Characterizing and Understanding Cloud Water and Radiation Budget Biases in CMIP3/CMIP5 GCMs, Contemporary GCMs and Reanalyses

Jui-Lin (Frank) Li
California Institute of Technology


We have utilized a considerable heritage of knowledge and experience (e.g., Li et al, 2007, 2008; Waliser et al., 2009) to perform a robust evaluation of the cloud ice water content (CIWC) and cloud ice water path (CIWP) from 20th century CMIP5 simulations and compare these results to the same analysis on CMIP3 (Li et al., 2012a). To account for observational uncertainty we use three different CloudSat+CALIPSO+MODIS retrieval schemes for tropospheric ice water content and two very different methods to remove the contribution from falling hydrometeors (i.e. Waliser et al. 2009 and Chen et al. 2011) so that a robust estimate for CIWC and CIWP - with quantified uncertainty - can be obtained for comparison to the GCMs. In addition, as an additional benchmarking measure, we include parallel evaluation of two reanalysis data sets (i.e. MERRA and ERA-Interim). For the same group of the models, we also conducted an observational based evaluation and analysis of TOA and surface radiation budgets (Li et al., 2012b) utilizing a number of contemporary satellite measurements for the TOA and surface (e.g. CERES-Aqua, CERES-Terra) and observationally/satellite-constrained model calculations for the surface (e.g. CERES product, CloudSat product).

The analysis and results show that in general the CMIP5 ensemble of models performs a bit better than CMIP3. For example, considering the Taylor Plot framework (see attached figures), the multi-model mean CIWP from CMIP5 is more than 50% closer to the reference value than the multi-model mean from CMIP3. One systematic bias across models appears to be from an overestimate of CIWP/CIWC in the extra-tropics and storm tracks. In addition, a persistent bias in most of the models and the multi-model mean is the overestimate of OLR and surface shortwave as well as underestimate reflected SW at TOA in the strongly convective region of the tropics. We suggest that at least part of this persistent bias stems from GCMs ignoring the effects of precipitating/convective core ice and liquid in their radiation calculations.

We perform a series of model sensitivity tests in order to examine the impacts of exclusion of the precipitating hydrometeors on atmospheric radiative fluxes and heating rates, as well as surface precipitation and/or dynamics using an offline radiative transfer model with CloudSat IWC and other satellite input data (Waliser et al., 2011) and with the ECMWF IFS (Li et al., 2012c). The offline sensitivity tests indicate that the systematic biases in the radiation budget are likely in part due to the exclusion of the precipitating components in the radiation calculations. The interactive tests with the ECMWF IFS also indicate there may be feedbacks from these impacts that might result in excessive of updraft velocity, convective intensity and convective precipitation over convective active regions.