Object-Oriented Inversion Method And Application Of Grassland Vegetation Variables

Grassland is an important component of terrestrial ecosystem,which plays an important role in energy flow and material circulation,as well as the survival and development of human beings.The growth status of grassland vegetation can be directly or indirectly reflected by the biological,physical and chemical parameters(such as leaf area index(LAI)and canopy water content(CWC)).Therefore,it is of great scientific significance and practical value to study the vegetation parameters to dynamically monitor the ecological environment of grassland.It also provides scientific decisions for relevant departments.In this study,we use multi-source remote sensing data and field measurements to retrieve the LAI and CWC in Qinghai Lake watershed by using the object-oriented method.To demonstrate the feasibility and effectiveness of the proposed method,a traditional(pixel-based)physical model and the neural network inversion methods have also been conducted.Meanwhile,this thesis uses the domestic satellite data(HJ-1 and GF-1)to retrieve the grassland vegetation LAI,which aims to explore the feasibility and effectiveness of the application potential of the domestic data and the object-oriented method.The major works and accomplishments are listed as follows:(1)The Landsat-8 OLI remote sensing data are used to retrieve the LAI and CWC of grassland vegetation in Qinghai Lake watershed based on the object-oriented method and the look-up table(LUT)algorithm.This method can effectively improve the inversion accuracy by taking into account the spectral information of neighboring pixels.In this thesis,sensitive analysis of model input parameters is carried out from the perspectives of quantitative and qualitative in order to solve the ill-posed inversion characteristic and the grassland vegetation heterogeneity of the retrieval model.The study area is classified into two parts.One part is considered as sparse area and the other one is classified as dense area.Two LUTs have been established for these two vegetation areas,respectively.The retrieved results are very promising in comparison with the field measured LAI and CWC.The coefficient of determination(R2)of LAI and CWC reached up to 0.88 and 0.81,and the root-mean-square-error(RMSE)is 0.59 and 67.31 g/m2,respectively.They both show a higher inversion accuracy,which verifies the validity of the proposed method.(2)The same remote sensing data(Landsat-8 OLI)are used to retrieve the LAI and CWC of grassland vegetation in Qinghai Lake watershed based on the pixel-based physical model method and the neural network method.In this thesis,these two methods are compared with the object-oriented method,which shows the superiority of the latter.The comparisons between the retrieved results of the two methods and the field measures show: in the pixel-based method,the R2 of the measured LAI and CWC between retrieved LAI and CWC reached to 0.87 and 0.78,respectively,and the RMSE was 0.62 and 80.11 g/m2,respectively;as to the neural network method,the R2 reached to 0.84 and 0.72,respectively,and the RMSE was 0.65 and 99.95 g/m2,respectively.Compared to the retrieved results of the proposed method,it is shown thats object-oriented method has better retrieve accuracy.(3)The remote sensing satellite data of China,including HJ-1 and GF-1,are used to retrieve the LAI in the same study area based on the previous three methods.On one hand,it is used to explore the data quality and application potential of the domestic satellite data and foreign remote sensing data.On the other hand,it is also intended to make further efforts to verify the feasibility and effectiveness of the proposed method.When using the same remote sensing data,the results show that the object-oriented method gives the best retrieve performance,while the pixel-based physical model method comes second.When using the same retrieval method,it shows that the application potential of the Landsat-8 is highest,while GF-1 is in the next place.