AnnaLs rif Glaciology 25 1997 Cl;; Internationa l Glaciologica l Society

Arctic precipitation as represented in the NCEPjNCAR reanalysis

MARK C. SERREZE,I JAMES A. MASLANIK2

ICoopemtive Ins/itll/ejor Research in Environmen/al SCIences, Division rifO)'osjJheric and PoLar Processes. CamjJZlS Box 449, University rifColomdo, Bouldel; CO 80309 -0449, USA.

2Colomdo CenlerJor Aslrod)mamlcs Research, Campus Box 431, Universi(y rifColomdo, BOllLder, CO 80309 -0..)31, USA.

ABSTRACT. Arctic prec ipita tion as depicted in the l\'a tiona l Center for Environ­mental Prediction (l\CEP) and the National Center for Atmospheric Research (NCAR) reana lysis effort is evaluated using 6 hourly model output for the period 1986- 93 in con­junction with ga uge-corrected climatologies. Climatologica l field s fi'om the model agree favo rably with observations in terms of general spatio-tempora l pa tterns, but with some notable diITerences. In pa rticul a r, the precipitat ion max imum over the centra l Arctic (the region north of 70° N ) is depicted inJuly, onc month too ea rly. Values arc too low from August through D ecember, resulting in underes timates of annua l precipitation of about 40 mm. Despite these shortcomings, the modeled precipita tion fi elds appear to be suffi­ciently reali stic to represent a base for blending with other data to prov ide g ridded fi eld s suitable for use in climate studies and sea-ice models.

INTRODUCTION

The Arctic sea-ice cover requires maintenance of a low-sa linity surface layer. Surface runoIT provides the largest single­ocean fresh-water source of about 35 cm a I (i\agaa rd and Carmack, 1989; Iva nov and Yankina, 1991). Following fresh­water import through the Bering Strait, precipitation less evaporation (P - E ) over the Arctic O cean itself provides the next la rges t contribution of \6- \7 cm a I (Serreze and others, \995). C hanges in precipitation over watersheds draining into the Arctic O cean and over the Arctic O cean itself may impact on sea-ice thickness and ice a rea, inOuen­cing ocean-to-atmosphere heat and Ill oisture Ou xes and, potentia ll y, cloud cover and cyclone activit y. The 0 ux of sea ice through Fram Strait a nd the Canadia n Arctic Archi pe­lago influences the oceanic convective regime, impacting on the g loba l thermoha line circulat ion (Za ucker and others, 1994). Arctic ma rine life is organized in to compl ex food webs, conditioned by sea ice, nutrient a\'ailability and water densit y. Changes in these factors may i m pac t on mari ne ecosys tems and the biochemical cycling of essenti a l nutri­ents. Changes in the terrestr ia l hydrologic cycle may a lso alter soil moisture, influencing plant cOlllmunities and their grazers. Arctic soil s and peatl and s serve as so urces and sinks of atmospheric trace gases, and appear to respond sen­sitively to changes in soil moisture and temperature (McCauley and Meier, 1991).

Addressing these issues requires improved databases of precipitation for model input and validation, as well as for di ag nostic climate studies. Prec ipitat ion data from Arctic land stations (Vose and others, 1992) suffer from severe under catchment in areas of blowing snow (\ Voo and others, 1983). Even in summer, estimates may be low due to neglect of occult precipitation (fog, dew) over tundra a reas (Ding­man a nd others, 1980), trace val ues a nd wett ing losses. As

most Arctic stations are at low altitudes, the tendency for

prec ipitation to increase with elevation also implies an underes timate of regional averages when stat ion data a re used. It is possible to adjust station data for exposure, winds and gauge type, but the network is still relatively sparse. Databases for the Arctic O cean a re prim arily for long-term mean a nnu a l and monthl y va lues (e.g. Bryazgin, 1976; Go rshkov, 1983; Legates and Wilmott, 1990). Aerological a nalyses using the reasonably dense network of Arctic rawin­sonde stations can provide estimates of P - E, but onl y la rge a rea l averages can be obtained.

