Cloud-SST feedback in southeastern tropical Atlantic anomalous events



Similar documents
ENSO: Recent Evolution, Current Status and Predictions. Update prepared by: Climate Prediction Center / NCEP 29 June 2015

ENSO Cycle: Recent Evolution, Current Status and Predictions. Update prepared by Climate Prediction Center / NCEP 9 May 2011

Evolution of Model Systematic Errors in the Tropical Atlantic Basin from the NCEP Coupled Hindcasts

Reply to No evidence for iris

Guy Carpenter Asia-Pacific Climate Impact Centre, School of energy and Environment, City University of Hong Kong

The Influence of the Mean State on the Annual Cycle and ENSO Variability: A Sensitivity Experiment of a Coupled GCM

Comment on "Observational and model evidence for positive low-level cloud feedback"

Observed Cloud Cover Trends and Global Climate Change. Joel Norris Scripps Institution of Oceanography

Indian Ocean and Monsoon

Near Real Time Blended Surface Winds

How to analyze synoptic-scale weather patterns Table of Contents

Temporal variation in snow cover over sea ice in Antarctica using AMSR-E data product

THE CURIOUS CASE OF THE PLIOCENE CLIMATE. Chris Brierley, Alexey Fedorov and Zhonghui Lui

IGAD CLIMATE PREDICTION AND APPLICATION CENTRE

What the Heck are Low-Cloud Feedbacks? Takanobu Yamaguchi Rachel R. McCrary Anna B. Harper

TOPIC: CLOUD CLASSIFICATION

Relationship between the Subtropical Anticyclone and Diabatic Heating

A SURVEY OF CLOUD COVER OVER MĂGURELE, ROMANIA, USING CEILOMETER AND SATELLITE DATA

ENVIRONMENTAL STRUCTURE AND FUNCTION: CLIMATE SYSTEM Vol. II - Low-Latitude Climate Zones and Climate Types - E.I. Khlebnikova

Decadal/Interdecadal variations in ENSO predictability in a hybrid coupled model from

Radiative effects of clouds, ice sheet and sea ice in the Antarctic

Huai-Min Zhang & NOAAGlobalTemp Team

Chapter Overview. Seasons. Earth s Seasons. Distribution of Solar Energy. Solar Energy on Earth. CHAPTER 6 Air-Sea Interaction

Fundamentals of Climate Change (PCC 587): Water Vapor

VOCALS-CUpEx: The Chilean Upwelling Experiment

Comparison of Cloud and Radiation Variability Reported by Surface Observers, ISCCP, and ERBS

Chapter 6: Cloud Development and Forms

A Comparison of the Atmospheric Response to ENSO in Coupled and Uncoupled Model Simulations

Seasonal & Daily Temperatures. Seasons & Sun's Distance. Solstice & Equinox. Seasons & Solar Intensity

8.5 Comparing Canadian Climates (Lab)

Scholar: Elaina R. Barta. NOAA Mission Goal: Climate Adaptation and Mitigation

SPATIAL DISTRIBUTION OF NORTHERN HEMISPHERE WINTER TEMPERATURES OVER THE SOLAR CYCLE DURING THE LAST 130 YEARS

Daily High-resolution Blended Analyses for Sea Surface Temperature

Supporting Online Material for

SPOOKIE: The Selected Process On/Off Klima Intercomparison Experiment

Formation & Classification

A decadal solar effect in the tropics in July August

Coupling between subtropical cloud feedback and the local hydrological cycle in a climate model

Precipitation, cloud cover and Forbush decreases in galactic cosmic rays. Dominic R. Kniveton 1. Journal of Atmosphere and Solar-Terrestrial Physics

Interhemispheric Influence of the Atlantic Warm Pool on the Southeastern Pacific

SST-Forced Atmospheric Variability in an Atmospheric General Circulation Model

IMPACTS OF IN SITU AND ADDITIONAL SATELLITE DATA ON THE ACCURACY OF A SEA-SURFACE TEMPERATURE ANALYSIS FOR CLIMATE

Mechanisms of an extraordinary East Asian summer monsoon event in July 2011

Arctic Cloud Changes from Surface and Satellite Observations

DIURNAL CYCLE OF CLOUD SYSTEM MIGRATION OVER SUMATERA ISLAND

Chapter 6 - Cloud Development and Forms. Interesting Cloud

Hurricanes. Characteristics of a Hurricane

Trimodal cloudiness and tropical stable layers in simulations of radiative convective equilibrium

Global Seasonal Phase Lag between Solar Heating and Surface Temperature

Clouds, Circulation, and Climate Sensitivity

How To Model An Ac Cloud

Intra-seasonal and Annual variability of the Agulhas Current from satellite observations

The Next Generation Flux Analysis: Adding Clear-Sky LW and LW Cloud Effects, Cloud Optical Depths, and Improved Sky Cover Estimates

Can latent heat release have a negative effect on polar low intensity?

Seasonal Temperature Variations

Improving Hydrological Predictions

Limitations of Equilibrium Or: What if τ LS τ adj?

The West African Monsoon Dynamics. Part I: Documentation of Intraseasonal Variability

Evalua&ng Downdra/ Parameteriza&ons with High Resolu&on CRM Data

Atmospheric Dynamics of Venus and Earth. Institute of Geophysics and Planetary Physics UCLA 2 Lawrence Livermore National Laboratory

A review of the fall/winter 2000/01 and comparison with

Joel R. Norris * Scripps Institution of Oceanography, University of California, San Diego. ) / (1 N h. = 8 and C L

Thompson/Ocean 420/Winter 2005 Tide Dynamics 1

Jessica Blunden, Ph.D., Scientist, ERT Inc., Climate Monitoring Branch, NOAA s National Climatic Data Center

Impacts of Pacific and Indian Ocean coupling on wintertime tropical intraseasonal oscillation: a basin-coupling CGCM study

South Africa. General Climate. UNDP Climate Change Country Profiles. A. Karmalkar 1, C. McSweeney 1, M. New 1,2 and G. Lizcano 1

How To Understand Cloud Radiative Effects

IMPACT OF SAINT LOUIS UNIVERSITY-AMERENUE QUANTUM WEATHER PROJECT MESONET DATA ON WRF-ARW FORECASTS

Changing Clouds in a Changing Climate: Anthropogenic Influences

Ecosystem-land-surface-BL-cloud coupling as climate changes

Geography affects climate.

Research Objective 4: Develop improved parameterizations of boundary-layer clouds and turbulence for use in MMFs and GCRMs

RaysWeather.Com Winter Fearless Forecast

Develop a Hybrid Coordinate Ocean Model with Data Assimilation Capabilities

USING THE GOES 3.9 µm SHORTWAVE INFRARED CHANNEL TO TRACK LOW-LEVEL CLOUD-DRIFT WINDS ABSTRACT

Frank and Charles Cohen Department of Meteorology The Pennsylvania State University University Park, PA, U.S.A.

The ozone hole indirect effect: Cloud-radiative anomalies accompanying the poleward shift of the eddy-driven jet in the Southern Hemisphere

The Oceans Role in Climate

Month-Long 2D Cloud-Resolving Model Simulation and Resultant Statistics of Cloud Systems Over the ARM SGP

Relation between Indian monsoon variability and SST

The relationships between Argo Steric Height and AVISO Sea Surface Height

The impact of parametrized convection on cloud feedback.

