Abstract
An experiment was carried out to collect statistics on the observability of severe-storm infrasound at three stations during the 1973 storm season. The results have been evaluated with the help of four “indices of usefulness”:
•False-alarm rate, which tells how often infrasound from other sources is mistaken for that from storms. We devised a sorting procedure that reduces the false-alarm rate to 15–20%, and still lower rates seem achievable.
•Detection rate, which tells what fraction of severe storms are detected. Here the big problems are defining what we mean by “severe storm” and verifying their occurrence; we estimate a 65% detection rate for tornadic storms, a 31% detection rate for tornadoes themselves, and a 33% detection rate for storms with radar tops above 50,000 ft.
•Timeliness, which tells how much advance warning the waves give compared to dangerous storm effects. It was practical to consider only tornadoes from this viewpoint, and we found that the emissions tend to precede tornado onset by an hour or so.
•Location accuracy, which tells how well the emissions can be used to locate and track storms. This index is hard to evaluate quantitatively, as illustrated by the six cases where storms were seen at all three stations. Propagation effects and measurement uncertainty presently prevent positive identification and tracking of a particular storm, but we see ways to improve this.
The answer to the title question is that the emissions show promise as a supplement to the present warning system. A question remains about the cost-effectiveness of doing the additional required research and deploying an operational sensor network.
