Estimation of soil organic carbon from airborne hyperspectral thermal infrared data: a case study

Airborne hyperspectral imagery in the thermal infrared region (8.0-11.5 mu m) was acquired with a thermal airborne spectrographic imager (TASI-600) on bare soils in agricultural fields in Pontecagnano (southern Italy). These data were related to ground sampling in order to assess the capability of this technology to predict topsoil properties at the field scale. Emissivity spectra were used to calibrate prediction models for the prediction of clay, sand and soil organic carbon (SOC) by partial least squares regression (PLSR) and Cubist regression techniques. The TASI-600 predictive models were validated by leave-one-out cross-validation. The results were compared with those obtained under laboratory conditions using both Fourier transform infrared (FT-IR) and visible and near-infrared (VNIR) spectroscopy. The accuracy of the predictive models was assessed in terms of R-2, root mean square error (RMSE), the ratio of the performance to deviation (RPD) and the ratio of performance to interquartile range (RPIQ). In the laboratory, results obtained from the Fourier transform infrared were better than those obtained from visible and near-infrared (VNIR) for the prediction of specific soil characteristics (sand, clay and SOC). The use of airborne data resulted in less accurate predictions than the FT-IR data obtained in the laboratory (resampled to the TASI-600 spectral characteristics). For TASI-600 airborne data, SOC was predicted more accurately (RPIQ = 1.96; RMSE = 0.26%) than clay and sand content. The results obtained in this study demonstrate the good potential of long-wave infrared (LWIR) remote sensing data for the quantitative estimation of the SOC content of topsoil.