Prediction of the Development of an MCS over the Continental US - PowerPoint PPT Presentation

A Comparison of GSI-based EnKF and 3DVar for Multi-scale Data Assimilation and Prediction of the Development of an MCS over the Continental US Aaron Johnson and Xuguang Wang University of Oklahoma, Norman, OK, USA Acknowledgement: Jacob Carley (NCEP) Lou Wicker, Chris Karstens, Nai Kang, Jian Zhang (NSSL) 6 th WMO Data Assimilation Symposium College Park, MD 7-11 October 2013 1

Background  The GSI-based EnKF-Var hybrid DA system showed significant improvement compared to GSI 3DVar and became operational on May 22, 2012 for GFS.  It has also been extended to a 4DEnsVar hybrid and showed further improvements.  Efforts are being conducted to integrate, further develop and research GSI based EnKF-Var hybrid DA for operational regional forecast systems, e.g., GSI-based 3DVar/EnKF/Hybrid Hurricane- NAM Rapid Refresh GFS WRF (HWRF) (WRF-NMM) (WRF ARW) 2

GSI hybrid for HWRF: Retrospective and real time tests • Model: operational HWRF Sandy • Observations: radial velocity 2012 from Tail Doppler Radar (TDR) onboard NOAA P3 aircraft • Initial and LBC ensemble: GFS global hybrid DA system • Ensemble size: 40 Xu Lu Poster 3

Track and MSLP forecasts for 2012-2013 cases with NOAA P3 missions MSLP Track 4

Background and Motivation  System development/enhancement for radar DA over CONUS • This study aims to extend the GSI-based 3DVar/EnKF/Hybrid system for WRF ARW to add capability to directly assimilate ground based radar data (e.g., add hydrometeors control/state variables for reflectivity DA). • The development takes advantages of existing capabilities in GSI and therefore directly integrates with the operational DA system. 5

Background and Motivation • This study assimilates both synoptic/regional scale observations and storm-scale radar observations to compare the performance and understand the differences of GSI based 3DVar/EnKF/hybrid at multiple scales (e.g., GSI-EnKF vs. GSI-3DVar). • Prediction of a complex MCS with multiple storm modes and interactions, and a largely heterogeneous environment remains a challenge. • Successful prediction of such complex system with multi-scale interactions requires accurate estimate at multiple scales. • Comparison of Var with EnKF or EnsVar hybrid for such complex cases with convection permitting/resolving resolution is still limited. 6

20 May 2010 case overview • A broad, deep layer cyclone; well defined cold www.mmm.locust.ucar.edu front, warm front and dry line on surface. • Complex evolution of multiple interacting storms with upscale growth into an MCS. • Johnson et al. (2013) using wavelet analysis showed particularly strong sensitivity of precipitation forecast to both small scale and large scale perturbations. 7

Experiment Design • WRF ARW • Outer Domain – Assimilate operational conventional surface and mesonet observations, RAOB, wind profiler, ACARS, and satellite derived winds to define synoptic/mesoscale environment • Convection-permitting Inner Domain – Assimilate velocity and reflectivity from WSR88D radar network WSR88D 8

Mesoscale analysis and forecasts • GSI-EnKF shows better forecast except low level temperature. 9

Need for storm scale analysis Inner domain PQPF (>12.7mm/h) initialized with outer domain EnKF • Mesoscale environment defined in outer domain EnKF analysis was able to support the development of a realistic-looking MCS. • However, spin-up time and resulting westward displacement throughout storm period suggests need for storm scale data assimilation. 10

GSI-based EnKF vs. 3DVAR for storm scale analysis GSI-EnKF KTLX Observation GSI-3DVar 11

Probabilistic precipitation forecasts Prob. (> 12.7mm/h) GSI-EnKF after radar DA • Phase lag is largely improved after radar DA. • 3DVar worse than EnKF The storm was GSI-3DVar after radar DA  pushed out to the east by the cold pools faster than observed;  out to the south too much than observed;  dissipated too quickly than observed 12

BS for probabilistic precipitation forecasts Prob. (> 12.7mm/h) BS: the lower the better 13

Impact of mesoscale analysis on storm scale forecast EnKF • The better evolution of MCS at 4-8 h suggests the environment was better analyzed by EnKF than 3DVar • Mesoscale warm front in 3DVar eastern OK was analyzed too far south for GSI 3DVar, explaining the forecast differences after the first few hours. 14

V r vs. V r +dBZ for GSI 3DVAR • The forecast with reflectivity is subjectively worse than velocity-only • Including reflectivity caused much more expansive cold pools (not shown) • This may be a result of the deficiency of the 3DVar static background error covariance -- added hydrometeors largely evaporating and falling out since without adjusting temp/wind fields consistently. V r only V r with dBZ 15

V r vs. V r +dBZ for GSI EnKF • Improved V r only forecast at 2-3 h due to better forecast of central OK storm V r with dBZ 16

Summary and Conclusions  GSI-based 3DVar/EnKF/Hybrid system is further developed/extended to directly assimilate radar data for WRF ARW.  Implemented for a real data, MCS case study with complex and heterogeneous environment assimilating data at multiple scales. Performance was evaluated at both mesoscale and storm scale.  Comparison between GSI-3DVar and GSI-EnKF suggests a . using the flow-dependent ensemble can improve the analysis at both meso and convective scales. b . Forecast quality is sensitive to both the quality of the storm scale analysis and the mesoscale environment analysis.  Improvement of static covariance for storm scale assimilation is needed.  Further compare to understand GSI-based Var, EnKF and hybrid (3DEnsVar, 4DEnsVar) for convective scale DA.  Systematic tests with more cases. 17

Posters • Xu Lu : Assimilation of Airborne Doppler Radar observations using the GSI- based hybrid ensemble-variational data assimilation system to improve high resolution hurricane forecast by HWRF • Nick Gasperoni : Improving Ensemble-based Observation Impact Estimate using a Group Filter Technique • Terra Thompson : Multi-Scale Data Assimilation of the June 13, 2010 VORTEX2 Tornadic Supercell • Ting Lei : GSI-based Hybrid Data Assimilation for NCEP GFS: How Is the Dual Resolution Hybrid Compared to the Single Resolution Hybrid? • Kutty Govindan : A Comparison of Impacts of Radiosonde and AMSU Radiance Observations In GSI-based Hybrid and 3DVar Data Assimilation Systems for NCEP GFS Acknowledgement: Local organizing committee 18

20 May 2010 case overview Johnson et al. (2013) using wavelet analysis showed particularly strong sensitivity of precipitation forecast to both small scale and large scale perturbations. Johnson et al., 2013: Multiscale characteristics and evolution of perturbations for warm season convection-allowing precipitation forecasts: Dependence on background flow and method of www.mmm.locust.ucar.edu perturbation. Mon. Wea. Rev. , in review. 19