Development of a calibrated proxy for thunderstorm occurrence using lightning and reanalysis data

The aim of our study is to model the occurrence of severe convective weather events such as large hail, tornadoes and severe wind gusts using climate models and reanalysis data. A common approach is to use predictors such as CAPE and shear-related parameters. The big unknown of this methodology is the factor of convective initiation. Storms may not occur even though the atmosphere is in an area of CAPE/shear space that would support severe storms once initiated. Hence, we would like to take convective initiation into account explicitly. The probability of a severe storm is the product of the probability that a storm forms and the likelihood of it becoming severe:

P(severe storm) = P(storm initiation) x P(severe|storm initiation)

In our study, we focused on Central Europe and calculated several parameters related to instability, moisture or shear in the time period 2007-2013 from the ERA-Interim global atmospheric reanalysis. We set out to find a best function for P(storm initiation). Therefore, we scrutinized the connection between the derived parameters and storm initiation using lightning data from the EUropean Cooperation for LIghtning Detection (EUCLID).

Our results include the following findings:

First, higher lightning probabilities can be observed with increasing CAPE values with a saturation, or even slight decrease setting in from approximately 800 J/kg, followed by strong increase at values in excess of 2500 J/kg.

Second, for a given CAPE value, lightning probability is relatively large both for very low (< 5 m/s) and high values (> 15 m/s) of 0-6 km bulk shear. Apparently there are two shear regimes that favor storm initiation and/or sustenance. We found similar findings for low level 0-1 km bulk shear with higher lightning probabilities for values below 2 m/s and above 5 m/s. Furthermore, we find that thick shallow CAPE produces less lightning than long thin CAPE.

Last, considering CAPE and CIN, lightning probability is highest for large CAPE and small CIN values but for small CAPE values, higher probabilities are found in midrange CIN values.