Satellite Assessment of Divergent Water Vapor Transport from … · 2008-04-29 · Journal of the...

Journal of the Meteorological Society of Japan, Vol. 85, No. 5, pp. 615--632, 2007 615

Satellite Assessment of Divergent Water Vapor Transport from NCEP,

ERA40, and JRA25 Reanalyses over the Asian Summer Monsoon Region

Seong-Chan PARK, Byung-Ju SOHN

School of Earth and Environmental Sciences, Seoul National University, Seoul, Korea

and

Bin WANG

International Pacific Research Center, School of Ocean and Earth Science and Technology, University ofHawaii at Manoa, Honolulu, Hawaii, USA

(Manuscript received 4 January 2007, in final form 21 June 2007)

Abstract

The divergent component of water vapor transports was constructed using evaporation, precipitation,and total precipitable water estimated from the Special Sensor Microwave Imager (SSM/I). The SSM/Imoisture budget parameters were then compared with those from the National Centers for Environmen-tal Prediction (NCEP), the European Centre for Medium-Range Weather Forecasts (ECMWF) 40-yearReanalysis (ERA40), and the Japanese 25-year Reanalysis Project (JRA25) data over the Asian monsoonregion for the May to September (MJJAS) period from 1988 to 2000.

The climatology of SSM/I water vapor transports clearly indicates three major water vapor sourcesfor the Asian monsoon, i.e., the subtropical Indian Ocean and Pacific Ocean in the Southern Hemisphere,and the North Pacific Ocean. In contrast, sinks are located in the Asian summer monsoon trough, theequatorial convective zones over the Indian and western Pacific Oceans, and over the East Asian mon-soon region from the northern tip of Philippine Sea to the Kuroshio extension region. These sources andsinks are linked to well-known large-scale rotational circulation features, i.e., the cross-equatorial flowassociated with monsoon circulation over the Indian Ocean, the anticyclonic circulation along the west-ern periphery of the western North Pacific High, and the cross-equatorial flow north of New Guinea. Inconjunction with the fluctuation of these sources and sinks, the northward propagating climatologicalintraseasonal oscillations in water vapor flux convergence are evident in the South and East Asian mon-soon regions in the SSM/I data.

From the comparison of water budget parameters of NCEP, ERA40, and JRA25 reanalysis withSSM/I-derived features, we found that the general features of all three reanalyses are in good agreementwith those from SSM/I; however, the magnitudes of water vapor transports are comparatively weaker inall three reanalyses than what SSM/I measurements suggested. In addition, much weaker water vaportransports in three reanalyses are found in the intraseasonal oscillation signals with less distinctpatterns, compared to what are inferred from the SSM/I measurements.

1. Introduction

Global reanalysis data produced indepen-dently by major operational centers, such as theNational Centers for Environmental Prediction(NCEP) and the European Centre for Medium-

Corresponding author: Byung-Ju Sohn, School ofEarth and Environmental Sciences, Seoul Na-tional University, NS80, Seoul, 151-747, Korea.E-mail: [email protected]( 2007, Meteorological Society of Japan

Range Weather Forecasts (ECMWF) have beenwidely used to examine the Earth’s global hy-drological cycle and to understand its associ-ated atmospheric circulation (Trenberth andGuillemot 1995; Mo and Higgins 1996; Tren-berth and Guillemot 1998; Cohen et al. 2000;Roads et al. 2002). These reanalyses are pro-duced by assimilating the models’ first guessfields to various meteorological observations.Some of the variables such as pressure, temper-ature, wind, and humidity are directly ana-lyzed by the assimilation system, while others,like precipitation and evaporation, are derivedproducts which strongly depend on the modelphysics and thus the model parameterization.In other words, those derived variables aremodel-dependent and are not fully determinedby the observations. Furthermore, the watervapor balance is not warranted.

Naturally we expect uncertainties in the hy-drological cycle in the reanalysis data, due toinsufficient and inaccurate observations usedfor the reanalyses as well as deficiencieswithin the assimilation system. For instance,Rasmusson and Mo (1996) demonstrated thatthe hydrological cycle from NCEP is useful fordescribing general features of the atmosphericbranch of the hydrological cycle. At the sametime, they also showed that there are limita-tions in the study of continental-scale moisturebudgets because of large regional terrain-related systematic biases. Consistent with thispoint, Trenberth and Guillemot (1998) noted anegative bias in tropical precipitation in NCEPproducts, suggesting a weak divergent circula-tion over the Tropics.

Consequently uncertainties in moisturebudgets remain among various reanalyses (An-namalai et al. 1999; Newman et al. 2000) al-though intercomparison itself provides a meansfor diagnosing the performance of global anal-yses and climate models. To evaluate the qual-ity of global reanalysis data, or to better under-stand the role of the hydrological cycle in theEarth’s climate, quantitative ‘‘ground-truth’’ in-formation is demanded.

The recent advent of satellites such as theSpecial Sensor Microwave Imager (SSM/I) on-board the Defense Meteorological SatelliteProgram (DMSP) satellite can provide atmo-spheric water budget parameters, i.e., watervapor, cloud liquid water, and precipitation, to

examine the water budget in the atmosphere.The spatial and temporal homogeneity and ac-curacy of satellite observations offer an oppor-tunity for a satellite-based study of the hydro-logical cycle, which in essence, captures all theimportant water budget components and watervapor transport features. Thus, it is worth-while to compare well-known global reanalysisagainst SSM/I-derived water budget data, andshed light on the strengths and weaknesses ofthe reanalyses, as well as the ways, by whichthe model physics might be improved.

In this paper, we compare the divergent com-ponent of water vapor transports derived fromreanalyses against SSM/I-derived parametersover the Asian summer monsoon region. Themoisture sources and sinks, and their associ-ated transport features, are important ele-ments for understanding the monsoon pre-cipitation and circulation. In doing so, weconstructed the moisture transport field usingthe SSM/I retrieved precipitation and evapora-tion datasets for the period of 1988–2000, byapplying the methodology developed by Sohnet al. (2004). The satellite-derived moisturebudget parameters will then be compared withthose from NCEP, ECMWF 40-year reanalysis(ERA40), and Japanese 25-year ReanalysisProject (JRA25) data. In this study, we aim toidentify common features and discrepanciesamong various reanalyses and satellite-basedresults, and attempt to explain what mightgive rise to such discrepancies.

