| Over the past few decades, along with the rapid and continuous development of economy in our country, the environment and spatial structure of coastal waters have been changed greatly, by both climate and human activities, which has deeply affected the service and ecological balance of the sea environment. Especially, the Bohai bay’s capability of circulation and self-purification is lower than other seas, due to its semi-closed geographical location. So monitoring the water quality in Bohai bay is of great significance in estimating its ecological environment and the level of development and utilization of resources reasonably. Measuring the chlorophyll-a concentration in Bohai bay can help to evaluate the state of health of water quality, aquatic resources distribution and pollution degree. The application of remote sensing technology can make up for the shortage of the traditional survey methods, and monitor the water quality parameters continuously in a wide range and a long time. Though many empirical models or semi-empirical and semi-analytical models has been established to retrieve the chlorophyll-a concentration in ocean, they may not be applied to the turbid water in Bohai bay. Therefore, a monitoring model, based on the condition of Bohai bay, should be established.According to researches of the characteristics of the spectral reflectance of Bohai bay coastal water, on in-situ measurements of chlorophyll-a concentration and quasi-synchronous landsat-7 ETM images were used to build up remote sensing retrieval models in this paper. On the basis of demand for quantitative remote sensing retrieval, a series of preprocessing work was done to implement the preparation of information extraction from remote sensing images accurately. By analyzing the correlation between the bands or band combinations of ETM images and the chlorophyll-a concentration in water, the band combinations with strong sensitivity were selected as variables. Stepwise regression analysis was used to find out the quantitative relationship between these variables and the chlorophyll-a concentration, then a statistical model was built up with the R2 0.864, RMSE 0.957. To improve the retrieval accuracy, the artificial neural network technology was applied. By comparing the training results of networks with different hidden nodes, a BP neural network model with 3 layers was constructed. The network model was consisting of 4 input nodes(the reflectance of the first 4 bands of ETM), 8 hidden nodes and 1 output node(the chlorophyll-a concentration measurements) while the R2 is 0.956 and RMSE is 0.856.The results indicated that the BP neural network model has better fitting effect and it provides a reliable basis and method for monitoring water chlorophyll-a concentration by remote sensing technology quickly and accurately.Images of chlorophyll-a concentration distribution from 2007 to 2010 were produced by applying the statistical model to find out the spatial distribution and temporal change characteristics of Bohai bay. Through analysis, the chlorophyll-a concentration in Bohai bay was found to have the spatial characteristics that the chlorophyll-a concentration of southern sea is higher than that of northern sea, and it decreased obviously from nearshore to far offshore. And no obvious difference with the season change was observed while the chlorophyll-a concentration in autumn is lower than that in spring and summer due to various factors such as precipitation and water temperature. |
PDF Full Text Request