Hailstorm at 19 August over Berlin and Issues

Hailstorm at 19 August over Berlin and Issues of Verification of Deep Convection Matthias Jaeneke, DWD 1

Outline Introduction ê Why Case-Studies in Verification ? The Berlin Hailstorm and Forecast-Verification ê Local Data, Synoptics, Short-Range Development ê Time-Series of Model Weather Verification What does the Berlin Hailstorm Case teach us for Deep Convection Case Verification ? 2

Introduction • Why Case-Studies in Verification ? ê Case-studies provide insight into strengths and weaknesses of forecasts of the single weather situation ê Case-studies are able to stratify verification into typical classes of weather cases, for instance high impact situations ê Case-studies are more adapted to problems and needs of the operational forecaster and are able to show up the real range of day to day variability of forecast quality ê Case-studies are therefore integral part of all verification and represent all qualitative aspects of forecasts 3

Direction of Wind Maximum Gusts Visibility MSLP Temperature 2 m Temperature 5 cm Precipitation (1 min. ) 4

Total Precipitation Amount (mm) during the Hailstorm 5

B Geopot. 500 T 850 6

B Geopot. 500 KO-INDEX : Generalized vertical gradient of Theta. E 7

B Geopot. 500 IR-SAT 8

9

Surface Map 19 -08 -2000 12 UTC Meteor. Institut Free University Berlin 10

B NOAA-IR-DIAG T 2 m 11

B NOAA-IR-DIAG T 2 m 12

B NOAA-IR-DIAG T 2 m 13

B NOAA-IR-DIAG T 2 m 14

B Radar-Reflectivity Warning-Markers 15

B Radar-Reflectivity Warning-Markers 16

B Radar-Reflectivity Warning-Markers 17

B Radar-Reflectivity Warning-Markers 18

B Radar-Reflectivity Warning-Markers 19

B Radar-Reflectivity Warning-Markers 20

B Radar-Reflectivity Warning-Markers 21

B Radar-Reflectivity Warning-Markers 22

B Radar-Reflectivity Warning-Markers 23

B Radar-Reflectivity Warning-Markers 24

B Radar-Reflectivity Warning-Markers 25

B Radar-Reflectivity Warning-Markers 26

B Radar-Reflectivity Warning-Markers 27

B Radar-Reflectivity Warning-Markers 28

B Radar-Reflectivity Warning-Markers 29

B Radar-Reflectivity Warning-Markers 30

B LM: Grid-Weather 1 h Lightnings 31

B LM: Grid-Weather 1 h Lightnings 32

B LM: Grid-Weather 1 h Lightnings 33

B LM: Grid-Weather 1 h Lightnings 34

B LM: Grid-Weather 1 h Lightnings 35

B LM: Grid-Weather 1 h Lightnings 36

B LM: Grid-Weather 1 h Lightnings 37

B LM: Grid-Weather 1 h Lightnings 38

B LM: Grid-Weather 1 h Lightnings 39

B LM: Grid-Weather 1 h Lightnings 40

B LM: Grid-Weather 1 h Lightnings 41

B LM: Grid-Weather 1 h Lightnings 42

B LM: Grid-Weather 1 h Lightnings 43

B LM: Grid-Weather 1 h Lightnings 44

B LM: Grid-Weather 1 h Lightnings 45

B LM: Grid-Weather-Observations 46

B LM: Grid-Weather-Observations 47

B LM: Grid-Weather-Observations 48

B LM: Grid-Weather-Observations 49

B LM: Grid-Weather-Observations 50

B LM: Grid-Weather-Observations 51

B LM: Grid-Weather-Observations 52

B LM: Grid-Weather-Observations 53

B LM: Grid-Weather-Observations 54

B LM: Grid-Weather-Observations 55

What does the Berlin Hailstorm Case teach us for Deep Convection Case Verification ? • There are three questions to be discussed in connection with the Berlin Hailstorm case : ê What is concerning verification the characteristics of deep convection and its NWP forecasts ? ê Are standard measures of contingency table statistics applicable in single case verification ? ê What was the quality of the NWP forecast for the Berlin Hailstorm Case and what do forecasters want to have ? 56

What does the Berlin Hailstorm Case teach us for Deep Convection Case Verification ? • What is concerning verification the characteristics of deep convection and its NWP forecast ? ê Deep convection bears, though in nature deterministic, in its realization some stochastic characteristics ê Deep convection forecasts are, like in the DWD-LM, expressed mostly as categorical grid-point „thunderstorm-activities“. Therefore they are also more probability than categorical type. ê Also the Berlin Hailstorm shows : NWP forecasts provide „carpets“ of convection, in contrast to that real thunderstorms are structured in clusters and cells. 57

What does the Berlin Hailstorm Case teach us for Deep Convection Case Verification ? • Are standard measures of contingency table statistics applicable in single case verification ? ê In situations of deep convection three different cases may occur: (1) There are „yes“-forecasts and „yes“-observations (2) There are „yes“-forecasts but no „yes“-observations (3) There are „yes“-observations but no „yes“ forecasts ê Case (1) may show overlapping of forecast and observations areas with no problems for POD, FAR, BIAS and CSI. If no overlapping occurs (no hits) evaluation results in POD = 0, FAR = 1, CSI = 0 and BIAS is determined by the ratio of both „yes“-areas. 58

What does the Berlin Hailstorm Case teach us for Deep Convection Case Verification ? • Are standard measures of contingency table statistics applicable in single case verification ? ê For Case (2) with no „yes“-observation and therefore also no hits, POD and BIAS are not defined, FAR = 1 and CSI = 0 ê In case (3) with no „yes“ forecast and again no hits, POD =0, BIAS = 0 , CSI = 0 and FAR is not defined ê In cases 1 a, 2 and 3 the contingency table measures reveal bad forecast accuracy, partly quality assessment even is not possible. ê Though this was partly true also in the Berlin hailstorm the forecaster recognized spatial and temporal „near-by“ cases to say : „The forecasts were not too bad, I got signals for severe convection. “ 59

What does the Berlin Hailstorm Case teach us for Deep Convection Case Verification ? • What was altogether the NWP forecast quality in the Berlin hailstorm case? What do forecasters want to have ? ê The LM forecasts of significant weather in the Berlin Hailstorm case gave reasonably good synoptic scale indication of what could happen in Central Europe at that day. ê Despite systematic differences in daily run and mesoscale structure between forecast („carpet“) and observation (clusters and cells) the forecaster was able to extract an appropriate forecast signal. ê The most valuable forecast signal was the mesoscale indication of hailstorm-pixels (intensity), despite of some errors in position and time. The forecaster wants this signal. The NWP forecast was useful for him. 60