We describe the methodology and datasetsin Section 2. Section 3 compares the derivedmoisture flux divergence from the satellitemeasurements and reanalysis data, along withradiosonde observations. In Section 4 we pres-ent geographical distributions of precipitation,evaporation, and associated divergent com-ponent of water vapor transports from theSSM/I, and from NCEP, ERA40, and JRA25reanalyses. Climatological intraseasonal varia-tions associated with the water vapor transportand divergence field are discussed in Section 5and conclusions follow in Section 6.

2. Methodology and description of useddatasets

The methodology used here for deriving watervapor transports over the global oceans fromsatellite-retrieved precipitation ðPÞ and evapo-

616 Journal of the Meteorological Society of Japan Vol. 85, No. 5

ration ðEÞ is described in detail in Sohn et al.(2004). Briefly, for a given atmospheric column,the water balance is achieved as follows:

qW

qtþ ‘ � Q ¼ E � P; ð1Þ

where W is the total amount of precipitablewater, E and P are evaporation and precipita-tion, respectively. Q is the horizontal watervapor transport vector defined by

Q ¼ 1

g

ð ps

p0

qV dp; ð2Þ

where g is the acceleration of gravity, q is spe-cific humidity, V is the horizontal wind vector,ps is the pressure at the surface, and p0 is thepressure at the top of the atmosphere.

In the case of SSM/I, by separating the watervapor transport vector ðQÞ into rotational ðQRÞand divergent ðQDÞ components (Rosen et al.1979; Chen 1985; Sohn et al. 2004) and intro-ducing the water vapor transport potentialfunction ðFÞ we effectively remove the rota-tional component ðQRÞ, i.e.:

‘ � QD ¼ E � P � qW

qt¼ �‘2F ð3Þ

QD ¼ �‘F: ð4Þ

Equations (3) and (4) are solved for F by us-ing a spectral method on a global domain withinputs of evaporation, precipitation, and totalprecipitable water. The used pentad meanSSM/I data are given in a 2:5� � 2:5� grid for-mat. Solving Eq. (3) for the SSM/I water vaportransport potential function ðFÞ requires knownboundary conditions for a given domain. Toavoid the difficulty in specifying boundary con-ditions, we solve Eq. (3) over the whole globeusing spherical harmonics in which boundaryconditions vanish at both poles.

As in Eq. (3), the local change of W needs tobe calculated for solving the water vapor fluxdivergence. The local change term qW

qt

� �is ex-

plicitly calculated using pentad means of totalprecipitable water. For the calculation of SSM/I divergent component of water vapor trans-ports in a global domain, land areas void ofSSM/I precipitation and evaporation are filledwith NCEP data.

The global mean of SSM/I E � P � qWqt

� �must be zero in order to solve Eq. (3) over the

global domain since the area integral of‘ � QD of Eq. (3) over the global domain iszero. In order to satisfy the balance require-ment for solving Eq. (3), the global mean biasesare subtracted from the E � P � qW

qt

� �field to

yield a zero global mean while preserving thespatial gradient of E � P � qW

qt

� �. However, in

the case of SSM/I, (1) water vapor divergencesover the land are filled with NCEP values be-cause SSM/I values are available only over theocean; (2) adjustment is made only over theocean to achieve a zero global mean by sub-tracting a fixed 0.67 mm day�1 from eachSSM/I grid value over the ocean. The adjust-ment only over the ocean would result in aweaker SSM/I water vapor transport from theocean to the land than without adjustment.The large positive adjustment in the SSM/I of0.67 mm day�1 is due to larger evaporationthan precipitation over the ocean, as indicatedby the mean precipitation (2.5 mm day�1) andmean evaporation rates (3.3 mm day�1) aver-aged over the ocean.

In the reanalyses, however, water vaportransports are directly calculated from q and Vdata for more realistic depiction of water vaportransport rather than indirectly calculatedfrom the E � P field. It is because q and V fieldsare produced by assimilating the model outputsto various observations. In doing so, verticallyintegrated water vapor transport vectors ðQÞare obtained by applying Eq. (2) to reanalysisdata, and then a similar methodology used forthe SSM/I calculation is applied for calculatingthe potential function, i.e.:

‘ � Q ¼ �‘2F: ð5Þ

The same spherical harmonics approach is em-ployed to avoid the boundary value problemsfor solving potential functions. The associateddivergent component of water vapor transportðQDÞ in the reanalsyes is then obtained by tak-ing the gradient of potential function as in Eq.(4).

Examining the long-term water budget overthe Asian monsoon domain, SSM/I-derived pre-cipitation and total precipitable water (Wentzand Spencer 1998) for the May–September pe-riod of the years from 1988 to 2000 were ob-tained (from http://www.ssmi.com). The evapo-ration data were obtained from the productsbased upon an algorithm developed by Chou

October 2007 S.-C. PARK et al. 617

(1993), in which the stability-dependent trans-fer coefficients were solved for the bulk aerody-namic formula iteratively, with latent, sensible,and momentum fluxes. Because SSM/I datawere the main source for the evaporationretrieval, we refer to Chou’s products as theSSM/I evaporation. Chou et al. (2003) vali-dated the SSM/I daily evaporation data againstvarious field experiment results. It has been re-ported that SSM/I daily evaporation productsshow the bias up to 0.09 mm day�1 with a cor-relation coefficient of 0.80 over the tropicaloceans. The root mean square error (RMSE)on a monthly time scale was claimed to beabout 0.19 mm day�1, indicating a relativelysmall error ðA6%) in comparison to themonthly mean evaporation rate of 3.21 mmday�1.

The hydrological variables (i.e., E and P), spe-cific humidity ðqÞ, and horizontal wind fieldsðVÞ were obtained from the NCEP (Kalnayet al. 1996) and the ERA40 (Simmons and Gib-son 2000) for a comparison with satellite-basedresults over the Asian summer monsoon regionin the same 1988–2000 period. The NCEP re-analysis consists of 8 levels for q and 17 levelsfor V, and for calculating the vertically inte-grated water vapor transport ðQÞ, only thelower 8 levels (1000, 925, 850, 700, 600, 500,400, 300 hPa) of wind are used in conjunctionwith the q profiles. On the other hand, theERA40 provides q and V at 23 levels from1000 to 1 hPa, but only the lower 11 levels(1000, 925, 850, 775, 700, 600, 500, 400, 300,250, 200 hPa) are used since most of thewater vapor in the atmosphere resides below200 hPa. In addition, we employed the Japa-nese 25-year Reanalysis Project (JRA25) data,from the ongoing cooperative project betweenthe Japan Meteorological Agency (JMA) andthe Central Research Institute of the ElectricPower Industry (CRIEPI) for the target periodof 1979–2004 (Onogi et al. 2005). The verticallyintegrated JRA25 water vapor fluxes aredirectly obtained from the JRA25 Web site(http://jra.kishou.go.jp).

