Link Between the Double-Intertropical Convergence Zone Problem and Cloud Biases Over the Southern Ocean

Link between the double-Intertropical Convergence Zone problem and cloud biases over the Southern Ocean

Yen-Ting Hwang1 and Dargan M. W. Frierson

Department of Atmospheric Sciences, University of Washington, Seattle, WA 98195-1640

Edited by Mark H. Thiemens, University of California at San Diego, La Jolla, CA, and approved February 15, 2013 (received for review August 2, 2012) The double-Intertropical Convergence Zone (ITCZ) problem, in which climate models show that the bias can be reduced by changing excessive precipitation is produced in the Southern Hemisphere aspects of the convection scheme (e.g., refs. 7–9) or changing the tropics, which resembles a Southern Hemisphere counterpart to the surface wind stress formulation (e.g., ref. 10). Given the complex strong Northern Hemisphere ITCZ, is perhaps the most significant feedback processes in the tropics, it is challenging to understand and most persistent bias of global climate models. In this study, we the mechanisms by which the sensitivity experiments listed above look to the extratropics for possible causes of the double-ITCZ improve tropical precipitation. problem by performing a global energetic analysis with historical Recent work in general circulation theory has suggested that simulations from a suite of global climate models and comparing one should not only look within the tropics for features that affect with satellite observations of the Earth’s energy budget. Our results tropical precipitation. A set of idealized experiments showed that show that models with more energy flux into the Southern heating a global climate model exclusively in the extratropics can Hemisphere atmosphere (at the top of the atmosphere and at the lead to tropical rainfall shifts from one side of the tropics to the surface) tend to have a stronger double-ITCZ bias, consistent with other (11). They also showed that extratropical cloud responses recent theoretical studies that suggest that the ITCZ is drawn to- are important in determining the magnitude of the tropical pre- ward heating even outside the tropics. In particular, we find that cipitation shift (12). The physical mechanism for extratropical cloud biases over the Southern Ocean explain most of the model- connections is based on energetic constraints (11, 12) and essen- to-model differences in the amount of excessive precipitation in tially argues that tropical precipitation shifts toward whichever Southern Hemisphere tropics, and are suggested to be responsible for hemisphere is heated more at the surface or top-of-atmosphere. this aspect of the double-ITCZ problem in most global climate models. This theoretical framework is consistent with recent simulations of Last Glacial Maximum conditions (13), water-hosing experi- tropical precipitation | model biases | cloud radiative forcing | ments (14, 15), and aerosol cooling simulations (16), which also atmospheric energy transport | general circulation show that tropical rainfall shifts away from high latitude cooling (see also a recent review (17)). A global climate model inter- recipitation is essential to life, with its variation tightly linked comparison study (18) applied this framework to interpret Pto water and food security. Providing the best estimate of tropical rainfall shifts in global warming simulations from a suite future trends in precipitation has always been a primary goal of of global climate models, showing that tropical rainfall always global climate models. For this reason, global climate models are shifts toward the hemisphere with more heating, with a particu- closely scrutinized not only on their ability to simulate large-scale larly important role played by extratropical clouds. dynamics but also on their skill in simulating precipitation dis- In this paper, we use the energetic theoretical framework to tributions at regional scales. One naturally only trusts model evaluate the degree to which extratropical biases contribute to forecasts of precipitation if there is substantial fidelity in simu- the double-ITCZ problem. We first analyze zonal mean tropical

