*Article* **Satellite-Observed Global Terrestrial Vegetation Production in Response to Water Availability**

**Yuan Zhang 1,2, Xiaoming Feng 1,\* , Bojie Fu <sup>1</sup> , Yongzhe Chen 1,2 and Xiaofeng Wang <sup>3</sup>**


**Abstract:** Water stress is one of the primary environmental factors that limits terrestrial ecosystems' productivity. Hense, the way to quantify gobal vegetation productivity's vulnerability under water stress and reveal its seasonal dynamics in response to drought is of great significance in mitigating and adapting to global changes. Here, we estimated monthly gross primary productivity (GPP) first based on light-use efficiency (LUE) models for 1982–2015. GPP's response time to water availability can be determined by correlating the monthly GPP series with the multiple timescale Standardized Precipitation Evapotranspiration Index (SPEI). Thereafter, we developed an optimal bivariate probabilistic model to derive the vegetation productivity loss probabilities under different drought scenarios using the copula method. The results showed that LUE models have a good fit and estimate GPP well (R<sup>2</sup> exceeded 0.7). GPP is expected to decrease in 71.91% of the global land vegetation area because of increases in radiation and temperature and decreases in soil moisture during drought periods. Largely, we found that vegetation productivity and water availability are correlated positively globally. The vegetation productivity in arid and semiarid areas depends considerably upon water availability compared to that in humid and semi-humid areas. Weak drought resistance often characterizes the land cover types that water availability influences more. In addition, under the scenario of the same level of GPP damage with different drought degrees, as droughts increase in severity, GPP loss probabilities increase as well. Further, under the same drought severity with different levels of GPP damage, drought's effect on GPP loss probabilities weaken gradually as the GPP damage level increaes. Similar patterns were observed in different seasons. Our results showed that arid and semiarid areas have higher conditional probabilities of vegetation productivity losses under different drought scenarios.

**Keywords:** LUE-GPP; SPEI; copula function; conditional probability
