A Hybrid Deep Learning Model CNN-GRU for Crop Yield Prediction Under Climate Impacts: A Case Study of Rice Yield Prediction in India

Rice is one of the most important crops grown throughout India and has a significant impact on the socioeconomic status of the country. The Indian Summer Monsoon (ISM) provides a crucial water supply that supports rice production. However, climate change has led to erratic weather patterns that exacerbate the intensity of climate variables. Given the relationship between climate and rice production, it is important to understand and predict rice yields to strengthen the resilience of the socioeconomic sector. Recent improvements in deep learning models show greater predictive capabilities in extracting spatial or temporal features. This study aims to develop a hybrid deep learning-based model for rice yield prediction in India by integrating CNNs with GRU and using climate projection data from CMIP6 and AgMIP crop models. The model shows good prediction accuracy but tends to slightly overestimate yields, especially in northeast India. The predictions indicate a decline in rice yields, with significant losses expected under the SSP585 scenario.