lating the current precipitation climatology. precipitation biases in the global climate models included in EARTH, ATMOSPHERIC, Because precipitation features are related with processes oc- Phase 5 of the World Climate Research Program Coupled AND PLANETARY SCIENCES curring at a tremendous range of time and spatial scales, their Model Intercomparison Project (CMIP5). We then show that simulation remains challenging. The main precipitation feature model biases outside the tropics have a significant effect on the that most global climate models have difficulty capturing is the interhemispheric temperature asymmetry, tropical circulations, Intertropical Convergence Zone (ITCZ) in the deep tropics at and tropical rainfall. Last, we show that cloud biases over the around 6°N, a narrow latitude band with some of the most in- Southern Ocean introduce anomalous warming in the Southern tense rainfall on Earth. Despite decades of work by modeling Hemisphere and are the primary contributor to the double-ITCZ centers around the world, the double-ITCZ problem, in which bias in most global climate models. excessive precipitation is produced in the Southern Hemisphere tropics resembling the stronger Northern Hemisphere ITCZ, Double-ITCZ Problem remains the largest precipitation bias of most state-of-the-art Comparing Fig. 1 A and B shows a range of precipitation biases global climate models. There has been little progress in reducing in the multimodel mean (see Fig. S1A for the difference between this bias over recent years (1–3) (Figs. 1 A and B and 2A). Fig. 1 A and B). Although it should be noted that these biases The double-ITCZ bias is most apparent in the strip 5–15°S vary considerably from model to model, we briefly summarize over the central and east Pacific, and a similar feature is visible in the important features of tropical precipitation biases. Models the Indian and Atlantic Oceans in most models. Most of the proposed reasons for tropical precipitation biases involve local mechanisms within or close to the tropics, for example, warm sea Author contributions: Y.-T.H. and D.M.W.F. designed research; Y.-T.H. performed surface temperature errors in the coastal upwelling region off research; Y.-T.H. analyzed data; and Y.-T.H. and D.M.W.F. wrote the paper. Peru (4, 5), often compounded by a deficit of stratocumulus (1, 4, The authors declare no conflict of interest. 6). Previous studies (e.g., ref. 2) link the cold sea surface tem- This article is a PNAS Direct Submission. perature bias especially along the equatorial Pacific with overly 1To whom correspondence should be addressed. E-mail: [email protected]. fi strong trade winds, excessive evaporation, and insuf cient solar This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. damping by clouds. Sensitivity experiments with individual global 1073/pnas.1213302110/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1213302110 PNAS | March 26, 2013 | vol. 110 | no. 13 | 4935–4940 AB

Fig. 1. Latitude–longitude maps of precipitation and shortwave cloud radiative forcing in observations and global climate models. Annual mean precipitation, 1985–2004 from (A) the Global Precipitation Climatology Project (GPCP), version 2.1, and (B) the ensemble mean of historical simulations of 20 CMIP5 global climate models. (C) Shortwave cloud radiative forcing from satellite observations [Cloud and the Earth’s Radiant Energy System (CERES)], 2001–2010. (D)Biasesin shortwave cloud radiative forcing in the ensemble mean of historical simulations of 20 CMIP5 global climate models (departure from observations). tend to overprecipitate over all ocean basins in the Southern much rainfall. Many of these features are similar to the CMIP3 Hemisphere tropics. The Northern Hemisphere tropical ocean biases shown in the Intergovernmental Panel on Climate Change basins behave differently: a negative precipitation anomaly exists (IPCC) Fourth Assessment Report (3) and are described in in the Atlantic, whereas the Indian and Pacific have positive detailed in SI Text. anomalies that are smaller than the positive anomalies south of As shown in Fig. 2A, the biases in zonal mean precipitation in the equator. In the immediate vicinity of the equator in the CMIP3 global climate models (2) still exist in CMIP5 global Pacific Ocean, there is a large negative precipitation anomaly. climate models. First of all, comparing with Global Precipitation Central America, the Amazon, and India all have large deficits Climatology Project (GPCP) observational data, most models of rainfall in the multimodel mean, whereas Indonesia has too simulate excessive precipitation in the entire tropics. However,

Fig. 2. Zonal mean and hemispheric asymmetry of precipitation, temperature, and shortwave cloud radiative forcing. Annual mean zonal mean of (A) precipitation, (C) surface air temperature, and (E) shortwave cloud radiative forcing. (B, D, and F) The interhemispheric asymmetry (NH minus SH) of A, C,and E, respectively. Each line is colored according to the atmospheric cross-equatorial energy transport in each global climate model. The black lines are from GPCP and CERES observations. The gray lines represent the SD of year-to-year variability in observations.

