Arctic High-Resolution Elevation Models: Accuracy in Sloped and Vegetated Terrain

Abstract

New high-resolution elevation models for Alaska have recently been released; they were created using interferometric synthetic aperture radar (IFSAR) and automated matching of high-resolution optical satellite stereo imagery (OSSI). These products promise to fill a void in available digital elevation models (DEMs) for the Arctic. However, the effective use of these models requires knowledge of their expected accuracy, and to date, a detailed analysis of these models in remote Arctic locations has not been undertaken. Expected accuracy is necessary to gauge the uncertainty of any scientific conclusions based upon analysis of these DEM sources. To that end, both aforementioned DEM techniques were compared to airborne LiDAR (light detection and ranging) in the area surrounding Sitka, Alaska. It was found that both the IFSAR and OSSI DEMs provide vertical accuracy at the 2–4-m level (1 σ) in flat and open terrain but perform significantly worse in areas of vegetation cover with standard deviations increasing to ∼7–12 m. The DEM errors were found to have a strong positive correlation with vegetation height, and the overall error pattern suggests that neither OSSI nor IFSAR accurately model either the ground or top of the tree canopy, instead representing a surface between the canopy and topographic elevation.