Crop Prediction Uncertainty Maps

Crop rotation refers to planting different crops at different times in the same field. Farmers rotate crops to enhance soil fertility, structure, and biodiversity. Scientists and policy makers track crop rotations for environmental assessment, acreage, yield forecasting, and many other applications. Crop prediction, which refers to forecasting the next crop to be planted, is important for statistical applications, agri-business, and climate modeling. Crop prediction is commonly conducted using transition probability models. However, the study of the uncertainty associated with the categorical choice of crops is often neglected. This paper presents two measures of uncertainty for a comprehensive assessment of crop prediction quality. The main objective is to attribute the uncertainty of the prediction errors to specific aleatoric, epistemic, or mechanistic factors. Preliminary results show that these uncertainties are higher for fields with irregular crop rotations.