Using Spectroradiometry to Measure Organic Carbon in Carbonate-Containing Soils

Area of Science:

• Soil Science
• Analytical Chemistry
• Remote Sensing

Background:

• Accurate quantification of soil organic carbon (SOC) is crucial for soil health and environmental monitoring.
• Carbonate-rich soils present analytical challenges for traditional SOC assessment methods.
• Visible near-infrared spectroscopy (VIS-NIR) offers a rapid and non-destructive approach for soil analysis.

Purpose of the Study:

• To evaluate the feasibility of using VIS-NIR spectroscopy for SOC analysis in carbonate-rich soils.
• To develop and compare various modeling pipelines for optimizing SOC prediction.
• To identify the most effective spectral preprocessing and modeling techniques for this soil type.

Main Methods:

• Development of 22 distinct modeling pipelines by combining different datasets (spectral data, spectral data with carbonate content).
• Utilized various feature groups: raw reflectance, first-derivative (FD), and second-derivative (SD) reflectance.
• Applied multiple variable selection methods (Spearman correlation, VIP, Rfrog) and regression models (PLSR, RFR, SVR).

Main Results:

• The study identified that the first-derivative (FD) spectral data preprocessing method, when combined with Random Forest (RF) regression, yielded the most robust and stable model.
• This specific pipeline demonstrated high potential for accurate SOC quantification in soils with high calcium carbonate content.
• Comparison of 22 pipelines highlighted the superiority of FD-RFR for carbonate-rich soil analysis.

Conclusions:

• Visible near-infrared spectroscopy is a viable technique for analyzing soil organic carbon in carbonate-rich soils.
• The combination of first-derivative spectral transformation and Random Forest regression provides a reliable and stable analytical model.
• This approach offers a promising alternative for efficient and accurate SOC assessment in challenging soil environments.