Efficient Assimilation of Radar Data at High Resolution for Short-Range Numerical Weather Prediction Keith Brewster, Ming Hu, Ming Xue and Jidong Gao Center for Analysis and Prediction of Storms University of Oklahoma USA WSN05 6 Sep 2005 Toulouse, France Radar Analysis & Assimilation Research Topics in CAPS • Single-Doppler Velocity Retrieval (SDVR) • Bratseth-type Successive Correction Analysis (ADAS) • 3DVAR at Storm Scale • Cloud & hydrometeor analysis with latent heating adjustment • Phase/Position error correction methods • Ensemble-Kalman Filter at Storm Scale WSN05 6 Sep 2005 Toulouse, France Radar Analysis & Assimilation Research Topics in CAPS • Single-Doppler Velocity Retrieval (SDVR) • Bratseth-type Successive Correction Analysis (ADAS) • 3DVAR at Storm Scale • Cloud & hydrometeor analysis with latent heating adjustment • Phase/Position error correction methods • Ensemble-Kalman Filter at Storm Scale WSN05 6 Sep 2005 Toulouse, France CAPS 3DVAR Radar Assimilation Flow Chart Radar 1 Radar 2 External Model Interpolator Radar QC & Remapper Radar 3 Radar 4 Radar N Aircraft Multi-scale 3DVAR Rawinsondes AIRS Soundings Mesonets Wind Profilers Cloud Analysis & Latent Heat Adjustment ARPS NWP Model METAR Sat IR Satellite Remapper ARPS-to-WRF WSN05 6 Sep 2005 Toulouse, France Sat Vis WRF NWP Model Radar Quality Control & Remapping • Quality Control – AP & Clutter detection – Doppler radial velocity unfolding • Remapping – – – – Matches data spacing to model resolution Eases reflectivity mosaicking Can be viewed as a form of “superobbing” Local least-squares interpolation/smoothing Quadratic in horizontal, Linear in vertical WSN05 6 Sep 2005 Toulouse, France Remapping to Dx = 2 km WSN05 6 Sep 2005 Toulouse, France CAPS 3DVAR System • General form 1 1 b T 1 b o T J (x) x x B x x H x y R 1 H x y o J c x 2 2 • Rewritten in incremental form • Error correlation implemented by means of a recursive filter. • Can be applied in multi-grid fashion • Dynamic constraint: weak constraint: anelastic mass continuity 1 2 2 J c c D 2 w u v D y z x WSN05 6 Sep 2005 Toulouse, France Radar Ingest- Reflectivity • Cloud analysis system – Remapped Satellite Images (Vis and IR) – Surface observations of cloud bases – Reflectivity converted to hydrometeors Rain, hail, dry snow, wet snow • Cloud water quantity and latent heating estimated using a lifted-parcel with entrainment WSN05 6 Sep 2005 Toulouse, France 3DVAR Applied to Fort Worth Tornadic Storm • Fort Worth, Texas area tornadoes of 28 Mar 2000 • 3-km ARPS Forecast 23 UTC-06 UTC nested in 9-km forecast 18 UTC – 06 UTC • Six 10-min analysis cycles (1 hour) using NEXRAD data 22 UTC-23 UTC. • Experiments: – Wind and Cloud Assimilated – Wind Alone – Cloud Alone Ming Hu et al. papers submitted to MWR WSN05 6 Sep 2005 Toulouse, France 00:30 UTC Radar Reflectivity 1.5 h Forecast Wind & Cloud Assim WSN05 6 Sep 2005 Toulouse, France 1.5 h Forecast Cloud Only Assim 1.5 h Forecast Wind Only Assim WSN05 6 Sep 2005 Toulouse, France 00:30 UTC Radar Reflectivity 1.5 h Forecast Surface Vorticity Wind & Cloud Assim WSN05 6 Sep 2005 Toulouse, France 1.5 h Forecast Surface Vorticity Cloud Only Assim 1.5 h Forecast Surface Vorticity Wind Only Assim WSN05 6 Sep 2005 Toulouse, France Fort Worth Case Summary • Similar situation observed for second tornado about 15 min later. • Good forecast results for this case primarily due to cloud & diabatic portion of analysis. • Winds provide improvement to forecasted vorticity. • Applicable to on-going convection; other case studies show utility of radial wind assimilation in convection-initiation forecast situations. WSN05 6 Sep 2005 Toulouse, France 1-hour Forecast (1-hr Accum Precip)17-May-2004 01:00 WRF IC: Eta Interp WRF IC: ADAS w/Radar Radar Precip Obs WSN05 6 Sep 2005 Toulouse, France 2004 Real-time Use Summary • Spin-up at 4-km is largely eliminated using radar and satellite data. • Good results even with a static analysisinitialization. WSN05 6 Sep 2005 Toulouse, France Sample of Ongoing & Future Work with These Tools • Testing different lengths of assimilation cycle and total assimilation window length • Will also test using 3DVAR output in Incremental Analysis Updating • More real-time high-resolution test periods in collaboration with SPC/NSSL • Smaller-domain real-time system run daily http://www.caps.ou.edu/wx WSN05 6 Sep 2005 Toulouse, France Credits • CAPS Research Scientists – Ming Xue, Jidong Gao, Dan Weber, Kelvin Droegemeier • CAPS Model and Real Time System Support – Kevin Thomas and Yunheng Wang • CAPS Students – Ming Hu, Dan Dawson • WSN05 Conference Travel Support OU School of Meteorology WeatherNews Chair funds WSN05 6 Sep 2005 Toulouse, France