Assessment of the Quality of Thunderstorm Data at First-Order Stations

Abstract

Climatological assessments of past fluctuations in thunderstorms and other weather extremes require high-quality records. Data on thunder-day occurrences exist at U.S. first-order stations (FOS) since 1894 and represent the only long-term data available for temporal thunderstorm assessments. A thunder day is based on hearing thunder, and hence audibility is a major factor affecting record quality. Assessment of the potential conditions that affect audibility and of other factors that influence record quality identified four factors. Changes in audibility can occur from structural changes at the station, altered levels of noise such as increased aircraft operations, and relocations of stations to sites with different audibility conditions. Another data-limiting factor is local changes in storm activity due to urban influences on the atmosphere. Station relocations in areas where sharp natural differences in storm frequency exist such as in coastal and mountainous sites are a third factor affecting local values, and uncertainties related to record keeping are the fourth. The quality of the thunder-day records at the 130 U.S. FOS with 90 or more years of data during 1896–1995 were evaluated using a five-step process. First, the historical thunder-day frequencies were compared with the station relocations to detect any major changes after site changes. Second were tests of homogeneity using data from surrounding stations. The third step involved comparison of each station’s historical fluctuations with those of surrounding stations to identify anomalous values that persisted for 20 years or more. Step four involved assessment of possible urban effects on local storm frequencies using past findings and assessments of where and when urban storm activity may have changed in relation to the locations of the local weather station. Fifth, reported thunderstorms were compared with nearby cloud-to-ground lightning strokes to assess unreported thunder occurrences. Homogeneity testing revealed that 10 FOS had questionable records resulting from major multidecadal shifts in values, often a result of station relocations. Six other stations had urban influences on local storm activity that made their records unrepresentative of regional climate conditions, and four stations had anomalously low values of storms based on comparisons with nearby lightning activity. In all, 20 of the 130 stations with long records were found to have records unsuitable for long-term climatic assessments.