STATISTICS OF TRAJECTORIES IN THE NORTH ATLANTIC STORM TRACK

Aspects of the Lagrangian behaviour of the North Atlantic storm track are studied by calculating 3-D back trajectories using winds from ECMWF analyses and climate runs of the UK Met Office Unified Model (UM). The origins of trajectories over whole seasons are examined using probability density estimates which are quite insensitive to errors in the trajectory calculations. These estimates can therefore differentiate the impacts of storm track regimes on atmospheric tracers. The representation of Lagrangian behaviour in the UM has a strong influence on the distribution of water vapour which in turn impacts on the response of the model climate to imposed changes, such as double carbon dioxide.

This trajectory data set is also being employed to examine statistics of dispersion including cluster displacement, spread and shape, and the average strain rate experienced along trajectories. The aim is to provide quantitative climatological information on the origins of air and the typical rates at which air masses are stirred together. Preliminary results suggest that interannual variability is particularly strong for trajectories arriving over the European side of the Atlantic.

Back trajectories have been routinely calculated at the ECMWF for the whole of 1995, 1996 and 1997. The trajectories are 5 days long and arrive at 900hPa in three clusters centred over western Europe, the mid-Atlantic storm track region and the eastern USA. Throughout 1995 the trajectories were released daily, at 12UT, and for 1996 the release frequency was increased to once every 6 hours. Since the end of September 1996 the trajectories have been calculated using a new package, called ``Offline'' (see Methven (1997), UGAMP technical report 44, for more details). Now meteorological fields, including potential temperature, specific humidity and potential vorticity, are interpolated to the particle positions in the same manner as the winds. All these trajectories have been archived at the British Atmospheric Data Centre (BADC) and are available to any U.K. researchers.

Trajectories arriving at one instant over a particular region may have disparate origins with very different chemical signatures. Moreover, trajectories in the North Atlantic storm track region vary greatly from day to day. The aim here is to quantify the probability that air arriving over western Europe has originated from a given area when averaging over a season. The simplest method would be to count the number of trajectories originating from each lat-long grid box. However, here we use kernel estimation methods which have been developed by Kevin Hodges at the Environmental Systems Science Centre (ESSC). These have two major advantages over grid box counting methods: there is no area bias due to box definition and the estimation is much more robust and smooth than ``binning''. The number density of particles is estimated by summing kernel functions which are placed at the position of each trajectory at a given time before arrival. Each kernel has a finite spread which smooths the density estimate and this spread is wider where density is low. The spread is determined by a maximum likelihood estimation which aims to keep both the bias and variance of points about the estimate to a minimum. The procedure is similar to finding a weighted average of a set of points, where a larger kernel spread is used where uncertainty is greater. The density function is normalised so that its integral over all solid angles equals one.

Figure 1 (Postscript version) shows the probability densities for air origin at 2 days and 5 days before arriving over Western Europe. These distributions have been calculated using all trajectories which arrive in the European cluster, daily, for winter (DJF) 1995/1996, amounting to some 15288 trajectories. One can immediately see that this winter was one where air typically originated from central Europe, although at 5 days back the spread of origin is wide and extends right across the Atlantic.

In contrast, the previous winter (1994/1995) was typified by a predominance of westerley flow. The maximum probability density is situated on the far side of the Atlantic at only 2 days back; at 5 days back the density is spread widely but is a maximum over NE Canada (see Figure 2 (Postscript version) ).

The NERC Thematic Programme Atmospheric Chemistry Studies in the Oceanic Environment (ACSOE) held two major field campaigns on the west coast of Ireland (June-July-August 1996, April-May 1997). The campaigns involved the ground based Mace Head Atmospheric Research Station, the NERC Reseach Vessel Challenger and the Cranfield Jetstream Aircraft. Trajectories derived from ECMWF forecast data were used to decide where to locate the ship in the Atlantic, the flight track of the aircraft and the mode in which instruments were run. Trajectory data derived from ECMWF analyses were input into models to investigate the chemical development along their paths. This type of modelling was carried out both during and after the campaigns using the Offline trajectory code. The trajectories are still being used in the scientific interpretation of the collected chemical data, providing information on the origin of the air and the processes, such as deposition and emissions, which are likely to have affected it along its path to Mace Head.

Return to DYNAMICAL PROCESSES home page.

Return to John Methven's research page.