Lightning data provide valuable information on the existence of thunderstorms and can be used to augment radar reflectivity data for the purpose of initializing ongoing convective systems within numerical weather prediction models. Radar data are generally not available over oceanic areas, terrain-blocked areas, and rural areas; regions where lightning data can provide additional information.

In the Rapid Update Cycle (RUC) cloud analysis and diabatic digital filter initialization (DFI) procedure, lightning data are converted to reflectivity using a simple assumed relationship between flash density and reflectivity and then used within the DFI to induce convective-scale features within the model initial fields. This storm-scale initialization capability (using both reflectivity and lightning data) has yielded impressive improvements in RUC model predictions of ongoing convection. In addition, 3-km High Resolution Rapid Refresh (HRRR) forecasts initialized with the RUC radar/lighting DFI procedure have shown marked improvement over parallel 3-km forecasts without the special assimilation procedure. This same storm-scale initialization capability is being ported to the Rapid Refresh (RR) system, which will then replace the RUC as the source for HRRR initial fields. With the greatly expanded RR domain (covering all of North America and adjacent waters compared to the CONUS RUC domain), the importance of lighting data as a convective indicator will increase substantially.

In order to obtain a more realistic relationship between lightning flash rate and reflectivity, we have calculated a statistical relationship between the number of flashes per grid column and the corresponding column maximum of model grid-box averaged reflectivity (obtained from the NSSL 1-km radar reflectivity mosaic product). A preliminary relation based on a short period was introduced in the 2nd International Lightning Meteorology Conference and here we will present the results from a longer period, which yields a more realistic relationship between lightning and radar reflectivity.

We will then present preliminary results from the use of lighting data within the RUC/RR domains, including the use of Alaska BLM data over a radar data-sparse portion of the RR domain. The lightning observations also provide important information for deducing the cloud types in the cloud analysis. This information has been used in a complex cloud analysis to improve the hydrometoer distributions and experiments showing results from assimilating lightning data as a supplement of reflectivity and convective system through the cloud analysis of RR are also introduced.