an evaluation of the proposed mechanism of the adaptive...

An Evaluation of the Proposed Mechanism of the Adaptive Infrared Iris HypothesisUsing TRMM VIRS and PR Measurements

ANITA D. RAPP, CHRISTIAN KUMMEROW, WESLEY BERG, AND BRIAN GRIFFITH

Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado

(Manuscript received 10 September 2004, in final form 15 January 2005)

ABSTRACT

Significant controversy surrounds the adaptive infrared iris hypothesis put forth by Lindzen et al.,whereby tropical anvil cirrus detrainment is hypothesized to decrease with increasing sea surface tempera-ture (SST). This dependence would act as an iris, allowing more infrared radiation to escape into space andinhibiting changes in the surface temperature. This hypothesis assumes that increased precipitation effi-ciency in regions of higher sea surface temperatures will reduce cirrus detrainment. Tropical RainfallMeasuring Mission (TRMM ) satellite measurements are used here to investigate the adaptive infrared irishypothesis. Pixel-level Visible and Infrared Scanner (VIRS) 10.8-�m brightness temperature data andprecipitation radar (PR) rain-rate data from TRMM are collocated and matched to determine individualconvective cloud boundaries. Each cloudy pixel is then matched to the underlying SST. This study examinessingle- and multicore convective clouds separately to directly determine if a relationship exists between thesize of convective clouds, their precipitation, and the underlying SSTs. In doing so, this study addressessome of the criticisms of the Lindzen et al. study by eliminating their more controversial method of relatingbulk changes of cloud amount and SST across a large domain in the Tropics. The current analysis does notshow any significant SST dependence of the ratio of cloud area to surface rainfall for deep convection in thetropical western and central Pacific. Results do, however, suggest that SST plays an important role in theratio of cloud area and surface rainfall for warm rain processes. For clouds with brightness temperaturesbetween 270 and 280 K, a net decrease in cloud area normalized by rainfall of 5% per degree SST wasfound.

1. Introduction

Anthropogenic effects on climate have become anincreasingly important topic in atmospheric and envi-ronmental sciences. Projected climate change has sig-nificant implications for socioeconomic activity and en-vironmental policy. In the Third Assessment Report bythe Intergovernmental Panel on Climate Change(IPCC; Houghton et al. 2001), globally averaged sur-face temperature is projected to rise between 1.4° and5.8°C over the next 100 yr. This increase in temperatureis due largely to the current and predicted emissions ofgreenhouse gases such as carbon dioxide. Carbon diox-ide, however, is only indirectly responsible for the pro-jected increase in temperature. By fixing the atmo-spheric lapse rate, relative humidity, stratospheric tem-

perature, and cloudiness, Rasool and Schneider (1971)computed only a 0.8°C temperature increase in a car-bon dioxide doubling experiment. The real atmosphereresponds to changes in the surface temperature withfeedbacks that further modify the initial forcing causedby the increased carbon dioxide. The large uncertaintyin the IPCC projected temperature increase arisesmostly from the unknown response of these feedbacksin the climate system, especially those due to watervapor and clouds. Raval and Ramanathan (1989), In-amdar and Ramanathan (1994), and Jackson andStephens (1995) all showed that column water vaporincreases with increasing sea surface temperature(SST). Water vapor, much like carbon dioxide, has apositive feedback on temperature because it serves totrap terrestrial radiation in the atmosphere. This watervapor feedback is thought to account for more than halfthe projected warming simulated by climate models(Manabe and Weatherald 1967; Ramanathan 1981; Halland Manabe 1999). Increased water vapor, in turn, islikely to impact clouds whose effect on the climate sys-

Corresponding author address: Anita D. Rapp, Dept. of Atmo-spheric Science, Colorado State University, Fort Collins, CO80523-1371.E-mail: [email protected]

15 OCTOBER 2005 R A P P E T A L . 4185

© 2005 American Meteorological Society

JCLI3528

tem is even less certain. More importantly, convectiveclouds have been postulated (Ramanathan and Collins1991; Lindzen et al. 2001) to have a large impact onregulating the west Pacific SSTs, which rarely exceed303 K (Fu et al. 1986; Waliser 1996). In this region,convection is strongly coupled to the SST (e.g.,Bjerknes 1966; Krueger and Gray 1969; Graham andBarnett 1987; Fu et al. 1994) because of the fact thatwater vapor and latent energy of an air parcel increasewith temperature and allow the parcel to overcome thegravitational potential energy and rise to the upper tro-posphere. The tropical Pacific is therefore ideal to in-vestigate cloud and water vapor feedbacks because ofthis strong relationship between the surface tempera-ture and radiative feedback effects through convection.

