

The advection of a particle or puff is computed from the average of the threedimensional velocity vectors at the initialposition P(t) and the firstguess position P'(t+Δt). The velocity vectors are linearly interpolated in both space and time. The first guess position is
The integration time step (Δt) can vary during the simulation. It is computed from the requirement that the advection distance per timestep should be less than the grid spacing. The maximum transport velocity is determined from the maximum transport speed during the previous hour. Time steps can vary from 1 minute to 1 hour and are computed from the relation,
The integration method is very common (e.g. Kreyszig, 1968) and has been used for trajectory analysis (Petterssen, 1940) for quite some time. Higher order integration methods will not yield greater precision because the data observations are linearly interpolated from the grid to the integration point. Trajectories are terminated if they exit the meteorological data grid, but advection continues along the surface if trajectories intersect the ground. In the horizontal, the integration of the position vector is done in grid units, while in the vertical, a normalized sigma coordinate system is used, where sigma is defined by
The animation shows the sequence of model computations of a single trajectory (red) from 750 m AGL with the associated 900 hPa height contours (black). Above the boundary layer, when the winds become geostrophic, the trajectory path is parallel to the height contours. It should be noted that the variable time step can at times lead to what might appear as inconsistencies in the calculation results. This may be particularly evident when there are strong gradients in the wind field (in space or time) that may not be adequately represented by the resolution of the gridded meteorological data. For instance, when calculating two or more trajectories at the same time compared with their calculation individually, because the maximum wind speeds for each trajectory may differ, the time steps could be different for the individual calculations, resulting in slightly different trajectories. This is not an error but should be treated as a cautionary note that less confidence should be placed on an individual trajectory. Some of the issues relating to trajectory uncertainty will be examined in more detail in subsequent sections. 