Use of High-Resolution WRF Simulations to Forecast Lightning Threat

1. Use of High-Resolution WRF Simulations to Forecast Lightning Threat E. W. McCaul, Jr. USRA Huntsville/SPoRT National Space Science and Technology Center Huntsville, Alabama, USA Science Advisory Council 13 June 2007 Photo, David Blankenship Guntersville, Alabama

2. Premises and Objectives Given: Precipitating ice aloft is correlated with LTG rates Mesoscale CRMs are being used to forecast convection CRMs can represent many ice hydrometeors (crudely) Goals: Create WRF forecasts of LTG threat, based on ice flux in layers near -15 C, and on simulated reflectivity aloft. WRF: Weather Research and Forecast Model CRM: Cloud Resolving Model Additional Forecast Interests CI - convective initiation Ti - First lightning (35 dBZ at -15C, glaciation) Tp - Peak flash rate Tf - Final lightning

3. 0 oC Flash Rate Coupled to Mass in the Mixed Phase Region Cecil et al., Mon. Wea. Rev. 2005 (from TRMM Observations)

4. WRF Lightning Threat Forecasts:Methodology Use high-resolution (2-km) WRF simulations to prognose convection for a series of case studies Develop diagnostics from model output fields to serve as proxies for LTG: graupel fluxes, reflectivity profiles, etc. Calibrate WRF LTG proxies with actual total LTG rates from HSV LMA and reflectivity/LTG from TRMM PR/LIS 4. Draw maps of WRF LTG threat based on calibrated proxies Assess WRF capabilities for forecasting LTG threat

5. Compute field of graupel flux at -15C Plot peak values against observed peak LTG flash rates on the same grid, for high-LTG, low-LTG cases; empirically estimate the nature of the relationship between observed LTG, forecast proxy; estimate any coefficients needed to convert gridded proxy to gridded LTG rate Here, we find good fit from F1 = 0.06 w qg (see next page) With histogram truncation, could match areal coverage to observations; not done here because of loss of some simulated storms which appear to be important After applying conversion function to WRF proxy field, obtain field of LTG flash rates F1 in units of fl/(5 min)/grid column LTG Threat 1 derived from WRF graupel flux at -15C

6. Prototypical Calibration Curves Threat 1 (Graupel flux)

7. Compute field of reflectivity Z1 at -15C Compute reflectivity field Z2 at 3 km above -15C Regress LIS flash rates against TRMM PR Z1 and Z2 for SE USA warm season; obtain three coefficients, one for Z1, one for Z2, and a threshold; this yields an uncalibrated reflectivity profile shape parameter Z that can be evaluated in WRF output: Z = aZ1 + bZ2 + c, where a=1.0856, b= 0.5196, c = -49.49 Plot peak values of LMA flash rate density vs peak gridded Z, for low-LTG and high-LTG cases; empirically estimate functional relation; plot shows possibly quadratic functional relationship, F2 = 0.005Z2 (see next page) Apply functional equation to Z data to obtain flash rate density field F2 in terms of fl/(5 min)/grid column LTG Threat 2 derived from WRF reflectivity profile shape

8. Prototypical Calibration Curves Threat 2 (dBZ profile)

9. Methods based on LTG physics, and should be robustly useful Methods supported by solid observational evidence Can be used to obtain quantitative estimates of flash rates Methods are fast and simple; based on fundamental model output fields; no need for complex electrification modules Methods can be revised to include thresholds, which will allow truncation of proxy histograms so as to allow matching of areal coverages of predicted and observed LTG activity LTG Threat Methodology: Advantages

10. Methods are only as good as the numerical model output; model used here (WRF) has shortcomings in several areas Inherent ambiguity in interpretation of threats (extent density, etc.) Small number of cases means uncertainty in calibrations; choices of coefficient values may be off by factor of 2, but such errors are small compared to gross model errors in storm placement Calibration requires successful simulation of both high- and low-LTG cases; the latter can be hard to find; in most cases, there are at least a few moderate or strong-LTG storms present in the domain Calibrations must be redone whenever model is changed or upgraded LTG Threat Methodology: Disadvantages

