Mapping peat layer properties with multi-coil offset electromagnetic induction and laser scanning elevation data

•Non-invasive mapping of peat properties by electromagnetic induction is feasible.•Performance was the best for dry bulk density and soil organic carbon (SOC) content.•SOC stock estimates were still reliable but less accurate.•Ancillary laser-scanning elevation data improved model performance.•Predicted spatial patterns showed substantial similarities to former land use.

Peatlands store large amounts of soil organic carbon (SOC). Depending on their present condition, they act as a source or sink of carbon dioxide. Therefore, peatlands are highly relevant for climate change investigations and there is considerable interest to assess spatial heterogeneity of peat soil properties in order to assess the total amount of stored carbon. However, reliable information about peat properties remains difficult to obtain at the field scale. A potential way to acquire this information is the indirect mapping of easily recordable physical variables that correlate with peat properties, such as the apparent electrical conductivity (ECa). In this study, we aim to explore the potential of multi-coil offset electromagnetic induction (EMI) measurements to provide spatial estimates of SOC content, bulk density, and SOC stock for a highly variable and disturbed peatland relict (~ 35 ha) with a remaining peat layer thickness of less than 1 m. EMI measurements comprised six integral depths that varied from 0-0.25 to 0-1.80 m. In combination with ancillary laser-scanning elevation data, a multiple linear regression model was calibrated to reference data from 34 soil cores that were used to calculate integral properties of the upper 0.25, 0.5, and 1 m layer, as well as for the total peat layer. Leave-one-out cross-validation for the different depth ranges resulted in a root mean square error of prediction (RMSEP) between 1.36 and 5.16% for SOC content, between 0.108 and 0.183 g cm− 3 for bulk density, and between 3.56 and 9.73 kg m− 2 for SOC stocks, which corresponds to roughly 15%, 10%, and 20% of the total field variability, respectively. The selection of explanatory variables in the regression models showed that the EMI data were important for accurate model predictions, while the topography-based variables mainly acted as noise suppressors. The accuracy of the SOC content estimates roughly equalled the quality of SOC content predictions obtained in previous field applications of the visible-near infrared technique (vis-NIR). The spatial variation of the predicted peat layer properties showed similarities to the former land use distribution. Overall, it was concluded that EMI measurements offer a useful alternative to the established vis-NIR method for SOC content mapping in carbon-rich soils.