• AchutaRao, K., , and K. Sperber, 2006: ENSO simulation in coupled ocean–atmosphere models: Are the current models better? Climate Dyn., 27, 115, doi:10.1007/s00382-006-0119-7.

    • Search Google Scholar
    • Export Citation
  • Arawaka, A., , and W. Schubert, 1974: Interaction of a cumulus cloud ensemble with the large-scale environment, Part I. J. Atmos. Sci., 31, 674701.

    • Search Google Scholar
    • Export Citation
  • Bellucci, A., , S. Gualdi, , and A. Navarra, 2010: The double ITCZ syndrome in coupled general circulation models: The role of large-scale vertical circulation regimes. J. Climate, 23, 11271145.

    • Search Google Scholar
    • Export Citation
  • Berg, W., , C. Kummerow, , and C. Morales, 2002: Differences between east and west Pacific rainfall systems. J. Climate, 15, 36593672.

  • Biasutti, M., , A. Sobel, , and Y. Kushir, 2006: AGCM precipitation biases in the tropical Atlantic. J. Climate, 19, 935958.

  • Bony, S., , J. Dufresne, , H. Le Treut, , J. Morcrette, , and C. Senior, 2004: On dynamic and thermodynamic components of cloud changes. Climate Dyn., 22, 7186.

    • Search Google Scholar
    • Export Citation
  • Cai, W., , A. Sullivan, , and T. Cowan, 2009: Rainfall teleconnections with Indo-Pacific variability in the WCRP CMIP3 models. J. Climate, 22, 50465071.

    • Search Google Scholar
    • Export Citation
  • Christensen, J., and Coauthors, 2007: Regional climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 847–940.

    • Search Google Scholar
    • Export Citation
  • Gregory, D., , and P. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability dependent closure. Mon. Wea. Rev., 118, 14831506.

    • Search Google Scholar
    • Export Citation
  • Huffman, G., , R. Adler, , M. Morrissey, , D. Bolvin, , S. Curtis, , R. Joyce, , B. McGavock, , and J. Susskind, 2001: Global precipitation at one-degree daily resolution from multisatellite observation. J. Hydrometeor., 2, 3650.

    • Search Google Scholar
    • Export Citation
  • Kanamitsu, M., , W. Ebisuzaki, , J. Woollen, , S.-K. Yang, , J. Hnilo, , M. Fiorino, , and G. L. Potter, 2002: NCEP-DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 16311643.

    • Search Google Scholar
    • Export Citation
  • Lin, B., , B. Wielicki, , P. Minnis, , L. Chambers, , K.-M. Xu, , Y. Hu, , and A. Fan, 2006: The effect of environmental conditions on tropical deep convective systems observed from the TRMM satellite. J. Climate, 19, 57455761.

    • Search Google Scholar
    • Export Citation
  • Lin, J.-L., 2007: The double-ITCZ problem in IPCC AR4 coupled GCMs: Ocean–atmosphere feedback analysis. J. Climate, 20, 44974525.

  • Martin, E., , and C. Schumacher, 2011: The Caribbean low-level jet and its relationship with precipitation in the IPCC AR4 models. J. Climate, 24, 59355950.

    • Search Google Scholar
    • Export Citation
  • Mestas-Nuñez, A., , D. Enfield, , and C. Zhang, 2007: Water vapor fluxes over the Intras-Americas Sea: Seasonal and interannual variability and associations with rainfall. J. Climate, 20, 19101922.

    • Search Google Scholar
    • Export Citation
  • Misra, V., , S. Chan, , R. Wu, , and E. Chassignet, 2009: Air–sea interaction over the Atlantic warm pool in the NCEP CFS. Geophys. Res. Lett., 36, L15702, doi:10.1029/2009GL038737.

    • Search Google Scholar
    • Export Citation
  • Moorthi, S., , and M. Suarez, 1992: Relaxed Arakawa-Schubert: A parameterization of moist convection for general circulation models. Mon. Wea. Rev., 120, 9781002.

    • Search Google Scholar
    • Export Citation
  • Neelin, J., , M. Munnich, , H. Su, , J. Meyerson, , and C. Holloway, 2006: Tropical drying trends in global warming models and observations. Proc. Natl. Acad. Sci. USA, 103, 61106115.

    • Search Google Scholar
    • Export Citation
  • Plant, R., 2010: A review of the theoretical basis for bulk mass flux convective parameterization. Atmos. Chem. Phys., 10, 35293544.

  • Rayner, N., , D. Parker, , E. Horton, , C. Folland, , L. Alexander, , D. Powell, , E. Kent, , and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, doi:10.1029/2002JD002670.

    • Search Google Scholar
    • Export Citation
  • Sobel, A., , C. Burleyson, , and S. Yuter, 2011: Rain on small tropical islands. J. Geophys. Res., 116, D08102, doi:10.1029/2010JD014695.

  • Turner, A., , and J. Slingo, 2009: Uncertainties in future projections of extreme precipitation in the Indian monsoon region. Atmos. Sci. Lett., 10, 152158.

    • Search Google Scholar
    • Export Citation
  • Waliser, D., , N. E. Graham, , and C. Gautier, 1993: Comparison of the highly reflective cloud and outgoing longwave radiation datasets for use in estimating tropical deep convection. J. Climate, 6, 331353.

    • Search Google Scholar
    • Export Citation
  • Wang, C., , and D. Enfield, 2001: The tropical western hemisphere warm pool. Geophys. Res. Lett., 28, 16351638.

  • Wang, C., , S.-K. Lee, , and D. Enfield, 2008: Climate response to anomalously large and small Atlantic warm pools during the summer. J. Climate, 21, 24372450.

