Lightning Threat Forecasting Background
Many parts of society have an interest in improved short-term forecasts of lightning threat. These include the aviation, spaceflight, military, and outdoor recreation sectors, some of which depend on the National Weather Service (NWS) for guidance. To address this need and, ultimately, to provide the NWS with better tools for anticipating lightning threat, SPoRT researchers have brought together their experience in lightning observations and mesoscale modeling to devise techniques that can provide quantitatively useful lightning forecast fields based on reliable proxy fields from cloud-resolving mesoscale models. The goal of this research effort is the creation of products that can be used by operational NWS and other forecasters in real-time. The only guidance available currently from NWS is an uncalibrated probability of lightning product.
Figure 1. Low-level radar reflectivity from KHTX Doppler radar at 0400 UTC 30 March 2002 (grayshades), and LMA-derived integrated flash extent density (contours) for a 5-min period at the same time. Flash extent density is depicted for clarity, although all calibrations and predictions of lightning threat are based on observations of flash origin density.
Research into lightning forecasting at SPoRT has relied on recent global observational findings from the TRMM-LIS satellite platform that show strong correlations between storm flash rates and other storm characteristics, such as the amount of large precipitating ice lofted by updrafts above the -15°C level, and the shape of the reflectivity profile. Aware of these findings, SPoRT researchers have performed high-resolution (2-km) simulations of a number of North Alabama storm cases using the WRF model, and have generated model-derived time- and space-dependent proxy fields expected to be proportional to total storm flash rate. To convert these proxy fields into WRF-based quantitative estimates of the flash rate, a calibration step is incorporated in which the case-by-case regional peak values of the simulated proxy fields are compared to the corresponding regional gridded peak lightning flash rates observed by the North Alabama Lightning Mapping Array (LMA). The LMA system is comprised of a network of 10 sensors that detect radiation from lightning channels, infer the times and locations of the radiation events, and allow the three-dimensional mapping of lightning channel segments in storms. Other algorithms then aggregate these segments into discrete lightning flashes, whose rate of occurrence is what is compared to the WRF-simulated proxy fields.
Figure 2. WRF-derived surface-based pseudoadiabatic CAPE at 0400 UTC 30 March 2002 (grayshades), and WRF-predicted flash origin density (contours) for a 5-min period at the same time, based on WRF graupel flux at the -15°C level. Instantaneous aerial coverage of predicted flash density is printed at the bottom of the figure.
The calibration of WRF data against LMA observations has been conducted for a series of case studies that cover the range of intensity of storms found in the Tennessee Valley area. The lightning threat forecasts obtained in this research appear reasonable, and are achieved without the expense of adding expensive electrification schemes to the model. However, models such as WRF still cannot reliably pinpoint exactly where convective storms will form, and how strong each individual storm will be. Current results are most useful in anticipating lightning events on the 100-km scale, up to 12 h into the future. Improvements in the specificity of WRF lightning forecasts will require higher resolution model runs with more sophisticated ice microphysics, and more accurate initial mesoscale analyses used in starting the model.
Figure 3. WRF-derived reflectivity at the -15°C level at 0400 UTC 30 March 2002 (grayshades), and WRF-predicted flash origin density (contours) for a 5-min period at the same time, based on WRF two-level reflectivity profile parameter anchored at the -15°C level. Instantaneous aerial coverage of predicted flash density is printed at the bottom of the figure.
Update for 2010
Although the calibrations for the LMA cases studied in our first publication were derived from a variety of
different weather scenarios, all were derived from cases that occurred only in the Tennessee Valley area. To check
the robustness of our findings, we are beginning to examine cases from other times and other locations. The North
Alabama LMA-based algorithm has been coded into WRF models that are being run on additional storm cases on a
nationwide basis, and validated against new LMA data for Tennessee Valley thunderstorm events. Preliminary findings
appear promising, and the algorithm is now being prepared for inclusion in the operational WRF runs to be conducted
by NOAA researchers at the Storm Prediction Center in Norman, OK, beginning in Spring 2010. Test cases from these
forecast runs will provide an abundance of opportunities to validate the algorithm against a variety of storm systems
sampled by several LMA networks nationwide.
