| 引用本文: | 侯毅凯,张安兵,吕如兰,等.基于多源数据的河道水质遥感反演研究[J].灌溉排水学报,2023,42(11):121-130. |
| HOU Yikai,ZHANG Anbing,LYU Rulan,et al.基于多源数据的河道水质遥感反演研究[J].灌溉排水学报,2023,42(11):121-130. |
|
| 摘要: |
| 【目的】探讨基于不同平台获取的遥感数据对河流水质反演的适用性。【方法】以邯郸市滏阳河为研究对象，分别采用Sentinel-2卫星影像、无人机（UAV）搭载多光谱传感器和ASD地物光谱仪3类遥感平台获取了不同季节的光谱数据，引入随机森林（RF）算法对不同水质参数建立反演模型，并评价模型的预测效果。【结果】①基于不同平台数据源的RF模型预测效果具有明显的季节特征，整体表现为，夏季优于春季，冬季最差，其中，3类平台遥感数据均可对夏季的各水质参数进行反演，春季可以完全由UAV遥感技术完成；②反映水体浑浊程度的Turb和SS，在不同季节均可由基于无人机多光谱数据的RF模型反演预测，尤其是冬季，利用Sentinel-2卫星和无人机（UAV）搭载多光谱传感器这2类面源监测遥感平台优势明显；③基于高光谱数据的RF模型对TN的检测效果最佳，春季的预测R2超过0.9，且泛化能力很好；④不同季节的高光谱反射率曲线显示，因季节因素造成污染程度不同，反射率分布曲线存在差异，夏季不同区位反射率差距明显，相同波长，不同样本的反射率高值与低值变化幅度大，且吸收、反射变化明显，春季和冬季的反射率曲线形态类似，但春季各波长反射率变化较冬季明显。【结论】利用不同平台的遥感技术可以为河道水体水质监测提供丰富数据，为实现河道水质信息实时监测提供新的技术手段。 |
| 关键词: 多源遥感；随机森林；中小河流；水质参数 |
| DOI：10.13522/j.cnki.ggps.2023187 |
| 分类号: |
| 基金项目: |
|
| Assessing River Water Quality Using Different Remote Sensing Technologies |
|
HOU Yikai, ZHANG Anbing, LYU Rulan, XUE Xiaoshuang, ZHANG Yanping, PANG Jiyu
|
|
1. Hebei University of Engineering, Handan 056038, China; 2. Hebei Water Ecology and Social Governance Research Center,
Handan 056038, China; 3. Handan Polytechnic College, Handan 056001, China; 4. Hebei Branch of Construction and
Administration Bureau of South-to-North Water Diversion Middle Route Project, Handan 056006, China
|
| Abstract: |
| 【Objective】Remote sensing has become a prominent tool for monitoring and evaluating surface water quality across catchment and basin scales. This paper compares different remote sensing technologies for assessing water quality of terrestrial rivers. 【Method】We took the Fuyang River in Handan City as an example. Spectral data in different seasons were acquired from three distinct remote sensing platforms: Sentinel-2 satellite images, unmanned aerial vehicles (UAVs) equipped with multispectral sensors, and ASD field spectrometers. The random forest (RF) algorithm was used to derive the inversion model from each platform to assess water quality.【Result】① The results obtained from RF models for different platforms showed noticeable seasonal variation. Overall, the calculated results were most accurate in the summer, followed by spring, and were least reliable in winter. During summer, remote sensing data obtained from all three platforms can accurately estimate water quality parameters. In spring, UAV remote sensing images were sufficient. ② The RF model derived from the UAV multispectral data can effectively predict turbidity (Turb) and suspended solids (SS) of water i the river, regardless of seasons. ③ The RF model derived from the hyperspectral data was most accurate for estimating total nitrogen, with R2 >0.9 in spring. ④ Hyperspectral reflectivity curves for different seasons showed distinct variation due to the change in pollution level. In summer, reflectivity distribution curves display markedly spatial variation. The reflectivity values at each wavelength varied significantly, with pronounced changes in absorption and reflection. The reflectivity curves in spring and winter were similar, but showed pronounced changes in spring, regardless of the wavelength. 【Conclusion】Remote sensing images obtained from different platforms offer a wealth of data for river water quality monitoring. They provide a new avenue for real-time river water quality assessment. |
| Key words: multi-source remote sensing; random forest; small and medium-sized rivers; water quality parameters |
