Hydrocarbon exploration in the Arctic environment will very much depend onour ability to continuously track ice floes and forecast ice events that maygenerate dangerous loads on exploration and production infrastructure. Wepresent a first-of-its-kind computational framework which is centered aroundnear-real-time satellite imagery and incorporates real-time metocean data, providing automated analysis of such hazards in regions where moving ice ispresent. Our automated framework carries out several ongoing operations: icedetection and classification from satellite images, floe tracking from oneimage to the next, forecasting of floe trajectories beyond the observed tracks, and estimation of an uncertainty cone around the trajectory forecast. Weutilized the IBM InfoSphere™ Streams real-time analytics platform to deploy oursoftware, which made it possible for us to concentrate exclusively onprototyping algorithms, taking for granted the streaming infrastructure neededfor real-time data ingestion and flow between operators. Given our experiencedeveloping this prototype we conclude that a production-worthy, automatedtracking and forecasting capability is computationally feasible and within ourreach.


Exploration in the Arctic will lean heavily on our ability to continuouslytrack ice floes and to forecast ice events that may generate dangerous loads ondrilling and production infrastructure. The engineering challengesnotwithstanding of deploying data acquisition systems in the Arctic, in thisproject we investigated analytical problems that surround continuous trackingand forecasting of ice floe trajectories. Here we introduce a first-of-its-kindcomputational prototype for real-time analysis of ice floes and icebergs. Ourprototype centers around our proven ability to acquire near-real-time satellite(Synthetic Aperture Radar) imagery and to incorporate real-time metocean datastreams, a powerful combination of data for automated analysis and modeling ofhazards in regions that feature moving ice. Key metocean data streams forforecasting ice movement include wind velocity measurements and forecasts, aswell as water current models and velocity measurements. We collected both SARand metocean data over a ConocoPhillips block in the Beaufort Sea (Figure 1)for this investigation. Our near-real-time SAR imagery stream over theblock (Figure 2) provided the primary data for tracking.

This content is only available via PDF.
You can access this article if you purchase or spend a download.