| High concentration of nutrients from agriculture and urban runoff, or those produced by offshore upwelling, are causing algal blooms in many lakes and reservoirs, rivers and streams, estuaries and coastal waters. Algal blooms induce eutrophic conditions, depleting oxygen levels needed by organic life, limiting aquatic plant growth by reducing water transparency, and producing toxins that can harm fish, benthic animals, and humans. The magnitude and frequency of cyanobacteria blooms in Lake Taihu have increased significantly in recent decades, which has attracted the attention of environmental agencies, water authorities and human health organizations. Reliable mapping of the amount of cyanobacteria is especially important in the case of Lake Taihu, where cyanobacterial blooms occur every year. Satellite or airborne measurement of spectral reflectance is an effective method for detecting and monitoring algal by its proxy, concentration of chlorophyll-a, the green pigment. Dectection of cyanobacteria blooms from space has been of interest since1980s. Methods using single band, band ratios between near-IR and red bands, normalized difference vegetation index (NDVI), enhanced vegetation index (EVI) or floating algae index (FAI) have been proposed to detect intense surface cyanobacteria blooms from limited field and satellite data for Lake Taihu. The aim of this study is to demonstrate an approach, which can effectively map the distribution of cyanobacteria blooms in turbid case II waters.Firstly, radiative transfer model and bio-optical model were introduced. Based on the simulated data of remote sensing reflectance simulated by the optical simulation software HYDROLIGHT, we generated a large number of spectra data with different combinations of water quality parameters (WQPs) of Lake Taihu under different external environment. Then the relational curves between Rrs (1)(645nm, MODIS band1)～Rrs(2)(800nm, MODIS band2) were analyzed, each curve representing chlorophyll-a and NPSS contour. The lookup table for chlorophyll-a concentration and ao threshold was obtained by linear fitting of Rrs(1) and Rrs(2). Through statictics of α0threshold gradient difference over pixels in the MODIS image of May19th,2008, ao threshold of cyanobacteria blooms was determined, as well as the corresponding chlorophyll-a concentration. To evaluate the effectiveness and applicability of ao threshold, our method was applied to three MODIS image data in Lake Taihu and the detecting results were compared with false color synthesis pictures. Finally, validation and comparison between α0detection mode and other methods, i.e., single band mode, band ratio Rrs(2)/Rrs(1), band ratio Rrs(2)/IRrs(4), NDVI, EVI and FAI were analyzed. The landsat ETM+false color synthesis image was used to evaluate α0detection mode for mapping the detail distribution of cyanobacteria blooms. The main conclusion of the study can be summarized as follows:(1) The bio-optical model describes the relationship between IOP and AOP. The one order bb/(a+bb) model-deduced equation can be used to demonstrate the relation between remote sensing reflectance of NIR and red band, for MODIS sensor, the equation is as below:1/Rr(2)=α0/Rrs(1)+1-α0/g (2) The optical simulation software HYDROLIGHT is based on the theory ofradiative transfer model. It can be used to simulate remote sensing reflectance without the interference of changing environmental and observing conditions during field measurements. We designed a cross simulation table for Lake Taihu to analyse the variation features and influential factors of the remote sensing reflectance. One is the influence of single parameter (solar zenith angle, pure water, chlorophyll-a, CDOM and NPSS) and the other is the mixed spectra of the combination of four water quality● parameters. According to the simulation results, solar zenith angle has little influence on the growth of remote sensing reflectance relative to the four water quality parameters. The reflection spetra of CDOM present exponential decrescence with wavelength. During400-700nm, remote sensing reflectance decreases with the increasing of chlorophyll-a concentration as a result of the strong absorption of chlorophyll-a. The reflection peak near the wavelength of680nm gradually surpasses that in the wavelength of540nm along with the growth of chlorophyll-a concentration. For NPSS, its variation has enormous impact on remote sensing reflectance, especially on the wavelength of visible band.(3) In our study, algal waters were classified into high reflectance algal waters and low reflectance algal waters. We use1/Rrs (1) and1/Rrs (2) threshold to map the distribution of Cyanobacteria blooms of the former one. ao threshold was used to detect low reflectance algal waters. Statictics of ao threshold gradient difference over pixels in the MODIS image of May19th,2008were analysed. Accroding to the lookup table for chlorophyll-a concentration and ao threshold, ao threshold of cyanobacteria blooms was determined. Then by comparing the detecting result with false color synthesis images, the threshold varied between0and1.2001after calibration through trial and error with careful visual inspection. The ao detection mode could precisely map the cyanobacteria blooms of low reflectance excluding the impact of high turbid water, which was applicable for Lake Taihu.(4) A window for detecting cyanobacteria blooms in Lake Taihu in the domain of Rrs(1) and Rrs(2) was proposed in our study. Contours of chlorophyll-a and NPSS concentration in the domain could be used to judge the effectiveness of a certain detection mode. In addition, a fast detection method based on the2D scatter plots for MODIS band1and band2reflectance were created, combined with our α0detection mode.(5) Single band mode, band ratio Rrs(2)/Rrs(1), band ratio Rrs(2)/Rrs(4), NDVI, EVI and FAI were the commonly used algal detecting mode. We applied these modes to the MODIS surface reflectance data image of May19th2008, August13th2010and December31th2010. The mapping results were compared with that of ao detection mode and false color synthesis images. In summer with relatively low suspended sediments in Lake Taihu, the results of single band mode, band ratio Rrs(2)/Rrs(4) mode and NDVI were less than normal as they were unable to distinguish low reflectance algal bloom from clear waters. The result of FAI was higher than normal as it contained some non-algal waters. Nevertheless, in winter with high suspended sediments, except our α0detection mode, NDVI and EVI mode, other methods wouldn’t distinguish cyanobacteria blooms from turbid water and cannot be fit for detecting cyanobacteria blooms from turbid waters. In conclusion, it has proved that α0detection mode is better than the existing detection methods.(6) Finally, Landsat7ETM+high resolution data image of September24th,2011was used to evaluate the accuracy and effectiveness of ao detection mode. Band3,2,1of Landsat7ETM+were used to creat true color synthesis picture, while Band5,4,3were used to creat false color synthesis picture, where the bloom texture can be clearly visualized and therefore distinguished. Our ao detection mode can effectively represent the distribution of cyanobacteria blooms in Lake Taihu, for example, the algae slicks and patches were clearly visible in the detecting result. However, the limitation of our mode is that it is not useful for studying cyanobacteria blooms with biomass concentrations too low to form surface scums or algae blooms of noncyanobacterial taxa, where the algae particles are mixed with the water molecules. |
PDF Full Text Request