    • Search Google Scholar
    • Export Citation
  • Zhang, C., 1993: Large-scale variability of atmospheric deep convection in relation to sea surface temperature in the tropics. J. Climate, 6, 18981913.

    • Search Google Scholar
    • Export Citation
  • Zhang, G., , and N. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model. Atmos.–Ocean, 33, 407446.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Annual mean SST (from HadISST), beginning at 26.5°C with intervals of 1°C. Black boxes indicate averaging regions for the western Pacific and Caribbean.

  • View in gallery

    Scatterplot of (a) Caribbean (10°–25°N, 55°–90°W) and (b) western Pacific (10°S–5°N, 130°–165°E) area-averaged sea surface temperature (°C) and precipitation (mm day−1). Observations (GPCP and HadISST) are labeled OM, with the CMIP multimodel ensemble mean marked as CM and the AMIP multimodel ensemble mean shown by AM. Shapes indicate model type (CMIP vs AMIP) and colors indicate convective parameterization type as shown. Horizontal and vertical lines indicate observed mean precipitation and SST values respectively. Note that different ranges for precipitation are used in (a) and (b).

  • View in gallery

    Regime sorting analysis of Caribbean area-averaged (10°–25°N, 55°–90°W) precipitation by SST for (left) AMIP and (right) CMIP models and observations (solid black line). (a),(b) PDF of SST. (c),(d) Precipitation composited by SST. (e),(f) Composited precipitation weighted by the PDF of SST. Convective parameterization group multimodel means indicated by colored lines as shown in (a). Multimodel ensemble mean shown by dashed black line. Gray shading indicates plus/minus one standard deviation from observations.

  • View in gallery

    As in Fig. 3, but for the western Pacific (10°S–5°N, 130–165°E).

  • View in gallery

    As in Fig. 3, but replacing regime sorting by SST with ω500 (hPa day−1).

  • View in gallery

    As in Fig. 4, but replacing regime sorting by SST with ω500 (hPa day−1).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 72 72 5
PDF Downloads 69 69 5

The Relationship between Tropical Warm Pool Precipitation, Sea Surface Temperature, and Large-Scale Vertical Motion in IPCC AR4 Models

View More View Less
  • 1 Department of Atmospheric Sciences, Texas A&M University, College Station, Texas
© Get Permissions
Full access

Abstract

A regime sorting analysis is used to identify Caribbean and western Pacific precipitation, sea surface temperature, and large-scale vertical circulation relationships and biases within coupled and uncoupled Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC AR4) general circulation models. This analysis shows that an oversensitivity of precipitation to both SST and vertical circulation (as represented by ω500) is inherent in the atmospheric models in both regions, with models using a spectral-type convective parameterization performing best in the Caribbean, but less separation between convective parameterization groups is seen in the western Pacific. The error in magnitude of precipitation for a given SST and vertical circulation causes uncoupled models to overestimate Caribbean and western Pacific mean precipitation. In coupled models, however, errors in the frequency of occurrence of SSTs (the distribution is cold biased in both regions) and deep convective vertical circulations (reduced frequency) lead to an underestimation of Caribbean and western Pacific mean precipitation. In the western Pacific, increased frequency of subsidence regimes in coupled models leads to an overestimation of precipitation at ω500 values above 0 hPa day−1. The varied ability of convective parameterization groups in the two warm pool regions suggests that deficiencies in parameterization groups differ between the two regions, with improvements needed particularly in the deep convective regime in the Caribbean and subsidence regimes in the western Pacific.

Current affiliation: Department of Atmospheric and Environmental Sciences, University at Albany, State University of New York, Albany, New York.

Corresponding author address: Elinor R. Martin, Department of Atmospheric and Environmental Sciences, University at Albany, State University of New York, Albany, NY 12222. E-mail: emartin@atmos.albany.edu

Abstract

A regime sorting analysis is used to identify Caribbean and western Pacific precipitation, sea surface temperature, and large-scale vertical circulation relationships and biases within coupled and uncoupled Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC AR4) general circulation models. This analysis shows that an oversensitivity of precipitation to both SST and vertical circulation (as represented by ω500) is inherent in the atmospheric models in both regions, with models using a spectral-type convective parameterization performing best in the Caribbean, but less separation between convective parameterization groups is seen in the western Pacific. The error in magnitude of precipitation for a given SST and vertical circulation causes uncoupled models to overestimate Caribbean and western Pacific mean precipitation. In coupled models, however, errors in the frequency of occurrence of SSTs (the distribution is cold biased in both regions) and deep convective vertical circulations (reduced frequency) lead to an underestimation of Caribbean and western Pacific mean precipitation. In the western Pacific, increased frequency of subsidence regimes in coupled models leads to an overestimation of precipitation at ω500 values above 0 hPa day−1. The varied ability of convective parameterization groups in the two warm pool regions suggests that deficiencies in parameterization groups differ between the two regions, with improvements needed particularly in the deep convective regime in the Caribbean and subsidence regimes in the western Pacific.

Current affiliation: Department of Atmospheric and Environmental Sciences, University at Albany, State University of New York, Albany, New York.

Corresponding author address: Elinor R. Martin, Department of Atmospheric and Environmental Sciences, University at Albany, State University of New York, Albany, NY 12222. E-mail: emartin@atmos.albany.edu

1. Introduction

The correct simulation of Caribbean climate—in particular, its moisture budget and sea surface temperature (SST)—is important not only for the Caribbean but also for weather and climate in the United States. The amount of moisture transport to the central United States by the Great Plains low-level jet is strongly influenced by the Caribbean moisture budget (Mestas-Nuñez et al. 2007; Wang et al. 2008) and hence the correct simulation of precipitation by general circulation models (GCMs) in the Caribbean is critical for accurate simulations of regional climate.

