Building a Vis-NIR spectral database for rapid determination organic carbon and nitrogen in forest soils of the southern Calabria (Italy)

Forest soils play an important role in the carbon stock; in particular, the evaluation of soil organic carbon (SOC) content is an important step for carbon sequestration studies. The project LIFE09 ENV/IT/078 Managing forests for multiple purposes: carbon, biodiversity and socio-economic wellbeing (ManFor C.BD.), among other objectives, is aimed to evaluate carbon sequestration taking into account forest management for conserving and enhancing carbon stocks, and increase carbon sequestration.Conventional laboratory analyses for the determination of soil properties as organic carbon (OC) and nitrogen (N) are expensive and time-consuming. Visible-near infrared (Vis-NIR) spectroscopy in combination with chemometrics techniques is claimed to be a rapid, cost-effective and non-destructive method for measuring soil properties.The objective of this study was to build Vis-NIR soil spectra database and develop prediction models for OC and N in a representative forest area of the Biogenetic Nature Reserve "Marchesale" located in Calabria region (south Italy).A set of 265 soil samples were collected within the study area. Soil samples were air dried, sieved at 2 mm and analyzed to estimate OC and N content. Subsequently, the Vis-NIR reflectance of each soil sample was measured in laboratory, using an ASD FieldSpec IV 350-2500 nm spectroradiometer.In order to develop models based on soil spectra and reference laboratory data of OC and N, partial least squares regression (PLSR) was used.To evaluate the accuracy of the PLSR models, the dataset was randomly separated into two subsets: calibration set (70%, n=180) for developing the prediction model and validation set (30%, n=80) to test the models accuracy. Several calibration models were built and compared by cross-validation. The predictive ability of the cross-validation models was evaluated by the coefficient of determination (R2) and the root mean square error (RMSE) of calibration.Results revealed a high level of agreement between measured and predicted values with high R2 and low RMSE values. The best calibration model obtained for OC show a R2 of 0.91 and a RMSE of 0.56% while for N, the R2 was equal to 0.81 and the RMSE to 0.50%.Good results of validation were obtained for both OC (R2=0.87 and RMSE=0.76%) and N (R2=0.79 and RMSE=0.76 %).The results indicate that Vis-NIR spectroscopy is a reliable alternative technique to determine OC and N in forest soils.Finally, the spectral database reported in this study could be used to support soil survey in other areas of the Calabria region.