Robust Filtering of Airborne Laser Scanner Data for Vegetation Analysis

Airborne laser scanning (ALS), often referred to as lidar or laser altimetry, is a remote sensing technique which was originally designed to measure the topography of the Earth’s surface. While the first commercially available airborne laser scanners recorded only the time of one backscattered pulse, state-of-the-art systems record several echoes for each emitted laser pulse. Thereby a 3D data cloud is obtained which conveys valuable information about the vegetation canopy. For the retrieval of vegetation parameters the most common procedure is to 1) calculate a digital terrain model (DTM) by filtering last-pulse ALS data, 2) form a digital surface model (DSM) from first-pulse ALS data to represent the topmost surface (top of vegetation, building roofs, etc.), and 3) calculate a normalised digital surface model (nDSM) to represent the height of the vegetation. The so derived vegetation height model can be used as input for further vegetation analysis. In this study we investigate the quality of DTMs and nDSMs derived from first/last-pulse ALS data in an alpine environment. We use a hierarchic robust filtering technique for DTM generating from ALS data obtained during leaves-off (December) and leaves-on (July) conditions. The derived DTMs compare well for flat, nonvegetated terrain. Over forested terrain it is found that the penetration rates are much higher in winter compared to summer, even for forest patches dominated by spruce. As a result of the low forest penetration in summer, high differences between the summer and winter DTMs were observed over forested, steep terrain. Another consequence of the different penetration rates is that only the summer DSM correctly represents the top of canopy. This shows that, ideally, a vegetation height model is obtained by subtracting a winter DTM from a summer DSM.