Download
the cobel model and processing numerical model output for weather forecasting n.
Skip this Video
Loading SlideShow in 5 Seconds..
The COBEL model and processing numerical model output for weather forecasting PowerPoint Presentation
Download Presentation
The COBEL model and processing numerical model output for weather forecasting

The COBEL model and processing numerical model output for weather forecasting

241 Views Download Presentation
Download Presentation

The COBEL model and processing numerical model output for weather forecasting

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

  1. UQAM Université du Québec à Montréal National Scale C&V Science Meeting, NCAR, April 19, 2001 The COBEL model and processing numerical model output for weather forecasting Dr. Peter Zwack Robert Tardif UQÀM Atmospheric Sciences NCAR, April 19 2001

  2. The COBEL model • 1D (column) model • ensemble-average 1D BL equations • moist physics for stratiform clouds • coupled with surface • High vertical resolution for “shallow” phenomena • surface inversions • nighttime LLJs • stratocumulus • stratus, fog etc... • 3D to 1D “nesting” for mesoscale influences • horizontal pressure force • horizontal advections • vertical motion UQÀM Atmospheric Sciences NCAR, April 19 2001

  3. The COBEL model (cont’) • Developed at Paul Sabatier Univ. to simulate the nocturnal boundary layer (Estournel and Guédalia, 1987) • Turbulence parameterization for very-stable stratification • Detailed parameterization for longwave radiative transfer • Further developed for radiation fog forecasting (Bergot and Guédalia, 1994) • Moist physics (fog) • Visibility • Adapted for marine stratus “burnoff” very short-term forecasting for west coast (UQAM) UQÀM Atmospheric Sciences NCAR, April 19 2001

  4. The COBEL model (cont’) Radiation IR : 232 channels (Vehil et al., 1989) Solar : Fouquart and Bonnel (1980) Turbulence (1.5 order closure) : Neutral : Delage (1974) Stable and very stable: Estournel and Guédalia (1987) Unstable: Bougeault and Lacarrère (1989) Soil : Ts through diffusion eq. (Bergot, 1993) Moisture following Mahrt and Pan (1984) Microphysics : Gravitational settling velocity (cloud) constant Precip. (drizzle) following Kessler (1969) External forcings(obs. or mesoscale model): Horiz.pressure force, horiz adv. (T and Hum), vertical motion and pressure tendencies UQÀM Atmospheric Sciences NCAR, April 19 2001

  5. What has been achieved... Fog onset forecasting (an example): Visibility (Kunkel,1984) • North of France - flat & homogeneous terrain • Winter, anticyclonic circulation (calm winds) • Moist soil • COBEL profiles initialized from “on-site” sounding • COBEL driven with mesoscale advections from surface mesonet • Soil temp. from “on-site” sensors (taken from Bergot, 1993) Good at forecasting fog onset when accurate initial cond. and advections UQÀM Atmospheric Sciences NCAR, April 19 2001

  6. Regional Wx Center COBEL + meso. model Probability of Detection 67% 85% False Alarm Ratio 25% 57% Most errors could be related to mesoscale model error (erroneous advections + mid-level & high-level cloudiness) What has been achieved ...(cont’) Pre-op. evaluation: Init. cond. & advections from operational mesoscale model Statistical verification of fog forecasts. 229 days, 54 observed fog events (Lille field experiments - 3 winters) (from Bergot and Guédalia, 1996) UQÀM Atmospheric Sciences NCAR, April 19 2001

  7. What has been achieved ...(cont’) Morning BL transition forecasting • Adapted COBEL for daytime conditions • Very short-term forecasts • Coupling with Meso-ETA forecasts (advections) • Init. cond. from “on-site” soundings and in-situ sensors • Real-time prototype system deployed at DFW Sept.-Oct. 1997 (wake vortex experiment) ETA 12hr fcst COBEL 3hr fcst • Able to forecast short-term morning evolution of BL • Soil moisture important • Better structure than available op. meso. model UQÀM Atmospheric Sciences NCAR, April 19 2001

