2702 VOLUME 17J O U R N A L O F C L I M A T E

q 2004 American Meteorological Society

Climate Effects of the Deep Continental Stratus Clouds Generated bythe Tibetan Plateau

RUCONG YU

LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

BIN WANG

Department of Meteorology and IPRC, University of Hawaii at Manoa, Honolulu, Hawaii, and LASG, Institute of Atmospheric Physics,Chinese Academy of Sciences, Beijing, China

TIANJUN ZHOU

LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

(Manuscript received 6 May 2003, in final form 28 January 2004)

ABSTRACT

Evidence is presented to show that the maximum annual mean cloud optical depth between 608S and 608Nis located on the lee side of the Tibetan Plateau. This largest cloud optical depth is produced by persistent deepstratus clouds (primarily the nimbostratus and altostratus) during winter and spring. These deep stratus cloudsare generated and maintained by the frictional and blocking effects of the Tibetan Plateau. The plateau slowsdown the overflow, inducing downstream midlevel divergence; meanwhile it forces the low-level flows toconverge downstream, generating sustained large-scale lifting and stable stratification that maintain the thickstratus clouds.

These stratus clouds produce extremely strong cloud radiative forcing at the top of the atmosphere, whichfundamentally influences the local energy balance and climate change. Analysis of the long-term meteorologicalstation observations reveals that the monthly mean anomalous cloudiness and surface temperature vary in tandem.In addition, the surface warming leads to destabilization and desaturation in the boundary layer. This evidencesuggests a positive feedback between the continental stratus clouds and surface temperature through changinglower-tropospheric relative humidity and stratification. It is shown that the positive feedback mechanism is morerobust during the period of the surface cooling than during the surface warming. It is suggested that the positiveclimate feedback of the continental stratus cloud may be instrumental in understanding the long-term climatictrend and variations over East Asia.

1. Introduction

Clouds are important modulators of climate. Distri-bution of clouds has profound impacts on radiative en-ergy balance and in turn on the atmospheric circulationsand climate (Schneider 1972; Hartmann and Short 1980;Hartmann et al. 1992). The effect of clouds on radiationbudget is measured by cloud radiative forcing (CRF),which represents the difference between cloud-free ra-diative fluxes and the average of all-sky observations(Ramanathan et al. 1989). The net CRF at the top ofatmosphere (TOA) depends on the balance between thecloud albedo and greenhouse effects.

The effects of clouds depend on their properties (Slin-

Corresponding author address: Dr. Rucong Yu, LASG, Instituteof Atmospheric Physics, Chinese Academy of Sciences, Beijing100029, China.E-mail: yrc@lasg.iap.ac.cn

go 1989). Deep convective and cirrus clouds play animportant role in regulating sea surface temperature(SST) over the tropical warm pool (Arking and Ziskin1994; Ramanathan et al. 1995; Lau et al. 1997; Waliser1996). Over the warm pool, both shortwave and long-wave CRFs are strong, but the net CRF at the TOA isnegligible because of cancellation between them (Kiehl1994). The low-level marine stratus clouds, on the otherhand, produce net radiative cooling due to the domi-nance of the negative shortwave CRF. These low-levelclouds play a critical part in establishing the coldtongues of SST and the equatorial asymmetry of thetropical convergence zone in the tropical Pacific andAtlantic Oceans (Philander et al. 1996; Yu and Mechoso1999).

So far much attention has been paid to the CRF ofhigh and low clouds, but the impacts of middle stratusclouds have received little attention. Klein and Hart-mann (1993) noticed that all of the regions with a sig-

1 JULY 2004 2703Y U E T A L .

nificant fraction of stratus clouds are located over coldoceans except over China. Distinctive from low marinestratus clouds, the stratus clouds over eastern China areprimarily middle clouds (Yu et al. 2001). In this study,we focus on the climatic impacts of the middle stratusclouds downstream of the Tibetan Plateau. We will de-scribe its unique cloud radiative characteristics and itsformation mechanism.

Another motivation of the present study concerns theclimate feedback of the continental stratus clouds. With-in a climate system the feedback processes that involveclouds and water vapor have foremost influences on theclimate system in response to changes in external forc-ing. However, our current knowledge of cloud feedbackremains inadequate. The current climate model resultssuggest that on a global scale clouds have a weak neg-ative feedback (Cess et al. 1996); yet, the earlier ver-sions of these models yield large inconsistencies in thenature of cloud feedbacks with a majority of the modelssuggesting a positive feedback (Cess et al. 1990). Pre-vious studies on cloud feedback were primarily focusedon tropical ocean regions. Study of the cloud feedbackover midlatitude land area is rare. The Tibetan Plateauprovides a natural laboratory for understanding the feed-back between continental stratus clouds and surfacetemperature. One of the major purposes of the presentstudy is to investigate the nature of this feedback.

After a brief description of the data used in this studyin section 2, we will describe unique features of thecontinental stratus deck downstream of the Tibetan Pla-teau (section 3), elaborate mechanisms responsible forthe formation of these continental stratus clouds (section4), and investigate the nature of the stratus cloud–cli-mate feedback (section 5), as well as its impacts on themean climate and climate variation (section 6). In thelast section, we summarize the major points and discussthe implications of present results on global warming.

