Abstract: The spatial pattern of phenology reflects long-term plant adaptation to local environments, yet the drivers of these patterns remain poorly understood. Using satellite data from 2001 to 2018, this study employed the normalized difference vegetation index for vegetation structural greenness and solar-induced chlorophyll fluorescence for vegetation functional photosynthesis to analyze spring phenology on the Qinghai-Tibetan Plateau (hereafter, QTP). A machine learning method, Boosted Regression Trees (BRT), was applied to evaluate the contributions of 19 abiotic and biotic factors to the spring phenology. The results showed that both the spring leaf phenology (SOS_NDVI) and photosynthesis phenology (SOS_CISF) exhibited a delayed trend decreasing from east to west across the QTP. BRT analysis demonstrated shortwave radiation or/and elevation as key drivers, with higher radiation or elevation associated with more delayed spring phenology spatially, likely due to the constraints of extreme radiation and elevations on spring phenology. Furthermore, we also noted that plants were acclimated to strong radiation to some extent with increasing elevation, namely declined negative effect of radiation/elevation on spring phenology. This acclimation likely enhances plant fitness in the harsh environments of the QTP. Our study provides novel insights into plant phenology on the QTP and highlights the importance of integrating spatial and temporal analysis to improve the localization of phenology models.

Key words: spatial pattern, boosted regression trees, strong radiation, acclimation, phenology