The use of a multifactor Southern Oscillation Index for the estimation of annual hailstorm frequencies in the Sydney area

Abstract

Relationships between the Southern Oscillation Index (SOI) and hailfall activity in the Sydney area were investigated with the help of a newly compiled Natural Hazards Research Centre (NHRC) hailfall database. The original hail data, which contain a record of all hailstone sizes for the period 1791–1994, were stratified into the form of (monthly, seasonal) annual hailday sums for the period 1935–1994 and cross-correlated with the SOI values stratified accordingly. With the exception of a marginally significant value for the month of November (r=0.26) no other significant simultaneous or lagged correlations were found between these two types of variables. A multifactor SOI was then developed as a possible indicator of the annual (June to May) hailstorm frequencies. The new three-factor SOI is composed of the sum of two SOI gradients (the differences between the October and June SOI values, and the November and February SOI values, respectively), minus the absolute value of the SOI average for the second half of the year. It is shown that high annual hailstorm activity tends to be associated with a high value of this index. High index values are generally reached during years marked by a rising SOI from June to November and a trend towards negative SOI values thereafter. It is suggested that this reversal of the relationship between the hail activity and the SOI can be attributed to similar changes observable in (low and high level) geopotential height anomalies over the area. The multifactor SOI leads to correlations of 0.6 (0.8) with the 1935–1994 (1985–1994) time series of annual hailday numbers. The three-factor index has more of a monitoring than a predictive quality (as it relies on the SOI values for the June to February period), whereas a subset index, consisting of the first and third component only, can have a predictive character. The latter partial index relies on the June to October data only and leads to nearly as high correlations with the annual hailday sums as the entire three-factor index (r=0.4–0.8). The stepwise development of the multifactor SOI is based on a multiple categorisation of the annual (seasonal) hailday numbers into several broad SOI categories according to the magnitude of the concurrent SOI values. Further details about the characteristics of these indices and their association with hailfall activity are described. © 1998 Royal Meteorological Society