SEIS program being able to better fit the storm patterns
and the storm tracks becoming more consistent. Starting at
138 h, there is little difference in the SST or the total
heat flux for the two model runs. The sizes of forecast
storms were considerably smaller than the analyzed storm, but
there is little difference between the two forecasts. The
central pressures in both forecasts are higher than the actual
storm, with the control run being closer to the analysis.

Storm A2 was first processed by SEIS at 102 h (Fig. 26} .
It began as a small low over the middle Atlantic states and
traveled to the northeast. When it crossed into the Gulf of
St. Lawrence at 162 h, the additional energy available to
the storm from the surface heat flux resulted in a rapid
deepening of the storm by 10 mb in 24 h. In both model fore-
casts this storm developed over the ocean just off of Cape
Cod 24 h later. The model storm tracks parallel the actual
track, which indicates that after the initial error in the
development position, the storms were correctly moved by the
model .

The derived storm parameters (Fig. 27) show that this
storm was relatively well forecast. There are higher SST
early in the SST run but as the storm moves north the SST
differences in the two model runs are smaller. The control
run has higher heat flux until 174 h when the heat flux
becomes larger in the SST run. The central pressure of both
model runs is lower than that of the actual storm. The lower

pressure in the forecasts may be due to the additional
development since the cyclone was developed over the ocean.
The control run forecast is closer to the analysis than the
SST run forecast. There is little difference in the storm
radius of the two forecasts and both are generally smaller
than the actual storm size.

The final storm, A3, developed at 114 h in the middle
of the Atlantic Ocean (Fig. 2 8) and tracked eastward before
dying at 186 h. The two model runs handled this storm differ-
ently. The SST run forecast deepening 12 h earlier than the
actual storm, moved the center to the northeast, and maintained
it until 186 h. The control run developed this storm 24 h
later and forecast it to remain active until 174 h. The SST
run was a better forecast in terms of the forecast development
time and life cycle of this storm.

Storm parameters for storm A3 (Fig. 29) show some agree-
ment between the three storms . The SST and heat flux were
lower for the SST run. The radius of the storm for both fore-
casts was larger than that for the analysis. The forecast
central pressures were initially too low and the storm began
to fill before the actual storm which resulted in higher
pressures . The SST run was closer than the control run to
the observed behavior over most of the forecast.

The first four storms — Al, PI, P2 and P3--developed
earlier in the forecast period than those discussed. The
effect of the surface heat flux on the cyclogenesis is

normally minimal compared to dynamical effects during short
time periods. Introducing the time-dependent SST into the
model should not result in large changes in these storm
forecasts.

Storm Pi was present in the Gulf of Alaska at the initial
time. The storm track (Fig. 30) was to the southwest before
ending at 54 h, while the forecasts terminated the storm at
42 h. The agreement between the two model runs was very good,
although both were ahead of the analyzed position. Storm
parameters (Fig. 31) also show this agreement. The SST for
the control run was lower than the SST run at 30 h. The
total heat flux was initially lower for the SST run, but then
was almost the same in the two forecasts. The small changes
in these inputs resulted in only small changes in the forecast
storm. Both the radius and the central pressure were not
changed. As expected, early in the model run the time-dependent
SST did not cause a change in the model response.

During this early period, the Atlantic Ocean was dominated
by the subtropical high and no storms were in this area.
Two storms, Al and P2, developed after the 66 h into the
forecast. At the time SEIS began to track Storm P2 , it was
already in a mature stage and began to occlude and fill.
Storm P2 was followed until 126 h (Fig. 32) . During th-is time
it was almost stationary in the Bering Sea. The storm centers
in the two model runs were to the southeast of its observed
location. The track for the control run began to turn to the

southeast while the track for the SST run moved into the
vicinity of the actual storm. The storm parameters (Fig. 33)
further indicate the SST run provided a marginally better
forecast for this storm. The SST run continued the storm
to the same time as the analysis, whereas the control run
ended the storm 24 h early. The storm radius for the SST
run was also closer to the analyzed size. However, the error
in central pressure for the SST run was greater than for the
control run. The storm center in the control run remained
farther to the south over warmer water. The total heat flux
at the storm center in the control run was larger, which is
consistent with the lower central pressure.

As storm Al developed over Scandinavia (Fig. 34) , the
time-dependent SST should not have a large effect on the
cyclogenesis . The storm traveled to the northeast, while the
two model run tracks were toward the south. The storm
parameters (Fig. 35) show the SST run had less error. The
storm in the SST run remained active for as long as the
analyzed storm. The radius for the SST run was also closer
to the actual storm radius . The magnitude of the central
pressure error was about the same, although it was too low
in the control run and too high in the SST run. This is most
likely due to the storm in the control run being located over
the Baltic Sea, while the storm in the SST run was over land.
The surface temperature and heat flux reflect this difference
in location.

The next storm that developed was storm P3 (Fig. 37) at
90 h of the forecast. This storm in the Gulf of Alaska
moved to the north across Alaska. The two forecast storm
tracks seem close to this path, but with very different
timing (Fig. 38) . The control run forecast storm development
24 h early and dissipation 12 h after the storm actually
started. The SST run began the storm at the same time as
the analysis but had predicted the end of the storm 36 h
too soon. Comparisons of central pressures, storm radius,
SST and heat flux (Fig. 38) are difficult due to the differ-
ent periods these storms were active. In general, the fore-
cast central pressure was too high, with the SST run having
greater errors than the control run. The SST for the control
run was higher and the heat flux was higher in the early
stages of the storm, which would account for the lower central
pressure.

C. FORECAST DIFFERENCES

The forecast position, intensity and movement of a cyclone
are among the most important parameters an operational
meteorologist can use to determine future weather conditions
for an area. These parameters were objectively determined
by the SEIS program for each period a storm was active. Dif-
ferences in these parameters were then summarized to determine
changes in the cyclogenesis as forecast by the two model runs.
Position differences are described in terms of latitude and
longitude and the total distance between two positions. The

latitude and longitude difference can also be used to deter-
mine if there is a consistent error in the forecast movement
of a storm. Storm intensity was determined from the central
pressure and the radius of the storm.

A summary of the differences between the two model runs
and the analysis and each of the two model runs is given in
Table 1. Means and standard deviations for each parameter
of each storm were computed. It should be noted that the
largest number of forecast times (12 h intervals) is only
eleven. Thus it is difficult to determine the statistical
significance of the average difference values.

