Sensitivity of WRF Rainfall Simulations to Physical Parameterization Schemes over Tropical Peatlands in the Maritime Continent

Abstract

Mitigating hydrometeorological hazards, particularly peatland fires, requires reliable rainfall simulations over the Maritime Continent (MC). However, accurately simulating rainfall remains challenging because of complex interactions among convection and land–atmosphere processes in tropical peatland regions. This study evaluates the sensitivity of rainfall simulations to different physical parameterization schemes in the Weather Research and Forecasting (WRF) model, focusing on two microphysics schemes (WRF Single-Moment 3-class (WSM3) and WRF Single-Moment 6-class (WSM6)), three cumulus schemes (Betts–Miller–Janjic (BMJ), New Tiedtke (NT), and Kain–Fritsch (KF), and two planetary boundary layer (PBL) schemes (Yonsei University (YSU) and Mellor–Yamada–Janjic (MYJ)). These configurations were evaluated over southeastern Borneo during the 2019 El Niño dry season. Model performance was assessed using the probability of detection (POD), proportion correct (PC), mean absolute error (MAE), and Spearman correlation coefficient (ρ). Rainfall simulations were highly sensitive to the choice of cumulus and PBL parameterizations, with convective precipitation (RAINC) accounting for an average of 82.5% of total simulated rainfall. BMJ performed best in detecting rainfall events, whereas NT showed superior performance under no-rain conditions. Both schemes generally performed better with the YSU scheme, and the BMJ–YSU combination exhibited relatively balanced performance across four evaluation metrics, with no pronounced degradation in any individual metric. Statistical analyses showed no significant differences between WSM3 and WSM6, whereas cumulus and PBL parameterizations significantly affected rainfall simulation performance. These findings provide guidance for selecting WRF physical parameterization schemes for tropical peatland rainfall simulations under dry-season conditions within the tested model configuration.