ABSTRACT Against the backdrop of global change and efforts to combat land degradation, cropland abandonment (CA) has continued to occur in agricultural regions, altering land cover and ecological processes and thereby affecting land productivity (LPD) patterns and the efficiency of allocating restoration resources. However, dynamic cropland boundaries can readily introduce bias into abandonment statistics, and event‐based quantitative frameworks that distinguish long‐term productivity trends from the effects of abandonment events remain limited. Using the Huai River Basin (HRB) as a case study, this research identified CA events from 2000 to 2020 (continuous ≥ 3 years) based on annual land‐use trajectories under constraints imposed by dynamic cropland boundaries, and quantified LPD dynamics using net primary productivity (NPP). The Theil–Sen slope estimator and Mann–Kendall test were applied to characterize the magnitude and significance of LPD trends from 2001 to 2020. To address the lack of event‐based frameworks that separate long‐term trends from abandonment‐event effects, we anchored analyses to the abandonment start year ( t 0 ) and constructed (ΔNPP) to represent responses to abandonment events; CA–LPD coupling types were then delineated, restoration priority areas were identified using a rule‐based approach, and XGBoost–SHAP was introduced to interpret differences in drivers across “abandonment occurrence—post‐abandonment recovery—priority level”. The results show that: (1) CA was patchily distributed with pronounced interannual fluctuations (mean annual newly abandoned area: 613.78 km 2 ; reaching 1587.55 km 2 in 2020), while long‐term LPD change was dominated by increases (increase classes: 78.71%; significant/very significant increases: 43.09%; significant decreases: 1.93%). (2) Within abandonment‐event extents, coupling types were dominated by the recovery‐potential type (40.74%), whereas types associated with degradation constraints were rare (types 1 + 2 totaling 0.32%); high‐priority restoration areas were small in extent but spatially explicit (0.26%). (3) XGBoost–SHAP indicated systematic differences in the ranking and response forms of dominant factors across the three outcomes: topographic factors contributed most during the occurrence stage; the importance of hydrothermal and vegetation factors increased during the recovery stage; and, in the integrated priority stage, joint constraints from socioeconomic conditions and hydrothermal regimes became more prominent. This study provides a transferable spatial methodological framework and decision support for diagnosing abandonment–productivity degradation/recovery and identifying restoration priority areas in major grain‐producing regions under dynamic cropland patterns.
Zhang et al. (2026) studied this question.