2. Data

The data used in this study include meteorologicalstation observations, reanalysis data, and satellite ob-servations. The multidecadal time series of monthlymean total cloud fraction and surface temperature areobtained from the weather database provided by theChinese Meteorological Administration (CMA). To fa-cilitate climate studies, 160 surface stations, which aredistributed reasonably evenly over mainland China, areselected. The monthly data include precipitation, surfacetemperature, and cloud fraction. (The locations of 145stations in eastern China are marked by heavy dots inFig. 7a.) These station data were then interpolated to 183 18 horizontal grids by weighted averaging the stationdata with weights proportional to the inverse of thesquare distances between the grid and stations. The at-mospheric general circulation data are derived from theNational Centers for Environmental Prediction–Nation-al Center for Atmospheric Research (NCEP–NCAR) re-

analysis data (Kalnay et al. 1996) at a horizontal res-olution of 2.58 3 2.58 and a vertical resolution of 17pressure levels. The NCEP–NCAR reanalysis data arecurrently available for the period starting January 1948.However, because the quality of the analysis over Asiamay be low before the 1960s (Yang et al. 2002), thedata used in this study are for 1963–2000.

The shortwave and longwave fluxes at the TOA aretaken from the Earth Radiation Budget Experiment(ERBE) (Barkstrom 1984). The data archive consists ofmonthly mean all-sky and clear-sky longwave and short-wave radiative fluxes at the TOA with 2.58 3 2.58 res-olution available from February 1985 to December1989. Uncertainties in the ERBE data arise mainly frominstrument errors, sampling biases, and errors from theconversion of radiance to irradiance using bidirectionalmodels. Barkstrom et al. (1989) gave an error bound of5 W m22 for the monthly regional fluxes. A more de-tailed analysis of the uncertainties in ERBE data hasbeen made by Rieland and Raschke (1991). Part of theerrors, such as sampling errors and the bidirectionalmodel errors, is random. The impact of random errorsis negligible when time and space averages are per-formed (Yu et al. 1999). The International SatelliteCloud Climatology Project (ISCCP) products (Rossowand Schiffer 1991, 1999) were also used, which providecloud optical thickness and cloud amount for variouscloud types. These cloud types are defined by the visiblespectrometry/infrared (VIS/IR) cloud-top pressure andoptical thickness or by the IR cloud-top pressure alone.Based on cloud top-pressure, clouds are classified intolow-level cloud (including cumulus, stratocumulus, andstratus), middle-level cloud (altocumulus, altostratus,and nimbostratus) and high-level cloud (cirrus, cirro-stratus, and deep convection). The specific cloud typeslisted above are determined further by classifying op-tical thickness into three categories, below 3.6 for cu-mulus, altocumulus, and cirrus, above 23 for stratus,nimbostratus, and deep convection, and between 3.6 and23 for stratocumulus, altostratus, and cirrostratus. TheISCCP D2 dataset (Doutriaux-Boucher and Seze 1998),used to display the cloud properties related to cloudtypes, is available as monthly means from 1984 to 2000at 2.58 3 2.58 resolution.

The surface-observed cloud amounts suffer from ar-tificial errors and the accuracy of the ISCCP cloud prop-erties are limited by intersatellite calibration error andby errors from estimating cloud properties from radi-ances. Changes in the ISCCP prcessing produced sig-nificant differences between versions of the ISCCP da-taset (Doutriaux-Boucher and Seze 1998), and a numberof artifacts remain in the D2 data. We have examinedthe consistency between the cloud amounts observedfrom the surface station and from satellites. The inter-annual variations of total cloudiness obtained from theISCCP data and from the station observations are ingood agreement. Although systematic deviation be-tween the two datasets exists, it is well within the normal

2704 VOLUME 17J O U R N A L O F C L I M A T E

FIG. 1. Annual-mean total cloud fraction (percentage) averagedover the Sichuan basin (278–328N, 1038–1088E) for the period 1984–2000. Solid line represents cloudiness observed by weather stations.Dashed lines represent the total cloud amount derived by summingall individual clouds in the ISCCP data.

uncertainty contained in current cloud observations.Figure 1 shows the annual variations of total cloud per-centage from 1984 to 2000 over the Sichuan basin,which is adjacent to Tibet (278–328N, 1038–1088E). Thetotal cloud amount derived by adding all individualclouds in the ISCCP data is systematically (about 2%)larger than the total cloudiness observed by surface me-teorological station (Fig. 1). The correlation coefficientbetween them is 0.90. This consistency adds confidenceto our analysis of the long time series of the surface-observed cloudiness data, although the consistency be-tween cloud amount from surface stations and satellitedata is not enough to entirely dispel the limitation ofboth data. The agreement in the interannual variationsbetween the two cloud datasets should result from thetime and space averages, which result in random errorsbeing negligible. Previous studies have also shown thatthese two independent cloud datasets could producecomparable climatological patterns in cloud distributionand cloud movement in low and middle latitudes (Lauand Crane 1997; Hahn et al. 2001).

3. Cloud–radiative forcing downstream of theTibetan Plateau

Based on ISCCP data, 1991–2000 mean cloud opticalthickness and fractional amount of the nimbostratus andaltostratus clouds are shown in Figs. 2a and 2b, re-spectively. The maximum cloud optical depth in theglobal Tropics and midlatitudes between 608S and 608Nis located downstream of the Tibetan Plateau (Fig. 2a).Note that the maximum total amount of clouds is notlocated downstream of the plateau. In fact, the annual-mean total cloud fraction over eastern China (about65%) is less than those over the western Pacific warmpool (where deep cumulus/anvil clouds prevail) andover the eastern Pacific (where marine stratus persists);both are about 75%. Why is the cloud optical depth overthe eastern flank of the Tibetan Plateau the largest?

We found that the largest cloud optical depth down-stream of the plateau is primarily attributed to its cloudproperties. Downstream of the plateau, nimbostratus andaltostratus clouds prevail (Fig. 2b). The fractional cov-erage of nimbostratus and altostratus clouds reaches amaximum just east to the Tibetan Plateau.

