Differences Of High Clouds In Reanalyses In The Tropical Upper Troposphere

Clouds play an important role in tropical weather and climate and have substantial impacts on the tropical radiation budget and the atmospheric water cycle.Clouds also have strong relationship with other variables,such as tropical SST(SST gradiant),relative humidity and temperature.In atmospheric reanalysis,these variables are either boundary conditions or influenced by data assimilation.In general,different reanalysis have similar distributions of clouds,but there are still many differences in temporal and spatial distributions.Differences in cloud fields among reanalysis may therefore be influenced not only by the physical parameterizations used in the forecast model(including convection schemes,non-convective cloud schemes),but also by differences in the type of data assimilation(3D-VAR,3D-FGAT,4D-VAR).We present an evaluation and intercomparison of estimates of high cloud fraction and related variables in the tropics(clouds upon 500 hPa,within 30°S–30°N)from five reanalyses datasets(JRA-55,ERA-Interim,MERRA,MERRA-2 and CFSR).We identify large differences in geographic distribution,vertical profile,joint distribution with related variables seasonal cycles,and interannual variability of high cloud fraction.For example,for vertical profile,cloud fraction in the deep tropics peaks at higher altitudes in JRA-55 and ERA-Interim than in MERRA and MERRA-2,implying substantial differences in deep convective detrainment.MERRA and MERRA-2 produce larger high cloud fractions than the other reanalysis,especially in convective areas,while JRA-55 is the smallest in the tropical mean,primarily due to smaller high cloud fractions in the maritime continent region.The 4D-Var data assimilation used by ERA-Interim produced more realistic high cloud when compared with CERES.CFSR and CFSv2(using different model and data assimilation method from CFSR)show large discontinuities in time.The amplitude of the seasonal cycle in high cloud fraction also differs among reanalyses.