  8. Init. with OAK sounding + local data (sodar, surface T, Td) burnoff Re-init. with updated local data (sodar, radiometer, surface T, Td) Re-init. with updated local data (sodar, radiometer, surface T, Td) What has been achieved ...(cont’) Marine stratus burnoff forecasting for SFO Hourly cloud water forecasts (an example) SFO August 12 1997 UQÀM Atmospheric Sciences NCAR, April 19 2001

  9. COBEL forecasts-summer 1999, 47 cases RMS Mean error (hours) advections important ! Mean error (COBEL - obs.) Forecast init. times Use of op. models (ETA, Canadian model…) advections deteriorated results ! What has been achieved ...(cont’) SFO “batch” results (without advections) UQÀM Atmospheric Sciences NCAR, April 19 2001

  10. T at 10m Ongoing development Horizontal advection retrieval methodology temperature and humidity advections Difference between model & obs : Horizontal advection + model error • Iterative method: Which advection so model converges toward obs. ? • Constrain model with obs. to minimize error • Do for T and Hum. UQÀM Atmospheric Sciences NCAR, April 19 2001

  11. Temperature advection retrieval Temperature advection (C/h) Hour (UTC) Ongoing development Horizontal advection retrieval methodology UQÀM Atmospheric Sciences NCAR, April 19 2001

  12. 10% forecast error reduction Ongoing development Horizontal advection retrieval methodology Correlation with COBEL burnoff forecast error Rel. hum. “advection” = Fct (T advection, Hum. advection) Summer 1999 (37 days) UQÀM Atmospheric Sciences NCAR, April 19 2001

  13. Ongoing development (cont’) • Navy’s COAMPS high-resolution (nest 3 ~ 5km) forecasts for horizontal advections & vertical motion • Coupling strategy with 3D mesoscale model for coastal clouds forecasting • COBEL initialization with in-situ measurements • Coupled COBEL/COAMPS system better than COBEL alone and COAMPS alone for cloud ceiling & burnoff forecasts ?? 5x5 high-res. grid UQÀM Atmospheric Sciences NCAR, April 19 2001

  14. Ongoing development(cont’) Coupling with COAMPS model for coastal clouds forecasting over Pt. Mugu California • COBEL/COAMPS “combo” can improve results • Burnoff forecasts • Ceiling forecasts • Meso. model must do arelatively good job at cloud forecasting... • If no burnoff over whole grid -> weak horiz. gradients -> weak advections -> no improvement over COBEL without advections (BO too early) METAR COBELCOAMPSCOBEL/COAMPS 2000Z 1812Z none 1845Z Ceiling RMSE : June 20th 2000 COAMPS : ~200m, COBEL/COAMPS : ~75m UQÀM Atmospheric Sciences NCAR, April 19 2001

  15. Nest 3 00Z “cold start” Vertical motion every 2 time steps • Model spin-up 0 24 • Fluctuations in derived output June 27 2000 Coupling with 3D mesoscale models Processing numerical model output Be careful about : UQÀM Atmospheric Sciences NCAR, April 19 2001

  16. Coupling with 3D mesoscale models Processing numerical model output Other lessons learned : System sensitivity to : • filtering options : 1hr burnoff forecast difference obtained with COAMPS advections filtered using 2 different filter configuration ! • 3D model spatial variability 1hr burnoff forecast difference obtained with COAMPS advections from 2 adjacent grid points ! UQÀM Atmospheric Sciences NCAR, April 19 2001

  17. Ongoing development • Mixed-phase microphysics being added to extend range of applications (available MM5 code of Rasmussen parameterization) • Adapt code to 1D model • Test various code configuration • Validation on idealized winter BL cloud • Simulation of fall/winter (near & below freezing) stratocumulus clouds over eastern US UQÀM Atmospheric Sciences NCAR, April 19 2001

  18. Looking ahead... • Areas for further R&D(for improved forecasting ability): • Microphysics of low clouds-fog • drop size distribution characterization and evolution during burnoff (has impact on cloud burnoff forecasts) • Better representation of local advections • Improve retrieval technique using more complete data sets from field campaigns - define data required • Coupling with RUC II and/or COAMPS - define improved synergy between models • Improve surface characterization • Soil temperature-moisture • Influence of surface heterogeneity on evolution of local atmosphere • Data assimilation ! UQÀM Atmospheric Sciences NCAR, April 19 2001

