Exploring the Impact of Different Input Data Types on Soil Variable Estimation Using the ICRAF-ISRIC Global Soil...

Area of Science:

• Soil Science
• Remote Sensing
• Geospatial Analysis

Background:

• Accurate soil variable estimation is crucial for agriculture and environmental management.
• Traditional soil analysis is often time-consuming and expensive.
• Developing cost-effective, accurate soil assessment methods is a priority.

Purpose of the Study:

• To develop and compare models for estimating soil variables using different data inputs.
• To assess the accuracy of soil variable estimation using site data, visible-near-infrared (Vis-NIR) spectroscopy, and red-green-blue (RGB) color data.
• To determine the cost-effectiveness of different data input combinations for soil analysis.

Main Methods:

• Utilized the ICRAF-ISRIC global soil spectroscopy database.
• Developed predictive models using five input data types: site data only, Vis-NIR spectroscopy only, combined site and Vis-NIR data, RGB color data only, and combined site and RGB data.
• Evaluated model performance based on estimation accuracy (r² values) for various soil variables.

Main Results:

• Models combining spectroscopy and site data yielded the highest estimation accuracy for most soil variables.
• Combined site and RGB data provided useful estimates (r² > 0.7) for major elements (Ca, Si, Al, Fe), organic carbon, and cation exchange capacity.
• Spectroscopy plus site data achieved high accuracy (r² > 0.9) for soil health indicators like cation sum, electrical conductivity, and various oxides.

Conclusions:

• Visible-near-infrared spectroscopy combined with site data is highly effective for detailed soil characterization.
• Red-green-blue color data, when combined with site information, offers a viable alternative for estimating certain essential soil properties.
• The study provides a framework for selecting cost-effective data input strategies for soil variable estimation based on required accuracy levels.