RD-Ai52 620 SIMULATION OF A SYNCHRONOUSLY COUPLED ATMOSPHERE-OCERN ANPREDICTION MODEL(U) NAVAL POSTGRADUATE SCHOOL MIONTEREYCA P J ROVERO SEP 84

UNCLASSIFIED F/G 4/2 ML

mhhhhhhhhhhmhumhhhhhhhhhhhhumhhhmhhhmhhhhlEIIIIIIIIIIIIImmhmhhmhhmhhhuEhhhhlllhIIIl

'lii -3 22IIIIIo

1.25 .1111_14 .6

MICROCOPY RLSOLUlION I[SI CHARI

d'

0

N Monterey, California

APR4 18

THESISA

C),

er

*85 03 1 i(t04

SEZCRIT y _ASS; I A-ICN T.S PAGE (lon Date Entered)

REPORT DOCUMEN4TATION PAGE iREAD INSTRUCTIONSBEFORE COMPLETING FORM

REPCR' NmJaEQ b GOVT ACCESSION NO.; 3 RECiPiENI'S CATA-OG -4UMBER

4 Tt - E an' 5bw,! 5 SPE OF REPORT & PERIOD COVERED

a, ic of a w2ynchronously Coupled .- astr's Thesis;< 2-Pere-(cean Prediction :-odel September, 195

6PERFORMING ORG REPORT NUMBER

"7 A--.3R IS aCONTRACT OR GRANt NUMBER',)

.- :. 2s:7ua kovero

9 PER -I N7 ORS&N ZA'Z, NAME 4ND ADDRESS 10. PROGRAM E .EMENT. PROJECT, TASKAREA a WORK UNIT NUMBERS::.-, - s:radua~e Jchool

4;.r' " a: fornia 9394353 NA-- AN'-V AD)[RESS 12 REPORT DATE.---ra auate -School September 19-E14

.c If:~r~ia939L.313 NUMELER OF PAGES97

V:% NC AE '. AV 5, 5 & ADDRESS'l! different from Controlling Office) 15 SECURITY CLASS so' rhis report)

Jrnclassified

15. DEC.ASSI iCATI N DOWNGRADINGSCHEDULE

( . .n.. .. , u release; dis-ributi on unlimited.

" '- - - . S-" - S s.,; entered In Block 20, If different Irom Report)

IS*!

I-

. ......... and bc nbe

a,- r -- s, -n r

DD -0 4od3

SE U I * CL SIIC TO ', '5IA E )f n~~

Li. . . _ 3 .. . ..::o u L e.-;aties'5 . oc nub

• ::- .- .f :.z :esarc " iyne l!re the n.eed for*. -: :. ..- _ ... :c :.. e. s at ,os)hric-- modei

• ,- " :.:::, ' ar~i r.lr~e-ra ' ' -s " -"

" -. .. ."..' . .- S :L a-e u .' fin -;'

] AML 1473 -" . . . ., -. c ss. - .. .. - ' ._

4SECUIIT', CLASSTlrICAION C€"

*osS lAG! - I".s- Dot. Bt,trereo

SECuRITY CLASSIFICATION OF THIS PAGE (Whon Data RnIeOM

r ...- I . dol usl r, i r, *e,"olatiors of 1.-hourly SS.e d eoas are compared to determine

c '' - rao Io - o t ,ie t..-; de pendentCL " '. s arc, --,,alvzd, one duri. .. ovemier 19- and

• .: 0 e- e"~f " tac ime-dependen-., sn :e in te for-ecast fields of

e I xc. . precpitation wnich were physically...... ~ ~ -.. ,... . . .,:: , g rend. A.nal,1sis of _ storm

rme . ; ficarnt change of storm tracK,- C c e:..eIs . only 4 cases. :;hree of tnese

,W,_r tne rine-layer version ofano one case was forecast by the.- r..- r;£ I.. I eU [° r c st L hSr. ovember period.

*/

Ct

FAl

C

SfC jqllm L ASSICATIlO OFr

TI"S PAGE'W er. Dicta E'iteed)

.4

Apiroved for public release; distribution unlimited.

Simulation of a Synchronously CoupledAtmosphere-Ocean Prediction Model

by

Peter Joshua RoveroLieutenant, United States Nav

B.S., University of 6ichigan, 1977

SuLmitted in partial fulfillment of tnereguirements for the degree of

MAS.LR OF SCIENCE IN METEOROLOGY AND OCEAN GEAPHY

from the

NAVAL POSTGRADUATE SCHOOLSejptember 1984

Auth or: ii

A. prove2 by: -------

... ..~ ~ ~ ~ ~ ~ ~ a --U Z - -- -. -,-. -E - e

Department of 4eteorology

Zean of .ciet.ze ani er.giq.eering

4

43

ABSTRACT

The jpurpose of this research is to explore the need for

tiame-depeident sea-surface temperatures in atmospheric model

iredictions to 10 days. Six and nine-layer versions of the

Navy Operational Gloial Prediction System (NOGAPS) are used

in this study. Control forecasts were made in which the

sea-surface temperature (SST) is tixe in time. Test hind-

casts were made in which the SST was updated at each time

step of the atmospheric model usinj interpolations of

12-hourly SST analyses. The 10-day prediztions are compared

to determine any improvement or degradation due to the time-

dcper dent SST. Two cases are analyzed, one luring Nov mber

1933 arnd another during Aril 1984. Use of the time-

dt_rendent SST's resulted in significant ctanges in the fore-cast fieids of surface heat fluxes and recipitation which

Were physicaLly consistent with tne SST trend. Andlysis of

15 storm forecasts revealed sigLificant changes of storm

track, durdtion or cyciogemesis in only 4 cases. Three oftLese cases were forecast by the nine-layer version of

tiN[APS during the April eriod and one case was forecast by

the six-layer NOGAPS during the November period.

'4

TABLE OF CONTENTS

INTRCDUCTION ........ ................... 12

II. BACKGROUND .......... .................... 14

A. COUPLING SCHEMES ...... ............... 16

B. PEEVIOUS STUDIES ..... .............. 16

III. PROCEDURE ......... .................... 18

-V. HEAT FLUX ANALYSIS ................ 21

A. SEA-SURFACE TEMPERATURE CHANGES .. ....... 21

E. SURFACE HEAT FLUX CHANGES ... .......... 22

1. Sensitle Heat Flux .... ............ 22

2. Latent Heat Flux ..... ............ 23

3. Total Heat Flux ..... ............. 23

C. SUMMARY .......... ................... 24

V. PRECIPITATION ANALYSIS ..... .............. 26

A. CUAULUS PRECIPITATION .... ............ 26

5. LARGE-SCALE PRECIPITATION ... .......... 27C. SUMMARY ........ ................... 27

V7. ANALYSIS OF S'ORM TRACKS ..... ............. 28

A. S_:DFMS DUFING NOVEMIBER ..... ........... 29

. S7OMS DURING APRIL .... ............. 30

1. Storm AA3 ...... ................ 30

2. Storm AA4 ....... ................ 31

3. Storm AP3 ....... ................ 31

C. SUMMARY ........ ................... 32

V:I. CC14CLUSIONS ........ ................... 33

APPENDIX A: STOEM T.ACK DATA ..... .............. 35

A. S:JE.S DURING NOVEMBER .... ............ 35

1. Storm NA. ....... ................ 35

2. Storm NA2 ....... ................ 35

3. Storm NA3 ....... ................ 35

4. Storm NA4 ....... ................ 36

5. Storm NPI ........ ................ 36

6. Storm NP2 ....... ................ 36

7. Storm NP3 ....... ................ 37

F. STORMS DURING APRIL .... ............. 37

1. Storm AA1 ....... ................ 37

Storm AA2 ....... ............... 37

3. Storm API ....... ................ 37

4. Storm AP2 ........ ................ 38

AP??ENDIX B: FIGURES ....... .................. 39

REFEEE CES ............ ....................... 94

::J:IAL DIS1 rIBUTION LIS! ....... ................ 96

0 : . . ... . i.. .. .. .. .. .. ... ... .. ... .... .i i ?