3. Comparison of moisture fluxdivergence with radiosondeobservation

We first compare satellite-estimatedE � P � qW

qt

� �and moisture flux divergence

ð‘ � QÞ of reanalyses against ‘ � Q from radio-sonde observations. Also compared are betweenwater vapor divergence from E � P of reanalysesand that from q and V profiles of reanalyses.This comparison offers a way of examining therelative accuracy of the water budget derivedfrom satellite measurements and reanalysisdata, and of examining the calculation methodfor the flux divergence from reanalysis data.

Flux divergence from radiosonde observa-tions is based on the line integral of the watervapor flux into the domain of interest. Thismethod has been successfully implemented tocalculate the vertically integrated moisturebudget on a regional basis, such as over theAustralian Gulf of Carpentaria (McBride et al.1989). Green’s theorem is applied over the do-main, yielding the area mean moisture flux di-vergence over the domain:

‘ � Q ¼ 1

AREA

ð200 hPa

1000 hPa

þqVn � dl

� �dp

g

¼ E � P � qW

qt; ð6Þ

where q is the water vapor mixing ratio, Vn isthe wind component normal to the boundarysegment dl, g is the acceleration of gravity,and AREA is the domain area. Line integralsare performed along the boundaries of the do-main at every 50 hPa level and vertical inte-gration is done from 1000 to 200 hPa.

Two oceanic areas surrounded by radiosondeobservation sites were chosen for the analysis(Fig. 1). Radiosonde observations were made ateach site with 6-hourly soundings during theMay–June 1998 period as a part of the GlobalEnergy and Water cycle Experiment (GEWEX)Asian Monsoon Experiment (GAME), i.e.,GAME/SCSMEX (South China Sea MonsoonExperiment)—see Lau et al. (2000) for the de-tailed information on this experiment. Verti-cally integrated water vapor fluxes onto tworegions are obtained from daily averages of ra-diosonde data by applying Eq. (6) to enable thecomparison with E � P � qW

qt

� �from SSM/I, and

with the water vapor flux divergence ð‘ � QÞfrom reanalysis.

Figure 2 displays comparisons of SSM/IE � P � qW

qt

� �and NCEP, ERA40, and JRA25-

derived moisture flux divergence ‘ � Qð Þ withthe moisture flux divergence from radiosonde


observations over two regions in Fig. 1. Alsocompared are the E � P of reanalyses againstflux divergences ð‘ � QÞ estimated from reanal-yses q and V profiles.

On a daily basis, SSM/I-derived flux diver-gence is in good agreement with that suggestedfrom radiosonde observations, showing a corre-lation coefficient of 0.88. The NCEP, ERA40,and JRA25 results based on the explicit mois-ture flux calculation in Eq. (2) show the correla-tion coefficients of 0.69, 0.67, and 0.75, respec-tively. Amongst three reanalyses, the JRA25seems to have a largest mean bias although itholds highest correlation coefficient. The com-parison of reanalysis E � P with radiosonde-based flux divergence shows correlation co-efficients of 0.67, 0.49, and 0.79 for NCEP,ERA40, and JRA25, respectively. Similar meanbias between E � P and ‘ � Q is found forJRA25. In general, the explicit calculationseems to result in flux divergences more compa-rable to radiosonde-derived values.

It is worthwhile to note that discrepanciesare larger for reanalyses when radiosonde-derived convergence is large, reflecting the

influence of few heavy rainfall events sincethe temporal and spatial variations of E � Pare predominantly controlled by variation inprecipitation. In other words, water budgetcharacteristics for heavy rain events duringthe summer over the experiment domain werewell captured by radiosonde observations, butnot well represented by NCEP, ERA40, andJRA25 precipitation. Those discrepancies forheavy rain events tend to be smaller when ana-lyzed q and V fields are used for the flux diver-gence calculation, suggesting that the data as-similation brought into a better agreementwith radiosonde-based water vapor divergence.

4. Properties of 13-year mean watervapor transport fields

Here thirteen year (1988–2000) summer(May–September) mean divergent componentof water vapor transports are examined alongwith precipitation and evaporation over theAsian monsoon region. Since the SSM/I-derived products are available only over theocean, the NCEP, ERA40, and JRA25 P and Eare compared over the oceanic region of analy-sis domain.

4.1 PrecipitationThirteen year (1988–2000) May–September

(MJJAS) mean precipitation derived from theSSM/I and its differences from the reanalysesare shown in Fig. 3. The salient features foundin the SSM/I precipitation (Fig. 3a) includeheavy rainfall patterns located in the equato-rial regions associated with strong convectionin the monsoon trough extending from theBay of Bengal to the Philippine Sea and the In-tertropical Convergence Zone (ITCZ). High pre-cipitation zones are also found in the equatorialeastern Indian Ocean slightly south of theequator and the South Pacific ConvergenceZone (SPCZ). Also noted is the heavy precipita-tion along the East Asian subtropical front ex-tending from the northern tip of the Philip-pines to the Kuroshio extension region alongthe northwest rim of the western Pacific Sub-tropical High. On the other hand, precipitationof less than 2 mm day�1 is seen over the sub-tropical Indian and the North Pacific Oceans,which are under the direct influence of the sub-sidence associated with the subtropical highpressure system.

Fig. 1. Two areas in the South China Seawhere divergence of water vapor fluxesare computed from radiosonde data col-lected during the Intensive ObservingPeriod (IOP) of the South China SeaMonsoon Experiment (SCSMEX) fromMay 1 to June 30, 1998.


Noticeable differences in precipitation existamong the three reanalyses and between thereanalyses and the satellite observation. Oneof the major differences is found over the EastAsian subtropical front, extending from theEast China Sea to the North Pacific, wheresignificant underestimation of precipitation isfound in all three reanalyses. By contrast, over-

estimates by reanalyses are commonly foundover the North Pacific high area.

