Prediction of Total Lightning Behavior in Colorado Thunderstorms from Storm Dynamical and Microphysical Variables

Accurate prediction of lightning flash rate is essential to successfully parameterize the production of nitrogen oxides by lightning (LNO_x). In this study, new flash rate parameterization schemes are developed using observations of 10 Colorado thunderstorms from the Deep Convective Clouds and Chemistry (DC3) and the CHILL Microphysical Investigation of Electrification (CHILL-MIE) field projects. Storm total flash rates were determined by automated clustering of lightning mapping array (LMA)-detected radiation sources. Storm parameters, including hydrometeor and reflectivity echo volumes, ice masses, and measurements of updraft strength were obtained from polarimetric radar retrievals and dual-Doppler derived wind fields. Echo volumes exhibited a particularly strong correlation to flash rate (R² = 0.78 for 30-dBZ echo volume).

It is shown that existing flash rate schemes tend to underestimate flash rate for the storms in this study. New parameterizations developed based on the graupel echo volume, the precipitating ice mass (graupel and hail), and the 30-dBZ echo volume within the mixed-phase region of storms predicted flash rate trends reasonably well. However, there were sometimes large errors in the prediction of flash rate magnitudes, possibly due to fluctuations in storm updraft intensity. Updraft-based flash schemes were developed but significantly underestimated flash rate for the storms studied. It is shown that very high flash rates correlate differently to updraft strength than low flash rates. We hypothesize why this behavior is observed. The use of multiple storm parameters to predict flash rate was also investigated, and the results are improved somewhat compared to single-parameter schemes. New flash schemes were tested for storms outside of Colorado to examine their potential regional dependence. Finally, observations of the relationship between flash rate and flash size are discussed, with implications for the improved prediction of LNO_x.