Crop disease surveillance through integration of machine and deep learning in the face of climate change

Early detection of potato plant diseases is vital for agricultural productivity, plants, economy, and human health. Preventing disease spread minimizes agricultural losses and safeguard food production. Potatoes are a major food source worldwide, and reduced production could contribute to food insecurity. Precise and automatic plant disease detection represents a fundamental research topic. This study focuses on detecting potato disease risk associated with Early Blight (EB) and Gray Mold (GM) spores under the influence of dynamic climate. The aim of this study is to utilize Deep Learning (DL) and Machine Learning (ML) technologies for the disease detection of infected and healthy potato plants in Atlantic Canada, for Prince Edward Island (PEI) and New Brunswick (NB); and United States (U.S.), for Maine, to mitigate agricultural losses. The novel potato dataset used in the study comprised 5630 instances, with 1482 instances from PEI, 3032 from NB, and 1116 from Maine. A hybrid Artificial Neural Network - Random Forest (ANN-RF) is implemented model for binary disease classification, where the presence of EB or GM spores is considered as a single ‘diseased’ class. The hybrid ANN-RF model yields testing accuracy of 91 % for PEI, 86.5 % for NB, 88.4 % for Atlantic Canada, 84 % for Maine, and 87.5 % for combined data, with training accuracy of 97 %, 94.7 %, 91 %, 94.5 %, and 91 %, respectively, and log loss values under 0.4. These results highlight the potential of the developed model for disease detection in potato crops based on historical spore and weather data. The research findings examine potato diseases across multiple regions, uncovering geographical dissimilarities that shape disease prevalence, patterns, and management effectiveness. The study emphasizes the effectiveness of the hybrid model for predicting the likelihood of potato disease based on pathogenic spores, demonstrating the importance of AI in enhancing agricultural productivity, pesticide inventory control, and promoting sustainable farming practices.