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

Area of Science:

• Environmental Science
• Soil Science
• Remote Sensing

Background:

• Anthropogenic climate change necessitates robust carbon credit security and stable carbon sinks.
• Topsoil organic matter (SOM) is key for carbon sequestration but vulnerable to environmental factors and erosion.
• Effective topsoil management requires comprehensive data on environmental factors and composition, particularly in erosion-prone regions like the Republic of Korea.

Purpose of the Study:

• To investigate the spectral characteristics of topsoil in the Republic of Korea.
• To develop and evaluate a machine learning-based model for predicting SOM content using spectroscopic techniques.
• To identify key wavelengths associated with SOM for improved topsoil monitoring.

Main Methods:

• Collection of 138 topsoil samples from four major rivers in the Republic of Korea.
• Measurement of topsoil reflection spectra (350-2500 nm) using a spectroradiometer.
• Development and evaluation of partial least squares regression and support vector regression models for SOM prediction.

Main Results:

• A machine learning model achieved a coefficient of determination (R²) of 0.706 for predicting SOM content in total topsoil.
• Identified specific wavelengths crucial for determining SOM content through spectroscopic analysis.
• Demonstrated the predictability of SOM content using spectroscopic technology.

Conclusions:

• Spectroscopic techniques combined with machine learning offer a viable method for rapid and accurate SOM content prediction.
• The developed model and identified wavelengths can support the creation of a national topsoil database.
• Findings contribute to enhanced topsoil management, erosion mitigation, and carbon sink stability.