An Intercomparison of Semi-Eulerian and Lagrangian Based Cyclone Tracking Methods for the North Pacific and Alaskan Regions

Wednesday, 17 December 2014
Norman J Shippee and David E Atkinson, University of Victoria, Victoria, BC, Canada
The idea of considering the “end user perspective” regarding storm activity and objective tracking methods used to compile information on their behaviour is particularly important in the Alaskan region. Annually, coastal regions in the North are exposed to stormy conditions, though most impacts occur during periods where multiple storms track over the same area in a short period of time (serial cyclones) or where strong storms occur without the presence of a protective sea ice buffer. From a fixed perspective (i.e. Eulerian), a storm may be identified more by the impacts that it generates at that location (winds, sea state, erosion). From a Lagrangian (tracking) view, the intensity, duration, and characteristics of the synoptic environment may prove more relevant for understanding. The overall “effectiveness” of an objective tracking method depends on the intended use of the provided information. While pitting different methods against each other is not necessarily a fruitful exercise (Mesquita et al. 2009), the reality is that one method may better reflect the reality of storm activity and impacts to those experiencing the weather first hand.

One of the more subtle points in extra-tropical cyclone tracking and comparison work is the method by which a storm is defined. Most cyclones are analyzed on MSLP fields; others define a cyclone by relative vorticity (ζ) maxima at 850 hPa (NH) and minima (SH). Storms can also be defined by wind events, or even impacts, at a location. Using counts of strong wind events at a grid point or location can account for pressure gradients both associated with storms and absent of a synoptic event.

Three separate tracking algorithms are analyzed to determine the method most likely to produce a long-term homogeneous dataset that can be used to train a statistical seasonal prediction method. These methods include the Serreze algorithm, Hodges TRACK algorithm, and Atkinson algorithm. Both the Serreze and Hodges methods provide a tracking perspective while the Atkinson algorithm provides a Eulerian view using wind speed returns at grid points throughout the study area. Unique to this study is a view of how all three methods perform during strong storms (such as the 2004 Nome, AK storm) and weaker events where wind events are still triggered.