In Situ VIS-NIR Spectroscopy for a Basic and Rapid Soil Investigation

Area of Science:

• Soil Science
• Spectroscopy
• Geospatial Analysis

Background:

• Visible and near-infrared (VIS-NIR) spectroscopy is widely used in soil science for property prediction, primarily under laboratory conditions.
• In situ measurements often require contact probes and time-consuming methods, yielding spectra that differ from remotely acquired data.

Purpose of the Study:

• To evaluate the effectiveness of direct VIS-NIR spectral measurements using a fiber optic or a 4° lens on bare soils for predicting soil properties.
• To develop and compare partial least-square (PLS) and support vector machine (SVM) regression models for soil carbon (C), nitrogen (N) content, and soil texture (sand, silt, clay).

Main Methods:

• Acquisition of reflectance spectra directly from bare, untouched soils using a fiber optic or a 4° lens.
• Development of prediction models for C, N content, and soil texture using PLS and SVM regression.
• Incorporation of soil moisture and temperature as auxiliary data to potentially improve model performance.

Main Results:

• Satisfactory prediction models were achieved for C content (R² = 0.57) and N content (R² = 0.53) after spectral pre-processing.
• Model performance showed improvement when auxiliary data like moisture and temperature were included.
• Generated maps of C, N, and clay content using both laboratory and predicted values.

Conclusions:

• VIS-NIR spectra acquired with a bare fiber optic or a 4° lens are suitable for building prediction models for preliminary soil composition information at the field scale.
• The developed prediction maps are valuable for rapid, albeit approximate, field screening of soil properties.