Using hyper-spectral indices to detect soil phosphorus concentration for various land use patterns.

Area of Science:

• Environmental Science
• Soil Science
• Remote Sensing

Background:

• Accurate soil phosphorus data is crucial for managing nonpoint source pollution.
• Remotely sensed data offers potential for soil phosphorus assessment, though direct spectral features are unclear.
• Previous research indicates indirect detection via Normalized Difference Vegetation Index (NDVI) is feasible.

Purpose of the Study:

• To estimate total phosphorus and Olsen-P concentrations using an optimized index in RED and near-infrared (NIR) wavelengths.
• To evaluate the prediction accuracy of soil phosphorus across different land use patterns.
• To investigate the influence of land use on the effectiveness of remote sensing for soil phosphorus estimation.

Main Methods:

• Utilized an optimized spectral index in RED and NIR wavelengths for phosphorus estimation.
• Applied Normalized Difference Vegetation Index (NDVI) for indirect phosphorus detection.
• Analyzed prediction accuracy on mixed and categorized land use datasets.

Main Results:

• Prediction accuracy for mixed land use was moderate (R² ≈ 0.6), with poor performance for forests.
• Accuracy significantly improved when samples were separated by land use, exceeding R² = 0.8 for tea plantations.
• Variability in soil phosphorus prediction was linked to vegetation health, soil organic matter, and soil composition.

Conclusions:

• Optimized spectral indices combined with land use categorization enhance soil phosphorus prediction accuracy.
• Remote sensing, particularly NDVI-based approaches, shows promise for soil phosphorus assessment, but land use context is critical.
• Factors like vegetation condition and soil properties influence the efficacy of remote sensing for soil phosphorus mapping.