438
1. The back-trajectories model
The computer model that produced the atmospheric trajectories for this study was
developed by Harris (1982). The primary input for the trajectory program consists of
griddled wind components at mandatory pressure levels, produced at the National
Meteorological Center (NMC), Suitland, MD, U.S.A.,by a global atmospheric circulation
model. The resolution of the NMC data grids used in this study is 2.5° of latitude and
longitude. Meteorological observations from around the globe constrain the atmospheric
model that produced the griddled data for 0000 Z and 1200 Z each day. Coverage of
meteorological observations is naturally sparse over the poles and oceans compared with
well-populated land areas, but aircraft reports and satellite data supplement for areas where
the regular rawinsonde observations are not possible.
A 5-day trajectory consists of 40 individually computed 3-h trajectory segments
placed end to end. To compute each segment, the trajectory program interpolates in time
between two NMC grids (0000 Z and 1200 Z) to the midpoint of the 3-h period in question.
The winds at the previous trajectory segment endpoint are calculated from the winds at the
four grid points surrounding the endpoint. A first-guess trajectory segment is computed,
assuming that this wind is constant for the 3-h duration. At the midpoint of the first-guess
segment the winds are again calculated by bilinear interpolation. These latter winds are then
used to compute the final trajectory segment because they are assumed to be more
representative of winds for the 3-h segment. This iterative approximation technique is
called the modified Euler method (Stark, 1970).
The trajectories produced by this model are best used as an indication of the general
airflow pattern. Errors that may be introduced by interpolation, poor data coverage or the
model's approximation of the real atmosphere could be compounded as trajectory segments
are added further back from the destination. Thus, the apparent origin of a 5-day backtrajectory should be considered an approximation, and, as indicator of source regions,
should be supported by other atmospheric data or, as is done here, used in a climatological
sense.
Despite their limitations, atmospheric trajectories have been frequently employed in
studies of long-range transport of gases and aerosols. Using back-trajectories, Herbert et al.
1. The back-trajectories model
The computer model that produced the atmospheric trajectories for this study was
developed by Harris (1982). The primary input for the trajectory program consists of
griddled wind components at mandatory pressure levels, produced at the National
Meteorological Center (NMC), Suitland, MD, U.S.A.,by a global atmospheric circulation
model. The resolution of the NMC data grids used in this study is 2.5° of latitude and
longitude. Meteorological observations from around the globe constrain the atmospheric
model that produced the griddled data for 0000 Z and 1200 Z each day. Coverage of
meteorological observations is naturally sparse over the poles and oceans compared with
well-populated land areas, but aircraft reports and satellite data supplement for areas where
the regular rawinsonde observations are not possible.
A 5-day trajectory consists of 40 individually computed 3-h trajectory segments
placed end to end. To compute each segment, the trajectory program interpolates in time
between two NMC grids (0000 Z and 1200 Z) to the midpoint of the 3-h period in question.
The winds at the previous trajectory segment endpoint are calculated from the winds at the
four grid points surrounding the endpoint. A first-guess trajectory segment is computed,
assuming that this wind is constant for the 3-h duration. At the midpoint of the first-guess
segment the winds are again calculated by bilinear interpolation. These latter winds are then
used to compute the final trajectory segment because they are assumed to be more
representative of winds for the 3-h segment. This iterative approximation technique is
called the modified Euler method (Stark, 1970).
The trajectories produced by this model are best used as an indication of the general
airflow pattern. Errors that may be introduced by interpolation, poor data coverage or the
model's approximation of the real atmosphere could be compounded as trajectory segments
are added further back from the destination. Thus, the apparent origin of a 5-day backtrajectory should be considered an approximation, and, as indicator of source regions,
should be supported by other atmospheric data or, as is done here, used in a climatological
sense.
Despite their limitations, atmospheric trajectories have been frequently employed in
studies of long-range transport of gases and aerosols. Using back-trajectories, Herbert et al.