Improved derivation of forest stand canopy height structure using harmonized metrics of full-waveform data

Abstract

A variety of prototypical metrics of ICESat-1 GLAS full-waveform data have been applied to estimate the forest canopy height over a footprint area and presented using the values of maximum canopy height, crown-area-weighted canopy height, or basal-area-weighted canopy height. The full-waveform pattern of a large footprint is changeable due to complicated interaction of canopy elements and terrain relief, consequently a noticeable bias may be introduced to the estimation. This paper begins with the approach of area-based growing stock volume calculation, developing harmonized metrics to integrate versatile signals in a denoised multimodal waveform of ICESat-1 data. These novel metrics were applied to derive a harmonized canopy height model for estimating the canopy height at footprint level and the distribution of canopy height at stand level. The performance of this model was evaluated with respect to the prototypical metrics derived canopy height model based on airborne LiDAR data determined canopy heights of a secondary evergreen mixed forest in a steep rugged terrain in central Taiwan. Footprint-level canopy height was evaluated via accuracy measures using a cross-validation method. Appropriateness of canopy height structure was examined via a post hoc test of the shape similarity for both ICESat-1 data-derived and airborne LiDAR-derived canopy height distribution models at stand level. Results showed that the harmonized canopy height model was able to achieve an accuracy of RMSE = 3.13 ± 0.08 m and RMSE% = 22.40 ± 0.55% as well as a significant reduction of bias. The canopy height distribution of the forest stand derived by harmonized metrics of ICESat-1 full-waveform data is statistically the same as that determined by airborne LiDAR data. Harmonized metrics provide compensatory information of trailing-edge extent and terrain index for describing canopy height. In contrast to the canopy height model using only prototypical waveform metrics, the harmonized metrics based canopy height model can mitigate the potential influence of terrain effects and dense canopy coverage on waveform pattern. Because a priori knowledge of terrain relief data is not required for modeling, the harmonized algorithm is useful for retrieving the trend of canopy height growth over the terrestrial forest ecosystem.