  19. Looking ahead... Integrated “1D array” system for C&V forecasting in precipitating systems UQÀM Atmospheric Sciences NCAR, April 19 2001

  20. UQÀM Atmospheric Sciences NCAR, April 19 2001

  21. What has been achieved ...(cont’) Evolution of profiles during morning BL transition (in support of wake vortex prediction system) • Added components to COBEL for daytime simulations (solar radiation, soil model for surface evaporation…) • Evaluation using data from Memphis Wake Vortex field experiment • Initial conditions from on-site sounding • Low soil moisture • External forcings from Canadian regional operational forecast model UQÀM Atmospheric Sciences NCAR, April 19 2001

  22. What has been achieved ...(cont’) Marine stratus burnoff forecasting for SFO Developments • Initialization procedure • Oakland sounding + SFO sodar BL height + SFO obs. T & Td at surface  COBEL initial profiles • COBEL “re-initialization” methodology • sodar BL height + obs. T & Td at surface + obs. incoming solar rad.  updated initial cond. for updated hourly forecasts • Horizontal advection retrieval methodology UQÀM Atmospheric Sciences NCAR, April 19 2001

  23. Initialization UQÀM Atmospheric Sciences NCAR, April 19 2001

  24. sounding Oakland + local data (12Z) Very short-term forecasts 12 Z forecast 13 Z forecast 14 Z forecast Re-initialization 15 Z forecast Re-initialization 16 Z forecast profiles COBEL + local data (13Z) profiles COBEL + Local data (14Z) time UQÀM Atmospheric Sciences NCAR, April 19 2001

  25. Re-initialization UQÀM Atmospheric Sciences NCAR, April 19 2001

  26. “advRH” vs COBEL error (COBEL - obs) Summer 1999 (37 days) Retrieval Forecasts 14Z 15Z 16Z 17Z 12-14Z -0.17-0.20 13-15Z -0.40-0.45 14-16Z -0.28 -0.19 15-17Z -0.04-0.28 If advRH (+) (RH ), COBEL without advection should be too early (error “-”) Ongoing development Horizontal advection retrieval methodology Correlation with COBEL burnoff forecast error Rel. hum. “advection” = Fct (T advection, Hum. advection) UQÀM Atmospheric Sciences NCAR, April 19 2001

  27. METAR COBEL alone COAMPS alone COBEL/COAMPS from 12Z from 00Z combo 2000Z 1812Z none 1845Z Ceiling RMSE COAMPS : ~200m COBEL/COAMPS : ~75m ceiling Ongoing development(cont’) Coupling with COAMPS model for coastal clouds forecasting over Pt. Mugu California …an example... June 20th 2000 burnoff UQÀM Atmospheric Sciences NCAR, April 19 2001

  28. -> Filtering required -> (40 min centered average) Coupling with 3D mesoscale models Processing numerical model output Be careful about : • Unwanted fluctuations in model derived fields (such as advections) ~ -2 C/h UQÀM Atmospheric Sciences NCAR, April 19 2001

  29. COBEL with COAMPS advections cloud burnoff forecasts 45 min. 180 min. 19990913 1719Z 1718Z (1745Z) 19990927 2225Z 2128Z (none) Coupling with 3D mesoscale models Processing numerical model output • Other lessons learned : System sensitivity to : Filtering options (averaging period) COAMPS output filtered (in time) using 45 min. or 180 min. centered moving average UQÀM Atmospheric Sciences NCAR, April 19 2001

  30. COBEL with COAMPS advections cloud burnoff forecasts (2,2) (3,3) 19990913 1719Z >2400Z 19990914 1605Z 1620Z 19990927 2225Z >2400Z Coupling with 3D mesoscale models Processing numerical model output • Other lessons learned : System sensitivity to : 3D model spatial variability (advections from various grid points) 5x5 high-res. grid -> Need for “ensemble” forecasting ? UQÀM Atmospheric Sciences NCAR, April 19 2001