The NCEP (Fig. 3b) generally underesti-mates precipitation over the highly convectiveareas such as the equatorial Indian Ocean, theITCZ along the equatorial western Pacific, andthe East Asian monsoon region, whereas itoverestimates over the drier regions such as

Fig. 2. Scatterplots for radiosonde-derived water vapor flux divergence vs. (a) SSM/I, (b) NCEP, (c)ERA40, and (d) JRA25-derived water vapor flux divergence, respectively. In (b)–(d), black andwhite dots indicate ‘ � Q and E � P of reanalyses, respectively.


subtropical oceans and the western IndianOcean. The precipitation deficit along the ITCZhas been noted by Trenberth and Guillemot(1998), who pointed out that such a bias couldbe due to deficient precipitable water in theTropics.

In contrast, the ERA40 in Fig. 3c shows anoverestimation over most equatorial latitudescompared to SSM/I-derived precipitation. Thelargest discrepancies are found over the trop-ical western Pacific Ocean between 10�S and20�N. This finding is consistent with the resultsof Bengtsson et al. (2004) who compared theERA40 with two sets of precipitation climatol-ogy derived from the Global Precipitation Cli-

matology Project (GPCP) and the CPC MergedAnalysis of Precipitation (CMAP). This exces-sive precipitation over the Tropics seems dueto the excessively strong circulation over thetropical convective regions in the ERA40 assim-ilation system (Holm et al. 2002). Again, under-estimation of precipitation is seen over the EastAsian summer monsoon region.

While the difference pattern in the JRA25(Fig. 3d) is similar to those of the NCEP andthe ERA40, the magnitudes of the discrepan-cies in the JRA25 in the area off the east coastof Africa and over the equatorial western In-dian Ocean are smaller. In general, the JRA25shows better agreement with the SSM/I than

Fig. 3. Thirteen year (1988–2000) MJJAS mean (a) precipitation derived from SSM/I, and precipita-tion differences between SSM/I and (b) NCEP, (c) ERA40, and (d) JRA25. The contour intervalsare 2 mm day�1. Shaded area represents precipitation greater than 6 mm day�1 (a) and negativevalues (b–d).


NCEP and ERA40 reanalysis do. The re-sults of this comparison suggest that the long-term mean precipitation produced by theJRA25 in the Asian summer monsoon and theIndo-Pacific warm pool regions matches satel-lite observations better than those of theNCEP and the ERA40, even though all threereanalyses show significant underestimation ofprecipitation over the East Asian monsoon re-gion.

4.2 EvaporationThe distributions of evaporation are given in

Fig. 4. In the SSM/I distribution (Fig. 4a)values larger than 5 mm day�1 are found inthe trade wind zone centered at about 20�Sover the Indian and Pacific Oceans of the

Southern Hemisphere. The reason is that inthose regions high surface winds are coupledwith a large surface humidity gradient (Chouet al. 2004; Feng and Li 2006). On the otherhand, over the heavy precipitation areas, inparticular in the equatorial oceans, evaporationrates are generally below about 4 mm day�1

probably due to the weak winds despite highsurface air humidity over those regions. Rela-tively low evaporation is found in the NorthernHemisphere high latitudes because of the lowsea surface temperature.

The seasonal mean distributions of evapora-tion fields reproduced by the NCEP, ERA40,and JRA25 are generally consistent withSSM/I estimations. However, significant differ-ences are found in some regions. Figure 4b

Fig. 4. Same as in Fig. 3 except for evaporation. The contour intervals are (a) 1 mm day�1 and (b)–(d) 0.5 mm day�1, respectively. Shaded areas represent evaporation greater than (a) 5 mm day�1

and (b)–(d) 1 mm day�1, respectively.


shows that the NCEP evaporation is strongerthan the SSM/I counterpart over the equato-rial Indian Ocean and SPCZ. This is due tothe fact that the effects of increased surface hu-midity gradients and moisture transfer coeffi-cient override the effect of reduced surfacewinds in the NCEP (Chou et al. 2004). Theyfurther pointed out that the NCEP reanalysissystem has a smaller surface humidity differ-ence as well as a smaller moisture transfer co-efficient over the rest of the area, especially inthe southern Indian Ocean, which overridesthe surface wind effect.

In the case of the ERA40 (Fig. 4c), the differ-ence pattern in evaporation are qualitativelysimilar to that of the NCEP except for thewestern part of the Indian Ocean from Mada-gascar to the Arabian Sea, where lower evapo-ration is found in comparison to other rean-alysis products. Larger evaporation over thewestern Pacific Maritime continent to the Bayof Bengal appears to be associated with largersurface humidity gradients in the ERA40 thanthe NCEP. Feng and Li (2006) asserted thatthe evaporation difference over the westernPacific is mainly due to the surface humiditydifference rather than the surface wind magni-tude.

On the other hand, the JRA25 shows greaterevaporation rate over most of the tropicaloceans, especially over the northern IndianOcean and the tropical western Pacific (Fig.4d). Apart from the NCEP and the ERA40,however, the analysis of surface humidity gra-dient versus wind speeds indicates that thesespatial patterns can not be solely explained bythe wind or humidity alone, rather, they are ac-counted for by the combined wind-humidity ef-fect (not shown). The aforementioned resultssuggest that the details and accuracies of eachvariable in the bulk aerodynamic formula andthe moisture transfer coefficients all contributeto improvement of the quality of evaporation.

