The nature of detecting thunderstorm occurrences based on audibility has the potential for inconsistent record keeping. Since 1947, records were kept at FOS of the time when thunder was first heard and the time 15 min after the last peal of thunder, and each such period was defined as a “thunder event” (Changery 1981). A thunder event can and often does occur more than once a day. The average number of events per thunder day is 1.6 at points in the central United States, diminishing to 1.1 on the West Coast and to 1.4 on the East Coast (Changnon 1988b).
The recording of a thunder day at first-order stations of the National Weather Service has been done since the 1870s. Beginning in 1894, a day with thunder was one when one or more distinct peals of thunder were heard, but before 1894 a thunder day was recorded only when it was accompanied by rain, and this shift is a major discontinuity in the records. Hence, a day when thunder was heard at a first-order station has been defined as a “thunder day” since 1894 (Weather Bureau 1951), and these data serve as the only long-term records of thunderstorm activity in the United States. Volunteer observers at cooperative substations have also been asked to record a thunder day when they heard thunder, but this was a voluntary endeavor and not all observers have kept records of thunder occurrences.
The thunder data from both types of stations need to be evaluated for their accuracy. A procedure for evaluating cooperative substation records of thunder-day data has been devised, tested, and found useful for detecting periods with quality substation values (Changnon 1967). This paper describes an assessment of the quality of the historical thunder-day data collected at 130 FOS in the contiguous United States with long records covering all or most of the 1896–1995 period.
