Optimized MaxEnt Modeling and Environmental Suitability Assessment of Musa spp. in the Amazonas Region, Peru

Identifying environmentally suitable areas for banana and plantain cultivation is relevant to agricultural planning in the heterogeneous landscapes of Amazonas, Peru. We assessed the current suitability of cultivated Musa spp. using an optimized maximum entropy (MaxEnt) model and a complementary GIS-based model parameterized from MaxEnt response curves. MaxEnt was developed from 461 filtered georeferenced occurrence records, with 70% assigned to training and 30% to internal testing. Of 37 bioclimatic, edaphic, topographic, and spectral predictors, 14 were retained following correlation analysis, preliminary jackknife assessment, and multicollinearity screening. We evaluated 42 combinations of feature classes and regularization multipliers. Although LQPTH–RM0.5 had the highest composite score (0.9315), LQPH–RM1 was selected for its comparable score (0.9082), minimum corrected Akaike information criterion (AICc = 9708.41), parsimonious structure, and response curves consistent with the environmental gradients of the study area (mean test AUC = 0.9715 ± 0.0026). Temperature seasonality (BIO4) contributed most during model fitting (62.5%), whereas total soil nitrogen had the highest permutation importance (47.8%). MaxEnt identified highly suitable areas mainly in Bagua and Utcubamba, while the GIS-based model represented broader gradients of environmental favorability. The resulting maps provide spatial information for preliminary agricultural zoning, although independent field validation remains necessary.