MODELED PRECIPITATION FIELDS

Recognition of the need for g ridded fi elds oC Arctic precipi­tation has led to interest in the evalu ation of output from numerical weather prediction (NWP) models (\VCRP, 1994). Although data assimil ation/forecast sys tems vary between d ifle rent N \ VP centers, the general procedure is to sta rt with a pre\'ious forecast as a fi rst "guess" of present at mospheric conditions. Obsen 'ed data (e.g. from rawin­sondes, d ropsondes a nd satell ites ) a re assi m i lated to adj ust the forecas t field s, providing ana lyses representing the best est i mate of th e current true state of the atmosphere, used to produce the next forecas t. Assimilation data are primarily free-atmosphere variables. Archived fields represent a blend of analyses based on the first guess a nd assimilation (e.g. pressure heights), a nd modeled surface \'a ri ables, including precipita tion.

Modeled surface field s for the Arctic have suffered from a number of problem s. These include the scarcity of sound­ing data over the central Arct ic Basin, difficulti es in quality control and in determining the appropri ate weights for observations vs first-guess fields over such data-sparse a reas, unce rta inti es in satellite-derived soundings, simplistic repre-

429

Serre;:;e and Maslanik: Arctic precipitation in NCEP/NCA R reana01sis

sentation of sea ice (genera ll y taken as a slab with fi xed thickness a nd 100% concentration), inadequate physical pa rameteri zations (pa rticul a rly with respect to clouds), radi a ti on a nd boundary-layer processes, and problems with filtering fi elds at the northernmos t rows of gridpoint models. A furth er problem is a lack of temporal consistency in archived field s due to changes introduced in models a nd data assimilation systems (WCRP, 1994).

Limited cases studies of operational NvVP output for recent yea rs a re nevertheless encouraging. Precipitation field s for April 1993 from a regiona l model of the Canadi an Atmospheric Environment Service showed very good agreement with observations (Environment Canada, 1993). In a recent WCRP Arctic C lim ate Sy tem (ACSYS) work­shop (held September 1995), it was demonstrated th at preci­pitati on fields from United Kingdom M eteorological Office (UKMO ) and Europea n Center for Medium-Range Weather Forecas ts (ECMvVF) model s provide realistic spa­ti o-tempora l patterns in qua li tati ve agreement with ex isting clim atologies (H. Cattle, persona l communication, 1995).

"Reanalysis" projects underway at several agencies a re prov iding internally con istent g ridded fi elds. Although still prone to many of the problems outlined above, including si mpl istic treatments of sea ice, they should provide improv­ed a rchived fi elds through: I) the elimination of tempora l discontinuiti es through the use of "frozen" state-of-the-art data ass imilation/forecast systems; and 2) the assurance that all available historical data a re used in the assimilations, a nd a re subj ected to strict qualit y control. The most compre­hensive of these projects represents a cooperati ve efTort between the Ja tional Center for Envi ronmental Prediction (NCEP form erly the Nationa l Meteorological Center) and the la tiona l Center for Atmospheric Research (NCAR).

The NCEP/NCAR proj ect a imed to complete the reana lysis for a 40 year time period (1957- 96) by the end of 1996. De­ta il s of the re analysis system a re prov ided by Kalnay and others (1996).

The qua lity of reana lysis prec ipita tion fi elds remains

la rgely untested for the Arctic. A case stud y using NCEP/ NCAR output over the M ackenzie watershed for 1985 a nd 1986 (J. Walsh, personal communication, 1995 shows the seasona l cycles and spati al pa tterns of precipita tion to be reali stic, but with the model totals g reater than from uncor­rected station data . As station va lues a re considered to be

underestimates, this difference is in the "correct" direction. However, the la rgest di screpa ncies occur during spring a nd summer, when most prec ipitati on would be in the liquid form. Precipitation patterns over Greenl and and Antarctica, examined using 6 years of ECMvVF rean alysis data, appear reali stic (Genthon and Braun, 1995).

H ere, we provide an initial evaluation of Arctic precipi­tation as depicted in the NCEP/NCAR reanalysis over the 8 yea r period 1986- 93 for the region north of 60° N . We use the 6 hourl y instantaneous precipitati on ra tes (in kg m 2 S i),

which a re provided over a 192 x 94 gaussian grid. The in­stantaneous rates were converted into monthl y precipitation

tota ls by month and year. Comparisons are made with severa l gauge-corrected cli matologies, including the map of a nnual tota ls provided by Bryazgi n (1976), based on data from fi xed sta tions, a ircra ft landings a nd the Russian No rth Pole series of manned drifting camps (1916- 74). We al so use monthly means north or70° N based on the Gorshkov (1983) a tl as, which according to Burova (1983), rely on Bryazgin's (1976) results, as well as a recentl y acquired set of gauge-

430

cor rected monthly maps by Bryazgin (1976), updated using data through 1990.

