1. 1 st UNSTABLE Science Workshop 18-19 April 2007 Science Question 3: Science Question 3: Numerical Weather Prediction Aspects of Forecasting Alberta Thunderstorms Jason Milbrandt Recherche en Prévision Numérique [RPN] (Numerical Weather Prediction Research Section), Meteorological Research Division, Environment Canada

2. 1 st UNSTABLE Science Workshop 18-19 April 2007 OUTLINE of PRESENTATION 1.ISSUES related to high-resolution NWP 2.DEMONSTRATION of high-resolution simulations 3.SUB-QUESTIONS 4.CONCLUSION

3. Premise: NWP part of UNSTABLE should focus on maximizing and improving the usefulness of high- resolution Canadian models as forecasting tools for severe convection (with emphasis on storms that originate along the foothills) Science Question 3

4. Science Question 3: To what extent can high-resolution NWP models contribute to forecasting the initiation and development of severe convective storms that originate in the Alberta foothills? That is: 1.How can the usefulness of the current GEM-LAM (2.5 km) forecasts be maximized? 2.How can the forecasts from the GEM-LAM be improved? Science Question 3

5. ISSUES Pertaining to Science Question 3

6. Issues Pertaining to Science Question 3: (and high-resolution NWP in general) DATA - initial conditions of model are crucial  high-resolution data assimilation is a complex topic  research on development of high-resolution analyses is possible in context of UNSTABLE - IOP data useful for verification (examination of sub-questions) Science Question 3

7. Issues Pertaining to Science Question 3: (and high-resolution NWP in general) DATA COMPUTATIONAL RESOURCES -high-resolution NWP is very expensive -resources are increasing  This is a logical time to examine Science Question 3 Science Question 3

8. Issues Pertaining to Science Question 3: (and high-resolution NWP in general) DATA COMPUTATIONAL RESOURCES MODEL DETAILS -GEM is an appropriate tool for this job non-hydrostatic, limited-area model (LAM), sophisticated physics -To be learned: 1. modelling strategies 2. improvements to physics parameterizations Science Question 3

9. Issues Pertaining to Science Question 3: (and high-resolution NWP in general) DATA COMPUTATIONAL RESOURCES MODEL DETAILS DETERMINISTIC vs. PROBABILISTIC -Use of any single model run is inherently deterministic -Analysis errors mean that a single high-resolution NWP run is doomed  Ensembles of runs will ultimately need to be considered for high- resolution NWP Science Question 3

10. EXAMPLES of simulated thunderstorms

11. Estimated from WMI radar: 4-km GEM simulation:* * Source: Erfani et al. (2003), Meteorol. Atmos. Phy. 14 July 2000 Pine Lake Storm Accumulated Precipitation RADAR

12. mm 40 30 25 20 16 13 10 8 6 4 RADAR 33 mm 1-km CNTR 8:00 pm 1-km MC2 simulation:* 50 km N Estimated from WMI radar: 8:00 pm 50 km N 30 25 20 15 10 5 mm 14 July 2000 Pine Lake Storm Accumulated Precipitation * Source: Milbrandt and Yau (2006a), J. Atmos. Sci.

13. dBZ 1-km MC2 Simulation* using detailed microphysics 40 km 16 km 40 km 16 km Maximum: 60–65 dBZ COMPOSITE Maximum: 63.6 dBZ 750 hPa RADAR: 0030 UTC [6:30 pm] SIMULATION: 4:30 h [6:30 pm] N N 65 60 57 54 51 48 45 42 39 36 33 30 * Source: Milbrandt and Yau (2006a), J. Atmos. Sci.

14. Reflectivity CAPPI (2 km) 10 km Equivalent Reflectivity (750 hPa) 10 km RADAR: 0030 UTC [6:30 pm] 1-km SIMULATION: 4:15 h [6:15 pm] N N Source: Milbrandt and Yau (2006a), J. Atmos. Sci. 1-km MC2 Simulation* using detailed microphysics

15. Local time: 6:30 pm (Simulation time: 4:30 h) SENSITIVITY EXPERIMENTS: Variations in microphysics scheme TRIPLE-MOMENT SINGLE-MOMENT-A DOUBLE-MOMENT Diagnosed  DOUBLE-MOMENT Fixed  700 hPa: Z eh [dBZ] Source: Milbrandt and Yau (2006b), J. Atmos. Sci.

16. 10 9 9 6-h ACCUMLATED SOLID PRECIPITATION (Hail) [mm] 13 25 35 34 14 23 TRIPLE-MOMENT DOUBLE-MOMENT Diagnosed  SINGLE-MOMENT DOUBLE-MOMENT Fixed  Source: Milbrandt and Yau (2006b), J. Atmos. Sci. SENSITIVITY EXPERIMENTS: Variations in microphysics scheme

17. SENSITIVITY OF MICROPHYSICS SCHEME 4-category SINGLE-moment vs. 6-cateogory DOUBLE-moment: MC2_2.5 km: Milbrandt-Yau scheme (nested from GEM15 output) MC2_2.5 km: Kong-Yau scheme (nested from GEM15 output) 429 mm h -1 112 mm h -1 [g kg -1 ] RAIN “GRAUPEL” GRAUPEL HAIL RAIN

18. SUB-QUESTIONS for Science Question 3

19. SUB-QUESTION A: What defines a “success” for a high-resolution simulation in terms providing useful numerical guidance from the current GEM-LAM-2.5 configuration? The simulated storm structure may be realistic, but its track displaced; The location of CI may be incorrect, but the type of storm that is forecast may be similar to the observed  Success or a bust? Science Question 3

20. SUB-QUESTION B: How can the model’s ability to accurately simulate the general nature of the observed convection be quantified? Once a “success” is defined, how can the general skill of the model be measured? Important for evaluating the effects of changes to the model configuration Science Question 3

21. SUB-QUESTION C: Can the atmospheric state be classified a priori as “predictable” or “non-predictable” in terms of recommended use of the GEM-LAM-2.5 run to guide the forecast? Science Question 3

22. SUB-QUESTION D: How realistic are the simulated storm structures and microphysical fields? Comparison to radar observations (related to definition of “success”?) Science Question 3

23. SUB-QUESTION E: How realistic is the evolution of the boundary layer and surface processes in the foothill regions for the high-resolution model simulations? Comparison to IOP observations Sensitivity tests for hind-cast simulations Science Question 3

24. SUB-QUESTION F: Can deficiencies in the current physical parameterizations be identified? Comparison to IOP observations Sensitivity tests for hind-cast simulations Science Question 3

25. SUB-QUESTION G: What would be the effect of performing a subsequent nest to a higher-resolution (e.g. 1-km) grid, driven from the 2.5-km run? dx = 2.5 km is INSUFFICIENT to fully resolve individual storms Would the value added by a 1-km grid be worth the cost? Science Question 3

26. SUB-QUESTION H: Can an ensemble of high-resolution runs improve the prediction of convective initiation? Science Question 3 Modified ICs (different members from different 15-km driving runs) Modified physical parameterizations/settings for 2-5 km runs

27. SUB-QUESTION I: Can a high-resolution analysis, created using the additional observations, improve the numerical prediction of convective initiation and subsequent storm development? Science Question 3

28. 1 st UNSTABLE Science Workshop 18-19 April 2007 CONCLUSION 1.Emphasis of NWP component of UNSTABLE is on maximizing and improving the utility of high-resolution NWP output as a forecast tool 2.Real-time 2.5-km runs will be done (and archived) by CMC for summers of 2007 and 2008 (and hopefully beyond…) 3.Investigation of several sub-questions can begin any time