Simulação do balanço hídrico e rendimento de grãos da soja em condições irrigada e de sequeiro no Rio Grande do Sul

Abstract

Central pivot sprinkler is the most widespread irrigation method among medium and large farms in southern Brazil. Understanding the water dynamics in this context is difficult and expensive, however, using modern tools such as modeling, irrigation water management (rain + irrigation), can be optimized at the level of farmers. Therefore, the objective of the resent study was to simulate the soil water dynamic and its relation with crop grain yield in an irrigated and rainfed area, as well as to evaluate water productivity (WP), identify the yield response factor (Ky) and to use Stewart's empirical model for predicting yield. We used the SIMDualKc model, which calculates crop evapotranspiration (ETc) using the dual culture coefficient (dual Kc) methodology, separating soil water evaporation (Es) from crop transpiration (Tc). The experiment was conducted in the municipality of Júlio de Castilhos - RS (2017/18 season), in an irrigated area under a central pivot of 29 ha and a rainfed area of 7.7 ha. The water content in the soil was monitored in both areas, using watermark, installed in the intermediate layer of 0-20, 20-40, 40-60 cm depth. The physical-water characterization of the local soil was performed prior to the installation of the experiment. In addition, periodic morphological evaluations of the culture were performed. The meteorological information was collected from an automatic meteorological station, belonging to the National Institute of Meteorology (INMET), located approximately 30 km from the property, and irrigation monitoring was performed by the Irrig System®. The observed data were used to calibrate and validate the soil water balance model. The initial, mean and final Kcb values adjusted for the local conditions were, respectively, 0.10; 1.07 and 0.35. The observed soil water content (ASW) was compared to the simulated by the model using statistical indicators (b0: regression coefficient; R2: determination coefficient; RMSE: root mean square error; AAE: average absolute error; ARE: average relative error; EF: modeling efficiency; and PBIAS: percentage of bias). The results showed a good association between the observed values of available soil water (ASW) and those simulated by the model, with regression coefficient (b0 = 1) in both areas, and RMSE in relation to irrigated and rainfed area was 4.43 and 5.05 mm, respectively. The water use efficiency was low, as a result of the high rainfall during the crop development cycle, and a mean ETc of 534 mm (irrigated area) and 517 mm (rainfed area), in addition to the losses by deep percolation and surface runoff observed results of the soil water balance. The difference between the average grain yield between the areas was statistically similar. From the Stewart function, the yield response coefficient (Ky = 0.9) was generated to predict the grain yield of the soybean. The prediction of yield using the Stewart-SIMDulKc approach generated a RMSE of 1245 kg ha-1