Combining Field and Imaging Spectroscopy to Map Soil Organic Carbon in a Semiarid Environment

Abstract

Semiarid regions are especially vulnerable to climate change and human-induced land-use changes and are of major importance in the context of necessary carbon sequestration and ongoing land degradation. Topsoil properties, such as soil carbon content, provide valuable indicators to these processes, and can be mapped using imaging spectroscopy (IS). In semiarid regions, this poses difficulties because models are needed that can cope with varying land surface and soil conditions, consider a partial vegetation coverage, and deal with usually low soil organic carbon (SOC) contents. We present an approach that aims at addressing these difficulties by using a combination of field and IS to map SOC in an extensively used semiarid ecosystem. In hyperspectral imagery of the HyMap sensor, the influence of nonsoil materials, i.e., vegetation, on the spectral signature of soil dominated image pixels was reduced and a residual soil signature was calculated. The proposed approach allowed this procedure up to a vegetation coverage of 40% clearly extending the mapping capability. SOC quantities are predicted by applying a spectral feature-based SOC prediction model to image data of residual soil spectra. With this approach, we could significantly increase the spatial extent for which SOC could be predicted with a minimal influence of a vegetation signal compared to previous approaches where the considered area was limited to a maximum of, e.g., 10% vegetation coverage. As a regional example, the approach was applied to a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">${320\;\text{km}}^{2}$</tex-math> </inline-formula> area in the Albany Thicket Biome, South Africa, where land cover and land-use changes have occurred due to decades of unsustainable land management. In the generated maps, spatial SOC patterns were interpreted and linked to geomorphic features and land surface processes, i.e., areas of soil erosion. It was found that the chosen approach supported the extraction of soil-related spectral image information in the semiarid region with highly varying land cover. However, the quantitative prediction of SOC contents revealed a lack in absolute accuracy.