Evaluating Areal Errors in Northern Cascade Glacier Inventories

Abstract

This thesis evaluates spatial errors, inherent to image-based analysis, that affect our ability to detect areal changes in glaciers. Analysis of areal differences of 742 North Cascade glaciers, observed in A.D. 1958 and A.D. 2006 inventories, suggests change detection is limited by linear errors of 5 m, 7 m, and 184 m caused by the imagery, digitization operators, and snowpack variability, respectively. Using the cumulative error of 196 m as a threshold for detecting areal change over the 48-year period, only 250 glaciers were detectable outside of the error threshold. Of those detectable changes, 240 glaciers were decreasing in area and 10 were increasing in area. Coupling the A.D. 2006 inventory with a contemporary 15 m resolution digital elevation model allows for correlation of areal changes with common geometric, geographic, and hypsometric derivatives. Generally, correlations are poor between the measured variables for shrinking glaciers, but the large number of glaciers suggests that warming or drying trends dominate the study period. For the 10 growing glaciers, the skewness has a strong positive correlation coefficient with kurtosis and AABR, as well as strong negative correlations with ELA and HI. These relationships suggest common geometries that are advantageous for high-altitude accumulation areas. The intrinsic errors in the analysis of these images are comparable to the observed changes over the past half century for a large fraction of the glaciers.