LIST OF FIGURES

1. Methods for couvlin atmospheric and oceanic modelsSiniEaI fee bac (b) non-synocranous andc synchronous. - -- - - - - - - - - - -- - - - - - - - -

- -he difference between the SST field at day 10 of themodel run and the initial SST field for the Novembercase. Contour interval is 0.4 C. Thin solid linesare hi.?er SST, thick solid line is no change anddasetd !ines are lower SS.- ------------------------------ 40

3. 7-h difference between tie SSI field at dav 10 of themodel ru±: and the initial SST field for the AoriiCdst. Contour interval is 0.4 C. Thin solid' lines4re hij!.er SST, thick solid line is no change anduasned lines are lower SS. ------------------------------ 41

4. Filid-Wcan SST ifferences ( C) 3urin, Aril in1,) Pacific and (b) Atlantic and during ovemberin (c) iacific and J) Atlantic. Dashed lines areiinear rejression of mean SST difference. 2-- ---- 2

5. Field statistics for sensible heat flux during NovEmberfor the Atlantic region. (a) lield-mean differencein ;-cal/cm2 -h. (b) Ratio of field-mean differenceto :ield-mean control. (c) Ratio of ;eak differenceto i-eak control. (d) Ratio of pear, difference toweaL control. - ------------------------------------------- 43

6. Similar to Fig. 5, except for Pacific region. ------------ 44

7. Simar to Fig. 5, except for Atlantic during Airil. ---- 45

E. Similar to Fig. 7, except for Pacific. - 46

9. The difference between the latent heat flux fieldsof the SST and ccntrol runs at day 10 for the Novembercase. Contour interval 5 qm-cal/cm2 -h. Thin solid.ines are hijher LHF, thic solid line is no change aiALasheu -ines are lower LEF ------------------------------ 47

10. Field statistics for latent heat flux dun November in..n the Atlantic region. (a) Field-mean difference inIm-cal/c. 2 -h. ( r atio of field-mean difference to~iied-mean contrc1. I=) Ratio of peak difference to peakcontrol. (d) Ratio of peak difference to mean controi. - 48

11. Similar to Fig. 10, except in Pacific region. 49

12. Similar to Fig. 10, excej't luring April. ----------------- 50

13. £imilar to Fig. 12, except in Pacific rejion. 51

0I

14. Field statistics for total hea. flux during Noverber inthe Atlantic reqion. (a) Field-mear, differenice inm-cal/cm2 -h. 1L) Ratio of Zield-mean difference to

field-zmean contrcl. (c) datio of paK difference to peakcontrol. (d) Ratio o teak diffLtence to mean controi. - 52

1j. Siilar to Fig. 14, except in ?aclfic regio.. 53

1.. Similar to Fij. 14, except for Aprii. 54

17. Sim.,dr to Fig. 16, except for Pacific region. 551h3. 7ield statistics for cumulus precipitation during

Novenber in the Atlantic re on () Field-meandifference in cm/day. (b) .atio of ield-mean differenceto field-mean control. (c) Ratio of peak differeace topeak control. (d) Ratio of peak difference to meancontrol. ---------------------------- 6

19. Similar to Fig. 18, except in the Pacific region. 57

20. Similar to Fig. 18, except during April. 58

21. Similar to Fig. 20, except in the Pacific region. 59

Field statistics for large-scale precipitation duringNovember in the Atlantic reqion. (a) Field-meandifference in cm/day. (b) Ratio oz field-mean differ-ence to field-mean control. (:, Ratio of peak differ-ence to peak control. (d) Ratio of peaK ifference toen con ,a ontrol.o.a cnrol.----------------------------------------------6 J

23. Similar to Fig. 22, except in the Pacific region. 61

24. Similar to Fig. 22, except during April. -- 2

25. Similar to Fig. 24, except for the Pacific regio-. 63

Storm tracks for storm NA5 which formed at 72 hours.Svmbols mark 12-hour analysis and forecast positions.SLiid line is based on th analyses, dashed line isfor control forecast and dotted line is for theSST forecast. --------------------------------------------- 64

7. a) Ampiitude () angle, c) radius aLd (d) ellipticitystorm NA5 based on ana ±vss(solid), cont rol forecast

(dashed) ahd SST forecast (aotted). The x-ixis units arerLuMer of 12-hour forecasts, 1 beiig the first detectionin either the analysis or forecasts. --------------------- 55

Similar to Fi _ 26, except for storm AA3 whichformei at 84 hours. -6--------------------------------------6

Simiar to Fig. 27, except for storm AA3.--7

33. Similar to Fi . 26, except for storm AA4 w9icnI orIa*d dit 12 Oh ours. ------------------------------------- 5

(C31. 5 imilaL to Fig. 27, except for storm AA4 - -- -- -- -- -- -- -- --- 93'. Similar to Fig. 26, except foL storm AP3 4ihich

foru.ed at 60 hours. ------------------------------------- 70

33. Fimilar to Fig. 27, except for storm AP3. 71

j 34. Similar to Fig. 26, except for storm NIAI whichwas resent in the initial conditions. ------------------- 72

35. Similar to Fig. 27, except for storm NA1. - --------------- 73

3t. Similar to Fig. 26, except for storm NA2 whichwas p)resent in the initial conlitions.------------------- 74

37. Similar to Fig. 27, except for storm NA2. ---------------- 75

'o. Similar to Fig 26, except for storm NA3 whichformed at 72 Lours.- ----------------------------------------- 76

3q. Similar to Fig. 27, except for storm NA3.--------------- 77

43. Similar to Fig. 26, except for storm NA4 whichformed dt 72 'hours. - ------------------------------------ 78

41. Similar to Fig. 27, except for storm NA4. 79

42. Similar to Fig. 26, except for storm NP1 which'uas present in the initial cohiitions. ---- 0

4'. Similar to Fi3. 27, except for storm NPI. -i

4. Similar to Fig. 26, except for storm NP2 whichformed at 24 hours. ------------------------------------- 82

45. Similar to Fig. 27, except for storm NP2. 83

4u. Similar to Fig. 26, except for storm NP3 whichformed at 123 hours. ----- 34

-£7. Similar to Fig. 27, except for storm NP3. ---------------- 85

4 Z. Similar to Fig. 26, except for storm AA1 whichwas Lresent in the initial con-itins. ------------------- 36

1. -Similar to Fig. 27, except for storm AA1. ----------------- 87r). imilar to Fig. 26, except for storm AA2 which

• formic at 12 hours. - ------------------------------------ 83

TI. .tiar t Fiu. 27, except for storm AA2. ----------------

C

Z.Sialiar to Fig. 26, LXC(J't fOL StJrM AP1 wiliCliformed at 24 hours. -------------------- 9

C r3. Silrilr to Fig. -Z7, excej~t for stormn AP1.---------- 9154 iiriJar tzo Fig. 26, exce~jt for storm AP2 waich

I OiuMt3 at 48 hurs.-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- - -- -- - --- 92

f. iaiarE t3 Fig. 27, exz.ept for storm AP2.------------3

L

ACKNOWLEDGEMENTS

i.,uld like to cxp~ress ny gratitiu]r- to Profussors E.

7uriic tihis research. mr. P. Harr, Dr. T. Tsai and Dr. T

,'Osmold jrovid.>d eX jertise dlid aSSiSta n..: WI tn rur.n1iny tl-le

S-'=) ar.J NG7APS trograms which was greatly a jreciated. I

Y~l'uildso like to thank LCDE P. i . Eanelii, wiaio jzoviu-ed I',Lsistunce il. the early stages of this proje t., an d tn e4

FICet Nancrical Oceanography Ceinter and Naval Postgraduate

SL Jol., whicri provided the consideranle computer r'esource2s

4 -2 ~ in tie courscz of this research.

I. INTRODUCTION

Atmosphere and ocean interaction is recojnized as an

iportailt component of tile atmospheric circulation on time

scales of a month and longer. Air-sea interactions onsnorter time scales, such as tilose over wnih numerical

weatLer prediction (NWP) models are integrdted, are less

well understood. Operation~i atmospheric prediction molds

generally eErloy one-way ocean to air influence derived from

a tine-independent sea-surface temperature (SS1) field. Tne

yuality of forecasts from such models declines rapidly when

extended beyond five or six days. Many factors contribute

to this observed degradation: inaccurate specification of

thz initial conditions; inadequate horizontal and vertical

resolutions; inaccurate or simplistic parameterizations of

atmospheric processes; and lack of two-way interactioL

4etween the atmosphere and ocean.Sea-surface temperature changes alter the Doundary

forcing of atmospheric prediction models through the surface

fluxes of sensible heat, latent Leat and momentum. 1he heatfluxes are functions of the dir-sea temperature difference

a..d plai.etary boundary layer (PBL) conditions. These fluxes

can also affect the radiation baiince tnrougn parameteriza-

tions of cloud cover and water vapor content. -he magnitudeof the atmospheric response to changing SST's should be

related to the magnitude and locations of SST changes. Thenonlinear character of these relationships reiuires the use

of a glolal atmospheric prediction model with sophisticated

pdrameterizations of the interactions between surfaceluixes, clouds, precipitation and radiative fluxes.

