Soil Nitrogen Content Detection Based on Near-Infrared Spectroscopy

Area of Science:

• Agricultural Science
• Analytical Chemistry
• Environmental Science

Background:

• Traditional soil nitrogen detection is slow and pollutes the environment.
• There is a need for rapid, non-polluting soil nitrogen analysis techniques.

Purpose of the Study:

• To develop a fast, easy-to-use, and environmentally friendly method for soil nitrogen detection.
• To utilize near-infrared (NIR) spectroscopy and random forest (RF) regression for accurate soil nitrogen content measurement.

Main Methods:

• Collected 143 soil samples from the Xunsi River area.
• Acquired NIR spectral data and determined soil nitrogen content using chemical and physical methods.
• Preprocessed spectral data, developed an RF regression model, and optimized it using ∆Gini, comparing it with BP and SVM models.

Main Results:

• The RF model achieved a prediction R²C of 0.921 and RMSEC of 0.115 on the modeling set, and R²P of 0.83 and RMSEP of 0.141 on the test set.
• The RF model exhibited higher prediction accuracy than Back Propagation (BP) and Support Vector Machine (SVM) models.
• Optimization using ∆Gini effectively extracted spectral information for soil nitrogen and reduced data redundancy.

Conclusions:

• Near-infrared spectroscopy combined with the random forest algorithm provides an accurate and efficient method for soil nitrogen detection.
• The developed RF model offers a significant improvement over traditional methods and other machine learning models like BP and SVM.
• The study successfully demonstrated the potential of NIR-RF for rapid, non-polluting, and precise soil nitrogen analysis.