Figure 3 further compares two longitudinal cross sec-tions of cloud fraction for different cloud types, one isalong 298N from 1008 to 1208E and the other along theequator from 1408E to 908W. Along 298N and to theeast of the Tibetan Plateau, especially over the Sichuanbasin (1038–1088E), the amount of nimbostratus andaltostratus clouds exceeds the amounts of all other typesof clouds that have large optical thickness (e.g., deepconvective, low-level stratus, and stratocumulus clouds)(Fig. 3a). In contrast, along the equator the marine stra-tus (stratus and stratocumulus) dominates in the easternPacific, whereas the cirrus and cirrostratus clouds pre-vail in the western Pacific (Fig. 3b).

The nimbostratus and altostratus clouds are respon-sible for the extremely large cloud optical thickness ofthe Tibetan Plateau stratus cloud deck. This propositionis further confirmed by the results shown in Figs. 4 and5. As shown in Fig. 4, the nimbostratus and altostratusclouds exhibit the best correlation with the total cloudoptical thickness. Figure 5 compares zonal variations ofcloud optical depth and nimbostratus–altostratus cloudamount along 298N from 1008 to 1208E. Obviously, thenimbostratus and altostratus clouds dominate the zonalvariation of cloud optical depth (Fig. 5a). Over the Sich-uan basin (278–328N, 1038–1088E), the annual variationof the thick stratus cloud amount is in tandem with thatof the cloud optical thickness (Fig. 5b), confirming thatthe nimbostratus and altostratus clouds are major con-tributors to the large cloud optical thickness. The cor-relation coefficient between annual-mean variations ofthe nimbostratus–altostratus cloud amount and the cloudoptical thickness in Sichuan basin is 0.674 from 1984to 2000.

4. Formation mechanism of the nimbostratus andaltostratus clouds

Nimbostratus and altostratus clouds generally covera vast area and have a bulky vertical extent. The nim-bostratus is sometimes thick enough to block out en-tirely the direct solar beam. The nimbostratus cloudshave low cloud base and considerable vertical devel-opment, bringing the tops into the middle troposphericlevel. The nimbostratus and altostratus often form whenstably stratified moist air is forced by steady mechanicallifting over a large area. This often happens at a warmor cold front with gentle slope, occlusion, or in thepresence of other large-scale forcing. Usually the at-mosphere is stably stratified, and turbulence mixing inthe clouds is weak.

We put forward that the persistent nimbostratus andaltostratus clouds over subtropical East Asia result from

1 JULY 2004 2705Y U E T A L .

FIG. 2. 1991–2000 (a) mean cloud optical thickness with contour interval 3, (b) the fractionalamount of the nimbostratus and altostratus clouds with contour interval 5%, and (c) the 1985–1999 mean net CRF at the TOA with contour interval 10 W m22. The data used in (a) and (b)are derived from the ISCCP and the data used in (c) are from the ERBE.

the blocking and frictional effects of the Tibetan Plateau.During most of the year, in particular from Novemberto May, the Tibetan Plateau is continuously exposed totropospheric westerlies. The elevated plateau bifurcatesupstream low-level westerly flows and forces the sur-rounding flows to converge downstream. Meanwhile,the plateau also slows down the midtropospheric west-erlies that flow over its mountainous surface, resultingin downstream midlevel divergence. The low-level con-vergence sustains large-scale steady lifting, while themiddle tropospheric divergence confines the lifting tothe lower troposphere.

Based on NCEP–NCAR reanalysis data, Fig. 6ashows the 1991–2000 annual mean westerly speed at500 hPa and winds at 850 hPa. It is clear that the west-erly flow is considerably slowed down when flowingover the Tibetan Plateau, which induces a strong di-vergence in the middle troposphere leeward of the pla-teau. The 850-hPa surrounding flows converge down-stream of the plateau. The 1991–2000 annual-mean ver-tical profile of divergence averaging in the region (278–328N, 1038–1188E) is shown in Fig. 6b, which clearlydisplays the low-level convergence and middle and hightropospheric divergence downstream of the Tibetan Pla-

2706 VOLUME 17J O U R N A L O F C L I M A T E

FIG. 3. Longitudinal cross sections of cloud fraction for differentcloud types along (a) 298N from 1008 to 1208E and (b) along theequator from 1408E to 908W in units of percentage, which includemiddle stratus clouds (the nimbostratus and altostratus) (crosses),low-level stratus and stratocumulus (filled squares), deep convectiveclouds (filled circles), cirrus and cirrostratus (open circles), and cu-mulus and altocumulus (open squares), derived from the ISCCP D2data, averaged from 1991 to 2000.

FIG. 4. Scatterplots of the fractional amount (percentage) of (a) themiddle-level nimbostratus and altostratus clouds, (b) low-level stratusand stratocumulus clouds, and (c) the sum of deep convection andcirrostratus as functions of cloud optical depths over eastern China(258–358N, 1038–1188E) derived from the monthly ISCCP data from1991 to 2000.

teau. Figure 6c shows the relationships among the totalamount of nimbostratus and altostratus clouds and themidlevel divergence. Both variables are presented interms of their mean annual cycles and averaged overthe lee side of the plateau (278–328N, 1038–1088E). Themidtropospheric divergence coincides very well withthe amount of nimbostratus and altostratus cloudsthroughout the year. Results shown in Fig. 6 suggestthat mechanical forcing by the plateau provides a fa-vorable large-scale environment for formation of deepstratus clouds. Figures 5b and 6c also indicate that thecloud optical thickness and nimbostratus–altostratuscloud amount reach their extremes in the cold seasonwhen the westerlies are strongest.

In addition to the mechanical forcing by the plateau,the southern branch of the low-level westerly flows (Fig.6a) is recharged with moisture due to their trajectoriespassing across the warm Indian subcontinent and theBay of Bengal. The moist, southern branch of the low-level westerly is constantly uplifted by the Yun-Gui Pla-teau, a highland extending from the southeast corner ofthe Tibetan Plateau to Indochina. The sustained ascent,increased moisture transport, and stable stratification onthe lee side of the plateau, together provide a suitablelarge-scale condition for maintenance of nimbostratusand altostratus clouds.