_ hiz work, is the second in a series o, c se studies

desLr.e to stuiy the necessity and feasitility of coupling

atmospheric and oceanic models. The first case study

(Finclli, 1984) examined the response of 10-day atmosphcric

forecasts to oLserved SSI changes introduced through simu-

lated Lor.-synchronous coupling. Tne temperatures at the sea

surface were updated Ly replacing the initial SST's with the

Fleet Numerical Oceanography Center (FNOC) SST analyses

valid at 12-haur intervals during the forecast period. In

tLis second step, simulated syncnronous coupling is used to

introduce observed SST's into the atmospheric model. Two

cases, one eazh in the spring and autumn transition periods,

dre investigated. While a global model As used, the anal-

ysis is restricted to the norther hemisphere mid-latitude

oceanic areas and the effects of changing SST's on the

dtmospheric madel 10-day predictions.

13

II. BACKGROUND

Currcrt uperatioibal N P model3 fall far short of the

aPp.:-)ximateI, 15-day theoretical limit of atmospheric

predlctatility (Fosmond et al. 1983). Many methods are

ieing IroposLd and tested to incrementally improve existing

Ioels. These include imiroved numerical differencing

schemes, spectral model formulation, increased vertical and

horizontal resolutioi,, improved initialization technigues,

an, more accurate parameterization of diabatic processes.

:nile efforts in these and other areas will ultimately be

ruired in the process of NWP model improvement, the

present study will concentrate on the effects of diabatic

processes in NWP models.

Liabatic processes become more important in determining

atmospheric circulations at long time scales. At shorter

. time scales (1-3 days), dynamic processes seem adequate as

the primary forcing in NWP models. Dibatic processes begin

to beccme important in medium time scales (5-15 days), and

tend to dominate the long time scale (15-30 days and beyond)

forecast problem. Eecause the ocean is a major heat and

moisture source, the air-sea fluxes are very important in

modling diabatic processes.

Latent arL sensible heat fluxes across the sea surface

ace modelled as functions of air-sea temperature differ-

e:ces. Wiile Sandgathe (1981) concluded that maritimecyclogenesis reuired accurate specification of air-seafluxes, no operational model includes time-iependent SS:.

C.e European Certer for Aedium-range Weather Forecasts

(::::-F) and the UJnited States National Meteoroiogical Center

0;:1 ) ust ti.e IYC weekly global SST analysis in initializa-

tion. Te -N!OC NOGAPS model is initialized by 12-hourly SST

r

14i

a;,a 1 sN produced by the Thermodynamic Ocean Prediction andExpanded Ocean Thermal tructure Systems (TCPS-EOIS)

utscriLtid y Clancy and Pollack (1983) Without a time-

dCipendent specification of SST, the model representation of

air-sea interface fluxes and cyclojenesis is bound to dete-

riDrate wit, increasing forecast time.

Significant SST changes ca L occur during 5-15 day

j eriods. Atmospheric forcing causes large SST changes

dui ,g the spring and autumn transition periods (Camp andElslerry, 1978; Elsberry and Camp, 1978; Elsberry and Raney,

1978). Lurin 9 these periods, warm shallow oceanic mixed

layers are mechanically mixed, deepen and cool. Winter SST

changes are generally smaller than transition changes

hezause of a deeper ocean mixed layer. Summer SST changes

are smaller due to decreased mechanical mi.cing (fewer and

less intense storms). The intensity and orientation cf SST

gradients associated with western boundary currents are

inportant i. determining cyclogenasis in tht western ocean

basins (Sander s and Gyakum, 198j) . Significant SST

increases ir tropical/equatorial regions are manifest in

local enhancements in the convective cloud amounts on short

(1-2 days) time scales.

7he specification of a time-varying SST to a NWP model

would reuire a corresponding oceanic prediction model. The

inclusion of atmospheric forcing influences on SST chainjesouLd require an input to the oceanic model from the atmos-

Ai eric model. The sensitivity of Lotn models to the irput

from the other must be well understood before truly coueled

atmspheric-oceanic models become operat.ional.

15

A. COUPLING 5CHEMES

ILree methods of model couplinj dere lescrined by

Eiu~rry et al. (1982)

1. minimal feedback;

2. non-synchronous coupling; and

3. synchronous coupling.

.iniLai feedlck (Fig. 1a) is tne method used by current

ou,&ratioral m-dels, it involves siecification of SST at the

i .itial time and no time- de pendence of the SST field.

N;on-synchronous coupling (Fig. Ib) involves SST input from

r. essentially independent ocean prediction model.

SyncLronous coupling (Fig. 1c) involves concurrently running

oceanric and atmospheric models waicb provide boundary

forcing to each other.

AL understanding of the effects of more sophisticated

coupling schemes on N4P models is required before these

schemes are implemented in o~erational models. The addi-

tional degrees of freedom afforded by the new time dependen-

cies iii Lour.dary forcing can possibly degrade model forecast

ski4.i, especially that achieved tnrougri "tuning" of parame-

terized processes. Atmospheric model biases may result from

0iases in the oceanic prediction models or improper coupling

between models. A large number of model verification

studies will be required to identify these biases with any

statistically sijnificant degree of certainty.

B. PREVIOUS STUDIES

rary previous studies have concentrated on long time

scile (clilatalogical) effects of SS cranges on the atmos-

vaErE. A review of these studies is contained in Elsterry

i 1. (19%2).

. 16

0

Arpe (151) examinined the effect of artificial 1SST anona-

lies on the performance of the ECLWF modei. ic showcd mrodel

s, rnsitivity at forecast times of six days aiid increased

cY:l o-,enesis with large-scale positive SSI anomalies.

Fanelli (198) examined the effects of observed SST devia-

tions (wit! respect to tLe initial conditions) on the NOGAPS

to&f . Citar differences in model behavior were noticed

over a 10-d -,y forecast. The model euployin 9 the time-

Ljrendent SST seemed to show improved skill in predicting

mir-iatitude maritime storm lifetimes.

1

17

III. PROCEDURE

The exerimental procedure was similar to that of

raFelli (19E,4) Two versions of NO3APS were used in tnis

rtsearch. in each of two cases, two N03APS integrations to

10 ys forecast time were made. The first of the intejra-

ti0.s was the control ruL in w.ich the SST was helCd

constant. In the second integration, designated the SST

run, jiobal SST's were updated at every time step (4

inutes) of the model. The model atmospheric responses of

each forecast were analyzed for differences. The analysis

was restricted to the following fields:

a. surface pressure;

b. sea-surface temperature;

c. surface sensible heat flux;

d. surface latent heat flux;

e. surface total heat flux (includes radiation);

convective precipitation; and

g. large-scale precipitation.

7he NOGAPS model was made available by Dr. I. Posmond of

t Na ndval Environmental Prediction Research Facility

(UEP F). The general cnaracteristics of the model were

descriled by Fanelli (1984). Durinj the course of this

* Experiment, major changes were made to NOGAPS. The November

1983 case study was run witi a six-layer version of NOGAPS

vcry similar to that used by Ranelli (1984). The April 1984

case was rur. with the new nine-laver version of NOGAPS. Indilitio. to vertical resolution, tne two model versions

differed irn the method by wLich diabatic heati;g was

a:,lied, in the earlier version, tae dianatic processes

%cre calculated and applied once evt-r" 13 model time steps

18

(LJ nin.utes). TLL new version also calculates the diabatics

evtlrV 19 model time steis but applies tibs hOtii;g uiifoLi-lyat each tim. step. Consequently, the dianatic Effects will

Ic ajpiied smoothly with less shock between tie time steps

in whicl diabatic effects are and are not introduced. The

effects of time-dependernt SST will also be introduced more

sLoothly in the new NOGAPS. Since additional model levels

are included, the coupling between the planetary boundary

:ayer and the free atmosphere should also be enhanced.