An Analysis of the Rossby Wave Theory

RADIATION IN THE TROPICAL ATMOSPHERE and the SAHEL SURFACE HEAT BALANCE. Peter J. Lamb. Cooperative Institute for Mesoscale Meteorological Studies

Parameterization of Cumulus Convective Cloud Systems in Mesoscale Forecast Models

Decadal predictions using the higher resolution HiGEM climate model Len Shaffrey, National Centre for Atmospheric Science, University of Reading

Name Period 4 th Six Weeks Notes 2015 Weather

Heavy Rainfall from Hurricane Connie August 1955 By Michael Kozar and Richard Grumm National Weather Service, State College, PA 16803

Sea Surface Temperature Biases under the Stratus Cloud Deck in the Southeast Pacific Ocean in 19 IPCC AR4 Coupled General Circulation Models

Tropical Stationary Wave Response to ENSO: Diabatic Heating Influence on the Indian summer monsoon

A Project to Create Bias-Corrected Marine Climate Observations from ICOADS

Goal: Understand the conditions and causes of tropical cyclogenesis and cyclolysis

Comparing Properties of Cirrus Clouds in the Tropics and Mid-latitudes

CLIMATOLOGICAL DATA FROM THE WESTERN CANADIAN ODAS MARINE BUOY NETWORK

Diurnal Cycle of Convection at the ARM SGP Site: Role of Large-Scale Forcing, Surface Fluxes, and Convective Inhibition

Climatology of Surface Meteorology, Surface Fluxes, Cloud Fraction and Radiative Forcing Over South-East Pacific from Buoy Observations

Why aren t climate models getting better? Bjorn Stevens, UCLA

Convective Clouds. Convective clouds 1

Improving Low-Cloud Simulation with an Upgraded Multiscale Modeling Framework

Transcription:

Click Here for Full Article JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112,, doi:10.1029/2006jc003626, 2007 Cloud-SST feedback in southeastern tropical Atlantic anomalous events Bohua Huang 1,2 and Zeng-Zhen Hu 2 Received 6 April 2006; revised 3 July 2006; accepted 3 August 2006; published 23 March 2007. [1] Using satellite-based cloud measurements for 1984 2004, the interannual variability of the low-level cloud cover over the tropical Atlantic Ocean in austral winter (June July August, JJA) is examined. It is found that the leading pattern of the low-cloud anomalies in this season is a modulation of the climatological center of the cloud cover over the southeastern tropical Atlantic Ocean off the Angola and Namibia coasts. The fluctuation of cloud amount there occurs on both interannual and longer timescales. The relationship between this low-cloud anomalous pattern and basinwide ocean-atmosphere anomalies is studied through a composite analysis based on the objectively selected major low-cloud deficit and excess years. For the composites we intentionally use data sets mainly based on satellite measurements for the past one to two decades to minimize the potential influences of the bias in the model-based ocean-atmosphere analyses. The composites show that the JJA anomalous cloud pattern is strongly influenced by the sea surface temperature (SST) anomalies of the equatorial and southeastern tropical Atlantic Ocean in the previous summer. The anomalous surface warm events off the southwestern coast of Africa near 15 S in January and February are usually initiated dynamically by remote forcing from the westerly wind anomalies over the western equatorial Atlantic, preceding the cloud anomalies. During the next few months, the warm water is spread into the southeastern tropical Atlantic Ocean and is conducive to deficit low-cloud cover in subsequent JJA. The reduced low-cloud cover in turn forces positive SST tendency in a larger area of the southeastern Atlantic Ocean by changing the amount of the local solar radiation reaching the sea surface. In June and July, this process moves the major center of the SST anomalies away from the coast and closer to the equator when the coastal process weakens. The low cloud-radiation-sst feedback also plays a role in the slow westward expansion of the SST anomalies in late austral winter and spring. Overall, the influence of the cloud fluctuation is an important component in the evolution of the southeastern tropical Atlantic anomalous events. Citation: Huang, B., and Z.-Z. Hu (2007), Cloud-SST feedback in southeastern tropical Atlantic anomalous events, J. Geophys. Res., 112,, doi:10.1029/2006jc003626. 1. Introduction [2] Stratus cloud decks often develop over cold waters off the eastern boundaries of the tropical and subtropical oceans. In the southeastern tropical Atlantic Ocean, extensive annual mean amount of low clouds with top height below 680 hpa pressure level has a center of maximum coverage above 60% of the sky near 10 15 S and the Greenwich longitude. Seasonally, the cloud cover in this region starts to increase in May from its minimum of around 50%. By austral winter (June July August, JJA), a welldefined center of maximum cloudiness is formed off the 1 Department of Climate Dynamics, College of Science, George Mason University, Fairfax, Virginia, USA. 2 Center for Ocean-Land-Atmosphere Studies, Institute of Global Environment and Society, Calverton, Maryland, USA. Copyright 2007 by the American Geophysical Union. 0148-0227/07/2006JC003626$09.00 coasts of Angola and Namibia with coverage above 60% (Figure 1a). The cloud amount continues to grow during the next few months and peaks at above 70% in austral spring (September October November, SON, Figure 1b). Because the increasing marine cloudiness blocks incoming solar radiation from reaching the surface, a local minimum of the net surface shortwave radiative heat flux appears in the southeastern Atlantic Ocean at the latitude band between 10 15 S during these two seasons (Figures 1c and 1d). The eastern part of the basin receives 60 80 Wm 2 less solar radiation than the western basin does. [3] The substantial blocking of solar flux during austral winter and spring by the cloud apparently contributes to the nearly simultaneous seasonal enhancement of the coastal and equatorial cold tongue in the eastern Atlantic Ocean. Besides, the longwave radiative cooling at the cloud top generates strong local southerly surface winds that cause rapid sea surface cooling during late boreal spring [Nigam, 1of19

Figure 1. Seasonal climatologies of low-cloud cover and net shortwave radiative heat flux into the surface for austral winter (June August, JJA) and spring (September November, SON) in the tropical Atlantic region. The percentage of low-cloud cover is shown in (a) for JJA and (b) for SON with contour interval 10%. Regions with cloud cover between 50 60 are lightly shaded and those beyond 60 are darkly shaded. The corresponding shortwave flux is shown in (c) for JJA and (d) for SON with contour interval 20 Wm 2. Regions with shortwave flux less than 180 Wm 2 are lightly shaded and those larger than 240 Wm 2 are darkly shaded. These climatologies are based on ISCCP monthly data sets from July 1983 to December 2004, to be described in detail in section 2. 1997]. On the other hand, the cloud related climate effect is not just from the atmosphere to the ocean. Klein and Hartmann [1993] suggest that the seasonal growth of the subtropical low-level stratus cloud cover is partly controlled by the increasing stability of the lower atmosphere above the sea surface. In particular, they find that the atmospheric stability off the Namibian coast is mainly driven by the seasonal cooling of the local sea surface temperature (SST), which implies an SST contribution to the marine stratus clouds there. Based on these observational evidences, Philander et al. [1996] propose a positive feedback between the stratus cloud and SST due to their respective influences on the incoming solar radiation and the stability in the lower atmosphere. This feedback partly accounts for the asymmetry of the mean SST state and the strong annual cycle in the eastern equatorial Pacific and Atlantic [Xie, 2004]. It should be pointed out that this cloud-sst feedback over cold water is different from another important air-sea feedback process involving cloud in the tropical Atlantic Ocean, which occurs over the warmer water further to the west. In the Atlantic side of the Western Hemisphere warm pool, an SST increase induces stronger convection and cloudiness, which in turn result in less long-wave radiative heat loss from the sea surface and promote further SST increase [Wang and Enfield, 2001, 2003]. [4] The southeastern tropical Atlantic Ocean is a region of intensive interannual SST variability [Servain et al., 1982; Zebiak 1993; Enfield and Mayer, 1997; Huang et al., 2004; Huang and Shukla, 2005; Hu and Huang, 2005]. In particular, major anomalous warming and cooling events appear off the coast of southwestern Africa in austral summer (December February, DJF) and fall (March April May, MAM), which are referred to as the Benguela Niños [e.g., Hirst and Hastenrath, 1983; Shannon et al., 1986; Florenchie et al., 2004]. The Benguela Niños usually expand northwestward toward the equator after peaking near the coast [Florenchie et al., 2004], which enhance 2of19

