Estimativa de produtividade da soja utilizando dados espectro-agroclimáticos de sensoriamento remoto no modelo WOFOST

Abstract

The soybean crop is currently the main crop produced in Brazil, being responsible for a significant part of the national economy and generating income for producers. The use of remote sensing techniques contributes to achieving this context, because through its utilization producers can improve the use of their resources, generating greater profitability. Thus, the objective of this work was to estimate soybean yield using the technique of assimilation of agrometeorological data with the World Food Studies (WOFOST) crop growth model, at the field level for areas located in the municipalities of Castro and Piraí do Sul, state of Paraná. For this purpose, the WOFOST model was used associated with leaf area index data, from vegetation index calculations using Sentinel-2 satellite images, and climate data obtained through the NasaPower platform. The results show that spatial and soybean yield changes occur over the years. When comparing the estimated yield with the field yield, values of coefficient of determination (R²) of 0.5 and 0.6, RMSE of 679.36 and 346.95 kg ha-1 were obtained for the municipalities of Castro-PR and Piraí do Sul-PR, respectively. The accuracy of the model was calculated using the improved index of Willmott (2012) and presented satisfactory results for both municipalities, while for the evaluation of performance [Pi] the municipality of Castro-PR (Dr: 0.523; Pi: 0.369) was classified as tolerable and Piraí do Sul-PR (Dr: 0.700; Pi: 0.544) as good. The use of the WOFOST model allowed to estimate soybean yield at the pixel level, for plots of varying sizes of areas, providing results that allow further studies.