Ctserved sea-surface temperatures were usei to simulate

tLa input from an ocean prediction model. The SST analyses

are produced twice daily for valid times of 0300 GMI and

123 GMI. These analyses were used to create fields of

12-iour forward difference SST change. NOGAP3 was modified

to iput these 12-hour forward difference SST changes and

apiy ai, incremental portion of the SST change at each time

step. This process simulates a synchronous coupling of

N0 .PS to a "perfect 1prog" ocean model.

The NOGAPS initial conditions used for the case studies

were identical to those used for the operational forecasts

with tne same starting times. The case starting times were:

a. 1200 G4T 31 October 1983; and

b. 1200 GMT 21 April 1984.

ie model output parameters were analyzead by three

methods. Forecast and analyzed vortices were objectively

tracked using the Systematic Errors Identification System

(SEIS) developed by Brody, et a.l. (1984) at NEPFF. SEIS is

baseC on the vortex representation system of ,illiamson

(19 1). The system tracks vortices in a 900 grid point

'window' ir. the FNOC 63x63 polar stereograpnic grid.

epa rate SZS3 runs were reluird to track Pacific and

AtLL1rtic stor s. 7ie otier uodel output fields were

a iyzt d ii. Atlantic ar, Pa ci fc 'windows' of the FNOC

19

72x144 SifieriCal jri6 which Extended trom 20N to 60N. TheSST arid coi-trol rui, forecast parameters were diferenced

C(SST-cortrol) ard Lofth catour plots and field statisticswere usCed to determine the relationsLeips betweer the fore-

cast differuences and the time-dependent SST.

02

IV. HEAT FLUX ANALYSIS

he iitroduction of time-dependent sea-surface temFera-

tures into the NOGAPS model results in immeliate changes in

tha surface fluxes of sensible and latent heat. These

fluxes in 1rOdAPS ar parameterized according to Deardorff

(1972). Because the PEL and convective parameterization

schemes ii. NOGAPS are intimately coupled, the changes in

air-sea fluxes may change the diabatic forcinj tnroaghout

t. troposphere.

A. SEA-SURFAZE TEMPEEATURE CHANGES

-n both the November and April cases, significant SST

c;.aLSes evolved luring the 10-day forecast period. The SSI

difference field (day 1D - initial conditions) for the

Ll ovef ,Lr case. is shown in Fig. 2. Larle-scale negative SST

c.anJes are evident in the western Pacific Dzean, Sed Of

Ckhotsk, easterr. Pacific Ocean, and mid-Atlantic Ocean.

gLere wele small areas of weak positive SSI change to the

cst of tht British Isles and to the east of the norrn-

easterL Ur.itul States. The SS: difference fields for the

KLril case are shown in Fig. 3. Phile both positive andrug ativt SS- changes are shown, the overall mean SST differ-

iE positive. Large positive differences are founld in

tjt Sea of Japarn, Yellow Sea, eastern Pacific and western

Atlar.ic Oceans. Negative differences (decreasing SST with

time) occur in the western Pacific (east of Japan) and north

Atlantic (south of Greenland). Thie evolution of the field

1iu'ua SS differences for eacL. ocean nasin ii both cases is

stown i:. Fi;. 4. The t3-day trends in the mean SST differ-

. rc , ar, esseitially linear. Crrelation cofficients of

21I

linear cgrt_,ssions on the field-mean SST diffe:eric2s were

letwten 0. 86 and 0. gq. Some of this trend is due to the

actual SS observations. I: addition, the TGPS-EGTS

forecast/analysis schene: includes a time-depende.t effect in

t:.e first guess and a reversion to ciimatology in data-

sl'arse areas. Each of these effects will also contribute to

a gteneral trend.

B. SURFACE HEAT FLUX CHANGES

4.s in Fa7,elili (lq84), the surface heat fluxes wereresronsive to the SST changes. However, the spatial

patterns o heat flux differences were not always related

directly to the patterns of 5S3 difference. The heat flux

difference Catterns often took the form of couplets of posi-

tive and negative differences.

1. Sensible Heat Flux

:The field mean heat flux differences did show trends

siailar to the SST differences. Tae 12-hour fields of

sunsiLle heat flux differences were extcemely noisy.

Features in the difference fields were rarely stable for 48

hours. The largest differences, both positive and negative,

were observed in the western ocean basins in the November

case an] in the Pacific Ocean south of 30 N for the April

The 2evelopment of the mean sensible heat flux

differences is shown in Figs. 5 tnrough 8. While the fieid-

L ai. di:frences reach 5-20 percent of tue control forecast

LI- En se1 si.l heat flux, :,eak values in tne difference

fih !ds exceed 50 percent of peak values in the control fore-

ca st fields. Peak differences exceed the mean control

uXCs 1,y factors of 3-15. Cozrelation coefficients of

i:ri az Ci: re !osZio1,s of tl,_e fie d-nean differences were =ro:-,

C. 1- to C.

. t rt iedt Flux

S1I Iate;t ,cat flux dif:erunce fiels showed less Sn iit. A:.L mire stability thar. those of sensible heat flux.

" direct correlation of the latent flux dif erence fields

If S: d ifference fields was riot obvious. Representativeiatent i eat flux difference fields are sLjwn in Fig. 9. in 0

t..c Novumber case, strong positive-nejative couplets devel-

oped ii. the western Atlantic rejioi,. The Pacific Occan

areAs south of 30 N exhibited the largest differences in the

A;,ril case. Analysis of the field statistics (Figs. 10

trour, 13) shows mean differences reachiing 5-12 percent of

the coi trol means. Peak values in the difference fields

reach 25-75 percent of the peak values of the corresponding

control latent heat flux fields anj ex:-eed field-mean fluxes

i. the control run by factors of 2-6. Linear regression of

ti,e field-means produced correlation coefficients of

0.75-0.88 in the November case and only 0.22-0.60 in the

A;EIi case.. The differences in latent heat fluxes after 10

days during November are at least 10 times as large as the

_::crences in sensible heat fluxes, and represent consider-

ii reductioLs in evaporation over the entire field.

43. otal Heat Flux

bCG6APS total heat flux includes the sensible and

it.;.t ":uxes lus the net radiative flux at the surface.

:aditio:. is affected by model cloudiness. Fields of

t,)ta! ;,at -iux difference tend to be correlated with fields

u idt-nt hcat flux difference. Fieil statistics (Figs. 14

through 17) snowed mean diffeLences of 3-20 perzent over the

1,.- y forecast periods. Peak differences at specific times

43 1-00 ercnt of the control peak flu.es and exceeded

co:.trol fieju-neans by factors of 2-13. Linear regressiont th :i 2l-means ir ou 1c- correla tio. coefficients of

23

S

0. 4Ls-0. c4 i1, the November case aad 3.1 -0 .46 in the Ali il

case. zsI.cially ii. the Novembtr case, these differences

atter 10 days represent significarnt snfa, es £n tnP Surack

hLat budijet terms wiich would be used to drive the ocean

moiel.

C. SUMMARY

-he introduction of time-deptndent sea-surface tempera-

turts resusated in significant field-mean sensible, latent

'11 total heat flux differences of up to 20 percent over

13-Cay forecast periods. Maxima in the flux differences

:Ctwee. the SS1 and control runs after 10 dcvs were of the

snme order of magnitude as peak values in tLe control flux

fields. Tus, the surface fluxes at individual points ma.

differ significantly from the control run when time-

-oendent SST's are specified. Stable featares in the

di ffert nce fields of latent and total heat flux were

ot served to ii.tensify and move consistently durinj tne

stcord Lalf of the forecast period. The far jest changes

wer noted in the western Atlantic in tie November case ard

ii the Pacific (south of 30 N) for the April case.

TIe forecast differen~ces in heat flaxes are of such

n] ej itude tLat they can 12 expected to chanje the nature of

tlh,_ atmosrheric (and oceanic) forecast at some future time.

Local changes in the heat fluxes coujd also be expected to

s~riously affect model-Jerived forecasts, such as tnose

ij'luct -by mod el-output statistics and systems similar to

t:.e Nav.0 Oerational Local Atmospheric Prelition System

(UOLA.PS) , since these ,roducts depend o:i the redictei

surface fluxes.

