Towards Mapping of Soil Crust Using Multispectral Imaging

Soil crusts and surface roughness are properties which are highly dynamic in both space and time that change in response to biotic processes, meteorological conditions and farming operations. These factors, however, are difficult to quantify and are usually described using simplified expert-based classes. This hampers a clear identification of the controlling factors and their relation to soil erosion and sediment generation processes. The availability of new small portable multispectral cameras offers the potential to study soil surface dynamics at a high spatial and temporal resolution. The objective of this study was to analyse the relationship between soil crusting, represented by cumulative rainfall kinetic energy, and soil surface reflectance, as derived from vis-NIR multispectral imaging. We designed a series of rainfall-soil surface experiments to disentangle the effects of soil crusting on spectral reflectance factors from those related to surface micro-scale roughness. Partial least squared regression (PLSR) models were developed to predict both kinetic energy and roughness from multispectral images. We evaluated different roughness removal methods which were based on the transformation of reflectance through standard normal variate (SNV) and roughness thresholding using high resolution digital elevation models. Furthermore, we assigned the light scattering effect related to roughness in the multispectral spatial domain by calculating the inter-quantile range of the reflectance values in a kernel. Our experiments and workflow demonstrate that it is possible to model crust development, using rainfall kinetic energy as a proxy, from vis-NIR based multispectral imaging.