The Intraseasonal-to-Daily Variance Ratio Shapes the Spatial Patterns of Subseasonal Predictive Skill for Asian Summer Land Precipitation

Skillful subseasonal prediction of Asian summer land precipitation (ASLP) is crucial for sustainable societal development, yet remains a significant challenge in climate science. Here, we investigate the drivers and limitations of subseasonal prediction skill across 12 subseasonal-to-seasonal (S2S) models, among which ECMWF demonstrates consistently superior performance. We identify a typical spatial pattern in prediction skill, with the highest skill over Pakistan and western-central India, followed by Korea–Japan, the Tibetan Plateau, and eastern China. We find that both the spatial pattern of prediction skill and its inverse decay rate—that is, the persistence of skill—are intrinsically linked to the intraseasonal-to-daily variance ratio (IDVR), defined as the ratio of intraseasonal variance with periods longer than 22 days to total daily variance, and serving as a measure of intrinsic predictability at the intraseasonal time scale. The IDVR, governed by internal atmospheric dynamics, indicates the extent to which future atmospheric states can be anticipated from observed past behavior and structure. The IDVR outperforms other predictability measures—including the weighted permutation entropy of observed precipitation and the signal-to-noise ratio—in explaining the models’ predictive skill. The Tibetan Plateau, characterized by high intrinsic predictability, exhibits relatively low overall prediction skill, suggesting significant potential for model improvement in this topographically complex region. These findings open a new avenue for estimating subseasonal predictability and advancing our understanding of model predictive capabilities.