基于归一化遥感反射比反演水中石油含量模式研究

Research on an Inverse Model of Petroleum Content in Water Bodies Based on the Normalized Remote Sensing Reflectance

自然水体中石油污染含量的时空间分布是数字地球领域水体环境研究的一个重要参数,利用遥感技术提取水中石油含量也是数字地球研究领域的一个热点问题。基于2009—2013年进行的六次石油类污水野外配比实验数据,结合2013年6月—2014年4月晴天获取的中国辽宁盘锦辽河油田Landsat 8/OLI遥感数据,在对配比实验测定值、Landsat 8/OLI传感器模拟值和Landsat 8/OLI传感器过境值等三类遥感反射比数据进行归一化处理的基础上,通过对波段特征进行分析,发现含油水样与无油水样在绿色、红色和近红外波段的遥感反射比存在明显的差异,由此提出了水体石油污染归一化遥感反射比指数NDPRI(Normalized Difference Petroleum Remote Sensing Reflectance Index),该指数可用来判断水体是否存在油污染,进一步建立利用NDPRI来反演水中石油含量遥感模式。结果表明:(1)含油水样和无油水样在蓝色波段的归一化遥感反射比差别不大;(2)含油水样在绿色和红色波段的归一化遥感反射比明显低于无油水样;(3)无油水样在近红外波段的归一化遥感反射比呈现下降趋势,而含油水样的归一化遥感反射比随着石油类浓度含量的增加而明显抬升,呈现出较好的线性关系;(4)含油水样,其NDPRI小于0.45,此阈值作为是否有油污染的判断值。最后,利用2014年4—8月污水厂获取的现场实测数据对模式进行验证,相对误差为13.5%,表明所建立的模式具有良好的石油含量反演精度。

The temporal and spatial distribution of petroleum content in contaminated water is an important parameter in the research on digital earth water environment, and petroleum content extraction through the remote sensing technology is also a focus issue in the digital earth sector. In this work, combining the data from 6matching experiments of petroleum contaminated water between 2009 and 2013, and the remote sensing data of Liaohe Oilfield acquired by Landsat 8/OLI（Operational Land Imager） under clear sky from June 2013 to April2014, this paper firstly normalizes the remote sensing reflectance data from the matching experiments, the estimated Landsat 8/OLI sensor values and the observed Landsat 8/OLI sensor values. Through analyzing the characteristics of wavebands, it is discovered that the remote sensing reflectances of water samples with and without petroleum are obviously different at the green, red and near-infrared bands. Based on this result, this paper proposes the Normalized Difference Petroleum Remote Sensing Reflectance Index（ NDPRI） to determine whether the water is contaminated by petroleum, and a remote sensing model is then established to inverse the petroleum content using NDPRI. The results show that compared with water samples without petroleum, the normalized remote sensing reflectance of water samples with petroleum is similar at the blue band, obviously lower at the red and green bands, and significantly rises with increasing petroleum concentration and indicates good linear relation as against the declining trend of water samples without petroleum in the near-infrared band. The NDPRI of water samples with petroleum is lower than 0.45, and this threshold value is used to determine whether oil pollution occurs. The model is validated using the in-situ data of sewage plants from April to August in 2014, with the relative error of13.5%, indicating a reasonable accuracy of the model in inversing the petroleum content.