4936 | www.pnas.org/cgi/doi/10.1073/pnas.1213302110 Hwang and Frierson recent studies suggest that global precipitation products such as Energetic Analysis GPCP observational data and Tropical Rainfall Measuring The tropical precipitation asymmetry indices of CMIP5 models Mission satellite data may underestimate tropical precipitation are correlated with their interhemispheric temperature asym- from low-topped clouds (19–21), so we do not focus on the ex- metry (Fig. 3A; R = 0.73) and interhemispheric tropical tem- cessive precipitation bias due to the observational uncertainties. perature asymmetry (R = 0.89), consistent with previous studies The second primary bias in zonal mean precipitation that we that identify that sea surface temperature has a strong control on instead focus on is that the excessive precipitation in the tropical precipitation (22, 23). In addition, from the theoretical Southern Hemisphere tropics is much more pronounced than in energetic framework (11, 12, 18, 24), the tropical precipitation the Northern Hemisphere tropics. asymmetry indices are also expected to be anticorrelated with The tendency for too much precipitation especially in the the atmospheric cross-equatorial energy transports (Fig. 3B; Southern Hemisphere tropics can be seen best in Fig. 2B, which R = −0.89). Energy transport between the hemispheres is primarily shows Northern Hemisphere minus Southern Hemisphere pre- performed by the upper branch of the Hadley cell, whereas cipitation. Nearly all models lie below the observed curve, in- moisture is concentrated near the surface and is thus transported dicating too much precipitation falling on the south of the in the opposite direction, by the lower branch of the Hadley cell. equator. Four blue-colored models are an exception to this and actually have too much rain in the Northern Hemisphere tropics. As shown in Fig. 4, one would expect anomalous southward To quantify the double-ITCZ bias, we calculate a tropical moisture transport along with anomalous northward energy precipitation asymmetry index, deﬁned in Materials and Methods, transport across the equator in the models that simulate exces- which removes the effect of overall excessive precipitation. As sive precipitation in the Southern Hemisphere tropics. shown in x axis in Fig. 3 A and B, all but four models’ tropical The relationship between tropical precipitation, surface air precipitation asymmetry indices are lower than observed, and temperature, and atmospheric cross-equatorial energy transport three models have a negative asymmetry index, indicating that can be seen in the zonal mean precipitation and surface air tem- B D B D they have more precipitation in the Southern Hemisphere tropics perature plots (Fig. 2 and ). In Fig. 2 and , it is clear that than in the Northern Hemisphere tropics. The tropical pre- the models with insufﬁcient southward energy transport at the cipitation asymmetry index is correlated with the latitude of the equator (red colors) are the models that have a more severe center of Hadley circulation in these models (Fig. S2), with models double-ITCZ problem (smaller tropical precipitation asymmetry having lower tropical precipitation asymmetry index also having index) and a smaller interhemispheric temperature asymmetry. a southward displacement of their Hadley circulation center. In Models that receive more precipitation in northern tropics (larger addition, our asymmetry index represents the tropical precipita- precipitation asymmetry index) have a larger interhemispheric tion biases in most regions in terms of both the multimodel mean temperature difference and a southward transport of energy at the and the models’ spread (Fig. S1 A and B and Table S1). equator (blue colors).

AB E EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES

Fig. 3. Global energetic analysis of the tropical precipitation asymmetry. (A) Interhemispheric temperature asymmetry versus tropical precipitation asymmetry index. (B) Atmospheric cross-equatorial energy transport (northward transport is positive) versus tropical precipitation asymmetry index. (C) Asymmetry in extratropical shortwave cloud radiative forcing (20°N∼North Pole minus 20°S∼South Pole) versus atmospheric cross-equatorial energy transport. (D) Asymmetry in extratropical shortwave cloud radiative forcing versus tropical precipitation asymmetry index. (E) Attribution of tropical precipitation asymmetry index based on the energetic framework (Materials and Methods): from Left to Right, the hemispheric asymmetry of extratropical shortwave cloud radiative forcing (CS), longwave cloud radiative forcing (CL), the noncloud shortwave effect (NS), the noncloud longwave effect (NL), and the upward surface flux (O). Each dot is colored according to the atmospheric cross-equatorial energy transport in each global climate model (y axis in B). The open circles in A–D are from the global climate model with the largest bias in surface flux (see O term in E). The black Xs are observations. The gray lines and gray Xs represent the SD of year-to-year variability in observations. Some of the SDs are too small to be visible in the figures.