The feedback of cirrus clouds produced by convec-tion has been the subject of many studies, although theproposed mechanisms for this feedback differ. The twomost prominent and disputed studies utilizing satellitedata both offer hypotheses for this cirrus cloud feed-back, which this study will term the “thermostat hy-pothesis” (Ramanathan and Collins 1991) and the “irishypothesis” (Lindzen et al. 2001). The thermostat hy-pothesis suggests that as the greenhouse effect in-creases with surface temperature, deep convection pro-duces more reflective cirrus clouds. These more reflec-tive cirrus clouds shield the surface from solar radiationand act like a thermostat to regulate the SST. Ra-manathan and Collins (1991) attempted to predict theclimate implications of the thermostat hypothesis andcalculated a maximum SST between 303 and 305 K, inagreement with observations. They also suggested thatthis thermostat would require carbon dioxide concentra-tions to increase by an order of magnitude before the SSTcould be substantially increased from its current value.

The proposed thermostat hypothesis was met withsignificant criticism. A number of papers (Fu et al.1992; Lau et al. 1994; Wallace 1992; Hartmann andMichelsen 1993) suggested that convection is muchmore sensitive to the large-scale circulation than SST.Most of these papers also suggested that evaporationalcooling is a much more likely candidate for the regula-tion of SST. Pierrehumbert (1995) found that neitherclouds nor evaporational cooling could provide regula-tion of SST. He reported instead on a strong connectionbetween SST and the distribution of dry and moist re-gions in the Tropics.

As suggested by Pierrehumbert (1995), Lindzen et al.(2001; hereafter LCH) investigated changes in the rela-tive areas of high- and low-humidity regions with SST.Dry regions typically occur in regions of large-scalesubsidence, while the moist regions are associated withascent that is concentrated in convective “hot towers”

(Riehl and Malkus 1958). Near the top of the convec-tive tower, ice particles are detrained by the upper-levelair motions to form the cirrus anvil. These ice particlesare detrained downstream of the original cumulustower, and the anvil cirrus cloud grows quite large rela-tive to the size of the cumulus updraft. It is with theseclouds that LCH found an inverse relationship betweencloud area and the cloud-weighted SST in the tropicalwestern and central Pacific. The production of cirrusdepends on how much water is available for detrain-ment into the anvil, which in turn can be thought of asa competition for cloud water between detrainment andprecipitation. Citing earlier findings from Sun andLindzen (1993) that raindrop growth rate increases withtemperature, LCH hypothesized that increases in sur-face temperature result in increased precipitation effi-ciency in the cumulus tower, which leads to a decreasein cirrus detrainment. LCH assumed that on average,the longwave effect of the cirrus would dominate.These cirrus clouds are more transparent to solar ra-diation than the cumulus towers and allow some of theincoming shortwave radiation to reach the surface.Since they exist in the cold, upper levels of the atmo-sphere, however, they emit less radiation to space andtrap terrestrial radiation in the atmosphere, increasingthe greenhouse effect. A decrease in the amount ofdetrained cirrus with increasing temperature wouldtherefore decrease the cirrus cloud greenhouse effectby allowing more outgoing longwave radiation (OLR)to escape into space and thus provide a negative feed-back to regulate the underlying SST. This dependenceof cirrus detrainment on temperature that adapts byopening and closing dry regions to inhibit changes inthe surface temperature, is likened to the eye’s iris,which adapts to different light intensities by increasingor decreasing the size of the iris.

