Variability of soil hydraulic properties and its impact on agro-hydrological model predictions

• Ano de defesa: 2019
• Autor(a) principal: Oliveira, Thalita Campos
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Análise de incerteza; Inverse modelling; Modelagem inversa; Modelo SWAP; Propriedades hidráulicas do solo; Realização estocástica; Soil hydraulic properties; Stochastic realization; SWAP model; Uncertainty analysis
• Link de acesso: http://www.teses.usp.br/teses/disponiveis/64/64134/tde-08102019-172326/

Agro-hydrological models have been widely used to predict and simulate soil water balance components and crop yield with reliable results. These models provide detailed water and energy balances and enables simulating scenarios with distinct land management strategies, environmental and climate conditions. However, they require many input parameters, especially those related to soil water retention and hydraulic conductivity functions. These input parameters are prone to variation due to the determination methods, related errors and uncertainties, and soil variability. In this thesis we aimed to (1) analyze the suitability of inverse modelling as an alternative to traditional methods to estimate soil hydraulic properties using water content data obtained with Frequency Domain Reflectometry (FDR) sensors in a field experiment; (2) analyze the influence of the Mualem-van Genuchten parameters (M VG) uncertainty on water balance components and crop yield predicted by the SWAP model for a soil under rainfed maize crop by uncertainty analysis using two sampling methods. One method used Monte Carlo Random Sampling from normal distribution based on standard errors of the hydraulic parameters obtained from inverse modelling (MCRS), and the other used Monte Carlo Latin Hypercube Sampling (MCLHS). Our results from the inverse modelling showed that n and Ks from both horizons, and ?r from the Bt horizon, were estimated with low accuracy. Low values of field water contents in the A horizon led to a lower estimate of ?r compared to the laboratory method. In the Bt horizon, the small observed range of field water contents contributed to an unreliable estimation of parameters ?r and n. The MCRS and MCLHS sampling methods provided distinct ranges and probability density distributions shape for n parameter, and simulates runoff, soil evaporation and bottom flux. The M VG parameters from MCRS may enhanced the uncertainty of simulated results, whereas MCLHS provided more reliable M VG parameters combinations, and therefore, simulated results. The uncertainty analysis may provide useful information about the uncertainties of model SWAP predictions and should be preferred over a mere deterministic approach, which often provided results diverging those obtained from probabilistic methods. Moreover, the uncertainty analysis is a key tool for more reliable interpretation of the water balance and crop yield in agro hydrological systems and should be considered in agro modelling studies