Presentation is loading. Please wait.

Presentation is loading. Please wait.

Forecast Sensitivity from new Models and Special Observations Jan Paegle (U. of Utah), Lee A. Byerle (USAF) Celeste Saulo, Juan Ruiz (U. Buenos Aires)

Similar presentations


Presentation on theme: "Forecast Sensitivity from new Models and Special Observations Jan Paegle (U. of Utah), Lee A. Byerle (USAF) Celeste Saulo, Juan Ruiz (U. Buenos Aires)"— Presentation transcript:

1 Forecast Sensitivity from new Models and Special Observations Jan Paegle (U. of Utah), Lee A. Byerle (USAF) Celeste Saulo, Juan Ruiz (U. Buenos Aires) Julia Nogues-Paegle (U. of Utah)

2 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman OUTLINE.Previous work by Gonzalo Miguez-Macho.Recent work by Jan Paegle, Lee Byerle, Juan Ruiz, Celeste Saulo, Julia Paegle

3 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman BACKGROUND Gonzalo estimated initial state uncertainty from the difference of two equally credible analyses (EC and NCEP) He initialized the Utah P.E. model with each analysis => 2 forecasts of same case And studied the differences of these forecasts Next 4 slides from his dissertation

4 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman

5

6

7 “predictability” curves are useful to determine fcst sensitivity to new analyses or observing systems

8 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Gonzalo: Area integrated results for N. America suggest that: Enhanced obs over N. America more valuable than those outside N. America until about 30h After 30h enhanced obs outside N. America are more important

9 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Remaining questions.Do results: depend on forecast model?. on initial state uncertainty estimate?. on region and season? Present talk employs.newer models (regional WRF, Utah global).and special observations to get obs uncertainty.Over a relative data void (S. America, summer).Use time evolution of area integrated results

10 OUTLINE.Much previous work has emphasized the impact of init. state changes over N.H..Most of the testing done with PE models.Here we discuss forecast sensitivity to special obs over S. America.Use recent Euler models WRF New Global Euler

11 Operational S. American radiosonde network pre-SALLJEX

12 S. American radiosonde network (pre-SALLJEX) positioned over N. America

13 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman An example of radiosonde coverage at 00UTC

14 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Example of Aircraft coverage

15 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman SALLJEX obs enhancement (blue) Upper air enhancement 12/02-2/03

16 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Dirceu and Kousky (2007) assimilated SALLJEX obs using Operational NCEP GDAS system: EXPERIMENT control GDAS assimilations w/o SALLJEX obs: CONTROL The subsequent slides show impact of these obs in Forecast models that are initialized with both analyses

17 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman |V(exp)-V(con)| at 500 mb, 24 Jan 2003, 00UTC (analyses) 10 m/s

18 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Area ave RMS diff (exp-con) at 500 mb Area rms |V(exp)-V(con)|, 500mb Analysis 2 m/s

19 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman WRF Euler model forecast response res: 50 km Regional dmn

20 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman (48h response)/(00h response) for WRF Euler model < 2 at all times Ave response amplifies by About 20% in 48h 2 1

21 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Utah Global Euler Model Our Euler model runs stably without horiz diffusion (K H ) in troposphere

22 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Evolution of |V(exp)-V(con)| at 500 mb for Utah Euler model 16 Jan -31 Jan 2003 Day 0Day 4 Day 9 Day 14

23 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Area average of global Euler model sensitivity vs fcst day for selected domains Much more sensitive (400% growth in 48h) than WRF model (20% growth in 48h) Why? 16m/s 0 m/s

24 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Tools: uniform resolution global model Uniform resolution grid in Euler model:

25 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Tools: rotated, uniform res, global model Mathematical pole rotation and variable resolution in Euler model:

26 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Tools: stretched grid, rotated, global model Increased resolution over the area of interest in Euler model:

27 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Selected Case Mesoscale convective system observed on January 17 th 2003 January 17 at 17:45 UTC January 17 at 23:45 UTC January 18 at 05:45 UTC January 18 at 17:45 UTC

28 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman |V(exp)-V(con)| 500 mb for Variable res. Utah Euler model (S. Am) w. diff, w. precip no diff, no precip no diffusion, with precip Uniform res

29 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman INTERIM CONCLUSIONS Moist, non diffusive variable resolution Euler model most responsive Dry, non-diffusive case less responsive Moist, diffusive case also less responsive But all are more responsive than WRF or PE models Could it be that the moist non-diffusive Euler model simply generates noise? Next examine forecast accuracy of Euler cases

30 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Anomaly correlation Euler model no diffusion with precip. S. H. skillful for 144h

31 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman, Anomaly correlation Euler model with diffusion with precip SH skillful for 130h

32 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman Anomaly correlation Euler model no diffusion no precip SH skillful for 126h

33 1020 April 2004 Intercomparison of Global Research and Operational Models Jennifer C. Roman CONCLUSIONS.fcst. response to obs depends on model.Utah PE model least responsive.Var. res. global Euler model most responsive.Regional WRF intermediate.Precip processes, diffusion are important.Most responsive global model is most accurate.More study needed on FCST ERROR


Download ppt "Forecast Sensitivity from new Models and Special Observations Jan Paegle (U. of Utah), Lee A. Byerle (USAF) Celeste Saulo, Juan Ruiz (U. Buenos Aires)"

Similar presentations


Ads by Google