[Hyper spectral estimation method for soil alkali hydrolysable nitrogen content based on discrete wavelet transform...

Area of Science:

• Agricultural Science
• Soil Science
• Remote Sensing

Background:

• Accurate estimation of soil alkali hydrolysable nitrogen is crucial for sustainable agriculture.
• Hyperspectral remote sensing offers a non-destructive method for soil analysis.
• Traditional methods often involve complex sample processing and can be time-consuming.

Purpose of the Study:

• To develop and validate a novel method for quantitative estimation of soil alkali hydrolysable nitrogen content.
• To improve the efficiency and accuracy of soil nutrient assessment using hyperspectral data.
• To explore the effectiveness of data compression and variable selection techniques in improving soil analysis models.

Main Methods:

• Soil samples were collected from Qihe County, Shandong Province, China.
• Hyperspectral reflectance measurements were taken and processed using first deviation transformation.
• Discrete Wavelet Transform (DWT) was employed for spectral denoising and compression.
• Genetic Algorithms (GA) were used for selecting optimal variables.
• Partial Least Squares (PLS) regression was applied to build quantitative estimation models (DWT-GA-PLS).

Main Results:

• The DWT-GA-PLS method effectively compressed spectral variables and reduced model complexity.
• The developed models achieved high accuracy in estimating soil alkali hydrolysable nitrogen content.
• Models based on low-frequency coefficients of DWT showed comparable or superior prediction accuracy to full spectra.
• The best model, utilizing second-level low-frequency coefficients, achieved a predicting R2 of 0.85, RMSE of 8.11 mg x kg(-1), and RPD of 2.53.

Conclusions:

• The DWT-GA-PLS method is effective and accurate for estimating soil alkali hydrolysable nitrogen content.
• Data compression and variable selection significantly enhance the performance of hyperspectral-based soil nutrient models.
• This approach provides a promising tool for large-scale, efficient soil nutrient monitoring in precision agriculture.