Most models from the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) predict a decrease in precipitation across the Caribbean region while underestimating current precipitation amounts (Christensen et al. 2007; Neelin et al. 2006). The Caribbean Sea is part of the Atlantic warm pool (AWP) with SSTs exceeding 28.5°C. The AWP has maximum extent in boreal summer, when it is positively correlated with precipitation not only over the Caribbean, but also in Central America, southeastern Pacific, and the United States (Wang and Enfield 2001; Wang et al. 2008). However, Misra et al. (2009) show that in eight of the IPCC AR4 coupled simulations for summer, the SSTs have a large cold bias in the AWP. Although coupled models underestimate Caribbean precipitation, it has been shown the uncoupled models forced with observed SSTs overestimate Caribbean precipitation (Biasutti et al. 2006; Martin and Schumacher 2011).

In the western Pacific warm pool (WPWP), SSTs are higher and have a greater spatial extent than the AWP, as seen in Fig. 1. The WPWP has not only higher SSTs, but also high relative humidity in the lower free troposphere and weaker surface winds (Sobel et al. 2011). The warm SSTs of the WPWP are associated with large amounts of water vapor, column ice, and a larger areal coverage of rainfall events. Clouds are deeper and higher in the WPWP than in the eastern Pacific and Caribbean and much of the rainfall is convective (Berg et al. 2002). Both the western Pacific and Caribbean contain a large number of islands and show enhanced rain frequency and total rainfall over larger islands compared to the surrounding ocean (Sobel et al. 2011).

Fig. 1.
Fig. 1.

Annual mean SST (from HadISST), beginning at 26.5°C with intervals of 1°C. Black boxes indicate averaging regions for the western Pacific and Caribbean.

Citation: Journal of the Atmospheric Sciences 69, 1; 10.1175/JAS-D-11-0104.1

Modeling results similar to those for the Caribbean, although less extreme, have been shown in the western Pacific warm pool (Lin 2007). Although most models get the general large-scale structure in the western Pacific, the well-known double ITCZ structure over the tropical Pacific and associated westward extension of the eastern Pacific cold tongue is a major problem in the IPCC AR4 simulations (Lin 2007). The WPWP is also more directly influenced by ENSO than the Caribbean, and hence the poor simulation of ENSO in the IPCC AR4 models (AchutaRao and Sperber 2006) may impact the WPWP more than the AWP. Despite the similarities between the regions, the IPCC AR4 report predicts precipitation to increase in the western Pacific (Christensen et al. 2007). Further differences and similarities between the regions in observations and IPCC AR4 models will be investigated in this study.

In addition to SSTs, the connection between precipitation and the large-scale vertical circulation (as represented by ω500) is also of importance. Correctly representing large-scale vertical circulations is essential for correctly reproducing heat and moisture transport and thus impacts on stability and precipitation. The relationship between precipitation and both SST and ω500 will be investigated using the regime sorting (or compositing) technique of Bony et al. (2004) to determine the source of the precipitation errors.

2. Data and models

Monthly mean precipitation, SST, and ω500 from one run of 20 coupled ocean–atmosphere GCMs [the Coupled Model Intercomparison Project (CMIP), version 3] and 11 atmospheric-only GCMs [the Atmosphere Model Intercomparison Project (AMIP)] are used (Table 1; all model names are given in the appendix). For the CMIP simulations, the last 30 yr of the climate of the twentieth-century simulations are used. For the AMIP simulation, the years 1979–2000 are used [except for the Geophysical Fluid Dynamics Laboratory (GFDL) Climate Model version 2.1 (GFDL 2.1) and the National Center for Atmospheric Research Parallel Climate Model (NCAR PCM), which were available for 1980–99 and 1979–99, respectively]. The models are a subset of those used as part of the IPCC AR4 and are available through the Program for Climate Model Diagnostics and Intercomparison (PCMDI). (Further model details can be found online at http://www-pcmdi.llnl.gov.)

Table 1.

Convective parameterizations of models used in this study. Model names in italics indicate that both CMIP and AMIP data were used. An asterisk indicates models where ω was not available. ZM = Zhang and McFarlane. All model names are given in the appendix.

Table 1.

To determine whether the type of convective parameterization scheme significantly influenced relationships among precipitation, SST, and the large-scale vertical circulation, models were grouped by three different convective parameterization types as shown in Table 1. Bulk parameterizations (e.g., Gregory and Rowntree 1990) represent an ensemble of convective clouds with a single cloud type. Spectral parameterizations (e.g., Moorthi and Suarez 1992) are similar to Arawaka and Schubert (1974), which parameterizes deep convection with ensembles of cloud types. A number of models use the Zhang and McFarlane (1995) scheme (herein ZM), which applies techniques from both the spectral and bulk methods.

For comparison with model output, precipitation data from the Global Precipitation Climatology Project (GPCP; Huffman et al. 2001) at 2.5° resolution along with Hadley Centre Sea Ice and Sea Surface Temperature dataset 1 (HadISST1) at 1° resolution (Rayner et al. 2003) are used. The validation of ω uses National Centers for Environmental Prediction (NCEP)–Department of Energy (DOE) Reanalysis II data (Kanamitsu et al. 2002) at 2.5° resolution. All observations, reanalysis, and model output are regridded to the GPCP 2.5° horizontal resolution grid.

