从高光谱卫星数据中提取植被覆盖区铜污染信息

重金属矿山开采活动对生态环境造成了复杂的影响，而植被的生长状况是矿山周围生态环境变化的重要指示因子。常规的对植被中重金属元素富集程度的估计方法往往依赖于大量的野外调查数据以及实验室化学分析方法，需要耗费大量的物力与时间。而通过光谱分析方法能够非接触地获取植被中的重金属含量信息。然而，从高光谱遥感卫星数据中反演植被重金属含量信息仍然是个比较困难的工作。从众多的卫星数据波段中找到对特定重金属元素敏感的波段是植被重金属含量提取的关键。该工作探索了一种基于统计分析的方法从国产卫星（HJ-1）高光谱数据中实现特征波段发现与重金属含量提取，该方法基于野外实测数据集与环境星高光谱遥感图像，选取叶片铜离子含量作为重金属铜污染程度的指示因子，利用44组德兴铜矿植被覆盖区的叶片铜离子含量数据与逐步多元线性回归与交叉验证方法，实现了环境星高光谱反射光谱与叶片铜离子含量的统计建模。将统计模型应用于德兴铜矿研究区，生成了德兴铜矿植被覆盖区叶片铜离子含量分布图。结果表明，该工作建立的重金属含量估计模型具有明显的统计学意义，叶片铜离子含量对516 nm附近的冠层反射光谱最为敏感。矿区铜含量制图结果表明，德兴铜矿植被覆盖区铜离子含量的分布范围介于0～130 mg·kg-1之间，铜离子含量在德兴市东南角和铜矿周围的植被覆盖区域含量最高，均介于80～100 mg·kg-1之间。制图结果可以为德兴铜矿重金属铜污染的预防与治理提供基础数据支持。

Retrieval of Copper Pollution Information from Hyperspectral Satellite Data in a Vegetation Cover Mining Area

Heavy metal mining activities have caused the complex influence on the ecological environment of the mining regions. For example, a large amount of acidic waste water containing heavy metal ions have be produced in the process of copper mining which can bring serious pollution to the ecological environment of the region. In the previous research work, bare soil is mainly taken as the research target when monitoring environmental pollution, and thus the effects of land surface vegetation have been ignored. It is well known that vegetation condition is one of the most important indictors to reflect the ecological change in a certain region and there is a significant linkage between the vegetation spectral characteristics and the heavy metal when the vegetation is effected by the heavy metal pollution. It means the vegetation is sensitive to heavy metal pollution by their physiological behaviors in response to the physiological ecology change of their growing environment. The conventional methods, which often rely on large amounts of field survey data and laboratorial chemical analysis, are time consuming and costing a lot of material resources. The spectrum analysis method using remote sensing technology can acquire the information of the heavy mental content in the vegetation without touching it. However, the retrieval of that information from the hyperspectral data is not an easy job due to the difficulty in figuring out the specific band, which is sensitive to the specific heavy metal, from a huge number of hyperspectral bands. Thus the selection of the sensitive band is the key of the spectrum analysis method. This paper proposed a statistical analysis method to find the feature band sensitive to heavy metal ion from the hyperspectral data and to then retrieve the metal content using the field survey data and the hyperspectral images from China Environment Satellite HJ-1. This method selected copper ion content in the leaves as the indicator of copper pollution level, using stepwise multiple linear regression and cross validation on the dataset which is consisting of 44 groups of copper ion content information in the polluted vegetation leaves from Dexing Copper Mine in Jiangxi Province to build up a statistical model by also incorporating the HJ-1 satellite images. This model was then used to estimate the copper content distribution over the whole research area at Dexing Copper Mine. The result has shown that there is strong statistical significance of the model which revealed the most sensitive waveband to copper ion is located at 516 nm. The distribution map illustrated that the copper ion content is generally in the range of 0～130 mg·kg-1 in the vegetation covering area at Dexing Copper Mine and the most seriously polluted area is located at the South-east corner of Dexing City as well as the mining spots with a higher value between 80 and 100 mg·kg-1. This result is consistent with the ground observation experiment data. The distribution map can certainly provide some important basic data on the copper pollution monitoring and treatment.