Hwang and Frierson PNAS | March 26, 2013 | vol. 110 | no. 13 | 4937 Hemisphere in these models is radiated back to space and does not influence the atmospheric cross-equatorial transport. The asymmetry of longwave cloud radiative forcing has very little spread among models (CL in Fig. 3E), as longwave cloud radiative forcing is generally smaller than shortwave cloud radiative forcing in the extratropics. Biases in the noncloud shortwave asymmetry are not negligible (NS in Fig. 3E). Most of its spread is dominated by biases in sea ice around Antarctica (60°S∼80°S), with some additional contribution from midlatitude planetary albedo (surface albedo or aerosol scattering effect) (Fig. S3 C and D). In most global climate models, biases in surface flux (ocean circulation) are relatively small compared with other terms (O in Fig. 3E), although we find this term to be dominant in causing the observed hemispheric asymmetry of precipitation, a point we are Fig. 4. Schematic of the proposed mechanism for the double-ITCZ bias. The currently examining in a related study. anomalous energy fluxes and circulation in multimodel mean relative to observations are plotted. Most models simulate too much incoming short- Shortwave Cloud Radiative Forcing Bias wave radiation over the Southern Ocean due to cloud biases, which results in With their strong radiative effects, clouds have the potential to anomalously high temperatures in the Southern Hemisphere midlatitudes. influence circulations both locally and nonlocally by affecting Similar cloud biases exist in Northern Hemisphere midlatitudes, but to atmospheric energy transports (25, 26). Fig. 1C shows the a much smaller degree. The anomalous heating in the Southern Ocean is shortwave cloud radiative forcing in satellite observations. Mul- spread into the Southern Hemisphere tropics by baroclinic eddies. An timodel mean biases are in Fig. 1D and zonal mean shortwave anomalous Hadley circulation is induced to transport energy from the cloud radiative forcing in each model compared with satellite Southern Hemisphere to the Northern Hemisphere (the red arrow across the E equator), and keep the tropical tropospheric temperatures relatively flat. observations is in Fig. 2 . The largest spread of global climate Because water vapor is concentrated in the lower troposphere, this anom- models’ shortwave cloud radiative forcing is in the deep tropics alous Hadley circulation transports moisture southward (the blue arrow and in the Southern Hemisphere midlatitudes. The former is across the equator), and results in excessive precipitation in Southern a result of the double-ITCZ problem, whereas the latter is re- Hemisphere tropics. Other biases in energy fluxes at the top of the atmo- lated with cloud biases over the Southern Ocean. Based on our sphere or at the surface that have hemispheric asymmetry can also affect energetic analysis, cloud biases over the Southern Ocean are the tropical precipitation through this mechanism. main cause of biases in atmospheric cross-equatorial energy transport and the tropical precipitation asymmetry index. In Given the high correlations in Fig. 3B and the clear relationships addition to their effect on cross-equatorial energy transport and seen in Fig. 2 B and D, we can gain some insight into the sources of tropical precipitation, these cloud biases have also been shown to lead to biases in jet stream latitude (27) and total meridional the double-ITCZ problem by understanding the causes of the biases energy transport (28). in atmospheric cross-equatorial energy transports and interhemi- Clouds over the Southern Ocean make a significant contribu- spheric temperature asymmetry. We decompose the atmospheric tion to the top-of-the-atmosphere radiation balance; however, cross-equatorial energy transports into different terms, using the they are often poorly simulated by global climate models (29, 30). methodology described in Materials and Methods, and claim that In most models, these biases lead to anomalous heating, i.e., the whatever causes anomalous heating in the Southern Hemisphere simulations have too low cloud optical thickness and/or too low atmosphere relative to the Northern Hemisphere atmosphere is cloud fraction. Similar magnitudes of shortwave cloud radiative also what causes the anomalous northward cross-equatorial energy forcing biases are reported in reanalyses and in some case study transport and the double-ITCZ problem in the models. simulations, in which sea surface temperature and large-scale The results of the energetic analysis for each model are shown fi E dynamics conditions are constrained (29, 31). These ndings in Fig. 3 , comparing with observations shown in black crosses. suggest that biases are more likely due to problems with cloud Although every component contributes to some of the hemi- parameterizations or missing local processes, although the causes spheric asymmetry and cross-equatorial energy transport, short- are far from being fully understood and may be model dependent. wave cloud radiative forcing is clearly responsible for most of the E spread and bias (CS in Fig. 3 ). The correlation between the Conclusions and Discussions asymmetry of extratropical shortwave cloud radiative forcing and We have introduced a framework to identify the cause of the dou- − C atmospheric cross-equatorial energy transport is 0.80 (Fig. 3 ), ble-ITCZ problem in global climate models and have found that and between the asymmetry of extratropical shortwave cloud cloud biases over the Southern Ocean are largely responsible. Many radiative forcing and the tropical precipitation asymmetry index models struggle to produce enough cloud fraction and thick enough is 0.80 (Fig. 3D). Global climate models with more severe dou- clouds over the often-overcast Southern Ocean. The lack of cloud ble-ITCZ problems (lower tropical precipitation asymmetry in- results in anomalously high temperatures over the Southern dices) tend to have too little reflection from extratropical clouds Hemisphere as a whole, and also a southward shift in tropical pre- and do not simulate the observed asymmetry of extratropical cipitation, reflected in the double-ITCZ problem. It has been con- shortwave cloud radiative forcing (red models in Fig. 2 E and F, firmed in a number of previous studies that heating anomalies of the more in the next section). magnitude and location of that induced by the shortwave cloud The asymmetry of extratropical noncloud longwave effect has radiative forcing biases over the Southern Ocean can lead to sig- E a large amount of spread as well (NL in Fig. 3 ), however this is nificant ITCZ shifts (11, 12). Models with less of the cloud bias over anticorrelated with the asymmetry of extratropical shortwave the Southern Ocean have less of a double-ITCZ problem. cloud radiative forcing and the tropical precipitation asymmetry One always has to be cautious when inferring causality by index. This is because models with deficient shortwave reflection analyzing correlations. In this case, however, we think it is very from clouds over Southern Ocean (the red models in Fig. 2 E and unlikely that the double ITCZ would be the cause instead of the F) also have a comparatively warmer Southern Hemisphere (Fig. result of the shortwave cloud radiative forcing bias over the 2D) and radiate more energy to space. In other words, some of the Southern Ocean (see additional discussion and experiments in additional solar radiation that reaches the surface in the Southern SI Text and Fig. S4). Rather, cloud parameterizations are likely

