OBJECTIVE VERIFICATION OF A RADAR-BASED OPERATIONAL TOOL FOR IDENTIFICATION OF HAILSTORMS

1. OBJECTIVE VERIFICATION OF A RADAR-BASED OPERATIONAL TOOL FOR IDENTIFICATION OF HAILSTORMS I. San Ambrosio,F. Elizaga and F. MartínAnalysis and Forecasting Techniques Department (STAP)INM, SPAIN e-mail: ismael@inm.es

2. CONTENTS • Introduction. • Hail data collection. • Calibration of the hail module. • Verification methodology. • Parameters performance. • Verification of the operational thresholds. • Conclusions.

3. INTRODUCTION (I) • INM radar network: • 14 C band radars. • Normal (N) and Doppler (D) modes. • Normal mode: cartesian volume with 12 CAPPIS; Resolution: 2x2 km and 10 min. • Doppler mode: 6 CAPPIS; Resolution: 1x1 km and 10 min. Radar sites Radar coverage in normal mode

4. INTRODUCTION (II) • Operational aim of the Hail Module: support indecision making processes for nowcasting activities. • Based upon: • Operational application for convective monitoring. • HIRLAM (INM) model data. • Hail algorithms used by this module: • Density of VIL (DVIL). • Technique of Waldvogel, to obtain POH (Probability Of Hail of any size). • Hail Detection Algorithm (HDA), to obtain POSH (Probability Of Severe Hail). Severe hail is larger than 19 mm in diameter.

5. HAIL DATA COLLECTION (I) • Information about hail storms during the years 2001, 2002 and 2003: • Place, time of the hailstorms. • Hail size: split into two categories, severe or non-severe. • Sources of this information: • Different media: press, radio, TV, internet. • Atmospheric Research group (University of León). • Events summaries of ADV (“Agrupació de Defensa Vegetal de les Terres de Ponent”). Network with 170 hailpads in the SW of Lleida (“Zona d’Actuació”).

6. HAIL DATA COLLECTION (II):

7. HAIL DATA COLLECTION (III):thunderstorms without hail • It is very difficult to obtain reliable information about these events. • Summaries elaborated by ADV have been used for this purpose. • Thunderstorms without hail: • take place in the region of interest of ADV. • there are not any hail information about them.

8. CALIBRATION OF THE HAIL MODULE (I) • For this process a 144 thunderstorms data base from the 2001 and 2002 campaignswas used: • 29 with severe hail. • 52 with non-severe hail. • and 63 without hail. • Taking into account contingency tables and different skill measures a calibration of this module was made to identify two categories of hailstorms. • In the calibration process a set of thresholds were tuned.

9. CALIBRATION OF THE HAIL MODULE (II) • These are the thresholds finally obtained to identify the two hail categories: • Probability of severe hail: DVIL  1,5 and POSH  10. • Probability of hail (any size): DVIL  1,3 or POH  20.

10. METHODOLOGY OF VERIFICATION (I):analysis of the hailstorms • A completelife cycle of every 3D cell is obtained. • Use of the three different algorithms to analyze the selected parameters (POH, DVIL and POSH) and their evolution along every life cycle. • During the calibration process, a Criterion for Temporal selection, CT50, was established to extract information related with: • 40 minutes before every hail event. • During the hail event itself. • And 10 minutes after it.

11. CT50 Data

12. METHODOLOGY OF VERIFICATION (II): • The same criterion, CT50, was also used in this verification process. • Other criteria, CT30 (data during the hail event and 30 minutes before it) and CT00 (data during the hail event), have been also established to analyze the performance of the hail module.

13. CT30 Data

14. CT00 Data

15. METHODOLOGY OF VERIFICATION (III): • The different criteria (CT50, CT30 and CT00) have been applied to the 2003 global set of data (from March to September). • The same analysis has been made for Spring thunderstorms, those that have taken place between March and June, both included, and also for Summer ones (July, August and September).