the warming in the equatorial Atlantic during austral winter [Hu and Huang, 2005]. Previous studies have established that these southeastern Atlantic anomalous events are initiated by the anomalous zonal winds over the western equatorial ocean as early as austral spring [Hirst and Hastenrath, 1983], which remotely induces dynamical adjustment of the thermocline and upwelling in the equatorial and eastern coastal Atlantic [e.g., Hirst and Hastenrath, 1983; Zebiak, 1993; Carton and Huang, 1994; Huang and Shukla, 1997; Florenchie et al., 2004]. The process bears some resemblance to the air-sea feedback associated with the El Niño-Southern Oscillation (ENSO) in the tropical Pacific [Bjerknes, 1969]. Given the relatively high skills of the ENSO prediction by current coupled ocean-atmosphere general circulation models (CGCM) [e.g., Schneider et al., 2003; Saha et al., 2006], however, it is puzzling that these CGCMs ability to predict the southeastern Atlantic anomalous events is generally quite dismal. In fact, the southeastern Atlantic has the lowest predictive skill in the tropical Atlantic basin according to our evaluation of the National Centers for Environmental Prediction s (NCEP) state-of-the-art climate forecast system (CFS) hindcast [Hu and Huang, 2007]. [5] In this study, we examine whether the interannual low-cloud fluctuations in this region contribute to the evolution of these anomalous events. There have been quite extensive previous global-scale studies on the general relationship between the stratus cloud and SST on interannual scales [e.g., Weare, 1994; Norris and Leovy, 1994]. Based on historical surface observations, Tanimoto and Xie [2002] found that the increased (decreased) cloudiness over the cold (warm) lobe of the tropical Atlantic dipole pattern tends to enhance the persistence of this pattern on decadal time scales. Examining the cloud outputs from the NCEP/ NCAR (National Center for Atmospheric Research) reanalysis, Hu and Huang [2005] showed statistically that the low-cloud anomalies in the southeastern ocean contribute positively to the warming events in the southeastern ocean during austral fall and winter. The cloud cover from the reanalysis, however, is a parameterized model variable only indirectly influenced by the atmospheric data assimilation. Its reliability for climate study is largely unknown. [6] In this study, we have used high quality satellite observations for more than twenty years to examine the role of cloud more directly. The satellite-based observations collected from measurements are described in section 2. Section 3 examines the major pattern of the interannual variability of the low cloud in the tropical Atlantic region and its relationship with major regional SST indices. A composite analysis is conducted in Section 4 to describe the evolution of the anomalies of cloudiness and its connection with major ocean-atmosphere variables. The main results of this study are summarized and discussed in Section 5, where the NCEP/NCAR reanalyzed low-cloud product is also evaluated against the satellite data. 2. Data Sets [7] A set of 21-year monthly cloud coverage from July 1983 to December 2004 is produced by the International Satellite Cloud Climatology Project (ISCCP), which infers the global distribution of clouds and their properties from diurnal to interannual time scales using a series of satellite radiance measurements [Rossow and Dueñas, 2004]. The cloud properties used in this study include the percentages of sky coverage by total cloud, as well as by the low-level, midlevel, and high-level clouds individually. These clouds are defined by cloud top pressure, categorized as below the pressure level of 680 hpa, between 680 hpa and 440 hpa, and above 440 hpa. Roughly, the low-level clouds are composed of stratus, stratocumulus, and shallow cumulus clouds. The midlevel clouds are altocumulus, altostratus, and nimbostratus clouds while the high-level clouds are cirrus and cirrostratus clouds, as well as cumulus clouds formed by deep convection. The monthly data are on a 2.5 lat 2.5 lon grid, converted from an equal-area grid of 280 km resolution. A corresponding set of monthly surface net shortwave and long-wave radiative fluxes at the same spatial resolution of the cloud datasets and the same time period is also used, which has been derived from a comprehensive radiation dataset calculated by Zhang et al. [2004] using a complete radiative transfer model with input from surface and atmospheric observations and the ISCCP cloud properties. [8] Observed SST, precipitation, surface wind stress, surface latent and sensible heat fluxes, and sea level height from independent sources are also analyzed in this study. The monthly mean SST fields for July 1983 to December 2004 are from the version 2 of the optimal interpolation (IO) analysis, which is produced by the Climate Prediction Center (CPC) using satellite and in situ SST measurements on a 1 1 grid [Reynolds et al., 2002]. The monthly surface wind stress fields for the same period are obtained from the NCEP/NCAR reanalysis on a Gaussian grid [Kalnay et al., 1996] because the satellite based scatterometer wind data are too short for this study. Since the reanalyzed wind stress is model-dependent, we also analyzed a complimentary set of surface wind stress from the version 2 of the Goddard Satellite-based Surface Turbulent Fluxes (GSSTF2). The GSSTF2 wind stress is calculated from vector surface winds with the wind speed of the Special Sensor Microwave/Imager (SSM/I) measurements on board several satellites while its direction is adjusted to those of the NCEP/NCAR reanalyzed 10-m winds and other buoy and ship measurements using a minimization scheme [Atlas et al., 1996]. The GSSTF2 data also provide surface latent and sensible heat fluxes derived from SSM/I 10-m wind speed and specific humidity as well as SST and 2-meter air temperature from the NCEP/NCAR reanalysis. This dataset spans from July 1987 to December 2000 over the global ocean on a 1 1 grid [Chou et al., 2003]. Finally, the 11-year monthly sea level height fields on a 1 1 grid from January 1993 to December 2003 are processed by Chambers et al. [2003] from the satellite altimeter measurements from TOPEX/Poseidon and subsequent JASON-1 missions. 3. Low-Cloud Variability and Its Relation to SST Indices [9] The basic patterns of the anomalous low-cloud variability for 1983 2004 are derived through an empirical orthogonal function (EOF) analysis for the tropical Atlantic basin with 30 S 15 N and 60 W 15 E. The chosen domain 3of19

Figure 2. The 1st EOF mode of seasonally averaged anomalies of the low-level cloud cover for austral winter (JJA). Panel (a) shows the EOF spatial pattern. The contour interval is 3%. Regions below 3% (above 3%) are lightly (darkly) shaded. The corresponding principal component (PC) is the solid curve in panel (b). The PC is normalized by its maximum value. The two dashed lines represent the upper and lower positions at 75% of its standard deviation. The points marked by open and closed circles represent specific positive and negative years used for a composite analysis to be described in section 4. is skewed to the southern hemisphere to accommodate the fact that major low-cloud covers concentrate over the southeastern basin (Figures 1a and 1b). Moreover, given the facts that the cloud amount varies considerably from season to season, it is expected that the cloud cover anomaly be seasonally modulated, like other major features of the tropical interannual variability. Therefore, the EOF modes are calculated separately for cloud anomalies averaged for each season. [10] Figure 2 shows the 1st EOF mode of the seasonally averaged low-cloud cover for JJA during 1984 2004. This mode accounts for over 25% of the total variance of lowcloud fluctuation in this season and is well separated from the 2nd mode. Its spatial structure is dominated by negative cloud anomalies in the southeastern Atlantic centered within 10 20 S and 10 W 10 E (Figure 2a). The anomalies are weakened over the equator but enhanced again further north along a zonal belt around 5 N. This anomalous pattern is largely monopole, with only weak anomalies of opposite sign in the western equatorial Atlantic. Compared with the mean state of this season (Figure 1a), the 1st EOF mode apparently represents an interannual modulation of the main center of cloud during this season. [11] The principal component (PC) of this mode shows that the anomalous cloud pattern fluctuates on two different timescales (Figure 2b). As a relatively long-term change, the PC is largely negative (excessive cloud cover) from 1985 to 1995 followed by mostly positive values (deficit cloud cover) from 1996 to 2004. The long-term change is possibly associated with the decadal mode identified by Tanimoto and Xie [2002]. However, the time series of the satellite cloud data is too short to make any inference on decadal modes based on it. Superimposed on this lowfrequency change is a quite active year-to-year fluctuation, which generates some major peaks, both positive and negative, during this period. The major deficit years are 1984, 88, 2003, and a continuous episode from 1998 to 2001, while the major excessive years are 1987, 90, 92, 94, 97 and 2002. [12] The spatial patterns of the leading EOF modes derived from the low-cloud data of other seasons are somewhat similar to the one shown in Figure 2a with certain shifts of the major centers of variability from season to season. The correlation coefficients of the JJA PC with those from other seasons are between 0.45 and 0.48, which are above the 95% significance level and suggest a certain degree of persistence from season to season. However, the correlations from season to season are not high, also suggesting that the contribution of intraseasonal fluctuations is not negligible. The JJA mode is chosen here partly because it is the only one that shows relatively high correlations to all other seasons, implying a larger content of low-frequency signals. [13] We have correlated the JJA PC with three indices of averaged SST anomalies in different areas of the tropical Atlantic Ocean, as well as an ENSO index, on monthly 4of19

