OK, I fixed the plotting inaccuracy, to within +/- 1 pixel. So now the tornadoes that are over land will be... over land, not water. :) The plotting on the Unisys maps still needs work. (I don't know if you use them or not, but in the outlook editor, you can select the map type, where you get options for 850 mb, 300 mb, and RH/Lift. These maps are from Unisys.) They use a different projection, and I haven't figured out the parameters for that yet.

 

The pick-points are still a bit rough. Basically, the map is divided up into an 8-pixel grid, and wherever you click, it finds which 8x8 square you clicked in. If I re-write the point-picker in JavaScript, I can get it accurate to the pixel, but that will take more time.

 

BTW, I was thinking about altering the scoring algorithm. There have been a number of very high-quality outlooks, but they got bad scores because of a few tornadoes on the other side of the country, so there was a big "dst missed" factor. So I'm thinking that a few stray tornadoes shouldn't hurt a good outlook, and that perhaps I should do something like disregard a certain number (or percentage) of the tornadoes that were the furthest away. So I'll look into that.