A climatology, synoptic assessment, and thermodynamic evaluation for cloud-to-ground lightning in Georgia: a study for the 1996 Summer Olympics

Abstract

A lightning climatology within 50 km of nine outdoor venue locations for the 1996 Summer Olympics is produced. Spatial and temporal patterns were analyzed for July and August from 1986 to 1993. Active and inactive lightning days were isolated and thermodynamic parameters examined. Logistic regression routines were developed using the aforementioned parameters to differentiate active days from inactive days. A linear regression routine was developed to predict the lightning flash category within 50 km of Athens, Georgia based on the thermodynamic variables. At the inland locations, no pattern was found in the spatial distribution of cloud-to-ground lightning: the strike locations were random. At the coastal location, an inland maximum in ground flash density was observed. The patterns of ground flash density exhibited large year to year variability. Although there was great day to day variability, there was a diurnal progression of lightning with a broad minimum from 0600 UTC to 1400 UTC and a sharp maximum near 2200 UTC. Composite synoptic charts were produced for eight selected active days and eight selected inactive days. At the 500 hPa level the composited dewpoint depression was approximately 8 C less on active days than on inactive days. At the 850 hPa level the vector averaged windfields on active days revealed weakly anticyclonic southwesterly flow throughout Georgia. On inactive days the vector averaged winds exhibited a large anticyclone centered in northern Georgia. Strong correlation was found between lightning activity and several of the thermodynamic variables. The most highly correlated was a variation of convective available potential energy (CAPE) called CAPE 2000 with a correlation coefficient of .70. Showalter stability index and K index had correlation coefficients of .60 and .56, respectively. Logistic regression was used to predict active days versus inactive lighting days. The critical success index is nearly 80% and the probability of detection is over 93%. A model for independent data--inactive days versus all other days--had a CSI of 50.3%, but a false alarm rate of 46.6%. The linear regression model was not an exceedingly strong predictor. The coefficient of determination for the model was .3243.