ANALYSIS OF LIDAR DATA DURING RAPID SCIENTIFIC RESPONSE TO THE JANUARY 12, 2010 HAITI EARTHQUAKE

Within 10 days of the Haiti earthquake, ~2.7 billion point measurements of surface topography were gathered using airborne LiDAR. These data are the first large-footprint (850 km²) LiDAR terrain map obtained during rapid response to an earthquake, motivating us to develop an efficient analysis workflow that relies upon the human capacity to visually identify meaningful patterns embedded in noisy data. We used open-source software Crusta and LiDAR Viewer to conduct interactive visual analysis of the bare-earth digital elevation model (DEM) and point cloud data, respectively, in a 4-sided, 800 ft³ CAVE immersive visualization environment. Our workflow involved 6 steps: We visualized the bare-earth DEM along the ~50 km-long reach of the active, left-slip Enriquillo fault in the vicinity of the epicenter using Crusta and GIS. Although we found no evidence of 2010 surface rupture, this visualization did reveal clear evidence of past surface rupture, including apparent landform offsets as small as ~10 m. Using LiDAR Viewer to visualize the full point cloud revealed numerous topographic details absent from the bare-earth DEM, including an additional landform offset. Quantitative comparison of the full and classified point clouds using a pointset comparison tool in LiDAR Viewer indicated that only ~10% of the LiDAR returns were classified as bare ground. Visualizing this classification using LiDAR Viewer and the CAVE revealed numerous points within the ground plane misclassified as vegetation, explaining the DEM degradation observed with Crusta. To obtain the most accurate bare-earth DEMs for the landform offsets, we used LiDAR Viewer in the CAVE to manually subset and classify portions of the cloud. Using a virtual paintbrush selection tool to identify, select, and remove vegetation in these areas led to ~70% of the points being classified as bare earth. We then gridded these results and used Crusta to remotely measure landform offsets and associated uncertainties and conduct surface geologic mapping to document field relationships. We anticipate that workflows supporting remote, interactive visual analysis of LiDAR data, such as that developed here, will transform the use of high-resolution topographic data in both relief efforts and the rapid scientific response following future natural disasters.