The forecast central pressures were generally higher than
those for the analysis. The exceptions were early in the
forecast for storms Pi, Al and for storm A3 which developed
over land. Both the control run and the SST run underf ore-
cast the central pressure relative to the analysis, but the
error was smaller for the control run than for the SST run.
The central pressures of storms Al and A3 are better forecast
by the SST run. Recall that storm A3 was poorly forecast
by the control run. Storm A2 was better forecast by the con-
trol run. This storm developed over land while the model
placed it over the ocean.

The radius differences of the storm vary widely and have
large standard deviations. In almost all cases, the storm
radius forecast in the SST run was closer to the analysis
than was the control run.

The latitude differences indicate the SST run consistently
placed the storm farther to the north than the control run.
This may be due to the warming of the sea surface due to the
time-dependent SST. The longitude error shows storms in the
SST run were generally to the east of the storms in the con-
trol run. The combined latitude and longitude differences
indicate that the storms in the SST run had moved faster than
in the control run. Distance differences in the position of
the storm for the two runs vary. The most important result
is for storm P4 . This storm had a small distance difference
even though it had a long life cycle and was present in the
later stages of the forecast. Storms A2 and A3 also had
small differences in position. When the two model run posi-
tions are compared with the analysis, the positions from the
control run forecasts are better. However, it is difficult
to determine the significance of this result due to the large
standard deviations.

D. SUMMARY

Seven storms were identified during the 10-day forecast
period and objectively analyzed using SEIS . Storms Al and
Pi appeared in the early forecast period before the effect
of the time-dependent SST was large. Storm P3 was generally
over Alaska and the effect of a time-dependent SST was also
small. The remaining storms P2, P4, A2 and A3 were the storms
most affected by the time-dependent SST, although the changes
are not very large. In general, the central pressure

forecast by the control run was closer to the analysis than
was the SST run forecast. The positions of the storm for
the control run were also closer to the analyzed position.

The biggest improvement in the SST run was that the life
of the storm was closer to the actual life of the storm.
This change was most apparent for storm A2 , but could be
seen for other storms, particularly Al and P2 .

-p

•H

5-1

-P

5-1

•H

-P

•

•H

-P

•H

u X

CD
T3

•

T3 CO

5-1 -H

•H

<TJ CO

CO T3 >i

>iCH

03 H

fO (d

-P G

CO (d

CD CD

-P

o x: x:

CO x:

-P 4->

rH O

4->

-H

CO 5-1

W TJ

■H 0

J G

OQ -H

5-1 T3

M-l

c tu

Eh Eh

0 H

U 03

CO CO

•

a> s

> -C

-p -p

■p

Q) TJ T3

-P

c c

(d a;

(13

•H

iTJ T3

O. G

•H

5-1 rH

(d jG

-p

•H

•

<JJ

-P

«H

•H

-j

■p

•H

-H

CN H H

r-» vo

CTi CTi

CN <Ti

• • •

vo • •

CTi • •

•H CO

• • CO

00 rH CN

• rH 00

• co co

^J* • •

CN in •

r» "<3* >H

O O ^

CN CO O

• vd in

o <h r»

rH CN CN

kO H H

00 rH CN

o ^r m

CN rH CO

roro co

f^ 00 rH

in cri co

rH VD cn

vo co rr

• • •

r» VD rH

«* 00 rH

r>CM CT\

<y\ cti ^

CN •^ <J\

CN O CN

cn cn r-»

H rH in

H vd r^

^ o r*»

(NCOT

oo oo

h in m

H "=r co

e'-
en r» •

oo O cn

CN o <-)

CO CO ^

-rr CO CN

• • m

• • ■

• • •

• • a

• • •

UJCOH

o ^ "31

CO «3< VO

O rH rH

CO VD VD

t co co

m r- cn

cn r> a\

rH 00 CTi

CT> CN a\

• • •

m co r*»

O U)fO

"Sf

o o o

rH ^ VD

1 rH rH

1 1 1

1 1

■^

• •

IX) 00 rH

O VO <T\

a\ rr co

■ •

G\ 00 C\

• • •

cn o n

H O rH

O CN CN

O <H rH

CN CO CN

rH O 00

LO LD rH

cn m r*

rH rH CN

co m r~

• • •

• • «

• • •

CN "31 oo
1 1

h t in

h m vd
i i i

o o o
1 1 1

rH ^r CN

1 1

£.

co r- co

• •

00 00 CN

• •

O <H CN

• •

CT> O CO

-P

• • •

-P

• • •

-P

« • •

-p

CO rH •

• • •

5-4

CN cn co

5-1

r^ «sO ko

5-1

co r-~ ^r

5-1

• • vD

5-1

co 'D in

o o •*

CN X) CTs

CO CN CO

cn in o

cn oo r-

-p

conn

-p

H CN rH

rH CM CN

-P

CO CO rH

-p

CN rH rH

vo r- o

vo us in

^r h in

rH vd in

cn cn in

• • •

• • a

cn t m

CN "31 rH

VD CO CN

VOf^H

rH r-~ r*

cn cn m
1 1 1

CN ID rH
0O CO

CO O H

CN H

i i

H m ^
1

r» cn i

H rH
i

1 1

cn oo ld

in co rH

m oo m

CN CO 00

o ^ 00

• • •

• • 1

• ■ m

• • •

• •

O -H CN

••

■H *3" -^r

CN CO CO

• •

O rH rH

CO CO rH

p* rr vo

N r- VO

5-1

r^ oo m

5-1

rH VD [""•

5-1

r~ o\ rr

• • •

• ■ •

■ • ■

• • •

-P

co cn o

-P

"^ rH O

•P

■H r-» <y\

-P

O rH rH

-p

H r- cn

l l

1 t 1

1 1 1

T ro>

4J rH rH

c (d td

0 G G
u < <

1 I I

Eh G Eh

CO 0 CO

CO U CO

•q* -^ VD

-P rH rH

■v>

< <

I 1

■si

r-l

-Cont
-Anal

03
G
<

Eh
CO

TT "^ ■**

-U

O < <

1 ' *

vo co -^

CO U CO

CO O CO

— 1

0 G

1 i

i
G

cn in

iH m

. VD •

CX* • •

, — .