Regions with equal numbers of grid points were chosen to represent the Caribbean (AWP) and western Pacific (WPWP) and are shown by the boxes in Fig. 1. The Caribbean is defined as 10°–25°N, 55°–90°W and the western Pacific as 10°S–5°N, 130°–165°E. These regions were chosen as they are both influenced by the surrounding land and/or islands.

3. Results

a. Sea surface temperature climatology

The climatology of annual mean Caribbean area-averaged precipitation and SST is shown in Fig. 2a for observations (OM) and model output. The strong cold bias of the CMIP models is evident, exceeding 1.5°C in nine of the models. This bias occurs throughout the annual cycle (not shown). The CMIP model results do not show a simple relationship between SST and precipitation, as models with the largest SST bias (over 2°C) do not necessarily produce the lowest mean precipitation amounts. The CMIP multimodel ensemble mean (CM) SST is biased cold (26.1° vs 27.5°C) and the multimodel ensemble mean precipitation is biased dry (2.38 vs 2.71 mm day−1). This oversensitivity of precipitation to SST is demonstrated in the AMIP model output also shown in Fig. 2a. In this case, and much of the later discussion, oversensitivity is regarded as too much precipitation for a given value of SST, unless otherwise specified. Despite forcing with observed SST, the range of precipitation produced by the AMIP ensemble is larger than the CMIP ensemble, with the AMIP multimodel ensemble mean (AM) 0.86 mm day−1 larger than observed.

Fig. 2.
Fig. 2.

Scatterplot of (a) Caribbean (10°–25°N, 55°–90°W) and (b) western Pacific (10°S–5°N, 130°–165°E) area-averaged sea surface temperature (°C) and precipitation (mm day−1). Observations (GPCP and HadISST) are labeled OM, with the CMIP multimodel ensemble mean marked as CM and the AMIP multimodel ensemble mean shown by AM. Shapes indicate model type (CMIP vs AMIP) and colors indicate convective parameterization type as shown. Horizontal and vertical lines indicate observed mean precipitation and SST values respectively. Note that different ranges for precipitation are used in (a) and (b).

Citation: Journal of the Atmospheric Sciences 69, 1; 10.1175/JAS-D-11-0104.1

The impact of the convective parameterization type is more obvious on precipitation than SST. Models using bulk parameterizations tend to produce higher than average precipitation and models using spectral parameterizations produce lower precipitation amounts, for both CMIP and AMIP models. Models using the ZM scheme tend to overestimate precipitation in AMIP models and underestimate it in CMIP models. When analyzing by mean SST in which convective parameterization is not the main factor in determining value, it seen that the four models with the best Caribbean averaged SST values (CSIRO 3.5, INGV, MPI, and UKMO HadCM) all use bulk parameterizations. However, the three models with the lowest Caribbean averaged SST values (CSIRO 3.0, GISS-ER, and CNRM) also use the bulk parameterizations, suggesting no simple relationship between convective parameterization type and SST in the region, as may be expected due to the complex interactions and feedbacks between precipitation and SST.

The climatology for the western Pacific is shown in Fig. 2b and shows similarities with the Caribbean although it is both warmer and wetter. The majority of the CMIP models produce SSTs that are biased cold, although unlike the Caribbean most of the models are within 1°C of the observed mean. The CMIP models also produce dry conditions in the western Pacific and wetter than observed conditions in the AMIP simulations. In addition to the weaker cold bias in the western Pacific, the main difference between the climatology of the regions is the impact of the convective parameterizations. Again, the bulk parameterizations produce both the warmest and coldest annual mean SSTs, but the impact of parameterization on precipitation is reversed from the Caribbean. The spectral parameterization group produces the most rainfall, with the ZM and bulk groups producing less. This result suggests that processes controlling deep convection differ in the two regions and hence the performance of the parameterizations is not consistent in the two warm pool regions.

b. Sea surface temperature regime sorting

The role of convective parameterizations and the relationship between precipitation and SST is further analyzed by exploring the annual mean precipitation–SST space by regime sorting. This involves three steps: 1) calculating the probability distribution function (PDF) of SST, 2) compositing precipitation by SST, and 3) weighting the composite by the PDF to calculate the regime sorted (or weighted) precipitation. This allows the error in precipitation to be associated with errors affecting either the frequency of SST or the magnitude of precipitation associated with a given SST (Bony et al. 2004; Bellucci et al. 2010). Regime sorting is performed for all grid points and all months before averaging to create Caribbean and western Pacific area-averaged values.

The three stages of the regime sorting analysis for the Caribbean and western Pacific are shown in Figs. 3 and 4 for AMIP and CMIP simulations, in addition to the multimodel ensemble mean and observations. Multimodel mean values for each convective parameterization group (bulk, ZM, and spectral) are shown in the regime sorting analysis. The standard deviation of the precipitation for each SST is calculated and plus/minus one standard deviation is indicated by gray shading in the appropriate plots. The PDFs of AMIP SSTs (Figs. 3a and 4a) have slight variations from the HadISST data; however, these do not affect the results.

Fig. 3.
Fig. 3.

Regime sorting analysis of Caribbean area-averaged (10°–25°N, 55°–90°W) precipitation by SST for (left) AMIP and (right) CMIP models and observations (solid black line). (a),(b) PDF of SST. (c),(d) Precipitation composited by SST. (e),(f) Composited precipitation weighted by the PDF of SST. Convective parameterization group multimodel means indicated by colored lines as shown in (a). Multimodel ensemble mean shown by dashed black line. Gray shading indicates plus/minus one standard deviation from observations.

Citation: Journal of the Atmospheric Sciences 69, 1; 10.1175/JAS-D-11-0104.1

Fig. 4.
Fig. 4.

