AI-Powered Forecasting for Precision Agriculture: Predicting Crop Yield from Satellite Imagery and Weather Data

Abstract

Precision agriculture represents the forefront of sustainable farming, where data-driven insights guide resource optimization and yield enhancement. In this study, we propose an AI-powered crop yield prediction framework that integrates satellite imagery with temporal weather data to accurately forecast agricultural productivity. Our approach combines remote sensing indices such as the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and soil moisture maps with historical weather variables, including daily temperature, rainfall, humidity, and solar radiation. Machine learning models namely random forest, Long Short-Term Memory (LSTM) networks, and hybrid convolutional neural network + LSTM (CNN-LSTM) architectures are trained and evaluated using publicly available Indian datasets like PRISM, MODIS, and NASA POWER. Results demonstrate that the CNN-LSTM model consistently outperforms traditional baselines, achieving high R² scores and low RMSE, particularly in early-stage yield estimation. Our method not only facilitates timely agricultural planning but also enables stakeholders farmers, policymakers, and agritech companies to make informed decisions regarding crop selection, irrigation scheduling, and food supply forecasting. This research underscores the potential of AI in transforming traditional agriculture into a resilient, smart system responsive to climatic and environmental variability.