Soybean yield monitoring system for Brazil with FAO-AEZ crop model

Resumo

Year-to-year soybean yield variability in Brazil affects the national economy and food supply. National yield prediction systems for soybeans are important to anticipate possible disruptions in the supply chain. This study aimed at developing a soybean yield monitoring system based on a crop model. Previously calibrated and validated FAO Agro-Ecological Zone (FAO-AEZ) crop model was used to simulate soybean yields for 59 agroclimatic homogenous zones distributed across the Brazilian territory. The spatial performance of FAO-AEZ to simulate soybean yield at homogenous zones level was computed by comparing simulated median soybean yield with observed soybean yield data from national records after applying a trend correction on the time series of 2011 to 2020. To monitor the soybean crop in-season performance, we computed the soybean crop yield anomaly representing the in-season yield variation in relation to five previous cropping seasons with the same sowing dates. The in-season analysis was limited to the most recent cropping seasons reported by the national bulletins (2021 and 2022). The FAO-AEZ crop model performance between 2011 and 2020 reached an r² of 0.59 and an RMSE of 402 kg ha^-1 when compared to historical national statistics. In turn, the in-season analysis for 2022 revealed the model’s capacity to anticipate signs of negative yield trends three months in advance compared to national bulletins. National yield estimations are possible at earlier months of the season further improving the prediction performance when approaching the end. FAO-AEZ crop model only uses commonly available inputs including daily weather data, simple crop management coefficients, and soil water holding capacity parameters, which makes this method easy to adapt to other custom needs such as crops, geographical extent, and prediction resolution.