Estimation of soil organic matter content based on CARS algorithm coupled with random forest

Area of Science:

• Agricultural Science
• Soil Science
• Spectroscopy

Background:

• Soil organic matter (SOM) is crucial for soil fertility, nutrient availability, and precision agriculture.
• Accurate and efficient estimation of SOM content is essential for effective farmland management.

Purpose of the Study:

• To develop a rapid and efficient method for estimating soil organic matter (SOM) content in farmland using Vis-NIR spectroscopy.
• To select optimal characteristic spectral variables and build a robust predictive model for SOM.

Main Methods:

• Collected 190 farmland soil samples and their corresponding Vis-NIR spectroscopy data.
• Employed the Competitive Adaptive Weighting Algorithm (CARS) for characteristic wavelength selection.
• Utilized Random Forest (RF) regression to establish and validate the SOM predictive model.

Main Results:

• The CARS algorithm identified optimal spectral variable subsets, enhancing model accuracy.
• The CR-RF model achieved high prediction accuracy, with R² of 0.96 and RPD of 3.02 on the validation set.
• The method demonstrated accurate prediction of SOM in the cultivated layer.

Conclusions:

• The combined CARS and RF approach provides an efficient and accurate method for estimating soil organic matter content.
• This technique supports precision agriculture by enabling rapid soil fertility assessment.
• The study successfully realized accurate prediction of SOM in the Jingbian County farmland.