bases. The averaging domains of the tropical Atlantic SST indices are shown in Figure 3a, superimposed on the field of the standard deviation of the monthly SST anomalies from July 1983 to December 2004. The ABA (Angola-Benguela area) index defined by Florenchie et al. [2004] is the averaged SST anomalies within the area from 10 20 S and 8 E to the African coast. The ATL3 index defined by Zebiak [1993] is the averaged SST anomalies in the area within 3 S 3 N and from 20 W to Greenwich lo ngitude. T he domains of both indices are located in maximum centers of the SST standard deviations, with ABA characterizing the Benguela Niño and ATL3 the SST variability in the equatorial central and eastern Atlantic. The CLD index, defined here as averaged SST anomalies within 3 15 S and 5 W 8 E, is not directly related to local maximum of the SST standard deviation. Instead, it corresponds to the region of large low-cloud anomalies in austral winter with potentially large SST response, based on our preliminary visual analysis of the fields of SST anomalies. Its role will be seen more clearly later on. [14] Besides the tropical Atlantic SST indices, we also examined the relationship between the JJA PC of the lowcloud anomalies and the NINO3 index, which is defined as the averaged SST anomalies with 5 S 5 N, 90 150 W and characterizes the ENSO cycle in the Pacific Ocean [Zebiak and Cane, 1987]. The ENSO connection is examined here because it is one of the major remote forcing factors to the tropical Atlantic variability [e.g., Xie and Carton, 2004]. [15] For the calculation, the monthly time series of an index is first grouped into subsets of time series with the same calendar months. The JJA PC is then correlated with each of these subset time series for respective months from preceding October to following November. The correlation coefficients for a given index are presented as a sequence spanning from October to the next November, as shown by the curves in Figure 3b. For each curve, the segment contained within the shaded time interval (June to August) in Figure 3b can be seen as roughly simultaneous correlations with the PC. The segment before or after the shaded interval shows correlations with SST leading or lagging the JJA PC. [16] The highest ABA-PC correlation (thick solid curve) has the maximum in February (0.8) and a secondary peak of 0.7 in July. The former leads the JJA anomalous cloud pattern by 5 to 6 months while the latter is nearly simultaneous. The correlations are above the 95% significance level from previous November to August. It surpasses the 99% level (the long dashed horizontal line) from January to March. It dips in March and April but increases again in May and June, before decreasing quickly in August and September. The double-peak structure can be explained as a sequence of forcing and feedback between the SST and cloud anomalies. The ATL3 index is also significantly correlated with the PC in February and March (thin solid curve). Since then, the correlation holds between 0.4 and 0.5 until September. This relation, however, is weaker than that between the ABA index and the cloud pattern. [17] The correlation with CLD index, on the other hand, increases since February and peaks in August at 0.8. The relatively fast increasing rate from June to August suggests stronger response to the cloud forcing during this period. As shown in Figure 3c, the ABA in February (thin solid curve), the CLD in August (thin dashed curve), and the JJA PC1 (thick solid curve, reprinted from Figure 2b) are consistent among each other. The statistical analysis establishes that warm SST anomalies in the Angola and Namibia coastal region during austral summer precede the deficit low-cloud cover in the tropical southeastern Atlantic Ocean during subsequent winter. The regional deficit cloud cover is associated with nearly simultaneous, or slightly delayed warming up of the SST in the southeastern tropical Atlantic (i.e., the CLD region), further to the northwest from the coastal region represented by the ABA index. A major deviation of the ABA index from other variables appears in 1995 when the major ABA anomaly does not provoke significant subsequent response. [18] Besides the regional correlations, remote forcing from the Pacific ENSO also influences the low-cloud anomalies in the southeastern tropical Atlantic, which can be seen by its significant negative correlation with the NINO3 index since June (dotted line, Figure 3b). Although the JJA PC is uncorrelated to the equatorial Pacific SST anomalies in the previous DJF, a negative correlation is built up quickly in MAM and reaches its maximum ( 0.65) in June. This implies that the low-cloud deficit in JJA over the southeastern tropical Atlantic is usually also associated with nearly simultaneous cold SST anomalies in the equatorial Pacific Ocean. In fact, an out-of-phase relation can be seen clearly between the NINO3 index in June (Figure 3c, dotted line) and the JJA PC of the low cloud (Figure 3c, thick solid curve). This relation suggests that warm SST anomalies in the equatorial Pacific can remotely enhance the marine low cloud in the southeastern tropical Atlantic, probably by changing the atmospheric circulation cells to enhance subsidence over the region. Wang [2002a, 2002b] have examined the atmospheric circulation cells associated with ENSO and the tropical Atlantic variability. The negative correlations between the JJA PC and the NINO3 index in the subsequent months of the year are generally high and close to 99% significance level because the SST anomalies in the Pacific are highly persistent. Although a comprehensive examination of this remote connection is beyond the scope of this paper, one should keep its potential influences in mind when examining the regional SST-cloud feedback. The physical process accounting for the regional relationship is demonstrated in the next section through a composite analysis. 4. Composite Analysis [19] To further understand the tropical Atlantic large-scale ocean-atmosphere circulation patterns associated with the cloud fluctuation, a composite analysis is conducted for various variables. The anomalous years chosen for composite are based on the JJA low-cloud PC (Figure 2b) with the following two criteria: First, the PC magnitude should be larger than 75% of the PC standard deviation marked by the two long dashed horizontal lines in Figure 2b. Second, all the chosen years should be local peaks (maxima or minima) of the PC time series shown in Figure 2b. Based on these two criteria, five positive (deficit cloud) and negative (excessive cloud) years are chosen for the period. The positive years are 1984, 1988, 1999, 2001, and 2003. The negative years are 1987, 1992, 1994, 1997, and 2002. They 5of19

Figure 3. (a) Averaging domains of the SST indices ABA (10 20 S, 8 E, coast), ATL3 (3 S 3 N, 0 20 W), and CLD (3 15 S, 5 W 8 E), superimposed on the standard deviation of the monthly SST anomalies from July1983 to December 2004. The contour interval is 0.1 C with areas between 0.6 0.8 C lightly shaded and those beyond 0.8 C darkly shaded. (b) Correlation coefficients of four SST indices for each calendar month with the PC of the 1st low-cloud EOF mode for JJA from the preceding October to following November. The thick solid curve shows the correlations for ABA; thin solid curve for ATL3, thin short dashed curve for CLD; and the dot curve for NINO3. The long dashed lines show the 99% significance level. The shaded area marks the time interval of PC1. (c) The time series of low cloud JJA PC1 (thick solid curve), the ABA index for the preceding February (thin solid curve), the CLD index for subsequent August (short dashed curve) and the NINO3 index for June (dot curve). Unit of the indices is C and the PC1 is nondimensional. 6of19

