Operational iceberg drift forecasting in Northwest Greenland

Abstract

An operational iceberg drift forecast model was developed and used in 2012 to support Shell Greenland A/S, which was leading a consortium of companies to conduct a scientific coring campaign offshore northwest Greenland. The nine-week program took place in August–October 2012 in Melville Bay where there is a very high spatial density of icebergs. In order to increase operational safety and efficiency, the iceberg drift model presented here was fully integrated within the ice management strategy and its decision protocols. The model was optimized to fulfill the operational requirements in terms of (1) the format and content of the input/output, (2) the calculation of the iceberg alert zones, and (3) the derivation of the associated Closest Point of Approach (CPA) and time to CPA (T-CPA). To achieve this, the iceberg drift forecast model used as near real-time input: in-situ measured metocean parameters, observed iceberg drift and size data, tidal currents, and weather forecasts. The model forecast the 24-hour iceberg drift trajectory including the CPA and T-CPA. It also derived the T-Time (e.g., Total Time, or the time necessary to lift the coring pipe to 50m below the seabed) alert zones based on the observed mean iceberg drift speed. The iceberg trajectory forecast was optimized by running a hindcast simulation of the observed iceberg trajectory. In this simulation, the air and water iceberg form drag coefficients were tuned to minimize the difference between the hindcast and the observed iceberg drift trajectories. The forecast iceberg track, beginning at the end of the observed iceberg record, was subsequently generated using the air and water form drag coefficients derived from the hindcast simulation. The model also incorporated a method to correct for the lack of robust near real-time in-situ ocean current measurements. During this campaign, the model was used to forecast the drift of 73 icebergs and exhibited good performances in locations with strong and persistent non-tidal currents. However, further offshore where the ocean currents were dominated by tidal and inertial forcings, the model exhibited poorer performance. Nevertheless, as shown for two icebergs in this paper whose drift was dominated by tidal and inertial oscillations, model performance could be significantly improved in the hindcast mode by using the vessel's wind data and the ocean current data retrieved from a mooring deployed in the area. This illustrated the importance of using adequate near real-time ocean current and wind measured and forecast input data.