co • in

• 00 rH

r^ vd co

vo -=r r-~

CN ^ CN

CN rH CO

»•*

co ^ r-

r-> cn vd

• • •

■ « •

•^r "^ in

(Ti cn r^

•H

HCO (N

vd in rH

(N in 15

co in

-p

c
o
u

in co

CO • •

O cn •

• vo co

• • CN

T rH rH

rH ^ rH

cn ^0 in

in 0 0

• • •

VD
rH

• •

0 cn 0
1 1

CO
CN

•a

OH^O
1 1

in cn "^

r-» w

• • •

^ 0 m

0 m^o

co cn *x>

m ^ cn

CO VX5 CO

• • •
O O rH

r* co in

in r-»

00 • •

cn cr> r»

• 0 0

•— »

O O T

0 0 co

rH ^ CO

VO ^ CN

\Sv\

-p

CO CO 0

rH r- CO

•H

• • •

rH G\ <S\

•^ ^ r^

cn r^ ^0

co rH r-

1 rH

•^

rH CO CN

1 1

** CN <J\

Mnvo

• • •

• •

CN CN CN

■■

rH rH CN

—

cn in r^

rH "=T CO

• • •

4->

CN cn in

rH rH CO

co co r^-

O rH rH

<
u

-Cont
-Anal
-Anal

1 1 1

En C Eh

05 0 cn

■ 1 ■

Eh C Eh

cn 0 cn

cn U cn

cn u cn

VI. CONCLUSIONS

The goal of this research was to determine the response
of an atmospheric forecast model to a time-dependent sea
surface temperature (SST) . The results were surprising in
that there was little change between the two model runs .
Specifically, the following conclusions were reached:

(1) The total heat flux was strongly dependent on the
solar radiation. This was indicated by the small
differences between the mean total heat flux for the
two model runs and the large standard deviations.

(2) The differences in the latent and the sensible heat
fluxes were also small but could be attributed to the
change in SST.

(3) NOGAPS has the ability to generate and maintain a
cyclone beyond five days forecast time. This is an
important ability for a numerical model to possess
before the forecast period can be extended into the
10-15 day range.

(4) Differences between the SST run and the control run
were small compared to the standard deviation of the
changes. Thus, the statistical significance and any
improvement in forecast skill cannot be determined with
confidence .

(5) In general, the control run forecast central pressure
and position of storms was better than the SST run.
The life cycles of individual storms were better fore-
cast by the SST run.

This study was limited in its application and results.

The initial conditions were chosen to be able to determine

the changes in the atmospheric model during the spring

transition in the ocean. A decrease in the SST was expected

in the wake of a cyclone which would have decreased the amount

of cyclogenesis in the SST run. However, a general increase

in SST was observed and this hypothesis could not be tested.
Additionally, SST fields were only available for the first
7.25 days of the model run. This may have reduced the size
of the differences between the two model runs.

It is recognized that one case study is not a statistically
significant sample. Additional case studies following the
same approach need to be conducted before the full impact of
a time-dependent SST on an atmospheric model can be deter-
mined. These studies should be expanded to include the
following: (1) A model with higher vertical resolution should
be used. Specifically, the nine-level version of NOGAPS which
is now available will serve to improve the propagation of
the effects of the changing surface heat fluxes into the
atmosphere. (2) The analysis should be extended to include
additional model variables and geographical areas. Cloud
patterns, precipitation and other diabatic effects need to
be examined to determine their impact on the model response.
Additionally, model changes in the equatorial and tropical
regions will need to be examined. (3) The role of fluxes
across the air-sea interface requires additional study with
the goal of improving the parameterization of the fluxes.
It is felt that the small changes in fluxes in this study
resulted in part from limitations in the parameterization
method.

APPENDIX A
THE NAVY OPERATIONAL GLOBAL ATMOSPHERIC PREDICTION SYSTEM

The Navy Operational Global Atmospheric Prediction System
(NOGAPS) used at the Fleet Numerical Oceanography Center
(FNOC) is a slightly modified version of the UCLA general
circulation model. NOGAPS has been the Navy's operational
atmospheric forecast model since August 19 82. The following
sections describe the various features of NOGAPS as used
during the experiment. The complete model has been des-
cribed by Rosmond (19 81) .

A. DYNAMICS

The dynamics of the UCLA GCM are described in detail by
Arakawa and Lamb (19 77) and are only discussed briefly here.
NOGAPS is a primitive equation model. The prognostic varia-
bles are horizontal velocity, V, temperature, T, surface
pressure, p , and specific humidity, q. Additional prognostic
variables associated with the planetary boundary layer (PBL)
will be described below. The finite difference scheme used
has a spatial resolution of 2.4° lat by 3.0° long. The
variables are staggered in the horizontal according to Arakawa
scheme C (Fig. 38) . The center grid point contains the T
value. The meridional wind component, v, is carried at points
north and south of the center point andthe zonal wind com-
ponent, u, is carried at points east and west of the center

point. The numerical differencing scheme is both energy and
enstrophy conserving.

NOGAPS uses a sigma coordinate system in the vertical
defined as:

a = (p - p±)/i\

where :

p. = 50 mb and tt = p - p. ,
i ^s *1 '

p is pressure and p is surface pressure. There are six
model layers in the vertical with the top of the model atmos-
phere at 50 mb . All prognostic variables except vertical
velocity, a, are carried at the middle of each layer.
Vertical velocity is carried at the layer interfaces (Fig. 39) .

NOGAPS uses a second order (leapfrog) time difference
scheme with a four minute time step. Model diabatics are
executed every forty minutes. A Matsuno time step is used
every fifth time step. This is used to control the compu-
tational mode and to assist in the assimilation of the diabatic
effects. In regions above 60° latitude, a special Fourier
filter is used to avoid an extremely short time step. Whereas
a simple three point filter is used equatorward of 60 deg.
This filtering reduces the amplitudes of the zonal mass flux
and pressure gradients and maintains computational stability.

