Estimation of Soil Moisture by Deterministic Models in the Root Zone of Soybean

Abstract

The soil water content in the root zone during planting and crop development is a key point for the success of the agricultural harvest. Knowing this, we evaluated the performance of the GIOVANNI-NASA (GN) and water balance models proposed by Thornthwaite-Mather (TM), compared to the Decision Support System for Agrotechnology Transfer (DSSAT) standard, to estimate soil moisture in the root zone of soybean ( Glycine max ) in Brazil. We evaluated both models for 18 years (2001-2018) according to soybean planting seasons, for 64 locations, comprising the five most important regions of Brazil. The TM model was calibrated considering soybean crop requirements and soil conditions at each site. The GN model was obtained from public NASA-Power data, using only geographic coordinates of the locations. The results indicated that TM is able to estimate soil moisture in all regions of Brazil, and obtained high accuracy (RMSE= 0.06) and higher precision (R 2 = 0.76). Despite the GN model estimates the surface water content of a crop directly, this model only had a higher accuracy (R 2 = 0.75) for the southeast and midwest regions of Brazil (a region that represents the regions of dry winter and rainy summer in a tropical climate), despite of high precision for some locations, also had low accuracy for locations with high water content in the soil (MAPE > 31%). In contrast, we show for the first time that GN has great importance for regions of climatic extremes, with high values of water deficiency or surplus, mainly for the arid regions. Results indicated that the use of the climatological model proposed by Thornthwaite-Mather obtain high performance to estimate soil moisture in the root zone of soybean. However, we do not rule out the use of GIOVANNI-NASA to estimate soil moisture in agricultural crops, because of its ease of use and performance in tropical microclimates. Overall, these models calibrated and tested may be useful for planting decision makers and provide a new perspective for applying modeling to estimate soil moisture during the agriculture cycle.

Keywords: crop model, soil water balance, Thornthwaite-Mather, DSSAT, GIOVANNI-NASA, Python