Validation of the North American Ice Service iceberg drift model using a novel database of in-situ iceberg drift observations

<p>Icebergs calved from high-latitude glaciers and ice shelves pose a threat to vessels and offshore infrastructure at a time when Arctic shipping and offshore resource exploration are increasing. Knowledge of the location of potential ice hazards is therefore critical to ensure safe and efficient operations in this remote region. The Canadian Ice Service provides information to stakeholders on the observed and predicted distribution of icebergs in Canadian waters by combining iceberg observations with forecasts from the North American Ice Service (NAIS) iceberg drift model. The NAIS model estimates the forces acting on an iceberg to predict its future position and velocity and is widely used for the East Coast of Canada. However, the model is unproven for areas >60&#176;N and suffers from insufficient validation due to a lack of reliable <em>in-situ</em> observations of iceberg drift. In this study, we use a newly compiled iceberg tracking beacon database to assess the skill of the NAIS iceberg model's predictions of iceberg drift and investigate sensitivity to morphology and environmental forcing (e.g., ocean currents, winds).</p><p>Hindcast simulations of the observed tracks of 44 icebergs over the period 2008-2019 were run using ocean currents from three ocean models (CECOM, GLORYS and RIOPS) and wind and wave inputs from the ERA5 reanalysis. Comparisons of several distance error metrics between observed and modelled drift tracks indicate that the NAIS model produces realistic simulations of iceberg drift in Baffin Bay. The root mean square error after the initial 24-hour hindcast period ranged from 18-22 km and increased at a daily rate of 11-13 km, which is comparable to operational forecasts elsewhere. Improved model performance was observed for longer (>250 m) and deeper-keeled (>100 m) icebergs, which appears to counteract the model&#8217;s tendency to overestimate drift by reducing the influence of stronger surface ocean currents acting on the iceberg. Ocean current direction, wind direction, and iceberg keel geometry were identified by a sensitivity analysis as the model parameters and environmental driving forces that have the greatest influence on modelled iceberg drift. These results emphasize the need for accurate environmental information and underscore the importance of properly representing the physical characteristics of icebergs in drift models.</p>