B. MODEL DIABATICS

The sophisticated model diabatics contained in NOGAPS

is an important component in this experiment. This treatment

of the diabatic processes is necessary to adequately simulate

fluxes across the air-sea interface and to propagate the full

effect of these changes throughout the atmosphere. NOGAPS

directly computes the physical processes for:

dry convective adjustment
large scale precipitation
diagnosis of stratus cloud depth
mid-level convection
ground hydrology
surface friction

horizontal diffusion of momentum
radiative transfer processes
cumulus convection

1 . Planetary Boundary Layer

The planetary boundary layer (PBL) is defined as a
well mixed layer in moisture, moist static energy and momen-
tum. It is assumed to be capped by discontinuities in tempera-
ture, moisture and momentum. The PBL treatment in this model
follows Deardorff (19 72) and has been formulated for the UCLA
GCM by Randall (19 76) . It allows for interaction between the
PBL and cumulus cloud ensembles and/or a stratus cloud layer
at each grid point. Surface fluxes are determined using a
bulk Richardson number based on the values of the sea surface
temperature and the values of V, T and q from the adiabatic
portion of the model. These values are then used to predict
a new PBL depth and thre strength of the inversion jumps.

The NOGAPS PBL is constrained to remain in the bottom
sigma level of the model. This differs from the original
formulation of the UCLA GCM, in which the PBL was allowed
to pass out of this layer. An overly deep PBL can result in
serious computational problems with the model . Constraining
the PBL this way imposes a maximum depth of about 20 0 mb on
the PBL.

2 . Cumulus Parameterization

Cumulus parameterization in NOGAPS follows the scheme
of Arakawa-Schubert (19 74) as introduced into the model by
Lord (19 7 8) . In the model, cumulus clouds must have their
bases at the top of the PBL. Cloud tops can be at all sigma
levels above the PBL. Cumulus clouds are modeled as entrain-
ing plumes in which environmental air is mixed with the PBL
air from which the cloud originated. Tendencies of moisture,
temperature and momentum are diagnosed as well as the cloud
mass flux. The cloud base mass flux removes mass from the
PBL, which decreases the PBL depth. Condensation occurs at
each grid point where the air becomes supersaturated. A
moist convective adjustment procedure removes convective
instability between mid-tropospheric layers that is not
eliminated by clouds originating from the PBL.

3. Radiation

The radiation parameterization follows Katayama (19 72)
and Schlesinger (1976) . It includes both a diurnal variation
and interaction with the cloud distribution. Radiative

transfer processes for incoming solar radiation are computed.
Effects of water vapor, Rayleigh scattering by air molecules
and absorption and scattering by water droplets in clouds are
included. Reflection due to clouds is also calculated. The
model cloud cover predicted by the PBL, the cumulus parameteri-
zation and large scale precipitation interact with the long
wave radiation. The net surface heat flux is computed as a
function of the incoming solar heat flux, long wave radiation
and sensible heat flux. In the present model, this affects
only the surface temperature over bare land and ice and has
no effect on sea-surface temperature.

APPENDIX B

THERMAL OCEAN PREDICTION SYSTEM-EXPANDED
OCEAN THERMAL STRUCTURE (TOPS-EOTS)

The Navy began using the TOPS-EOTS system as the opera-
tional ocean thermal analysis and forecast system in March
19 83. The objective analysis component is a modified version
of the conventional EOTS analysis for the northern hemisphere.
The forecast component, TOPS, is a synoptic mixed layer
model .

The Expanded Ocean Thermal Structure (EOTS) [Mendenhall ,
et al . , 1978; Holl, et al . , 1979J has been the Navy's opera-
tional ocean thermal analysis system for the past several
years. It is used to objectively analyze the approximately
200 XBT and 20 00 surface ship observations reported to FNOC
in real time each day [Clancy, 19 81J. With some modification,
the conventional EOTS analysis has become the objective analy-
sis component of TOPS-EOTS. The analysis is performed on the
FNOC 63x63 hemispheric polar stereographic grid. The EOTS
analysis is performed for the Northern Hemisphere only. Due
to the small number of available subsurface temperature
profiles, a sea-surface temperature analysis only is performed
in the Southern Hemisphere.

The EOTS analysis is carried out using a Fields by Infor-
mation Blending (FIB) methodology [Holl and Mendenhall, 19 71J .
This falls into the broad category of objective analysis

known as successive corrections. Twenty-six ocean parameters
are analyzed in the upper 400 m on the vertical grid shown in
Fig. 40. Parameter one is the primary layer depth (PLD) ,
which is approximately the depth of the seasonal thermocline.
The remaining parameters are temperatures and vertical tempera-
ture derivatives. Parameters 2-8 are analyzed at floating
levels defined relative to the PLD and parameters 9-26 are
associated with fixed levels.

The first guess field is the previous 24 hour TOPS fore-
cast. The first guess field is horizontally blended, with
the observations available for each of the eighteen fixed
levels parameters. The analysis is performed over a three-
cycle assimilation using reevaluated weights at each grid
point during each cycle. The floating parameters are analyzed
in the same manner, but with the added complication of deter-
mining the PBL. Next, a vertical blending process is per-
formed. Vertical blending minimizes inconsistencies in the
vertical in a weighted least squares sense. This is completed
in one step as opposed to the three-cycle analysis used in
the horizontal blending. The sea surface temperature is
given an extremely high weight, which effectively anchors the
upper part of the thermal profile to this field.

The forecast component, which is designated as the
Thermal Ocean Prediction System (TOPS) [Clancy and Martin,
1979; Clancy, et al., 1981], is a synoptic mixed layer model
that employs the Mellor and Yamada (19 74) Level-2 turbulence

parameterization scheme. It includes advection by instan-
taneous wind drift and climatological geostrophic currents.
The horizontal grid used is the FNOC 63x63 Northern Hemis-
phere polar stereographic grid. The values of the mean
temperature, T, mean salinity, S, and mean north-south and
east-west currents, v and u, are carried at each grid point.

Additionally, the advection currents u and v are carried at
J a a

grid points displaced one-half grid length in the x- and y-

directions. The vertical grid includes 18 levels between the

surface and 500 m. The variables T, S, u, v, u and v are

a a

carried at each level. The vertical eddy fluxes and vertical

advection velocity, w , are carried at the mid-levels.

The initial conditions for the temperature fields are
provided from the EOTS analyses. An initialization algorithm
is used to match the EOTS analysis to the vertical levels
used in TOPS. Salinity is determined by interpolation of
monthly climatology. Wind velocity, surface heat flux (sensi-
ble heat, infrared radiation and latent heat) are provided
from NOGAPS every six hours and are linearly interpolated to
each time step of the model run.