As in Fig. 3, but for the western Pacific (10°S–5°N, 130–165°E).

Citation: Journal of the Atmospheric Sciences 69, 1; 10.1175/JAS-D-11-0104.1

The observed composite of precipitation at a given SST in the AMIP simulations is shown in Figs. 3c and 4c by the solid black line. As expected, convection increases with SST but this increase is not constant across the range of SSTs observed in the Caribbean (Fig. 3c). Precipitation is relatively constant with SST below 27°C but increases rapidly above this value, similar to the results of Waliser et al. (1993), Zhang (1993), and Lin et al. (2006). This observation is also seen in the western Pacific (Fig. 4c) although the increase begins at 26°C and increases more rapidly than in the Caribbean. Above 30°–31°C in each case, the precipitation drops quickly to zero as SSTs rarely occur at such high values. In the CMIP models (Figs. 3d and 4d) this SST dropoff value varies among the models and hence the multimodel ensemble mean and even the convective parameterization group means are not beneficial above 29°C in the Caribbean and 31°C in the western Pacific.

Figures 3c and 4c shows the overestimation of precipitation at a given SST by the atmospheric component of the model. The overestimation in the Caribbean is smallest below 26°C but rapidly increases above this value, with maximum overestimation greater than 2 mm day−1 between 28° and 30°C (Fig. 3c). The multimodel mean shows an increase of precipitation with SST that is too large in the models above 26°C. The regime sorted weighted precipitation is shown in Fig. 3e, with the greatest contribution to the observed precipitation at 28°–29°C. This is also the range where the AMIP models have the greatest oversensitivity to SST.

In the western Pacific, where the SSTs are warmer and the SST distribution is narrower (Fig. 4a), similar results for the AMIP precipitation composite (Fig. 4c) are seen. However, as the extremely high SSTs do not occur frequently and the precipitation composite is not as consistently high biased as the Caribbean, the weighted regime sorted precipitation in Fig. 4e shows excellent results, particularly in comparison to the Caribbean. These results show that the majority of the overestimation of precipitation by the AMIP models in the western Pacific is at SSTs above 30°C, but this overestimate is small in comparison to the Caribbean.

In the Caribbean, the AMIP models that utilize spectral convective parameterizations produce the regime sorted precipitation that is closest to observations with the bulk and ZM schemes most oversensitive to SST (Figs. 3c,e). However, in the western Pacific, all schemes perform equally, although the bulk scheme is most oversensitive to precipitation at high SSTs (Figs. 4c,e). Bulk parameterizations use adiabatic ascent from the lowest model level and hence SST is a much stronger constraint than the spectral schemes that can initiate plumes from multiple levels (Turner and Slingo 2009). By initiating an ensemble of plumes from multiple levels (including the lowest level) the spectral parameterizations produce a better precipitation–SST relationship in the AMIP models in the Caribbean, but not in the western Pacific, suggesting differences between the way deep convection is structured and initialized between the two regions. The bulk and spectral parameterizations not only differ in the levels of initialization, but also in the entrainment/detrainment profiles used which may also be influencing their simulation ability (Plant 2010).

The results of the regime sorting analysis using the CMIP models are more complex (particularly when analyzing the impact of the convection parameterizations). The Caribbean SST PDF in Fig. 3b shows the cold bias of the coupled models as seen in Fig. 2a. The bulk group mean appears to be shifted to warmer temperatures because the four models with the best SST PDF are all in this group (as discussed previously). The multimodel ensemble mean maximum is over 2°C colder than the observed maximum. The precipitation composite in Fig. 3d shows similar results to the AMIP composite (Fig. 3c), although given the infrequent occurrence of high SSTs (above 29°C) in the models, the multimodel ensemble mean and parameterization means appear lower than the observations. This is to be expected as the amount of precipitation at a given SST is controlled by the atmospheric component of the coupled model system.

In the western Pacific, the CMIP models are cold biased (Fig. 4b) as expected but still produce a much narrower distribution than in the Caribbean. The bulk model mean appears broader, not because the individual models produce a broader distribution but rather because two models are centered at SSTs much higher (30°–31°C) than the observed mean and two are centered lower (27°C). It is noted that the models have a longer tail at high SSTs than the observations, which leads to a much slower dropoff of precipitation at high SSTs (Fig. 4d). The precipitation composite is similar to that of both the Caribbean and the western Pacific AMIP models, considering the shifted SST distribution. In both the Caribbean and western Pacific precipitation continues to occur at larger SSTs than observations, despite the rare occurrence of these SSTs in the CMIP simulations (Figs. 3d and 4d). Although the error in the regime sorted precipitation is small at these high SSTs in the Caribbean, it is a larger problem in the western Pacific, suggesting problematic relationships between precipitation and CAPE in this region.

The regime sorted precipitation in both the Caribbean (Fig. 3f) and western Pacific (Fig. 4f) shows the underestimation of the precipitation by the CMIP multimodel ensemble mean and a shift of the distribution to lower SSTs, with the largest shift occurring in the Caribbean. The overestimation of precipitation at a given SST that is an inherent part of the atmospheric model in both locations is overcompensated by the cold SST bias in the CMIP models, leading to an underestimation of the precipitation, particularly at SSTs greater than 27.5°C in the Caribbean and between 28° and 30°C in the western Pacific. The individual models that produce the best distribution in the Caribbean and western Pacific use bulk parameterizations because these models have the best SST distribution; however, those that are most different from the observations also use bulk parameterizations (as seen in Fig. 2), illustrating the difficulty of extracting such relationships in coupled model simulations.

c. Vertical large-scale circulation

In addition to precipitation regime sorting by SST, the same methodology was applied to regime sorting by ω500 (Fig. 5 for the Caribbean and Fig. 6 for the western Pacific). Figure 5c shows that the AMIP models produce too much rain for a given vertical circulation compared to observations/reanalysis in the Caribbean. This is less evident in the CMIP models (Fig. 5d) because of the severe underestimation of the deep convective (negative) values of ω500 (Fig. 5b), particularly those less than −50 hPa day−1. Individual models that simulate these strong upward motion regimes overestimate precipitation at these values (not shown), but the multimodel ensemble mean is shifted to lower values of precipitation by averaging these models with those that do not produce these strong upward motion events and hence have composite precipitation values close to zero.