4.3 Divergent component of water vaportransports

It is of interest to examine how the water va-por transports in reanalyses differ from SSM/Ibecause water vapor transport implies energytransports shaping the atmospheric circulationpattern. Thirteen year (1988–2000) MJJASmean potential function ðFÞ, divergent mois-

ture flux ðQDÞ, and its convergence ð‘ � QDÞare presented in Fig. 5. In the SSM/I (Fig. 5a),the water vapor source (i.e., ‘ � QD > 0) regionsare located over the Southern Hemisphere sub-tropical Indian and Pacific Oceans and theNorth Pacific Ocean, whereas sink regions (i.e.,‘ � QD < 0) are in the equatorial convectivezones over the Pacific Ocean and the areafrom the western Pacific to the northeast of Ja-pan along the rim of the western Pacific sub-tropical High, associated with the East Asiansummer monsoon. The ‘ � QD (AE � P) pat-terns are similar to those shown in precipita-tion. It is because precipitation behaves in alocalized fashion while evaporation showslarge-scale behavior, implying that spatial pat-terns of the vapor flux convergence (or diver-gence) are mainly controlled by the spatialdistributions of precipitation. The resultant di-vergent water vapor transport ðQDÞ derivedfrom the SSM/I is characterized by cross-equatorial transports from the Southern Hemi-sphere. These are consistent with well-knownlarge-scale circulation features over the Asiansummer monsoon region (Murakami and Mat-sumoto 1994; Li and Yanai 1996; Sohn et al.2001); i.e., (1) the cross-equatorial flow associ-ated with monsoon circulation over the IndianOcean, (2) the southeast flow associated withthe anticyclonic circulation over the westernPacific, and (3) the cross-equatorial flow northof New Guinea. These transports then result ina strong water vapor convergence over the Bayof Bengal, the tropical western Pacific, and theEast Asian monsoon region.

The general features of moisture transportsfrom the three reanalyses are in good agree-ment with those from SSM/I; however, signifi-cant differences are found in some areas (seeFigs. 5b–d). The common features found inthe three reanalyses are significantly weakernorthwestward transports over the PacificOcean north of 10�N, in comparison to SSM/I.Thus, the strong convergence over the EastAsian monsoon region, from northern Philip-pines to east of Japan noted in SSM/I, is notwell depicted in the three reanalyses. In thecase of NCEP (Fig. 5b), the magnitudes of thepotential function and moisture transportsare substantially smaller than those of SSM/Iover the entire oceanic area, indicating weakmoisture convergence and divergence in the


NCEP. No distinct moisture convergence re-gions are found in the equatorial regions fromthe Bay of Bengal to the Philippine Sea, the In-tertropical Convergence Zone (ITCZ) over theequatorial western Pacific, the South PacificConvergence Zone (SPCZ), and the East Asianmonsoon region along the northwest rim ofthe western Pacific Subtropical High, wherestrong water flux convergences are noted inthe SSM/I.

On the other hand, the magnitude of mois-ture transports of the ERA40 is larger thanthose of NCEP, but substantially smaller than

suggested from SSM/I (Fig. 5c) although thegeneral patterns of flux convergence are quitesimilar to SSM/I. In contrast much similarpatterns as well as magnitudes are found be-tween the ERA40 and the JRA25 (Fig. 5c andFig. 5d).

Summarizing the different features discussedfor Fig. 5, the 13 year (1988–2000) MJJASmean F, QD, and ‘ � QD differences betweenthe SSM/I and three reanalyses are shown inFig. 6. The difference map between NCEP andSSM/I (i.e., NCEP minus SSM/I) depicts addi-tional southward moisture transports over the

Fig. 5. Thirteen year (1988–2000) MJJAS mean divergent component of water vapor transports½QD� (arrow) derived from (a) SSM/I, (b) NCEP, (c) ERA40, and (d) JRA25, respectively. The con-tours indicate the water vapor transport potential function ½F� with the intervals in1:0 � 108 kg s�1. ‘ � QD is shown in shaded area with the intervals in 2 mm day�1.


entire ocean, implying that the NCEP producesweaker northeastward transport over the In-dian Ocean and northwestward transport overthe Pacific Ocean compared with the SSM/I—Fig. 6a. On the other hand, the ERA40 minusSSM/I (Fig. 6b) shows smaller spatial gra-dients of F fields over most of the IndianOcean, suggesting that the ERA40 moisturetransports over the Indian Ocean more resem-ble those inferred from SSM/I. It is noted thatdifferential southward transports are preva-lent over the southern hemispheric Indian andPacific Oceans, indicating that northward

transports in the ERA40 are much weakerover those oceanic areas than those inferredfrom SSM/I. The JRA25 (Fig. 6c) also showsvery similar differential patterns to those inthe ERA40, resulting in weaker northwardwater vapor transports over the southernhemispheric oceans and weaker northwest-ward transports over the North Pacific.

5. Water vapor transports in theclimatological intraseasonal variation

One of dominant phenomena over the Asianmonsoon regions is the climatological intra-

Fig. 6. Thirteen year (1988–2000) MJJAS mean divergent component of water vapor transport ½QD�differences between SSM/I and (a) NCEP, (b) ERA40, and (c) JRA25, respectively (arrow). The con-tours indicate differences of the water vapor transport potential function ½F� with the intervals0:5 � 108 kg s�1. Differences of ‘ � QD are shown in the shaded area with the intervals in2 mm day�1.


seasonal oscillations (CISO) (Nakazawa 1992;Wang and Xu 1997; Kang et al. 1999; LinHoand Wang 2002). The CISO, which differs fromtransient intraseasonal oscillation such as theMadden-Julian Oscillation (Madden and Julian1971), results from phase-lock to the seasonalcycle of the transient intraseasonal variation.It may be viewed as a result of the regulationof the transient intraseasonal oscillation bythe annual cycle (Wang and Xu 1997). It iswell known that because of the intraseasonaloscillation, the South Asian summer monsoonexhibits substantial subseasonal variability,often referred to as the active/break phases ofthe monsoon. The intraseasonal variations inthe East Asian monsoon region are best re-flected into the northward migration of the ma-jor subtropical rain band. The latter is directlyconnected to the convergence or divergence ofwater vapor transport over the East Asian re-gion. Thus, it is of great interest to examinehow the climatological intraseasonal oscillationof water vapor transports from reanalysisproducts compare with SSM/I-inferred fea-tures in the Asian monsoon region. In doing sowe use the 20–95 day filtered data becausethere are various intraseasonal time scales[e.g., 20–80 days (Lau and Yang 1996) and15–95 days (Jiang et al. 2004)], being domi-nant in the Indian and East Asian monsoonregions.

Examining the latitudinal changes with timeover the Indian monsoon region, ‘ � QD wasaveraged over the (80�–90�E) belt and is pre-sented as the time-latitude diagrams in Fig. 7.The SSM/I water vapor divergence field showsthat there is an alternating divergence and con-vergence of water vapor mainly over the oceanfrom 10�S to 20�N (Fig. 7a), with northwardmoving patterns. From May to September, fourdominant northward propagating moisture con-vergence (divergence) cycles are noted, at about30-day time intervals.