LCH used infrared (IR) brightness temperatures(Tb) from the Geostationary Meteorological Satellite-5(GMS-5) operated by the Japan Meterological Agencyat two different wavelengths, 11 and 12 �m, to deter-mine high cirrus clouds over the tropical west and cen-tral Pacific (30°S–30°N, 130°E–170°W). Satellite 11-�mpixels less than 260 K were considered to be filled withhigh clouds. If the difference between the 11- and 12-�m Tbs was less than 1.5 K, then the high cloud was alsoconsidered to be thick (i.e., a cumulus tower). To esti-mate the cumulus area of a cloud, an 11-�m Tb thresh-old of 220 K was used. The mean cloud amount, Ac, fortheir entire domain of study was represented by

Ac � � An cos�n�� cos�n, �1�

where A is high cloud area and � is the latitude, andsubscript n represents each 1° � 1° region. The cloud-

4186 J O U R N A L O F C L I M A T E VOLUME 18

weighted SST (CWT) for a grid box was then calculatedby

CWT � � AnTn cos�n�� An cos�n, �2�

where Tn is the SST in each grid box. LCH found thathigh-cloud area, cumulus-cloud area, and high-cloudarea normalized by cumulus area all have negative cor-relations with CWT. From their results, they estimatedthe amount of cirrus decreases by 17%–27% per 1°Cdecrease in CWT. This negative correlation led to acooling of �0.45° to �1.1°C in the model used by theauthors.

Both the methodology and the radiative calculationsin the LCH study have been criticized. The criticisms ofthe radiative calculations (e.g., Lin et al. 2002; Cham-bers et al. 2002) focus on the radiative fluxes’ input tothe climate model that LCH used to assess the iris feed-back. Of more interest to the present study are themethodological criticisms. In a series of papers, Hart-mann and Michelsen (2002a,b,c, hereafter HM) sug-gested that the negative correlation between SST andcloud area found by LCH is due to the use of CWT andnot from a negative climate feedback. Using the samedata as LCH, HM showed that deep convective coresare separated from the subtropical cloud variations thatare producing the changes in CWT. This problem arisesbecause the study by LCH was limited to IR data, andthe use of Tb 220 K to represent the convective corearea is inadequate, especially in the subtropical regionswhere convective cores tend to be warmer.

To address criticisms of the methodology adopted byLCH, this study examines individual clouds determinedto be convective using Tropical Rainfall Measuring Mis-sion (TRMM; Simpson et al. 1996) satellite pixel-leveldata, not gridded cloud fraction. The same domain as inLCH is used. Identification of convective cells elimi-nates the possibility that anvil cirrus can be decoupledfrom the convective cores. Rain-rate information fromthe TRMM precipitation radar is used to determine ifthe cloud has a convective core. LCH hypothesized thatthe decrease in the size of the cloud is due to an in-crease in precipitation efficiency with SST. Not onlydoes the rain-rate information allow the determinationof convective activity, but it also allows examination ofthe specific mechanism of the iris using rain rate andcloud area as a proxy for precipitation efficiency. Itshould be noted that this is only a proxy for precipita-tion efficiency and not the ratio of moisture source toprecipitation. The definition used in this paper goestoward examining the specific mechanism of the irisand trying to show that the cloud area and precipitationare related as suggested by LCH. Precipitation effi-ciency is italicized when used in this context throughout

the paper to remind the reader of this definition. Be-sides addressing the issue of convective cloud identifi-cation, this study eliminates the arguable cloud-weighted SST. By developing a dataset of convectiveclouds with information on the cloud size, convectiveactivity, rainfall, and underlying SST of each cloud, thecriticisms by HM of LCH are reexamined, and the hy-pothesized mechanism for the iris is explored. It is notthe aim of this study to quantify any radiative feedbackeffects by the iris, only to determine whether such afeedback exists.

