Linking Multidimensional Cropland Ecosystem Health Assessment, Spatialized Dominant-Driver Diagnosis and Mechanism-Informed Zoning in the Middle and Lower Yellow River, China

CONTEXT Cropland systems in river-basin agricultural regions are jointly shaped by hydro-climatic constraints, cropland configuration and human–land development pressure. Existing assessments often separate productive, structural and dynamic properties and rarely localize dominant drivers for differentiated regulation. OBJECTIVE This study developed a framework linking multidimensional cropland ecosystem health assessment, spatialized dominant-driver diagnosis and mechanism-informed zoning in the middle and lower reaches of the Yellow River, China. METHODS Cropland ecosystem health was evaluated from vitality, organizational stability and resilience using environmental, land-use and socioeconomic data harmonized to a 1 km grid for 2000–2011 and 2012–2023. Random Forest and XGBoost quantified driver relationships, while local TreeSHAP contributions, current CH grade and standardized diagnostic scores supported zoning. RESULTS AND CONCLUSIONS Cropland ecosystem health was relatively favourable in Henan and Shandong and lower in the western and northern sections. Between the two periods, 57.39% of grids shifted to lower vitality grades, whereas 93.45% and 64.64% shifted to higher organizational stability and resilience grades, respectively. Integrated health nevertheless declined in 54.62% of the grids. Precipitation remained the leading predictor, accounting for 38.0% and 35.4% of total TreeSHAP importance. The contributions of cropland density and GDP density increased from 10.0% to 16.1% and from 9.2% to 16.5%, respectively. Spatialized dominant-driver diagnosis supported the delineation of five regulation zones: mountain ecological conservation, extensive cropland maintenance, terrain–hydrology constraint regulation, cropland-intensive coordination and development-pressure control. SIGNIFICANCE The framework reveals asynchronous changes within integrated cropland health and links regional driver shifts and local dominant factors with differentiated management priorities.