Soil Nutrients Inversion in Open-Pit Coal Mine Reclamation Area of Loess Plateau, China: A Study Based on ZhuHai-1 Hyperspectral Remote Sensing

Abstract

Soil nutrients are crucial to assess land reclamation quality, and the use of various types of remote sensing data for soil nutrient inversion has been a key focus for soil monitoring. However, fewer studies have been conducted using satellite-based hyperspectral remote sensing. To explore the potential of satellite-based hyperspectral remote sensing in soil nutrient monitoring, this study selected soil organic matter, total nitrogen, available phosphorus, and available potassium content data from 83 sample sites using ZhuHai-1 hyperspectral data. After spectral transformation and feature extraction, various inversion models were constructed, including partial least squares regression, support vector machine, recurrent neural network, and random forest. After verification by accuracy, the best spectral-model combination was used for inversion. The results showed that the R-squared range of the inversion models was 0.67748–0.78115. High content areas of soil organic matter and available potassium exhibited concentrated and contiguous features, while high content areas of total nitrogen and available phosphorus were more fragmented and fine-grained. Alfalfa grassland plays a vital role in improving reconstructed soil in the early reclamation stage, and agricultural activities have differential impacts on soil nutrient accumulation. This study provides a theoretical basis for verifying the application capability of ZhuHai-1 hyperspectral satellite data in soil monitoring.