4938 | www.pnas.org/cgi/doi/10.1073/pnas.1213302110 Hwang and Frierson the primary cause of these biases, although there certainly may minus precipitation in Southern Hemisphere tropics (equator to 20°S) nor- be coupled aspects to this process as well, e.g., higher temper- malized by the tropical mean precipitation (20°S∼20°N). atures leading to further burn-off of low clouds. Future modeling Interhemispheric temperature asymmetry. Interhemispheric temperature asym- fi experiments are needed to quantify the exact contribution of metry is de ned as follows: the difference between the hemispheric mean fi surface air temperatures, north minus south. When calculating tropical clouds in each model, and ultimately modi cations of parame- temperature asymmetry, averages are taken only to 20°. terizations will be necessary to see whether this can lead to al- Latitude of the center of Hadley circulation. The latitude of the center of Hadley leviation of biases. circulation is defined as follows: the latitudinal location of the zero crossing It should also be pointed out that solving the tropical pre- of the 500-hPa stream function in between the two Hadley cells. cipitation asymmetry problem would not be sufficient to eliminate Shortwave (longwave) cloud radiative forcing. Shortwave (longwave) cloud ra- all tropical precipitation biases. The double-ITCZ problem has diative forcing is defined as follows: the top of atmosphere downward a rich longitudinal structure. As shown in previous studies (1, 2), shortwave (longwave) flux in all sky minus clear sky. The shortwave forcing is part of the double-ITCZ problem involves the fact that most always negative, because clouds reflect shortwave, and the longwave forcing global climate models do not simulate the correct tilt of the South is always positive, because clouds reduce the outgoing longwave radiation at Pacific convergence zone, a band of precipitation extending from the top of the atmosphere. fi Noncloud shortwave (longwave) effect. Noncloud shortwave (longwave) effect is the west Paci c warm pool southeastward toward French Poly- defined as follows: the net downward shortwave (longwave) flux at the top nesia, which we have not addressed in this study. In addition, even of the atmosphere in clear sky. The shortwave effect is determined by in- in the zonal mean, there are other tropical precipitation biases in solation, aerosols, water vapor, and other atmospheric properties, and the CMIP5, namely, the excessive precipitation problem and the longwave effect is related with temperature at the surface and in the at- tendency for precipitation to minimize too much on the equator. mosphere, and other atmospheric properties that affect the emissivity of Although the former will require better observations to solve, the the atmosphere. latter is a problem which is not addressed by our framework (and Surface flux. Surface flux is defined as follows: the net energy flux from the may be more tropical in origin). surface to the atmosphere, which is equal to the convergence of ocean heat Besides highlighting the importance of simulating clouds over transport plus the change in ocean energy content. fi Atmospheric cross-equatorial energy transport. We calculate the atmospheric the Southern Ocean with delity, this study introduces a frame- cross-equatorial energy transport in reanalysis data by integrating the moist work to interpret results from modeling experiments that affect static energy divergence in ERA-Interim to obtain the meridional moist static tropical precipitation biases. Modifying a parameterization or local energy flux at the equator. The atmospheric cross-equatorial energy trans- process has a global effect, and tropical precipitation biases could port in each global climate model is calculated by integrating the atmospheric be reduced through unexpected processes outside the tropics. energy budget as follows:

