Constraining plant functional types in a semi-arid ecosystem with waveform lidar

摘要: Abstract Accurate classification of plant functional types (PFTs) reduces the uncertainty in global biomass and carbon estimates. Airborne small-footprint waveform lidar data are increasingly used for vegetation above-ground estimates at a range spatial scales woody or homogeneous grass savanna ecosystems. However, gap remains understanding how features represent ultimately can be to constrain PFTs heterogeneous semi-arid This study evaluates performance six major PFTs, including shrubs trees, along with bare ground Reynolds Creek Experimental Watershed, Idaho, USA. Waveform were obtained NASA Snow Observatory (ASO). From these we derived two (1 m 10 m rasters) by applying Gaussian decomposition frequency-domain deconvolution. An ensemble random forest algorithm was assess select most important features. Classification models developed outperformed those 1 m (Kappa (κ) = 0.81–0.86 vs. 0.60–0.70, respectively). At resolution, height improved PFT accuracy 10% compared analysis without inclusion heights other decreased 4%. Pulse width, rise time, percent energy, differential target cross section, radiometrically calibrated backscatter coefficient both scales. A significant finding is that clearly differentiated from using pulse width. Though overall ranges between 0.72 0.89 across scales, shrub showed 0.45–0.87 individual success 1 m, while tree high (0.72–1.0) 10 m. We conclude characterize this similar ecosystems resolution. Furthermore, such as width terrain modeling environments where returns close time space. The dependency on resolution plays critical role tree-shrub co-dominant