Forest Structure Parameters Estimation Based On Airborne X-Band Single-pass InSAR Data

Forest height and forest above-ground biomass(AGB) are important structure parameters of forest. Its accurate measurements would benefit the activities of forest production and operation, study of global climate change and global carbon cycle, analysis of the role of forests in carbon storage and moisture change, and so on. SAR data has been an important data source for forest structure parameters estimation. Based on the wavelength, it can be divided into two categories: the long wavelength SAR and the short wavelength SAR. In previous works forest parameters have been successfully retrieved based on the long wavelength SAR data, and the potential on forest parameters estimation for the short wavelength SAR has also been shown. Due to the poor penetration and tendency to interacting with forest canopy, the short wavelength SAR data is prone to reflecting canopy information. However, the phase center of the ground topography could not be detected, thus it is against forest height estimation. In addition, it could not interact with the trunk which is the body of the forest AGB, thus it is also against forest AGB estimation. In view of the problems of forest parameters estimation with X-band data, this paper proposed the methods of estimating forest height and forest AGB with the airborne X-band dual-antenna InSAR data with different incidence angles. The major research contents include the following aspects:1) The relationship between penetration depth of X-band and the incidence angleConsidering that for the poor penetration, X-band data have advantages in obtaining the phase center of canopy but could not easily obtain the phase center of ground topography, this paper proposed the method of estimating forest height based on the phase difference. Forest height was obtained by making the DSM(Digital Surface Model) acquired from InSAR data minus the DEM(Digital Elevation Model) acquired from LiDAR(Light Detection and Ranging). The Accuracy of the estimated forest height was validated with the LiDAR CHM(Canopy Height Model). The results show that the highest R2 is 0.82 and the lowest RMSE is 1.1 m. With the analysis of all the estimated forest heights with different incidence angles, we find that the height of the phase center is affected by incidence angle. The height of the phase center is higher for the bigger incidence angle, while the height of the phase center is lower for the smaller incidence angle.The results show that for the sparse and lower forest in the test site, the X-band microwave has a certain penetration and the height of its phase center is lower than the canopy top height even its penetration is poor. In addition, since X-band microwave is easy to scattering and sensitive to forest structure, the penetration decreases with the incidence angle increase.2)Forest height estimation based on coherence without high-precision DEMThe high-precision DEM is needed for forest height estimation based on phase difference, which restricts the application of X-band in forest height estimation. Ignoring the phase information, forest height estimation method based on coherence amplitude had been studied. For the X-band single-pass InSAR data is not influenced by temporal decorrelation and the ground under forest is hard to observe, we could assume that the most contributions of interferometric coherence comes from the volume decorrelation. Thus the forest height estimation model based on coherence amplitude is deduced from the vegetation coherence model by using Fourier-Legendre polynomials and selecting the zeroth-order Legendre polynomial only. The accuracy of forest height estimation was then validated with the test samples. It is shown that the estimation accuracy of this method is similar to the method based on phase difference. And the estimated results are in accordance with the true forest height and needn't calibration.The results show that we could obtain the estimation result with a higher accuracy only based on coherent amplitude of X-band single-pass InSAR data. In addition, the estimated height needn't to be calibrated with true data and the high accuracy DEM is not required. This method is more applicable and generalized for the forest in the northeast Daxinganling of China.3) Forest height estimation from multi-track InSAR data without high-precision DEMSince the penetration decreases with incidence angle increases, this paper proposed the forest height estimation method with multi-track InSAR data. The forest height was obtained by making the height close to canopy minus the height close to ground. The accuracy validation shows that the estimation accuracy within a pixel for statistical analysis was lower, while the accuracy within a certain size window was higher. The accuracy went higher and tended to be saturation, with the size of the window enlarging, since the heights of canopy and ground would be close to the true height.The results show the feasibility of the estimation method with multi-track InSAR data. But due to the difference of the incidence angle is not significant, there is a bigger error in the estimated results. This method should be further studied using simulation data.4) Forest AGB estimation with multi-track InSAR dataSince the penetration decreases with incidence angle increases, this paper proposed the forest AGB estimating method with multi-track InSAR data. This method built the forest vertical structure profile using the retrieved heights from different incidence angles, and extracted the profile parameters to estimate forest AGB. As the different profile structures were obtained from different angle combinations and affected the correlation between the profile parameters and the true forest AGB, this paper analyzed and selected the best incidence angle combination. The profile structure parameters were extracted and the correlation between the profile parameters and LiDAR forest AGB were analyzed. And then the forest AGB estimation model was built based on the profile parameters. The estimated results are validated against the LiDAR forest AGB. It is shown that the estimated results with a higher accuracy could be obtained by the profile parameters compared to the result only based on the forest height.The results show that the features of forest vertical structure distribution could be obtained with multi-track InSAR data and be used to estimate forest AGB. The accuracy of the estimated forest AGB based on the vertical structure profile parameters is higher than that based on forest height. The estimation accuracy could be improved by combining the features of forest vertical structure distribution.The innovations in this paper lie in three aspects. Firstly, the relationship between the penetration of the X-band microwave and the incidence angle were analyzed based on different incidence angles. The sensitivity of the X-band data to forest structure has been revealed, which would contribute to understanding the scattering mechanism of the X-band data in forest. Secondly, based on the characteristics that the penetration is different with different incidence angles, the forest height estimation method was proposed with multi-track InSAR data. It provided a new solution for forest height estimation without the high precision DEM. Thirdly, based on the characteristics that the forest structure features are different with different incidence angles, the forest vertical structure profile was built with multi-track InSAR data. And the profile parameters were defined and used to estimate forest AGB. This method provided a new idea for forest AGB estimation with short wavelengh InSAR data and provided a base for developing the forest AGB estimation model with multi-track InSAR data.