Usefulness of Storm-Scale Model Guidance in SPC Fire Weather Events with Dry Thunder

Wildfires are a serious threat to life and property in the United States. To mitigate this threat, the National Weather Service Storm Prediction Center (SPC) is charged with creating national fire weather guidance in the form of daily fire weather outlooks. These outlooks are typically issued where there is a risk of:¨1) very low relative humidity, strong winds, and dry fuels, or 2) dry thunderstorms (hereafter “dry thunder”), which are a primary ignition source for wildfires. Dry thunder, defined as a convective storm that contains lightning but produces less than 0.1 inches of rain in a given location, is particularly difficult to predict accurately. While precipitation has been a common and long-standing output of numerical models, explicit lightning output (lightning threat) has only become available in the past few years to SPC forecasters through a total lightning algorithm (McCaul et al. 2009) that has recently been applied to several convection allowing models, including the experimental Advanced Research WRF (ARW) run daily by the National Severe Storms Laboratory (NSSL WRF).

The purpose of this study is to determine how well the NSSL WRF predicted dry thunder during the summer (June-August) of 2011 using different parameter thresholds to identify a dry thunder event. Experimental forecasts of dry thunder were created by utilizing several threshold values for lightning threat, using various methods to account for spatial uncertainty, and incorporating different constraints for environmental moisture and precipitation. Observed NLDN cloud-to-ground (CG) lightning activity, precipitation, and analyzed precipitable water fields were then used to objectively verify these experimental forecasts. Overall, forecasts were found to have a high false alarm ratio (~0.70) and a low critical success index (~0.20). However, the high FAR may be at least partially attributable to the use of CG lighting as the verifying event, which occurs less frequently compared to total lightning. It was also found that with improved tuning of parameter thresholds, some of the forecasts produced a probability of detection of 0.75 or greater with lower bias scores. This suggests that these forecasts may provide useful first-guess guidance to assist fire weather forecasters.