Aim Most airborne LiDAR sensors additionally record the backscattered signal amplitude (also referred to as “intensity”) for each reflection. The signal amplitude contains valuable information about the scattering characteristics of the scanned surface in the near- infrared wavelength (1064 nm) of the laser scanner (e.g. reflectance; cf. Fig.1). We aim at mapping glacier surface features by jointly using geometry and radiometry provided by airborne LiDAR. Glacier surface feature detection and classification from airborne LiDAR data B. Höfle 1 , R. Sailer 2 , M. Vetter 2 , M. Rutzinger 3 , N. Pfeifer 1 1 Institute of Photogrammetry and Remote Sensing (IPF), Vienna University of Technology, Austria 2 Institute of Geography, University of Innsbruck, Austria 3 International Institute for Geo-Information Science and Earth Observation (ITC), The Netherlands REFERENCES: [1] Höfle, B., Vetter, M., Pfeifer, N., Mandlburger, G., Stötter, J. (2009): Water surface mapping from airborne laser scanning using signal intensity and elevation data. Earth Surface Processes and Landforms. submitted. [2] Höfle, B., Pfeifer, N. (2007): Correction of laser scanning intensity data: data and model-driven approaches. ISPRS Journal of Photogrammetry and Remote Sensing, Vol. 62 (6), 415-433. [3] Höfle, B., Geist, T., Rutzinger, M., Pfeifer, N. (2007): Glacier surface segmentation using airborne laser scanning point cloud and intensity data. International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. 36 (Part 3/W52), pp. 195-200. We present a fully automatic procedure for glacier surface classification, working directly on the LiDAR point cloud (Fig.2). Firstly, signal amplitudes are corrected [2] (Fig.3) and laser shot dropouts are modeled (Fig.4), i.e. where no laser point could be measured [1]. A point cloud based segmentation delineates connected, homogeneous areas (i.e. low variation in amplitude within a segment and no abrupt change in elevation). Based on the segment feature statistics (e.g. mean amplitude) the surface objects are classified into snow, firn, ice or surface irregularities (e.g. crevasses). Methods Fig.2: Workflow of glacier surface mapping starting with the original ALS point cloud, resulting in a GIS vector of the classified regions (cf. [3]). Fig.1: Cross-section of corrected intensity along different surfaces. Moving average line clearly shows steps in intensity between the surface classes (cf. [2],[3]). As a first result corrected signal amplitudes could be derived (Fig.3), which are proportional to surface reflectance [2] in the laser‘s wavelength (1064 nm). Secondly, the modeling of dropouts (Fig.4) provides important information about areas with very low reflectance (mainly ice and water). The point cloud segmentation and object-based classification was performed for a representative test site in the upper part of the Hintereisferner (Fig.3). An overall classification accuracy of >90% could be reached (Tab.1), which exceeds the accuracy of a concurrent orthophoto segmentation and classification (86.7%). The final classified GIS vector polygon layer is shown in Fig.5. Results Tab.1: Point-based error matrix resulting from LiDAR point cloud segmentation and classification. Contact : bh@laserdata.info www.ipf.tuwien.ac.at Fig.4: Modeled laser shot dropouts at the terminus of Hintereisferner. Proglacial braided river is also clearly emphasized by dropouts (cf. [1]). laser shot dropout Fig. 3: Resulting corrected LiDAR intensity image of glacier Hintereisferner, Austria (cf. [2],[3]). Flight campaign was performed August 12, 2003, with an Optech ALTM 2050 sensor. Average point density is 1.7 pts/m 2 and average flying height above ground is 1150 m. Fig.5: (a) Delineated point cloud segments colored by mean intensity (black: low intensity, light gray: high intensity, white: unsegmented); (b) final glacier surface classification [3]. Hintereisferner Ötztal/Austria