2. Data and methods

Twenty months (January 1998–August 1999) ofTRMM satellite measurements from the same regionused in the LCH study, 30°S–30°N, 130°E–170°W, areexamined to investigate the iris hypothesis. The TRMMsatellite operates in a circular precessing orbit provid-ing only instantaneous “snapshots” of clouds, whichprohibits the examination of the full life cycle of thecloud. However, since the sampling is random and oversuch a long time period, this dataset should providesampling over the entire system’s life cycle and reducethe temporal sampling issues. Pixel-level Visible Infra-red Scanner (VIRS; Kummerow et al. 1998) 10.8-�mbrightness temperatures and precipitation radar (PR)2A25 rain-rate data (Iguchi et al. 2000) are used todetermine individual convective cloud boundaries. TheVIRS instrument has a 2.11-km field of view at nadirand a swath width of 720 km, while the PR has a 4.3-kmfield of view at nadir and 215-km swath width. SinceVIRS and PR have different fields of view, the data arecollocated and matched before analysis. Because it isimpossible to determine the convective activity outsideof the PR swath, all clouds intersecting an edge of theswath or edge of the domain are eliminated from theanalysis. Also, because clouds over land likely respondto more complicated dynamics, they are also elimi-nated. Once the matched PR rain-rate and VIRS IRbrightness temperature dataset is produced, the dataare examined for convective clouds. A pixel is deter-mined to be cloudy if its brightness temperature (Tb) isless than a defined threshold. In the original study byLCH, the existence of the iris was not dependent on theTb threshold chosen. However, since the aim of thisstudy is to reexamine the validity of the iris hypothesisitself, thresholds other than the 260-K value used byLCH are examined. Four Tb thresholds at 250, 260, 270,and 280 K are used to identify discrete cloud clustersand determine if the existence of the iris is dependenton the threshold chosen for cloud identification. De-spite the availability of higher temporal and spatial


resolution SST datasets, the National Centers for En-vironmental Prediction (NCEP) Reynolds Optimal In-terpolation SST (Reynolds and Marsico 1993; Reynoldsand Smith 1994) dataset is used to maintain consistencywith the original study by LCH. Comparing a sample3-day average TRMM Microwave Imager (TMI) SST(Wentz and Meissner 2000) to the correspondingweekly NCEP SST shows that over the entire domainthere is a root-mean-square error of 0.43, which is onlya 1%–2% error depending on the SST. NCEP SSTs areaveraged within each cloud boundary as determined byVIRS to calculate the mean underlying SST for everycloud. This eliminates the need for CWT.

After all the clouds within a swath have been iden-tified, each cloud is examined to determine the exis-tence of contiguous convective cores within the cloud.For this study, convective pixels are defined as thosemeeting a rain-rate threshold of 10 mm h�1 or greater.While this rain rate may seem low, studies usingprofilers and disdrometers (e.g., Tokay et al. 1999)have shown almost all cases of rain rates greater than10 mm h�1 to be convective. It should also be notedthat the PR rain rate is defined for a much larger area(16 km2) than what would be seen using a profiler anddisdrometer. A PR rain rate of 10 mm h�1, therefore,represents a conservative estimate of convection. Forthis study, a convective core is defined by at least onepixel that meets the rain-rate threshold. In cases wherethere is more than one pixel in the cloud meeting thisthreshold, each contiguous area of pixels meeting thethreshold is considered a convective core. A sample ofresults from the cloud and convective core identifica-tion algorithm are shown in Fig. 1. The total rainfall foreach cloud is then calculated by summing the instanta-neous PR rain rates matched to that cloud.

Having identified clouds and their convective cores,they are divided into two categories: single-core andmulticore systems. Examining single-core convectivesystems allows a direct investigation of the proposedmechanism for the iris hypothesis. Since the suggestedmechanism for the iris is that the size of the detrainedanvil cloud decreases due to an increase in precipitationefficiency, it is necessary to isolate the convective pre-cipitating area from the nonconvective precipitatingand nonprecipitating area of the cloud. In examiningmulticore systems, there is no quantitative method toisolate which part of the anvil cirrus results from anygiven convective core. Single-core convective systems,therefore, provide a way to look at this hypothesizedmechanism without being concerned by ambiguitiescaused by systems with more than one convective core.It should be noted that an unintentional result of ex-amining single-core clouds could be to limit results to

convective systems that are not fully developed. How-ever, the numbers of samples, as well as examination ofother systems, should limit the problems associatedwith the spatial sampling of the instruments used in thisstudy.

FIG. 1. (a) VIRS brightness temperature, (b) matched PR rainrates, (c) cloud identification number, and (d) number of convec-tive cores in the cloud.


Fig 1 live 4/C

Though single-core convective systems, as defined bythis study, are ideal for examining the mechanism of theiris, they only account for about 4.5% of rainfall in thedomain of interest. Multicore systems account forabout 91% of the rainfall in this domain and play amuch larger role in the hydrologic cycle, as well as be-ing the dominant source of cloud feedbacks in theTropics. Therefore, these are examined separately. Tointerpret the results of this study, a regression analysisas in LCH is performed.