16. METHODOLOGY OF VERIFICATION (IV): • For the global period 246 thunderstorms have been analyzed (39 with severe hail, 120 with non-severe and 87 without hail). • Spring period: 78 thunderstorms (10 with severe hail, 52 with non-severe and 16 without hail). • Summer period: 168 thunderstorms (29 with severe hail, 68 with non-severe and 71 without hail). • In these three periods and with the three temporal criteria, we have evaluated: • Performance of the parameters (DVIL, POH and POSH). • Verification of the operational thresholds by means contingency tables and verification indexes.

17. PARAMETERS PERFORMANCE • The most remarkable difference appears when DVIL values are analyzed in Spring and Summer thunderstorms, ones apart from the others. There are small differences between the three criteria considered for temporal selection.

19. PARAMETERS PERFORMANCEEvents with Severe Hail • Another important difference appears when the number of events is represented depending on DVIL and POSH. In Spring thunderstorms the cuasi-linear relationship disappears. The shorter is the criteria for temporal selection (CT00), the less clear is the maximum located in lower values of DVIL and POSH (as can be seen in CT50).

21. PARAMETERS PERFORMANCEEvents with Non Severe Hail • In these hailstorms, a clear difference can be seen when the number of events is represented depending on DVIL and POH, and specially when CT00 is used. In Spring hailstorms a maximum number of events is located on DVIL values lower than 1.0 g/cm3. For Summer cases, almost all the hailstorms have DVIL values larger than 1.0 g/cm3.

23. PARAMETERS PERFORMANCEThunderstorms with No Hail • A slight difference can be seen when the number of events is represented depending on DVIL and POH. In Summer thunderstorms the maximum is very well located on DVIL values lower than 0.5 g/cm3 and POH smaller than 10%. For Spring cases, this maximum spreads up to 1.5 g/cm3 of DVIL.

25. VERIFICATION OF THE OPERATIONAL THRESHOLDS (I) • For the global period of verification (Spring and Summer 2003), and also for the split seasons, and with the three criteriaCT50, CT30 and CT00, the next skill measures have been worked out from contingency tables: • POD, Probability Of Detection. • FAR, False Alarms Rate. • CSI, Critical Success Index. • HSS, Heidke Skill Score. • OR, Odds Ratio.

26. VERIFICATION OF THE OPERATIONAL THRESHOLDS (II) • Next, different graphics comparing the values obtained for the skill measures can be seen. These graphics are bar charts to compare the global, Summer and Spring value of every verification index, and for every temporal criterion, with the 2001-2002 calibration process value.

27. SEVERE HAIL THRESHOLD PERFORMANCE OF THE PROBABILITY OF DETECTION (POD). HAIL OF ANY SIZE THRESHOLD

28. SEVERE HAIL THRESHOLD PERFORMANCE OF THE FALSE ALARMS RATE (FAR). HAIL OF ANY SIZE THRESHOLD

29. SEVERE HAIL THRESHOLD PERFORMANCE OF THE CRITICAL SUCCESS INDEX (CSI). HAIL OF ANY SIZE THRESHOLD

30. SEVERE HAIL THRESHOLD PERFORMANCE OF THE HEIDKE SKILL SCORE (HSS). HAIL OF ANY SIZE THRESHOLD

31. SEVERE HAIL THRESHOLD PERFORMANCE OF THE ODDS RATIO (OR). HAIL OF ANY SIZE THRESHOLD

32. CONCLUSIONS (I) • This HAIL MODULE is operational since 2003, and has shown its utility for convective events nowcasting. • Combining DVIL and HDA seems accurate to detect severe hail, in the some way DVIL and technique of Waldvogel does for hail of any size. • The global verification carried out with 2003 data base has given quite good results. • The shorter are the criteria for temporal selection, the better are the results. • This different performance can appear due to lots of data contained in the CT50.

33. CONCLUSIONS (II) • The skill of the module is better in Summer hailstorms than in Spring ones. • The performance of the considered parameters (DVIL, POH, POSH) is related with summer deep convection processes. • It seems necessary to analyzeseparately Spring and Summer hailstorms in more detail, to tune more accurate procedures and thresholds for different seasons (or months) of the year.