1 JULY 2004 2707Y U E T A L .

FIG. 5. (a) Longitudinal cross section of cloud optical thickness(solid) and the percentage of nimbostratus and altostratus cloudamount (dashed) along 298N from 1008 to 1208E; (b) annual cyclesof the averaged cloud optical thickness (solid), and the percentageof the nimbostratus and altostratus clouds (dashed) averaged over theSichuan basin (278–328N, 1038–1088E) from 1991 to 2000. The dataare derived from the monthly ISCCP data.

5. Stratus cloud–climate feedback over the lee sideof the Tibetan Plateau

Over subtropical eastern China, the cloud amount andsurface temperature exhibits a pronounced negative cor-relation. This can be seen from Fig. 7a, which showsthe spatial pattern of the correlation coefficients betweenthe observed anomalous monthly cloud fractions andthe anomalous surface temperatures from 1951 to 2000.The most significant negative correlation is found alongthe Yangtze River valley. Over the Sichuan basin (278–328N, 1038–1088E) where the cloud–radiative forcing isstrongest (Figs. 2, 3, and 5), the correlation coefficientof annual mean cloud amounts and surface temperaturesreaches 20.62 for the 50-yr period. Figure 7b showsthe 10-yr running mean surface temperature and cloudfraction for the Sichuan basin from 1956 to 1995. Overthe Sichuan basin (and the southeastern China in gen-eral) the surface temperature variation has a strong op-posite tendency from that of the total cloudiness onmonthly to decadal time scales.

The negative correlation between the surface tem-perature and stratus cloud amount suggests a couplingbetween the surface temperature and the cloud radiativeforcing. It is well understood that the clouds can affectsurface temperature through changing CRF (Chen et al.2003). How does surface temperature affect the conti-nental stratus clouds?

Analysis of the NCEP–NCAR reanalysis data andChinese surface station data reveals that the surface tem-perature is highly correlated with the temperature below850 hPa (correlation coefficient reaches 0.72 from 1963to 2000), but is nearly uncorrelated to the temperatureabove 600 hPa. This implies that a surface warmingwould destabilize the low troposphere and reduce therelative humidity in the boundary layer because the in-crease of water vapor in the air is slower than the in-crease of the saturation vapor over most of the landarea.

The above analyses lead to two hypothetically posi-tive feedback processes between the cloud physics andlarge-scale dynamics, that is, the stratus–surface tem-perature feedbacks through changing stability and rel-ative humidity. When surface temperature rises, the re-duced relative humidity would reduce stratus cloud frac-tion, allowing more radiative fluxes into the earth systemand resulting in further surface warming. Meanwhile,the surface warming would also reduce the lower-tro-pospheric static stability, which in turn reduces the po-tential for stratus cloud formation and favors furthersurface warming. In addition, the reduction in staticstability could stimulate stronger mixing with drier airabove and could further induce the reduction in low-level humidity, which could enhance the feedback.

To confirm the proposed positive cloud feedbackmechanisms, we examine interannual variations. For theperiod from 1963 to 2000, the correlation coefficientbetween the surface temperature and the mean relativehumidity in the layer between 925 and 700 hPa is 20.48;the correlation coefficient between the surface temper-ature and 850–500-hPa differential potential tempera-tures is 20.42. The annual mean total cloudiness ispositively correlated with the mean relative humidityand the differential potential temperature (stability) withthe correlation coefficients being 0.73 and 0.50, re-spectively.

Figure 8 shows 10-yr running mean relative humidityand the differential potential temperature. Comparisonof Figs. 8 and 7b indicates that a significant negative(positive) correlation between the surface temperature(total cloud amount) and lower-troposphere relative hu-midity or static stability exists on decadal time scales.The strong coupling among the surface temperature,clouds, relative humidity, and static stability indicatesthe important contributions of the stratus cloud feedback(through changing relative humidity and static stability)to the climate variations on the lee side of the plateau.

To better understand the effects of positive stratuscloud feedback on the surface temperature variation, wepresent an example to illustrate how the cloud radiativeforcing changes in response to surface temperaturechanges. Combining the ERBE data, ISCCP data, andChinese station observation data, the surface tempera-ture induced changes in cloud radiative interaction isexamined from 1985 to 1989. Although the ERBE datais available only for this short period, it nevertheless

2708 VOLUME 17J O U R N A L O F C L I M A T E

FIG. 6. (a) The 1991–2000 annual-mean westerly wind speed at 500 hPa (contour interval is 2m s21) and the winds at 850 hPa (vectors with maximal length for 6 m s21). The dashed contoursdenote wind speed below 10 m s21. The shading denotes mountain heights, and the edges of lightand dark shading denote the contours of 2000 and 3500 m, respectively. (b) 1991–2000 annual-mean vertical profile of horizontal divergence averaging in the region (278–328N, 1038–1188E)in units of 106 s21. (c) 1991–2000 mean annual cycles on the lee side of the plateau (278–328N,1038–1088E), including nimbostratus–altostratus cloud amount (percentage) (solid), which is de-rived from the monthly ISCCP data, and the divergence in the midlevel (600–500 hPa) in unitsof 106 s21 (dashed). The circulation data are derived from the NCEP–NCAR reanalysis dataset.

contains a sufficiently large signal in the surface tem-perature that may be associated with the 1986/87 ElNino. As shown in Fig. 9a, changes in net CRF at theTOA positively correlated with changes in the surfacetemperature, and both of them negatively correlate withtotal cloud amount (Fig. 1). Figure 9b shows the yearlymean variations of the stratus cloud amount with largecloud optical depths (including nimbostratus, stratus,altostratus, and stratocumulus), and the associated short-wave CRF at the TOA. Comparison of the results shownin Figs. 9a and 9b suggests that the shortwave radiationforcing of the stratus clouds dominates the changes innet CRF at the TOA. Since the amount of solar radiationenergy absorbed by the atmosphere is small, the surface

radiative forcing change is almost the same as that atthe TOA.

