Effectiveness of different approaches for in situ measurements of organic carbon using visible and near infrared spectrometry in the Poyang Lake basin area

Abstract

AbstractResearch on soil carbon dynamics and content in the Poyang Lake basin area is important because of their effects on the soil carbon pool and soil carbon cycle. Visible and near infrared (VNIR) diffuse reflectance spectroscopy can provide analytical dense soil data reflecting multiple physical and chemical properties of soil under in situ conditions. The effective use of in situ VNIR in the Poyang Lake basin area would greatly improve the collection of high‐resolution and spatially explicit soil data and, thus, hasten the monitoring process. We collected two data sets S1 and S2 in 2016–2018. A dataset from S1 was used to develop the in situ correction matrix using laboratory and in situ spectra. Spectra from the Chinese Soil Spectroscopic Database (CSSD) and dataset of S1 in dry laboratory conditions were used for calibration (CSSD‐local calibration). The dataset from S2 was utilized as the validation dataset. Four in situ correction methods, external parameter orthogonalization (EPO), piecewise direct standardization, direct standardization, and generalized least squares weighting, were used to remove the in situ effect on the spectra. In addition, partial least squares regression and support vector machine (SVM), were used to test the effectiveness of the prediction. We also compared the prediction results by calibration of the CSSD, CSSD‐local and CSSD combined with four in situ samples from the validation datasets with extra‐weighting (CSSD‐EX). The results showed that EPO was the most effective method for removing the in situ effect. EPO and SVM with the calibration of CSSD‐EX yielded the best prediction result with the lowest root mean squared error at 1.71 g kg−1 and the highest Lin's concordance correlation coefficient at 0.88. The CSSD with EPO and SVM would allow the in situ estimation of soil organic carbon in the Poyang Lake basin area at a reasonable cost.