Materials and Methods Z0 Z2π Zπ=2 Z2π ðϕ = Þ = 2 ϕ λ ϕ = − 2 ϕ λ ϕ; Climate Model Simulations. The simulations analyzed in this study are from FA 0 QAa cos d d QAa cos d d asuiteofclimatemodelsincludedinPhase5andPhase3oftheWorldClimate −π=2 0 0 0 Research Program Coupled Model Intercomparison Project (CMIP5 and CMIP3) ðϕ = Þ ϕ multimodel datasets, which will be used in the IPCC Fifth Assessment Report where FA 0 is the atmospheric cross-equatorial energy transport, is λ (AR5) and were used in the IPCC Fourth Assessment Report (AR4), respectively. latitude, is longitude, a is radius of the Earth, and QA is the atmospheric Results from CMIP5 are presented in the main text and the comparison of CMIP3 energy budget, which is the sum of the shortwave cloud radiative forcing, and CMIP5 are shown in SI Text (Fig. S5). The models are listed in Tables S2 and the longwave cloud radiative forcing, the noncloud shortwave effect and S3. We only include simulations that provide sufﬁcient data for the analysis. All the noncloud longwave effect. analyses are performed by examining 20-y averages from the last 20 y Asymmetry of extratropical shortwave cloud radiative forcing. Asymmetry of (1985∼2004) of historical integrations from CMIP5 and the last 20 y (1980∼1999) extratropical shortwave cloud radiative forcing is deﬁned as follows: short- of the 20th century climate integrations from CMIP3. The results are not sen- wave cloud radiative forcing in the Northern Hemisphere extratropics sitive to the time periods; model biases in the quantities we analyze are an (20°N∼North Pole, area weighted) minus shortwave cloud radiative forcing order of magnitude larger than interannual variability or long-term trends. in the Southern Hemisphere extratropics (20°S∼South Pole). The same cal- culation is applied for the asymmetry of extratropical longwave cloud ra- Climate Data. The precipitation data are from the GPCP (32, 33), from year diative forcing, the noncloud shortwave effect, the noncloud longwave EARTH, ATMOSPHERIC,

1985 to year 2004. Energy fluxes at the top of the atmosphere are from effect, and the atmospheric energy budget. A positive sign in these terms AND PLANETARY SCIENCES Clouds and the Earth’s Radiant Energy System (CERES) Energy Balanced and implies a southward contribution to the cross-equatorial energy transport. Filled data set, covering the time period of 2000–2010 (34). We calculate the implied surface flux as in ref. 35 using CERES data and European Centre for Energetic Analysis. For the 10- and 20-y averages we consider in this study, the Medium-Range Weather Forecasts Re-Analysis Interim (ERA-Interim) data atmospheric budgets are very close to steady state. In steady state, an energy (36) over the same time period of 2000–2010. deficit or surplus at the top of the atmosphere and the surface at a given latitude is balanced by divergence of atmospheric energy transport into or Definitions and Calculations of Different Terms. Tropical precipitation asymmetry away from that latitude. This framework can be applied to decompose at- index. The tropical precipitation asymmetry index is defined as follows: pre- mospheric cross-equatorial energy transport. For example, if the surface flux cipitation in Northern Hemisphere tropics (equator to 20°N, area-averaged) in the Northern Hemisphere is larger than in the Southern Hemisphere in

Table 1. Correlation coefﬁcients between terms considered in the study

Asymmetry Index PrI ITTA ITA FA (0)

PrI ITTA 0.89 (0.89) ITA 0.73 (0.73) 0.66 (0.67)

FA (0) −0.88 (−0.89) −0.75 (−0.77) −0.77 (−0.78) Asymmetry in extratropical 0.64 (0.80) 0.52 (0.70) 0.64 (0.76) −0.72 (−0.80) shortwave cloud radiative forcing

Correlation coefﬁcients calculated when excluding IPSL-CM5A-LR, the model with anomalous surface ﬂux, are in parentheses. The

correlations that are signiﬁcant with P value smaller than 0.01 are in bold. FA (0), atmospheric cross-equatorial transport; ITA, interhemispheric surface air temperature asymmetry; ITTA, interhemispheric tropical surface air temperature asymmetry; PrI, precipitation asymmetry index.

