A new algorithm for estimating ground elevation and vegetation characteristics in coastal salt marshes from high-resolution UAV-based LiDAR point clouds

Salt marshes are transitional zones between ocean and land, which act as natural buffers against coastal hazards. The survival of salt marshes is governed by the rate of organic and inorganic deposition, which strongly depends on vegetation characteristics, such as height and density. Vegetation also favours the dissipation of wind waves and storm surges. For these reasons, an accurate description of both ground elevation and vegetation characteristics in salt marshes is critical for their management and conservation. For this purpose, airborne LiDAR (light detection and ranging) laser scanning has become an accessible and cost-effective tool to map salt marshes quickly. However, the limited horizontal resolution (similar to 1 m) of airborne-derived point clouds prevents the direct extraction of ground elevation, vegetation height and vegetation density without the coupling with imagery datasets. Instead, due to the lower flight altitude, UAV (unmanned aerial vehicle)-borne laser scanners provide point clouds with much higher resolution (similar to 5 cm). Although methods for estimating ground level and vegetation characteristics from UAV LiDAR have been proposed for flat ground, we demonstrate that a sloping ground increases prediction errors. Here we derive a new formulation that improves the estimation by employing a correction based on a LiDAR-derived estimate of local ground slope. Our method directly converts the 3D distribution of UAV LiDAR-derived points into vegetation density and height, as well as ground elevation, without the support of additional datasets. The proposed formulation is calibrated by using measured density and height ofSpartina alterniflorain a marsh in Sapelo Island, Georgia, USA, and successfully tested on an independent dataset. Our method produces high-resolution (40 x 40 cm(2)) maps of ground elevation and vegetation characteristics, thus capturing the large gradients in the proximity of tidal creeks. (c) 2020 John Wiley & Sons, Ltd.