Subseasonal Asian summer monsoon (ASM) rainfall prediction is crucial for disaster risk reduction and sustainable development, yet remains challenging for dynamical prediction systems. This study demonstrates that refined Tibetan Plateau surface soil temperature (ST-TP) initialization effectively boosts subseasonal prediction skill for ASM rainfall, by comparing two parallel 15-year (2010–2024) fully coupled prediction experiments with/without refined ST-TP. The ST-TP initial assimilation substantially improves one-month lead rainfall prediction over both TP and core ASM regions, cutting the 15-year mean July rainfall bias by ~23%. It also yields significant improvements in interannual variability and extreme rainfall occurrence, with the largest gains over southern flank of TP and Middle–Lower Yangtze River Region. These improvements arise from a physically coherent pathway: correcting ST-TP initial biases enhances surface latent heat release and ascent motion in early July, driving a circulation response with an intensified South Asian High and a southward-displaced Western Pacific Subtropical High that better matches observations, thereby reshaping moisture transport and rainfall distribution across the ASM regions. These results confirm that surface soil temperature with distinct subseasonal inertial memory is a crucial source of subseasonal predictability, while underscoring the imperative of assimilating land surface parameters to enhance subseasonal predictive skill.

次季节