The global mining sector generates billions of tons of tailings stored in thousands of tailing ponds. The occasional spills of tailings resulting from dam failures or pipe damage can have devastating consequences, threatening nearby human populations and ecosystems, particularly those located along river corridors. Satellite remote sensing technology is a vital supplementary method to traditional field methods for monitoring and evaluating the water pollution caused by spilled tailings. The researchers have developed workflows to evaluate the effect of tailing spills on water quality using low and medium-spatial satellite imagery from satellite sensors. Due to insufficient spatial resolution, these workflows were hard to apply to monitor the water pollution caused by spilled tailing in small rivers. Using valuable on-site data from a river water pollution incident caused by spilled tailing, a workflow utilizing high-resolution satellite imagery (GF1) was developed. This workflow incorporates a machine learning algorithm (improved DeepLabV3+) to extract water masks first and a novel spectral index method to determine TSM concentrations. The improved DeepLabV3+ algorithm can obtain an accurate water mask no matter the water pixels, whether influenced by the tailing spills from GF1 imagery with IoU of 82.66%, Precision of 93.21%, Recall of 87.96%, and F1-score of 90.51%. A new spectral index combination algorithm that provides reliable TSM products for an extensive TSM magnitude range was presented to assess the level of water contamination. The strong correlation (R2 = 0.97) between in situ TSM and Mo concentrations suggests that the retrieved TSM products are suitable for assessing the water pollution caused by the spilled tailing. This workflow provides a method for monitoring and evaluating water pollution resulting from spilled tailings in small rivers. It utilizes high-resolution satellite data to observe and analyze the pollution levels.
Read full abstract