Stochastic assessment of soil water balance components: development and application in a hardsetting soil scenario

• Autor(a) principal: Santos, Arthur Klebson Belarmino dos
• Data de Publicação: 2022
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/11/11140/tde-23032023-153059/
• topic: Amostragem de solo; Componentes do balanço hídrico; Condutividade hidráulica; Hydraulic conductivity; Método estocástico; Retenção de água no solo; Soil sampling; Soil water retention; Stochastic method; Water balance components

Hydrological modelling is an essential tool for understanding the processes that occur in the soil vadose zone. These processes are dependent on soil water retention parameters and hydraulic conductivity, which are normally determined using replicas of undisturbed soil samples. However, this method is not effective when trying to represent large areas, since an average of all replicas is usually performed. Based on this, we present a method to merge the hydraulic soil parameters of all replicas into a final set of data, with their associated statistics (standard errors and correlation matrix). To do so, we used VGM parameters obtained at sample scale in three replicas from a Brazilian savanna hardsetting soil from the eastern part of Maranhão state, Brazil through inverse modelling of laboratory evaporation experiments. The effectiveness and representativeness of the proposed methodology were evaluated by observing the frequency distribution of the output parameters, and comparing individual and merged sample properties (VGM parameters, retention, and hydraulic conductivity characteristics together with soil water balance components) stochastically predicted by a hydrological model. With the established method, a 31-year historical data set was analysed for three samples collected at three depths (0-15, 15-30, and 30-45 cm) in the hardsetting soil. The stochastic method allowed obtaining the variability of the combined replicas for the water balance components. Applying the technique to the 31 years, some generated VGM parameter sets, as well as the rainfall accumulation and distribution during the crop cycles, showed to be the determining factors for the dispersion of the simulated water balance components. Except for transpiration data, the other water balance components (bottom flux, evaporation, and runoff) showed a good correlation with the accumulated precipitation. A significant reduction in the dispersion of the transpiration rate was observed in high precipitation years. In general, using the mean hydraulic property parameter values to deterministically predict water balance components may yield values that are substantially different from the median values of stochastic realizations. This suggests that these values may generate unrepresentative results in hydrological modelling, showing the important role of stochastic analysis.