Soil Properties Mapping Using Proximal and Remote Sensing as Covariate

Abstract

Obtaining knowledge about the distribution of spatial variability of soil properties is crucial to the proper site-specific management. One way to improve the quality of soil mapping is by using auxiliary information (covariate). The objective of this study was to test whether remote and proximal sensing data can assist in soil properties mapping through geostatistical prediction. We worked in an experimental area cultivated with sugarcane located in Sao Paulo State, Brazil, and selected five soil properties: organic matter, CEC, base saturation, K and P availability. Two covariates often used to express soil variation were chosen, one obtained by remote sensing (SWIR2 band) and the other by proximal sensing (apparent soil electrical conductivity -ECa). These covariates were individually and together used in geostatistical interpolation method (kriging with external drift). We found that ECa is a more promising covariate than SWIR2 band from orbital imaging. Such proximal sensing can identify the soil short-range spatial variability. However, when the soil property variability is well explained by the sampling procedure, multivariate geostatistical methods may not improve the mapping accuracy.