3. Results

Since LCH assumed that the clouds are respondingto the underlying SST, it is necessary to test the sensi-tivity of the iris to changes in SST regime. SSTs canrespond to locations of the clouds, as well as large-scaleseasonal and interseasonal regime changes. To test thesensitivity of the iris to SST regime, the regressionanalyses are performed on a variety of temporal scales,from monthly to the entire 20-month period. The re-gression analysis is performed for clouds identified ateach of the cloud-top temperature thresholds. As thethreshold is increased, the cloud areas identified bylower Tb thresholds are included, as well as more ofthe cloud edge area. In addition, the LCH domain is

also broken into smaller geographic subdomains, �10°and �20°.

Figure 2 summarizes the monthly slopes and corre-lation coefficients for the four different Tb thresholdsused for cloud identification. This figure shows that therelationship between cloud size normalized by rainfalland SST is extremely variable from month to month.Note that the 280-K threshold is the only one that isfrequently negative. The other thresholds show primar-ily positive correlations, most being less than 0.15, butopposite in sign to the correlations presented by LCH.These consistently negative correlations for 280 K andpositive correlations at the other thresholds indicatethat some type of cloud is observed using the 280-Kthreshold that is not observed by the other thresholds.Because the 280-K threshold is inclusive of all theclouds identified with the lower thresholds, this impliesthat the observed cloud temperature must be warm,between 270 and 280 K, and that the cloud is rainingheavily enough to be considered convective by thisstudy’s definition. One possible scenario that could ac-count for this observation is that as SST increases, theamount of cirrus between 270 and 280 K associatedwith a convective core decreases. However, a compari-son of the percentage of observed cloud area between270 and 280 K shows no substantial decrease with SST.

FIG. 2. (a) Slope and (b) monthly correlation coefficients between SST and single-core convectivecloud size normalized by rainfall for 30°S–30°N, 130°E–170°W.


Fig 2 live 4/C

The single-core clouds examined in this study show thatthis percentage of cloud area is around 90% of the totalarea regardless of the SST. The other possibility is thatheavily raining warm clouds with cloud-top tempera-tures between 270 and 280 K, possibly cumulus conges-tus, are included in this analysis and may be driving thisnegative correlation. Since these warm clouds have alarger effect in the shortwave than in the longwave, apositive feedback would likely result from a reductionin these clouds.

Regressions of SST and cloud size normalized byrainfall amount for the entire 20-month time period areseen in Fig. 3. Again, the 280-K threshold for the wholedomain has a negative correlation. All the correlationcoefficients are near zero regardless of the domain orthreshold. Examination of the confidence interval ofthe slopes shows that the given slopes are valid to�0.002 at the 95% confidence level, indicating that thesign of the slopes and correlations is valid, despite beingvery small. An examination of the three latitudinalbands is shown in Fig. 4 for clouds identified with a Tb

threshold of 260 K. Not surprisingly, this figure showslittle variation with latitude in the weak slopes and cor-relations. Separate examination of data for 1998 and1999 shows similar results despite the end of the strongEl Niño occurring in early 1998. These results, from

Figs. 2–4, thus show no inverse relationship betweenSST and single-core convective cloud size normalizedby rainfall amount on time scales of a year or longer forthe entire domain, as well as the subdomains.

The negative correlation found at the 280-K thresh-old in the overall statistics is examined next. Figure 5shows regressions of SST and single-core clouds iden-tified by this study with all Tbs between 250 and 260 K(Fig. 5a), 260 and 270 K (Fig. 5b), and 270 and 280 K(Fig. 5c). The relationship between these heavily rain-ing warm clouds and SST is much stronger than thatobserved when analyzing all of the convective clouds.This finding is consistent with that of Lau and Wu(2003), who showed that precipitation efficiency in-creases with SST for warm rain systems. The currentstudy suggests that warm cloud area, normalized byrainfall, decreases by approximately 5% per degreeSST. While not shown here, similar results were foundfor the geographical subdomains, �20° and �10°.These results, along with the fact that 76% of the warmrain systems are occurring between �20°, also suggestthat these changes are occurring in the Tropics and arenot being driven by warm subtropical systems.