The latitude-time evolutions of the 20–95day filtered water vapor flux divergence overthe 80�–90�E longitude were also obtainedfrom three reanalyses—see Figs. 7b–d. TheCISO patterns from 10�S to 20�N shown inNCEP (Fig. 7b) are in general agreement withthose from SSM/I, clearly showing the featuresof northward propagation. But the signals aremuch weaker in comparison to what is sug-

gested from SSM/I. Furthermore, the signalsappearing in SSM/I over the equatorial lati-tudes are not clear in the NCEP reanalysisproducts.

The ERA40 (Fig. 7c) also exhibits well-defined northward propagating features ofvapor flux divergence/convergence, similar tothose noted in the SSM/I. It can be seen thatthe spatial and temporal patterns of the CISOin the region between 10�S and 20�N areclearer than those in NCEP throughout theanalysis period. Nonetheless the magnitudesare still much smaller when compared withSSM/I.

As shown in the 13 year mean water vaportransports, the JRA25 (Fig. 7d) appears tohave similar CISO patterns to those in theERA40. In particular, a strong divergence re-gion propagating from 10�S to 30�N, in the pe-riod from mid June to late July, is similar tothe ERA40, although such a pattern is notshown in SSM/I. Less prominent CISO signalpropagating northward is found during Mayand in the period from late July to August. Themagnitudes of flux divergence in JRA25 appearto be weaker than those of SSM/I, but similarto the ERA40.

Comparisons of water vapor flux divergencesð‘ � QDÞ were also made over the East Asianmonsoon region. In doing so, flux divergenceswere averaged over the 125�–135�E longitudebelt and the results are given as time-latitudediagrams in Fig. 8. It shows that the CISO sig-nals of water vapor flux divergence are also ev-ident over the East Asian region. In the SSM/I(Fig. 8a), the northward propagating CISOsignals appear to extend from 5�S to 40�N dur-ing the summer. Three dominant northwardpropagating moisture convergence (divergence)cycles are found during the May–Septemberperiod. From May to mid-July a strong mois-ture convergence area propagates northwardfrom 10�N to 40�N latitude. The next conver-gence area propagates north from about theequator to 40�N during the early July-late Au-gust period. The third one seems to propagatefrom mid-August to late September with a com-paratively weak magnitude. Among those threenorthward moving convergence patterns, thefirst one seems to be related to the evolutionof the East Asian monsoon rain band associ-ated with the Mei-Yu and Baiu around mid-


June and the Changma in late June (Lau andLi 1984; Lau et al. 1988; Sohn et al. 2001). Ithas been known that along the longitude band122.5�–132.5�E there are four major episodesof northward propagation of outgoing longwave

radiation (OLR) anomalies from May to Octo-ber (Wang and Xu 1997). Among those four epi-sodes, the first three occur in the May to Sep-tember period, and are consistent with thosedescribed in the SSM/I.

Fig. 7. The Hovmoller diagram of ‘ � QD for (a) SSM/I, (b) NCEP, (c) ERA40, and (d) JRA25. Valueswere averaged in the longitudinal belt from 80�E to 90�E. The contour interval is in 0.5 mm day�1

and the shaded area represents ‘ � QD less than �0.5 mm day�1.


Similar features of the CISO are found inthe NCEP along the latitude from 10�N to40�N (Fig. 8b). However, the moisture flux di-vergence and convergence are much smallercompared to the SSM/I values. The generalfeatures of the CISO in both ERA40 andJRA25 are similar to each other and are in

good agreement with those of SSM/I (Figs. 8c–d), although the JRA25 CISO signal tends tobe weaker than ERA40. Three dominant north-ward propagating moisture convergence (diver-gence) cycles are clearer than those in NCEP,but both ERA40 and JRA25 show relativelyweak northward propagating signals compared

Fig. 8. Same as in Fig. 7 except for averaged from 125�E to 135�E.


with SSM/I values. In particular, weak signalsare evident in the later part of August and dur-ing September. These weak signals may be inpart due to characteristic differences betweensatellite observation and reanalysis. Since thisperiod is subject to large interannual variabil-ity of typhoon-related precipitation, reanalysismay not well represent such mesoscale heavyrain events that can be well depicted by satel-lite observations.

It should be noted that the CISO signals de-rived from a 13-year climatology should be rig-orously tested. In Wang and Xu (1997) (WX97hereafter), three statistical methods (sign test,t-test, and Monte Carlo test) were applied toshow the statistical significance of the CISOderived by a 17-year mean OLR annual cycle.Further, they showed that the CISO revealedby OLR has a coherent dynamic structurewith the CISO derived from NCEP reanalysis.The four CISO cycles during May to Octoberdetected in WX97 have been confirmed later byusing longer climatological data sets (Wang etal. 2005). While the statistical test was not re-petitively performed here, the consistence be-tween the SSM/I derived moisture flux conver-gence and the results of WX97 derived fromOLR and wind fields lends support to the cred-ibility of the SSM/I analysis here.

In summary, it is shown that the moistureconvergence and divergence from SSM/I mea-surements clearly depict northward propagat-ing features in the CISO time scales over theIndian summer monsoon region as well as inthe East Asian monsoon region. The NCEP,ERA40, and JRA25 reanalyses show north-ward propagating features similar to thosefound in the SSM/I; however, the magnitudesof moisture flux divergence/convergence in thethree reanalyses appear to be generally weak.Amongst three reanalyses, the ERA40 seemsto be most comparable to the SSM/I.

6. Summary and conclusions

Water cycle parameters (precipitation, evap-oration, and water vapor transport) in theatmospheric branch from NCEP, ERA40, andJRA25 reanalysis were compared with SSM/I-derived estimates over the Asian summermonsoon region, in terms of moisture sourcesand sinks, and their associated transport fea-tures. Satellite-derived SSM/I products were

considered to be a reference for the comparison.We have identified the common features anddiscrepancies among various reanalyses andsatellite-based results, and explained why suchdiscrepancies may exist.