Figure 4. Difference between composite low cloud anomalies of deficit and excessive years for (a) February, (b) March, (c) April, (d) May, (e) June, (f) July, (g) August, (h) September and (i) October. The contour interval is 3% with zero lines omitted. Regions above 95% significance level are shaded. are marked in Figure 2b with solid and open circles respectively. [20] This list of events is slightly different from the major deficit and excessive cloud years discussed in the last section based on a visual examination of the PC amplitude. The years 1998 and 2000, both above 75% standard deviation criterion, are not chosen as the positive events because they are not peak years of the PC time series in Figure 2b. The peaking year of 1990 is not chosen as a negative event because its magnitude is slightly below the amplitude criterion while the year 1988 is selected because its amplitude is slightly above the criterion. Since the amplitude criterion (75% of the standard deviation) is subjective, we have also made the composite analyses using other amplitude criteria of 50%, 85% and 95% of the standard deviation. With these different criteria, the selected major events show some differences. The composite results, however, are qualitatively similar to those to be shown in this section. Therefore, we believe that the composite results to be reported here is robust. [21] In the following discussion, the difference maps of the positive minus negative composites are examined for the 9-month period from February to October, based on the correlation pattern demonstrated in Figure 3b. The statistical significance between the difference maps is tested with a student-t test except for SSH and latent heat flux. Their composites are constructed from fewer cases because the lengths of the data are shorter. [22] Figure 4 shows the composite evolution of the lowcloud anomalies from February to October. In February and March, there are cloud deficits in the open southern ocean (Figures 4a and 4b). However, it is not clear whether these early cloud anomalies, including those closer to the equator developed in subsequent April (Figure 4c), significantly affect later seasons. The more organized development occurs in May when negative anomalies of marginal significance appear near the eastern boundary in 10 20 S (Figure 4d). This development coincides with the seasonal increase of low-cloud cover both temporally and geographically. In the following June (Figure 4e) and July (Figure 4f), the anomalous cloud pattern similar to that shown in Figure 2 is established, which signifies a major center in the southeastern tropical Atlantic and a zonal belt to the north of the equator from central ocean to the northern coast of the Gulf of Guinea. By August, the cloud anomalies shift southward (Figure 4g) and start to decay afterwards. Since 7of19

Figure 5. Difference between composite SST anomalies from low cloud deficit and excessive years for (a) February, (b) March, (c) April, (d) May, (e) June, (f) July, (g) August, (h) September and (i) October. The contour interval is 0.25 C with zero lines omitted. Regions above 95% significance level are shaded. the southern anomalies are weakened more quickly, the anomalies over the central equatorial ocean become dominant in September (Figure 4h). The anomalies are further weakened in October and few areas pass the 95% statistical significance level by then. [23] On the other hand, the preconditioning of the summer low-cloud anomalies can be found from the SST composite (Figure 5). From February to May, the SST anomalies in the Atlantic are generally warm to the south of the equator (Figures 5a 5d), which may be in balance with the lowcloud deficit over the open ocean during this period [Tanimoto and Xie, 2002]. However, the largest SST anomalies appear near the southeastern boundary as well as the central equatorial ocean (Figures 5a and 5b) and should not be directly related to the contemporary cloud anomalies. In fact, these anomalies are more likely generated dynamically. According to a composite of the sea surface height (SSH) anomalies using three positive (1999, 2001, and 2003) and three negative (1994, 1997, and 2002) anomalous years from the TOPEX/POSEIDON and JASON-1 altimeter measurements (Figure 6), the SSH anomalies are positive in February from central equatorial ocean to the Gulf of Guinea and farther to the south around the coast of Angola and Namibia (Figure 6a). The magnitude of anomalies is above 7 cm near the coast. The anomalous SSH increase corresponds to the deepening of the thermocline, forced by the anomalous westerly surface wind stress over the central and the equatorial ocean, based on the composite of the NCEP/NCAR reanalysis data using all chosen anomalous years for 1984 2002 (Figure 6a). The equatorial wind anomalies are a part of a broader weakening of the southeast trade winds over the basin. The equatorial air-sea process can be traced back to the late months of last year when the equatorial westerly anomalies forces adjustment of the equatorial zonal thermocline slope (not shown), which is consistent with the remote forcing mechanism first proposed by Hirst and Hastenrath [1983] and reconfirmed more recently by Florenchie et al. [2004]. The SST response to the anomalous thermocline deepening is most efficient near the eastern boundary and the central equatorial ocean because the mean depth of the thermocline is shallow there. The wind and SSH anomalies are persistent in the next two months with slight weakening by April (Figures 6b and 6c). A composite of the GSSTF2 wind stress anomalies with two positive (1988 and 1999) and three negative (1992, 1994, and 1997) events (not shown) gives the 8of19

Figure 6. Difference between composite NCEP/NCAR reanalyzed wind stress and TOPEX/Poseidon and JASON-1 sea surface height anomalies from low cloud deficit and excessive years for (a) February, (b) March, (c) April, (d) May, (e) June, (f) July, (g) August, (h) September and (i) October. The wind stress vector of 0.02 Nm 2 is shown at the upper left corner. The contour interval of sea surface height is 1 cm with zero lines omitted. Regions with anomalies larger than 1 cm (less than 1 cm) are darkly (lightly) shaded. features qualitatively consistent with the NCEP/NCAR data. Although the dynamically generated SST anomalies along the southeastern coast are weakening in the austral fall (MAM), its slow spread into the open ocean provides warmer SST anomalies in the southeastern tropical Atlantic Ocean. They in turn influence the stability of the lower atmosphere over the ocean and cause the seasonal enhancement of the low stratus cloud over a larger basin. [24] The wider coverage of the low-cloud anomalies, on the other hand, provides a significant forcing to the SST anomalies in austral winter (JJA) in the open ocean. From May to October, the negative low-cloud anomalies developed over the southeastern Atlantic force warm local SST anomalies, as shown by the good correspondence between the negative cloud anomalies (Figures 4d 4i) and the positive SST anomalies (Figures 5d 5i) in the southeastern ocean during these months. This relationship can be explained by the cloud-induced surface heat flux change. The composite of the net downward shortwave radiative flux shows a positive center from May to July (Figures 7d 7f) in the southeastern Atlantic where there are major negative cloud anomalies. The domain with anomalous flux larger than 20 W/m 2 has been enlarging significantly during these three months. This solar radiation actually enhances the anomalous SST tendency away from the coast, which has been steadily decaying from April to June (Figures 5c 5e), and reestablishes the center of SST anomalies at the location defined by CLD index in July (Figures 4f, 5f, and 7f) and August (Figures 4g, 5g, and 7g). The subsequent westward extension of the anomalies south of the equator from August to October also shows clear cloud-radiation-sst coherence, suggesting that it is accompanied by the cloud-radiation-sst interaction (Figures 4g 4i, 5g 5i, and 7g 7i). [25] There is also sign that the equatorial westerly wind anomalies are reinvigorated during May (Figure 6d) and June (Figure 6e), which has been waning in April (Figure 6c). As a result, the equatorial SSH anomalies are also enhanced in June (Figure 6e) and July (Figure 6f). Whether this wind change is directly related to the cloud-induced SST change is not yet clear. We also notice that the centers of the SSH 9of19

Figure 7. Difference between composite net surface shortwave radiative flux anomalies from low cloud deficit and excessive years for (a) February, (b) March, (c) April, (d) May, (e) June, (f) July, (g) August, (h) September and (i) October. The contour interval is 5 Wm 2 with zero lines omitted. Regions above 95% significance level are shaded. anomalies are away from the equatorial ocean from August to October (Figures 6g 6i), demonstrating some characteristics of the off-equatorial Rossby waves. As mentioned before, this westward propagation involves several ocean-atmospheric variables, as can be seen more clearly from the averaged composite SST (Figure 8a), shortwave radiative flux (Figure 8b), low-cloud cover (Figure 8c), and SSH (Figure 8d) anomalies within 5 10 S during their evolution. In the first four months of the year, positive SST and SSH signals are confined closely near the eastern boundary while the anomalies of cloud and shortwave radiation are relatively small except for an isolated enhancement in March April (Figures 8b and 8c). On the other hand, there are clear indications of westward propagation in all variables with nearly the same phase speed from May to November, which can be seen from the westward migration of the signals above the 95% significance level (shaded areas in Figures 8a 8c). The major SSH anomalies also expand westward following the same track (Figure 8d). Whether this westward propagation is driven by the off-equatorial Rossby waves, as suggested by the SSH anomalies (Figure 8d), needs to be further investigated. [26] Another question that needs to be addressed is why the reduction of low-cloud cover to the north of the equator does not generate similar response in the solar radiation and SST. In fact, the solar heat flux is reduced over the equatorial zone from February to July (Figures 7a 7e) although the statistically significant signals appear only in July over the western equatorial ocean (Figure 7f). Therefore, the increased equatorial SST anomalies during this period are mostly dynamically generated and not directly related to the reduced low cloud during May July (Figures 5d 5f). This conclusion is consistent with the relatively low correlation between ATL3 and PC1 during this period (Figure 3b). The reduction of the solar radiation over the equatorial zone from April to July (Figures 7c 7f) is caused by the increasing total cloud cover, corresponding to an anomalous southward migration of the intertropical convergence Zone (ITCZ) in response to the warm equatorial SST anomalies. As shown in Figure 9, the total cloud is increased significantly over the equatorial zone from 10 of 19

