Trajectory Cluster Analysis

An air parcel trajectory shows the path of an air parcel as it is transported through the atmosphere by the wind. HYSPLIT is one such model that computes trajectories. One application of a forward in time trajectory, i.e. computed using a meteorology forecast, is to estimate where a hazardous substance in the air, such as radioactivity, will be at a later time. One example of a backward in time trajectory is to estimate the source region(s) of atmospheric pollutants occuring at a given location (see NARE, the North Atlantic Regional Experiment). An individual trajectory gives only a general description of the wind field because it does not account for atmospheric processes such as vertical and horizontal mixing and diffusion. Dispersion models, which include atmospheric mixing processes and may have the same advection algorithm as trajectory models, are usually run to provide a more complete representation of atmospheric transport.

Given a large set of trajectories, say one back-trajectory beginning each day precipitation occurred at an AIRMoN-wet site over the period of a year, one method to analyze the atmospheric transport associated with the AIRMoN samples is to cluster the trajectories. Trajectory clustering is a process of grouping similar trajectories together whereby differences among individual trajectories in a cluster are minimized and differences among clusters are maximized. Ideally, each cluster represents different classes of synoptic regimes over the duration of the trajectories.

Trajectory clustering at ARL (Silver Spring) has been done for the following projects.

• To assess the quality of forecast trajectories (the attached example shows the trajectories in 5 distinct clusters),

• To form the basis of the criteria used to select case studies to evaluate fine scale Eulerian mesoscale models (e.g. Chesapeake Bay study). Models such as the Regional Atmospheric Modeling System, (RAMS), and the Regional Acid Deposition Model, (RADM), can be used to predict or analyze deposition velocities for anthropogenic substances known to contribute to eutrophication, the most serious threat to ecosystem habitat in the Chesapeake Bay and other regions.

• To investigate effects of the 1990 Clean Air Act Amendments-mandated sulfate emissions reductions (see below).

Research Summary

AIRMoN data from 1995 show generally lower sulfate concentrations in precipitation than in 1993, presumably because of the 1990 Clean Air Act Amendment-mandated emissions reductions. However, meteorological and chemical factors, among others, also contribute to the sulfate concentrations. An investigation of the meteorological factors affecting sulfate concentration in precipitation is currently underway. In a preliminary study, HYSPLIT back-trajectories from one AIRMoN site (State College, PA) for two years, 1993 and 1995, were computed beginning at the midpoint of the 24-h AIRMoN precipitation sample and at an elevation of 2000 m, assumed to be representative of atmospheric transport leading up to a precipitation event. Trajectories associated with small precipitation amounts (less than 5 mm) were removed from further analysis because very high sulfate concentrations typically occurred with the very low precipitation amounts, an artifact of dilution. Results from a cluster analysis found generally lower sulfate concentrations, by cluster, in 1995 as compared to 1993 (see figures).
Precipitation volume-weighted mean SO4 concentration by cluster for 1993 (left) and 1995 (right).

Presentation

Stunder, B.J.B. and R.S. Artz, A comparison of 1993 and 1995 AIRMoN precipitation chemistry measurements using HYSPLIT trajectories, 1996 NADP Technical Committee Meeting, October 21-24, 1996, Williamsburg, VA.

Stunder, B.J.B., 1996: An assessment of the quality of forecast trajectories, J. Appl. Meteor., 35, 1319-1331.