Precision agriculture has substantial hurdles in wheat crop disease because of the complicated environmental variability. This study presents a novel method by fusing Random Forest models optimized using nature-inspired algorithms like Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Grey Wolf Optimization (GWO), Differential Evolution (DE), and Honey Badger Algorithm (HBA) with explainable artificial intelligence (XAI). Nine important climatic and biophysical factors are assessed, such as precipitation (Pr), maximum and minimum temperatures (Tmax and Tmin), the Green Chlorophyll Index (CGI), the Normalized Difference Chlorophyll Index (NDCI), the Modified Soil-Adjusted Vegetation Index 2 (MSAVI2), the Normalized Difference Vegetation Index (NDVI), soil organic carbon (SOC), and total nitrogen (TN). Multicollinearity analysis and Boruta were used to evaluate feature importance. SSP2-4.5 and SSP5-8.5 scenarios from CMIP6 were used to predict climate projections using GEE (1990–2030) across three GCMs (EC-Earth3, NorESM2-LM, and MIROC6). Blight, rust, fusarium wilt, and powdery mildew disease predictions were verified using 10-fold cross-validation and field data observed by farmers. Based on current research, the best AUC for powdery mildew illness (AUC = 0.946) was represented by RF-GWO, the highest AUC for rust (AUC = 0.897) was recorded by the RF-DE model, and the highest AUC values for blight (AUC = 0.833) and fusarium wilt (AUC = 0.836) diseases were created by RF-GA. RF-GA performs the best on average. The method's importance for sustainable agriculture and SDG accomplishment was supported by the XAI interpretation, which identified temperature and precipitation as major disease-promoting factors. Furthermore, it is a cutting-edge decision-support system that transforms Morocco's wheat disease management by fusing XAI with cutting-edge nature-inspired random forest optimization (NIRFO). • A novel framework on integrating nature-inspired random forest optimization and explainable artificial intelligence . • The present & future wheat crop disease has been predicted to be using various random forest optimization methods, including GA, PSO, GWO, DE, and HBA. • Climate change scenarios were generated using of the integrated three GCMs model (EC-Earth3, NorESM2-LM, and MIROC6) of CMIP6 Shared Socioeconomic Pathways dataset. • SHAP was used to analyze how different factors affect wheat crop disease prediction. • Perturbation sensitivity analysis was applied for the importance of the influencing parameters.
Sahoo et al. (Wed,) studied this question.