2. A physical initialization algorithm for non-hydrostatic NWP models using radar derived rain rates Günther Haase Meteorological Institute, University of Bonn German Weather Service

3. Günther HaaseHelsinki, 3 October 20022 area: Northern Germany ~ 400x400 km 2  x = 7 km time period: May, 2000 (  t = 1 h) radar  LM

4. Günther HaaseHelsinki, 3 October 20023 LM Configuration non-hydrostatic model (V 1.27) Arakawa-C-grid:  x = 2.8 km vertical hybrid-coordinate: 35 levels time step:  t = 30 s prognostic variables: u, v, w, T, p, q v, q c,... diagnostic variables: rain- and snowflux,... no convection parameterization initial and boundary fields from LM (  x = 7 km, one-way nesting)

5. Günther HaaseHelsinki, 3 October 20024 PIB Algorithm preprocessing of reflectivity measurementsreflectivity measurements determine pre-forecast period, conversion efficiency and mean cloud top height compute present LM-CCL (cloud base)cloud base precipitation analysissaturation adjustment modify LM variables w, qv, and qc

6. Günther HaaseHelsinki, 3 October 20025 DWD radar network 16 C-band radars: = 5.3 cm cartesian grid:  x = 4 km temporal resolution:  t = 15 minutes 6 reflectivity classes fixed Z/R relation

7. Günther HaaseHelsinki, 3 October 20026 Cloud top height Derivation of cloud top heights from averaged LM cloud water profiles Alternative: application of a cloud initialization method using Meteosat measurements (F. Ament)

8. Günther HaaseHelsinki, 3 October 20027 Cloud base height CCL LCL open symbols: LM without convection parameterization closed symbols: LM with convection parameterization

9. Günther HaaseHelsinki, 3 October 20028 Modifications RR RAD > 0.1 mm/h RR RAD  0.1 mm/h

10. Günther HaaseHelsinki, 3 October 20029 Vertical wind (PIB) assumptions: only two hyd. components: water vapor and rain only two cloud processes: condensation and evapo- ration closure: conversion efficiency of saturated air into rain Computation of vertical wind profiles using a 1-dim cloud model C = 0.1

11. Günther HaaseHelsinki, 3 October 200210 Case study: 13 July 1999 Exp.pre-forecast periodforecast CTL –12 h PIB12 – 13 UTC12 h LHN12 – 13 UTC12 h

12. Günther HaaseHelsinki, 3 October 200211 PIB sensitivity study

13. Günther HaaseHelsinki, 3 October 200212 LHN sensitivity study

14. Günther HaaseHelsinki, 3 October 200213 CTLLHNPIBradar 13 UTC 14 UTC 15 UTC

15. Günther HaaseHelsinki, 3 October 200214 CTL with convection parameterization

16. Günther HaaseHelsinki, 3 October 200215 Hydrology a)Area averaged hourly accumulated precipita- tion b)LWP and IWV: PIB generates more clouds than the control run (CTL)

17. Günther HaaseHelsinki, 3 October 200216 a)Hit Rate b)False Alarm Rate c)Kuipers Score (measures the skill of a forecast relative to a random forecast) Objective skill scores

18. Günther HaaseHelsinki, 3 October 200217 Scale-dependency of the RMSE PIB provides better forecasts than the control run (CTL) on all scales LHN is better on large scales

19. Günther HaaseHelsinki, 3 October 200218 a)vertical wind (500 hPa) b)surface pressure c)absolute surface pres- sure tendency Noise

20. Günther HaaseHelsinki, 3 October 200219 Summary (1) PIB is suitable for nowcasting of convective precipitation events on the meso-  -scale reduction of spinup and position errors in the precipitation forecast over a couple of hours closure of the information gap between LM forecasts and nowcasting based on observations PIB uses only operational radar products as input

21. Günther HaaseHelsinki, 3 October 200220 low computational costs compatible with future model developments  column approach prevents the method from initializing large-scale precipitation events  PIB reacts very sensitive on variations of the input data (quality control!) Summary (2)

22. Günther HaaseHelsinki, 3 October 200221 combination with a cloud initialization method using Meteosat measurements using 3-dim reflectivity fields (on model levels) forcing a dynamic balance between mass and wind fields application of a modified LM precipitation scheme Future research