Since most of the rainfall in the Tropics is from mul-ticore convective systems, not single-core convection,these multicore systems have the most impact on the

FIG. 3. Single-core convective cloud size normalized by rainfall regressed against SST for cloudsidentified with Tb thresholds of (a) 250, (b) 260, (c) 270, and (d) 280 K for Jan 1998–Aug 1999. Theslopes of the regression lines, the correlation coefficients (R), and the number of points included in theregressions are labeled in each panel.


hydrologic and energy cycles and are also investigated.The narrow width of the PR swath limits the number ofmulticore convective systems available for use in thisstudy since 75% of them intersect the edges and it is notpossible to determine the convective activity outside ofthe swath. Because of this, one plot for the entire pe-riod is created.

The multicore convective cloud size normalized bythe amount of rainfall from the cloud regressed againstSST is shown in Fig. 6. This figure shows slightly posi-tive correlations between SST and multicore cloud sizenormalized by rainfall amount for all thresholds and allregions except for clouds with Tb less than 280 K for thewhole domain, and even this correlation is almost zero.

FIG. 4. Single-core convective cloud size normalized by rainfallregressed against SST for clouds identified with a Tb threshold of260 K for (a) 10°S–10°N, (b) 20°S–20°N, and (c) 30°S–30°N. Theslopes of the regression lines, the correlation coefficients (R), andthe number of points included in the regressions are labeled ineach panel.

FIG. 5. Single-core convective cloud size normalized by rainfallregressed against SST for clouds with mean Tbs between (a) 250and 260 K, (b) 260 and 270 K, and (c) 270 and 280 K for Jan1998–Aug 1999. The slopes of the regression lines, the correlationcoefficients (R), and the number of points included in the regres-sions are labeled in each panel.


Examination of the confidence intervals of the slopes atthe 95% level shows that the slopes are valid to within�0.004, which means that the sign of the slopes andcorrelations are valid, except at 280 K. This indicatesthat not only is the size of multicore convective cloudsnot decreasing with SST, but the amount of rainfallfrom the cloud, or precipitation efficiency, is not wellcorrelated with the cloud size.

Results from both the single-core and multicore con-vective cloud analyses do not support the adaptive in-frared iris hypothesis except for the possible cumuluscongestus clouds (270 K Tb 280 K). Results fromcolder clouds show almost no correlation, or even aslightly positive correlation, between the size of a cloudnormalized by its rainfall amount and its underlyingSST. While not shown here, regressing cloud area withSST or cloud area normalized by the core area with SSTresults in similar lack of correlation.

4. Conclusions

The adaptive infrared iris hypothesis suggests thatmoist and dry regions in the Tropics open and close toregulate the outgoing longwave radiation and provide anegative feedback on the sea surface temperature.LCH proposed that as the SST rises, precipitation effi-ciency of convective clouds increases, which results in adecrease in the amount of cirrus detrainment. Using the

TRMM dataset of IR brightness temperatures and ra-dar rainfall estimates, this hypothesis has been directlyexamined.

Defining the effective size of a cloud as the area ofcloud divided by the total rainfall from that cloud, thisstudy finds very weak correlations between the effec-tive size and the underlying SST. Examination of theconfidence interval of the slopes indicates that the signof the slopes and correlations is valid. Since the major-ity of the results shows a positive correlation, this indi-cates that the interaction between the SST and effectivecloud size may even have a slight positive relationship,not the inverse relationship suggested by LCH. Theeffective size is negatively correlated with SST only forclouds identified using the threshold of 280 K. Furtherexamination of the systems driving this negative corre-lations reveals that cloud size and rainfall from warmcloud systems are more highly correlated with SST thanin cold, deep convection. These warm systems show a5% decrease in the ratio of cloud area and rainfall perdegree rise in SST. These results suggest it is possiblethat interactions between precipitation efficiency andcloud size with SST are acting on shallow, warm clouds,not deep convection. This is supported by the Lau andWu (2003) finding that precipitation efficiency in warmclouds is more sensitive to the underlying SST thandeep, cold clouds, in which strong updrafts dominatethe conversion of cloud water into precipitation, not

FIG. 6. Same as in Fig. 3, but for multicore convective cloud size.