In the 13 year (1988–2000) MJJAS meanclimatology from the SSM/I, the water vaporsource regions are located over the subtropicalIndian and Pacific Oceans in the SouthernHemisphere, and over the North Pacific Ocean.On the other hand, moisture sink regions arelocated in the Asian monsoon trough, the equa-torial convective zones over the Pacific Oceanand the East Asian monsoon region from thenorthern Philippines to east of Japan. Thus,the resultant divergent water vapor trans-port ðQDÞ is characterized by cross-equatorialtransports of water vapor from the subtropicaloceans in the Southern Hemisphere and watervapor transports along the southwesternboundary of the North Pacific high. These areconsistent with well-known large-scale circula-tion features over the Asian summer monsoonregion, i.e., (1) the cross-equatorial flow associ-ated with monsoon circulation over the IndianOcean, (2) the southeast flow over the tropicalwestern North Pacific, and (3) cross-equatorialflow from the east of Australia. In conjunctionwith these water vapor sources and sinks, theclimatological intraseasonal variations of themoisture flux convergence over the Asian sum-mer monsoon regions clearly show northwardpropagating patterns in the South Asia andthe East Asian sectors.

The general long-term summer mean fea-tures found in all three reanalyses are in goodagreement with those from SSM/I; however,much weaker northwestward transports overthe Pacific Ocean north of 10�N are found incomparison to SSM/I. Thus, the strong conver-gence over the East Asian monsoon regionfrom the western Pacific to the east of Japanshown in SSM/I is not well described in any ofthree reanalyses. The salient findings are sum-marized as follows:

(1) NCEPIn the summer mean climatology, the mag-

nitudes of the moisture transports are muchsmaller than those of SSM/I over the wholeocean area, and thus indicate weaker moistureconvergence and divergence. It is also shown


that the CISO magnitudes are much weaker,even though the northward propagating pat-terns are similar to those found in the SSM/Iover the Asian summer monsoon region.

(2) ERA40The magnitudes of the moisture transports

shown in the summer mean climatology arecomparable to those of SSM/I. General pat-terns of flux convergence are quite similar toSSM/I. The CISO patterns are also similar tothose in SSM/I, and magnitudes of the north-ward propagating signals are more comparableto SSM/I, in comparison to NCEP and JRA25reanalyses.

(3) JRA25The JRA25 is much similar to the ERA40 in

its spatial patterns as well as in the magnitudeof its divergent component of water vaportransports. The northward moving patterns ofCISO are also similar to the ERA40 over theAsian summer monsoon region.

It is important that the reanalysis products,particularly hydrological variables such asprecipitation and evaporation, and moistureflux divergence, are compared and assessedagainst satellite-based observations or someother ground-based observations. This paperpresents a new effort to describe the mean fea-tures of divergent components of water vaportransport and its climatological intraseasonalvariation over the Asian summer monsoonregion from a satellite-based observationalapproach.

The ramification of the results here is farreaching. It implies that the water vapor bud-get drawing upon the reanalysis datasetsmay have considerable uncertainties. Cautionshould be exercised when quantitative assess-ment is attempted. In a recent analysis of thewater vapor budget in the East Asian summermonsoon region (10�N–50�N, 90�E–145�E) forthe period of May–August 1998, Wang andYang (2006) found that the moisture transportfrom the Bay of Bengal and from the Philip-pine Sea into the East Asian domain differ by20–30% between the NCEP/DOE reanalysis-2and the ERA40 datasets; further, the totalmoisture convergence into the East Asian sum-mer monsoon domain differ by 47% betweenthe two reanalyses. These uncertainties result

in remarkable uncertainties in the regional cli-mate modeling of the East Asian summer mon-soon in which the regional climate model solu-tion critically relies on the large scale lateralboundary forcing fields.

Acknowledgements

The authors would like to thank Dr. YukiHonda of the Japan Meteorological Agency,the editor of the JMSJ, and two anonymousreviewers for their constructive and valuablecomments which led to improvements in thequality of the paper. This research has beensupported by the SRC program of the KoreaScience and Engineering Foundation, and alsoby the BK21 Project of the Korean Govern-ment. This research was partially performedwhile the first author was resident at the Inter-national Pacific Research Center, University ofHawaii at Manoa. Bin Wang acknowledges thesupport from the International Pacific ResearchCenter which is in part supported by the Fron-tier Research Center for Global Changes. Theauthors wish to thank NCEP/NCAR andECMWF for releasing reanalysis data. TheJRA25 data were obtained from the Japanese25-year Reanalysis Project Web site (http://jra.kishou.go.jp).

References

Annamalai, H., J.M. Slingo, K.R. Sperber, and K.Hodges, 1999: The mean evolution and vari-ability of the Asian summer monsoon: Compar-ison of ECMWF and NCEP-NCAR reanalyses.Mon. Wea. Rev., 127, 1157–1186.

Bengtsson, L., K.I. Hodges, and S. Hagemann, 2004:Sensitivity of large-scale atmospheric analysesto humidity observations and its impact on theglobal water cycle and tropical and extratrop-ical weather systems in ERA40. Tellus, 56A,202–217.

Chen, T.C., 1985: Global water vapor flux and main-tenance during FGGE. Mon. Wea. Rev., 113,1801–1919.

Chou, S.H., 1993: A comparison of airborne eddycorrelation and bulk aerodynamic methods forocean-air turbulent fluxes during cold-air out-breaks. Bound.-Layer Meteor., 64, 75–100.

Chou, S.H., E. Nelkin, J. Ardizzone, and R.M. Atlas,2004: A comparison of latent heat fluxes overglobal oceans for four flux products. J. Climate,17, 3973–3989.

Chou, S.H., E. Nelkin, J. Ardizzone, R.M. Atlas, andC.L. Shie, 2003: Surface turbulent heat and


momentum fluxes over global oceans based onthe Goddard Satellite Retrievals, Version 2(GSSTF2). J. Climate, 16, 3256–3273.

Cohen, J.L., D.A. Salstein, and R.D. Rosen, 2000: In-terannual variability in the meridional trans-port of water vapor. J. Hydrometeor., 1, 547–553.

Feng, L. and J. Li, 2006: A comparison of latent heatfluxes over global oceans for ERA and NCEPwith GSSTF2. Geophys. Res. Lett., 33, 10.1029/2005GL024677.

Holm, E., E. Andersson, A. Beljaars, P. Lopez, J.F.Mahfouf, A. Simmons, and J.N. Thepaut, 2002:Assimilation and Modeling of the HydrologicalCycle: ECMWF’s Status and Plans, ECMWFTechnical Report 383. ECMWF, Reading, UK.