L The cna.Jes in the surface heat fluxes were greater than

tn. ose oL served - y ?,Atdei 1 (1 94). -ncee Kev differences in

t e : i Ltr !,t jrocC',u zes a ccou!.t i r the , ff er i1ces in

2 4

0 . ... ... .... . _ o _ . _ ..__

i., at lU X 0 :L VIt iOT-'-,. A 1u l I i s D :-T c~i'e w r

-it 7 ~~~~~~clu -llcryIdl(2 l tu- ti

u he nc ) -dtLcef- tF

SS-Lsadte pdii 1 vr1;Utt VS- - iU S ff r~ ~ ie

V. PRECIPITATION ANALYSIS

Char ges 'n the surface heat fluxes ir.duced by time-

d~pc rdent sea-surface temperatures can be expected to affect

convective and large-scale precipitation forecasts. A study

Ly Bosse (1984) showed that surface heat fluxes could be

related to NOGAPS large-scale precipitation fields through

layer cloud instability.

A. CUMULUS PRECIPITATION

The development of the cumulus precipitation difference

fields during tne 10-day forecasts is shown in Figs. 19

through 21. Over both the North Pacicic and io:th Atlantic,

there are 25 to 50 percent decreases in mean cumulus precip-

itation during the November forecast. In the April fore-

cast, z can cumuLus precipitation over the Pacific decreased

anmost 15 percent while cumulus precipitation over the

Atlantic increased 20 percent. Peak difference values

exceeded 100 percent of peak at some time during each of the

forecasts. The decrease in cumulus precipitation durin the

Novetmber period seems logical in view of the Jezreasiag mean

sea-surface temperatures and latent neat fluxes. However,

tie magnitude of the changes is surprisingly large.

Correlation coefficients of 0. 74 were obtained in linear

rejression of both the Atlantic and Pacific mean cumulus

iZecip itation 1differences. The magnitudes of the SS': anIlatent heat flux field-m~ean diffEcernces were much less in

ti.e April case than in the Novemoer case. _11a variance of

lot! mean SET change and mean flux change was greater in the

Airi_ case thin in the November case due to tae presence of

'1e-is Of strng ipositive and negative SST change in motL

26

0

1i! C i U Z. itatn Iurin the April forecast.

C,:t _ait:. ci

VI. ANALYSIS OF STORM TRACKS

The SEIS program (Brody, et al., 1984) was used to otjec-

tivcly analyze the position, size, share and amplitude of

low pressure systems in the control and SST forecasts and

tle analyses. The analysis was performed on surface pres-

sure only.

Ll SEIS storm parameters are defined in a waLner

similar to that of Williamson (1981). Amplitude is the

difference Letween the storm central pressure and the

.it-tuiI.Ully-averaged surface pressure. Radius is the size

of tie storu expressed in terms of tae spacing of the F '0C

Nirt'.trn Herisphere polar stereographic grid. Ellipticity

is the square root of the ratio of the semi-major and semi-

a.inor axes off the ellipse which best fits the aaalyzed storm

srle. Angle is the anige between the x axis and the storm

tlijsE major axis, measured counterclockwise from the x

axis.

ir the SEIS program, the forecast low pressure systems

A"e analyzed first, then low pressure systems in the anal-

yses, tLen forecast and analyzed systems are matcned

dczordiny to set criteria. The nominal SEIS matching

criteria are plus or minus 2.5 grid points ir, both the x and

y 1irections and plus or minus 14 hours in time. SEIS is

rDutineiy used to evaluate 60 nour forecasts arid the

r atc.irg criteria are ojtimized for this period. As d

r-sult, tne vortex tracking ai.d descri. tior. portions of SETS

Ct-rfori satisfactorily th~ru.Lout tne 10-day forecasts, butt;.j autoratic rtch'n2 jortion of S bf'5 Decoies ineffectlve

a tr frv adiys. This is iri!-arily a functio of tie overly

17 trict'v' time Latchi;.J C riteria. i .spectn of the SEIS

out- ut ir.dicated tLat st~vera. Joo t 'oru t rz matches in

2b

!"

- - -- -- -. -- --- -- - -:6~~--

the 5-10 day portion of the forecast period could be made Af

the time :atching criteria were increased to plus or minus

4P hours. aianual mAtchi i with tne standard distance

criteria and 48 hour time criteria was used in the 5-10 day

portion of the forecasts.

Labels were assigned to the matzred storm tracks for

discussion and reference purposes. Tne storm labels consist

of 3 characters. The first cnaracter is the first letter of

the forecast month (N for Novecber, A for April); the second

character is the first letter of the ocean basin (A for

Atlantic, P for Pacific); and tne third character is a

r.umeral assijned on the basis of chror-olojical order of

storm identification within a particular forecast and ocean

Lasin.

A. STORMS DURING NOVEMBER

The November control and SST forecasts were made with

;t:e six-level version of N03APS. A total of 15 storms with

iifetines greater than 36 hours were tracke! in the surface

pressure analyses. Fifteen storms were present in the

control forecast while 16 were tracked in the SST forecast.

Matches were obtained for eight of the 15 analyzed storm

tracks.

reiatively minor differences between the control ani SST

forecast storm kositions and parameters were noted. These

differences grew with time. The tracks for storm NA5, which

s:.owed the largest track and parameter differences, are

shown in Fig. 26. This storm was first identified at 72

hours in the analyses and at 96 hours in ooth forecasts.

Tie analysis and control forecast stora lifetimes of 120

hours are identical w.ile tie SS: storm lifetime was 24

hours longer. The forecast storm tracks divrgje after the

t.ird foiecast Lut renain almost -araliel. The control run

29

S

storm remA~iLs guasi-stationary north of Great Britain four

days after lirst detection, while the SST ran storm stailsfor only 12 hours in the same iEeion and tAen. proceeds to

the northeast. The cyclogenesis and early storm tracks of

Ni. 5 are in the region of maximum SST cnarge, but the late

tracks and cyciolysis positions are in the region of minimum

Sas:: chanye. The SEIS parameters for storm NA5 are shown in

Fig. 27. The forecast amTiitude and period of maximum

radius la the analyzed parameters by 48 to b) hours. -he

control ru. was more accurate than the SST run in this

particular case, although the differences are much smaller

ti,a, the departures of the forecasts from the analyses. The

NA5 forecast was the only one of the eight November storms

for whici. one forecast could be judged Lore accurate thanth e other. The descriptions, tracks and parameters for the

other seven November storms are contained in Appendix A.

B. STORMS DURING APRIL

The April control and SST forecasts were made with the

nine-layer version of NOGAPS. A total of 11 storms with

lifetimes yreater than or equal. to 36 hours were tracked in

t..e surface pressure analyses. Fourteen were tracked in the

coi.trol forecasts, while 15 were present in the SST fore-

casts. natches were obtained for seven analyzed storm

tracks. Significant differences between control and SST

forecast tracks and parameters were found for three of the

! LVeL matched tracks. Those matched tracks witii significant

&iffere:.ces are discussed here. Ihe remainder of the April

: to~n tracks are discussei in Appendix A.

1. Storr AA3

-tie tracks an SE5 parameters for storm AA3 are

_. iwr. F> LEs. and 2-. T;.is stora was fi:z3t detectel in

30

0I

tLe analyses at 84 hours and was forecast only by the SST

run at 108 hours. The forecast storm was weaker (5 niL) and

sl.orter-livel (36 vice 168 hours) than the actual storm, but

the cyclogenesis position and initial track were well

jredicted. This was the only case for which only one of the

forecast models predicted a matched storm track when the

other forecast model did not. The cyclogenesis position was

to the west of, and the storm track over, an area of maximuipositive SST change. The SST forecast was clearly superior

to the control run in this case, especially considering that

tie storm developed around the fourta day of the forecast.

2. Storm AA4

The tracks and SEIS parameters for storm AA4 are

s;.owl. in Figs. 30 and 31. Storm AA4 was forecast by both

the control and SST runs at 120 hours and was first detected

ir, the analysis 12 hours later. The cyclogenesis position

was in an area of maximum positive SST change (1.2 C). Both

forecasts overestimated the actual storm lifetime of 72

hours hy 48 hours. The forecast tracks were similar until

the sixth forecast (third day after formation), when the SST

stora began erratic ,ovement. The forecast SEIS parameters

] iLFig. 31 are similar, with significant differences in

radius and anole at the second forecast period and radius

arul elli'ticity at the eighth forecast period. The control

track forecast for AA4 was very sligntly more accurate tuan

the SST track forecast during tLe first three days. :he

differences between the forecasts again are muzr. less than

the departures of each forecast from the analvsis.

