Comparison of statistical iceberg forecast models

Abstract

Short-term iceberg drift prediction is challenging. Large uncertainties in the driving forces -current, wind and waves -usually prevent accurate forecasts. Recently several statistical iceberg forecast models have been proposed by the authors, which use iceberg position measurements to improve the short-term drift forecast. In this article these statistical forecast methods and models are briefly reviewed. An extensive comparison between the statistical models, in addition to a dynamic iceberg forecast model, is performed on several iceberg drift trajectories. Based on this comparison a new statistical forecast scheme is proposed that combines some of the advantages of the other methods.

Assessing Ice Island Drift Patterns, Ice Island Grounding Locations, and Gridded Bathymetry Products between Nares Strait and the North Atlantic

“…The drift and fracture of the three largest and most recent calving events from the PG caused a pulse of ice island fragments through eastern Canadian waters over the following years, while ongoing melt resulted in an input of freshwater over the ice islands' drift trajectories (Crawford et al, 2018b). The ice islands originating from the 2008, 2010, and 2012 PG calving events followed the same general drift trajectory as those previously reported from northern Greenland ice shelves and floating ice tongues (Higgins, 1989;Newell, 1993;Peterson et al, 2009) and icebergs originating from other glaciers in the north and west sectors of Greenland (Hansen and Hartmann, 1998;Andersson et al, 2018;Marson et al, 2018), though many ice islands were observed to drift into adjacent passages as well. Common deviations from the nominal trajectory were observed into the steeply sided Jones and Lancaster Sounds.…”

An iceberg forecast approach based on a statistical ocean current model

“…It has been observed that for the short-term forecasts (1 h to 24 h) of iceberg trajectories that methods incorporating past observations exhibit superior performance relative to simulations of mechanistic dynamic models (Andersson et al, 2018;Marko et al, 1988). This is not surprising since they work with more information.…”

“…Moreover, a relative performance index is introduced to compare the different forecast models on different icebergs (Andersson et al, 2018). It may be that the icebergs drift considerably differently (for instance, with different drift velocities).…”

“…In this situation, other approaches that include past information to forecast an iceberg trajectory are feasible (Andersson et al, 2018). In this article a simple transformation is introduces that allows to forecast the iceberg trajectory with a statistical ocean current model.…”

“…Figure1: Visualisation of the relative performance index(Andersson et al, 2018). If the end position of the iceberg forecast is encapsulated in the inner circle is defined as excellent, followed by two rings where the forecast is defined good and acceptable.…”