6. Impacts of the Tibetan Plateau stratus clouddeck on east China climate

a. Impacts on mean climate

The negative net CRF of the plateau continental stra-tus exhibits a maximum that is comparable with that ofmarine stratus clouds over the southeastern Pacific (Fig.2c). This is expected from the radiative properties ofthe nimbostratus and altostratus clouds. Correspondingto the largest cloud optical depth, the Tibetan Plateaustratus deck produces strongest shortwave CRF at the

1 JULY 2004 2709Y U E T A L .

FIG. 7. (a) Correlation coefficient between monthly mean anom-alous surface temperatures and total cloud fraction (after removingthe mean seasonal variation) in eastern China from 1951 to 2000.The heavy dots indicate the locations of the surface stations, fromwhich observations are used in this study. (b) Ten-year running meansurface temperatures in units of 8C (dashed line with filled circle)and the total cloud percentage (solid line with open circle) averagedover (278–328N, 1038–1088E) from 1951 to 2000, derived from sur-face station observations.

FIG. 8. Ten-year running mean 925–700-hPa averaged relative hu-midity (percentage) (dotted line with open square) and 500–850-hPavertical differential potential temperature (K) (dashed line with filledsquare) derived from NCEP–NCAR reanalysis products (1963–2000)averaged over (278–328N, 1038–1088E).

FIG. 9. (a) Surface temperature (8C) observed by Chinese weatherstations (solid line with open circle) and the net CRF at TOA (Wm22) derived from the ERBE dataset (dashed line with filled square).(b) Percentage of middle-level/low-level stratus clouds (includingnimbostratus, altostratus, stratus, and stratocumulus) in ISCCP ob-servations (solid line with open circle), referring to the scale outsidethe left ordinate, shortwave CRF at TOA (W m22) (dotted line withfilled square) in ERBE observations. Shown are the yearly meanvalues averaged over the Sichuan basin (278–328N, 1038–1088E) from1985 to 1989.

TOA (not shown). Although the corresponding long-wave CRF is more than twice larger than that of themarine stratus, the net CRF at the eastern flank of theplateau remains to be dominated by the shortwave CRF.Under the clear sky condition, the annual mean netdownward radiative fluxes in this region are around 40W m22. The all-sky net downward radiative fluxes arenegative with a minimum below 220 W m22, which islocated at the Sichuan basin, Therefore, the clouds pro-duce more than 60 W m22 radiative cooling (Fig. 2c).The zonal variation of shortwave CRF and the net CRFalong the Yangtze River coincides well with the zonalvariations in nimbostratus–altostratus clouds and thecloud optical thickness. Their annual cycles also matcheach other quite well (figure not shown), indicating thatboth the strongest shortwave CRF and net CRF are overthe Sichuan basin, especially during the cold season.

The extremely strong negative CRF of the plateaustratus deck has prominent impacts on the climate ineastern China. It affects profoundly the local energybalance. To compensate the radiative cooling inducedby the negative net CRF at the TOA, the atmospheric

2710 VOLUME 17J O U R N A L O F C L I M A T E

FIG. 10. Schematic diagram illustrating the stratus cloud feedbacksthrough change of low-level relative humidity and static stability.

column there must gain energy from the moist staticenergy convergence. Therefore, eastern China becomesan area of energy sink. This is in sharp contrast to othersubtropical regions where the atmosphere exports moiststatic energy (Yu et al. 1999).

b. Impacts on the surface temperature variation

The positive feedback between the surface tempera-ture and the Tibetan Plateau stratus deck may help toexplain the climate variation downstream of the plateau.It is conceivable that the cloud radiative feedback duringthe surface cooling period could be more robust thanthat in the warming period. When the surface cools, theincreased static stability favors stratus cloud formationwhile it restrains deep convection and related cirruscloud formation, which results in more intensified solarradiative cooling that dominates the cloud-inducedgreenhouse warming. However, when the surfacewarms, the induced unstable stratification might, in part,favor cumulus convective clouds that could weaken thepositive cloud feedback.

Figure 9 illustrates more robust positive cloud feed-back in the cooling period. During the warming periodfrom 1986 to 1987, the surface temperature increased0.788C; the net cloud radiative forcing increased by 6W m22 (the total cloud fraction decreased by 1%, andthe stratus cloud amount decreased by 3.3%). Therefore,the cloud radiative feedback is only 7.7 W m22 K21 ifchanges in cloud are entirely related to the changes insurface temperature. On the other hand, during the cool-ing period from 1987 to 1989, the surface temperaturedecreased 0.788C, and the net cloud radiative forcingdecreased by 13 W m22 (the total cloud fraction in-creased by 5.5% and stratus clouds increased by 6.2%).Thus, the cloud radiative feedback amounts to 16.7W m22 K21 if changes in cloud are entirely related tothe changes in surface temperature. Based on this es-timation, the positive cloud radiative feedback duringthe cooling period is obviously stronger than that duringa warming period. The relatively weak cloud radiativefeedback during the warming period could be due to thenonstratus cloud formation. During the warming periodof 1986–87, 70% of the stratus cloud decrease is bal-anced by nonstratus clouds, while during the coolingperiod 1987–89, only 11% of stratus cloud increase isoffset by nonstratus clouds. Thus the change of stratusclouds during the cooling period more effectively dom-inates the total cloudiness variation than that during awarming period.