RESULTS

Annual field

Figure I shows the spati a l di stribution of modeled annual prec ipita ti on averaged over the 8 years while Figure 2 pro­vides the a nnual gauge-corrected map of Bryazgin (1976). The reana lysis fi elds were interpolated to the NMC octago­na l grid using Cressman weights with a 500 km sphere of influence. This smoo thing eliminates loca l "bull's-eyes" in precipita tion that appear to result from spectral truncati on er rors. The m odel depicts the highest tota ls over the Atlanti c side of the Arcti c. Maximum va lues a re found off the so uth­east coas t of Greenl and (1400 mm ), with amounts decreas­ing to the northeast. Thi s pattern is physicall y consistent with the efTects of the frequent cyc lone acti vity associated with the mean Icelandic Low, and the poleward decay of the North Atlantic cyclone track (Serreze, 1995). High tota ls (up to 1200 mm ) a re a lso depicted near the edge of the map over southern Alas ka, which refl ects the influence of the Aleutian Low and orography. The lowest values (150-200 mm ) a re found over the Beaufort Sea, pa rts of the Canadian Arctic Archipelago, the East Siberi an Sea a nd northern Greenland. These a re regions of infrequent cyclone activity for most of the yea r, and where anticyclonic conditions are common during winter and spring (Serreze and others, 1993).

Fig. 1. Mean annual precipitation (1986- 93) fro m the NCEP/NCA R reanalysis ( in mm).

Although the Bryazgin (1976) climatology is based on data from a difTerent period a nd the gauge corrections employed a re not well documented, comparisons indicate that the model is capturing the mqjor spatial patterns of pre­cipitation reasonably well. In particul a r, the model cor­rectl y depicts precipitation maxim a over the Atl antic side of the Arcti c and southern Alas ka, and low central Arctic values. Nevertheless, there a re some notable differences. In particul ar, Bryazgin (1976) shows high precipitation values (> 400 mm) along the Atlantic side of the Arctic penetrat­ing as fa r as Novaya Zeml ya . The model also fa il s to capture

.'

.....• ~

Fig. 2. M ean annual gauge-corrected precipitation (1916- 74)

redrawn after B lyazgin (1976) (in mm).

precipita ti on maxima related to local orography, such as along the Scandinavian coast, as well as over Iceland, Sva l­bard a nd Novaya Zemlya. This is likely to be related to the rela ti vely low (T 62) model resolution, as well as to the smoothing employed here. Nevertheless, these initia l results must be considered encouraging.

Seasonality and inter-annual variability

Figure 3 shows the modeled average precipi ta tion fi elds for J a nuary andJuly. TheJanuary pattern (Fig. 3a) is simila r to tha t for the annual field , with the highest total s over the Atlantic side (up to 200 mm near the position of the Icelan­

dic Low) and southern Al aska. By contrast, July (Fig. 3b ) shows a reduction in Atlantic-side precipitation, coupled with increases over the central Arctic Ocean, Eurasia, central Alaska and Canada. These results a re consistent with seas­ona l changes in atm ospheric circulation (Whittaker and

H orn, 1984; Serreze and others, 1993; Serreze, 1995). During January, the primary North Atlantic cyclone

track and the Icelandic and Aleutian Lows are strong, with the a ffected areas receiving relatively abundant precipita­tion. During summer, the North Atlantic track a nd sub­pola r lows weaken, but with cyclone activity becoming common over northern Eurasia, Canada and Al aska. Eura­

sian systems in particul a r tend to migrate into the central Arctic O cean, where they subsequently occlude. While lows may enter the Arctic O cean from anywhere along the Eur­asian coast, a Laptev Sea track is preferred.

Bryazgin's (1976) updated monthly analyses agree with the model in showing aJanuary precipitation of 10- 20 mm over the centra l Arctic O cean increas ing towards the Atlan­tic side, but again di splay higher values along the Scandina­vian coast. Bryazgin's (1976) data also show a reduction in July precipitati on over the Atlantic side, with increases over the centra l Arctic O cean, but with lower amounts (30 mm)

than in the reana lysis (30- 40 mm ). Bryazgin's maps also di s­play sharp increases over land, especially the local maxima over Alas ka and eastern Eurasia, which are regions of fre­quent summertime cyclogenesis (Serreze, 1995).

