Soil degradation determined by temporal satellite images and environmental variables in São Paulo State, Brazil
Soil health is a major challenge in the 21st Century. Tropical regions are the ones with the strongest expansion in agricultural lands. Therefore, novel researches on the soil degradation process are imperative to prevent damages to social and environmental dynamics. The main goal of this research w...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Estado: | Versión publicada |
| Fecha de publicación: | 2022 |
| País: | Brasil |
| Institución: | Universidade de São Paulo (USP) |
| Repositorio: | Biblioteca Digital de Teses e Dissertações da USP |
| Idioma: | inglés |
| OAI Identifier: | oai:teses.usp.br:tde-11052022-145143 |
| Acceso en línea: | https://www.teses.usp.br/teses/disponiveis/11/11140/tde-11052022-145143/ |
| Access Level: | acceso abierto |
| Palabra clave: | Degradação do solo Digital soil mapping Imagens temporais de satélite Landsat Mapeamento digital do solo Monitoring Qualidade do solo Remote sensing Sensoriamento remoto Soil degradation Soil quality Time-series satellite images |
| Sumario: | Soil health is a major challenge in the 21st Century. Tropical regions are the ones with the strongest expansion in agricultural lands. Therefore, novel researches on the soil degradation process are imperative to prevent damages to social and environmental dynamics. The main goal of this research was to generate a Soil Degradation Index based environment co-variables acquired by remote sensing and processed with machine learning. The work was developed in all the complete agricultural area of São Paulo State, Brazil. We used a Landsat time-series (1985 to 2019) and determined the areas with exposed soil using the Geospatial Soil Sensing System methodology. Based on a dataset with soil samples (0-20 cm) we calibrated pixel images and generated thematic maps of clay, and cation exchangeable capacity, CEC. Organic matter was also determined but used for validation and not a co- variable. The specialization was performed using a random forest algorithm. Other co- variables were determined such as land use. The k-means clustering algorithm was used to overlay all variables including historical data of rainfall and surface temperature as well as terrain attributes in order to generate a Soil Degradation Index (SDI) (values from 1, very low to 5, very high levels of degradation). Finally, the model was validated using OM. There was an important relationship between the SDI and the spectral surface reflectance obtained by Landsat. Locations with less OM presented a higher degradation. Therefore, integrating multitemporal remote sensing data and environmental variables proved to be effective to assist the SDI, which allows for land use decision-making and public policies. |
|---|