Unsupervised detection and quantification of iceberg populations within sea ice from dual-polarisation SAR imagery
Accurate estimates of iceberg populations, disintegration rates and iceberg movements areessential to fully understand ice sheet contributions to sea level rise and freshwater and heatbalances. Understanding and prediction of iceberg distributions is also of paramount importancefor the safety of com...
| Main Authors: | , , , , , |
|---|---|
| Format: | Conference Object |
| Language: | English |
| Published: |
.
2022
|
| Subjects: | |
| Online Access: | https://meetingorganizer.copernicus.org/EGU22/EGU22-8267.html https://doi.org/10.5194/egusphere-egu22-8267 http://ecite.utas.edu.au/149344 |
| Summary: | Accurate estimates of iceberg populations, disintegration rates and iceberg movements areessential to fully understand ice sheet contributions to sea level rise and freshwater and heatbalances. Understanding and prediction of iceberg distributions is also of paramount importancefor the safety of commercial and research shipping operations in polar seas. Despite theirmanifold implications the operational monitoring of icebergs remains challenging, largely due todifficulties in automating their detection at scale. Synthetic Aperture Radar (SAR) data from satellites, by virtue of its ability to penetrate cloud coverand strong sensitivity to the dielectric properties of the reflecting surface, has long beenrecognised as providing great potential for the identification of icebergs. Many existing studieshave developed algorithms to exploit this data source but the majority are designed for openwater situations, require significant operator input, and are susceptible to the substantial spatialand temporal variability in backscatter characteristics within and between SAR scenes that resultfrom meteorological, geometric and instrumental differences. Further ambiguity arises whendetecting icebergs in dense fields close to the calving front and in the presence of sea ice. Fordetection to be fully automated, therefore, adaptive iceberg detection algorithms are required, ofwhich few currently exist. Here we propose an unsupervised classification procedure based on a recursive implementationof a Dirichlet Process Mixture Model that is robust to inter-scene variability and is capable ofidentifying icebergs even within complex environments containing mixtures of open water, sea iceand icebergs of various sizess. The method exploits freely available dual-polarisation Sentinel 1 EWimagery, allowing for wide spatial coverage at a high temporal density and providing scope fornear-real-time monitoring. It overcomes many of the limitations of existing approaches in termsof environments to which it may be applied as well as requirements for labelled training datasetsor determination of scene-specific thresholds. Thus it provides an excellent basis for operationalmonitoring and tracking of iceberg populations at a continental scale to inform both scientific andnavigational priorities. |
|---|