Retrieval of small-relief marsh morphology from Terrestrial Laser Scanner, optimal spatial filtering, and laser return intensity

Marshes are ubiquitous landforms in estuaries and lagoons, where important hydrological, morphological and ecological processes take place. These areas attenuate sea action on the coast and act as sediment trapping zones. Due to their ecosystem functions and effects on coastal stabilization, marshes are crucial structures in tidal environments, both biologically and geomorphologically, and are fundamental elements in wetland restoration and coastal realignment schemes. The spatially-distributed study of the geomorphology of intertidal areas using remotely-sensed digital terrain models remains problematic, owing to their small relief, often of the order of a few tens of centimetres, and to the presence of short and dense vegetation, which strongly reduces the number of resolvable ground returns. Here, we use high-resolution Terrestrial Laser Scanning (∼ 200 returns/m2) to retrieve a high-resolution and high-accuracy Digital Terrain Model within a tidal marsh in the Venice lagoon. To this aim we apply a new filtering scheme to Terrestrial Laser Scanner data which selects the lowest values within moving windows, whose optimal size is determined with the aid of a limited number of ancillary Differential GPS data in order to maximize resolution while ensuring the identification of true ground returns. The accuracy of the filtered data is further refined using classifications of the intensity of the returns to extract additional information on the surface (ground or canopy) originating the returning laser beam. Validations against about 200 reference Differential GPS ground elevation observations indicates that the best separation of canopy and ground signals is obtained using a low-pass filter with window size of the order of 1 m and the maximum likelihood classifier to further refine the detection of ground returns. In this case the average estimation error is about 1 cm (slight overestimation of ground elevation), while its standard deviation is about 3 cm. Our approach allows the separation of laser returns coming from the low marsh vegetation from those coming from the marsh surface. The overall result is a new observation technique producing Digital Terrain and Digital Surface Models in areas with very small relief, which is shown to provide unprecedented high-resolution and high-accuracy characterizations of marsh morphology.