Digital mapping of the soil available water capacity: insights for the resilience of agricultural systems to climate change

• Ano de defesa: 2023
• Autor(a) principal: Rico Gómez, Andrés Mauricio
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Aprendizagem de máquinas; Climate change; Ecosystem services; Machine learning; Mudanças climáticas; Remote sensing; Sensoriamento remoto; Serviços ecossistêmicos; Shapley value; Soil functions; Valor Shapley
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/11/11140/tde-03052023-104403/

Soil available water capacity (AWC) is a key function for human survival and well-being. However, its direct measurement is laborious and its spatial interpretation is complex. These difficulties have led to the use of indirect ways to estimate the AWC. Among them, digital soil mapping (DSM) techniques have emerged as an alternative to spatial modeling of soil properties. DSM techniques usually apply machine learning (ML) models, with a high level of complexity. In this context, we aimed to identify spatial patterns estimated by the Random Forest (RF) algorithm to predict AWC, and in a case study, to show that digital AWC maps can support agricultural planning in response to local climate change effects. To do this, a data-driven approach was applied using laboratory-determined soil attributes (clay, sand, and organic matter content), along with a pedotransfer function (PTF), remote sensing, DSM techniques, and meteorological data. The digital map of available soil water and weather station data were used to calculate climatological soil water balance for the periods 1917-1946 and 1991-2020. The selection of covariates contributed to the parsimony of the model, obtaining quality of fit metrics of R2 0.72, RMSE 16.72 mm m-1, CCC 0.83, and Bias of 0.53 on the validation set. The largest contributions to soil AWC prediction were multitemporal Landsat imagery with bare soil pixels, diurnal mean, and annual temperature variation. The present case study shows that climate change at the study site has modified the rainfall regime, increasing the amount of water retained in the soil during the dry period (from April to August). The methodology used provides parameters for the adaptation of agricultural systems to the effects of climate change.