| Summary: | 博士 === 國立成功大學 === 環境工程學系碩博士班 === 96 === Nutrient overenrichment (eutrophication) is the major cause of decline in overall water qualities of lakes and reservoirs. Since the watershed nonpoint sources (NPS) pollution is the primary source of nutrient and sediment loads, the study of the linkage between NPS pollution and lake/reservoir water quality is essential and of importance. With an intention to perform better assessments of the linkage, this study was divided into two parts: for a better estimation of NPS loads, collecting representative data and using appropriate models; and for a better evaluation of the water status, mapping the water quality parameters using remote sensing (RS) techniques. This new approach was carried out at Tsengwen Reservoir (TWR), the largest artificial freshwater source in Taiwan.
In the first part of this study, two sampling site in the TWR watershed, Tapu Dam and Takubuyan Dam, were selected to estimate the NPS loads, which represents the NPS loads from the entire TWR watershed and an undistributed forest watershed, respectively. The flow rates were measured with weirs and samples taken for water quality analysis in both non-rainy and rainy days for 2 years (2002 and 2003). The subroutine of the Hydrological Simulation Program - FORTRAN (HSPF) was used to simulate runoff for additional 3 years (2000, 2001 and 2004). Total annual loads of various water quality parameters were then estimated by a regression model. The results indicate that most of the parameter concentrations in both sites are higher during the rainy days. In Takubuyan Dam, the values of parameter concentrations and loads are typically higher as compared to data from other undisturbed forest areas. The fluctuation of annual load from TWR watershed is significant. For example, in Takubuyan Dam, six major events of the entire year, for which the total duration is merely 6.4 days, contribute 42% of the annual precipitation and at least 40% of the annual NPS loads.
In the second part of this study, two approaches had been developed for mapping water qualities from remotely sensed imagery. Whenever the hyperspectral image is available, e.g., images taken by MODerate resolution Imaging Spectroradiometer (MODIS) onboard AQUA satellite or the hyperspectral imager-ISIS from an airborne platform, a novel approach that integrates a semi-analytical (SA) model and a genetic algorithm (GA) to retrieve the constituents of water bodies from remote sensing of ocean color (GA-SA) was suggested. Following the same procedures the International Ocean-Color Coordinating Group (IOCCG) employed in evaluating various algorithms, this GA-SA approach is validated against a synthetic dataset (N=500) and an in-situ dataset (N=656) compiled by the IOCCG. The results of validation indicate that the GA-SA technique is accurate and robust, easy to use, and sufficiently fast to process satellite imagery on a regional scale. This novel approach is applied in processing the images taken by MODIS and generates maps of water constituents and IOPs, including concentrations of chlorophyll-a (Chl-a), non-algal particle (NAP), and the absorption coefficient of color dissolved organic matter (CDOM) at 443nm.
On the other hand, an empirical approach (FORTW-EM) that transform the ratio of Green to Blue band and Red to Blue band of the imagery to respective retrieve the concentration of Chl-a and suspended solids (SS) was suggested, whenever there are only multispectral images available, e.g., images taken by FORMOSAT-2. During the development of FORTW-EM, both the in-situ measurement of spectral reflectance in TWR and the techniques of image processing are applied to the time series of imagery collected by FORMOSAT-2. The result of FORTW-EM shows that the relative percentage differences (RPD) for retrieving Chl-a and SS from the FORMOSAT-2 imagery are 51% and 33%, respectively. The error might be caused by the different level of atmospheric effect on different date. Another possibility is that the FORMOSAT-2 imagery and the in-situ measurement were not collected on the same time.
Because the most appropriate hyperspectral image for mapping reservoir water quality is not available at present. Therefore, based on fifteen FORMOSAT-2 images in TWR taken in 2006, we apply the FORTW-EM approach to derive maps of Chl-a and SS. With the daily runoff loads derived by the NPS regression model and the ancillary information of averaged precipitation and effective storage volume of TWR, the imagery revealed surface dispersal patterns can be reasonably categorized as low-level, stormwater, full-level and water level decreasing periods. The results demonstrated that the different characteristics of Chl-a and SS make them ideal tracers for observing large-scale dispersal patterns. Mapping the water constituents from remotely sensed data enables a better understanding of the linkage of river-borne substances in TWR.
|
|---|
