Spatio-temporal yield variation and precipitation within a field

Abstract

In rainfed systems, precipitation significantly influence crop yield variability across years. Understanding the spatio-temporal yield variation within a field is crucial for enhancing farming production and profitability. This study aims to establish a data pipeline to quantify spatio-temporal yield variation within a field in response to precipitation, using multiple years of yield-precipitation records. Four field case studies – two with maize and two with soybean − each containing at least five years of yield monitor data of the same crop are used to evaluate the proposed data pipeline. Daily precipitation data of each case study were obtained from CHIRPS dataset and used to calculate derived variables representing cumulative precipitation over different periods. Generalized Additive Models (GAMs) were fitted to analyze the within field variation of yield response to cumulative precipitation across all possible cumulative precipitation periods. The optimal cumulative precipitation period was selected based on the Akaike Information Criterion. Spatially varying response parameters derived from the GAM with the optimal cumulative precipitation period were then used to classify the varying yield responses to precipitation within the field using a Classification and Regression Tree methodology. The most relevant outcomes of this study include: (i) the optimal period of cumulative precipitation impacting yield change by field and crop type; (ii) within a field, some areas exhibit greater yield variability than others; and (iii) areas with equal high yield variability within the field may respond differently to precipitation. Future work could include the utilization of the spatio-temporal variation within a field for guiding input recommendations.