A Study On Spectral Characteristics Of Typical Submerged Aquatic Vegetation In Shanghai And Its Application

Submerged aquatic vegetation (SAV) is a key component in the aquatic ecosystem. Re-establishment of SAV has been recognized as a valuable technique of ecological engineering for the restoration of eutrophicated lakes or rivers. Remote sensing technology can be used to monitor the distribution and abundance of SAV through time on a large scale. SAV monitoring could provide scientific basis for protecting and using SAV resources and assessing the effectiveness of the ecological restorations for aquatic ecosystem.Dianshan Lake, the largest natural freshwater lake in Shanghai suburbs, was chosen as the study area. This study was designed to investigate the spectral reflectance characteristics of several typical SAV plants in Shanghai using a FieldSpecTM Pro JR Field Portable Spectroradiometer and map SAV distribution in Dianshan Lake using a WorldView-2image. The among-special spectral characteristics were investigated by selecting and combing the best wavebands. By means of the unique optimal wavebands of SAV were explored, the classification model of SAV had been established to identify different SAV species from remote sensing data. To analyze how the factors of water condition affecting the spectral characteristics of SAV quantitatively or qualitatively, the spectral reflectance of SAV communities in different water condition i.e. the water depth from water surface to plant canopy (WDC), algal chlorophyll content (Chla), suspended content (SC) and substrate were measured in the control experiment. The response and the characteristic sensitive wavebands for the spectral reflectance of SAV with varied biophysical parameters (coverage/biomass) in situ and the control experiments were analyzed; and meanwhile, the most important factors affecting the spectral characteristics of SAV on these sensitive wavebands were screened from the other factors of water condition. The revised models, which could effectively reduce the influence of water condition on the deductive accuracy for the biophysical parameters of SAV, were established adopting both the key factors of water condition and spectral reflectance. The pre-processing of the WorldView-2image of Dianshan Lake got in September2011was performed to realize the conversion of DN to ground reflectance. The distribution and species composition of aquatic plants in Dianshan Lake area were interpret based on the classification models established from among-species spectral characteristics of aquatic plants. The biophysical parameters of SAV were predicted using the inversion models in Dianshan Lake area. On this foundation, the composition, distribution and biophysical parameters of SAV were mapped. The main results of this study are as follows:(1) The spectral reflectance at the visible and near infrared wavebands for SAV were a much lower than those for emerged or floating-leaved aquatic plants. There were clear reflection peaks around1670nm and2200nm in the spectral reflectance of emerged and floating-leaved aquatic plants, but the spectral reflectance of SAV approached0after1350nm. The spectral characteristics of middle-infrared, green and near infrared bands could be useful to discriminate SAV species from the emerged and floating-leaved species. The position of maximum value in the red edge for the first derivative curves of reflectance were observed near the band of720nm for emerged species, while were near the bands of700nm for floating-leaved and SAV species. So the position of maximum value in the red edge for the first derivative curves of reflectance could be used to discriminate the emerged aquatic species from the floating-leaved species. The vegetation index RVI, NDVI and the spectral index NGP were most sensitive characteristic index for identifying5emerged or floating-leaved species, Zizania caduciflora, Phragmites australis, Alternanthera philoxeroides, Lemna minor and Trapa incise, while spectral index NAV, REP, NGP were the most sensitive characteristic index for identifying6SAV species, Najas marina, Hydrilla verticillata, Myriophyllum spicatum, Ceratophyllum demersum, Vallisneria natans and Potamogeton malaianus. The classification functions based on vegetation indexes and spectral indexes had done well in identifying the aquatic plants species from field sampling.(2) The reflectance of SAV communities increased with their increasing coverage/biomass both at the visible band and the near infrared band in situ and the control experiments. The biggest and the bigger correlation coefficient could be observed at the band of700-900nm and520-620nm between the reflectance of SAV communities with their coverage, respectively. The correlation coefficients between the biomass and the reflectance were a little lower than those between the coverage and the reflectance. SAV plants had special spectral characteristics in different seasons, and the position of red edge/green peak and the reflectance of the near infrared band were changed regularly with seasons.(3) The reflectance of Cabomba caroliniana communities decreased with their increasing WDC both at the visible band and the near infrared band. The biggest correlation coefficient could be observed at the band of700-900nm and520-570nm. In the same condition of coverage and WDC, the reflectance of C. caroliniana communities increased both at the green band and the near infrared band with the increasing CHLA and SC in water condition. In contrast, the different substrates in water did not affect much on the reflectance of C. caroliniana communities.(4) The coverage of SAV and WDC were the most important factors affecting the spectral characteristics of SAV at the band of700-900nm. An inversion model of SAV coverage was established, which could effectively reduce the influence of water condition on the deductive accuracy for the coverage of SAV.(5) The WorldView-2image of Dianshan Lake was interpreted to different aquatic plants species using both a supervised classification and a decision tree classification. The decision tree classification was established to identify the land features types and aquatic vegetation species based on the characteristic bands and index. The decision tree classification model has a better results, the overall classification accuracy reached88.7%, Kappa factor of0.8306. Using the inversion model of SAV coverage, which had been optimized the factors of water condition affecting spectral characteristics of SAV, the coverage of SAV was deduced on a large scale based on the reflectance image performed from WorldView-2image of Dianshan Lake. The inversion precision was83.4%.In this research, the spectral characteristics of typical submerged aquatic vegetation in Shanghai and its application were investigated for monitoring SAV using remote sensing technology. The WorldView-2high resolution image was combined with the spectral information to explore the appropriate interpretation technique for the inversion of SAV distribution and biophysical parameters. The implications of this observation could provide scientific basis for protecting and using SAV resources, and offer a technological support for monitoring the distribution and dynamics of SAV on a large scale using remote sensing technology.