| Muddy tidal flats mainly composed of fine particles mud. It’s a typical coast type of North Jiangsu plain. Restricted by natural conditions, the use of conventional methods of flat elevation cost high, and the data replacement cycle very slow. Remote sensing have the characteristics of informative, synchronization in extent, multiscale in space and less restricted by ground conditions. It has a great advantage of acquire tidal flat profiles terrain information. The existing remote sensing research on tidal flat is given priority to multispectral remote sensing. As the ability of high spectral resolution within the scope of the visible light to near infrared wavelengths to get information, the hyperspectral remote sensing cause the attention of more and more researchers. The use of hyperspectral remote sensing technology to extract tidal flat information has become an important research direction of tidal flat remote sensing.Taking Wanggang tidal flat, Dafeng city, Jiangsu province as the study area. The spaceborne Hyperion hyperspectral image Lidar and sediment samples as the data sources. Study the quantitative retrieval of tidal flat beach surface elevation by Hyperspectral image. Firstly, the paper established the statistical correlation model between hyperspectral reflectance and the measured height, used the "reflectivity method" to elevation remote sensing retrieval; Secondly, considered the tidal flat sediments median diameter and water content are important factors which affect tidal flat reflectivity, sediment size and moisture content of the change have a close relationship between elevation. Based on the spatial distribution of the layer of median diameter and water content which was extracted by the remote sensing method, Two quantitative models between median diameter and elevation and water content and elevation were built respectively, realized the "median diameter method" and "water content method" of the remote sensing application. At the same time seek Mechanism explain for the inversion elevation on tidal flat for remote sensing method. Finally, under the premise of consistent in the sediment types, the frequency ratio between three kinds of elevations and sediment types, analysis the space distribution relations between remote sensing inversion elevation and sediment, the applicability of the three height inversion methods were proposed, and the feasibility of the remote sensing inversion terrain was clarified. The main content of this research and related conclusion are as follows: First, the method of elevation inversion via remote sensing on tidal flat was researched according to their surface reflectivity. Two kinds of inversion models namely, single-band linear regression model(SBLRM) and multi-band linear regression model(MBLRM) were built and the tidal flat elevation was inverted. Results showed that:(1) Surface reflectance in short-wave infrared ranges was sensitive to the change of elevation on the tidal flat. Those bands with central wavelengths at1164nm、1275nm and1336nm were ideal bands for the inversion of elevation. Using these ideal bands as inversion factors, SBLRM and MBLRM could be built to successfully simulate the elevation distribution character of the tidal flat. But compared with the SBLRM, MBLRM could not evidently improve the elevation prediction accuracy.(2) Topography such as sandy ridges and tidal creeks on the tidal flat would greatly impact the value of surface reflectance by influence the spatial distribution of sediment size and moisture content, so tidal flat needed to be divided into flat and undulating area spatially to build elevation reversion models separately. It is contrasted that before the partition tidal flat overall elevation modeling and verification accuracy were-0.302and-0.473, respectively, after the partition, the modeling and verification accuracy on north beach were0.559and0.526, on south beach were0.762and0.699, the height inversion accuracy was obviously improved after partition.(3)From the integral form contrast, reflectivity method inversion of tidal flat elevation and the measured elevation have same roughly trend that from land to the sea the elevation gradually reduced, but the height inversion of fine particles on the edge of the trench sedimentary area was poorer, and the inversion elevation was flat.(4) From the view of contrastive analysis for the form of section elevation, average elevation differences in study area was135.5cm, in the inversion elevation was76cm, mean absolute error(MAE) and mean relative error(MRE) on north beach were7.75cm,6.2%, and on south beach were36.75cm,141%, the root mean square error(RMSE) between remote sensing simulated and observed elevation was9.5cm on north beach, and45.2cm on the south beach. It proved that the north beach inversion cross-section was more close to real topography than south beach.Second, the method based on the median diameter remote sensing inversion study of tidal flat elevation. Measured161sampling points of the median particle size of sediment data, combined with high reflectivity spectrum image, established the inversion model of median diameter, and the tidal flat median diameter distribution was obtained. On this basis, section median diameter were extracted from remote sensing of the median diameter layer, the statistical model between median diameter and elevation was set up, and the tidal flat elevation inversion from remote sensing median diameter was realized. Results showed that:(1)The inversion median diameter on tidal flat by the model of single band that used the reflectivity of993nm, and the modeling and verification accuracy were0.891,0.842, the spatial distribution of median diameter was inverted effective. At the same time contrast the inversion effects of coarse particle size was superior to small particle size.(2)Application "median diameter method" inversion the tidal flat elevation, the modeling and verification accuracy on north beach were0.485,0.438, and0.627,0.636on the south beach.(3)The model accuracy of the "reflectivity method" and "median diameter method" were chose, it found that the inversion results of "median diameter method" inversion elevation was closer to the actual terrain on which the tidal flat deposits of fine grain.Third, the method based on the water content remote sensing inversion study of tidal flat elevation. Firstly,2282nm,1044nm and1296nm bands were selected to build NEWCI Index (Normalized Enhanced Water Content Index) and set up moisture content model, then applied in the tidal flat hyperspectral image, got the tidal flat sediments water content distribution; The water content section was extracted and built model with the measured elevation, obtained inversion elevation of tidal flat. Results showed that:(1) NEWCI was the effective factor of the inversion sediments moisture content on tidal flat, the modeling and verification accuracy of inversion water content was0.837,0.842. The inversion model was reliable and could carry out to invert the tidal flat sediments moisture content.(2)In the end, the modeling and verification accuracy of "water content method" inversion elevation were0.533,0.48on north beach, and0.71,0.716on the south beach.(3)The model accuracy of the "water content method" was slightly lower than the "reflectivity method", and a little higher than the "median diameter method", it found that the inversion results of "water content method" inversion elevation was closer to the actual terrain on which the tidal flat region of higher water content.Fourth, analysis the spatial distribution between remote sensing inversion elevation and sediment. ArcGIS spatial analysis technology was used to calculate frequency ratio between elevation and sediment, the spatial relationships from them were discussed. Results showed that:(1)A part of elevations for sediment type of sand, silty sand were reflected by three kinds of inversion elevation, only the "median diameter method" invert elevation reflects the measured elevation of-0.01-0.22m which located in the sandy silt.(2) Two area of elevation values of silt which located in-0.01-0.22m and more than1.31m did not reflected by the three height inversion methods. The sand content of silt was least, and the particle relative fine from four types of sediment. The composition of particle size and water content was more complicated than other types of deposits, so the inversion results were not better.(3) Silty sand has a high correlation with terrain in the measured elevation and three types of remote sensing elevation are the performance part of the silty sand area confirmed it. The differences part of correlation between elevation and sediment types was the error that from three kinds of inversion elevation to the measured.Fifth, in general, the inversion elevation precision of "reflectivity method"," median diameter method" and "water content method" were close. There was a certain degree of error due to the median diameter and moisture content modeling data which was obtained by hyperspectral imaging inversion. In the modeling process, the error may be involved in the transfer, and affect the inversion result. From the result of spatial processing and analyzing, the type of tidal flat sediments could be combined to elevation inversion in tidal flat. The analysis indicated that the remote sensing inversion method could effectively simulate average trend for tidal flat topography. It could get synoptic terrain information in this area and provide reliable basis for erosion and deposition analysis on extensiveness tidal flat. |
PDF Full Text Request