Fig. 5.
Fig. 5.

As in Fig. 3, but replacing regime sorting by SST with ω500 (hPa day−1).

Citation: Journal of the Atmospheric Sciences 69, 1; 10.1175/JAS-D-11-0104.1

Fig. 6.
Fig. 6.

As in Fig. 4, but replacing regime sorting by SST with ω500 (hPa day−1).

Citation: Journal of the Atmospheric Sciences 69, 1; 10.1175/JAS-D-11-0104.1

The overestimation of rainfall for a given convective event is the main contributor to the AMIP models overestimating Caribbean precipitation (Fig. 5e) as the ω500 PDF (Fig. 5a) is fairly well reproduced by the AMIP multimodel ensemble mean. The spectral parameterization model mean produces the least accurate PDF with overestimation of shallow convective events and underestimation of deep convective events whereas the bulk and ZM groups overestimate the frequency of deep convective events. Much of this overestimation of precipitation is in deep convective regimes between −10 and −50 hPa day−1. As for the regime sorting by SST, AMIP models using the spectral parameterizations outperform other parameterization groups in the Caribbean. However, this good performance of the spectral models is in part because the errors in the ω500 PDF (Fig. 5a) are compensated by the errors in the precipitation composite (Fig. 5c).

In the western Pacific, the AMIP models show similar results to the Caribbean, with the ω500 PDF (Fig. 6a) being well represented by the multimodel ensemble mean. However, the precipitation composite (Fig. 6c) shows an even larger overestimation and rate of change of precipitation with ω500, far surpassing one standard deviation of the observations below −75 hPa day−1. It is also extremely consistent between all three parameterization groups. The result of this overestimation in deep convective regimes is to produce a regime sorted weighted precipitation composite (Fig. 6e) that shows the overestimate of precipitation in the deep convective range (below −50 hPa day−1), as seen in the Caribbean.

The CMIP results are, again, more complex in both locations. While too much rainfall for a given vertical circulation is inherent in the atmospheric model in the Caribbean (Fig. 5d) and western Pacific (Fig. 6d), the CMIP models poorly represent the PDF of vertical circulation in both locations. In the Caribbean, the ω500 PDF (Fig. 5b) shows the multimodel ensemble mean, the parameterization group means, and the majority of individual models (not shown) below the observations in the deep convective regime (−10 hPa day−1 and below). Also apparent is the overestimation associated with shallow convection (0–20 hPa day−1) regimes. These two features essentially produce a narrower PDF in the model output in comparison to the observations in the Caribbean. In the western Pacific, a different problem is apparent in the ω500 PDF (Fig. 6b). While the width of the distribution is similar in the model output and the observations, the multimodel mean, bulk, and ZM parameterization group means are skewed toward downward motion, with maxima close to 10 hPa day−1 compared with −40 hPa day−1 in the observations. This suggests that the correct simulation of large-scale circulation is a major problem in many of the CMIP models over the western Pacific. This may be due to the poor simulation of ENSO and the eastern Pacific cold tongue, which is known to shift the rising branch of the Walker circulation westward (Cai et al. 2009).

Despite the atmospheric models producing too much rain for a given ω500 in both locations (noting the averaging effects in the Caribbean), the differing ω500 PDFs produce different regime sorted weighted precipitation composites. In the Caribbean, the underestimation of the frequency of all upward motion events is large enough that the underestimation of precipitation in the CMIP models is mainly during deep convective regimes (Fig. 5f). In the western Pacific, the reduced frequency of upward motion between 0 and −70 hPa day−1 leads to underestimation of precipitation in this range, which dominates the two overestimating regions above and below this range. The overestimate of downward motion in the western Pacific produces an overestimate of rainfall for positive values of ω500. At the largest upward motion values, below −80 hPa day−1, the oversensitivity of precipitation to ω500 dominates and produces too much rain at these values. Bellucci et al. (2010), who performed a similar analysis in the southern tropical Pacific (20°S –0°, 100°–150°W), show results that vary from the western Pacific and the Caribbean. Both AMIP and CMIP models overestimate precipitation in this region and this overestimation is seen in across the entire ω500 spectrum.

Once again, the impacts of convective parameterization are considerably more difficult to discern for the CMIP models and are not constant between the two regions. The performance of the convective parameterization groups for the regime sorted precipitation in the CMIP models (Figs. 5f and 6f) is dependent on their performance in representing the ω500 PDF.

4. Conclusions

The relationship between precipitation and SST and precipitation and the large-scale vertical circulation in observations and IPCC AR4 models was investigated using a regime sorting analysis for two tropical warm pool regions: the Caribbean and the western Pacific. The oversensitivity of precipitation to SST was the dominant factor in AMIP models overestimating Caribbean rainfall, with models using spectral-type convective parameterizations performing better than either bulk or ZM type parameterizations. While this oversensitivity of precipitation to SST was still present in CMIP models, the known severe underestimation of SST in the Caribbean leads to a large underestimation of precipitation (particularly at SSTs above 28°C) in the CMIP models.

The same oversensitivity of precipitation to SST was observed in the western Pacific, but the narrowness of the SST distribution led to only a small overestimation of precipitation in the AMIP models at high SSTs. The CMIP models in the western Pacific also showed a cold bias, although not as strong as in the Caribbean, leading to an underestimation of rainfall at SSTs between 28° and 30°C. Although the relationship between precipitation and SST was similar in the two regions, in both models and observations the shifting of the SST distribution to colder temperatures is the determining factor in the CMIP simulations. By stratifying the results by type of convective parameterizations, some insights can be gained regarding differences between the deep convective mechanisms in the two regions. The spectral group, which can initiate convection at a range of levels and uses different entrainment rates for different cloud types, performed best in the Caribbean. Although no group could be identified as performing best in the western Pacific, results from the bulk group, which initiates convective plumes from the surface, often performed worse than the two other groups. This stratification by convective parameterization suggests that a bulk approach may not be appropriate to describe warm pool convection and the spectral approach is more successful in the Caribbean than the western Pacific.

Not only was precipitation too sensitive to SST, it was also too sensitive to vertical motion in both regions, which has implications for heating and moisture profiles. The largest disparities were evident in the deep convective regimes (less than −10 hPa day−1) and were inherent to the atmospheric models. This oversensitivity is consistent with the results of Bellucci et al. (2010), who performed a similar analysis in the southern tropical Pacific. The AMIP models were shown to overestimate precipitation due to the incorrect magnitude of rainfall for a given convective event (oversensitivity). Contrary to this result, CMIP models underestimated precipitation due to the reduced frequency of deep convective events in the Caribbean. The results in the western Pacific were similar, but large problems with the ω500 distribution in CMIP models (shifted into a downward motion regime) led to an underestimation of rainfall between 0 and −75 hPa day−1. This regime sorting analysis produces similar results when performed by season, although the underestimation of deep convective regimes is further exaggerated in the Caribbean in September–November.

It should be noted that there are a variety of other factors that contribute to rainfall and deep convection in the tropics, not only SST and vertical circulation as discussed here. Precipitation can also be controlled by SST gradients, surface heat and moisture fluxes, and moisture convergence, making the evaluation of precipitation and its errors difficult. Stratification by convective parameterization showed some variation between different parameterizations but with such a small sample size (4–10) in each group, significance of the separation between the groups is likely minimal. This study does not claim that the convective parameterization is the only or leading cause of differences in the precipitation–SST–vertical circulation relationship and other parameterizations are likely to strongly influence convection, particularly the radiation scheme and the type of moisture trigger, which will be the subject of additional studies.

These results suggest that by coupling the models and including atmosphere–ocean feedback, errors occur in the frequency of occurrence of both SST and the large-scale vertical circulation in both warm pool regions, with a reduction in deep convective events and high SSTs, and an increase in low SSTs. The results of the AMIP simulations show that simply improving the SST climatology in future CMIP models will not be enough. Improvements in convective parameterizations are also necessary as the errors in precipitation magnitude associated with a given SST or vertical circulation are large.

Acknowledgments

This research is supported by NASA Grant NNX10AG89G. The authors appreciate comments from editor Ming Cai and two anonymous reviewers for improving the quality of this manuscript. The authors acknowledge the modeling groups, the PCMDI, and the World Climate Research Programme’s Working Group on Coupled Modelling (WGCM) for their roles in making available the WCRP CMIP3 multimodel dataset. Support of this dataset is provided by the Office of Science, U.S. Department of Energy. The monthly GPCP combined precipitation data were developed and computed by the NASA/Goddard Space Flight Center’s Laboratory for Atmospheres as a contribution to the GEWEX Global Precipitation Climatology Project. NCEP Reanalysis 2 data were provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA (from their Web site at http://www.esrl.noaa.gov/psd/).

Appendix

Model Names

CCCMA (T63) Canadian Centre for Climate Modelling and Analysis

CNRM Centre National de Recherches Météorologiques

CSIRO 3.0/3.5 Commonwealth Scientific and Industrial Research Organisation Mark version 3.0/3.5

GFDL 2.0/2.1 Geophysical Fluid Dynamics Laboratory Climate Model version 2.0/2.1

GISS EH/GISS ER Goddard Institute for Space Studies Model E-H/E-R

IAP FGOALS Institute of Atmospheric Physics Flexible Global Ocean–Atmosphere–Land System Model

INGV Istituto Nazionale di Geofisica e Vulcanologia

MPI Max Planck Institute

MIUB Meteorological Institute of the University of Bonn

MIROC3.2 (medres)/(hires) Model for Interdisciplinary Research on Climate 3.2, medium-resolution/high-resolution version

MRI Meteorological Research Institute

NCAR CCSM National Center for Atmospheric Research Community Climate System Model

NCAR PCM NCAR Parallel Climate Model

UKMO HadCM Climate configuration of the Met Office Unified Model

UKMO HadGEM Met Office Hadley Centre Global Environmental Model

REFERENCES

  • AchutaRao, K., , and K. Sperber, 2006: ENSO simulation in coupled ocean–atmosphere models: Are the current models better? Climate Dyn., 27, 115, doi:10.1007/s00382-006-0119-7.

    • Search Google Scholar
    • Export Citation
  • Arawaka, A., , and W. Schubert, 1974: Interaction of a cumulus cloud ensemble with the large-scale environment, Part I. J. Atmos. Sci., 31, 674701.

