Study On The Remote Sensing Models For Monitoring Riverine Water Environment

River is the key component of the earth's hydrologic cycle.Water environment of the river basin is an important basis for evaluating the function of the regional ecosystem,and is closely related to the human life and social development.Therefore,with the development of social economics,riverine water environment monitoring becomes more and more important.It is reliable basis for water resources management and socio-economic development decisions.However,the traditional ground monitoring method is time-consuming and exhausting,and the time and space coverage is also limited,which is not conducive to large-scale and continuous monitoring.Thanks to the development of earth observation technology,remote sensing has played an essential role in water environment monitoring,which has become a research focus.This study takes Yarlung Zangbo River,an alpine river with unique geographical features,and Yellow River Estuary,famous for its high turbidity,as research areas respectively.This study focus on two remote sensing monitoring scenarios: with field survey data and lack of sufficient field data.A semi-empirical inversion model and a self-organizing feature extraction model are constructed,in combination with multitemporal,medium and high-resolution,and multi-source remote sensing data.As a result,two different water environment remote sensing monitoring methods are discussed,and long time series monitoring models are set up.The long time series monitoring model for alpine rivers is especially explored to provide important information support for water environment monitoring and water resources management in the context of global change.The main research results of this paper include the following points:(1)The spectral curve of the plateau river appears to reach peaks or valleys near band in 580-700 nm,750nm and 810 nm.The spectrums covering less than the blue band of 450 nm,the red band of 650-870 nm and the near infrared band are more sensitive to the concentration of water environment parameters,and the green band near 550-580 nm is the most insensitive.These bands should be considered first when constructing the remote sensing monitoring models of water environment.(2)Based on Sentinel-2A data,a variety of inversion models for water environment parameters were built.After optimization and comparison,the ratio of red to green band,ratio of red edge to green band had better inversion effects.The power function model of red to green band ratio is of high precision,stable result and strong applicability,and can be extended to Landsat series data.It is suitable for long time remote sensing monitoring for water environment.(3)The turbidity of the middle streams of Yarlung Zangbo River is monitored with Landsat and Sentinel-2A remote sensing data over 2007-2017.The results showed that the turbidity of the area was declining in the past ten years,and the changes had obvious seasonal characteristics.The turbidity difference was significant in wet and dry season,which was higher in June to September.The turbidity difference of the different stream segments in wet season was also great,which was lower in the upstream and downstream,and higher in the middle section.And the difference in dry season is nonsignificant.(4)In the research scenario of lacking sufficient field data,a self-organizing mapping(SOM)neural network model is proposed to extract the typical patterns of water environment characteristics from long time remote sensing images,which is a supplement to the empirical model.Taking the Yellow River estuary as the research area,a SOM model was constructed to extract typical turbidity patterns of 2000-2015 with MODIS data.The results showed that turbidity has been decreasing in recent 15 years and annual variation had obvious seasonal characteristics.The turbidity in winter and spring was higher,and in summer and autumn was lower.The validation results indicated that typical features extracted from the SOM model were in good agreement with the empirical inversion results.It demonstrates that the SOM model is reliable to extract the water environment characteristics,and can be an important supplement for empirical model in water environment monitoring when field data is inadequate.Remote sensing technology is competitive in monitoring riverine water environment.With the improvement of spatial and spectral resolution of many remote sensors,remote sensed earth observation technology will play a more significant role in water environment monitoring.It is an essential method for comprehensive evaluation of regional water environment and overall ecosystem.