1. Forecasting Storm Duration Neil I. Fox David Jankowski, Elizabeth Hatter and Liz Heiberg Dept. Soil, Environmental and Atmospheric Science University of Missouri - Columbia

2. Considering rear edge propagation velocity in flash flood forecasting  Why worry about your rear  How your rear moves compared to your middle  Using knowledge of your rear to forecast rainfall totals  Stop the rear jokes

3. Why worry about your rear?  Current nowcasting tools (e.g. SCIT tracks) concentrate on arrival time  Excellent for Severe Weather warning  Flash flood forecasting: Interested in total duration of precipitation  Event management / Emergency services like to know end time

4. This study looked at  The use of three measures of storm velocity as indicators of flash flood potential 1/v c1/v c 1/v r1/v r (v c -v r )/v c v r(v c -v r )/v c v r  The last of these is defined as the ‘Storm Duration Factor’

5. Storm duration factor (SDF) Duration (D) over a point at distance x : Rainfall accumulation (R a ) at x assuming steady-state rainfall rate R:

6. Data  Initially data was taken from a number of cases where (flash) flooding occurred  A range of storm types, locations and situations  Not all storm cells observed caused flooding  Then more data for more cases

7. Analysis  Centroid velocities found using the NSSL algorithms (SCIT)  Rear edge velocities found by locating position from tracing centroid vector backward until Z falls below threshold

8. Analysis  The three measures were plotted against rainfall accumulations for the subsequent 60 minutesfor the subsequent 60 minutes 0 km and 25 km ahead of storm center location0 km and 25 km ahead of storm center location Greater distances saw very little rain (storms don’t move that fast or dissipate within the distance)Greater distances saw very little rain (storms don’t move that fast or dissipate within the distance)

9. Comparison of v c & v r

10. 1/v c & precip accumulation

11. 1/v r & precip accumulation

12. SDF & precip accumulation

13. Results  All correlation coefficients are horrible  If you squint you can kind of see what you want to see  More work required!!

14. Next  This could be because We don’t consider development/dissipationWe don’t consider development/dissipation We don’t consider size of stormWe don’t consider size of storm We don’t look at sensible distances or have good rainfall dataWe don’t look at sensible distances or have good rainfall data

15. Accounting for storm size Tried a “pure” measure Unsuccessful – so try measure based on velocity and storm size

16. Test  Rainfall accumulation versus duration based on centroid velocity (x+Δx/v c )centroid velocity (x+Δx/v c ) rear edge velocity (x+Δx/v r )rear edge velocity (x+Δx/v r ) Both (reduces to the others for x = 0)Both (reduces to the others for x = 0)

17. Rainfall total vs v c /Δx

18. Rainfall total vs v r /Δx

19. Problems  Radar rainfall accumulations use gaugeuse gauge  Rear edge velocity determination automate – make robustautomate – make robust  Mixture of storm types stratifystratify

20. Thanks  Parts of this work have been funded by the COMET Partners Program and The University of Missouri Research Council