| The coastal water and inland lakes support over 70% of the human activities. With global climate changing and human activities strengthen, the coastal water and inland lakes are suffering from a series of water environment problems, such as water quality deterioration and eutrophication. As an effective method for long-term and wide-range monitoring, the remote sensing technology has been applied in marine environmental monitoring successfully.Because of the relative small area, complex optical properties and dynamic spatial and temporal variations, remote sensing monitoring of coastal water and inland lakes require higher spatial, temporal and radiometric resolution. Although the spectral resolution and signal to noise ratio of conventional ocean color satellites show great advantages, the temporal and spatial resolution are insufficient for monitoring of coastal water and inland lakes. Since water environment monitoring with multi-source remote sensing data has become an irresistible trend, the inconsistency of radiometric characteristic and instability of long term remote sensing data pose a great challenge for the quantitative remote sensing monitoring of water environment. Aiming at the key issues for quantitative remote sensing of coastal water/inland lakes, this thesis conducted researches on quantitative requirements of spatial-temporal-radiometric characteristics of remote sensing data for highly dynamic water environment monitoring as follows:1. Spatial scales of coastal water/inland lakes and optimal spatial scale for remote sensing monitoringAiming at the highly dynamic water bodies, this study takes Poyang Lake, Taihu Lake, Bohai Sea, Yangtze Estuary and Pearl River Estuary as study areas. With the analysis of spatial semivariation and noise equivalent reflectance, the spatial variation scales of coastal water/inland lakes and open sea are below 150 m and higher than 300 m respectively, based on high resolution dataset of GF-1 WFI. Decreasing of spatial resolution increases the variability of the spatial information within pixel. Almost 80% spatial variations could effectively parsed with the spatial resolution of 30 m, while it decreases to 50% with spatial resolution of 256 m. Based on the continuity of spatial variation and Taylor series expansion of water environmental parameters, the thesis analyzed the correlation between the space scale error, spatial variation and non-linear retrieval model for water environmental parameters. With the reduction in spatial resolution, the standard deviation of spatial variation of water environmental parameters in coastal water/inland lakes increased dramatically. In addition, the mutual influence from non-linear retrieval model of water environmental parameters like suspended sediment concentration made the space scale error of suspended sediment gradually increased. The space scale error demonstrated typically regional characteristics. The uncertainties in water environmental quantitative products caused by space scale error could be neglected in calm open sea like South China Sea. In highly dynamic water, by comparison, the uncertainties caused by space scale error ended up to ±5%, ±10% at most. On the basis of point spread function, the thesis built a spatial scale transformation method to get a more realistic description of the conversion process from high spatial resolution to low spatial resolution. By comparison, the spatial pixel average method based on remote sensing reflectance (PSF-RRS) got better results than the method based on total suspended sediment (PSF-TSS).2. Time scale analysis of water environment based on high frequency observation of auto hydro-meteorological buoy system-taking Poyang Lake as an exampleThe highly dynamic characteristics of auto hydro-meteorological buoy system were analyzed with field measured data. The intraday and diurnal rate of change were 16.4 and 147.6 for turbidity; were 10.1 and 34.8 for Chlorophyll; were 17.7 and 28.2 for CDOM, respectively. Based on semi-variance time-scale analysis method, the temporal variant scale for turbidity, Chlorophyll and CDOM in Poyang Lake were 17.5 h,6.6-12.6 h and 8.4-9.4 h, respectively. Taking the highly dynamic characteristics into consideration, the study discussed the best observation strategy:(1) Conduct at least twice observations a day to guarantee the uncertainties of water environment parameters less than 30%; (2) Consider the best observation time with smallest errors when the multiple observations are not available. The network observation of Terra/Aqua could effectively improve the monitoring ability for water environment and made the uncertainties less than 10%. However, their inherent errors were still twice as large as that of best observation strategy. As a result, network observation of multi-sensor satellite remote sensing were required for better monitoring ability for water environment.3. The strategy analysis of observation time for coastal water/inland lakes based on geostationary satellite GOCISelecting about 2300 GOCI images (8 images a day) in 2014, the thesis studied the observation strategy for remote sensing monitoring of coastal water/inland lakes with statistical analysis technique. In general, the best observation time was around 14:30. But the observation time still showed regional features. The best observation time was 9:00-14:30 for Taihu Lake and Bohai Sea; 11:30-13:30 for Yangtze Estuary; 08:30-14:30 for South China Sea. As a result, selecting appropriate observation time was crucial to get optimal observation accuracy when conducting field measurement and remote sensing observation of water with different dynamic characteristics.4. Multi-sensor remote sensing in water environment remote sensingThe radiometric peformance of satellite sensors are assessed in terms of signal to noise ratio (SNR), radiometric sensitivity and radiometric uncertainty, including the Landsat TM/ETM+/OLI, HJ-1 CCD, GF-1 WFI and Terra/Aqua MODIS. Results demonstrate remarkable radiometric peformance of the new generation of satellite sensors including GF-1 WFI and Landsat 8 OLI, with higher than two-fold enhanced SNR and radiometric sensitivity to stable water environment variations. And also, the radiometric peformance of GF-1 WFI and Landsat 8 OLI are comparative to Terra/Aqua MODIS. Uncertainties in remote sensing data and onean color products associated with radiometric calibration are assessed using insitu dataset and MODTRAN simulation. Higher than 60%,30%,25% and 70% errors are pronounced in remote sensing reflectance of the blue,green,red and NIR bands due to ±5% calibration uncertainties, and more than 50% in water quality products like TSS. Radiometric consistency and stablibity of multi-source and multi-temporal data are critical for high precision water environment monitoring, especially for sensors without onboard calibration systems, such as Chinese HJ-1 CCD, which has been widely adopted in water issues for its high spatial-temporal resolution. Methods are proposed using radiometric stable calibration site as well as cross validation with stable sensors like Terra/MODIS and Landsat 7 ETM+. Significant improvements of multi-sensors consistency are achieved in terms of spatial distributions and long-term trend of water qualitu monitoring, with the bias of ±100% to-200% before correction, which are reduced to about 40% after radiometric correction. Results demonstrate great potential of combination from multi spatial,temporal and radiometric resolution remote sensing data for water environment monitoring where high dynamic spatial-temproal variations occurred. |
PDF Full Text Request