Crop production is susceptible to annual biophysical shocks from weather phenomena that are increasingly becoming more extreme and potentially more damaging, because of human-induced climate change. Such shocks result in crop yields being more variable between years and threaten the stability of the food supply. The analysis of year-to-year variations in agricultural output is thus even more crucial in the face of current and future climate change. Given the escalating frequency of these shocks, yield impact models' continued efforts to understand yield fluctuations due to climate variations have not yet completely matured. This, in particular, is limited due to the paucity of data, especially in developing nations. The goal of this analysis is to capture interannual variation in crop yield in data-poor regions by learning from data-rich regions and crops. To this end, we propose a novel transfer learning approach. We utilize Long Short-Term Memory (LSTM) to develop a transfer learning model that is trained on historical annual corn yield data (as an example) at the county-level in the USA and predicts district-level corn yields in India. We achieve promising results in predicting interannual variations in India’s corn harvests despite the significantly smaller amount of data available. This approach can be used in diverse contexts and has the potential to advance the understanding of key socioeconomic and environmental challenges in the human-Earth system, especially in areas with sparse data and high shock susceptibility.