Study On Remote Sensing Information Extraction And Landscape Pattern Dynamic Change Of Wetland In Nansihu Area

As an important and special ecological system on the earth, wetland is one of important living environment of human being in the transition zone between terrestrial ecosystem and aquatic ecosystem. A syetematic research of wetland is important to wetland resources protection and reasonable exploitation, maintaining fundamental eco-process of wetland, improving ecological environment,sustainable development of society and economy. Wetland ecological system is hard to approach and measuring on the ground, most of wetlands are difficult to fieldwork, and, wetland changed obviously, traditional research method of wetland cannot meet the requirements. Today, we have obtain the global coverage images of different periods, different kinds and different scales, so, remote sensing technology becomes the most ideal method of investigating wetlands and monitoring their dynamic change.Taking the wetland in Nansihu area, which is the largest lacustrine wetland in Shandong as the study case, we groped for information extracting technique of wetland in Nansihu area as well as landscape pattern change. The RS data we used is three Landsat TM images of 1987, 1997, 2004 and two SPOT images obtained in 2006, some related thematic information and survey data were used too.Based on completely analyzing the study area, this paper adopted the below methods:(1) Wetland information extraction based on TM images. Wetland information was extracted using decision tree classification based on three TM images of 1987,1997 and 2004. Decision tree was built by feature spectrum computation, eigen variable selection, K-T transformation, NDVI variable. This was applied to wetland information extraction, and the accuracy was evaluated. The classification accuracy of TM images of 1987,1997 and 2004 were 87.39%, 83.5%, 85.42%, and got the better classification results.(2) Wetland information extraction based on SPOT images. Wetland information was extracted using object-oriented approach based on two SPOT images of 2006. The studied region was divided into two sub-areas: paddy area and lake area, then they were classified by multi-subarea object-oriented approach. Using nearest neighbor rule and membership function rule synthetically, wetland information was extracted based on multi-scale segmentation, characteristic parameters selection and calculation. The classification accuracy of SPOT image of 2006 was 86.23%, and got the better classification results.Analyzed the landscape pattern dynamic change of wetland in Nansihu area according to the above classification results. Choosed 90m as the best spatial analysis scale based on landscape metrics calculation. Using SPOT images extraction result as the landscape pattern status analysis data, TM images extraction result as the landscape pattern dynamic change analysis data, this paper analyzed the landscape pattern of wetland separately from landscape level and class level. The results showed that, the areas of natural wetland had decreased in Nansihu area, the areas of constructed wetland increased, the areas of constructed wetland decreased, the landscape fragmentation increased in recent 20 years.Finally, this paper gave a brief analysis for the driving mechanism of landscape pattern from natural factors and human factors. It was indicated that, because of the significant changes in hydrological conditions and effects of human activities in recent years, large areas of wetland in Nansihu area were reduced, the landscape fragmentation and spatial heterogeneity were improved, eco-environmental problems were increasingly serious.