Indian Agriculture in a Changing Climate: Using CMIP6 Projections for Predicting Yields of Multiple Crops To 2100

Agricultural production is sensitive to short-term weather variability as well as long-term climate change, and predicting crop yields as a function of climate is getting increasingly important due to a rapidly changing environment. Future climate projections continue to improve every day, and the latest CMIP6 framework is the most recent step in this direction. Given the easy availability of high-quality data and advanced computational facilities in recent years, researchers now have a whole arsenal of statistical tools at their disposal. This study models crop yields as a function of climate using two different techniques: the popular OLS linear regression, as well as an advanced machine learning based technique that fits a more flexible non-parametric functional form to the complex relationship between yields and climate. An ensemble of models with various seasonal and sub-seasonal climate variables including temperature, growing degree days, precipitation amount and frequency, soil moisture, heat and drought events, etc, are combined with climate projections for 4 different Shared Socioeconomic Pathways from 13 GCMs from CMIP6, in order to predict yields for 9 different crops till 2100 for India. We discuss and contrast the yield sensitivity to climate change, both across regions and crops. Because precipitation variability and drought probability is expected to increase in the future, we also estimate the role irrigation can play in reducing yield vulnerability to climate change by running our models for different irrigation availability scenarios. This study improves our understanding of Indian agricultures vulnerability to a changing climate, and can assist policymakers in designing adaptation and mitigation strategies.