Assessment of drought and its impact on winter wheat yield in the Chinese Loess Plateau

Drought has pronounced and immediate impacts on agricultural production, especially in semi-arid and arid rainfed agricultural regions. Quantification of drought and its impact on crop yield is essential to agricultural water resource management and food security. We investigated drought and its impact on winter wheat (Triticum aestivum L.) yield in the Chinese Loess Plateau from 2001 to 2015. Specifically, we performed a varimax rotated principal component analysis on drought severity index (DSI) separately for four winter wheat growth periods: pre-sowing growth period (PG), early growth period (EG), middle growth period (MG), and late growth period (LG), resulting in three major subregional DSI dynamics for each growth period. The county-level projections of these major dynamics were then used to evaluate the growth period-specific impacts of DSI on winter wheat yields by using multiple linear regression analysis. Our results showed that the growth period-specific subregions had different major DSI dynamics. During PG, the northwestern area exhibited a rapid wetting trend, while small areas in the south showed a slight drying trend. The remaining subregions fluctuated between dryness and wetness. During EG, the northeastern and western areas exhibited a mild wetting trend. The remaining subregions did not display clear wetting or drying trends. During MG, the eastern and southwestern areas showed slight drying and wetting trends, respectively. The subregions scattered in the north and south had a significant wetting trend. During LG, large areas in the east and west exhibited wetting trends, whereas small parts in south-central area had a slight drying trend. Most counties in the north showed significant and slight wetting trends during PG, EG, and LG, whereas a few southwestern counties exhibited significant drying trends during PG and MG. Our analysis identified close and positive relationships between yields and DSI during LG, and revealed that almost all of the counties were vulnerable to drought. Similar but less strong relationships existed for MG, in which northeastern and eastern counties were more drought-vulnerable than other counties. In contrast, a few drought-sensitive counties were mainly located in the southwestern and eastern areas during PG, and in the northeastern corner of the study region during EG. Overall, our study dissociated growth period-specific and spatial location-specific impacts of drought on winter wheat yield, and might contribute to a better understanding of monitoring and early warning of yield loss.