Object-based evaluation of rainfall forecasts over eastern China by eight cumulus parameterization schemes in the WRF model

Abstract

Traditional pointwise verification is not fully accurate for the evaluation of rainfall forecasts due to the double-penalty problem. An alternative object-based metric, Method for Object-based Diagnostic Evaluation (MODE), was used in this study to evaluate the rainfall simulation performance of eight cumulus parameterization (CP) schemes in the Weather Research and Forecasting model. The average rainfall from May to August (MJJA) over eastern China from 2009 to 2018 and the sub-seasonal evolution of rainfall systems were mainly verified. The results showed that the heavy rainfall objects simulated by KFCuP, New Tiedtke and KSAS schemes are the closest to the observations over South China, Yangtze River Basin and North China, respectively. MODE could highlight the advantages of these schemes in the simulation of rainband physical attributes, such as centroid position, orientation, rainfall area, etc. At the sub-seasonal scale, KFCuP had the highest skill score for mimicking the observed heavy rainfall objects. To explore the reasons for the differences in rainfall simulations, the evaporation, water vapour inflow and precipitation efficiency were calculated for each scheme. It was found that members with higher precipitation generally corresponded to larger water vapour inflow and precipitation efficiency, and the water vapour inflow mainly came from the southern boundary. Further analysis found that there existed a significant positive correlation between the meridional wind and the water vapour inflow on southern boundary, which indicated that the meridional wind was an important factor affecting the regional rainfall simulation.