Evaluating the Effects of Climate Change on Agriculture: Deep Learning-Based Predictive Pest and Disease Severity Classification

The rising influence of climate change on the worldwide agricultural landscape necessitates the development of advanced forecasting technologies to evaluate its impact and aid adaptation. In order to predict the effects of climate change on agriculture, this study investigates the use of deep learning algorithms. To better understand the interplay between climatic variability and crop yields, we employ the ResNext50 architecture, which has been upgraded with an attention mechanism, in this study. One of the most pressing issues in agriculture brought on by climate change is the difficulty in classifying the severity of pest and disease infestations. Our study relies on a thoroughly curated collection of 2895 agricultural photos, all of which are labeled with severity annotations. The overall accuracy of the deep learning model is 97.37%, indicating a very high level of accuracy. With this precision rate, the model can be deemed effective in classifying severity levels starting with "No Pest" and up to "High Severity". Studies on interpretability, based on attention mechanisms, reveal the critical visual regions that greatly affect severity categorization, which lets use it to increase the accuracy of classification. This increases the implementation and applicability of the model to the stakeholders in the agricultural sector. This study offers findings that are significant for the topic of climate change adaptation in the agricultural setting. Prompt diagnosis and accurate assessment of the damages from a problem can be used as tools to equip farmers and policymakers with accurate responses when making decisions, which in turn, ensures a more efficient allocation of resources to reduce the damages on crop yields. Taking into consideration the changing nature of the climate cycle, our studies serve as a revolutionary discovery in the field of climate friendly farming. The acquired knowledge does have application other than just the classification of severity, but rather it gives stipulations of innovative techniques for the application of sustainable and changeable measures in light of climate unpredictability.