7. Conclusions and discussions

It is shown that on the eastern flank of the TibetanPlateau the annual mean cloud optical depth exhibits amaximum in the global Tropics and extratropics (be-tween 608S and 608N). This maximum cloud opticalthickness is due to the persistence of the thick nimbo-

stratus and altostratus clouds, in particular during winterand spring. The nimbostratus and altostratus clouds de-termine the cloud radiative properties of the continentalstratus cloud deck in the wake of the Tibetan Plateau.

We propose that the thick stratus cloud deck resultsfrom the blocking and frictional effects of the TibetanPlateau on the prevailing westerlies. The elevated pla-teau bifurcates the upstream low-level westerly flowsand forces the surrounding flows to converge on its leeside. Meanwhile, the plateau also slows down the mid-tropospheric westerlies that flow over its mountainoussurface, resulting in downstream midlevel divergence.As such, the plateau constantly generates and maintainsmiddle stratus clouds, resulting in the maximum cloudoptical depth.

The persistent plateau stratus deck generates strongestcloud radiative forcing at the TOA in the global Tropicsand midlatitudes. It produces about 60 W m22 radiativecooling at TOA over the Sichuan basin. This prominentradiative forcing makes eastern China an area of moiststatic energy sink. This contrasts with most other mid-dle–low latitude regions where the atmosphere exportsmoist static energy.

In the regions covered by the plateau stratus deck,the total cloud amount and the surface temperature ex-hibit a pronounced negative correlation. Surface coolingis found to stabilize the lower troposphere and increasethe relative humidity in the boundary layer and the lowertroposphere. The opposite is true for the surface warm-ing.

Based on these observed facts, we propose that thereexist two positive feedback processes between the sur-face temperature and stratus clouds: one through changeof the relative humidity and the other through changeof the static stability (Fig. 10). Rising surface temper-ature leads to a reduction of the stratification and de-crease of relative humidity; both suppress the formationof continental stratus clouds. Reduction of the stratusclouds would, in turn, reduce cloud radiative coolingand favor further surface warming. Similarly, surfacecooling would increase the amount of stratus clouds andfurther enhance the surface cooling.

The results shown in Fig. 9 suggest that the two pos-

1 JULY 2004 2711Y U E T A L .

itive feedback mechanisms are more robust during theperiod of surface cooling and stratus cloud increase,than during the period of surface warming and stratuscloud decrease. This is because, when the surfacewarms, change of the static stability might favor thedevelopment of cumulus convection (in particular dur-ing summer), which can potentially offset the effects ofthe stratus clouds. Of course, a longer dataset is requiredto confirm these findings.

Downstream of the Tibetan Plateau the moderate sur-face cooling in the last century tends to be at odds withmost of the rest of the world (Li et al. 1995; Folland etal. 2001; Houghton et al. 2001). Previous studies haveemphasized effects of anthropogenic aerosols (Lou etal. 2001; Menon et al. 2002). Considering the criticalrole of clouds in regulating climate and the persistenceof the continental stratus deck, we propose that the dis-tinctive climate trend may be attributed to the uniqueregional cloud radiative properties. Chen et al. (2003)showed a significant increase in the middle and lowcloud amount in the winter half-year over the east pe-riphery of the Tibetan Plateau responding to doubledCO2 simulation and mentioned that cloud radiative in-teraction is the primary cause of the change in surfacetemperature in the regions with elevations below 3 kmaround the Tibetan Plateau. The two positive feedbackmechanisms could be instrumental for understanding thefact that the cooling trends from the 1950s to mid-1980sin the Sichuan basin is much stronger than elsewherein the global Tropics and midlatitudes (Fig. 7b). Becausethe positive cloud radiative feedback during the coolingperiod is more effective, the cooling trends from the1950s to mid-1980s in the Sichuan basin were so strongthat the overall trend is flat during 1950–2000, whereasmost of the global Tropics and midlatitude regions arecharacterized by considerable warming during the sameperiod.

Although the analysis in the present study is focusedon the Yangtze River valley, the conclusions concerningthe cloud–climate feedback may highlight a mechanismthat may apply globally. In middle and low latitudes,the ongoing enhanced warming trend has been attributedto the greenhouse gas effects and water vapor feedback(Yang and Tung 1998; Schneider et al. 1999; Hall andManabe 1999; Larson et al. 1999; Slingo et al. 2000).The water vapor feedback is based on the assumptionthat during surface warming the relative humidity re-mains constant. This assumption is not always supportedby our diagnostic analysis. Although the global-meanwater vapor is positively correlated with surface tem-perature, the mean relative humidity is negatively cor-related with the lower-tropospheric temperature becausethe increase of atmospheric humidity is slower than theincrease of saturation water vapor, especially over themiddle–high latitude continent. Some previous studieshave shown that most existing climate models overes-timate the positive correlation between the surface tem-perature and the temperature above the boundary layer.

They also overestimate the correlation between watervapor and atmospheric temperature in climate variation(Sun and Held 1996; Sun et al. 2001, 2003; Ingram2002). Therefore, coupled ocean–atmosphere modelsmight typically overestimate the positive water vaporfeedback in amplifying the warming effect of increasedgreenhouse gas concentrations. However, Bauer et al.(2002) argued that the discrepancy between observedand simulated correlations between temperature and hu-midity are substantially reduced when consistent sam-pling is introduced, and Soden et al. (2002) providedquantitative evidence of the reliability of water vaporfeedback in current climate models by comparing watervapor feedback in the Geophysical Fluid Dynamics Lab-oratory model and observations following the eruptionof Mount Pinatubo. The uncertainty of humidity in theNCEP reanalysis is arguable to display the long-termglobal changes (Trenberth et al. 2001).