Figure 8. Time-longitude sections of the composite variables averaged within 5 10 S in a 1-year evolution for (a) SST, (b) shortwave radiative flux, (c) low cloud cover, and (d) SSH. All panels show the difference between the positive and negative composite events. The contour intervals are 0.25 Cin (a), 5 Wm 2 in (b), 3% in (c), and 1 cm in (d). Zero contours are omitted in all panels. The shadings in (a) (c) show the time-space positions as the differences pass the 95% significance level. In panel (d), regions with SSH larger than 2 cm are darkly shaded and those less than 2 cm lightly are shaded. February to June (Figures 9a 9e), which is contributed by the increases in both high and mid cloud from April to August (Figure 10). In fact, both the high (left panels) and mid (right panels) cloud changes are nearly in phase with each other within the equatorial zone (10 S 10 N). The increase of the mid and high clouds by enhanced deep convection more than offsets the reduction of the low cloud there. [27] To the south of the equator, however, the low cloud accounts for a major portion of the total cloud cover (comparing Figures 4 and 9). There are few high cloud anomalies in this region because the relatively low mean SST in the south makes deep convection hard to sustain. On the other hand, a modest increase of midcloud occurs between 10 20 S in July and August (Figures 10h and 10j) because the cloud types may transfer from stratocumulus to cumulus above positive SST anomalies. These cumuli may only last a short time and not influence the heat budget significantly [Xie, 2004]. [28] We also examined other components of the total heat flux. The net long-wave radiative heat loss at the surface is influenced by the total cloud cover (Figure 11, left). In general, the long-wave radiation works in the opposite direction of the short-wave radiation with a significantly reduced magnitude, as described by Hu and Huang [2005]. Although the composite long-wave radiation patterns seem to be meaningful in the southeastern basin, they are not as statistically significant as the shortwave radiation during most of the months, except in July (Figure 11g) at the peak of its evolution. This suggests that the net long-wave radiation is more variable and not as strongly linked to the SST as the net shortwave radiation does. The composites of the sensible and latent heat flux anomalies, which are based on more limited cases contained in the GSSTF2 data, also show damping influences to the SST anomalies in the southeastern tropical Atlantic Ocean during the austral winter months. The sensible flux anomalies are generally small and negligible (not shown). The latent heat flux (Figure 11, right), however, is substantial in July (Figure 11h) and August (Figure 11j) in the southeastern tropical Atlantic Ocean. We notice that, without counting the long-wave heat loss, it seems that the heat loss associated with evaporation alone surpasses the gain of the shortwave radiation in July (Figures 7f and 11h) and August (Figures 7g and 11j). This may be the reason for the quick decay of the SST anomalies in the latter half of the year (Figures 5g 5i) even though the shortwave radiation still contributes positively to the SST anomalies. However, it is not suitable to quantitatively com- 11 of 19

Figure 9. Difference between composite total cloud anomalies of deficit and excessive years for (a) February, (b) March, (c) April, (d) May, (e) June, (f) July, (g) August, (h) September and (i) October. The contour interval is 3% with zero lines omitted. Regions above 95% significance level are shaded. pare the composite of the shortwave radiation with that of the latent heat flux because they are derived from different cases. The precise role of the net heat flux on the SST tendency requires a heat budget analysis of the mixed layer. With available measurements, the net heat flux into the ocean cannot be estimated accurately as the residue of these opposing components. Qualitatively, the high coherence between the SST and shortwave radiation anomalies in the austral winter and spring, as demonstrated in Figures 5 and 7, suggests that the cloud-radiation-sst feedback does play an active role in determining the net heat flux into the ocean at different stages of the evolution of the southeastern tropical Atlantic anomalous events. This feedback needs to be more accurately represented in modeling and theoretical studies. [29] The above analyses are all based on the differences between the cloud deficit and excessive years. We have also done similar composites based on major anomalous SST events in the southeastern tropical Atlantic Ocean. Given the fact that Benguela Niños usually peak in March or April [Florenchie et al., 2004], a major warm (or cold) event should have its ABA index above 1 C (below 1 C) in either or both of these two months. Moreover, the ABA index should keep the same sign continuously for at least three months. Based on these two criteria, five major warm events (1984, 1995, 1996, 1999, 2001) and five major cold events (1985, 1987, 1992, 1997, 2004) are selected. There is a substantial overlap between the warm (cold) events and the major low-cloud deficit (excessive) years (three out of five in each category, see Figure 2b). Moreover, we do not see any case of a major warm event paired with excessive marine low cloud, or vice versa. The composites of both the low cloud and SST anomalies based on these warm and cold events give qualitatively similar evolutions to those of Figures 4 and 5 (not shown). This result suggests that the cloud-sst feedback is an important component in many anomalous SST events in the southeastern tropical Atlantic Ocean. [30] To examine the anomalous SST evolution without cloud feedback, we made composites from the major warm and cold events that do not coincide with major corresponding low-cloud anomalies in JJA. The warm events in this category are 1995 and 1996 and the cold events are 1985 and 2004. We find that, although the composite SST anomalies from these events can be intensive in March and April, they decay more quickly than the composites of both 12 of 19

Figure 10. Difference between composite high (left panels) and mid (right panels) cloud anomalies of deficit and excessive years from April to August. The contour interval is 2% with zero lines omitted. Regions above 95% significance level are shaded. the deficit/excessive cloud years (Figure 5) and all warm/ cold SST events (not shown). And this seems to be related to the lack of continuous marine low-cloud feedback. [31] Figure 12 shows the evolution of the differences of the composite SST and low-cloud anomalies between the warm (1995, 1996) and cold (1985, 2004) events. In April, the warm SST anomalies are large near the southeastern boundary (Figure 12a) and accompanied by major cloud deficit (Figure 12b). In the following months, however, the cloud deficit reappears only briefly in June (Figure 12f) after the usual interruption in May (Figure 12d). Hereafter, the cloud anomalies are variable, changing to the opposite sign in July (Figure 12h) and becoming very small in August (Figure 12j). As a result, the warm SST anomalies generally weaken to the south of the equator in the next four months (Figure 12, left) while they increase on the equator from May (Figure 12c) to June (Figure 12e). By August, all SST anomalies (Figure 12i) are much weaker than those in Figure 5g. It is not yet clear why the intensive SST anomalies in March and April do not provoke persistent marine cloud feedback in these particular cases. It appears that, except for the 1995 case, the major SST anomalies in these events developed late and have small amplitude in February (thin solid curve, Figure 3c). This is consistent with the correlation result that the low-cloud anomalies in JJA are more strongly linked to the SST anomalies in February, which are reduced perceptibly in March and 13 of 19

Figure 11. Difference between composite anomalies of the surface long-wave radiative (left panels) and latent (right panels) heat loss from low cloud deficit and excessive years from April to August. For the long-wave radiative flux (left panels), the contour interval is 5 Wm 2 with zero lines omitted. Regions above 95% significance level are shaded. For the latent heat flux (right panels), the contour interval is 10 Wm 2 with zero lines omitted. Regions with anomalies larger than 10 W m 2 (less than 10 Wm 2 ) are darkly (lightly) shaded. April. It is possible that other atmospheric processes also influence the marine cloud. These issues should be further investigated. 5. Summary and Discussion [32] In this study, we have examined the interannual variability of the low-level cloud cover over the tropical Atlantic Ocean in austral winter (JJA) using satellite-based observations for 1984 2004. It is found that the leading pattern of the low-cloud anomalies is a modulation of the cloud amount around the climatological center in this season located over the southeastern tropical Atlantic Ocean off the Angola and Namibia coasts on interannual and longer time scales. The relationship between this low-cloud anomalous pattern and basinwide ocean-atmosphere anomalies is studied through composite analysis based on major low-cloud deficit and excess years. [33] Our results show that the winter anomalous cloud pattern is strongly influenced by the SST anomalies of the equatorial and southeastern tropical Atlantic Ocean in the previous austral summer (DJF). For instance, the anomalous 14 of 19

