Diurnal cycle of rainfall as predicted by wrf model: Verification using model evaluation tools software

Abstract

The quality of sub-daily rainfall forecast from the Weather Research and Forecasting model during June to September, 2013 is verified at three hourly intervals for forecasts up to 72 hours against the corresponding data from the Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis (TMPA) dataset. The validation is done in two stages, in the first stage, the diurnal cycle of the 3 hourly accumulated seasonal totals of the verification rainfall dataset (TMPA) is compared with the corresponding 3 hourly accumulated seasonal totals of the rainfall forecasts from WRF model for forecast up to 72 hours. The discrepancy in the spatial and temporal maximum of model derived diurnal cycle of rainfall with respect to the TMPA dataset is then investigated quantitatively for every forecast-verification dataset pair using the MODE (Method for Object-based Diagnostic Evaluation) component of the Model Evaluation Tools (MET) software. The analysis demonstrates that the rainfall and convection is strongly forced by surface sensible heating in the model. This causes the diurnal peak of convection and rainfall over the entire subcontinent to be in phase with the diurnal peak of temperature. Hence, the diurnal cycle errors are highest over regions, where other forcing factors actually pre-dominate; especially along the coasts and the foothills of the Himalayas. The sensible heat forcing tends to increase the areas of convection in the WRF model, rather than the number of zones of genesis of convection, indicating that its influence is more towards cloud organization rather than genesis of new convection in the WRF model. There is a strong component of displacement error in the WRF model forecast, which may cause large errors in location specific forecasts. The increase in the object size in the model forecast close to the diurnal maximum of temperature, accompanied by a peak of the rainfall intensity range, suggests the dominance of parameterized convection in the model rainfall output.