Phragmites Australis Extraction And Aboveground Biomass Estimation In Typical Regions Of Northeast China, Using Remote Sensing Technology

Wetlands ecosystem is one of the most productive ecosystems of thenature and the most important living environment of mankind. As one ofthe most active component in wetlands ecosystem, wetlands vegetationplays an important role in energy flow and matter cycling of the wetlands.At present, based on remote sensing technology, classification andmapping of wetlands vegetation, together with its physiologicalparameters inversion, are the hotspot study in the field of wetlandresearch. The Northeast region of China is the sensitive area of globalchange. As one of the main types of wetlands in northeast China,Phragmites australis wetland, which is rich in bird resources, become themain breeding grounds for waterfowl in northeast Asia and have greatecological and economic value. So, based on remote sensing technology,extracting the distribution information of Phragmites australisefficaciously and inversing the biomass accurately, have greatsignificance to wetlands protection of northeast region. In this paper,combined with field observation data, HJ-1A/B CCD data,Landsat-TM/OLI data and MODIS-NDVI data, based on the technology of Remote Sensing, GIS spatial analysis method and SPSS statisticalprocessing, we extract the distribution information of Phragmites australisefficaciously firstly, then inverse its biomass accurately. There are somebasic conclusions as follows:(1) Based on the multi-temporal HJ-1A/B CCD data andobject-oriented classification method, we extract the distributioninformation of Phragmites australis. The results showed that: Theextraction accuracy of four typical area was between68%to89%.Among them, the western of Songnen plain and Liaohe river delta had thehighest accuracy (>85%), the followers were Hulunbeir grassland (81.6%)and Sanjiang plain (68.5%). By adopting the method of decision treeclassification, extraction accuracy of Phragmites australis in Sanjiangplain is79.63%, improved by10%than before. The area of Phragmitesaustralis in four regions is in the following sequence: western of Songnenplain (193943.5hm2), Liaohe river delta (58669.2hm2), Hulunbeirgrassland (27542.4hm2) and Sanjiang plain (11007.5hm2).(2) Based on accurate sampling model, it had an significant linearcorrelation between stems biomass and leaves biomass of Phragmitesaustralis (R2>0.90) and it was the same between stems or leaves biomassof Phragmites australis and its aboveground biomass. By the inaccuratesampling model, the linear correlation between stems biomass and leavesbiomass of Phragmites australis, together with that between stems biomass and aboveground biomass is not so significant（R2≈0.50）.However, it also presented an significant linear correlation between stemsbiomass and aboveground biomass of Phragmites australis. There is notan significant linear correlation between the stems biomass obtained byaccurate sampling model and that obtained by inaccurate sampling modeland it was the same to the aboveground biomass（R2<0.50）. However, interms of leaves biomass obtained by the two sampling model, there isalso an significant linear correlation (R2=0.97). In addition, it is the sameto the aboveground dry weight and wet weight (R2=0.83).(3) Based on accurate sampling model, both the SCRM andMLRM biomass inversion models achieved higher estimation accuracy(about80%), but MLRM model had less prediction error and higherestimation accuracy than SCRM model. The estimation accuracy ofSCRM and MLRM models based on inaccurate sampling were inferior tothat based on accurate sampling. Even so, the estimation accuracy of themodels for leaves biomass inversion in accurate sampling was close tothat in inaccurate sampling. For the two sampling types, the calculationresults of the aboveground biomass estimation models with leavesbiomass as the independent variable were very similar to the results ofSCRM and MLRM model. The results showed that the abovegroundbiomass estimation models with leaves biomass as the independentvariable can have a very acceptable accuracy. (4) The best estimation model of Phragmites australisaboveground biomass was the MLRM model with the independentvariables of Band3（B3）, Band （4B4）and NDVI（p＜0.01, R2=0.666,Accuracy=80.80%). The aboveground dry weight and wet weight ofPhragmites australis in four regions is in the following sequence: westernof Songnen plain (2641.40t.,5778.78t.), Liaohe river delta(1383.83t.,3326.51t.), Hulunbeir grassland (565.68t.,1731.12t.) and Sanjiang plain(274.66t.,1163.63t.).