Impacts of the stochastically perturbed parameterization on the precipitation ensemble forecasts of the Betts-Miller-Janji<acute accent>c (BMJ) scheme in Eastern China

This study investigated the impact of stochastic perturbations on precipitation forecasts generated by the BettsMiller-Janjic (BMJ) cumulus parameterization scheme. Two types of stochastic perturbation approaches were compared. The first approach involved perturbing the temperature and humidity tendencies derived from the BMJ scheme. The second approach focused on perturbing the reference profiles of temperature and humidity estimated within the BMJ scheme. These profile perturbations led to reference profiles becoming either warmer and wetter or colder and drier. Ten precipitation cases occurring in eastern China during the summer of 2019 were selected to evaluate the different perturbation methods. The default BMJ scheme exhibited a significant wet bias at the light rain threshold due to overestimating entropy change and a dry bias at the heavier rain threshold. The tendency perturbation approach, which perturbed the temperature and humidity, did not affect the entropy change and produced precipitation placement similar to its unperturbed counterpart. These attributes resulted in a small ensemble spread and relatively low forecast skill scores. Perturbing the reference profiles, conversely, influenced the entropy change in the BMJ scheme, leading to a larger ensemble spread and higher forecast skill scores regarding the spatial distribution of precipitation. Asymmetrically perturbing the reference profiles considered the wet bias, which increased the grid points with negative entropy change and outperformed the symmetric perturbation. The asymmetric perturbation also increased the chance of heavier rain because more water was retained in the air, alleviating the corresponding dry bias.