生态环境学报 ›› 2021, Vol. 30 ›› Issue (7): 1492-1502.DOI: 10.16258/j.cnki.1674-5906.2021.07.018

• 研究论文 • 上一篇    下一篇

基于Landsat 8影像的太湖生化需氧量遥感反演

梁永春1(), 尹芳2,*(), 赵英芬3, 刘磊1   

  1. 1.长安大学地球科学与资源学院,陕西 西安 710054
    2.长安大学土地工程学院,陕西 西安 710054
    3.中国四维测绘技术有限公司,北京 100094
  • 收稿日期:2021-02-25 出版日期:2021-07-18 发布日期:2021-10-09
  • 通讯作者: *尹芳,女,副教授,博士,主要从事遥感与GIS应用研究。E-mail: yinf@chd.edu.cn
  • 作者简介:梁永春（1994年生）,女,博士研究生,主要从事资源环境遥感应用研究。E-mail: 1359205300@qq.com
  • 基金资助:
    国家自然科学基金项目(42071258);中央高校基本科研业务费专项资金(300102270204);中央高校基本科研业务费专项资金(300102278303)

Remote Sensing Inversion of Biochemical Oxygen Demand in Taihu Lake Based on Landsat 8 Images

LIANG Yongchun1(), YIN Fang2,*(), ZHAO Yingfen3, LIU Lei1   

  1. 1. School of Earth Science and Resources, Chang’an University, Xi’an 710054, China
    2. School of Land Engineering, Chang’an University, Xi’an 710054, China
    3. China Siwei Surveying and Mapping Techology Co. Ltd., Beijing 100094, China
  • Received:2021-02-25 Online:2021-07-18 Published:2021-10-09

RichHTML

PDF

摘要：

生化需氧量（BOD）是监测水体有机污染的重要水质参数,运用遥感技术进行水体BOD监测具有快速、便捷的优势。利用2016年7月27日和2016年8月28日Landsat 8/OLI影像数据与2016年8月1—4日和2016年9月1—8日40个样点实测BOD浓度数据,对BOD浓度与Landsat 8像元光谱反射率进行相关分析并选取敏感波段,运用建模集数据构建偏最小二乘回归模型（模型决定系数为0.61,预测均方根误差为0.69 mg∙L-1,预测偏差比为1.61）进行数据验证,实测值和预测值的相关系数达到0.85,预测均方根误差为0.74 mg∙L-1,预测模型效果较好,说明应用Landsat 8/OLI数据进行太湖BOD浓度监测具有可行性。将得到的偏最小二乘回归模型运用至2016年7月27日和2016年8月28日Landsat 8影像得到BOD遥感反演图,遥感反演图中BOD空间分布特征与2016年8月1—4日和2016年9月1—8日样点BOD空间插值结果吻合较好。遥感反演图能更清楚地表征太湖BOD的分布情况,BOD浓度从太湖南部向北部逐渐增加,中南部区域BOD浓度低,西北部、边缘地区以及东南部区域BOD浓度较高。运用太湖其他时段Landsat 8影像与实测BOD浓度构建偏最小二乘回归模型,2016年4月和2017年3月建模效果均比较好,说明偏最小二乘回归方法具有较强的适用性。分析多时段反演结果,2016年中期太湖有机污染状况较初期严重,说明水体有机物质受季节和温度等因素的影响。综上,偏最小二乘回归模型能够较准确估算水体BOD浓度及其空间分布,可为太湖有机污染的评价和治理提供依据。

关键词: 太湖, 有机污染, 生化需氧量, Landsat 8 OLI遥感影像, 偏最小二乘法, 遥感反演, 空间分布

Abstract:

Biochemical oxygen demand (BOD) is the main indicator to monitor organic pollution, which can reflect the degree of water quality. Using remote sensing technology has advantages of fastly and conveniently monitoring BOD concentration in Taihu. In this paper, 40 samples with BOD concentration were sampled in August and September, 2016 in Taihu Lake and Landsat8/OLI remote sensing images of the similar day also were collected. The BOD remote sensing inversion model was established using the partial least squares method based on the correlation analysis between BOD concentration and pixel spectral reflectance. The Partial Least Square Model (PLS) showed that the coefficient of determination (R2) was 0.61, the prediction of root mean square error (RMSEP) was 0.69 mg∙L-1 and the residual prediction deviation (RPD) was 1.60. The partial least squares regression model was applied to the validation data to get the correlation coefficient (r) is 0.85, and the prediction of root mean square error (RMSEP) was 0.74 mg∙L-1. The result was good showing that the inversion based on Landsat8 remote sensing data and BOD concentration in Taihu Lake was feasible. By applying the developed model to Landsat 8 images on July 27, 2016 and August 28, 2016, spatial and temporal changes in satellite-derived BOD could be observed in Taihu Lake. The spatial distribution characteristics of the BOD concentration were in good agreement with the spatial interpolation results of the sample points and the former works better. The BOD concentration increased gradually from the south to the north of Taihu Lake with low concentration in the central and southern regions and high concentration in the northwest, marginal areas, and southeast regions. The partial least squares method was applied to the inversion of BOD in other periods of Taihu Lake. The prediction results in January 2016 and March 2017 were both good indicating that the partial least squares regression model had a strong applicability. Analysis of the multi-time inversion results reveals that organic pollution in the middle of 2016 was more serious than that in the early stage, which indicated that the organic matter in the water was affected by seasonal and temperature factors. The results show that the partial least squares regression model can accurately estimate the BOD concentration and its spatial distribution. The results of this study can provide a basis for the evaluation and treatment of organic pollution in Taihu Lake.

Key words: Taihu, organic pollution, biochemical oxygen demand, Landsat 8/OLI remote sensing image, partial least squares regression, remote sensing inversion, spatial distribution

中图分类号: