Semi-empirical modeling of growth and yield of soybean cultivars under different irrigation levels

Abstract

Semi-empirical models are widely used to simulate crop growth and production variables. This study aimed to estimate plant length using the Logistic model, leaf area index (LAI) using the Peak Log-Normal model, and total dry matter (TDM) using the Sigmoidal model for two soybean cultivars subjected to different levels of replacement of the water evapotranspired by the crop, using field data obtained under the agrometeorological conditions of Rio Largo, Alagoas, Brazil. A randomized complete block design with split plots and four replications was used. Treatments consisted of fractions of crop evapotranspiration (35, 70, 105, 140, 175 and 210% of ETc) and two soybean cultivars. The six irrigation levels were assigned to the plots and the two cultivars to the subplots. Model accuracy was assessed using the root mean square error (RMSE), the index of agreement (d) of Willmott, the correlation coefficient (r) and the percentage deviation. In the treatments irrigated with 140, 175 and 210% of ETc, no water deficit was observed during the reproductive phase, because soil water storage fell below the readily available water only during the last 15, 13 and 11 days of the growing cycle, respectively, for both cultivars. The adjusted coefficients of determination for plant length ranged from 0.984 to 0.999, indicating that the model explained most of the variability of this variable. Pearson's correlation coefficient ranged from 0.906 to 0.986 and the maximum confidence index was 0.977, showing an excellent fit of the model to the LAI data. Irrigation depths significantly affected total dry matter, with a quadratic polynomial response. Irrigation depths did not impair the ability of the models to represent crop growth and yield.