SST. Sun and Lindzen (1993) used a simple Bowenmodel and Lau and Wu (2003) used TRMM retrievalscombined with model parameterization to calculate therate of conversion of cloud water to precipitation. Bothshow that raindrop growth is faster at higher SSTs. Thisshift in the drop size distribution with SST could ac-count for some of the reduction in the cloud area torainfall ratio observed in this study. These warm rainsystems are prevalent in the Tropics and are not onlyimportant to the hydrologic cycle, but could also pro-duce a considerable effect on the radiation balance. Areduction in the areal coverage of these systems wouldlikely yield a positive feedback on temperature sincethis would reduce the amount of reflected solar radia-tion.

A result similar to single-core clouds is found formulticore convective clouds. Complicated interactionsbetween convective cores and anvils occur within thecloud, making it more difficult to interpret results.However, it is important to investigate these cloudssince they account for the majority of deep convectionand precipitation in the Tropics. As with the single-coreclouds, effective cloud size for multicore convectiveclouds is not correlated with SST. SST explains lessthan 0.2% of the variance regardless of the cloud iden-tification threshold. However, because of swath widthlimitations by the PR, small multicore convectiveclouds are examined and should not be interpreted asbeing representative of the multicore convective cloudslarger than the PR swath, such as mesoscale convectivesystems.

The lack of any significant correlation between cloudsize normalized by rainfall amount and SST for bothsingle- or multicore convective clouds suggests that it islikely that the LCH use of CWT, as proposed by HM,was responsible for the negative correlations found inthe original iris study. Both single-core and multicoreresults indicate that changes in precipitation efficiencywith SST are not likely affecting the size of deep con-vective clouds. However, examination of warm rain sys-tems shows a much stronger inverse relationship be-tween rainfall, cloud size, and SST. Since these warmrain systems are so prevalent, it is possible that changesin these systems could have a large impact on the globalenergy budget. Further examination of the mechanismsaffecting multicore cloud size, as well as the interactionbetween precipitation efficiency, cloud size and SST forwarm, shallow clouds, is needed to fully understand thecomplex cloud radiative feedback effects in the Tropics.

Acknowledgments. This research was supported byGrant NA17RJ1228 from NOAA OGP and NASAGrants NAG5-12273 and NAG5-13694.

REFERENCES

Bjerknes, J., 1966: A possible response of the atmospheric Hadleycirculation to equatorial anomalies of ocean temperature.Tellus, 18, 820–829.

Chambers, L. H., B. Lin, and D. F. Young, 2002: Examination ofnew CERES data for evidence of tropical iris feedback. J.Climate, 15, 3719–3726.

Fu, C., H. F. Diaz, and J. O. Fletcher, 1986: Characteristics of theresponse of sea surface temperature in the central Pacificassociated with warm episodes of the Southern Oscillation.Mon. Wea. Rev., 114, 1716–1739.

Fu, R., A. D. Del Genio, W. B. Rossow, and W. T. Liu, 1992:Cirrus-cloud thermostat for tropical sea surface temperaturestested using satellite data. Nature, 358, 394–397.

——, ——, and ——, 1994: Influence of ocean surface conditionson atmospheric vertical thermodynamic structure and deepconvection. J. Climate, 7, 1092–1108.

Graham, N. E., and T. P. Barnett, 1987: Sea surface temperature,surface wind divergence, and convection over tropicaloceans. Science, 238, 657–659.

Hall, A., and S. Manabe, 1999: The role of water vapor feedbackin unperturbed climate variability and global warming. J. Cli-mate, 12, 2327–2346.

Hartmann, D. L., and M. L. Michelsen, 1993: Large-scale effectson the regulation of tropical sea surface temperature. J. Cli-mate, 6, 2049–2062.

——, and ——, 2002a: No evidence for iris. Bull. Amer. Meteor.Soc., 83, 249–254.

——, and ——, 2002b: Reply. Bull. Amer. Meteor. Soc., 83, 1349–1352.

——, and ——, 2002c: A two-box model of cloud-weighted seasurface temperature: The semiautomatic negative correlationwith mean cloud fraction. Bull. Amer. Meteor. Soc., 83(Suppl.), ES70–ES71.

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 Uni-versity Press, 892 pp.