11. WRF Lightning Threat Forecasts:10 December 2004Cold-season Hailstorms, Little LTG

12. 2-km horizontal grid mesh 51 vertical sigma levels Dynamics and physics: Eulerian mass core Dudhia SW radiation RRTM LW radiation YSU PBL scheme Noah LSM WSM 6-class microphysics scheme (graupel; no hail) 8h forecast initialized at 12 UTC 10 December 2004 with AWIP212 NCEP EDAS analysis; Also used METAR, ACARS, and WSR-88D radial velocities at 12 UTC; Eta 3-h forecasts used for LBC’s WRF Configuration (typical)10 December 2004 Case Study

13. Ground truth: LTG flash extent density, dBZ10 December 2004, 19Z

14. WRF fcst: LTG Threat 1, SBCAPE10 December 2004, 19Z

15. WRF fcst: LTG Threat 2, dBZ at -15C10 December 2004, 19Z

16. WRF Lightning Threat Forecasts:30 March 2002Squall Line plus Isolated Supercell

17. Ground truth: LTG flash extent density, dBZ30 March 2002, 04Z

18. WRF fcst: LTG Threat 1, SBCAPE30 March 2002, 04Z

19. WRF fcst: LTG Threat 2, dBZ at -15C30 March 2002, 04Z

20. Domainwide Peak Flash Density Time Series

21. Implications of time series:1. WRF LTG threat 1 peak values may have small pos. bias2. WRF LTG threat 1 peak values have proper t variability 3. WRF LTG threat 2 peak values may have bigger pos. bias4. WRF LTG threat 2 peak values have insufficient t variability5. WRF LTG threat mean biases are positive because our method of calibrating was designed to capture peak flash rates correctly, not mean flash rates6. WRF LTG threat t variability is too small for Threat 2, because of integrating effect of multi-layer methods7. WRF LTG threat 1 should give best performance

22. Conclusions 1:1. WRF forecasts of convection are useful, but of variable quality2. Timing of initiation of convection is depicted fairly well; however WRF convection is sometimes too widespread 3. Inclusion of WSR-88D velocity data helps; so does dBZ if storms already up at t=04. WRF wmax, dBZ values on 2-km grids sometimes too weak, relative to observed weather and 0.5-km RAMS simulations5. WSM6 microphysics still too simple; need more ice categories6. Finer model mesh may improve updraft representation, and hydrometeor amounts7. Biggest limitation is likely errors in initial mesoscale fields

23. Conclusions 2:1. Current LTG threat parameters provide more spatial coverage of threat as compared to actual LTG, but less than standard weather forecasts and less than area covered by CAPE>02. Coverage of graupel-flux LTG threat could be reduced by using a threshold value of graupel flux, but doing so eliminates some WRF convection that might be important3. Because of location, timing errors in WRF convection field, excess LTG threat spatial coverage may be desirable for now4. Graupel-flux LTG threat 1 shows large time rms, like obs; dBZ LTG threat 2, based on 2 levels, has small time rms

24. Future Work:1. Expand catalog of simulation cases to obtain robust statistics; consider physics/IC mini-ensemble simulation approach2. Continue refinement of dBZ profile LTG threat 2 method3. Test newer versions of WRF, when available: - more hydrometeor species - double-moment microphysics4. Run on 1-km or finer grids; study PBL scheme response5. Examine sensitivity of calibration constants to model physics and IC diversity6. In future runs, examine fields of interval-cumulative wmax, and associated hydrometeor and reflectivity data, not just the instantaneous values; for save intervals of 15-30 min, events happening between saves may be important for LTG threat7. Publish findings and collaborate with NWS, others

25. Acknowledgments:This research was funded by the NASA Science Mission Directorate’s Earth Science Division in support of the Short-term Prediction and Research Transition (SPoRT) project at Marshall Space Flight Center, Huntsville, AL. Collaborators:Kate LaCasse, UAH/SPoRTSteve Goodman, NOAA/NESDISDan Cecil, UAH