MARCH 2004 UPGRADE OF THE NCEP GLOBAL ENSEMBLE FORECAST by kimbrozic

VIEWS: 9 PAGES: 12

									      MARCH 2004 UPGRADE OF THE
NCEP GLOBAL ENSEMBLE FORECAST SYSTEM


                       Zoltan Toth
                      Yuejian Zhu
                     Richard Wobus

           Environmental Modeling Center
                 NOAA/NWS/NCEP



                 Acknowledgements:
         David Michaud, Yuqiu Zhu             NCO
         Hua-Lu Pan, Steve Lord               EMC

       http://wwwt.emc.ncep.noaa.gov/gmb/ens/index.html

                                                          1
            NCEP GLOBAL ENSEMBLE FORECAST SYSTEM

RECENT UPGRADE (Apr. 2003)
10/50/60% reduction                  NEW CONFIGURATION
in initial perturbation size over       MARCH 2004
NH/TR/SH


     CURRENT SYSTEM




                                                         2
              PLANNED CHANGES - SUMMARY


1. Extend T126 portion of forecast from 84 to 180 hours


2. Run 4 cycles per day, adding 06Z and 18Z forecasts


 3. Extend out to 180 hrs the high resolution enspost and ensstat
 files, and add new precipitation-related enspost and ensstat files


4. Shorten the time step for the T126 part of the forecast




                                                                      3
                       PLANNED CHANGES - 1


1. Extend T126 portion of forecast from 84 to 180 hours


This change is intended to improve ensemble support for
4-7 day forecasting, to allow tropical cyclone tracking
out to 7 days, and to provide boundary conditions for
SREF out to 90 hours


Rusults:
       More realistic forecast patterns between 84-180 hrs
       Improved skill beyond 84 hours
       Increased spread in 4-10 days period, better reliability

                                                                  4
                       PLANNED CHANGES - 2

2. Run 4 cycles per day, adding 06Z and 18Z forecasts


This change provides more timely information,
corresponding to the 4 GFS control forecasts each day, and provides
more ensemble members to contribute to medium-range and week-2
forecasts


Results:
    The new cycles increase the information content
    and utility of ensemble-based products

                                                                      5
                      PLANNED CHANGES - 3


3. improved enspost & ensstat output information
a) Extend out to 180 hrs the high resolution (1x1) files
b) Increase frequency of ensemble output from 12- to 6-hrly
c) Add precipitation type information


These changes improve ensemble support for
medium-range and week-2 weather forecasting


Results:
   Utility of ensemble forecast data enhanced

                                                              6
                       PLANNED CHANGES - 4

4. Shorten the time step for the T126 part of the forecasts


This change is intended to improve the quality of the ensemble
forecasts by using the same 10 min time step as used in the GFS control
instead of the longer (20 min) step that had been used in the ensemble


Results:
   This change improves the quality of the high resolution T126 portion
   of the ensemble forecasts that, in certain flow situations, has been
   degraded by the use of the longer time step. Occasional poor
   ensemble performance related to time step problem is eliminated

                                                                          7
   RMS ERROR, ENSEMBLE SPREAD, OUTLIER STATISTICS, NH

       RESULTS BASED ON 29-DAY NCO PARALLEL TESTING




RMS error slightly reduced      Number of cases when ensemble
Ensemble spread significantly     misses verification significantly
  increased                       reduced

                                                                      8
PATTERN ANOMALY CORREL, RANKED PROBABILITY SKILL SCORE, NH




PAC slightly improved          Probabilistic skill scores improve by
                                  0.5 day around day 7 lead time


                                                                   9
        PAC FOR SH, BRIER SKILL SCORE FOR TROPICS




Performance improves over SH    Both reliability and resolution
  extratropics as well            components of probabilistic
                                  performance improves over the
                                  tropics
                                                              10
HOW TO BEST UTILIZE INCREASED COMPPUTING POWER IN FUTURE?
 •    Run improved model                   •   Run more members        •   Make reforecast runs
       – Increase resolution                   – Increase                   – For improved post-
            • Horizontal              OR         membership       OR          processing
            • Vertical                         – Run ens. more
       – Use better physics                      often



• How do we evaluate test results?
     – Answer will depend on user procedures                                    Whitaker and Hamill, 2004
         • Evaluate before or after statistical post-processing
             – Effect of bias correction, smoothing, inter- and extrapolation etc can be crucial
     – Multifaceted problem, scientific results, user input, all need careful consideration
         • Trade-off



                                                                            Mullen & Buizza, 2002l




                                               Wei & Toth, 2003

                Toth & Kalnay, 1997



                                                                                                      11
                              NCEP PLANS
• Test effect of doubling membership
   – Still over somewhat steep curve regarding potential gain
   – 5% detection level for highly nonlinear aspects of forecast
       • For more linear aspects, extrapolation gives good probabilistic resolution
   – Can help with ET applications
       • Rescaling for generating initial perturbations
       • Targeted observation applications
   – Can be combined with MSC & other ensembles in future to gain 100+ members

• Increase resolution?
   – For first 84 hrs from T126 to T170?
   – Vertical resolution from 28 to 64 levels?
       • Coupled ocean-atm model uses T62-L64 version
   – For 2nd week, from T62 to T126?

• Keep room for operational generation of
      reforecast data for next configuration?
   – Use two-tier approach?
       • Update model used for 1st week semi-annually
       • Update only every 3 yrs for 2nd week?
                                                                    Whitaker and Hamill
                                                                                          12

								
To top