Abstract: Objective Forest structure characteristic is an important index to evaluate forest ecological function, which can directly present the growth status of forest. It is not only the research object in the fields of forestry, ecology and geoscience, but also an important input parameter of many terrestrial ecological models. How to obtain the characteristics of large-scale forest structure effectively has important scientific research value. Based on MISR(multi-angle imaging spectro-radiometer) satellite remote sensing data and UAV(unmanned aerial vehicle) photogrammetry data, SGM(simple geometric optical model) and the optimization methods of GRG(generalized reduced gradient) were used to obtain the key structure information of Populus euphratica forest in Tarim River, so as to provide a new method for the quantitative study of forest remote sensing.Method Taking the typical riparian forest near Yingsu in the lower reaches of Tarim River as the study area, based on field investigation, UAV remote sensing aerial survey, multi-angle satellite remote sensing, the inversion of SGM is carried out, the process is divided into three steps: firstly, the multi-angle reflectance data and forest structure parameters of the train dataset are input into the SGM model as known fixed variables, and the coefficients of Walthall model are retrieved by GRG optimization algorithm. Secondly, the regression equation of Walthall model coefficients is established by using the AMBRAL model coefficients of the train dataset and the multi-spectral reflectance data of MISR AN camera. Thirdly, an regression equation of the Walthall model coefficient is used to obtain the soil background reflectance of the test dataset, and the structural parameters of the test dataset are retrieved by GRG nonlinear optimization algorithm, Finally, the accuracy is verified and evaluated by UAV measurement data.Result The tree height, crown diameter and density data extracted by OBIA (object based image analysis) technology were compared with the measured samples. It was found that the R² of the height, crown diameter, density obtained by UAV oblique photogrammetry between the field-measured value was 0.90, 0.84, 0.94 and RMSE(root mean square error) was 0.45 m, 0.68 m and 4.25 trees·hm^-2 respectively. Compares the SGM simulated value with satellite observation reflection value of the train dataset, the maximum value of R² is 0.99, the minimum value of R² is 0.72 and the mean value of R² is 0.92. The tree height and coverage obtained from the inversion of the train dataset are consistent with the reference value. R² reaches 0.99 and RMSE is 0.73%. The consistency of tree height is slightly poor, R² is 0.66 and RMSE is 0.61m. c₁ parameter of Walthall model has the greatest impact on FVC and tree height, followed by c₄, c₂ and c₃. c₁-c₄ have relative little effect on the change of density. By establishing the multiple linear regression equation between the reflectance of AN camera multispectral band and AMBRALS model coefficients with c₁-c₄, the c₁-c₄ parameters of the train dataset are predicted. Compared with the inversion values, the adjusted R² of c₁-c₄ parameters are 0.78, 0.98, 0.59 and 0.75 respectively. Compared with the UAV measured data on the 275 m² scale, the R² of linear models of FVC, tree height, density and crown diameter of Populus euphratica obtained by MISR multi-angle satellite remote sensing were 0.54, 0.47, 0.41 and 0.24 respectively, and the RMSE were 3%, 0.76 m, 112 trees and 0.31 m respectively; MRE(mean relative error) was 24.7%, 8.9%, 22.5% and 10% respectively. When the power function model is used, the R² of FVC, tree height, density and crown diameter are 0.80, 0.53, 0.55 and 0.30; RMSE is 1%, 0.4m, 33 trees, 0.16 m; MRE is 10%, 5%, 6%, 6% respectively.Conclusion The structural data of Populus euphratica forest obtained by UAV can be used as the calibration parameters of the model and the accuracy verification instead of the field-measured data. SGM model can well simulate the anisotropic reflection characteristics of observation samples. The accuracy of background reflection simulated by Walthall model has a great impact on the final inversion results of structural parameters. The combination of MISR multi-angle satellite remote sensing and UAV technology can obtain more accurate forest structure information on a regional scale.

Key words: forest structural parameters, Populus euphratica, multi-angle satellite remote sensing, simple geometric-optical model (SGM)