Evaluation of Two Portable Hyperspectral-Sensor-Based Instruments to Predict Key Soil Properties in Canadian Soils.

Area of Science:

• Soil Science
• Analytical Chemistry
• Remote Sensing

Background:

• Classic bench-top spectrophotometers are limited for field soil analysis.
• Portable spectrophotometers offer rugged, cost-effective, and user-friendly alternatives for on-site soil property assessment.

Purpose of the Study:

• To evaluate two portable spectrophotometers for predicting soil texture (sand, clay) and soil organic carbon (SOC).
• To compare the performance of a prototype mid-infrared (mid-IR) spectrophotometer against a commercial visible-near-infrared (vis-NIR) spectrophotometer.

Main Methods:

• Diffuse reflectance spectroscopy (DRS) measurements were taken on 282 Canadian soil samples using both mid-IR and vis-NIR portable spectrophotometers.
• Partial least squares regression (PLSR) with leave-one-out cross-validation was employed to develop predictive models for soil properties.
• Calibration models were validated using 30% of the spectral data.

Main Results:

• The portable mid-IR spectrophotometer achieved high accuracy in predicting sand content (R²=0.82).
• The portable vis-NIR spectrophotometer demonstrated superior performance in predicting clay content (R²=0.82).
• Both instruments showed moderate accuracy for predicting soil organic carbon (SOC) (R²=0.54).

Conclusions:

• Portable mid-IR and vis-NIR spectrophotometers are effective tools for predicting soil texture in field conditions.
• The choice between mid-IR and vis-NIR depends on the specific soil property of interest (sand vs. clay).
• Further research is needed to improve SOC prediction accuracy with these portable devices.