A-Train Observational Metrics for CMIP5 Model Evaluations and Diagnoses

A-Train satellites provide a suite of measurements on atmospheric temperature, water vapor, cloud profiles and other tracer concentrations. These nearly simultaneous and collocated observations serve as important observational metrics for evaluations of CMIP5 model simulations. With extensive validation efforts by instrument teams, uncertainties of these measurements are well documented. Hence, a new era of model evaluation is emerging with significant advances from previous generations in that (a) multiple correlated parameters (instead of single parameter only) are examined at the same time; (b) vertical structures of cloud and water vapor (instead of column-integrated quantities only) are depicted; (c) process understanding is rendered through correlation analysis of multiple variables; (d) uncertainties (error-bars) in the observational measurements is clearly marked for quantitative model improvements. We have performed detailed evaluations for CMIP5 climate model simulations of clouds masses (ice and liquid) and atmospheric water vapor mixing ratios using A-Train data, and compared the model performance for CMIP3 and CMIP5. We have found improvements in modeled ice water path and liquid water path in the current CMIP5 versions, comparing to CMIP3, although the largest model-observation difference still stands as large as a factor of ~5, compared to a factor of more than 10 for CMIP3. The simulated water vapor path remains to be within 30% of observations for CMIP3 and CMIP5. The vertical distributions of clouds and water vapor are also compared with A-Train observations, and we found the largest spread among models and significant departures from observation are in the tropical upper troposphere. A grading scheme is developed to quantitatively assess the model performance for climatological mean clouds and water vapor distributions and the results will be reported.