Inter-Field Wheat Yield Variation Evaluated Using Time Series Satellite Imagery and Functional Data Analysis

The growing number of large agricultural corporations in Japan requires an efficient means to evaluate crop growth and yield formation for multiple fields. Using a combine harvester with a yield monitor sensor, we collected wheat yield data for 76 fields cultivated by an agricultural corporation. This study used functional regression and functional data analysis (FDA) to analyze the relation between inter-field wheat yields and growth using the time-series Normalized Difference Vegetation Index (NDVI) calculated from Sentinel-2 satellite imagery. The observed inter-field yield variation was 17–361 g m−2. FDA applied functional regression with two explanatory variables: the estimated functional NDVI and the sowing date. The inter-field yield variation was explained by the coefficient of determination (R2) = 0.39. The coefficient function indicated that functional NDVI affected yield. This study assessed the possibility of using satellite information for agricultural management and provided an alternative to analyzing time series data to infer crop growth and yield variation.