2. Storr AP3

The tracks and SEIS parameters for storm Ai3 are

shown ii, s. 32 and 33. Cyclogenesis occurred at C0 Lours

ir ioth fore'casts and tIe analysis. T ,e cv.iOeflsis dnd

31

. . ... 0__

.4

tricks were in an area of p ositiv S I c, al JL (J.4-0.8 C).

TLe control run forecast d weak storm wit!, A !ifctinm of

cnly 12 Lours whereas the SST rut. PrulictuJ d lifetime of 60

Lours. Since the analyzed storb lifetiwe . as 4 6 hours, the

SSF run more accurately predicted t:.e storz lifetime and

t;ack ir. this case. The amilitude (Fij. 33) if this rela-

ro tively weak storm was forecast ratntier dcc urdtely in the SST

run.

C. SUMMARY

S4gnificant differences in the fjrecast tracks and SEIS

[arameturs were noted in four of the 15 mitched stormtrazks. The control run was more accurate it two of the

cdses, correctiy forecasting November cyclolysis wnen the

S- rur. diL not and more closely matching an April storm

tL :k wre.n t e SSI run exhibited erratic movement. The SSTr.I". uas Lote accuratt in two Alril cases, forecasting an

:'.'L: 0 . :. SE wLici, the zontrol run. did not and

t. I tu r I Z t t ;,-e control run -zecast prema-

:..if ftr.t method of applying the

. , .i~it. z tI,. iZCEedSd vertical resolution of

, V. u - n t.e Airil runs made direct

e.:. . '. r ,t.: Ajril foreasts difficult.W....... t I , V v . A ':ij 1 r. it ad,eared that

. ., . . , i it v. to tne effe, cts of the

t. ; t V. ty trzt.Sldt-d into

. ", S

4

VII. CONCLUSIONS

-he introduction of time-dependent sea-surface tempera-

turuE into the NO3APS mod~l resulted in significant changes

i.. mean and peak surface heat fluxes and precifitation.

Forecdsts which use these fields directly, such as model-

output stdtistics, TOPS and NCLAPS-type models, may benefit

-I.E the incorporation of time-dependent sea-suirface

tFm pera tures.

Differeces between the control ani SST forecast storm

tricks and SEIS parameters were significant in a of the 15

natched storms. The April SST run, wich utilized the nine-

levei 146GAPS model, correctly predizted cyclogenesis and

cyzlone maintenance for two cases in which the control rundid not. The control run more accurately predicted a

Novenber cyclolysis *vent and one April storm rack. Only

ninor differences in forecast positions and SEIS parameters

were noted in the other 11 matchel storms. Differences in

the SEIS parameters were evident before those in storm posi-

tion. The new Line-level NOGAPS, with soihisticatic appli-

cation of diabatic forcing, seems sensitive to the effects

of time-dependent SST in a way which improves the accuracy

of some forecasts.

TLis study was limited in its ai4ication and results.

'a rt her stu lies oL the response of atmospheric prediction

.o_: s to tire-dep endent SST will b reuir ed to make

Zstjisticdily significarnt inferen-es iLout the effects of

sash. Loandary conditions on the ,uality of tnu numerical

for.casts. Additional case studies witn the analyses

+ Ytt .ud to c3ver the trop-cal and Soatnern Hemisphere ocean

va.. s;.o,.i- be conducted. Additional mod.el varianles, sucn

C cIou] ir. Ls,, P! thickness and static Liery and diahatic

33

SCUMuI Us heatinj should be arnalyzed. Furtiher attczits

sTh.Dad 1 e ILadze to verify tine Ileat f lux arid precipitationC :.r(-cdsts of amodties which in~corpocate time-depen.i.dent SST to

Q~t~Lmhemode:l sensitivity dnd biases. : he E2ftect of model

utput chcinyes on proposed long~ L ani2 odel-:3utput starnis-

.t, Ch an~d other modcl-derived f,)recasts should also bet evdiuateii.

0

0

A

(1

APPENDIX A

STORM TRACK DATA

. 11 iorecast storm tracks whicn were judged 1 0 show

1 .si:.i ficAt dif feier.ces are discussed i:. this ajpendix.

'j 5 ACra", storms which b, t r resent i:' t h forecast

!..... cconjitions or 2veloptd uLLly .n t;.e forecast periol

L~. Lot "IvtL op significant differences Letweern the control

a,;. £5: forecasts.

A. STORMS DURING NOVEMBER

1. Storm NAl

The tracks and SEIS parameters for storm NAl are

scow:. Fijs. 34 and 35. 1his storm was present in theinitial conditions of both forecasts. Only minor differ-

L setween the SS! and control forecast parameters

uevo after six to eight 12-Lour forecasts. Both forecast

t'orrs moved too slowly after 24 hours and iersisted longer

tl !. the analyzed storm.

LStorm NA2

.he tracks and parameters for storm :;, 2 are snowr, in

Ci.s. ard 37. In this case bota forecasts accuratelyIrelictk& storm lifetime while moving the storm too quickly

.cl inr a more zor.al path than,. was analyzed. iinor differ-

':,s in only two parameters, position and angle, can be

t> d the sixt; forecast period.

-tori. NA3

i.e tracks and iarameters for storm Nk3 are shown in7 lirt. 3 a.d 3 1. :.is storm formed at 72 hours into thi,

35

" .. ' - . .I. .. .• < , .. .i , i . . .... .............

-43-,our forecasts. Minor iiffencns ir the forecast storm

adrameters are evident from the timu of zyclojeiesis. Both

forecasts underestimated storm lifetime. The analyzed storm

rDovd erratically and regentrateU after the tenth forecast

period. I[either model was adLl to predict the storm

reueneratio,.

4. Storm NA4

The tracks and parameters for storm NA4 are siown ii1

iLjs. 41 and 41. This storm forr,.ed at 72 hour7, khile the

control and SST forecast parameters are similar, the control

storm lifetime was 36 hours greater than that of the SSTforecast. :be analysis aid both forecast tracks were

err at ic.

5. Storm NP1

:he tracks and parameters for storm NP1 are shown in

Figs. :u2 and 43. This Aleutian, low was present in the

initial condition of both forecasts and persisted for 7 days

in each case. A rather extreme southeastward deflection in

t.e analysis occurred between the third and fourth from the

erl positions. Differences in the SEIS forecast parameters* are apj arent after the seventh through ninth forecasts.

£niy miror forecast storm position differences are evident.

f. Storm NP2

The tracks and parameters for storm NP2 are shown in

is 44 a, d 45. This storm formed at 24 hours into the

2-.-Lour forecast p.eriod. Both forecasts underestimate

storm lifetime. No differenL-s in the forecast storu

* tracks, and oi.ly minor differences in the forecast SEiS

j-rameters, were noted.

36

[t.

i0

7. Storr P

-THe traz:ks a xLJ jarame'ters of storm Nt'3 ir, s:,cwi. i:. 0

Yij_;. 46 dli7 (47. Storm NP3 was ti, Pxtrdtr)-,cal traLlsitior.

ufL troj jcuj cvclone which was iLot ai~aiv.zed until (48 hours

i:.to the fortecast. SEIS did iLot tcitt,,js vortkex uitil 132

i.OuL.- bccause the storm size was less than 2 gr~id egisS

ot.forecasts predict cyclo genesis inr the are a of the

UC t Ia extratrojicai tzAhsition, a *Proximat__; (48 hours

Lif ore A. actaaly occurr d.

B. STORMS DURING APRIL

1. 3'torL AA1

,Le tracks arnd parameters for storm A are snowr. in,

YL+s 4c ns 9. This high latitUi,2 storM Wa S presen~t initie. initial conditiox.s of both forecasts. Despite the 129

tj 132 hour forzecast stozi lifetisa2, onlv minor lifferenrce3

4the foreczast eositio;.Z and acuieters arei notef 1. k&.alzeJ nc forecast storm' tracks are roply carallci to

t:.e cli:satic ice edge.

£torm AA2

71i.e tracks and parameters for Sto:'M AA_' are showi. in

50s.S a r, 51. This storm formed 121 iours -Lnro the fore-

cost- erio d 2Toti forec asts sefriously over-esti mated th e

,A: a 1e sto-m lifet iEe t,: of 36 iLours _,v72 to 34 nours.

hetracks and parametters for storm A'l are show. in.

