Monitoring grasshopper habitat in Hulunbuir using MaxEnt and Geodetector

Grasshopper outbreaks threaten grassland ecosystems, but habitat controls and long-term spatial trends remain poorly resolved. We combined 1,573 occurrence records from Hulunbuir (2008–2020) with remote-sensing, meteorological, soil, and topographic data. Twenty-six predictors represented overwintering, incubation, eclosion, and oviposition. Cross-year Pearson screening reduced redundancy, while SPEI-12-stratified Geodetector analysis tested factor stability across climate backgrounds. Independent annual MaxEnt models used linear and quadratic features, a regularization multiplier of 4.5, four-fold spatial block cross-validation, and 30 bootstrap replicates. Nine predictors were retained: PlantType, SoilType, OV_FVC, IP_Pr_mean, IP_Mois, OP_LST_min, EP_LST_mean, DEM, and Slope. Factor rankings remained stable across climate groups (Spearman ρ = 0.911–0.964), and IP_Pr_mean × PlantType was consistently the strongest interaction. Mean validation AUC was 0.806 ± 0.061, and the mean Continuous Boyce Index was 0.620 ± 0.163. IP_Pr_mean, EP_LST_mean, and OV_FVC had the highest multi-year permutation importance. Most and moderately suitable habitats clustered in western and southwestern Hulunbuir and changed through local expansion and contraction around a stable core. Sen–Mann–Kendall analysis found no significant trend in 88.84% of grassland; significant increases and decreases covered 2.15% and 4.30%, respectively. This framework supports annual habitat monitoring and targeted grasshopper surveillance.