2. Outline Background Highlights of NCAR’s R&D efforts A proposed 5-year plan for CWB Final remarks

3. Convective-scale DA: a brief look back 1990: Lilly’s motivational publication - NWP of thunderstorm – has its time come? 1990 -1996: studies on single Doppler retrievals 1997 - early 2000’s: studies on assimilation of radar observations into NWP models Late 2000’s - now: preliminary use of radar data in operational NWP systems

4. Radar data assimilation methods Warm-start algorithms using information derived from reflectivity - Cloud analysis - Latent heat nudging - DDFI – diabatic digital filter initialization - Assimilate estimated in-cloud humidity or vertical velocity Variational data assimilation Assimilating both radial velocity and reflectivity - 3D-Var (operational or semi-operational) - 4D-Var (semi-operational) Ensemble Kalman Filter (EnKF) Assimilating both radial velocity and reflectivity - Relatively new to radar data assimilation - Mostly used for research

5. High-resolution (<4km) RU (<3H) systems (as of 2012) SystemOrganizationRU cycleDA methodModel HRRRNOAA/GSDhourlyDDFIARW-WRF BJRUCNCAR/BMB3-hourly3DVarARW-WRF RTFDDANCARhourlyNewtonian nudging ARW-WRF LAPSNOAA/GSDhourlySuccessive correction ARW,MM5,or RAMS ADASCAPSN/A3DVarARPS AROME- France Meteo France3-hourly3DVarAROME COSMO-DEDWD3-hourlyNewtonian nudging COSMO UKVMet Office3-hourly3DVarUnified Model

6. What are the critical observations? Data impact study by Meteo France (Brousseau et al. 2014) Below 200km wave length, the reduction of error is only provided by radar, ground-based GPS, and aircraft Below 200km wave length, the reduction of error is only 10% - where most improvement is needed for heavy precipitation forecast

7. NCAR’s efforts on convective-scale DA: Overview 1992 – 2002: pioneer research and development work on radar DA using the 4DVar technique and a cloud- scale model - Sun et al. (1992), Sun and Crook (1994, 1997, 1998) 2003 – 2014: development of radar DA for WRF 3DVar - Xiao et al. ( 2005, 2007, 2008), Xiao and Sun (2008), Wang et al. (2012) 2010 – 2014: development of radar DA for WRF 4DVar - Wang et al. (2013), Sun and Wang (2013)

8. NCAR’s efforts on convective-scale DA : Current capability WRFDA 3DVar - 1h update cycles assimilating radial velocity and reflectivity in addition to all conventional observations - Indirect assimilation of reflectivity including an scheme for humidity adjustment - Options for choice of momentum control variables - Operational capability WRFDA 4DVar - One of a few for convective-scale - Use WRF tangent linear model as constraint - Can be run with multi-incremental option to save computation cost - Adjoint of Kessler warm rain microphysics and a simple cumulus scheme - Active research is being conducted to improve efficiency

9. Average score of 4 convective cases during summer 2009 5 mm hourly precipitation An example with WRF 3D-Var 23 July 2009 t = 3ht = 4h OBS No Radar Radar NCAR/BMB WRF 3DVar Pre-operational testing Assimilate both radial wind and reflectivity OPER (no radar) With radar

10. Recent improvement of WRF 3DVar OBS Old New t=1h t=6h New Old

11. Hourly Precipitation forecasts Obs WRF 3DVarWRF 4DVar 4DVar 3DVar w/o DI 3DVar with DI 6h fcst of 1h accum. rainfall 5mm Radius of Influence = 24km Comparison of WRF 3DVar with 4DVar  Diabatic initialization significantly improves QPF  Compared with 4DVar, 3DVar produces a slower squall line and the precipitation amount is ove-predicted with DI FSS

12. Hourly Precipitation forecasts 4DVar radar DA for Typhoon Fanapi Hourly precipitation at t = 6h OBS GTS3DVarRDR4DVar Surface analysis 4DVar successfully predicts the two low pressure centers with similar magnitudes as in the surface analysis

13. Hourly Precipitation forecasts If NCAR collaborate with CWB for the next fiscal 5 years, what will be a feasible operational goal?  An fully operational hybrid-3DVar convective-scale data assimilation system with hourly update cycles assimilating operational radar reflectivity, radial velocity, surface precipitation, and ground-based GPS  A semi-operational multi-incremental 4DVar system assimilating the same types of observations as 3DVar

14. Hourly Precipitation forecasts Year 1 Pre-operational testing of a radar QC system capable of detecting typhoon centers and performing accurate velocity dealiasing Pre-operational testing of a radar QC system capable of detecting typhoon centers and performing accurate velocity dealiasing Operation R & D Add a typhoon QC scheme to existing radar QC package 3DVar: combine DA of radar, surface, and ground GPS 4DVar: test the performance of physics Add a typhoon QC scheme to existing radar QC package 3DVar: combine DA of radar, surface, and ground GPS 4DVar: test the performance of physics

15. Hourly Precipitation forecasts Year 2 Pre-operational testing of 3DVar with 3-hourly cycles and radar data assimilation Pre-operational testing of 3DVar with 3-hourly cycles and radar data assimilation Operation R & D 3DVar: multiple case testing for operational configuration 4DVar: design optimal multiple-incremental scheme for convective-scale DA over Taiwan 3DVar: multiple case testing for operational configuration 4DVar: design optimal multiple-incremental scheme for convective-scale DA over Taiwan

16. Hourly Precipitation forecasts Year 3 Operational implementation of 3DVar with 3-hourly cycles and radar data assimilation Operational implementation of 3DVar with 3-hourly cycles and radar data assimilation Operation R & D 3DVar: testing of convective- scale hybrid-3DVar with hourly update cycles 4DVar: improve efficiency by coding improvement and model simplification 3DVar: testing of convective- scale hybrid-3DVar with hourly update cycles 4DVar: improve efficiency by coding improvement and model simplification

17. Hourly Precipitation forecasts Year 4 Pre-operational testing of hybrid 3DVar with hourly cycles and radar data assimilation Pre-operational testing of hybrid 3DVar with hourly cycles and radar data assimilation Operation R & D 3DVar: Pre-operation evaluation of the hybrid-3DVar 4DVar: comparison with the hybrid-3DVar over multiple cases 3DVar: Pre-operation evaluation of the hybrid-3DVar 4DVar: comparison with the hybrid-3DVar over multiple cases

18. Hourly Precipitation forecasts Year 5 Operational implementation of hybrid 3DVar with hourly cycles and radar data assimilation Semi-operational testing of 4DVar and comparison with the operational 3DVar Operational implementation of hybrid 3DVar with hourly cycles and radar data assimilation Semi-operational testing of 4DVar and comparison with the operational 3DVar Operation R & D 3DVar: evaluation of the performance of the operational hybrid-3DVar 4DVar: testing of the hybrid-4DVar 3DVar: evaluation of the performance of the operational hybrid-3DVar 4DVar: testing of the hybrid-4DVar

19. Hourly Precipitation forecasts Final Remarks  Improving convective-scale DA and forecast has a high society and economy impact and hence a priority for weather service centers worldwide  It is challenging and therefore a long term effort  Applying NCAR’s R&D experience in the convective-scale DA to CWB’s forecast operation will not only benefit CWB but also the overall advancement of high impact weather prediction  One of the most important aspects of the collaboration is to train the young meteorologists who will carry on the weather research and forecast to next generation