Nonlinear response patterns of grain yield to drought gradients and the underlying driving mechanisms

Drought is defined as a period with significant precipitation deficiency over one season or longer, leading to water shortage that constrains crop growth, agricultural production and ecological environment. However, most existing studies focus only on linear yield–drought relationships and rely on single analytical methods, with limited research on nonlinear yield variations along drought gradients and their driving mechanisms. This study used grain yield data and key indicators (Aridity Index (AI), Precipitation (P), Temperature (T), Soil Moisture (SM), Elevation (El), Crop Diversity (CD), Yield Stability (YS)) to explore nonlinear grain yield variations along the aridity gradient in Northeast China during 2010–2020. Using Geodetector and Structural Equation Models, it clarified the nonlinear driving mechanisms of grain yield and compared yield response patterns of soybean, rice, wheat and maize. Results showed (1) A drought threshold (AI = 1.52) was identified, where total grain yield relationship with AI shifted from positive to negative. (2) Total yield and pre-threshold yields were both dominated by El (negative), whereas post-threshold yields were dominated by P (positive). P was identified as the key driver underlying the nonlinear yield variation. (3) Soybean and wheat yields are primarily dominated by CD (positive) and T (negative); rice yield is mainly driven by P (positive) and T (positive); whereas maize yield is predominantly influenced by AI (negative) and El (negative). These findings provide a scientific basis for optimizing high-standard farmland and planting structures in Northeast China under future climate change.