Search

Soil organic carbon (SOC) is one of the key indicators of soil fertility, productivity and quality. Moreover, SOC variations are of critical importance for the ability of soils to sequester atmospheric CO2 derived from fossil fuel combustion and mitigateclimate change. Slight changes in the SOC pool could have a profound impact on the atmospheric CO2 concentration. Traditional methods for determining SOC content are expensive and time consuming. Soil reflectance spectroscopy is a rapid and cost-effective technique that can analyze many soil constituents simultaneously. The main objective of this study was investigated the feasibility of visible and near infrared (Vis-NIR) spectroscopy as a fast method for determining and monitoring the SOC content in a mountain area of Calabria region (southern Italy). In this study a multi-temporal data set including 280 soil samples was used: 210 soil samples were collected in 2012 whereas 70 in 2015. The soil samples were collected using a metallic core cylinder with a diameter of 7.5 cm and a height of 20 cm. Each sample was oven dried at 45° for 48 hours and sieved at 2 mm; soil samples were used for spectroscopic measurements and then analysed for SOC content. The Vis-NIR reflectance of soil samples was measured in laboratory, under artificial light, using an ASD FieldSpec IV 350 - 2500 nm spectroradiometer (Analytical Spectral Devices Inc., Boulder, Colorado, USA), while SOC was determined using a Shimadzu TOC-analyzer with a SSM-5000A solid sample module (Shimadzu Corporation, Kyoto, Japan). In order, to develop a prediction model based on soil spectra and reference laboratory data of SOC, the partial least squared regression (PLSR) analysis was applied. Spectra pre-treatment was carried out to minimize noise and optimize calibration accuracy before the calibration models development. A PLSR prediction model was developed using only the soil samples collected in 2012. The optimum number of factors to retain in the PLSR models was determined by cross validation. The prediction model was tested using soil data sampled in the 2015. Results revealed a high level of agreement between measured and predicted values with high coefficients of determination (R2) and low root mean square error (RMSE). These results indicated that Vis-NIR spectroscopy could be a promising method for determining and monitoring SOC content. In addition, a geostatistical approach could allow mapping and visualizing spatial and temporal variability of SOC.