As noted in the last section, the mid-boundary layer backward trajectories that used the data and divergence methods for vertical motion did follow the tracer plume and passed near the tracer release location.
- Another way to analyze the situation is to look at something similar to a weather map. Open the Meteorology / Display Data / Contour Map menu tab and select the global reanalysis file captex2_gblr.bin and the geopotential height HGTS field. The reanalysis is used for this example instead of WRF because it covers a much larger domain and it contains the height field. Above the boundary layer, the contours of the height field are parallel to the wind velocity vector (geostrophic) providing a quick visual of the flow.
α ∂p / ∂y = -f u
In addition, make the following changes to the contour map menu: select level number 4 which will be the 850 mb surface, set the time offset radio-button to 48 to draw the map at the end of the sampling period near the start time of the back trajectory, set the map center to 40N 80W, and increase the map radius to 25 degrees.
- The resulting 850 mb heights map shows a high pressure system centered over the eastern U.S. with a divergent region just off the coast. This region illustrates how the back trajectories easily got caught in different flow regimes, depending upon small differences in height between each method, resulting in very different horizontal trajectories primarily due to large variations of wind direction with height. The flow region to the north of the high pressure system contained most of the tracer.
In general, trajectory calculations should use the meteorological data's vertical velocity fields when available. Each of the computational approaches has its own limitations, isentropic in diabatic conditions (through clouds), isobaric over large terrain variations, and the divergence is sensitive to small differences in velocity gradients. Even the data method can become unreliable when meteorological grid sizes decrease but the data time interval remains unchanged, resulting in under-sampling a field that may change direction multiple times between output intervals.