Causal Inference of Nonlinear Vegetation Response to Earlier Seasonal Thaw in a Semi-Arid Mountain Permafrost Region | CU Experts

Abstract; Under rapid warming, the impact of advancing seasonal permafrost thaw on vegetation dynamics remains debated, particularly regarding whether earlier spring phenology induces late-season productivity deficits. Integrating multi-source remote sensing data over the Altai Mountains, this study employs a Double Machine Learning (DML) framework to quantify the causal effects of thaw timing on vegetation growth. The results indicate that mid-May acts as a critical phenological tipping point. Thaw onset prior to this date provides a net subsidy, significantly enhancing peak growing season Leaf Area Index (LAI) with a magnitude comparable to precipitation and solar radiation. Our causal inference and strict matching analysis confirm the positive stimulation effect during the peak growing season (July). It also confirms that the legacy effects of the spring thaw almost completely dissipate by the late growing season (October). These results suggest caution when applying rigid linear assumptions to complex multivariate systems as such assumptions may fail to capture the true underlying controls.

VIVO