Integrated Artificial Intelligence in Weather Forecasting for Agriculture: Opportunities, Challenges, and the Road Ahead

Abstract

Farming is very vulnerable to weather change, variable rainfalls, temperatures, humidity, and prevalence of extreme climatic conditions, which directly affect growth and yields, as well as livelihoods of farmers. Microclimate Small scale details about microclimate are generally missing in classical forecast systems - statistical models, numerical weather prediction (NWP) and expert forecasts due to the coarse resolution in space and time. In recent times, AI, and in particular, machine learning (ML) and deep learning (DL) is transforming agricultural weather forecasting, including processing massive volumes of data gathered by satellites, on-ground sensors, etc. This review evaluates the use of AI in prediction of rainfall, temperature, humidity, wind, and extreme events in which case studies within and beyond India show an increase in prediction accuracy, reduction in prediction error and lead time. The major progress is the hybrid AI-NWP models, the multimodal data fusion and the IoT-based sensor network, which allow utilizing real-life benefits in the area of irrigation scheduling, pest and disease management and disaster early-warning systems. Hyperlocal advisory platforms and edge computing are also capable of supporting real-time field level decision making precision farming. Nevertheless, there are still challenges (e.g. low data quality, high computational needs, poor rural-infrastructure, socio-economic restrictions) that hinder uptake. Researchers, policy-makers and technologists should be included to resolve these issues, to direct the design of technologies to user-friendliness and trust among farmers. The emerging pathways - adaptive AI, block-chain secured edge systems, and customized advisories, hold significant opportunities of creating resilience, resource-use efficiency, and food security in the context of increased climate variability.