Skip to main content
Academic

High resolution 3D mapping for soil organic carbon assessment in a rural landscape

Digital Soil Assessments and Beyond

Abstract

Soil Organic Carbon (SOC) stocks were mapped at 2 m resolution in a complex agricultural landscape (NW France). Soil properties were described and measured for the 200 profiles sampled using a conditioned Latin hypercube sampling design, and were used to reconstruct continuous soil profiles via equal-area splines. Calibration data were extracted from these profiles for 8 standard layers up to a depth of 105 cm, and were used in a data mining tool (Cubist) to build rule-based models that predict SOC content and bulk density. Predictive environmental data consisted of natural gamma radiometric emissions, geological variables and topographic attributes (from a LIDAR DEM). The predictive maps were evaluated with two independent datasets, focusing on landscape scale or hedgerow proximity. The respective RMSE for these datasets were 7.47 and 4.77 g kg-1 for SOC content, 0.11 and 0.21 g cm-3 for bulk density (BD). The best predictions were obtained for depths between 15 and 60 cm. The SOC stocks below 30 cm accounted for an average of 33% of the total SOC stocks.

Author(s): Lacoste, M; Michot, D; Viaud, V; Walter, C; Minasny, B; McBratney, AB

Journal: Digital Soil Assessments and Beyond

Year: 2012

Comments