Figure 12. Difference between composite anomalies of the SST (left panels) and low cloud cover (right panels) for warm and cold ABA years from April to August. These cases are chosen in the years when the low cloud cover is not strongly abnormal in the JJA seasonal mean. For SST (left panels), the contour interval is 0.25 C with zero lines omitted. Regions with SST larger than 0.5 C (less than 0.5 C) are darkly (lightly) shaded. For the low cloud cover (right panels), the contour interval is 3% with zero lines omitted. Regions with anomalies larger than 3% (less than 3%) are darkly (lightly) shaded. surface warm events near the southeastern coast centered at 15 S in January and February, which are usually initiated dynamically by remote forcing from the westerly wind anomalies over the western equatorial Atlantic, are able to cause deficit low-cloud cover in subsequent JJA. This is because the increased SST in the region reduces the stability of the lower atmosphere. In a normal year, the seasonal cooling at the sea surface increases the lower atmospheric stability in this region during austral winter and enhances the inversion at the top of the marine boundary layer. This creates favorable condition for the seasonal increase of the low-cloud cover. The low-cloud cover and the inversion are further enhanced by a positive feedback between the cooling near the cloud top, driven by the long-wave radiative heat loss, which enhances mixing within the cloud layer [Philander et al., 1996]. In an anomalous warm event, all these processes are weakened and the cloud cover is reduced. The reduced low-cloud cover in turn can generate positive SST tendency in the southeastern tropical Atlantic, away from the coast and closer to the equator by changing the amount of the local solar radiation reaching the sea surface. The low cloud-radiation-sst feedback may also 15 of 19

Figure 13. Seasonal climatology of the percentage of NCEP/NCAR reanalyzed low cloud cover is shown in (a) for JJA and (b) for SON with contour interval 10. Regions with cloud cover between 50 60 are lightly shaded and those beyond 60 are darkly shaded. The corresponding shortwave flux is shown in (c) for JJA and (d) for SON with contour interval 20 Wm 2. Regions with shortwave flux less than 180 Wm 2 are lightly shaded and those larger than 240 Wm 2 are darkly shaded. The correlations between the NCEP/NCAR and ISCCP cloud anomalies are shown in (e) for all months from July 1983 to June 2004 and (f) JJA seasonal means only. The contour interval is 0.1 with values within ±0.1 omitted. Areas with correlations larger than 0.2 (less than 0.2) are darkly (lightly) shaded. 16 of 19

Figure 14. Difference of composite low cloud (left panels) and shortwave radiation(right panels) anomalies from April to August between deficit and excessive low cloud years as shown in Figure 2b. The contour intervals are 3% for low cloud cover and 5 Wm 2 for shortwave radiation flux. Zero contours are omitted in all figures. Regions above 95% significance level are shaded. play a role in the slow westward expansion of the SST anomalies in late austral winter and spring. Overall, the influence of the cloud fluctuation is an important component in the evolution of the southeastern tropical Atlantic anomalous events. [34] The development of the composite anomalous event described here is largely consistent with the finding by Hu and Huang [2005]. Using historical SST and NCEP/ NCAR reanalyzed meteorological fields since 1950, they conducted a regression analysis of the seasonally averaged ocean-atmosphere data to a statistical mode of the southeastern tropical Atlantic anomalies, identified through a combined rotated EOF analysis of the anomalous SST gradient and wind stress. In particular, they notice that the low-cloud amount from the NCEP/NCAR reanalysis is reduced over the southeastern subtropical-tropical Atlantic during MAM and JJA when the coastal SST anomalies are warm. Our more direct analysis using higher quality cloud data largely confirms this finding. This is encouraging because it demonstrates that the NCEP/NCAR reanalysis cloud product, though a derived variable from model parameterization in a meteorological data assimilation system, may be useful in climate study, especially for the early periods when the satellite data are unavailable. [35] On the other hand, the reanalyzed product has its shortcomings. In this study, we further compared the 17 of 19