Jiang, X., T. Li, and B. Wang, 2004: Structures andmechanisms of the northward propagating bo-real summer intraseasonal oscillation. J. Cli-mate, 17, 1022–1039.

Kalnay, E. and Coauthors, 1996: The NCEP/NCAR40-year reanalysis project. Bull. Amer. Meteor.Soc., 77, 437–471.

Kang, I.S., C.H. Ho, Y.K. Lim, and K.M. Lau, 1999:Principal modes of climatological seasonal andintraseasonal variations of the Asian summermonsoon. Mon. Wea. Rev., 127, 322–340.

Lau, K.M. and M.T. Li, 1984: The monsoon of EastAsia and its global associations-A survey. Bull.Amer. Meteor. Soc., 65, 114–125.

Lau, K.M., G.J. Yang, and S.H. Shen, 1988: Seasonaland intraseasonal climatology of summer mon-soon rainfall over East Asia. Mon. Wea. Rev.,116, 18–37.

Lau, K.M. and S. Yang, 1996: Seasonal variation,abrupt transition, and intraseasonal variabil-ity associated with the Asian summer monsoonin the GLA GCM. J. Climate, 9, 965–985.

Lau, K.M. and coauthors, 2000: A report of the fieldoperations and early results of the South ChinaSea Monsoon Experiment (SCSMEX). Bull.Amer. Meteor. Soc., 81, 1261–1270.

Li, C. and M. Yanai, 1996: The onset and interan-nual variability of the Asian summer monsoonin relation to land-sea thermal contrast. J. Cli-mate, 9, 358–375.

LinHo. and B. Wang, 2002: The time-space struc-ture of the Asian-Pacific summer monsoon: Afast annual cycle view. J. Climate, 15, 2001–2019.

Madden, R.A. and P.R. Julian, 1971: Detection of a40–50 day oscillation in the zonal wind in thetropical Pacific. J. Atmos. Sci., 28, 702–708.

McBride, J.L., B.W. Gunn, G.J. Holland, T.D.Keenan, N.E. Davidson, and W.M. Frank,1989: Time series of total heating and moisten-ing over the Gulf of Carpentaria radiosonde

array during AMEX. Mon. Wea. Rev., 117,2701–2713.

Mo, K.C. and R.W. Higgins, 1996: Large-scale atmo-spheric moisture transport as evaluated in theNCEP/NCAR and the NASA/DAO reanalyses.J. Climate, 9, 1531–1545.

Murakami, T. and J. Matsumoto, 1994: Summermonsoon over the Asian continent and westernNorth Pacific. J. Meteor. Soc. Japan, 72, 719–745.

Nakazawa, T., 1992: Seasonal phase lock of intrasea-sonal oscillation during the Asian summermonsoon. J. Meteor. Soc. Japan, 70, 597–611.

Newman, M., P.D. Sardeshmukh, and J.W. Bergman,2000: An assessment of the NCEP, NASA, andECMWF reanalyses over the tropical west Pa-cific warm pool. Bull. Amer. Meteor. Soc., 81,41–48.

Onogi, K., H. Koide, M. Sakamoto, S. Kobayashi, J.Tsutsui, H. Hatsushika, T. Matsumoto, N. Ya-mazaki, H. Kamahori, K. Takahshi, K. Kato,T. Ose, S. Kadokura, and K. Wada, 2005:JRA-25: Japanese 25-year re-analysis project-progress and status. Quart. J. Roy. Meteor.Soc., 131, 3259–3268.

Rasmusson, E.M. and K.C. Mo, 1996: Large-scale at-mospheric moisture cycling as evaluated fromNMC global analysis and forecast products. J.Climate, 12, 3276–3297.

Roads, J., M. Kanamitsu, and R. Stewart, 2002: CSEWater and energy budgets in the NCEP-DOEreanalysis II. J. Hydrometeor., 3, 227–248.

Rosen, R.D., D.A. Salstein, and J.P. Peixoto, 1979:Variability in the annual fields of large-scaleatmospheric water vapor transport. Mon. Wea.Rev., 107, 26–37.

Simmons, A.J. and J.K. Gibson, 2000: The ERA-40project plan, ERA-40 Project Report Series No.1, ECMWF, Reading, UK, 62 pp.

Sohn, B.J., H.S. Chung, D.H. Kim, D. Perkey, F.R.Robertson, and E.A. Smith, 2001: Use ofsatellite-derived water vapor data to investi-gate northwestward expansion of North Pacificsubtropical high during 1995 summer. J. Me-teor. Soc. Japan, 79, 1059–1075.

Sohn, B.J., E.A. Smith, F.R. Robertson, and S.C.Park, 2004: Derived over-ocean water vaportransports from satellite-retrieved E � P data-sets. J. Climate, 17, 1352–1365.

Trenberth, K.E. and C.J. Guillemot, 1995: Evalua-tion of the global moisture budget as seen fromanalyses. J. Climate, 8, 2255–2272.

Trenberth, K.E. and C.J. Guillemot, 1998: Evalua-tion of the atmospheric moisture and hydro-logical cycle in the NCEP/NCAR reanalyses.Clim. Dyn., 14, 213–231.

Wang, B. and X. Xu, 1997: Northern Hemisphere


summer monsoon singularities and climatolog-ical intraseasonal oscillation. J. Climate, 10,1071–1085.

Wang, B., T. Li, Y. Ding, and R. Zhang, 2005: EastAsian–western North Pacific monsoon: A dis-tinct component of the Asian-Australian mon-soon system. In ‘‘The Global monsoon System:Research and Forecast’’, Ed. C.-P. Chang, B.

Wang, and G. Lau, World Meteorological Orga-nization, WMO TD number 1266, pp 545.

Wang, B. and H. Yang, 2006: Hydrological issues inthe regional modeling of the East Asian sum-mer monsoon. Submitted to Clim. Dyn.

Wentz, F.J. and R.W. Spencer, 1998: SSM/I rain re-trievals within a unified all-weather ocean al-gorithm. J. Atmos. Sci., 55, 1613–1627.


Satellite Assessment of Divergent Water Vapor Transport from … · 2008-04-29 · Journal of the...

Documents

Transcript of Satellite Assessment of Divergent Water Vapor Transport from … · 2008-04-29 · Journal of the...