Automated Coregistration of Repeat Digital Elevation Models for Surface Elevation Change Measurement Using Geometric Constraints

摘要

A large archive of digital elevation model (DEM) data spanning more than a decade is now available for a wide range of earth surface process studies. Precise coregistration is a critical step for obtaining accurate elevation change measurements from repeat DEMs. We present an algorithm for coregistering stereoscopic gridded DEMs that includes an automatic detection of control surfaces, a reliable estimation technique for defining a geometric relationship between DEMs through a 3-D conformal transformation, and a quality check (QC) for the confidence of the coregistration. For the control surface detection, we use a geometric constraint approach that uses similarities in slope, aspect, and height undulation as observation weights in an iterative transformation model solution. The QC is possible through the evaluation of the geometric similarities and the amount of detected relatively unchanged surfaces. We test our algorithm using a sample of repeat stereoscopic gridded DEMs from multiple sources over a rapidly changing glacier in Greenland. The algorithm results, on average, in a 24% decrease in the standard error, as obtained from elevation over stationary surfaces. The result is not sensitive to the distribution of the control surface. The QC metric provides a simple and effective criterion for parsing result quality in large data sets. Surface height decreases of ~ 110 and 200 m at the up-glacier and down-glacier ends of Kangerdlugssuaq Glacier in Greenland are observed between April 2001 and June 2010, respectively.