reanalyzed cloud products with the ISCCP measurements. Figures 13a 13d shows the mean low-cloud cover and net shortwave radiative fluxes for JJA and SON from the NCEP/NCAR reanalysis from July 1983 to June 2004, which correspond to the ISCCP results shown in Figure 1. The major difference in the southern part of the basin is a southwestward shift of the low-cloud center of the NCEP/ NCAR product during these two seasons so that the largest cloud cover moves away from the Angola-Benguela coast (Figures 1a 1b and Figures 13a 13b). Correspondingly, the minimum shortwave radiation also shifts southwestward. Interannually, the monthly NCEP/NCAR and ISCCP low-cloud anomalies show no correlation in the southeastern tropical Atlantic Ocean for all seasons (Figure 13e). The correlation is only slightly increased for the seasonally averaged anomalies in JJA (Figure 13f). [36] However, the composites of NCEP/NCAR low cloud and shortwave radiation (Figure 14) for the same cloud deficit and excessive years shown in Figure 2b show roughly similar features to those in the ISCCP data during austral winter season (Figures 4 and 5). This suggests that the cloud-sst relation does show some realistic characters in the NCEP/NCAR data during major anomalous events. The major NCEP/NCAR cloud reduction, however, appears earlier in April (Figure 14a) and is too confined to the south of the equator in subsequent months. This is clearly related to its bias in the seasonal means. The pattern of the lowcloud anomalies from the satellite data is different from that of NCEP/NCAR reanalysis in that, closer to the equator, the warm SST anomalies are associated with reduced low cloud in the former but increased one in the latter. A further analysis of the reanalyzed cloud may be needed to understand why such a discrepancy occurs and how further improvements can be made. Meanwhile, one should be cautious to use the reanalysis products as a substitute of the direct measurement. [37] The relatively high correlation between the February ABA index and the JJA low cloud PC suggests that there is predictability of the JJA cloud amount on seasonal scales. Moreover, we believe that a better understanding of the cloud influence is important to predict the southeastern Atlantic anomalous events on seasonal time scales. It has been noticed previously that deficiencies of the low-cloud simulation is a major obstacle for the current coupled oceanatmosphere general circulation models (CGCM) to simulate the mean climate over the tropical eastern Pacific [Ma et al., 1996; Philander et al., 1996; Yu and Mechoso, 1999; Gordon et al., 2000; Gudgel et al., 2001]. Analyzing the NCEP CFS hindcast for 1981 2003, B. Huang et al. [2007] find that the major systematic SST errors in the tropical Atlantic are in the southeastern ocean, which is closely tied to the model s inadequacy in simulating the seasonal enhancement of the low cloud. We speculate that the low skill of the current CGCMs in predicting the interannual variability in the eastern Atlantic region is partly related to their incapability to simulate the SST-radiation-cloud feedback correctly. Analyses are being conducted to further investigate this issue. [38] Acknowledgments. The financial support is provided by a research grant from the NOAA s CLIVAR Atlantic Program (NA04OAR4310115). The satellite cloud and radiation data are provided online by the International Satellite Cloud Climatology Program (ISCCP). The monthly mean satellite altimetry sea surface height fields from the TOPEX/Poseidon and JASON-1 missions are provided online by D.P. Chambers. We would like to thank R. Murtugudde and two anonymous reviewers for their very constructive comments and suggestions. We would also like to thank J. Shukla and J.L. Kinter III for their support of this research and D. Straus and an anonymous reviewer within COLA for their comments on an earlier version of the manuscript. References Atlas, R., R. H. Hoffman, S. C. Bloom, J. C. Jusem, and J. Ardizzone (1996), A multiyear global surface wind velocity dataset using SSM/I wind observations, Bull. Am. Meteorol. Soc., 77, 869 882. Bjerknes, J. (1969), Atmospheric teleconnections from the equatorial Pacific, Mon. Weather Rev., 97, 163 172. Carton, J. A., and B. Huang (1994), Warm events in the tropical Atlantic, J. Phys. Oceanogr., 24, 888 903. Chambers, D. P., S. A. Hayes, J. C. Ries, and T. J. Urban (2003), New TOPEX sea state bias models and their effect on global mean sea level, J. Geophys. Res., 108(C10), 3305, doi:10.1029/2003jc001839. Chou, S.-H., E. Nelkin, J. Ardizzone, R. Atlas, and C.-L. Shie (2003), Surface turbulent heat and momentum fluxes over global oceans based on the Goddard satellite retrievals, version 2 (GSSTF2), J. Clim., 16, 3256 3273. Enfield, D. B., and D. A. Mayer (1997), Tropical Atlantic sea surface temperature variability and its relation to El Niño Southern Oscillation, J. Geophys. Res., 102, 929 945. Florenchie, P., C. J. C. Reason, J. R. E. Lutjeharms, M. Rouault, C. Roy, and S. Masson (2004), Evolution of interannual warm and cold events in the southeast Atlantic Ocean, J. Clim., 17, 2318 2334. Gordon, C. T., A. Rosati, and R. Gudgel (2000), Tropical sensitivity of a coupled model to specified ISCCP low clouds, J. Clim., 13, 2239 2260. Gudgel, R. G., A. Rosati, and C. T. Gordon (2001), The sensitivity of a coupled atmospheric-oceanic GCM to prescribed low-level clouds over the ocean and tropical landmasses, Mon. Weather Rev., 129, 2103 2115. Hirst, A., and S. Hastenrath (1983), Atmosphere-ocean mechanisms of climate anomalies in Angola-tropical Atlantic sector, J. Phys. Oceanogr., 13, 1146 1157. Hu, Z.-Z., and B. Huang (2005), Physical processes associated with tropical Atlantic SST gradient. part II: The anomalous evolution in the southeastern ocean, COLA Tech. Rep. 194, 43 pp., Cent. for Ocean-Land- Atmos. Stud., Calverton, Md. (Available at http://www.iges.org/pubs/ tech.html) Hu, Z.-Z., and B. Huang (2007), The predictive skill and the most predictable pattern in the tropical Atlantic: The effect of ENSO, Mon. Weather Rev., in press. Huang, B., and J. Shukla (1997), Characteristics of the interannual and decadal variability in a general circulation model of the tropical Atlantic Ocean, J. Phys. Oceanogr., 27, 1693 1712. Huang, B., and J. Shukla (2005), Ocean-atmosphere interactions in the tropical and subtropical Atlantic Ocean, J. Clim., 18, 1652 1672. Huang, B., P. S. Schopf, and J. Shukla (2004), Intrinsic ocean-atmosphere variability of the tropical Atlantic Ocean, J. Clim., 17, 2058 2077. Huang, B., Z.-Z. Hu, and B. Jha (2007), Evolution of model systematic errors in the tropical Atlantic basin from coupled climate hindcasts, Clim. Dyn., doi:10.1007/s00382-006-0223-8. (Available at http://www. springerlink.com/content/3477115471337462/) Kalnay, E., et al. (1996), The NCEP/NCAR 40-year reanalysis project, Bull. Am. Meteorol. Soc., 77, 437 471. Klein, S. A., and D. L. Hartmann (1993), The seasonal cycle of low stratiform clouds, J. Clim., 6, 1587 1606. Ma, C.-C., C. R. B. Mechoso, A. W. Robertson, and A. Arakawa (1996), Peruvian stratus clouds and the tropical Pacific circulation: A coupled ocean-atmosphere GCM study, J. Clim., 9, 1625 1645. Nigam, S. (1997), The annual warm to cold phase transition in the eastern equatorial Pacific: Diagnosis of the role of stratus cloud-top cooling, J. Clim., 10, 2447 2467. Norris, J. R., and C. B. Leovy (1994), Interannual variability in stratiform cloudiness and sea surface temperature, J. Clim., 7, 1915 1925. 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 of the equator, J. Clim., 9, 2958 2972. Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang (2002), An improved in situ and satellite SST analysis for climate, J. Clim., 15, 1609 1625. Rossow, W. B., and E. N. Dueñas (2004), The International Satellite Cloud Climatology Project (ISCCP) Web site: An online resource for research, Bull. Am. Meteorol. Soc., 85, 167 172. Saha, S., et al. (2006), The NCEP climate forecast system, J. Clim., 19, 3483 3517. 18 of 19

Schneider, E. K., D. G. DeWitt, A. Rosati, B. P. Kirtman, L. Ji, and J. J. Tribbia (2003), Retrospective ENSO forecasts: Sensitivity to atmospheric model and ocean resolution, Mon. Weather Rev., 131, 3038 3060. Servain, S., J. Picaut, and J. Merle (1982), Evidence of remote forcing in the equatorial Atlantic Ocean, J. Phys. Oceanogr., 12, 457 463. Shannon, L. V., A. J. Boyd, G. B. Bundrit, and J. Taunton-Clark (1986), On the existence of an El Niño-type phenomenon in the Benguela system, J. Mar. Sci., 44, 495 520. Tanimoto, Y., and S.-P. Xie (2002), Inter-hemispheric decadal variations in SST, surface wind, heat flux, and cloud cover over the Atlantic Ocean, J. Meteorol. Soc. Jpn., 80, 1199 1219. Wang, C. (2002a), Atmospheric circulation cells associated with the El Niño Southern Oscillation, J. Clim., 15, 399 419. Wang, C. (2002b), Atlantic climate variability and its associated atmospheric circulation cells, J. Clim., 15, 1516 1536. Wang, C., and D. B. Enfield (2001), The tropical Western Hemisphere warm pool, Geophys. Res. Lett., 28, 1635 1638. Wang, C., and D. B. Enfield (2003), A further study of the tropical western hemisphere warm pool, J. Clim., 16, 1476 1493. Weare, B. C. (1994), Interrelationship between cloud properties and sea surface temperatures on seasonal and interannual time scales, J. Clim., 7, 248 260. Xie, S.-P. (2004), The shape of continents, air-sea interaction, and the rising branch of the Hadley circulation, in The Hadley Circulation, Present, Past, and Future, edited by H. F. Diaz and R. S. Bradley, pp. 121 152, Springer, New York. Xie, S.-P., and J. A. Carton (2004), Tropical Atlantic variability: Patterns, mechanisms, and impacts, in Earth s Climate: The Ocean-Atmosphere Interaction, Geophys. Monogr. Ser., vol. 147, edited by C. Wang, S.-P. Xie, and J. A. Carton, pp. 121 142, AGU, Washington, D. C. Yu, J.-Y., and C. R. Mechoso (1999), Links between annual variations of Peruvian stratus clouds and of SST in the eastern equatorial Pacific, J. Clim., 12, 3305 3318. Zebiak, S. E. (1993), Air-sea interaction in the equatorial Atlantic region, J. Clim., 8, 1567 1586. Zebiak, S. E., and M. A. Cane (1987), A model El Niño Southern Oscillation, Mon. Weather Rev., 115, 2262 2278. Zhang, Y., W. B. Rossow, A. A. Lacis, V. Oinas, and M. I. Mishchenko (2004), Calculation of radiative fluxes from the surface to top of atmosphere based ISCCP and other global data sets: Refinements of the radiative transfer model and the input data, J. Geophys. Res., 109, D19105, doi:10.1029/2003jd004457. Z.-Z. Hu and B. Huang, Center for Ocean-Land-Atmosphere Studies, Institute of Global Environment and Society, 4041 Powder Mill Road, 302, Calverton, MD 20705, USA. (huangb@cola.iges.org) 19 of 19