Modeling and assessing hydraulic properties of selected brazilian and australian soils

• Ano de defesa: 2024
• Autor(a) principal: Silva, Lucas de Castro Moreira da
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa; Engenharia Agrícola
• Palavras-chave em Português: Solos - Umidade - Medição; Solos - Umidade - Modelos matemáticos; Solos - Brasil; Solos - Austrália; Aprendizado do computador; Engenharia de Água e Solo
• Link de acesso: https://locus.ufv.br//handle/123456789/32312; https://doi.org/10.47328/ufvbbt.2024.098

Water availability is the main constraint for plants growth and current agricultural systems face the challenge of achieving high yields by optimizing the water use. The water dynamics in soils is governed by soil hydraulic properties and assessing these properties is essential for the efficient water use in agriculture. In this sense, several soil management practices have been employed to improve the soil water availability in agricultural fields, especially in water- limited regions. Thus, this study aimed to assess and predict soil hydraulic patterns in two different scenarios: Brazilian (First part) and Australian soils (Second part). In the first paper, a literature review was performed to bring to light what was done in ten years (2012-2021) regarding the prediction of soil hydraulic properties. In the second paper, machine learning models were developed to create regional pedotransfer functions for an important tropical agricultural center, the Mato Grosso state in Brazil. In the third paper, empirical models were tested for fitting water retention in Western Australian sandy soils modified by soil managements. In the fourth paper, popular practices employed to overcome sandy soils constraints were evaluated based on their effectiveness in enhance water availability. Results indicate that machine learning models are more accurate in predicting hydraulic properties compared to conventional methods. Regional-specific models were developed for soil hydraulic properties of Mato Grosso and are well calibrated for 91% of the state’s territory using basic predictors. However, additional predictors reduce their applicability. Brooks and Corey model showed the best performance and a consistent negligible bias in estimating soil water retention of Western Australian sandy soils. Adding subsoil clay significantly increased total porosity and microporosity of sandy soils but did not improve water availability. Keywords: Pedotransfer functions; Machine Learning; Soil hydraulic properties; Soil water retention; Available water