APPENDIX C
THE SYSTEMATIC ERROR IDENTIFICATION SYSTEM (SEIS)

SEIS is a tool to objectively analyze numerical model
predictions and produce error statistics for use by opera-
tional forecasters. It is presently being implemented for
operational use with NOGAPS by the Navy Environmental Predic-
tion Research Facility (NEPRF) . The system has been described
by Harr et al., (1982) . SEIS operates in a quasi-Lagrangian
frame with the reference center located at the center of the
storm.

The primary algorithm within SEIS is the vortex tracking
program (VTP) after Williamson (19 81) . The purpose of VTP is
to track synoptic-scale features and produce a listing of
operationally relevant parameters following the feature.
This program allows each vortex to be examined individually
and followed in time. The parameters chosen include ampli-
tude (A), ellipticity (£), radius (R) , orientation Cot). , and
position of the feature. Amplitude is the magnitude of the
vortex central pressure relative to the zonal mean pressure.
Ellipticity is a measure of the deviation of the shape of the
storm from circular. It is computed as the square of the
ratio of the semi-major and semi -minor axes. Orientation is
the angle between the x-axis and the semi-major axis, measured
counterclockwise from the positive x-axis. Position is

specified as either the model grid position or the geographi-
cal position.

The first step in the VTP is to extract the atmospheric
low pressure systems from the sea level pressure fields.
After removing the zonal mean pressure, a series of ellipses
is fit to the vortices to determine if the low pressure sys-
tems are generating, dying, merging or splitting. The original
SLP field has now been reduced to a set of parameters des-
cribing the ellipses which define the low pressure systems in
terms of A, R, e, a and position. Each low pressure system
is assigned a unique name.

After all maps during a forecast interval have been com-
pleted, the fitted parameters are transformed to raw verifi-
cation data and raw error statistics. The raw error statistics
are differences between the forecast and verifying analysis
values of a system's parameters, A, R, e, a and position.
Additional derived errors are produced as shown in Fig. 41.
Forecast error is the distance between forecast and verifying
positions. Track error is the shortest distance between the
forecast position and the track position. Timing error is
the hourly difference between the verifying position and the
position on the verifying track closest to the forecast posi-
tion. Speed error is the difference between the distance
traveled by the forecast and verifying centers divided by
the time increment. Heading error is the angle between the
forecast and verifying positions measured from the analysis
position.

Some modifications were made to SEIS for the purposes
of this study. The VTP analysis was extended from the
normal 48 h period to 10 days for the longer model forecasts
produced in this study. Storms generated during this period
required special fitting with the analysis. Also, SEIS was
originally designed to compare the forecast field with a
verifying field. It was adapted to compare either of the two
model runs to the analyzed fields, or one model run to a
second run, or all three possibilities at the same time.

APPENDIX D
FIGURES

(a)

SST
ANALYSIS

ATMOSPHERIC

MODEL

INITIALIZATION

TO 10 DAYS

ATMOSPHERIC FORECAST

(b)

OCEAN MODEL SST FORECAST

HISTORY FILE OF SST (PERFECT PROG)

TO 10 DAYS

ATMOSPHERIC

MODEL
UTILIZATION

ATM

OSPHERIC

FORECAST

(c)

ATMOSPHERIC FORECAST

INITIALIZATION

OCEANIC FORECAST

TO 10 DAYS

Figure 1.

Methods for coupling atmospheric and oceanic
models (a) minimal feedback, (b). non-
synchronous and (c) synchronous.

(a)

INITIAL

CONDITIONS

TO 240

HOI

r . , ,1

f i

r '

1 U 1

HISTORY FILES

(b)

INITIAL

CONDITIONS

SST HISTORY PILES
LRST INPUT RT 162 HOURS

TO 240

* i

f 1

r i

r 1

HOL

.. !

r 1

f <

r i

F <

f 1

r 1

! '

r '

' 1

i i

HISTORY FILES

Figure 2. Schematic of e xperiment design for (a) control
run (b) SST run. SST fields input every twelve
hours. History files output every six hours.

Figure 3. Sea surface-temperature fields used in the

initial conditions in the model run. Contour
interval is 2°C.

120* E 130* E MO* E 150* E 160* E 170' E 180% 170% 160* » 150* M HO* N 130* M 320* H

35 x 75 H 65 H

55* k 4S* H 35* M 25* » 15* « 5* H

Figure 4 .

The difference between the SST field input at
six hours of the model run and the initial SST
field. Contour interval is 0.5°C. Thin solid
lines are higher SST, thick solid is no change
and the dashed lines are lower temperature.

|20*C130*C 110* E ISO* E :6C' E 1 70 ' E I80* m I70' h Ififl" M ISO* M MO* U :ZZ* M J20* M

Figure 5. As in Fig. 4, except for 30 h

120* C 130* C HO'C ISO* C 160* E 170* E 180* * 170* H 163* k ISC* H 110* N 130* M 120* *

35* N 75* W 55* H 55* H «' W 35' N 25* H 15* W 5*W

Figure 6. As in Fig. 4, except for 54 h

:20* c 133* e mo': isc' : ieo' : :7c' c ibo' h i7c* k ieo'* iso' ^ ho* * iso" n ;2c'h

35' M 75-' M 55'w 55* N <S* N 35% 25* • IS* « 5* N

Figure 7. As in Fig. 4, except for 78 h

20* eijo'e ho* e iso* : i6o' e i7o* e ibg*h ito" m :60* k iso' m mc* h iw' * 120' m

85* H 75* H 6S' - SS* H IS* M 35* h 25 k 15 N

Figure 8. As in Fig. 4, except for 102 h

I20*CI3C*C HQ'E ISO1 E 160* C 170* I 160% J7C* k !6C* * ISO* ^ MO* * 130* k 120* k

'H 75* M 55* h S5*h 15* m 35s H 25 k 15 k 5 k

Figure 9. As in Fig. 4, except for 126 h

!20'E130'C HO' C ISO* C 160* £ !7C' C 180* fc 170* M 160* H ISO* H HO* M 130* H 120 ' H

^8S* N 75* * 65' « 55** 45* h 35* M 2S* N J5* M 5* N

Fig. 10. As in Fig. 4, except for 150 h

!20* E 13C" C MO' C 150s C 160s C 170* C I6C' « 17C" H 160* M 153* N ;«s' M 13C* K 120*1)

35' K 75'h 55* - 55* M 15* k 35* * 25* * 15

Figure 11.