Hwang and Frierson PNAS | March 26, 2013 | vol. 110 | no. 13 | 4939 a model, surface flux contributes to a southward atmospheric energy transport equivalently, the degree to which the Southern Hemisphere atmosphere is across the equator in this model. heated more than the Northern Hemisphere. From red to blue colors are the This same energetic attribution technique has been used in previous studies global climate models with the smallest to largest southward cross-equatorial (18, 28, 37, 38) for understanding the changes in atmospheric energy transport energy transport, and the model names in Tables S2 and S3 follow the with global warming. One difference between our analysis and those done in same order. The black lines and the black X symbols in figures are from previous studies is that we do not consider fluxes within the tropics. There are observations or reanalysis data. Error bars are estimated with the SD of energetic responses to ITCZ shifts within the tropics that we want to avoid in year-to-year variability. this energetic analysis to maximize the chances that the terms are a cause of There is one model that does not capture the observed hemispheric rather than an effect of the double-ITCZ problem. For example, models with asymmetry in surface flux, which is plotted as an open circle in Fig. 3 A–D.For a more severe double-ITCZ problem (lower tropical precipitation asymmetry this model, biases in oceanic transport offset some of the positive biases in indices) tend to have less outgoing longwave radiation in the Southern shortwave cloud radiative forcing asymmetry, which can be seen in Fig. 3E. Hemisphere tropics, due to more high clouds and stronger water vapor All of the correlation coefficients in the text are calculated excluding this greenhouse effect in the locations where they precipitate (Fig. S3 B and F). The model, and those calculated including all models can be found in Table 1. spread in the hemispheric asymmetry of the extratropical atmospheric energy budget explains most of the spread in the cross-equatorial energy transport, ACKNOWLEDGMENTS. We acknowledge John Fasullo for providing the with correlation coefficient 0.91 (the correlation coefficient between the ERA-Interim energy transports, Paulo Ceppi for assistance with the CMIP5 tropical asymmetry and the cross-equatorial energy transport is 0.76). In ad- outputs, Chris Bretherton and Dennis Hartmann for helpful conversions, and two anonymous reviewers for constructive comments. The Cloud and dition, the hemispheric asymmetries of the extratropical atmospheric energy the Earth’s Radiant Energy System data were obtained from the National ’ budget for both the multimodel mean and the models spread are both about Aeronautics and Space Administration Langley Research Center Atmospheric 1.7 times larger than the corresponding hemispheric asymmetries of the Science Data Center. We acknowledge the World Climate Research Pro- tropical atmospheric energy budget. gramme’s Working Group on Coupled Modelling, which is responsible for Fig. 3E presents the result of the energetic analysis. For each term, we CMIP, and we thank the climate modeling groups (listed in Tables S2 and S3) calculate the hemispheric asymmetry of the extratropical flux from that for producing and making available their model output. For CMIP, the US Department of Energy’s Program for Climate Model Diagnosis and Intercom- term (as defined above). parison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Figures and Tables. All figures in this paper are colored in order of the at- Science Portals. Y.-T.H. and D.M.W.F. are supported by National Science mospheric cross-equatorial energy transport in each global climate model, or, Foundation Grants ATM-0846641 and ATM-0936059.

