NOCN02 CWAO 181830
GENOT FAX NO. 15

CECI EST LA VERSION ANGLAISE DE GENOT FAX NO. 016.

REINTRODUCTION OF THE REGIONAL DATA ASSIMILATION SYSTEM
AT CMC.

AS OF WEDNESDAY SEPTEMBER 24 1997 AT 1200 UTC, THE
REGIONAL DATA ASSIMILATION SYSTEM (SPIN-UP) WILL BE
REINTRODUCED AT THE CANADIAN METEOROLOGICAL CENTRE. THE
3D-VAR (3-DIMENSIONAL VARIATIONAL) ASSIMILATION SYSTEM
THAT HAS BEEN IMPLEMENTED FOR GLOBAL DATA ASSIMILATION
AT CMC ON JUNE 18 1997 HAS BEEN ADAPTED FOR REGIONAL
DATA ASSIMILATION IN WHICH THE GEM MODEL IS USED TO
GENERATE THE FIRST-GUESS FIELDS. THE REGIONAL SPIN-UP
SYSTEM IS SET-UP IN A VERY SIMILAR WAY AS IT WAS
PREVIOUSLY WHEN THE RFE MODEL WAS USED AS THE CMC
REGIONAL MODEL. EVERY 12-HOUR A NEW REGIONAL CYCLE IS
STARTED FROM THE GLOBAL ASSIMILATION SYSTEM;  THE GEM
MODEL IS USED IN THIS CYCLE TO GENERATE THE HIGHER
RESOLUTION FIRST GUESS FIELDS (6-HOUR FORECASTS) NEEDED
TO PRODUCE THE NEXT REGIONAL ANALYSIS. THIS IS DONE
TWICE, FOR A TOTAL SPIN-UP TIME OF 12 HOURS. A REGIONAL
ANALYSIS WITH A DATA CUTOFF TIME OF 1:40 IS THEN
PRODUCED AND USED BY THE GEM MODEL TO GENERATE FORECASTS
OUT TO 48 HOURS. ALL THE REGIONAL UPPER AIR ANALYSES
WILL BE PRODUCED USING THE 3D-VAR SCHEME IN SPIN-UP
MODE. THEREFORE, THE ANALYSES DATA (CHARTS AND GRIB)
PRODUCED WITHIN THE MAIN REGIONAL RUNS AT THE CMC WILL
BE PRODUCED FROM A REGIONAL ANALYSIS AT THE END OF THE
12-HOUR SPIN-UP CYCLE.

THIS NEW REGIONAL DATA ASSIMILATION SYSTEM HAS BEEN RUN
IN PARALLEL SINCE THE END OF JULY, FOR BOTH THE 0000 UTC
AND 1200 UTC CYCLES. THE NEW REGIONAL ANALYSES, AND
FORECASTS FROM THE GEM MODEL INITIALIZED FROM THE
REGIONAL 3D-VAR ANALYSIS HAVE BEEN ASSESSED BOTH
OBJECTIVELY, AND SUBJECTIVELY BY CMC OPERATIONAL
METEOROLOGISTS.

PERFORMANCE EVALUATION:

ANALYSES: THE FIT OF GEOPOTENTIAL AND TEMPERATURE
ANALYSES TO OBSERVED DATA IS SLIGHTLY BETTER IN THE NEW
REGIONAL SYSTEM; MOISTURE ANALYSES, ON THE OTHER HAND,
ARE SLIGHTLY FURTHER AWAY FROM OBSERVATIONS. A
SIGNIFICANT IMPROVEMENT HAS BEEN NOTED FOR THE WIND
ANALYSES NEAR THE JET LEVEL WHERE THE WINDS IN THE NEW
REGIONAL ANALYSIS GENERALLY BETTER FIT THE OBSERVATIONS.
SUBJECTIVE EVALUATION HAS SHOWN THAT OFTEN, THE STRONGER
JETS ARE BETTER REPRESENTED IN THE ANALYSIS PRODUCED IN
THE SPIN-UP SYSTEM THAN IN THE CURRENT ONE. ABOVE 100
HPA, THE FIT IS NOT QUITE AS GOOD, BUT THIS IS NOT
DETRIMENTAL TO THE ACCURACY OF THE FORECASTS IN THE TIME
RANGE UP TO 48 HOURS. THE ANALYZED GEOPOTENTIAL IS
SLIGHTLY NOISIER THAN THE CURRENT OPERATIONAL ONE,
ESPECIALLY OVER THE WEST COAST OF NORTH AMERICA AND
UNDER SLACK UPPER AIR FLOW, WHERE THIS BEHAVIOR IS MOST
NOTICEABLE. THIS IS CARRIED BY THE TRIAL FIELD, AND EVEN
THOUGH NOT AESTHETIC, THE AMPLITUDE OF THESE
OSCILLATIONS REMAINS SMALL WITH TIME.

FORECASTS: GENERALLY THE SCORES FROM THE PARALLEL
SYSTEM INDICATE A VERY SLIGHT IMPROVEMENT OVER THE
OPERATIONAL SYSTEM EXCEPT FOR GEOPOTENTIAL AND
TEMPERATURE AT THE JET LEVEL WHERE THE IMPROVEMENT IS
SIGNIFICANT. IT IS REFLECTED AS A LARGE REDUCTION IN THE
BIAS AT THAT LEVEL. FORECASTS OF FIELDS SUCH AS MSLP AND
GEOPOTENTIAL AT VARIOUS LEVELS ARE GENERALLY SIMILAR IN
THE NEW AND OPERATIONAL SYSTEMS. AT TIMES WHERE THERE
HAS BEEN SIGNIFICANT DIFFERENCES, THE FORECASTS BASED ON
THE NEW ANALYSIS ISSUED FROM THE 12-HOUR SPIN-UP WERE
GENERALLY JUDGED BETTER AND CLOSER TO VERIFYING DATA.

PRECIPITATION: ONE OF THE MAIN REASON FOR IMPLEMENTING A
REGIONAL SPIN-UP SYSTEM IS TO IMPROVE SHORT-TERM
PRECIPITATION FORECASTS. AN EXAMINATION OF THE
PRECIPITATION RATES FROM THE GEM MODEL INITIALIZED FROM
THE SPIN-UP ANALYSIS CLEARLY INDICATED A VERY
SIGNIFICANT REDUCTION IN THE AMOUNT OF TIME TAKEN BY THE
MODEL TO REACH AN APPROPRIATE LEVEL OF PRECIPITATION
WITH BOTH MODELS PROVIDING VERY SIMILAR PRECIPITATION
RATES AFTER 12 HOURS. THE EFFECT IS MUCH MORE EVIDENT IN
THE CONVECTIVE PORTION OF PRECIPITATION. VERY OFTEN, THE
PRECIPITATION ACCUMULATION IN THE FIRST 6 AND 12 HOURS
OF INTEGRATION OF GEM MODEL BASED ON THE SPIN-UP CYCLE
WAS SIGNIFICANTLY MORE THAN IN THE CURRENT OPERATIONAL
SYSTEM, AND WAS CLOSER TO AVAILABLE PRECIPITATION
OBSERVATIONS. BEYOND 12 HOURS, PRECIPITATION FORECASTS
PRODUCED BY EITHER THE OPERATIONAL OR PARALLEL SYSTEMS
WERE CLOSER TO EACH OTHER, EXCEPT IN CONVECTIVE
SITUATIONS WHERE SIGNIFICANT DIFFERENCES WERE AT TIMES
NOTED OUT TO 48 HOURS.

MCBEAN/ADMA/TORONTO