I ~. .. ~ 2.TI'.is storL forzned 2-4 hoIurs irnto thc fore-t~njf.Eot:-, forecaists -,ived testorm too far in~lanrA

OV restmated th1e aotual lifet~ime of 2-4 hOlrS tj 4 "

ticC -djei .;..33 nt f t t_- dcl of tt~s f Lua

37

4A 2

-iIL~~~ tEdaF -i ramt-te:s :--r stormn AP2 are sihown, in

F I. ~~.. *wf Tis~ storL toznte at 46 hours iL the anal-

'.'i aLJ ti. forecas:tS. Both Lat5uo~sia~ tne

(lot"Idl s tor: J-1tiieo 168 iJUrS, L'y ' O LOUr7s. As sh owi. by

t te 1rawIu _: in" Fig. ~55 t f. s S is A :7Atrw.2 small Storm.At,.o u ti.e rc~ijus wa s we1 -,Leite for -7 12 - ouu r

~~r 1o0 , t>

( APPENDIX BFIGURES

t2'D oIii t j iu-c a ] o e n c m l l

:(a)fc: -Ii ac t h' I11 ,0us .

C*onps

Ali- - -- --- --

Z. --h di ffrence between the SST field at day 10 of themodel r and the initial SST field f or the Novembercase nitour interval is 0.4 C. Thin solid linesare i-i c SST, thick solid liue is no c.1-tange anddaslied lines are lower SS.

3. -fi difrec bewe -S ilda a 0o h

mo0liu- ndteiii 1STfil h pi

6 cas(I ) otu rtra S04C Ii oi ieare Lii.~ ST'tiksli il snocag n

daie -insrelowrST

6141

(a) (b)

- -

S(C) (d)

L -/

61

"---" .- -/ __..

r /

in (c) Pacific and (d) Atlantic. Das-ed lines arelinEar rerssiorn of mean SST difference.

t42

0 n , ,.n m ,=,

(a) (b)

I -j

L4

(c) (d)

0 . Field zt-itistics for sensible heat flux during Novemberfor thc Atlantic reqion. (a) Field-mean differencein -Cal/ ;M2-h. (L) Ratio of field-mean differenceto ficid-mean control. (c) Ratio :f peak differenceto peai control. (d) Ratio of peak difference tomean control.

0

43

0

I

(a) (b)

E

'-1 .-. - ~- C 1~

I

(c) (d)

C r -

0

6 - ~-/ C

~. Similat t~ Fig. 5, except for Pacific region.

I

6

6

*I

(a) (b)

/7

U - -

_ _ --- ---- -- _ _ -- - -_ . ~ ---

I(c) (d)I

I -

7. Sicilar ta Fig. 5, except for Atlantic during April.

45

(a) (b)

i A

64

RIHS

1 ,1IA

A~A

ca e -4ator ine vl 5 m- C -. Th n so i

iii~s a , hghe IHF' tic oli lin isno cang andasi~e linesare loer LHF

47=

6

(a) (b)

(c) (d)

S11j. Fie.d -;tatistics for latent heat flux durin November inin tLe Atlantic region. (a) Field-mean difierence in

n-c:alicm2-h.(b{ Patio of field-mean differenc:e toYield-rican :.ontr±. op) Patio of peak difference to aKcontrol. (d) Ratio ofpeak difference to mean contro

0

48

0I

(a)(b

(Cc) (d)

11. Similai tDL Fig. 10, except in Pacific region.

49

(a) (b)

r

S

(c) (d)

10

1.2. Scimat to Fig. 10, except during April.

5))

(a) (b)

.1

(c) (d)

04

13. SiaildI t3 Fig. 12, exz ept in Pacific rejion.

05

6i

(a) (b)

(C)

14. Fieic Statistics for total heat flux during November in

the Atlhftic rej ion.(a melJanaf difference ill:: - a!/,-M - I. atio Of field-mean difference to

le MWccl n nt r C -(c) (atio Df PeaK ditteLence to peak

cont-rol. (d) Ratio of peak difference to mean Control.

I i I

( S(a) (b)

I S

(

S

(c) (d)

( S

I S

S

1. ~ifl:>it t) Fi~. lLi, exzept in Pacific re~i~n.

I S

I S

I S

(a)(b

(C)-

16. ~ ~ ~ ~ ~ ~ ~ ~ T~ 2iia DFg 4 xetfrArl

05

(a) (h)

(c) (d)

17. Similai to Fig. 16, except for- Pacific regjion.

55

0.

(a) (b)

" - 2.-- \

Vl.~

4 7-

(c) (d)

a -

72

1 F. Field statistics for cumulus pcecipitation duringNovEmber in the Atlantic re ion. a) Field-meandifierence in cm/day. (b) atio o field-mean differenceto field-mean control (c) Fatio af peak difference topeay ccntrol. (d) Eatio of peak difference to meancont rol.

56

(a)(b

V.SimLlai to Fig. 18, except in the Pacific region.

57

(a) (b)

(c) (d)

20. Similar to Fig. 18, except during April.

6

58IS

Cb

(C (d)

21. cii id, 7 ig. 20, except iL the Pacific ragioL.

59

*1

(a)

, . " (b)

.1

- -, i

(c) (d)

/ \

.... C-.- \

- - 5;-* 22. Field statistics for large--scale prez:ipitation during

Novmber in the Atlantic reqion. (a) Field-meandiffere:xce in cm/day. (bI Ratio of field-mean differ-enc= to field-mea, control. (C) Ratio of peak differ-ence to peak control. (d) Ratio of peak aifference tomear co:trol.

06

66

, • .6,

/, (a) K (b)

I - ./

C(c) (d)

as

2. Sililar ta Fig. 22, except ir. the Pacific region.

61

-- i

(a) (b)

r

fL

(c) (d)

24. simiiar t Fig. 22, except during April.

62t"S

7.S imla -toFig 24,excpt or te P~ifi reion

( _ _ _ __ __63

IC

*26. Storm traicks for storm NA5 whi::h f orwed attau= 72 Lojurs. S ymbols mark 12-hour analysisand forecast posito Soli I line is based on theatidlyses, dashed ilnle is for coi~trol forecast anddotted line is for SST forecast.

66

£ (a) I(b)j

(C) (~d)

27 a) A T L it u 3 , WL angerdu and (d eliipticit(da.;he&) and SST forecast ( otted).

2'. Simili -.3 Fig. 26. except for storm AA3 whicnfozedl ,.it tau= 84 hours.

6 6

(a) (b)

0c

(c) I(d)

S i la: "o Fij. 27, exvept fo: storm AA3.

0

I..

0.

A

30. Similar t Fig. 26, except for st-rm AA4 whichformed at tau= 120 hours.

0

-."

ri.

[I,

" ' " 4"°' " " " " % " b 'd ' ' -- " * r K ' -' ' - -" M - ~ , -- m ---d * .. M . u ,

(a) (b)

66

( I

312. Simiiai to Fig. 26, except for storm AP3 whichformed at tau= 60 hours.

70

r--

(a)(b

44

717

._ ___-

I - -

72

C(a) (b)

(C) (d)

?.Simildi to Fig. 27, except for storm NAl.

0

73

- " -----

.'b,-

CR.

3A. Similar tj Fig. 26, except for storm NA2 whichwas pr-esent in the initial conditions.

74

. • .. o , - . . . • • , -

(a) (b)

4

Cc) (d)

. S

27. Similai to Fig. 27, except for storm NA2.

75

. . . . I. , , ., ... . ,S.. . . .. . : . -.. . . .. .

0I

FIN

3&. irilai to Fig. 26 except for storm NA3 whichf ornmed at tau= 7 2 Lours.

76

-- ----

(a) (b)

(C

Siia oUj 7 xetf0VsomN3

67

0

.78

Cz

L4. Similar t9 ?ig. 26, except for stornm NAL4 whichforred at tdU= 72 hours.

S

78

A - _ __ _ _ _ _ _ ___C )

( a r(oFgb2,ecptfrsor) A

4 -~79

42. Si.lit 2, exep fo str NP whic

II

L;:E -.7t-at in the initial conlditions.

I -

it

80i

i . , ."' ~ ~~~~~~~~~~~~~. " r " " " m ' " w ' - "

0

(a) (b)

I II I

I __________ I ___________________

0

I (c) Cd)

I0~ I

_____________________

_________________

4?. Sim±~ar t~ Fig. 27, except for storm NP1.

