Estimation Of The Forest Canopy Height Based On Multi-source Remote Sensing Data

The precise estimation of forest canopy height is of great importance because it can increase the accuracy of biomass estimation used in the study of the global carbon cycle,forest productivity,and climate change.Traditional manual methods of measuring forest structure parameters are labour-intensive and time-consuming,which makes it difficult to obtain forest parameters in regional scale.With the development of remote sensing technology,multi-source remote sensing data such as laser radar and optical remote sensing data have been used to estimate the regional scale forest structure parameters.Many research results have been obtained,but the accuracy of forest crown height estimation under complex terrain conditions still needs to be further improved.In this paper,based on the analysis of the feasibility of extraction of forest canopy height in complex topography condition using SPOT5,ZY3 and GLAS,the arithmetic of processing the GLAS waveform data was implemented and the model of estimation of canopy height within the footprints using GLAS data was established,which was then improved by adding optical remote sensing data,and finally the distribution map of forest canopy height with a resolution of 10 m was made.Results show that the combination of GLAS data and optical remote sensing data can be used to estimate forest canopy height in regional scale and the main results and conclusions are as follows:(1)The Gaussian low-pass filtering and wavelet denoising methods were used to denoise the GLAS waveform and the results show that the wavelet denoising preserves the waveform information better than the Gaussian filter.Waveform parameters including the waveform extent,the leading edge extent,the trailing edge extent and the half of median energy from the denoised GLAS waveform were extracted.Topographic index methods and non-linear regression methods were used to establish canopy-height estimation models.The model of coniferous forest and mixed forest had better simulation effect on canopy height and less effective for broad-leaved forest.Based on the topographic index model,the model built by introducing the leading edge and the trailing edge of the waveform has the best effect.The R~2of coniferous forest model increased to 0.950 and the RMSE was 1.466 m.The R~2 of coniferous and mixed forest increased to 0.911 and the RMSE was 2.202 m.This shows that based on the terrain index model,waveform leading edge extent and waveform trailing edge extent can effectively improve the estimation accuracy of the canopy height.(2)Optical remote sensing parameters were introduced to improve the canopy height estimation model and the results show that the accuracy of each model with SPOT5 texture parameters is improved.The coniferous forest and mixed forest had the best simulation results.The R~2 increased from 0.950 to 0.963 and the RMSE decreased from 1.412 to 1.228.Compared with the previous studies based on waveform parameters and the topographic index,the R~2 of the regression model increased from 0.763 to 0.798 and RMSE decreased from2.223 to 2.053.The estimated effect has been significantly improved.(3)Combined with GLAS data and optical remote sensing data,the extrapolation scale of forest canopy height was determined to be 50×50 m,and BP neural network model was used to estimate the forest canopy height and the distribution map of forest crown height at the scale of 50×50 m was made.The results of accuracy verification showed that R~2 of coniferous forest was 0.957,RMSE was 1.636 and R~2 of broadleaved forest was 0.814,RMSE was 2.110,R~2 of mixed forest was 0.964 and RMSE was 1.006.The multi-regression method was used to establish the GLAS crown height adjustment model within the footprint.The adjustment of canopy height of each forest type on a pixel-by-pixel basis was completed.Finally,the distribution map with a resolution of 10 m of forest canopy height was made.In summary,the results of this study can provide a methodological basis for the study of GLAS data combined with optical remote sensing data to estimate canopy height,which can be used to estimate forest biomass and carbon storage and provide more accurate scientific data for forest carbon cycle research.