Abstract
Geoscience laser altimeter system (GLAS) data have been widely used for forest aboveground biomass (AGB) estimation, but there is no consensus on the optimal metrics. To explore whether a few optimal GLAS metrics could generate accurate AGB estimates, we proposed five metrics and explored their combinations with ten existing ones. The importance of these metrics was measured according to their contributions to changes in the cross-validated mean-squared error. The two to eight most important metrics were then selected to develop AGB models, and their performances were evaluated using field AGB. The optimal combination of GLAS metrics was finally used for AGB estimation at GLAS footprints from 2004 to 2007 within a 2°×2° spatial extent in Tahe and Changbai Mountain, China. The results showed that four GLAS metrics, including our proposed metric CRH25 (25th percentile of canopy reflection heights) combined with Lead, quadratic mean canopy height, and H75, yield the best AGB estimates, with an R 2 of 0.61±0.15 and RMSE of 52.20±23.50 Mg/ha, and the inclusion of more GLAS metrics did not improve the results. The estimated forest AGB in Tahe was 89.03±19.16 Mg/ha and 103.07±23.42 Mg/ha in Changbai Mountain. In both regions, the annual average forest AGB estimates for 2005 were higher than the AGB estimates for 2004, 2006, and 2007. The results of this study suggested that a few waveform parameters could enable the accurate estimation of forest AGB. Moreover, this study indicated that GLAS data might be able to monitor forest AGB changes, which require further investigation.
Highlights
F OREST aboveground biomass (AGB) plays an important role in the global carbon cycle and climate change studies, but its magnitude, patterns, and uncertainties remain poorly quantified [1]–[3]
We evaluated the performances of Metrics II, Metrics IV, five importance metrics among the ten metrics (CRH25, CRH50, CRH75, mean canopy reflection (MCR), quadratic mean canopy reflection (QMCR), H25, H50, H75, mean canopy height (MeanH), and quadratic mean canopy height (QMCH)), and the optimal combination of geoscience laser altimeter system (GLAS) parameters from Metrics III in estimating forest AGB, using the optimal AGB modeling algorithm among random forest (RF), support vector regression (SVR)-linear, and SVR with the radial basis function kernel (SVR-RBF)
When Metrics II was added to Metrics I, no substantial changes in the accuracy of estimated AGB by RF, SVR-linear, and SVR-RBF algorithms are found for most of the 100 points, as shown in Figs. 3 and 4, indicating that it might be essential to perform feature selection or select optimal variables when GLAS parameters are used for predicting forest AGB
Summary
F OREST aboveground biomass (AGB) plays an important role in the global carbon cycle and climate change studies, but its magnitude, patterns, and uncertainties remain poorly quantified [1]–[3]. The science community has paid much attention to forest AGB estimates from multiple remote sensing datasets, including optical images, synthetic aperture radar (SAR), and light detection and ranging (LiDAR) data, or a combination of them [4]–[11]. Groundbased and airborne LiDAR has been successfully applied for individual tree species classification [19], deriving leaf area index [20], [21], analyzing forest canopies [22], [23], and estimating forest structure and biomass [24]–[26] but is only feasible at local scales at effective costs. Spaceborne LiDAR provides the solution for large-area or even global forest AGB estimates [27]. ICESat-2 and GEDI have covered a relatively shorter period compared with GLAS data
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
