Using Landsat and soil clay content to map soil organic carbon of oxisols and Ultisols near Sao Paulo, Brazil

Autor(es)

Padilha, Manuela Correa de Castro ^[1] ; Vicente, Luiz Eduardo ^[2] ; Dematte, Jose A. M. ^[1] ; Loebmann, Daniel Gomes dos Santos Wendriner ^[2] ; Vicente, Andrea Koga ^[2] ; Salazar, Diego F. U. ^[1] ; Guimaraes, Clecia Cristina Barbosa ^[1]

Afiliação do(s) autor(es)

^[1] Univ Sao Paulo, Luiz de Queiroz Coll Agr, Ave Padua Dias, 11 Cx Postal 9, BR-13418900 Piracicaba, SP - Brazil

^[2] Embrapa Environm, Low Carbon Agr Platform, Cx Postal 69, BR-13820000 Jaguariuna, SP - Brazil

• Tipo de documento: Artigo Científico
• Fonte: GEODERMA REGIONAL; v. 21, JUN 2020.
• Citações Web of Science: 0

Resumo

Quantification of soil organic carbon (SOC) is a low-cost and necessary practice to meet increasing agricultural demands. Studies show that remote sensing (RS) is important for SOC prediction and its use has become crucial in agricultural management. In this study, a Multiple Linear Regression (MLR) model was constructed to predict SOC in a site in Piracicaba, Sao Paulo, Brazil. As predictor variables, we used the optical-satellite data of OLI/Landsat-8 sensor (bands 5 and 7, specifically), clay concentration, and the Normalized Difference Vegetation Index (NDVI). We collected 218 samples at the sampling points in the field to quantify clay and SOC in the laboratory as a calibration procedure. An Exposed SoilMask (ESM) was created using the method GEOS3 technology, which showed pixels with greater variability of bare soil. The pixels were evaluated with their respective surface reflectance values obtained by the satellite sensor and their respective NDVI index values. We evaluated the model predictive performance based on the adjusted coefficient of determination (R-2), the Root Mean-Squared Error (RMSE), and the Ratio of Performance to Interquartile Range (RPIQ) obtained in data validation. The MLR model presented R-2 values 0.79 and 0.81 for calibration and validation, respectively. We obtained important RMSE and RPIQ values, 0.14 and 2.32, respectively. The high RPIQ indicated significative sampling distribution around the trendline. After construction, the model was applied to the C spatial distribution using the predictive variables as layers, predominant concentrations of 0.65 to 0.79 g. Kg(-1) in 51 (23.4%) soil samples. The analysis presented here offer possibilities for SOC prediction using Geographic Information Systems (GIS) tools. (C) 2020 Elsevier B.V. All rights reserved. (AU)

• Processo FAPESP: 14/22262-0 - Geotecnologias no mapeamento digital pedológico detalhado e biblioteca espectral de solos do Brasil: desenvolvimento e aplicações
• Beneficiário: José Alexandre Melo Demattê
• Modalidade de apoio: Auxílio à Pesquisa - Temático