Early leaf senescence under drought conditions in the Northern hemisphere

Changes in the dates of autumn foliar senescence (DFS) have significant impacts on regional carbon uptake, while current approaches for the estimation of DFS are still lacking. The most important issue is that there are complicated factors that affect the DFS, among which drought effects probably have contributed the most. Using long-term DFS observations derived from the third-generation normalized difference vegetation index dataset (NDVI3g), we found a wider spread of earlier DFS trends over the Northern Hemisphere from 1999 to 2015, three times larger than that from 1982 to 1998. The five multivariate analysis of variance approaches consistently suggest the key role of drought in regulating these changes. We therefore derived a new DFS algorithm with the standardized precipitation evapotranspiration index (SPEI) to characterize these drought effects, and validations from both NDVI3g and MODIS data demonstrated that our new algorithm provided significantly improved estimates of DFS for all plant functional types, with higher accuracy for water-limited ecosystems. We further applied this new algorithm to predict DFS under various shared socioeconomic pathways (SSPs) by the end of this century, and we found overall earlier DFS than the current expectations. Our results therefore highlight the importance of drought in the modeling of plant phenology using remote sensing observations and thus are highly important for understanding the relationships between land carbon sinks and climate change, especially given that droughts are projected to be more severe and frequent in the future.