Cloud effects in the current climate models exhibit arange of positive and negative feedback depending onthe physical parameterizations of models. Consequently,model-based predictions of global warming include botha strong amplifying effect of water vapor feedback anda significant uncertainty due to uncertainty of cloudfeedback processes (Schneider et al 1978; Shukla andSud 1981; Wetherald and Manabe 1988). Our study sug-gests that cloud radiative feedback could be positive inthe stratus deck region due to the associated changes instratification and relative humidity in the lower tropo-sphere. It is important to point out that the nature ofthe cloud radiative feedback on surface temperature de-pends on the properties of the clouds. The cloud–climatefeedback could be negative when the albedo effect ofhigh clouds dominates the greenhouse effects. The cloudliquid water feedback could also be important in somecircumstances. The complexity of cloud feedback in dif-ferent area arises from large different responses of dif-ferent cloud types to the changes in surface temperature.Over a cold oceanic area, for example, in the easternequatorial Pacific, the low-level stratus cloud dominatesthe total cloud (Fig. 3b), but the interannual variabilityof total cloud is regulated strongly by cirrus cloud (notshown). The El Nino warming reduces the static stabilityin the lower troposphere, which results in a low-levelstratus cloud decrease and cirrus cloud increase and thecirrus cloud–induced greenhouse effect dominates thechange in the net CRF at TOA. Although the 1987 ElNino warming also involved a positive feedback in theeastern equatorial Pacific; which is consistent with Sunet al. (2003), the physical process is very different fromthat downstream of the Tibetan Plateau. Therefore, re-alistic simulation of cloud properties and amount is crit-ical for correct representation of the cloud feedback inclimate variations.

The present hypothesis is preliminary because of thelimitations of the data used. The uncertainties in theabove analysis may arise from the fact that we did notconsider possible impacts of changes in greenhouse gas-

2712 VOLUME 17J O U R N A L O F C L I M A T E

es, aerosols, and atmospheric dynamics on the cloud–radiation and surface temperature variations. The hy-potheses raised in this study need to be further inves-tigated using numerical experimentations.

Acknowledgments. We would like to thank Leo J.Donner and three anonymous reviewers for their helpfulcomments and suggestions. Rucong Yu and TianjunZhou are jointly supported by the ‘‘Innovation Programof CAS’’ under Grant ZKCX2-SW-210 and the NationalNatural Science Foundation of China under Grants no40233031 and 40221503. Bin Wang acknowledge thesupported by the NOAA/PACS program.

REFERENCES

Arking, A., and D. Ziskin, 1994: Relationship between clouds andsea surface temperatures in the western tropical Pacific. J. Cli-mate, 7, 988–1000.

Barkstrom, B. R., 1984: The Earth Radiation Budget Experiment(ERBE). Bull. Amer. Meteor. Soc., 65, 1170–1185.

——, E. Harrison, G. Smith, R. Green, J. Kibler, R. Cess, and ERBEScience Team, 1989: The Earth Radiation Budget Experiment(ERBE) archival and April 1985 results. Bull. Amer. Meteor.Soc., 70, 1254–1262.

Bauer, M., A. D. Genio, and J. R. Lanzante, 2002: Observed andsimulated temperature–humidity relationships: Sensitivity tosampling and analysis. J. Climate, 15, 203–215.

Cess, R. D., and Coauthors, 1990: Intercomparison and interpretationof climate feedback processes in 19 atmospheric general cir-culation models. J. Geophys. Res., 95, 16 601–16 615.

——, and Coauthors, 1996: Cloud feedback in atmospheric generalcirculation models: An update. J. Geophys. Res., 101, 12 791–12794.

Chen, B., W. C. Chao, and X. Liu, 2003: Enhanced climatic warmingin the Tibetan Plateau due to doubling CO2: A model study.Climate Dyn., 20, 401–413.

Doutriaux-Boucher, M., and G. Seze, 1998: Significant changes be-tween the ISCCP C and D cloud climatologies. Geophys. Res.Lett., 25, 4193–4196.

Folland, C. K., and Coauthors, 2001: Global temperature change andits uncertainties since 1861. Geophys. Res. Lett., 28, 2621–2624.

Hahn, C. J., W. B. Rossow, and S. G. Warren, 2001: ISCCP cloudproperties associated with standard cloud types identified in in-dividual surface observations. J. Climate, 14, 11–28.

Hall, A., and S. Manabe, 1999: The role of water vapor feedback inunperturbed climate variability and global warming. J. Climate,12, 2327–2346.

Hartmann, D. L., and D. A. Short, 1980: On the use of earth radiationbudget statistics for studies of clouds and climate. J. Atmos. Sci.,37, 1233–1250.

——, M. E. Ockert-Bell, and M. L. Michelson, 1992: The effect ofcloud type on earth’s energy balance: Global analysis. J. Climate,5, 1281–1304.

Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van derLinden, X. Dai, K. Maskell, and C. A. Johnson, Eds., 2001:Climate Change 2001: The Scientific Basis. Cambridge Univer-sity Press, 881 pp.

Ingram, W. J., 2002: On the robustness of the water vapor feedback:GCM vertical resolution and formulation. J. Climate, 15, 917–921.

Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.

Kiehl, J. T., 1994: On the observed near cancellation between long-wave and shortwave cloud forcing in tropical regions. J. Climate,7, 559–565.

Klein, S. A., and D. L. Hartmann, 1993: The seasonal cycle of lowstratiform clouds. J. Climate, 6, 1587–1606.

Larson, K., D. L. Hartmann, and S. A. Klein, 1999: The role of clouds,water vapor, circulation, and boundary layer structure in thesensitivity of the tropical climate. J. Climate, 12, 2359–2374.