The spatial pattern of the month of the precipita ti on maximum (Fig. 4), calculated for each N MC gridpoint from

Serreze and Maslanik: Arctic precipitation in NCEP/ NCA R reanarysis

Fig. 3. Mean precipitation (1986- 93) from the NCEP/ NCA R reanafysis ( in mm) Jor ( a) J anua1), and (b) J ury.

the long-term monthly means, shows two maj or features: I) an O ctober-january maximum over the Atlantic side of the Arctic a nd the Baffin BayjDavis Stra it region; and 2) a June- September maximum over the remainder of the

Arctic, with a spatially homogeneous July peak [or the centra l Arctic O cean. A recent analysis (Cl a rk a nd others, 1996) o[ precipitati on frequencies over the Arctic O cean and its periphera l seas, evaluated from present weather reports conta ined in the Comprehensive Ocea n Atl11.osphere Da ta­set (COADS ) (Woodruff and others, 1987), shows this Atlan­tic-side win ter m aximum clearly. The timing of this maximum a lso agrees with the updated monthly maps of Bryazgin (1976). However, it appears that the model's July peak precipitation over the central Arcti c Ocean is in error by at leas t a month.

This is illustrated in Figure 5, which shows the long-term

(1986- 93) monthly reanalysis precipita ti on averaged for the centra l Arctic, taken as the region north of 70° N, the max­imum and minimum values based on individua l years, and the climatologica l monthly nl.eans from Gorshkov (1983). FromJanuary through April, the reana lysis means are low, but high from May throughJuly. Despite these differences,

431

Sareze and A1aslanik: Arctic precipitation in NCEPjNCA R rea1la01sis

./ .. ]

~ 9 7

t..9 .' 7'

.. ·7' " 6' ·T .. "]" ·· ·· ..

'r;" 6 '·. 6 6 7 7 7 .: 7 '7 ..

7 .7' " 7 '·

" ,.,' '2 '2

6 ,~~,.:\ . l?'\ 1 ' 1 ., 1

.. '{1~· 1 ··

;8\.<, 1

~12 \'~" 0 . " ~."

ri.g. .f.. Month of the precipitation maxil71umji'Olll the NCEPj . yeA R reana0lsis, based on long-term (1986- 93) monlh£JI means. The solid line sej)amles regions with October-.7anuar)1 andJlIIze-September maxima.

40

35

E 30 E ....., c: .Q !§ "6. "0 Q) "-

Cl..

25

2 4 6 8 Month

10 12

Fig. 5. l'vI.ean monthb precipitation jor the Tegion north ()[ 70° N as dep icted by Gorshkov (1983) (dashed line) and in the NCEPjNCAR reanaOlsis (solid line). The range in the reanaOlsis values over tlze 1986- 93 is shown by the vertical Lines ended by dots (all vaLues ill mm ).

the J anua ry- July reana lysis total s can be considered quite reasonable. H owever, the model shows a very sharp precipi­ta ti on peak inJul y, a month ea rlier than shown the data pre­ented by Gorshkov (1983), remaining far below hi s values

through December. Even the positive ex tremes of the reana­lysis precipitati on arc lower tha n the Gorshkov means from August through O ctober. As shown in the monthly time se r­ies ofmodeled precipi tation (Fig. 6), thi sJuly peak is persis­tent from year to year. Due largely to the underestimates from August through December, the model's a nnual-precipi­

tation total [or the centra l Arctic region is 250 mm, as com­pared to 293 mm from the Gorshkov a tl as. Interestingly,

432

....... E E -c o

""§ -

50r------------------------------------

40

;.

:: !

:9. 20 f\j\ o ~

Cl.. ~·V

o 86

:.'.: .:

Month

93

Fig. 6. Time series q/lI1onlh!)1 precipitation fin the period 1986- 93 ji-Oln the NCEPjNCA R reanalysis Jar the region north of 70° N ( in mm ).

nca rly a ll models reviewed as part of the Atmospheri c ~Iode l Tnlercomparison Prqjee l (AMIP) display an Aug ust prec ipit ati on maximum for the central Arctic (Kallsov a nd others, in press). Note a lso that the August peak depicted by these models occurs a month ea rli er than the September

peak in the water vapor nu x convergence a nd P - E [or the region north of 70° N as ana lyzed from rawinsonde data (Ser rcze and others, 1995).