The cumulative change in the SST field summed
over 24-h intervals. Contour interval is 2.0°C

20 ' E 130* E 110* E 150* C 160* E I7Q* C 180* N 170* H 160* K ISO* M HO* M 130* W 120* H

(a)

i2C' E13C'E MO* E :50' E 163' E '.70* E 190' w !70' M 16G* * 150** MO* * 130% 120%

Figure 12.

(b)

(a) Mean sensible heat flux for the Pacific
Ocean, SST run. Contour interval is 1 gm-
cal/cm^-h. Solid lines are positive (upward)
heat flux. Dashed lines are negative (down-
ward) heat flux. (b) Differences in the
sensible heat flux, control run minus SST run
Solid lines are positive indicating less
energy available to the atmosphere in the SST
run. Contour interval is 0.5 gm-cal/cm^-h .

(a)

(b)

Figure 13. As in Fig. 12, except for the Atlantic Ocean,

I2C* E 130* C I*' C 150*C l60'E 17C* C )80* m I/O* M IB* M ISO* M HO* M IX** 120* H

(a)

12C C ISO Z MO

: :sc* : i6s" c I7C* c isc'w :7c' n :m' •< ISO* M mo* * iso' * 120' X

Figure 14 .

(b)

(a) The standard deviations of the sensible
heat flux for the Pacific Ocean in the SST
run. Contour interval is 1 gm-cal/cia2-h .

(b) The differences in the standard deviations
of the two model runs, control run minus SST
run. Contour interval is 0.5 gm-cal/cm^-h .

85* H Ti' N 65* M 55* W «" *< 35* M 25* M IS* M 5* *

(a)

85'

75* « 65* * =5' ■ IS* ■ 3S' k 25* W i5* * 5* *

(b)
Figure 15. As in Fig. 14, except for the Atlantic Ocean

L20*C130*C HO* C 150' Z 150' E 170* £ 133* M 170" M 160* h I50'm H3* M 133* H 120* K

(a)

20* [ IK' E HO* C ISO' C 163' Z lV £ IK3* H 173* H 16C* h ISC* - MO' M 130* x 120" M

Figure 16 .

(b)

As in Fig. 12, except for the latent heat
flux. (a) Contour interval is 5 gm-cal/cm^-h
(b) Contour interval is 1 gm-cal/cm2-h .

5*H 75* N 55* N 55% <5* n 35* * 25* U IS' » 5' H

(a)

V - 75* M 85* h 55'- «' H 35* M 25*- 15% 5%

Figure 17.

(b)

As in Fig. 12, except for the latent heat
flux in the Atlantic Ocean. (a) Contour
interval is 5 gm-cal/cm^-h . (b) Contour
interval is 1 gm-cal/cm^-h .

>' r in' r ,..«» «■ .««•

120 E 130 C I«'C ISD'C IM'E 170* C :«•»* 170' - 160'- !30*M 140*

M 133* M 120' M

(a)

Figure 18.

(b)

As in Fig. 14, except for the latent heat
flux. (a) Contour interval is 3 gm-cal/cm^-h
(b) Contour interval is 1 gm-cal/cm^-h .

(a)

85* M 75' k 65'- 55' M 15* k 35' H 25 ' m IS* * S' M

Figure 19 .

(b)

As in Fig. 14, except for the latent heat
flux in the Atlantic Ocean. (a) Contour
interval is 3 gm-cal/cm^-h . (b) Contour
interval is 1 am-cal/cm^-h .

;•■?*•;• •! >..•,'.,

:20' £ !30' £ 110* C ISO' C I6C* E 170* C :8C' h 170* * 160* H 150* N HO* M 130* W 12C' M

(a)

120* ciao'c no* c iso' c ia* e ito* c iao' * i70* h ire* h jm* k i«'« isd'h 120* n

Figure 20.

(b)

As in Fig. 12, except for the total heat flux

(a) Contour interval is 5 gni-cal/cm2-h .

(b) Contour interval is 1 gm-cal/cm2-h .

8S' H 75* H 55* »< 55* M 15* H 35* N 2S* ■ 15* .4 S'w

(a)

85'N TS'm 65' M 55* M «* M 35* M 25* M 15* H 5* M

(b)

Figure 21.

As in Fig. 12, except for the total heat flux
in the Atlantic Ocean. (a) Contour interval
is 5 gm-cal/cm2-h. (b) Contour interval is
1 gm-cal/cm^-h .

L20* £ 130" E MC* C I5C* C 16C* E WO* C ISC* M 170* h 160% 150* K MO* N 130* w 120" K

(a)

12:' z :3c* «: ho* c :s:' c :e:'c :?:' c :93* n :tz* * :60* * :so* h no' a J30* h J20* m

(b)

Figure 22.

As in Fig. 14, except for the total heat flux

(a) Contour interval is 5 gm-cal/cm2-h.

(b) Contour interval is 0.5 gm-cal/cm2-h .

85* N 75* H 55% 55* k 45* M 35' N 25* h 1S%

la I

85* N 75' M 65* M 55% 15*- 35% 25% J5' M 5%

(b)

Figure 23.

As in Fig. 14, except for the total heat flux
in the Atlantic Ocean. (a) Contour interval
is 5 gm-cal/cm^-h . (b) Contour interval is
1 gm-cal/cm^-h .

115° E 125° E

' 145° E

Figure 24.

Tracks for storm P4 . Solid is analysis, dashed
is control run and dotted is SST run. "x"
indicates position at 126 h. "o" indicates
position at 138 h.

U">_

78 102 126 150 174 198 222

«— 1

•

.— •

"T" i i i i 1

78 102 126 150 174 198 222

(a)

(b)

•H

■

\
\

vi • /

< '• /

•— '

^ — u

/ \.'

I.
u. - * ■

■

/
/

o -

••

«*

mm *

oj i i r i

78 102 125 150 174 198 222

a) i i i i i i r
78 102 126 150 174 198 222

(d)

Figure 25.

Storm parameters for the storm P4 . (a) SST
at storm center in °C. (b) Heat flux at storm
center in gm-cal/cm2-h . (c) Radius of storm
in km. (d) Central pressure in mb . Solid is
analysis, dashed is control run and dotted
is SST run.

85° W 75° W 65° W 55° W 45° W 35° W

Figure 26.

