Results of a cloud-resolving model intercomparison study based on the TWP-ICE field campaign

The US Department of Energy Atmospheric Radiation Measurement (ARM) and GEWEX Cloud System Study (GCSS) programs coordinated a cloud-resolving model (CRM) intercomparison case study based on observations obtained under active and suppressed monsoon conditions during 16 days of the Tropical Warm Pool—International Cloud Experiment (TWP-ICE). Results from eight combinations of CRM dynamics and microphysics schemes were submitted, including two with 2D domains, and three with two-moment microphysics schemes. Sensitivity test results were also submitted for half of the eight, wherein nudging to observed thermodynamic fields with a six-hour timescale was included to minimize drift from observed conditions. Considering only the baseline simulations, predicted values of liquid and ice water path, surface precipitation rate, convective mass flux, cloud optical thickness and top-of-atmosphere albedo all vary by roughly a factor of two over the ensemble, indicating 50-100% uncertainty in simulated microphysical and radiative response to identically prescribed forcings. Considering both baseline and sensitivity test results, while precipitation rate is well correlated with liquid water path (LWP) across all models and times, the ratio of ice water path (IWP) to LWP in 2D results is roughly half that in 3D results during the active period, which seems related to a large, systematic reduction in convective mass flux in the 2D relative to the 3D simulations. Models with a higher convective mass flux appear to precipitate less efficiently, as defined by the ratio of surface precipitation rate to net condensation rate, but IWP does not correlate with precipitation efficiency across the ensemble. In 3D models, where convective and stratiform areas can be identified and compared with radar measurements, predicted convective areas vary by less than a factor of two, but deep stratiform areas (reaching 8 km) vary by a factor of six, and IWP tends to increase with increasing stratiform area, with some outliers. While cloud optical thickness is dominated by the liquid phase and is tightly correlated with total condensate path, LWP and IWP are respectively less correlated with liquid and ice optical depth, suggesting that differences in phase partitioning and ice properties explain some scatter in the latter relationships. The intercomparison specification, initialization and forcing data, submitted model results (including 3-hourly dumps of model fields), and processed field data sets are available at the ARM data archive for public use. Work described in other presentations at this conference (Varble et al., Smith-Mrowiec et al.) demonstrate specific applications of field data to probe simulated microphysics and precipitation fields and guide model improvement efforts.