DISCUSSION

Based on our initi a l study, the performance o[ the NCEPI NCAR reanalysis to depict Arctic precipitation could per­haps best be described as uneven. The spatial patterns of long- term annual and monthly precipitation agree reason­ably well with avail able climatologies, although there are

some notable differences. In particular, the annual peak in prec ipitation over the centra l Arctic O cean occurs too early in July with la rge underes tim ates from August through December. \Ve recogni ze the poss ibili ty that the 8 yea rs (1986- 93) examined here might not be sufficiently long enough [or effective comparisons with observed c1imatolo­

gies o r the AMIP outputs, wh ich span the period 1979- 88 (K attsov a nd others, in press).

The past decade has been anomalous in terms of Arctic circu lation (Nlas lanik a nd others, 1996; ' '''alsh and others, 1996). H owever, thi s per iod has been cha racterized by sharp increases in cyclone activity and accompanying recluctions

in sea-level pressure over the central Arctic O cean, which, a /)1'iol'i, suggest that precipitat ion over the centra l Arctic has recently been above norma l. Resolving this issue wi ll require further investigations using additiona l reanalys is data, as we ll as efforts to assess the model's treatment of high-l atitude

hyd rologic processes. D espite the shortcoming in the modeled fi elds, we echo

the recommenda ti ons of the WCRP (WCRP, 1994) to exam­ine methods of blending the reana lysis precipitation fi elds with observed data. Efforts a re underway by C. vVillmott at the University of Delaware to compute monthly gauge-cor­rected precipitation fi elds [or Arctic land a reas, bui lding on

techniques outlined by Legates a nd Wilmott (1990). Using such field s, along with the upda ted el imatological month ly

means from Bryazgin for the Arctic Ocean, and gauge­corrected station data, techniques such as optimal interpo­lation could be applied to obtain gridded fields for use 111

climate studies and sea-ice models.

ACKNOWLEDGEMENTS

Thi s study was supported by NSF grants ATM-9315351, OPP-950420l and OPP-9614297.

REFERENCES

Aagaard, K. a nd E. G Carmack. 1989. The role of sea ice a nd other fl'esh

water in the Arctic circul a tion.] Geophp. Res .. 94 (CI0]. Ic}.485 H.498.

Bryazgin ,:"J.:\. 1976. Srednegodol'oe koliches tvo osadkO\· I' Arktikr s uchclOm pogreshnost i osadkoll1 erOl ' rll lean a nnua l precipitation in the .-\rCl ic wit h consideration of ga uge biases 1- Tic , Irkl. :llIlmkl. Aallrilllo-Iss/{'(1. i llsl. 323,4·0 7+.

BurOl'a, L. P. 1983. FlagooborDl v allllosjere Arkliki .\/oislllre [xclwllg" illlhe Arclic almo5jJ/tere . Leningrad, Gidrolllclcoizclat .

Cl ark, 111. P. , M. G Scrreze a nd R. C. Ba IT Y. 1996. CharaCleristics of ArClic

O cean climate based on COADS data, 1980- 1993. Geoph}'S. Res. Lell. , 23 (15], 1953- 1956.

Dingll1an. S. L.. R. G. Barry, G. \\'ell er, G Bcnsnn, E. F. LeDrcw a nd G \\'. Goodwi n. 1980. Clinlatc, snow cover, 1l1icroclimate and hydrology. i ll Brown, j. , P. C. :'I I ill cr, L. L. Ti eszen and F L. Bunnell , eds. All Arclic ew­

. ,vslem: Ihe (oasf{d tUlldra al Barrow . . 1Iaska. Stroudsburg, PA, Dowden. Hutchinson a nd R oss, 30- 65.

EI1I 'ironmelll Canada. 1993. forecas t a lmost as good as observa tions? Clilllale Pm/Helives. 15,11.

Genthon, G a nd A. Brallll. 1995. EClIfWF a na lyses a nd predictions of' the surface c li ma te ofCreenland a nd Antarctica.] Climale, 8 (10), 232+--2332.

GorshkO\ ', S. C .. I'd. 1983. Arclic Ocean. Vo/. 3. I Vorld ocean alias. Oxford , etc .. Pergamon Press. U.S.S. R. Navy.

lI·a nOl'. VV and V A. 'l'<tnkina. 1991. Vod nye resurs), Arhiki [\\'ater resou rces of the Arctic]. Prabl. Ark. Anlark., 66.11 8 128. [In Russ ian .)

