Estimation of clay content by magnetic susceptibility in tropical soils using linear and nonlinear models

Author(s)

Filla, Vinicius Augusto ^[1] ; Coelho, Anderson Prates ^[1] ; Ferroni, Adrien Dorvalino ^{[1, 2]} ; Rabelo de Souza Bahia, Angelica Santos ^[1] ; Marques, Jr., Jose ^[1]

Affiliation

^[1] Sao Paulo State Univ Unesp, Dept Agr Sci, Res Grp CSME Soil Characterizat Specif Management, Jaboticabal, SP - Brazil

^[2] Aparecida S-N, Novo Horizonte, SP - Brazil

• Document type: Journal article
• Source: Geoderma; v. 403, DEC 1 2021.
• Web of Science Citations: 0

Abstract

The application of pedotransfer functions for soil attribute estimation has economic and environmental advantages. However, the appropriate choice of models is essential to obtain high accuracy. The aim was of this study was to evaluate the magnetic susceptibility (chi) potential to estimate soil clay content and to compare the estimation accuracy using linear (simple linear regression, SLR) and nonlinear models (power model, PM, and artificial neural networks, ANNs). An area of approximately 870 ha, cultivated with sugarcane in southeastern Brazil, was delimited and georeferenced for the collection of soil samples. Soil samples were collected every 2.5 ha (0.00-0.25 m), resulting in 372 points. A transect was directed in the area, and another 132 soil samples were collected at regular intervals of 30 m in the same layer. The samples were sent to the laboratory, and chi values were obtained for all points. Pedotransfer functions were developed to estimate the soil clay content using the chi values. The models were calibrated with the 372 points collected from the regular mesh of the 870-ha area, using the 132 points of the transect for validation. Regardless of the type of model applied to the pedotransfer functions, chi proved highly accurate for estimating soil clay content. However, the ANNs and PM increased the estimation accuracy, attenuated the error, and increased the estimation precision for soil samples with clay content below 200 g kg(-1). (AU)

• FAPESP's process: 13/17552-6 - Diffuse reflectance spectroscopy and magnetic susceptibility in prediction and mapping attributes of soil in different of landscape compartments.
• Grantee: Angélica Santos Rabelo de Souza Bahia
• Support Opportunities: Scholarships in Brazil - Doctorate
• FAPESP's process: 13/13978-9 - Iron oxides and clay emission of CO2 oxisols under cultivation of sugar cane
• Grantee: Vinicius Augusto Filla
• Support Opportunities: Scholarships in Brazil - Scientific Initiation