Severe drought events inducing large decrease of net primary productivity in mainland China during 1982-2015

The analysis of the impact of drought events on terrestrial net primary productivity (NPP) is significant to understand the effects of droughts on regional/global carbon cycling. During the past three decades, terrestrial ecosystems in mainland China have been frequently impacted by drought events. However, quantitative analyses of the variation of NPP induced by droughts are still not enough. Therefore, this study explored the response of NPP to drought events from 1982 to 2015 based on the standardized evapotranspiration deficit index (SEDI) and an NPP dataset obtained from the Carnegie-Ames-Stanford Approach model. We first identified drought events and analyzed the characteristics of drought events using a three-dimensional clustering algorithm. Subsequently, we determined the NPP variations in the drought-affected areas during the droughts and explored the correlation between the NPP variation and the drought characteristics. The results showed that 152 persistent drought events lasting at least 3 months were identified. Most events had durations between 3 and 5 months, and 19 events lasted >9months. A negative NPP was detected in >60% of the drought-affected areas during long-term (>6months) and severe (>4x10(6) km(2) month) drought events and the total NPP showed a clear decrease during these events. In general, strong drought events reduced the total NPP by >30 TgC in the Northern Region, South Region, Southwest Region, and Northeast Region. The substantial decrease was mainly caused by the NPP anomaly from April to September. The NPP responses to drought events exhibited differences due to different drought characteristics. Although a high proportion of the drought-affected areas experienced a decrease in NPP during most short-term (<5 months) and less severe droughts (>2 x 10(6) km(2) month), the total NPP did not exhibit a large change during these events. (c) 2019 Elsevier B.V. All rights reserved.