As in Fig. 24 except for storm A2 . "x"
indicates position at 126 h. "o" indicates
position at 162 h.

i r
102 126 150 174 198 222

102 125 150 174 198 222

o
o

U-)

S. '.'*./

\ £«N •' • /

\ k ^^\v y... ■■■:

\ / f .' \X ~ /

\/ >-■' x ' s '

,— '

V ,.-* * \ ,

/•' \ '

*' \l

LO-

• — •*'

CN.

ea-

<\J

1 1 1 1 1

CM
O

102 126 150 174 198 222

en i i i i r

102 126 150 174 198 222

Figure 27. As in Fig. 25 except for storm A2

in
in

LO
CO

45 W 35 W

25" W

15° W

Figure 28.

As in Fig. 24 except for
indicates position at 138 h.

storm A3. "x"

250 500 750 1000 1250 1500

980 990 1000 1010 1020

J ! L

i 10

15 2

OJ-

: •
/
; /

; /
. 1

cn-

1
1
1
1
1
1

i— »

'-.

Figure 29. As in Fig. 25 except for storm A3

zz.

_ . c— v.\. . n '.. /A

1 ^

a
O

■ — ■ r

V* \

i \x ;

* \ v,

1 ' T" —

170° W 160° W 150° W 140° W

130 W

Figure 30. As in Fig. 24 except for storm Pi
indicates position at 6 h .

'x'

"™T""

30
(a)

CNJ

a
o
o

^-1

~ ^ r% ?-.

*'--nn

•

co-
co

s
s
s

30
(d)

Figure 31. As in Fig. 25 except for storm Pi

170° E 180° W

Figure 32. As in Fig. 24 except for storm P2

— .

in-

s
s
/

1/7

«— •

1 1 l

90
(b)

114 138

" '

*— 1

% •* j»

>"" " -

cn-
cn

o
on

— i—

138 42

90
(d)

114 138

Figure 33. As in Fig. 25 except for storm P2 .

40° E

Figure 34. As in Fig. 24 except for storm Al

100

LO -

O-.

LO
I

(a)

102
(b)

126

Figure 35. As in Fig. 25 except for storm Al.

101

rs.

K J

. ?*

/ \

: o\a

rr v " \

- T

1 >

_j j

m T

180°1W3° WL60° W50° W140° W3Q° W

Figure 36

OAs in Fig. 24 except for storm P3
indicates position at 90 h.

'x'

102

LO-

\
\
\
\

"**-

■— •

"T"

i T~

162 66

114
(b)

138

162

Figure 37. As in Fig. 25 except for storm P3

103

2.4

Figure 38.

Horizontal distribution ^of model large-scale
prognostic variables. a which is not shown
is carried at T points. [Sandgathe, 19 81]

104

_ _ 7TCT = 0

100

200
5 400

CO
CO
LlI

cr
a.

600

800

1000

V, T,

7TCT

\V,T,

7TCT

W,T,

7TCT

W, T,

TTCT

W, T,

7TCT

\y,L

TOT =

Figure 39 . Vertical distribution of model large-scale

prognostic variables. Pressure values of sigma
levels vary with surface pressure. A surface
pressure of 1000 mb is assumed in this figure.
[Sandgathe, 19 81]

105

100

125

150
175

200

225
250

;

275

—

300

—

325

—

350

—

375

—

400

—

figure 40.

FIXED LEVELS

®
®

(22)

FLOATING LEVELS

T T' V

PLD-25 i—

PLD+50

— © ~ ®

The twenty-six TOPS-EOTS ocean thermal struc-
ture parameters. Labeled T, T ' and T" are
temperature, first vertical temperature
difference and second vertical temperature
difference, respectively. Parameters 2-8 are
associated with floating levels defined rela-
tive to parameter 1, Primary Layer Depth (PLD)
Parameters 9-26 are associated with fixed
levels. [Clancy and Pollack, 19 82]

106

Figure 41.

SEIS derived location errors measures
[Harr, et al . , 1983]

107

LIST OF REFERENCES

Arakawa, A., and V. Lamb, 1977: Computational design of the
basic dynamical processes of the UCLA general circulation
model. Methods in Computational Physics, 17, 173-265,
Academic Press, New York.

Arakawa, A., and W.H. Schubert, 19 74: Interaction of a

cumulus cloud ensemble with the large scale environment,
Part I. J. Atmos . Sci., 31, 674-701.

Arpe, K.V. , 1981: Impact of sea-surface temperature anomaly
on medium-range weather forecasts. Unpublished report,
European Centre for Medium-Range Weather Forecasts, 8
pages plus 14 figures.

Camp, N.T., and R.L. Elsberry, 1978: Oceanic thermal

response to strong atmospheric forcing II. The role of
one-dimensional processes. J. Phys . Oceanogr., £, 215-224

Clancy, R.M., and P.J. Martin, 19 79: The NORDA/FLENUMOCEANCEN
thermodynamical ocean prediction system (TOPS) : A techni-
cal description. NORDA Tech. Note 54, NORDA, NSTL
Station MS, 28 pp.

Clancy, R.M. , 1981: The Expanded Ocean Thermal Structure
(EOTS) Analysis: Description, critique and outlook.
Paper presented at Ocean Prediction Workshop. 29 April-
2 May, 19 81, Monterey, CA.

Clancy, R.M., P.J. Martin, S.A. Piacsek and K.D. Pollak,

19 81: Test and evaluation of an operationally capable
synoptic upper ocean forecast system. NORDA Tech. Note
92, NORDA NSTL Station, MS., 67 pp.

Clancy, R.M. , and K.D. Pollack, 1983: A real time Synoptic
ocean thermal analysis/forecast system. Prog. Oceanogr. ,
12_, 383-424.

Deardorff, J.W., 19 72: Parameterization of the planetary
boundary layer for use in general circulation models.
Mon. Wea. Rev., 100, 93-106.

Elsberry, R.L., and N.T. Camp, 1978: Oceanic thermal response
to strong atmospheric forcing. Part I. Characteristics
of forcing events. J. Phys. Oceanogr., 8_, 206-214.

108

Elsberry, R.L., and S.D. Raney , 19 78: Sea-surface tempera-
ture response to variations in atmospheric wind forcing.
J. Phys . Oceanogr., 8^ 881-887.