K a lnay, E. and 21 olhers. 1996. The ACEP/:\TCR 4·0-yea r reana lysis projec t.

Serreze and !llaslanik: Arclic jJrecipitation in NCEPINCAR reana£vsis

BIIII. Am .. \Ielforol. Suc., 77 (31, +37- J.71. Kattsol', V 11 I.. T. V PavlOl'a a nd V A. COI·orkOl·a. J n press. Heat andll'"tcr

budgets OITr the northern po la r region as est imated from I-f atmo­

spheri c genera l ci rcul atio n model s. III Arclic Clillla/e S)'slelll Slud)' (,-I(.'S1 S) COn/err'IICI" 199-1. Gdleborg, Sweden. Pmtm/ings. Ce t~CI·a. World !,deleorologica l Orga nizat ion. \' "orld Climale Progran1l11e.

Lega tes1 D. a nd C. " ·illm ott. 1990. ~dcan seasona l and spa ti a l variabilit y in gauge-correc ted globa l precipi tation. 1111.] Clilllalol., 10(2), III 127.

IIIaslanik. j. A., -"1. C. Scrreze and R. G. BatT\". 1996. Recent decreases in Arcti c summer ice COITr a nd li nkages ;0 atmospheric circu lat ion

anoma li es. Geop/~)'s. Res. Lell .. 23 (13 ), 1677- 1680.

IIleCauiev, L. L. a nd -" 1. F. -"Icier. eds. 1991. Arclit splem sciell(f lalld/ alll1o­sl)hne/ ice illlemffiolls: a 1)1011 Jar affioll. fairbanks, AK. Arcti c Resea rch Consortium ort he United States (ARCUS ).

Scrreze, :\1. C. 199:1. C lill1alological aspects of cyclone del'elopmcnt a nd decay in the Arctic. AllI1os/Jhere - Oreall. 33 (1), I 23.

Serrcze, \1. G. j. E. Box, R. C. BatTY a ndj. E. \\'a lsh.1993. Characteristics

of Arctic synoptic actil 'it y, 1952- 1989 . . l/eleoro!. Allllos. P/~)'s., 51. H7- 164. Serreze, ;'d. c.. R. C. Barry andj. E. \\'a lsh . 1995. Atmospheric water I'apor

characterist ics at 70 N.] Climale, 8 (,1'),719 731. Vose. R . S. and 6" olhns. 1992. The global hislorical dimalologl' lIelwork: IOllg-lerlll

1Il0nth[11 tempera/lire, jJrecijJitatioll . sea level jJressure, alld slalioll jJrfss lIre data. Oak Ridge. TN, National Laboratory. l Envi ro nmenta l Sciences Dil·i­sion Publicat ion 3912.)

Wa lsh, j. E., \\'. L. Chapman and T. L. Shy. 1996. Recent decreases of sea !n'cI pressure in the centra l Arctic.] Climale, 9 [2], C}80- 486.

\\'hi ltaker, L. i'd . a nd L. H. Horn. 1984. ;"iorthern H emisphere extra Jropi­ca l cyclone aCl ivit )' for four mid-season months. ] Climato/., 4. 297- 310.

\\'00, 1\1. -K .. R . H eron, P. illars h and P. Steer. 1983. Comparison of wc at her s la tion snow ralI with winter snow acculllulati on in High ."-\re lic basins . . I 1111 oS/Jit ere-Orerlll , 21 (3), ~ 1 2-325.

Wood ru lT, S. 0.. R .j. Slutz, R . L. J enne and Pill. Steurer. 1987. A COI11-

prchensil'c ocean atmosphere data set. BIIII. _JIII . • l/eleoral. Soc. , 68 (10), 1239 1250.

World C limate R esearch Progra mme (\\ 'CRP I. 199+ . Inlic Climate S)'slell1 Stll(!J ' ( , IC.S l :5"): illilial iml)lcmelllolioll ploll. Genel'a , \ \'o rld I\lcrcorol;gica l Organiza ti on Secretariat. (WCRP-85. \\']\10 TD 627.)

Zaucker. T , T. F. Stocker and \\'. S. Broecker. 1994. Atlllospheric fres hwater nu xes and tlwir dfect on the globa l thermohalinc circulation.}. (;eupl!),s.

Res., 99 (C6), 12.4+3 12,+57.

433