Lau, K. M., H. T. Wu, and S. Bony, 1997: The role of large-scaleatmospheric circulation in the relationship between tropical con-vection and sea surface temperature. J. Climate, 10, 381–392.

Lau, N. C., and M. W. Crane, 1997: Comparing satellite and surfaceobservations of cloud patterns in synoptic-scale circulation sys-tems. Mon. Wea. Rev., 125, 3172–3189.

Li, X., X. Zhou, W. Li, and L. Chen, 1995: The cooling of Sichuanprovince in recent 40 years and its probable mechanisms. ActaMeteor. Sin., 9, 57–68.

Lou, Y., L. Daren, X. Zhou, W. Li, and Q. He, 2001: Characteristicsof the spatial distribution and yearly variation of aerosol opticaldepth over China in last 30 years. J. Geophys. Res., 106 (D13),14 501–14 513.

Menon, S., J. Hansen, L. Nazarenko, and Y. Luo, 2002: Climateeffects of black carbon aerosols in China and India. Science,297, 2250–2253.

Philander, S. G. H., D. Gu, G. Lambert, T. Li, D. Halpern, N.-C. Lau,and R. C. Pacanowski, 1996: Why the ITCZ is mostly north ofthe equator. J. Climate, 9, 2958–2972.

Ramanathan, V., R. D. Cess, E. F. Harrison, P. Minnis, B. R. Bark-strom, E. Ahmad, and D. Hartmann, 1989: Cloud radiative forc-ing and climate: Results from the Earth Radiation Budget Ex-periment. Science, 243, 57–63.

——, B. Subasilar, G. J. Zhang, W. Conant, R. D. Cess, J. T. Hiehl,H. Grassl, and L. Shi, 1995: Warm pool heat budget and short-wave cloud forcing: A missing physics? Science, 267, 499–503.

Rieland, M., and E. Raschke, 1991: Diurnal variability of the earthradiation budget: Sampling requirements, time integration as-pects and error estimates for the Earth Radiation Budget Ex-periment (ERBE). Theor. Appl. Climatol., 44, 9–24.

Rossow, W. B., and R. A. Schiffer, 1991: ISCCP cloud data products.Bull. Amer. Meteor. Soc., 72, 2–20.

——, and ——, 1999: Advances in understanding clouds from ISCCP.Bull. Amer. Meteor. Soc., 80, 2261–2287.

Schneider, E. K., B. P. Kirtman, and R. S. Lindzen, 1999: Troposphericwater vapor and climate sensitivity. J. Atmos. Sci., 56, 1649–1658.

Schneider, S. H., 1972: Cloudiness as a global climate feedback mech-anism: The effects on the radiation balance and surface tem-perature of variation in cloudiness. J. Atmos. Sci., 29, 1413–1422.

——, W. M. Washington, and R. M. Chervin, 1978: Cloudiness as aclimatic feedback mechanism: Effects on cloud amounts of pre-scribed global and regional surface temperature changes in theNCAR GCM. J. Atmos. Sci., 35, 2207–2221.

Shukla, J., and Y. Sud, 1981: Effect of cloud radiation feedback onthe climate of a general circulation model. J. Atmos. Sci., 38,2337–2353.

Slingo, A., 1989: A GCM parameterization for the shortwave radi-ative properties of water clouds. J. Atmos. Sci., 46, 1419–1427.

——, J. A. Pamment, R. P. Allan, and P. S. Wilson, 2000: Water vaporfeedbacks in the ECMWF reanalyses and Hadley Centre climatemodel. J. Climate, 13, 3080–3098.

Soden, B. J., R. T. Wetherald, G. L. Stenchikov, and A. Robock,2002: Global cooling after the eruption of Mount Pinatubo: Atest of climate feedback by water vapor. Science, 296, 727–730.

Sun, D.-Z., and I. M. Held, 1996: A comparison of modeled andobserved relationships between interannual variations of watervapor and temperature. J. Climate, 9, 665–675.

——, C. Covey, and R. S. Lindzen, 2001: Vertical correlations ofwater vapor in GCMs. Geophys. Res. Lett., 28, 259–262.

——, J. Fasullo, T. Zhang, and A. Roubicek, 2003: A note on theradiative and dynamical feedbacks over the equatorial Pacificcold tongue. J. Climate, 16, 2425–2432.

1 JULY 2004 2713Y U E T A L .

Trenberth, K. E., D. P. Stepaniak, J. W. Hurrell, and M. Fiorino, 2001:Quality of reanalyses in the Tropics. J. Climate, 14, 1499–1510.

Waliser, D. E., 1996: Formation and limiting mechanisms for veryhigh sea surface temperature: Linking the dynamics and the ther-modynamics. J. Climate, 9, 161–188.

Wetherald, R. T., and S. Manabe, 1988: Cloud feedback processes ina general circulation model. J. Atmos. Sci., 45, 1397–1416.

Yang, H., and K. K. Tung, 1998: Water vapor, surface temperature,and the greenhouse effect—A statistical analysis of tropical-mean data. J. Climate, 11, 2686–2697.

Yang, S., K. M. Lau, and K. M. Kim, 2002: Variations of the East

Asian jet stream and Asian–Pacific–American winter climateanomalies. J. Climate, 15, 306–325.

Yu, J.-Y., and C. R. Mechoso, 1999: Links between annual variationsof Peruvian stratocumulus clouds and of SST in the eastern equa-torial Pacific. J. Climate, 12, 3305–3318.

Yu, R. C., M. Zhang, and R. D. Cess, 1999: Analysis of the atmo-spheric energy budget: A consistency study of available datasets. J. Geophys. Res., 104, 9655–9661.

——, Y. Q. Yu, and M. H. Zhang, 2001: Comparing cloud radiativeproperties between the eastern China and the Indian monsoonregion. Adv. Atmos. Sci., 18, 1090–1102.