Research On Estimation Of Coastal Wetland Vegetation Above-ground Biomass With Airborne Hyperspectral And LIDAR Data

A lot of reclamation and transformation was makes our coastal wetlands degradation, we lost a large number of local wetland vegetation and ecological environment continues to deteriorate since 1950. It was leading to alien species invasion disaster and coastal disasters frequently. Therefore, it has important significance of control the wetland vegetation distribution and changes effectively for the research and protection of coastal wetlands. Hard to get data, heavy workload and small scales of such problems exist in the traditional research methods of wetland vegetation; The remote sensing, as a new kind of wetland research means, it can be timely and accurate access to object feature in the target area. However, the growth is relatively exuberant and the density is bigger for most of the wetland vegetation in the mature stage. There were serious phenomenon of "with synonyms spectrum, foreign body with spectrum" and "high coverage saturation" in the category distinction and biomass inversion when we based on the method of Multi-spectral quantitative remote sensing. It makes the two-dimensional image information of ordinary multispectral and hyper-spectral spectrum has been unable to meet the requirements of precision.For the above questions, this Paper fully integrated information on the three-dimensional structure of the LIDAR data, hyperspectral data and detailed two-dimensional spectral information and rich field observation data of Wetland Vegetation in Da-Feng City and carried out a large number of experimental studies. Firstly, we distinguished the vegetation and non-vegetation by the features of the vegetation index, and we accessed to the vegetation of the study area distribution map for the fine classification of wetland vegetation. Secondly, we selected 21 spectral features and one highly characteristic by the characteristics of the target object, the cart decision tree classification algorithm used for the subdivision of wetland vegetation, and it contains reed, Artemisia halodendron and Spartina alterniflora. We accessed to the wetland vegetation classification map of the study area, and we set a contrast test to analysis the LIDAR data for the contribution of fine classification accuracy of wetland vegetation. Finally, we established the biomass inverse model of every class by the growth characteristics of wetland vegetation, and we obtained the biomass results of the reed, the Artemisia halodendron and the Spartina alterniflora. The main innovations and results of this dissertation are as follows:(1) We analyzed the effect of the 3D information integration of vegetation when we classed the wetland vegetation by the same classification algorithm (CART decision tree classification), classification sample and validation sample. It was found that the 3D information of vegetation can significantly improve the classification accuracy of wetland vegetation. It has significantly improved the phenomenon of "with synonyms spectrum, foreign body with spectrum". The overall classification accuracy has improved the 8.6%, the reed classification accuracy increased by 14.8% and the Artemisia halodendron classification accuracy increased by 4.4%, the classification accuracy of Spartina alterniflora increased by 8.9%.(2) We established the biomass inverse model of every class and the mixed biomass inverse model of the wetland vegetation by the same modeling sample data and verify the sampling data, respectively. Then, we verified the accuracy. It was found that the inversion accuracy of the classification biomass model was much higher than the total biomass model, when it compared with the classification biomass model and the overall biomass model. The classification biomass model has obvious advantages when we need to achieve the regional biomass of the multi-class.(3)It was found the results of this paper about the spatial distribution of wetland vegetation in the study area like the field research results when we analyzed the classification map of wetland vegetation. The reed space distribution was discrete, and the pixel continuity was small. The Artemisia halodendron space distribution was distributed in coastal and inland, the spatial distribution was more uniform, and the pixel continuity was good. The Spartina alterniflora was mostly distributed in the littoral coast, and the spatial distribution was uniform and continuous. However, the Spartina alterniflora space distribution was scattered in inland.(4)It was based on the pixel number when we analyzed the results of biomass by the statistical methods. We found that the reed biomass was mostly distributed between 0 kg/m2 to 31.25kg/m2, and the growth state was greatly difference. The total biomass of reed was 1866.86t. We found that the Artemisia halodendron biomass was mostly distributed between Okg/m2 to 6.25kg/m2, and the growth state was difference. The total biomass of reed was 1611.04t. We found that the Spartina alterniflora biomass was mostly distributed between Okg/m2 to 15.63kg/m2, and the growth state was uniform. The total biomass of reed was 681.18t.