1. ounding nalog etrieval ystem Ryan Jewell Storm Prediction Center Norman, OK SARS Sounding Analog Retrieval System

2. What is SARS? SARS is a sounding matching algorithm. Enables the forecaster to quickly compare soundings to historical proximity soundings associated with a particular type of severe weather. Uses key sounding parameters to find matching soundings within the historical database. Matching database includes ~1900 proximity soundings for tornadic and non- tornadic supercells as well as severe hail. (CONUS - All Seasons). SARS returns date, location and associated severe weather for each match. Example: (99050400.OUN SIGTOR) or (95050500.FWD 3.00” HAIL). SARS has also produce “probabilistic” forecasts for SIG Hail and Tornadoes, based upon the distribution of matches. More later.

3. Lets pretend it’s a forecast sounding 1) Load sounding, run SARS 3) Display SARS results. SARS Supercell / Tornado Matches  SARS Hail Matches  (Actual significant tornado proximity sounding from SARS Database) Probabilistic SARS Forecast of Tornadoes and Significant Hail If this was a forecast sounding, SARS would indicate a high likelihood of Tornadic Supercells, possibly F2+, and hail > 2.00” diameter. 2)SARS finds sounding matches based on various parameters. Hail: MUCAPE, MUMR, 0-6 SHR, 700-500mb LR, 500 temp

4. What can be said about the distribution of matches? Example: Forecast sounding yields 20 hail matches from sounding database. 10 are 0.75”, 5 are 1.00”, and 5 are 2.75”. What size of hail is most likely? Probabilistic SARS Forecast

5. Need to CALIBRATE SARS SARS has been calibrated to forecast the following… SIG vs. Non-SIG Hail (2.00” is threshold) SARS forecasts Significant hail if majority of matches are ≥2.00”. Tornadic vs. Non-Tornadic Supercells SARS forecasts Tornadic Supercells if majority of supercell matches are tornadic. Need to answer the following questions: What parameters should be used for matching? How large should the search ranges be for each parameter?

6. SARS Calibration Process Choose Matching Parameters – Relevant parameters associated with severe storms (various measures of instability and shear associated with hail, supercells, tornadoes). No “kitchen sink” method (yet). Define initial ranges for each parameter to be used in search. (Example +/- 500 CAPE, +/- 50 0-1km SRH, +/- 500 m LCL height). Test. Run soundings through the database, analyze matches. Adjust parameters and ranges until the desired result is received. Desired Result = The majority of matches agree on a particular type and magnitude of severe weather, and it verifies. If a sounding is associated with 3.00” hail, MOST of the SARS matches should be very large hail (rather than very small). If a sounding is associated with a tornadic supercell, MOST of the SARS matches should be tornadic supercells (rather than non-tornadic supercells).

7. Example – Calibration for hail forecast Test each sounding independently against the others. Determine if SARS produced a good forecast or not. Test various combinations of parameters and parameter ranges…see which one is the best. Start with list of 900+ historical proximity severe hail soundings.

8. SARS Matching Parameters (For THIS sounding database only ! ) Significant Hail Matching MUCAPE Mixing Ratio of MU Parcel 700-500 mb Lapse Rate 500 mb Temperature 0-6 km Bulk Shear Supercell / Tornado Matching 100 mb MLCAPE MLLCL Height (agl) 700-500 mb Lapse Rate 0-1 km SRH 0-6 km Bulk Shear

9. SARS Parameters Ranges (For THIS database only!) Significant Hail Ranges MUCAPE +/- 25-50% Mixing Ratio of MU Parcel +/- 2.0 g/kg 700-500 mb Lapse Rate +/- 1.5 C/km 500 mb Temperature +/- 4.5 C 0-6 km Bulk Shear +/- 6 m/s Supercell / Tornado Ranges 100 mb MLCAPE +/- 1500 J/kg MLLCL Height (agl) +/- 400 m 700-500 mb Lapse Rate +/- 1.0 C/km 0-1 km SRH +/- 30% 0-6 km Bulk Shear +/- 7 m/s

10. SARS Skill Scores Significant Hail Forecast HitMiss False Alarm Correct Null No Matches Found 457468733113 CSITSSPODFAR 0.7750.7000.9090.160 NOTE: Highest skill score AND highest % with matches

11. SARS Skill Scores Significant Hail Forecast - Filtered (≤ 1.5” vs. ≥ 2.5”) Removes “grey” area near 2” threshold CSITSSPODFAR 0.8590.7860.9560.098 HitMiss False Alarm Correct Null No Matches Found 43323482313 NOTE: Highest skill score AND highest % with matches

12. SARS Skill Scores Tornadic vs. Non-Tornadic Supercell Forecast CSITSSPODFAR 0.5880.3870.7850.309 HitMiss False Alarm Correct Null No Matches Found 3909917425010 NOTE: Highest skill score AND highest % with matches

13. SARS MATCHING EXAMPLES

14. F5 Tornado

15. F4 Tornado

17. F5 Tornado

18. GRIDDED SARS EXAMPLE

19. RUC Forecast Lines – Number of SARS matches for each grid point sounding. Color Fill - % of matches that had >2.00” hail.

20. SARS Strengths & Weaknesses SARS cannot predict whether storms will form (capping, forcing issues). Usefulness of SARS depends upon the accuracy of model forecasts. SARS can heighten awareness in severe weather situations which are not clearly evident. Can be adjusted to match different types of severe weather.

22. SPC NSHARP Sounding Display on Web Available only for observed soundings.

23. Ryan.Jewell@noaa.gov

24. 1.75” Hail

27. 5/19/60, Topeka KS, Long Track F4Date, Location, Event…Anyone?

28. Example – Calibration for hail forecast Remove 1 sounding for testing. This essentially leaves the dataset unchanged (removing 1 out of 900+ soundings). Run the sounding through the entire list, choose initial parameters (CAPE, shear, etc) and initial matching ranges for each parameter. Note the number and types of matches (hail < 2”, and hail ≥ 2”). Determine if Hit, Miss, False Alarm, Correct Null. Hit = Sounding associated with hail ≥ 2.00” and ≥50% matches ≥ 2.00” Miss = Sounding associated with hail ≥2.00” but < 50% matches ≥ 2.00” FA = Sounding associated with hail < 2.00” but ≥ 50% matches ≥ 2.00” CN = Sounding associated with hail < 2.00” and ≥ 50% matches < 2.00” Repeat for each sounding in the list. Calculate final skill scores for this calibration run (call it run 1). Start with list of 900+ historical proximity severe hail soundings. Example – Calibration for hail forecast Repeat the entire process using different combinations of parameters AND parameter ranges. Test thousands of combinations. Find optimal parameters and search range for those parameters which result in the highest skill score (CSI, TSS). Use this parameter set and calculated ranges for SARS guidance UNTIL…. More soundings are added, in which case the calibration process can be run again. SARS should get “smarter” as more cases are added to the database.