So far I’ve located three slightly different versions of the Atlantic track data:
Here are a few boring comments on these data sets for anyone who’s working on them.
The first two versions have the same number of lines and only 35 of 13085 lines differ. None of the differing lines record wind speeds, but appear to contain codes for landfalls and things like that, codes which I haven’t been using so far. Wind speeds are in knots.
The third dataset seems to be the most official. I haven’t fully reconciled the differences, but here’s what I’ve noticed so far. Wind speeds in the third data set are recorded in mph and kph. In spot checks, the mph and kph speeds are each calculated by multiplying the speed in knots by the appropriate conversion factor and then rounding to the nearest 5. So there’s not a lot of precision in the measurement.
In all but 3 storms, the number of measurements matches. Two storms – 1166- Barry and 1360 (Ophelia -2005) – have 1 (2 – Ophelia) extra reading at low wind speeds and the difference is irrelevant. For the other hurricane (Andrew- 1992), there was a material difference and it was very hard to figure out why there was a difference. Eventually I figured out that that the updated version had not recorded measurements in 6 hour intervals, but had included 4 or 5 measurements at intermediate times during very high wind speed intervals – contrary to the practice in the data set. One can easily work around the difference, but if you were assuming (reasonably) that the data was coming in 6-hour intervals, it would lead to an error.