1. Mechoso CR, et al. (1995) The seasonal cycle over the tropical Pacific in coupled 20. Trenberth KE, Fasullo JT, Kiehl J (2009) Earth’s global energy budget. Bull Am ocean–atmosphere general circulation models. Mon Weather Rev 123(9):2825–2838. Meteorol Soc 90(3):311–324. 2. Lin J-L (2007) The double-ITCZ problem in IPCC AR4 coupled GCMs: Ocean–atmosphere 21. Stephens GL, et al. (2012) An update on Earth’s energy balance in light of the latest feedback analysis. JClim20(18):4497–4525. global observations. Nat Geosci 5:691–696. 3. Solomon S, et al. (2007) Climate Change 2007: The Physical Science Basis (Cambridge 22. Dai A (2006) Precipitation characteristics in eighteen coupled climate models. J Clim Univ Press, Cambridge, UK). 19(18):4605–4630. 4. Gordon CT, Rosati A, Gudgel R (2000) Tropical sensitivity of a coupled model to 23. Xie S-P, et al. (2010) Global warming pattern formation: Sea surface temperature and specified ISCCP low clouds. J Clim 13(13):2239–2260. rainfall. J Clim 23(4):966–986. 5. Large WG, Danabasoglu G (2006) Attribution and impacts of upper-ocean biases in 24. Yoshimori M, Broccoli AJ (2008) Equilibrium response of an atmosphere-mixed layer – CCSM3. J Clim 19(11):2325 2346. ocean model to different radiative forcing agents: Global and zonal mean response. 6. Ma C-C, Mechoso CR, Roberton AW, Arakawa A (1996) Peruvian stratus clouds and J Clim 21(17):4399–4423. fi – the tropical Paci c circulation: A coupled ocean atmosphere GCM study. J Clim 9(7): 25. Hwang Y-T, Frierson DMW (2010) Increasing atmospheric poleward energy transport – 1635 1645. with global warming. Geophys Res Lett 37(24):L24807. 7. Frey H, Latif M, Stockdale T (1997) The coupled GCM ECHO-2. Part I: The tropical 26. Hwang Y-T, Frierson DMW, Kay JE (2011) Coupling between Arctic feedbacks and fi – Paci c. Mon Weather Rev 125(5):703 720. changes in poleward energy transport. Geophys Res Lett 38(17):L17704. 8. Zhang GJ, Wang H (2006) Toward mitigating the double ITCZ problem in NCAR 27. Ceppi P, Hwang Y-T, Frierson DMW, Hartmann DL (2012) Jet latitude biases in CMIP5 CCSM3. Geophys Res Lett 33(6):L06709. models linked to cloud shortwave forcing. Geophys Res Lett 39:L19708. 9. Hess PG, Battisti DS, Rasch PJ (1993) Maintenance of the Intertropical Convergence 28. Donohoe A, Battisti DS (2012) What determines meridional heat transport in climate Zones and the large-scale tropical circulation on a water-covered earth. J Atmos Sci models? J Clim 25(11):3832–3850. 50(5):691–713. 29. Trenberth KE, Fasullo JT (2010) Simulation of present day and 21st century energy 10. Luo JJ, Masson S, Roeckner E, Madec G, Yamagata T (2005) Reducing climatology bias budgets of the southern oceans. J Clim 23(2):440–454. in an ocean–atmosphere CGCM with improved coupling physics. J Clim 18(13): 30. Haynes JM, Jakob C, Rossow WB, Tselioudis G, Brown J (2011) Major characteristics of 2344–2360. Southern Ocean cloud regimes and their effects on the energy budget. J Clim 24(19): 11. Kang SM, Held IM, Frierson DMW, Zhao M (2008) The response of the ITCZ to ex- 5061–5080. tratropical thermal forcing: Idealized slab-ocean experiments with a GCM. J Clim 31. Bodas-Salcedo A, William KD, Field PR, Lock AP (2012) The surface downwelling solar 21(14):3521–3532. radiation surplus over Southern Ocean in the Met Office model: The role of mid- 12. Kang SM, Frierson DMW, Held IM (2009) The tropical response to extratropical – thermal forcing in an idealized GCM: The importance of radiative feedbacks and latitude cycle clouds. J Clim 25(21):7465 7486. convective parameterization. J Atmos Sci 66(9):2812–2827. 32. Adler RF, et al. (2003) The version 2 Global Precipitation Climatology Project (GPCP) – 13. Chiang JCH, Bitz CM (2005) Influence of high latitude ice cover on the marine monthly precipitation analysis (1979-present). J Hydrometeorol 4(6):1147 1167. Intertropical Convergence Zone. Clim Dyn 25(5):477–496. 33. Xie P, et al. (2003) GPCP pentad precipitation analyses: An experimental dataset – 14. Zhang R, Delworth TL (2005) Simulated tropical response to a substantial weakening based on gauge observations and satellite estimates. J Clim 16(13):2197 2214. ’ of the Atlantic thermohaline circulation. J Clim 18(12):1853–1860. 34. Wielicki BA, et al. (1996) Clouds and the earth s radiant energy system (CERES): An – 15. Broccoli AJ, Dahl KA, Stouffer RJ (2006) Response of the ITCZ to Northern Hemisphere earth observing system experiment. Bull Am Meteorol Soc 77(5):853 868. cooling. Geophys Res Lett 33(1):L01702. 35. Trenberth KE, Caron JM (2001) Estimates of meridional atmosphere and ocean heat 16. Rotstayn LD, Lohmann U (2002) Tropical rainfall trends and the indirect aerosol transports. J Clim 14(16):3433–3443. effect. J Clim 15(15):2103–2116. 36. Dee DP, et al. (2011) The ERA-Interim reanalysis: Configuration and performance of 17. Chiang JCH, Friedman AR (2012) Extratropical cooling, interhemispheric thermal the data assimilation system. Q J R Meteorol Soc 137(656):553–597. gradients, and tropical climate change. Annu Rev Earth Planet Sci 40(1):383–412. 37. Wu Y, Ting M, Seager R, Huang H-P, Cane MA (2010) Changes in storm tracks and 18. Frierson DMW, Hwang Y-T (2012) Extratropical influence on ITCZ shifts in slab ocean energy transports in a warmer climate simulated by the GFDL CM2.1 model. Clim Dyn simulations of global warming. J Clim 25(2):720–733. 37(1):53–72. 19. Schlosser CA, Houser PR (2007) Assessing a satellite-era perspective of the global 38. Zelinka MD, Hartmann DL (2012) Climate feedbacks and their implications for water cycle. J Clim 20(7):1316–1338. poleward energy flux changes in a warming climate. J Clim 25(2):608–624.

4940 | www.pnas.org/cgi/doi/10.1073/pnas.1213302110 Hwang and Frierson