Creating the Matsu Digital Elevation Model
by Rick Guritz
ASF is expanding its role as a resource to national, state, local and tribal governments interested in mapping applications using remote-sensing data.
The Matanuska-Susitna (Matsu) Borough in southcentral Alaska recently contracted ASF to produce a digital elevation model (DEM) of its more heavily populated areas.
The Matsu, a region about the size of West Virginia, is named after two principle river drainages—the Matanuska and the Susitna rivers. Extreme and diverse terrains, from mountain ranges, valleys and glaciers to wetlands, farmlands and pristine wilderness, are all found within the Matsu boundaries.
Consequently, the area presented some challenges not previously encountered while working with DEM test areas near Fairbanks and Delta Junction, Alaska.
In particular, difficulties exceeded expectations with phase unwrapping in regions of extreme topography. We also obtained results with poorer coherence than desired along the coast due to increased precipitation and warmer temperatures. We observed a –1 m average offset, a standard deviation of 7.68 m, and a root mean squared error of 7.71 m. Although these results for low to moderate topography are encouraging, the DEM mosaic accuracy may not be adequate in some areas of extreme topography.
The project included collecting geodetic control points for mosaic processing, kinematic GPS road surveying of major roads within the Borough, and assessing the accuracy of the resulting DEM mosaic, which was produced from eleven ESA Tandem Mission pairs.
Each individual DEM was produced using ASF-developed SAR Interferometry user tools. In addition, we used a hybrid minimum cost flow phase unwrapping software developed at Stanford University, and a mosaic software developed at the Jet Propulsion Laboratory.
To minimize the impact of phase unwrapping errors on baseline refinement, we used new quality assurance methods. These methods detect, quantify, and in most cases, correct the phase unwrapping errors by using existing reference topography data. This technique is referred to as bootstrapping for phase unwrapping.
Using these methods, we generated good quality DEMs for regions of low to moderate relief. We also developed a mask to delineate regions of high versus low confidence. The resulting DEM mosaic, in regions of high confidence, was evaluated with the kinematic GPS road survey data.
The team gained valuable insight into what it takes to produce high quality data in regions of extreme terrain. Some of the lessons learned provide a good foundation for two recently funded projects: 1) a National Geospatial-Intelligence Agency (NGA) University Research Initiative grant to study ICESAT data as a possible source of geodetic control for topographic mapping, and 2) an NGA Feasibility Study to evaluate alternative mapping methods to supplement Shuttle Radar Topography Mission (SRTM) data north of 60 degrees latitude.