Development of a Soil Organic Matter Content Prediction Model Based on Supervised Learning Using Vis-NIR/SWIR Spectroscopy

Abstract

In the current scenario of anthropogenic climate change, carbon credit security is becoming increasingly important worldwide. Topsoil is the terrestrial ecosystem component with the largest carbon sequestration capacity. Since soil organic matter (SOM), which is mostly composed of organic carbon, and can be affected by rainfall, cultivation, and pollutant inflow, predicting SOM content through regular monitoring is necessary to secure a stable carbon sink. In addition, topsoil in the Republic of Korea is vulnerable to erosion due to climate, topography, and natural and anthropogenic causes, which is also a serious issue worldwide. To mitigate topsoil erosion, establish an efficient topsoil management system, and maximize topsoil utilization, it is necessary to construct a database or gather data for the construction of a database of topsoil environmental factors and topsoil composition. Spectroscopic techniques have been used in recent studies to rapidly measure topsoil composition. In this study, we investigated the spectral characteristics of the topsoil from four major rivers in the Republic of Korea and developed a machine learning-based SOM content prediction model using spectroscopic techniques. A total of 138 topsoil samples were collected from the waterfront area and drinking water protection zone of each river. The reflection spectrum was measured under the condition of an exposure time of 136 ms using a spectroradiometer (Fieldspec4, ASD Inc., Alpharetta, GA, USA). The reflection spectrum was measured three times in wavelengths ranging from 350 to 2500 nm. To predict the SOM content, partial least squares regression and support vector regression were used. The performance of each model was evaluated through the coefficient of determination (R2) and root mean square error. The result of the SOM content prediction model for the total topsoil was R2 = 0.706. Our findings identified the important wavelength of SOM in topsoil using spectroscopic technology and confirmed the predictability of the SOM content. These results could be used for the construction of a national topsoil database.

Prediction of Soluble-Solid Content in Citrus Fruit Using Visible–Near-Infrared Hyperspectral Imaging Based on Effective-Wavelength Selection Algorithm

“…The actual SSC in the citrus fruits was compared with those predicted from the calibration (cross-validation) or independent validation datasets using the PLSR models. The performance of each model was evaluated by estimating the coefficient of determination of the calibration set (R c 2 ), the cross-validation set (R v 2 ), and the prediction set (R p 2 ), as well as root mean square error of calibration (RMSEC), cross-validation set (RMSEV), and prediction (RMSEP), in addition to the optimal factor (F) [ 18 , 27 ]. The model with the highest R v 2 and lowest RMSEV values was selected as the optimal model.…”

“…2 ), the cross-validation set (R v 2 ), and the prediction set (R p 2 ), as well as root mean square error of calibration (RMSEC), cross-validation set (RMSEV), and prediction (RMSEP), in addition to the optimal factor (F) [18,27]. The model with the highest R v 2 and lowest RMSEV values was selected as the optimal model.…”

Estimation of Total Nitrogen Content in Topsoil Based on Machine and Deep Learning Using Hyperspectral Imaging

“…Samples taken from five points in each area were mixed and homogenized as representative samples. The collected topsoil samples were brought to the laboratory, air-dried, and sieved through a 2 mm mesh to remove rock fragments and coarse roots [32].…”

“…It then evaluates each model against a calibration set. The optimal factors for PLSR were determined using leave-one-out cross-validation (LOOCV) as the validation method [24,30,32]. A plot of the LOOCV residual variance versus the number of LVs was tested to specify the optimal number of LVs (called optimal factors) for PLSR.…”

“…The regression coefficients (b coefficients) were applied to determine the effective wavelength band using the PLSR model. When the b coefficient exceeded the thresholds, which were set to the standard deviation of the values, the corresponding wavelength was considered important [32,38]. The PLSR model was generated using the Unscrambler X software (Ver.…”