Vol. 7, Issue 10, Part A (2025)

Climatic changes and rising pest attacks have been a very big issue to agricultural production, particularly in the environment when there are scarce resources and when the use of technology is not a key factor. In the proposed research, the researchers present a highly transparent framework based on explainable AI to combine remote sensing, environmental variables, and crop health indicators to predict the occurrence of disease outbreak and the risk of yield. The use of light ensemble models and SHAP interpretability techniques empowers the framework to provide farmers and agronomists to be able to see the most influential factors driving decisions, all the while making accurate predictions. Measurements that come in the UAVs, IoT sensors in the soil, and satellite feeds will be processed in the preprocessing phase where there is an optimization of features before being input in a predictive model that is a hybrid using XGBoost, LightGBM, and LSTM. The results of the models are made visual through an interactive dashboard that defines the areas that are most prone to the disease, and where yield losses might occur. The solution is optimized to support low-power devices to make it suitable to use in the countryside. Experimental testing proves good prediction performance in tight real-time field situations as well as being interpretable and responsive. The proposed solution would provide a flexible and transparent decision-support process, and eventually, it will provide agricultural stakeholders with data-driven interpretable information on how to reduce crop loss and become more resilient. The proposed system achieved 94.20%-point DR and 74.60%-point MSS, which were better than peers in the disease detection and predictive stability.