Digital mapping of soil attributes and functions: from farm to global scale

• Ano de defesa: 2025
• Autor(a) principal: Rosin, Nícolas Augusto
• 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: Fluorescência de raios X; Pedometria; Pedometrics; Remote sensing; Saúde do solo; Segurança do solo; Sensoriamento remoto; Soil health; Soil security; X-ray fluorescence
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/11/11140/tde-05112025-103640/

Producing highly detailed and accurate soil information in space and time are crucial for human well-being. In this way, the present thesis was divided into four chapters that aimed 1) To use soil point X-ray fluorescence (XRF) information to generate maps using remote sensing data to obtain valuable soil chemical element spatial data; 2) To test three strategies for soil organic carbon (SOC) stock spatial-temporal mapping at farm scale: a) a model for each year, b) a model fitted using only the data of one year and c) a multitemporal model; 3) To perform a high-resolution (30 m) spatial-temporal SOC stock mapping and evaluate changes in SOC stocks at different soil depths in the Brazilian territory, by biomes, states, and land use; 4) To assess the relationship between crop yield and soil attributes (in surface and subsoil layers) across the worlds ecoregions and to map the soil capacity to produce food and biomass worldwide using the soil security assessment framework (SSAF). To reach the first aim we measured the total concentrations of key chemical elements in the soil, such as Al, Si, Ti, and Fe using XRF. These data were then spatialized by digital soil mapping (DSM). The maps of XRF elements aligned well with existing pedological, and geological maps. The integration of spatialized XRF data with DSM and remote sensing techniques shows significant promise in assessing soil variations. To achieve the second aim, we proposed a 4D SOC stock mapping, comprehending space (two dimensions), plus depth and time dimensions. A grid sampling in 1997 and 2022 was conducted in a farm. Dynamic and static covariates, representing the soil formations were used to fit Cubist models to test the strategies. The strategy 1 and 3 produced more accurate and less biased maps, while the strategy 2 was not so efficient. To reach the aim 3, we utilized a combination of static and dynamic environmental covariates along with more than 50,000 observation points, achieving a SOC stock mapping at 30 meters resolution for Brazil. Over the 40 years, we have estimated an absolute gain of 0.80 Pg C in Brazil. Land use change from forest to anthropized uses was the primary drivers of SOC stock loss, whereas the conversion from pastures to croplands generally led to SOC gains. To reach the aim 4, the yield data of sugarcane, maize, rice, wheat and soybean were used as target indicator, while the clay, pH, SOC and plant available water were used as potential indicators in the SSAF. We obtained utility functions for each world ecoregion relating the yields with soil attributes. The final utility maps were predicted using DSM and allowed to downscale this information to 90 m and extrapolated the utility information for currently non-agricultural areas.