    • Search Google Scholar
    • Export Citation
  • Bellucci, A., , S. Gualdi, , and A. Navarra, 2010: The double ITCZ syndrome in coupled general circulation models: The role of large-scale vertical circulation regimes. J. Climate, 23, 11271145.

    • Search Google Scholar
    • Export Citation
  • Berg, W., , C. Kummerow, , and C. Morales, 2002: Differences between east and west Pacific rainfall systems. J. Climate, 15, 36593672.

  • Biasutti, M., , A. Sobel, , and Y. Kushir, 2006: AGCM precipitation biases in the tropical Atlantic. J. Climate, 19, 935958.

  • Bony, S., , J. Dufresne, , H. Le Treut, , J. Morcrette, , and C. Senior, 2004: On dynamic and thermodynamic components of cloud changes. Climate Dyn., 22, 7186.

    • Search Google Scholar
    • Export Citation
  • Cai, W., , A. Sullivan, , and T. Cowan, 2009: Rainfall teleconnections with Indo-Pacific variability in the WCRP CMIP3 models. J. Climate, 22, 50465071.

    • Search Google Scholar
    • Export Citation
  • Christensen, J., and Coauthors, 2007: Regional climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 847–940.

    • Search Google Scholar
    • Export Citation
  • Gregory, D., , and P. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability dependent closure. Mon. Wea. Rev., 118, 14831506.

    • Search Google Scholar
    • Export Citation
  • Huffman, G., , R. Adler, , M. Morrissey, , D. Bolvin, , S. Curtis, , R. Joyce, , B. McGavock, , and J. Susskind, 2001: Global precipitation at one-degree daily resolution from multisatellite observation. J. Hydrometeor., 2, 3650.

    • Search Google Scholar
    • Export Citation
  • Kanamitsu, M., , W. Ebisuzaki, , J. Woollen, , S.-K. Yang, , J. Hnilo, , M. Fiorino, , and G. L. Potter, 2002: NCEP-DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 16311643.

    • Search Google Scholar
    • Export Citation
  • Lin, B., , B. Wielicki, , P. Minnis, , L. Chambers, , K.-M. Xu, , Y. Hu, , and A. Fan, 2006: The effect of environmental conditions on tropical deep convective systems observed from the TRMM satellite. J. Climate, 19, 57455761.

    • Search Google Scholar
    • Export Citation
  • Lin, J.-L., 2007: The double-ITCZ problem in IPCC AR4 coupled GCMs: Ocean–atmosphere feedback analysis. J. Climate, 20, 44974525.

  • Martin, E., , and C. Schumacher, 2011: The Caribbean low-level jet and its relationship with precipitation in the IPCC AR4 models. J. Climate, 24, 59355950.

    • Search Google Scholar
    • Export Citation
  • Mestas-Nuñez, A., , D. Enfield, , and C. Zhang, 2007: Water vapor fluxes over the Intras-Americas Sea: Seasonal and interannual variability and associations with rainfall. J. Climate, 20, 19101922.

    • Search Google Scholar
    • Export Citation
  • Misra, V., , S. Chan, , R. Wu, , and E. Chassignet, 2009: Air–sea interaction over the Atlantic warm pool in the NCEP CFS. Geophys. Res. Lett., 36, L15702, doi:10.1029/2009GL038737.

    • Search Google Scholar
    • Export Citation
  • Moorthi, S., , and M. Suarez, 1992: Relaxed Arakawa-Schubert: A parameterization of moist convection for general circulation models. Mon. Wea. Rev., 120, 9781002.

    • Search Google Scholar
    • Export Citation
  • Neelin, J., , M. Munnich, , H. Su, , J. Meyerson, , and C. Holloway, 2006: Tropical drying trends in global warming models and observations. Proc. Natl. Acad. Sci. USA, 103, 61106115.

    • Search Google Scholar
    • Export Citation
  • Plant, R., 2010: A review of the theoretical basis for bulk mass flux convective parameterization. Atmos. Chem. Phys., 10, 35293544.

  • Rayner, N., , D. Parker, , E. Horton, , C. Folland, , L. Alexander, , D. Powell, , E. Kent, , and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, doi:10.1029/2002JD002670.

    • Search Google Scholar
    • Export Citation
  • Sobel, A., , C. Burleyson, , and S. Yuter, 2011: Rain on small tropical islands. J. Geophys. Res., 116, D08102, doi:10.1029/2010JD014695.

  • Turner, A., , and J. Slingo, 2009: Uncertainties in future projections of extreme precipitation in the Indian monsoon region. Atmos. Sci. Lett., 10, 152158.

    • Search Google Scholar
    • Export Citation
  • Waliser, D., , N. E. Graham, , and C. Gautier, 1993: Comparison of the highly reflective cloud and outgoing longwave radiation datasets for use in estimating tropical deep convection. J. Climate, 6, 331353.

    • Search Google Scholar
    • Export Citation
  • Wang, C., , and D. Enfield, 2001: The tropical western hemisphere warm pool. Geophys. Res. Lett., 28, 16351638.

  • Wang, C., , S.-K. Lee, , and D. Enfield, 2008: Climate response to anomalously large and small Atlantic warm pools during the summer. J. Climate, 21, 24372450.

    • Search Google Scholar
    • Export Citation
  • Zhang, C., 1993: Large-scale variability of atmospheric deep convection in relation to sea surface temperature in the tropics. J. Climate, 6, 18981913.

    • Search Google Scholar
    • Export Citation
  • Zhang, G., , and N. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model. Atmos.–Ocean, 33, 407446.

    • Search Google Scholar
    • Export Citation
Save