What would it take to reduce deaths from tornadoes in the US?

The annual rate of deaths from US tornadoes, on a per capita basis, dropped by over an order of magnitude from 1925 to the early 1990s. Since then, the rate has levelled off or, perhaps, increased. Assuming that society has, as a goal, reducing tornado deaths, several possibilities exist for accomplishing that goal:

1) Improving the quality of the forecasts (accuracy, specificity, lead time)

2) Improving communication of the forecast messages (sender, receiver)

3) Improving preparedness (education, safety options)

Challenges exist in all of these areas and in their interplay. In particular, the lack of understanding of response and needs of user communities makes prioritizing efforts difficult.

Finally, some thoughts will be offered on the existence and desirability of the assumed societal goal.