0

S

c~1

0

444. ~ ~ ~ ~ ~ ~ ~ ~ ~ -. > SVLa z i . 2,e c p o t r 2 hc

.4r~da au 4hus

C8

(a)(b

(C))

45. Simiir to Fig. 27, except for storm NP2.

83

rii

* K

C_&

. .. . . A

I

r4u. Similar t. Fig. 26, exct for storm 1123 which

forced ft tau = 120 hour'

I

. 84

I

-4t

(a) (b)

7/

C d

iS

P1

- A-

Cc) (d)

I --

47. Similar 0 Fig. 27, except for storm NP3.

85

. S

-.- - - - ----------..--- ~~-- p . ~ -

r .----- - -•r r ro - . - , . ,- - - w, - .

77

V...

'vl

L. 4-

--- / ' - , -..-.

-

whic

- . , ,-

(a) (b)

~i1 I

"- r

(c) (d)

I -

49. Siml.Iar to Fig. 27, except for storm AA1.

87

0 °

0I

-4-

0i

2-

r., SIi~LaU t Fig. 26 except for storm AA2 whichforr.EJ at tdu= 12 hOUrS.

0

i-

ha)

Cc) .(d)C.(C)

51. Simnilar tz) Fig. 27, except for Storm AA2.

- 39

C * *s.

rii

K--

52fiia Fg 6 xcp o tr P hc

zo e ttu 4hus

69O~

IJ

4.

(a) Cb)

C c) Cd)

--1 I

i" I

53. Siimilar tc) Fig. 27, except for storm APt.

91

I_ ' - _ . i ". .. •" , , . . , .. . .- . . . _ _ . .- . . . • L L "

re

54~. Si id -z -.j -6 xetfrstr P hc

0; ;

S9

0 iidtt i . 26 xe-tfrsoa P hc

forzced at tdu= a8 hours.

.1

92

*|

Ca) (b)

C. f (C) (d)1

EUSimiiai tz) Fig. 27, excejt for Storm AP2.

(93

C' REFERENCES

Ar pt K. V., 19 51: Im act of Sed-sUtf3Ce teM_)rLtUrE- cirnoWl.ozr L.e di u-za r. e weafher forecasts. uThp uL.i st. ed report,

~ryenCernter for Mdium-Fainge weatiher forecasts, 8 racesplu 14l~ figures.

ose,'.. 184: Estimation. of diabatic heating; for a Le OO~~e"-e~~r tnsadrit neci on&. t"I. S. 'Thesis Dept.

oi !Ieteoroo , Pva Potrduate Schaoo, Mor.terey, A.

SL~cl , L.L. , et al. 19 8L4 Atomated VerifiCdtO O.of nuiaEC-ical weater Te'dltion forecasts - A decision aid cor tiieoi eratioLal forecaster. Pre p rt oue 1tDrfcence on

.~ter Ar.a!Lvsis and Forecasting, Tamp a, FL. , i>lblisned byh e Aran 9eteorological Society, Boston, IA, 424-427.

Cal N.. and R.L. ElsLerry, 1973: JcealwiC t~ermal res~onseto trang ,"atmospheric forcing !I. The -oie Of one-

u~~1io1 processes. J. Phys. OceanDIE., 6, 2 15-224.

clancy'r F. a. r.d K. D. Pullack, 19 83: A real tiae synop'tic4 oceafl tlhCrn a dalysis/forecast system. P E 3 U Ccu-a n (-'f

1, 383-4+24.

DLEarOh - I J. W. ,1972: Paraneterizatior. c) the DiaLetarvLou:.arv la'er for use in getneral circ.ulation :on

___I. Levr 09, 93-16

_i L _Z ErV, F. . arid N. T. CadIP~ 19 78: uc::ar.ic thouna'Lrijs ,onrsei to strong atmospheric 9-orciogL .artCCharacterZistics of forcing events. J. :?hrc. 8raor, 320D6-2 14.

Biserr, f 1.~%d S.D. Faney, 1976: Sea-surfacectemp~-ra-ture resiornse vaitosin atnospneric Wind focig 3..Pi.yL;. GcearLojr., 8, 881-887.

I ls .k-;r ry , rL ., et alt 1982:e ac,/7o os~r TS1aosphere :,orecdst Sy stto Is: A State of

the Art eve. Te 1hrical Report ZR 6204, Systems and";)PLied Scien~ces Corporation, 510 Casanova Ave., M~oterey,0x. 79 .

F-n~l1 P. ii., 1934: Lesponse of an atMospIheric predictionZ.0 l o time-deperndent sea-surface te mpier atu rc-s.:.. S S, C. zt. Of Yeteorolojy, Nava.1 Postgraduate School,

0osaor~d, t.. t al., 1P33: Cou-led ocean.-u coDsi :.crcrl CiI. -o-dZ5 ay r-umerical. p r e i c:t o.i: wos'repoart. 'NE F 11F Tecin. F'ept. TI- 8305, NTEPF, uFe, CA,

f- i r a, ., en r., s . 113 umI 1980: SV 110 ti i %:,,Aa LL1C C :Idtoloy ofte "omb. lr. a.&v,1, 8 -6.

~ d~c .. , 191 unIricd! stud%, of th1.e r:o e ofdai: Cu a :1U X CS if;. extratrod ca cCd.ueess C, D.

i, De, t. of 1t'zjJ~,;LvIpst iUuate Sc)hou,

r

4

~i11iarnson, ~ L 1981: StorL tracK re;zesentatior. andv~rification.~eI1us, 33, 913-533.

I

(

I

C.

I

I

I

I

I

RD-Ai52 828 SIMULATION OF A SYNCHRONOUSLY COUPLED ATMOSPHERE-OCEAN 2/2PREDICTION MODEL(U) NAVAL POSTGRADUATE SCHOOL MONTEREYCA P J ROVERO SEP 84

UNCLASSIFIED F/G 4/2 NLEu..l

1111 1.0 5 L~ 11112-4 IIII II Jf 11111o 11111i== 2.2

1J111L 11161

MICROC OPY RESOLUTION tST CHARI

,o . .. . .-

INITIAL DISTRIBUTION LIST

No. Copies

1. Defense Technical Information Center 2Cameron StationAlexandria, VA 22314

2. Library, Code 0142 2Naval Postgraduate SchoolMonterey, CA 93943

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

*Naval Postgraduate SchoolMonterey, CA 93943

4. Chairman, (Code 68Mr)Department of OceanographyNaval Postgraduate SchoolMonterey, CA 93943

5. DirectorNaval Oceanography DivisionNaval Observatory34th ard Massachusetts Avenue, NWWashingtoa, D.C. 20390

6. CC¢aanderNaval Oceanography CommandNIE StationBay St. Louis, MS 39522

7. Ccmmanding OfficerNaval Oceanographic OfficeNSTI StationBay St. Louis, NS 39522

8. Ccmmanding OfficerFleet Numerical Oceanography ZenterMonterey, CA 93943

g. Ccmmanding OfficerNaval Ocean Research and

Develonment ActivityNSTI StationBay St. Louis, MS 39522

10. Ccmmanding OfficerNaval Environmental PredictionResearzh Facility

Monterey, CA 93943

11. Chairman, Oceanography DepartmentU.S. Naval Academy

0 Annapolis, MD 21402

12. Chief of 14aval Research800 N. Quincy StreetArlington, V 22217

96

13. Professor R.L. Elsberry, Code63Es 5Department of MeteorologyNaval Postg'~raduate SchoolMoi.terey, CA 939437

14. C.S. Liou Code 63Lu1Department. of MeteorologyNaval Postgqraduate SchoolMonterey, CA 93943

15. LCDE P.11. RanelliUSS New Jersey EB-62FPO San Francisco 96688-1110

16. LT P.J. Rovero 2Naval Oceanography Command DetachmentU.3. Naval Air StationFPO New York 09523

17. Dr. T.E. RosmondINaval Environmental PredictionEesearch Facility

Moniterey, CA 93943

18. Dr. T.L. Tsui1Naval Environmental Prediction

Research FacilityMonterey, CA 9 3943

19. Mr. Patrick Harr1Naval Environmental Prediction

iEesearzh FacilityMonterey, CA 93943

97

DTIC