Estimation of forest canopy height in mountainous areas using ICESat-GLAS data

Forest canopy height is an important input variable to derive a set of essential parameters of forest stands, which is yet costly and time consuming when measured based on ground surveys. The satellite-based laser scanner data from ICESat-GLAS provide a 3D representation of gorund objects by measuring the distance from spacecraft to the objects on the earth surface. By means of these data, this study aims to estimate forest canopy height in a portion of mountainous Kheyroud experimental forests in north of Iran. An ICESat-GLAS dataset was analyzed. Several metrics including waveform extent, lead-edge extent and trail-edge extent were extracted from waveform data, and a terrain index was additionally calculated based on a digital elevation model at the location of all laser footprints. Forest canopy height was retrieved by calculating difference between signal begin and ground peak (direct estimation) and regression models (indirect estimation). For fitting the regression, a number of 330 highest trees were measured in 33 circular plots (70 meter diameter) which were collocated with LiDAR footprints. The directly estimated height produced and RMSE values of 0.56 and 10.32 m, respectively. Compared to this, regression models based on combined waveform metrics and digital elevation model provided better results. Best model fit with lowest AIC= 204.55 was achieved using waveform extent and terrain index variables (=0.82; RMSE= 6.16m). The ICESat-GLAS therefore concluded to be able to retrieve a relatively accurate estimate of maximum forest canopy height in such steep mountainous area, especially on small scales. Better results are assumed to be achieved using other statistical methods, as well as by an improved waveform processing techniques.