Reconstructing topographic change from historical aerial images with photogrammetry and geostatistics

Abstract

Precisely measuring the Earth’s changing surface on decadal to centennial time scales is critical for many science and engineering applications, yet long-term records of quantitative landscape change are often temporally and geographically sparse. Archives of scanned historical aerial photographs provide an opportunity to augment these records and capture the historical state of the Earth’s surface at high resolution. However, historical images remain underutilized for quantitative analysis in Earth system science due to limitations of the input datasets and methodologies. In the first chapter of this dissertation, I developed an automated method to photogrammetrically reconstruct high-resolution digital elevation models (DEMs) from historical aerial photographs, with a particular focus on glacierized mountain ranges in the Western United States. Processing historical photographs at scale is challenging because images can be distorted due to film degradation and scanning artifacts, camera positions are often inaccurate or unknown, and the photographs were often collected with limited stereo overlap or inadequate stereo geometry. The method developed in this chapter addresses these challenges and is accompanied by publicly-released software that facilitates automated image standardization, block detection, multi-temporal camera model optimization, and robust DEM coregistration to high (e.g. 1 m) and low (e.g. 30 m) resolution reference terrain. In the second chapter, I designed a temporal Gaussian Process regression framework to interpolate spatially and temporally continuous surface elevation estimates from the DEMs produced in the first chapter. DEMs generated from stereo photogrammetry often lack data in areas where there are insufficient match points between images to confidently triangulate surface elevations. Data voids are problematic for glaciological analysis because a continuous estimate of the surface elevation change across the entire glacier is required to distinguish a change in mass from a kinematic redistribution of ice. The framework leverages the per-pixel temporal surface elevation covariance to predict an elevation estimate at any time step. Furthermore, the software that executes this framework is memory-efficient, user-friendly, and open-source, which enables processing at full resolution on any machine and thereby democratizes the methodology for the community. In my final chapter, I produced 70 years (∼1950 – 2020) of continuous topographic measurements and identified multiple periods of glacier advance, retreat, and stability. Mountain glaciers are perennial sources of freshwater that shape the natural landscape and provide important ecosystem services, such as aquatic stream habitat for keystone fish species and potential energy for hydropower generation. Over the past century, many glaciers in Western North America have lost mass and steadily retreated in response to long-term climate forcing. Some glaciers, especially those on stratovolcanoes with high-altitude accumulation areas and steep elevation gradients, present decadal climate-driven advances and kinematic movement on shorter (interannual) timescales. These changes can lead to significant sediment transport and devastating outburst floods that damage downstream infrastructure. To evaluate the method developed in the second chapter of this dissertation, I focused my analysis on three well-studied glaciers in the Pacific Northwest of the United States – South Cascade Glacier in the Cascade Range, Nisqually Glacier at Mount Rainier, and Easton Glacier at Mount Baker. The DEMs generated by this analysis characterize multiple periods of glacier advance, retreat, and stability at these sites, which can provide useful constraints for geophysical glacier model calibration. These long-term time series of high spatial and temporal resolution estimates produced by the methods developed in this dissertation can help constrain projections of future glacier mass change under different climate forcing and mitigate impacts on downstream water resources, infrastructure, and geohazard risk.