Enhanced Agricultural Forecasting: Climate Change and Crop Yield Prediction using Hybrid ML-DL Models

Abstract

Food security has become threatened owing to climate change being a negative influence on agricultural growth and its subsequent role in pressuring the availability of such essentials as water and soil nutrients, and, finally, in its role in pressuring crop productivity. Meeting this need requires building a precise model of the interaction between climate variability and crop production for practical, sustainable agricultural planning. So, this study has proposed a comprehensive deep-learning framework for analysing and predicting the impacts of climate change on crop production, using three architectures: Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Convolutional Neural Network (CNN) [5]. These historical climate datasets-temperature, rainfall quantity, humidity, and solar radiation-are combined with crop yield data for training and evaluating the models. To model long-term temporal dependencies within climate sequences and capture meaningful patterns of variability over time, LSTM and GRU architectures are implemented. CNN serves as a complementary model for extracting meaningful spatial and multidimensional features related to crop production. Thus, the integration of these architectures yields a stronger, more reliable prediction system that can balance between sequential learning and pattern recognition.