High-resolution DEM generation from spaceborne and terrestrial remote sensing data for improved volcano hazard assessment — A case study at Nevado del Ruiz, Colombia

Volcanoes with rugged terrain remain a challenging target for generating high-resolution digital elevation models (DEMs), especially in tropical areas with frequent cloud cover. Using Nevado del Ruiz volcano as an example, we combined DEMs from the TanDEM-X (TDX) satellite mission, terrestrial radar interferometry (TRI), and Structure from Motion (SfM), to generate a new DEM with 10-m spatial resolution. This is the first study combining satellite radar, ground-based radar, photography, and freely available global DEMs to generate a high-resolution DEM without data gaps. TDX data from ascending and descending orbits were combined to generate the base DEM. Instead of using a raster format to fuse DEMs generated from different data sets with different resolutions, we developed a methodology based on 3-D point clouds: 1) re-georeference the 5-m TRI and ~1-m SfM DEMs to the 10-m TDX DEM using the iterative closest point (ICP) algorithm to minimize the horizontal and vertical discrepancy between DEMs; then 2) merge the multiple point clouds to generate a final DEM without data gaps using an adaptive algorithm that uses two search distances to smooth the transition at the edges of different data sets. We assess the new 10-m DEM by comparing simulated inundation zones obtained with two volcano flow models, LaharZ (for lahars) and VolcFlow (for pyroclastic flows), and find significant differences with respect to the 30-m SRTM DEM. Our LaharZ simulation over the new DEM shows a longer lahar run-out distance. For pyroclastic flows, the VolcFlow simulation over the new DEM produces highly channelized flows over the steep portions of a river channel and gives a larger extent of thicker deposits compared to those obtained with the 30-m SRTM DEM. Quantitative and qualitative geomorphic analysis suggests that up-to-date DEMs with high spatial resolution (~ 10 m or even better) need to be generated to improve volcano hazard assessment for active volcanoes.