Abstract
This supplementary material provides additional detail on the derived orthophotomap models used to study Greenland's glacial evolution. Specifically, it offers a detailed summary of the individual structure-from-motion (SfM) models, including their technical specifications and error budgets.
Introduction
This document serves as the technical backbone for a study investigating how Greenland's glaciers have changed over the long term. While the main paper focuses on glaciological findings, this supplement details the rigorous photogrammetric process used to transform decades-old archival "snapshots" into scientifically valid 3D models. Researchers utilized Structure-from-Motion (SfM) photogrammetry to reconstruct historical glacier surfaces from archival aerial and satellite imagery dating back to the mid-20th century. This material provides the technical specifications for 14 distinct orthophotomap models, demonstrating that historical data can be anchored to modern coordinates with sufficient precision to measure decadal ice loss.
The Challenge of Historical Georeferencing
To understand glacier response to climate change, researchers require a "long view" extending beyond the modern satellite era. Greenland has been photographed from the air since the 1930s, but these archival images—often oblique, uncalibrated, and lacking GPS metadata—present significant measurement challenges.
The primary hurdle is georeferencing: aligning grainy photos from 1930s biplanes or 1960s spy satellites with modern maps. This study addresses the technical challenge of turning "messy" historical visual data into rigorous, measurable 3D digital elevation models (DEMs) and orthophotomaps.
Methodology: Structure-from-Motion (SfM)
The authors employed SfM to identify common features across multiple overlapping photos, calculating the 3D geometry of the scene and the camera positions. Key aspects of the approach include:
- Ground Control Points (GCPs): Stable features like rocky outcrops were identified in both archival photos and modern reference maps to anchor the models.
- Adaptive Density: GCP counts were varied based on terrain; additional points were manually added to models showing high error to force better alignment.
- Error Quantification: 2D and 3D residual errors, along with directional errors (X, Y, and Z), were tracked to ensure accuracy for glaciological analysis.
Datasets and Model Coverage
The study utilizes three primary sources of historical imagery:
- BAARE Mission: Archival aerial oblique photography.
- CORONA Satellites: Stereo pairs from first-generation US declassified spy satellites (1960s-70s).
- 1981 Aerial Surveys: Vertical aerial photography from missions flown in July and August 1981.
- Reference DEMs: Modern elevation data providing "ground truth" for GCP placement.
Key Insights and Results
The researchers successfully combined disparate imagery types into a single analytical framework. Despite the age of the material, 3D errors generally ranged between 13 and 32 meters. While larger than modern GPS errors, this precision is sufficient to detect the multi-hundred-meter changes occurring in Greenland’s glaciers over 40–80 years.
The supplement details 14 specific models:
- Coverage: Models range from small 77 km² glacier tongues to massive 2,898 km² regional surveys.
- Accuracy: Average 3D errors for BAARE (oblique) models were 13–26 meters. The 1981 vertical aerial models showed errors up to 45m but covered much larger areas.
- Consistency: By documenting "signed errors," the authors provide a roadmap for accounting for systematic biases in historical reconstructions.
Limitations and Future Outlook
GCP density was limited by the availability of identifiable features; areas of heavy snow or featureless ice remain difficult to anchor. This technical foundation allows scientists to move from observing that Greenland is changing to understanding how and why on a glacier-by-glacier basis, investigating the locally-heterogeneous responses of neighboring glaciers.
Summary updated