Soil Moisture, Organic Carbon, and Nitrogen Content Prediction with Hyperspectral Data Using Regression Models

Area of Science:

• Environmental science
• Remote sensing
• Agricultural science

Background:

• Traditional soil analysis methods are time-consuming and limited in scope.
• Hyperspectral (HS) imaging and machine learning (ML) offer advanced solutions for large-scale soil property estimation.

Purpose of the Study:

• To develop and evaluate a generalized approach for predicting soil moisture, soil organic carbon, and nitrogen content.
• To assess the effectiveness of various ML models combined with feature selection and dimension reduction techniques.

Main Methods:

• Utilized three benchmark HS datasets from Germany and Sweden.
• Applied principal component analysis (PCA) for unsupervised dimensionality reduction.
• Evaluated multiple ML regression models with band selection and PCA.

Main Results:

• Selected HS bands significantly improved soil content prediction accuracy.
• PCA further enhanced ML model performance for predicting soil moisture, carbon, and nitrogen.
• Soil temperature did not improve soil moisture prediction accuracy when combined with HS bands.

Conclusions:

• The combined approach of band selection and PCA-driven feature transformation offers superior accuracy for soil property prediction.
• This method outperforms existing estimation techniques, providing a robust tool for agricultural monitoring.