TORNADO-WARNING PERFORMANCE IN THE PAST AND FUTURE: A Perspective from Signal Detection Theory

摘要

The National Weather Service (NWS) issues tornado warnings and collects observations to evaluate those warnings. Historically, the evaluation has consisted of the probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI). These quantities can be derived from a 2X2 contingency table (lable 1). POD and FAR are clearly not independent of each other, and CSI provides no additional information. In practice, one could improve POD by issuing warnings on more storms, but thatwould almost certainly increase the FAR. Increasing POD while decreasing FAR at the same time requires improvements in scientific knowledge or technological application of that knowledge, or improvements in identifying events as tornadic or nontomadic. It would be nice to have a technique to estimate the effects of those changes of issuing additional warnings and improvements in science and/or technology.