Mapping within-field soil drainage using remote sensing, DEM and apparent soil electrical conductivity

In this study, we evaluated the capability of different datasets for soil drainage mapping within agricultural fields. The evaluated datasets include apparent soil electrical conductivity (ECa), remotely sensed high-resolution airborne hyperspectral reflectance (HR) and C-band synthetic aperture radar (SAR) backscattering coefficients, and a high precision digital elevation model (DEM) generated from GPS measurements. The study site was located in an experimental farm in Ottawa, Ontario, Canada. Three drainage classes representing moderately well drained, imperfectly drained, and poorly drained soils were identified during field surveys according to soil surveyor expert knowledge. Variables that significantly contributed to soil drainage classification were selected from the evaluated datasets with a stepwise discriminant analysis procedure. The selected variables were then used to classify soil drainage with a maximum likelihood classifier. A substantial agreement between the observed and classified drainage classes was achieved using the HR dataset, with a kappa coefficient (κ) of 0.68. Moderate agreement was achieved using the SAR and the ECa datasets, with κ = 0.52 and 0.55, respectively. The result obtained using the DEM-derived topographic variables showed only a fair agreement (κ = 0.31). Canonical analysis was also conducted to investigate the association between these datasets and field-observed soil water regime descriptors. This potentially provides an alternative way of drainage mapping using canonical variate. The canonical correlation between the water regime descriptors and the evaluated datasets was 0.81, 0.75 and 0.83 for the HR, SAR and soil ECa datasets, respectively. In this study, the topographic variables were not as efficient, but when combined with the SAR and soil ECa datasets, they improved soil drainage mapping.