矿区生态环境定量遥感监测研究进展与展望

矿区生态环境的科学有效监测是矿区生态环境保护与治理的前提，对于促进生态文明建设具有重要意义。遥感技术已经成为矿区生态环境监测的重要手段，特别是近年来遥感技术在数据、算法、算力方面的迅速发展极大地促进了国内外矿区生态环境定量遥感监测研究的进步，涌现出了一系列优秀的研究成果。从矿区地表要素类型识别以及矿区植被、土壤、水体、大气、生态系统的参数反演与监测方面归纳和剖析了矿区生态环境定量遥感监测的研究进展。结果表明：新兴遥感数据的应用提升了监测的时空分辨率，矿区地表要素识别以及参数反演方法得到了优化改进，提升了识别和反演的精度；深度学习和遥感云计算平台在矿区应用中崭露头角。但是也存在一些不足：①深度学习在矿区地表要素识别的应用尚未完全展开，缺乏矿区遥感地表分类体系标准与大规模高分样本库，矿区地表要素识别的自动化与智能化水平有待提高；②矿区定量遥感参数反演的广度有待拓展，反演与监测方法研究有待深入；③矿区多要素参数的中高分辨率、长时序、高频次同步观测与协同分析的研究还相对缺乏。在此基础上，对未来的研究方向进行了展望：①构建矿区遥感地表分类体系与大规模高分样本库，跟踪最前沿的深度学习算法，实现矿区地表典型要素的高精度识别；②研究矿区场景下的定量遥感物理机理建模方法，拓展遥感反演的矿区要素参数，提升反演方法的精度和稳定性；③融合矿区多源大数据，实现参数的中高分辨率、长时序、高频次的体系化同步定量遥感反演与监测。

Progress and Prospect of the Quantitative Remote Sensing for Monitoring the Eco-environment in Mining Area

Scientific and effective monitoring of the eco-environment in mining areas is the prerequisite for the protection and governance，which is of great significance to promote the ecological civilization construction.Remote sensing technology has become an important tool for monitoring the eco-environment in mining areas.Especially in recent years，the rapid development of remote sensing technology in data，algorithms，and computing power has greatly promoted the development of quantitative remote sensing of eco-environment in mining areas，which leads to a series of excellent research results.This paper summarizes and analyzes the progress of quantitative remote sensing for monitoring the eco environment in mining areas from the aspects of identifying the surface types，retrieval of the parameters about vegetation，soil，water，atmosphere，and ecosystem in mining areas.The results show that the application of new remotely sensed data has improved the temporal and spatial resolution;the methods for identifying surface types and retrieving parameters in mining areas have been optimized，and the accuracy of identification and inversion has been improved;deep learning and remote sensing cloud computing platforms are preliminarily used in mining areas.However，there are also some shortcomings:① The application of deep learning in the identification of surface types in mining areas has not yet been fully developed，and there is a lack of remote sensing surface classification system standards and large-scale high resolution sample databases in mining areas.The level of automation and intelligence for surface types identification in mining areas needs to be improved;② The breadth of the retrieval of parameters using quantitative remote sensing needs to be expanded，and the methods of retrieving parameters need to be deepened;③ The research on the moderate-high resolution，long time series，high-frequency synchronous observation and collaborative analysis of multi-parameters in the mining area is relatively lacking.On this basis，prospects for future directions are listed as follows:① A remote sensing surface classification system and a large-scale high-resolution sample database for mining areas should be built，and the state of art deep learning algorithms should be tracked to achieve high precision identification of typical surface types in mining areas;② Quantitative remote sensing methods for physical mechanism modeling should be conducted in the scene of mining areas，and the parameters retrieved by remote sensing should be expanded，to improve the accuracy and stability of the retrieving methods;③ Integrating the multi-sources big data in mining areas，the systematic synchronous retrieval and monitoring of parameters should be conducted with moderate high resolution，long time series，and high frequency.