I apologize in advance for the length of this reply...
Tornado paths are generated via the touchdown and liftoff latitude/longitude points given in the NWS historical data file, which can be found on Dan McCarthy's page here
http://www.spc.noaa.gov/wcm/ . Each line within the file represents either a unique tornado or a state specific segment of a multi-state tornado (actually, a 3rd type of line exists that is just continuation data for an event that takes up more than 1 line, but, for simplicity, we can exclude that from this discussion...) For the April 19, 2000 Parsons, Kansas tornado that "jketcham" mentions (shown here
http://www.tornadohistoryproject.com/map.php?i=40751) the
corresponding line in the NWS historical data file is:
2000.00,13.00,20.00,4.00,19.00,4/19/2000,19.50,3.00,709.00,1.00,,1.00,1.00,1.00,37.22,95.55,37.42,94.93,340.00,,1.87,5.60,,87.00,0.00,,71.42,125.00,99.00,133.00,37.00,"3",1.87,,"709"
Commas are used as delimiters to separate various pieces of information, but for the purposes of this discussion the relevant numbers are 37.22, 95.55, 37.42, 94.93. These numbers represent the touchdown lat/lon (37.22, 95.55) and the lift-off lat/lon (37.42, 94.93) as explained in the SPC format code instructions here
http://www.spc.noaa.gov/wcm/torforma.doc.
As for the actual plotting, the Google Maps API allows for the drawing of "polylines" between any number of points (polylines are the same things that show you your route when viewing driving directions....) In this case, there are only 2 points given in the NWS file: touchdown and liftoff. So,
necessarily, the path generated from these coordinates is a straight line (except for multi-state tornadoes, which have a new set of points at state lines...) Obviously, tornadoes don't move in a perfectly straight line, but zig and zag. The longer the path, the more potential for variation. Additionally, coordinates are only given to a precision of 2 digits. So, the coordinate (37.22, 95.55) could have in reality been (37.2167, 95.517) or something similar, which can probably add up to a difference of a couple hundred feet or so. But since the NWS only gives 2 digits we'll never know. Also, the width (thickness) of the polyline (tornado path) is currently a fixed number of pixels (although I could, in theory, adjust the thickness of the polyline to reflect the reported width of any specific tornado....) One must account for the actual width of the tornado on either side of the polyline. So, a tornado that appears to miss a town (such as Parsons, Kansas) on the map may look that way due to a combination of rounding errors in the coordinates and the non-displayed total path width.
Having said that, I suspect that a good deal of the discrepancies in the displayed paths stem from less than perfect record keeping by NWS, or just plain incorrect information given by NWS personell in their survey (remember, much of the coordinate information was generated in a pre-GPS world...) What I am absolutely sure of is that the information displayed on the map is consistent with the information given in the historical data file. Short of manually researching all 48,000+ tornadoes in the database or collecting eyewitness accounts, I think its the best we can hope for
Even taking the above comments into account, I know that some paths are spot on. For example, a tornado that hopped and skipped its way directly over my sister's house (shearing off the top of a tree in the process) is displayed exactly, and I mean
exactly over the location of the formerly healthy tree.
Speaking of eyewitness accounts, that's one of the reasons I have enabled users to leave comments, so that they can add their own experiences to the "official" information given by NWS. I encourage each of you that have noted path discrepancies to leave a comment on the appropriate tornado
details page in order to set the record straight.
On a related note, some of you may have noticed this disclaimer on the tornado details page:
"*The National Weather Service (NWS) data file contains touchdown latitudes and longitudes for all tornadoes. However, lift off latitutudes and longitutudes are not given on all tornadoes. When both are given, an approximate track is displayed. Note that since a maximum of only two sets of coordinates are given, the resulting approximate track must be drawn as a straight line. In reality, the actual path of the tornado between the touchdown and liftoff points can vary significantly, particularly for longer paths. Also note that the path displayed is only as accurate as the data within the NWS data file, which gives coordinates (latitude and
longitude) to 2 digits only."
Somewhere along the line this disclaimer has been inadvertently removed from the general tornado map page. I will be reinserting it shortly.
Thanks for your interest in the site. If anyone has suggestions for improving the site I welcome your comments (I also welcome criticisms
...)
Josh Lietz
