From satellites to yield: causal modeling of paddy rice production using sparse regression and dynamic remote sensing

This study presents a novel approach that combines agricultural census data with remotely sensed time series to develop accurate predictive models for paddy rice yield across the different regions of Peru. By leveraging sparse regression and Elastic-Net regularization techniques, the study uncovers causal relationships between key remotely sensed variables such as Normalized Difference Vegetation Index (NDVI), precipitation (PREC), temperature (TEMP), and agricultural yield. To further enhance prediction accuracy, first- and second-order dynamic transformations (velocity and acceleration) of these variables were applied to capture non-linear patterns and lagged effects on yield. The findings demonstrate improved predictive performance when integrating regularization techniques with climatic and geospatial variables, allowing for more accurate forecasts of yield variability. The results confirm the presence of causal relationships in the Granger sense, underscoring the value of this methodology to strategic agricultural management. This contributes to more efficient and sustainable production in paddy rice cultivation.