Plant functional types (PFTs) control the sensitivities of ecosystem models which predict the global carbon and water dynamics in response to climate change and disturbances. Vegetation vertical structure, density, and surface topography are significant indicators of different PFTs. Semi arid ecosystems with sparse and low-stature vegetation are continuously reshaped by climate driven disturbances (drought, fire, species invasion) resulting different PFTs across space and time and hence act as the controller of global carbon variability. High resolution waveform lidar have been proved as a successful tool differentiating PFTs in many forest ecosystems with its additional data matrices (pulse width, backscattered cross section, differential target cross section, rise time, percent energy at specific heights etc.) that capture the vegetation biophysical characteristics. In addition, waveform derived variables provide information of spatial density (ex: open, dense, and semi-dense), and surface topography (ex: elevation, slope, aspect, curvature etc.). To assess the spatial distribution of growth form (trees (aspen, juniper, Douglas fir), shrubs, and grass) and spatial density (open, dense, and semi-dense) based PFTs of Western US semi arid ecosystems, we applied a random forest based classification algorithm using waveform lidar derived biophysical, density, and topographic variables with field observations.
Field measured shrub heights highly correlated with waveform lidar derived pulse width deviation (R2 = 71, RMSE ~ 15 cm). Preliminary results of random forest PFT dissemination model showed a high performance rate (less than 10%error rate) discriminating shrubs, grass, aspen, and juniper from waveform lidar matrices. Backscatter cross section, Pulse width, amplitude, rise time, and number of echoes per waveform were the most significant variables differentiating PFTs. Backscatter cross section was also significant assessing spatial density (and percent cover) of vegetation. These results evaluate the PFT distribution across space and hence the ecosystem state and the habitat availability (Ex: for Pigmy rabbit and sage grouse). The results can also be used to constrain the model uncertainty that predicts the semi-arid ecosystem and similar environment responses to changing climate and land use, including future projection of terrestrial and atmospheric carbon fluxes and evaluation of hydrologic responses such as snow water equivalents (SWE), and snow melt timing. The results further provide basic information of thresholds and capacities of future sensors designed at coarser spatial resolutions (e.g. GEDI) over semi arid ecosystems.