Estimating soil organic matter Based on field spectroscopy data with boruta algorythm extructing optimal wavelengths number
Abstract
Measuring the spectral reflection of the soil surface has practical importance for solving many issues in soil monitoring and mapping. The aim of this paper was to test the Boruta algorithm capability to identify specific wavelengths that have the maximum relationship with soil organic matter content. The object of the study was an experimental agricultural field with an area of 5,437 hectares, located in the Vnukovo district, Moscow. During the field survey 81 spectra of the open soil surface were obtained, and 30 soil samples were selected for further laboratory analysis. Two algorithms were used for modeling: a random forest (RF) and an artificial neural network (NN). The artificial neural network showed higher accuracy (R2 = 0.96) compared to the random forest algorithm (R2 = 0.92). The analysis of the most significant wavelengths performed with the Boruta algorithm showed that 13 wavelengths can be used to predict the soil organic matter content without loss of accuracy. All of them were inthe visible range of spectrum, and the main set belonged to the beginning of the red region: 604–607 nm.Received: 01/26/2026
Accepted: 08/26/2026
Accepted date: 11/20/2026
Keywords: bare soil spectral reflection; random forest; Retisols
DOI: 10.55959/MSU0137-0944-17-2026-81-4-223-233
Available in the on-line version with: 18.11.2026
-
To cite this article:

This work is licensed under a Сreative Commons Atribiution - NonCommercial 4.0 International (CC BY-NC 4.0)

