| Weihe River is regarded as an important water sources in the central region of Shaanxiprovince. But along with the rapid development of our social economic and the ongoingexpansion of the scale of every city, Weihe River has been seriously polluted. As the riverwater quality has a direct impact on social production and people’s daily life, it is necessary tostrengthen the monitoring and management of Weihe River. The conventional water qualitymonitoring technology requires much time and effort, and it is difficult to reflect the status ofwater quality throughout the river. In recent decades, remote sensing technology has gainedrapid development, and it has the characteristics of wide range, fast and the ability of periodiccontinuous monitoring, which make up the defects of artificial water quality monitoring. So ithas a huge potential for application in water quality monitoring in large water area.In this paper, as the study area, Weihe River is located at Shaanxi Province, Xianyangcity. After the research progress of remote sensing technology at home and abroad and thewater quality characteristics of Weihe River were summarized, GA-SVM model was used toconstructed remote sensing inversion model for water quality index of Weihe River. And themain water quality index of transparency, chlorophyll a and total phosphorus were inversed.And then satisfactory results were obtained, which provides a new method for remote sensingmonitoring of water environment of Weihe River. Application of remote sensing technologyfor real-time, dynamic and accurate water quality monitoring has important theoretical andpractical value. The main research results are as follows:(1) Based on the analysis and summary of the basic principles and methods of remotesensing monitoring of water quality, parameters of remote sensing monitoring for waterquality related to this study were put forword. And these parameters mainly containchlorophyll a, transparency and total phosphorus. Chlorophyll a and transparency are essentialsubstances which could influence spectral characteristics of the water body.(2) After the basic theory of SVM and were studied, global optimization ability of GAwas used to optimize kernel function and parameters of SVM. And then a new supportvector machine model based on genetic algorithms was acquired which called GA-SVMmodel.(3) Landsat-7remote sensing images were atmospheric corrected, so as to avoid theatmospheric scattering, refraction, absorption and other factors affecting remote sensing data.Landsat-7remote sensing images were geometric corrected, so as to avoid topographic relief, earth rotation, earth surface curvature and other influence resulting in image distortion. Thecorrelated analysis of waveband and waveband, water quality parameters and water qualityparameters, waveband or waveband algorithm combination and water quality parameters wereconducted, which derived that correlation coefficient of chlorophyll a and B4/B3,transparency and B4/B2, total phosphorus and B4were respectively0.906**,0.877**,0.949**. And these values reach extremely significant correlation, so they could be variablesand dependent variable.(4) The training set data were respectively used to train SVM and GA-SVM model andinversion was conducted. The results showed that there was a significant correlation betweenthe results of the two models inversing chlorophyll a, transparency and ETM+, but nosignificant correlation between the results of the two models inversing total phosphorus andETM+. Although the result of SVM model fitting chlorophyll a, transparency and totalphosphorus was better than GA-SVM model, but its effect of inversing these water qualityparameters was much worse than the GA-SVM model. After through a rigorous refinementand testing, the chlorophyll a or transparency or total phosphorus GA-SVM modelconstructed in this paper could be all applied directly to the water quality monitoring of WeiheRiver in Xianyang City. |
PDF Full Text Request