The WRF Lightning Forecast Algorithm: Recent updates and results from convective ensemble forecasts

Three years ago, McCaul et al. (2009) devised and published a simple, empirical proxy-based method for converting selected fields from cloud-allowing models into two-dimensional fields of estimated lightning flash origin density. Since its development, interest from the community has led to its incorporation into a number of high-resolution forecast models, allowing more widespread testing and analysis of its performance across varying seasons and locations. To date, the Lightning Forecast Algorithm has been incorporated in the Weather Research and Forecasting (WRF) model runs performed at the National Severe Storms Laboratory, the WRF ensembles run by the Center for Analysis and Prediction of Storms at the University of Oklahoma, and, more recently, in the High Resolution Rapid Refresh runs disseminated by the Environmental Science Research Laboratory. Results have shown that the LFA produces generally reasonable lightning flash density fields, but that in 2010-2011 model runs, it tended to underforecast lightning rates in extreme storms, and to produce numerous false alarms in cold stratiform precipitation regimes. Simple changes to the LFA have been implemented for use in 2012 forward, including requiring that the vertical ice integral proxy have its peak value set equal to that from the graupel flux proxy before blending the two to get a final answer, and setting higher thresholds for screening out spurious weak threats. Analyses of LFA results from the 2011 CAPS WRF ensemble runs show that most combinations of physics packages employed yielded lightning flash rate estimates somewhat smaller than that from the default WSM6 configuration used for LFA development, and that there is a not unexpected sensitivity to variations in the microphysics schemes used. This sensitivity tends to increase slowly with the amounts of lightning forecast, such that the absolute variability is largest for very high lightning events, but the relative or fractional variability is largest for very low lightning events.