Incorporating Drought Thresholds Improves Model Predictions of Autumn Phenology in Tropical and Subtropical Forests

Drought dramatically influences vegetation phenology, thereby impacting terrestrial carbon and water cycles. However, the mechanisms by which drought drives changes in autumn phenology remain unclear, hindering the accurate simulation of these processes in phenology models. In this study, we employed ridge regression analysis to quantify the dynamic effects of intensifying drought on the end-of-photosynthetic-growing-season (EOPS) and identified the drought threshold at which the vegetation's response to drought shifts. We demonstrate that the response of EOPS in tropical and subtropical forests reverses from a delay to an advancement as drought intensity surpasses specific thresholds, with the average drought threshold across the study area corresponding to a standardized precipitation evapotranspiration index (SPEI) value of -0.9. Drought thresholds, however, vary geographically, increasing along the precipitation gradient, potentially due to variations in drought stress-related gene expression and tolerance strategies across different humidity environments. Therefore, we developed a new autumn phenology model (DMPD) by incorporating a drought threshold parameter that distinguishes contrasting drought effects and predicts future EOPS under two scenarios (SSP245 and SSP585). The DMPD model substantially enhanced the representation of EOPS, as evidenced by a lower root mean square error (RMSE), higher correlation, and a greater proportion of significant correlations with EOPS derived from GOSIF. By the end of the century, EOPS is projected to be consistently delayed under both moderate (SSP245) and high (SSP585) warming scenarios, with the rate of delay decelerating under SSP245 after 2066. Our study confirms that increasing drought intensity leads to contrasting shifts in the autumnal photosynthetic phenology of tropical and subtropical forests and highlights the potential of integrating these contrasting drought effects into phenology models to improve the accuracy of vegetation phenology predictions under future climate change scenarios.