| Lakes play important roles in human’s life, such as freshwater resources for drinking water, agriculture, industry, fishing, recreation, and tourism. In addition, they can regulate the regional climate and maintain regional ecosystem balance. In recent decades, for climatic changing and human activities intensified, the degradation of ecosystems, eutrophication and the reduction of storage capacity are widespread problems of lakes in China. These problems, especially accelerated eutrophication restrict the sustainable development and influence the lives of people around the basin. Taihu Lake, the third largest freshwater lake in China, plays a significant role in drinking water supply, restraint flood, irrigation, fishery and transportation. However, it is suffering from the change of natural environment and intensified human activities in recent years, the water quality becoming worse and worse, and algal blooms outbreak frequently, which severely restricted the sustainable development of the basin. In order to sustainable socio-economic development in the Taihu Lake Basin, it is crucial to continuous monitor the state eutrophication of Taihu Lake and control the outbreak of algal blooms. The most basic is to obtain the spatial and temporal distribution of algae in Taihu Lake.Chlorophyll a is an important indicator for measuring the primary productivity of water bodies. As such, it is usually to measure the concentration of chlorophyll a to map the water algal bloom of lakes. Conventional method is an effective way to obtain the water quality. However, it cannot truly characterize the spatial variability of the large lake such as Taihu Lake by a limited number of field samples. Remote sensing techniques, which have the inherent ability to provide spatial and temporal information about water, and pay less than conventional method, may be the only viable way to effectively monitor water quality in lakes. Combination of conventional and remote sensing monitoring technology provides strong technical support for this study, the purpose of which is to obtain the temporal and spatial variability of algal blooms in Taihu Lake.From the remote sensing perspective, waters can generally be divided into two classes:case I and case II waters. Case II waters are waters containing not only phytoplankton, but also suspended sediments (NPSS), coloured dissolved organic matter (CDOM), and anthropogenic substances which their concentrations are not correlated with chlorophyll-a concentration. Taihu Lake is a typical inland case II water, in which there is complex interactions of four optically active substances. Such that, remote sensing in it has been far less successful to retrieve chlorophyll a concentration by using the algorithms developed in case I waters.We propose a spectral decomposition approach for assessing chlorophyll a concentration in Taihu Lake. In this method, the mixed reflectance spectrum of a given pixel is considered as a linear combination of the basic components (also called end member), wherein the decomposition coefficients are related to the cover ratios for ground objects. Likewise, the spectral radiance/reflectance of a volume of Taihu Lake can also be conceptualized to be a representing thee weighted sum of the four basic components (clear water, non-phytoplankton suspended sediments (NPSS), coloured dissolved organic matter (CDOM) and phytoplankton). This approach is thought to be a better theory for retrieve the water quality parameters in the case II waters.In this study, a simulation experiment has been designed firstly to model the mixing spectral reflectance of four basic components through the Hydrolight software. After the spectral simulation experiment, the characteristics of the four basic optical components and the interaction between phytoplankton and non-phytoplankton suspended sediments and coloured dissolved organic matter have been analyzed.There are two steps to build spectral decomposition model for inversion of chlorophyll a concentration. First step is to obtain the standard spectral reflectance of the four basic components using the simulation spectral reflectance of these four end-members. The standard spectral reflectance obtained from simulation is more "pure" than obtained from field measurement or imagery directly. The second step is to establish four equations from different wavelengths basic on multi-band satellite data Landsat TM and MERIS to solve the decomposition coefficients Cp, Cn, Cc and Cw. There were one TM band combination and eight MERIS band combinations being selected to calculate the decomposition coefficients. After choosing the range of chlorophyll a concentration basic on the result of analysis the interaction of four basic optical components, the relationships between Cp and the chlorophyll a concentration have been analyzed to choose the best combination to build the model. The best combination is MERIS band3,5,8,9combination. And then, the model basic on the relation between Cp and chlorophyll a is built. The results show that the spectral decomposition model performs well, which has high R2(0.8892), low MAPE (34.28%) and RMSE (17.43ug/L).The simulation spectral reflectance and field measurement spectral data were used to valid the model. The results indicate the potential robustness of the spectral decomposition algorithm based estimation model. Furthermore, we applied our model to MERIS data on April,24th,2008. The whole lake was classified in to four classes using the method developed by Li Yunmei. After that, the model was applied to the whole lake (except the type A) and the synoptic distribution of chlorophyll a was obtained. Comparing the retrieval values and in situ chlorophyll a observations, the simulation values are almost the same as field measurements, which indicates that the model has high precision and applicability. Moreover, we built band combination models based on the studies of Li Yunmei and Zhou Lin. Then we apply these models to the MERIS image the same day, and obtain the MAPE and RMSE values. The results show that the spectral decomposition model is better than the other models, which has high inversion accuracy and MAPE and RMSE are31.06%and8.60ug/L, respectively.In addition, the model based on spectral decomposition algorithm in our study reflect the variety of chlorophyll a concentration much more accurately than the models based on band combinations of red and near infrared bands. The reason may be in the spectral decomposition model, the chosen bands are not only red and near infrared bands but also blue and green bands which are sensitive to the variety of chlorophyll a concentration. |
PDF Full Text Request