It's an approximation. It's definitely not exact.
1. I take the start and end points and assume the tornado is a straight line (which it isn't, but it's the best I can do with the data).
2. I "draw" the tornado onto an image where each pixel is 5km by 5km.
3. Next, make the assumption that the tornado is 1km wide.
4. Make the next assumption that the tornado goes down the center of the 5km grid cell, which isn't a stretch considered we assumed a straight line and then "drew" the tornado path.
5. The +/- 2km is the remaining portion of the 5km grid width. (2km left + 1km tornado + 2 km right = 5 km grid cell).
Is it exact? No. That's why I mention it's an approximation.