Forest height estimation and inversion study of satellite-based X-band InSAR data

  Objectives:   Based on TerraSAR/TanDEM-X satellite-based single-polarization InSAR data, we studied the effects of estimation algorithm selection and coherence coefficient calculation method on the accuracy of forest height estimation results based on InSAR technology at different spatial scales. The methods applied in this paper including DSM-DEM difference method and SINC model method for forest height estimation and inversion. It is clear that the penetration capability of X-band affects the forest height estimation results when comparing the difference before and after the calibration of the estimated results using DSM-DEM difference method. Through the calculation of coherence based on intensity and interferometric coherence images and the calculation of coherence based on interferometric coherence images, the influence of the two coherence calculation methods on the estimation results of the SINC model height was clarified. The influence of forest height estimation algorithm on the estimation results was clarified by comparing the forest height estimation results of DSM-DEM difference method and SINC model method.   Methods:   The forest height estimation results of DSM-DEM differential method were somewhat underestimated compared with LiDAR CHM and the estimation accuracy improved gradually with the increase of scale.   Results:   The R² value was 0.64, RMSE value was 2.00 m, and ACC. Value was 72.43% when the sample scale reached 60m×60m. When the sample size was 60m×60m, compared with LiDAR CHM H100, the estimation results of SINC model based on the coherence calculated by the traditional method severely overestimated the forest height with R²=0.64, RMSE=2.00m, and ACC=72.43%. The SINC model estimation results based on the coherence calculated by the traditional method were in good agreement with LiDAR CHM H100, with R²=0.74, RMSE=8.42m, and ACC.=67.95%. While the SINC model estimation results obtained by considering only the coherence calculated by the phase method were in good agreement with LiDAR CHM H100, with R²=0.78, RMSE=3.09 m, and ACC.=81.66%.   Conclusions:   Both DSM-DEM difference method and SINC model method can obtain a reliable forest height estimation results, but DSM-DEM difference method model method based on coherence neither needs actual measurement data calibration nor relies on high precision DEM, which has more extensive practical value. For the SINC model, the forest height estimation results obtained by considering only the coherence of phase calculation had higher accuracy and were more suitable for the forest height inversion at regional level or global level.