I am trying to better understand what the best approach is for calculating the trajectories I need. I have mostly looked at using the EDAS data, but it has a lot of holes in it for the period I am interested in, so we wind up with trajectories truncated or not calculatable. I tried to fill in the holes with the FNL data. However, in one case when I calculated 120-hour backward trajectories from Potsdam, NY at 1800 on April 8, 1999 using both the EDAS and FNL data bases, I found that the EDAS data suggested the air was in the upper midwest 5 days prior, but the FNL put it into northern Quebec. Why the difference?
Although all meteorological models use the same observations, there are three significant differences between EDAS and FNL: horizontal, vertical, and temporal resolution. EDAS has better resolution in all categories. For the case you sent me I re-ran the EDAS trajectory using starting heights of 300, 400, and 500 m agl. There are large differences in the first 24 hours. The 400 m trajectory corresponds more with your 500 m FNL. EDAS vertical resolution is every 25 hPa, FNL resolution is limited to mandatory surfaces: 1000, 925, 850 hPA etc. The usual procedure should be to run trajectories at several heights and locations about the point of interest — if the trajectories are similar then the confidence level is high — if they diverge then there is more uncertainty.