Iguchi, T., T. Kozu, R. Meneghini, J. Awaka, and K. Okamoto,2000: Rain-profiling algorithm for the TRMM precipitationradar. J. Appl. Meteor., 39, 2038–2052.

Inamdar, A. K., and V. Ramanathan, 1994: Physics of greenhouseeffect and convection in warm oceans. J. Climate, 7, 715–731.

Jackson, D. L., and G. L. Stephens, 1995: A study of SSM/I-derived columnar water vapor over the global oceans. J. Cli-mate, 8, 2025–2038.

Krueger, A. F., and T. I. Gray, 1969: Long-term variations inequatorial circulation and rainfall. Mon. Wea. Rev., 97, 700–711.

Kummerow, C., W. Barnes, T. Kozu, J. Shiue, and J. Simpson,1998: The Tropical Rainfall Measuring Mission (TRMM)sensor package. J. Atmos. Oceanic Technol., 15, 808–816.

Lau, K.-M., and H. T. Wu, 2003: Warm rain processes over tropi-cal oceans and climate implications. Geophys. Res. Lett., 30,2290–2294.

——, C.-H. Sui, M.-D. Chou, and W.-K. Tao, 1994: An inquiryinto the cirrus-cloud thermostat effect for tropical sea surfacetemperature. Geophys. Res. Lett., 21, 1157–1160.

Lin, B., B. A. Wielicki, L. H. Chambers, Y. Hu, and K.-M. Xu,2002: The iris hypothesis: A negative or positive cloud feed-back? J. Climate, 15, 3–7.


Lindzen, R. S., M.-D. Chou, and A. Hou, 2001: Does the earthhave an adaptive infrared iris? Bull. Amer. Meteor. Soc., 82,417–432.

Manabe, S., and R. T. Weatherald, 1967: Thermal equilibrium ofthe atmosphere with a given distribution of relative humidity.J. Atmos. Sci., 24, 241–259.

Pierrehumbert, R. T., 1995: Thermostats, radiator fins, and thelocal runaway greenhouse. J. Atmos. Sci., 52, 1784–1806.

Ramanathan, V., 1981: The role of ocean-atmosphere interactionsin the CO2 climate problem. J. Atmos. Sci., 38, 918–930.

——, and W. Collins, 1991: Thermodynamic regulation of oceanwarming by cirrus clouds deduced from observations of the1987 El Niño. Nature, 351, 27–32.

Rasool, S. I., and S. H. Schneider, 1971: Atmospheric carbon di-oxide and aerosols: Effects of large increases on global cli-mate. Science, 173, 138–141.

Raval, A., and V. Ramanathan, 1989: Observational determina-tion of the greenhouse effect. Nature, 342, 758–762.

Reynolds, R. W., and D. C. Marsico, 1993: An improved real-timeglobal sea surface temperature analysis. J. Climate, 6, 114–119.

——, and T. M. Smith, 1994: Improved global sea surface tem-

perature analyses using optimum interpolation. J. Climate, 7,929–948.

Riehl, H., and J. S. Malkus, 1958: On the heat balance in theequatorial trough zone. Geophysica, 6, 503–538.

Simpson, J., C. Kummerow, W. K. Tao, and R. F. Adler, 1996: Onthe tropical rainfall measuring mission (TRMM). Meteor. At-mos. Phys., 60, 19–36.

Sun, D.-Z., and R. S. Lindzen, 1993: Distribution of tropical tro-pospheric water vapor. J. Atmos. Sci., 50, 1643–1660.

Tokay, A., D. A. Short, C. R. Williams, W. L. Ecklund, and K. S.Gage, 1999: Tropical rainfall associated with convective andstratiform clouds: Intercomparison of disdrometer and pro-filer measurements. J. Appl. Meteor., 38, 302–320.

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

Wallace, J. M., 1992: Effect of deep convection on the regulationof tropical sea surface temperature. Nature, 357, 230–231.

Wentz, F. J., and T. Meissner, 2000: AMSR Ocean Algorithm:ATBD Version 2. NASA Reference Publication, 66 pp.[Available online at http://www.ssmi.com/papers/AMSR_Ocean_Algorithm_Version_2.pdf.]


an evaluation of the proposed mechanism of the adaptive...

Documents