Elsberry, R.L. , R.L. Haney, R.T. Williams, R.S. Bogart, H.D.
Hamilton and E.F. Hinson, 19 82: Ocean/troposphere/
stratosphere forecast systems: a state-of-the-art review.
Technical Report CR 8204, Systems and Applied Sciences
Corporation, 5 70 Casanova Ave., Monterey, CA., 79 pp.

Harr, P. A., T.L. Tsui and L.R. Brody, 1983: Model verifica-
tion statistics tailored for the field forecaster.
Preprint volume, Seventh Conference on Numerical Weather
Prediction, Omaha, NE . , published by the American
Meteorological Society, Boston, MA., 241-246.

Holl, M.M., and B.R. Mendenhall, 19 71: Fields by information
blending, sea level pressure version. Tech. Rept. M167,
Meteorology International Inc. , 2600 Garden Road, Suite
145, Monterey CA. , 71 pp.

Holl, M.M., M.J. Cumming and B.R. Mendenhall, 19 79: The
expanded ocean thermal structure analysis system: A
development based on the fields by information blending
methodology. Tech. Rept. M241, Meteorology International
Inc., 2600 Garden Road, Suite 145, Monterey, CA. , 216 pp.

Katayama, A., 1972: A simplified scheme for computing radia-
tive transfer in the troposphere. Technical Report No.
6, Dept. of Meteorology, UCLA.

Lord, S.J., 1978: Development and observational verification
of a cumulus cloud parameterization. Ph.D. Thesis,
Dept. of Atmos. Sci., UCLA.

Mellor, G.L., and T. Yamada, 1974: A hierarchy of turbulence
closure models for planetary boundary layers. J. Atmos.
Sci. , 31, 1791-1806.

Mendenhall, B.R., M.J. Cumming and M.M. Holl, 1978: The

expanded ocean thermal structure analysis system user's
manual. Tech. Rept. M2 32, Meteorology International
Inc., 2600 Garden Road, Suite 145, Monterey, CA. , 71 pp.

Randall, D.A., 1976: The interaction of the planetary boundary
layer with large scale circulations. Ph.D. Thesis, Dept.
of Atmos. Sci., UCLA.

Rosmond, T.E., 1981: NOGAPS : Navy operational global atmos-
pheric prediction system. Preprint volume, Fifth Con-
ference on Numerical Weather Prediction, Monterey, CA. ,
published by the American Meteorological Society, Boston,
MA., 74 79.

109

Rosmond, T.E., A.L. Weinstein and S.A. Piacsek, 1983:

Coupled ocean-atmosphere modeling for 3-15 day numerical
prediction: A workshop report. NEPRF Tech. Rept. TR
8305, NEPRF, Monterey, CA. , 81 p.

Sanders, F. , and J.R. Gyakum, 19 80: Synoptic dynamic

climatology of the "bomb." Mon. Wea. Rev. , 108, 1589-1606

Sandgathe, S .A. , 19 81: A numerical study of the role of
air-sea fluxes in extratropical cyclogenesis . Ph.D.
Thesis, Dept . of Meteorology, Naval Postgraduate School,
Monterey, CA.

Schlesinger, M.E., 1976: A numerical simulation of the
general circulation of the atmospheric ozone. Ph.D.
Thesis, Dept. of Atmos. Sci., UCLA.

Williamson, D.L., 19 81: Storm track representation and
verification. Tellus, 33, 513-530.

110

INITIAL DISTRIBUTION LIST

No. Copies

1. Defense Technical Information Center 2
Cameron Station

Alexandria, VA 22 314

2. Library, Code 0142 2
Naval Postgraduate School

Monterey, CA 9 3943

3. Chairman, (Code 63Rd) 1
Department of Meteorology

Naval Postgraduate School
Monterey, CA 93943

4. Chairman, (Code 6 8Mr) 1
Department of Oceanography

Naval Postgraduate School
Monterey, CA 9 3943

5 . Director 1
Naval Oceanography Division

Naval Observatory

34th and Massachusetts Avenue, NW

Washington, D.C. 20 390

6 . Commander 1
Naval Oceanography Command

NSTL Station

Bay St. Louis, MS 39522

7. Commanding Officer 1
Naval Oceanographic Office

NSTL Station

Bay St. Louis, MS 39522

8. Commanding Officer 1
Fleet Numerical Oceanography Center

Monterey, CA 9 3943

9 . Commanding Officer 1
Naval Ocean Research and

Development Activity
NSTL Station
Bay St. Louis, MS 39522

111

10 . Commanding Officer

Naval Environmental Prediction

Research Facility-
Monterey, CA 9 3943

11. Chairman, Oceanography Department
U.S. Naval Academy

Annapolis, MD 21402

12. Chief of Naval Research
800 N. Quincy Street
Arlington, VA 22217

13. Professor R.L. Elsberry, Code 6 3Es
Department of Meteorology

Naval Postgraduate School
Monterey, CA 9 3943

14. C.S. Liou, Code 63Lu
Department of Meteorology
Naval Postgraduate School
Monterey, CA 9 3943

15. LCDR P.H. Ranelli

USS New Jersey (BB-62)

Fleet Post Office

San Francisco, CA 96688

16. LCDR S.A. Sandgathe

Joint Typhoon Warning Center
CONN AVMARI ANAS Box 12
Fleet Post Office
San Francisco, CA 9 6630

17. LT. P.J. Rovero, Code 6 3
Department of Meteorology
Naval Postgraduate School
Monterey, CA 9 39 43

18. Capt. A.R. Shaffer, Code 63
Department of Meteorology
Naval Postgraduate School
Monterey, CA 9 3943

19. Dr. T.E. Rosmond

Naval Environmental Prediction

Research Facility
Monterey, CA 9 3943

20. Dr. T.L. Tsui

Naval Environmental Prediction

Research Facility
Monterey, CA 9 39 43

112

Thesis

R2124

Ranelli

Response of an atmos-
pheric prediction model
to time-dependent sea-
surface temperatures

207551

Thesis
R212U

Ranelli

Response of an atmos-
pheric prediction model
to time- dependent sea-
surface temperatures.

Internet Archive Audio

Featured

Top

Images

Featured

Top

Software

Featured

Top

Books

Featured

Top

Video

Featured

Top

Mobile Apps

Browser Extensions

Archive-It Subscription

Save Page Now

Full text of "Response of an atmospheric prediction model to time-dependent sea-surface temperatures."

See other formats