NOCN03 CWAO 091830
GENOT NO. 014

GENOT NO. 013 EST LA VERSION FRANCAISE DE CE GENOT.

ON TUESDAY DECEMBER 13, 2005, AT 12 UTC (EXACT TIME TO BE CONFIRMED),
THE CANADIAN METEOROLOGICAL CENTRE WILL IMPLEMENT A MAJOR CHANGE TO
THE ENSEMBLE PREDICTION SYSTEM (EPS).   

SOME NEW TYPES OF OBSERVATIONS WILL START BEING USED BY THE ENSEMBLE
KALMAN FILTER (ENKF) DATA-ASSIMILATION SYSTEM WHICH ITSELF WILL ALSO
BENEFIT FROM SOME STRUCTURAL IMPROVEMENTS. THE ENSEMBLE OF
FORECAST MODELS HAS BEEN REVISED TO ARRIVE AT A BETTER SIMULATION OF
THE DIURNAL CYCLE AND TO HAVE A REDUCTION IN MODEL BIASES. THE
FORECASTS WILL BE PERFORMED TWICE DAILY AND THE FORECAST PERIOD WILL
BE EXTENDED FROM 10 DAYS TO 16 DAYS.

MOTIVATION

THE REGULAR ADDITION OF NEW TYPES OF OBSERVATIONS IS NECESSARY IN
ORDER TO MAINTAIN A STABLE ANALYSIS QUALITY AS OLD SATELLITE-BASED
OBSERVING PLATFORMS ARE BEING REPLACED BY NEW ONES.  IN ADDITION,
CHANGES TO THE ANALYSIS ALGORITHM ARE NEEDED IN PREPARATION TO AN
UPCOMING ASSIMILATION OF OBSERVATIONS AT THE PROPER TIME (AS OPPOSED
TO ASSUMING THAT ALL OBSERVATIONS ARE VALID AT SYNOPTIC TIMES).

THE MODEL CHANGES HAVE BEEN MOTIVATED BY THE ESTABLISHMENT OF THE
NORTH AMERICAN ENSEMBLE FORECAST SYSTEM (NAEFS) IN WHICH THE
CANADIAN AND AMERICAN MEDIUM-RANGE ENSEMBLES WILL BE COMBINED INTO
A LARGER 'NORTH AMERICAN' ENSEMBLE.

TO PARTICIPATE IN THE NAEFS, WHICH WILL FEATURE FORECAST PRODUCTS
FOR WEEK TWO, WE MUST EXTEND THE FORECAST RANGE TO 16 DAYS AS WELL
AS REDUCE FORECAST BIASES IN ORDER TO MAINTAIN AN ACCEPTABLE QUALITY
IN THE SECOND WEEK.

ASSIMILATION

IN THE ENKF, THE FOLLOWING TYPES OF OBSERVATIONS WILL BE ADDED:
- MODIS WINDS FROM THE AQUA AND TERRA SATELLITES IN THE POLAR
REGIONS,
- AMSU-A RADIANCES FROM THE AQUA SATELLITE,
- DEWPOINT DEPRESSION AT THE SURFACE.

TO ARRIVE AT A BETTER BALANCE IN THE TRIAL  FIELDS, A FEW CHANGES
WILL BE MADE TO THE ENKF ALGORITHM.

FIRST, THE PARAMETERIZED ERROR FIELDS, WHICH ARE NOW ADDED TO THE
6H FORECASTS, WILL BE ADDED TO THE ANALYSIS. CONSEQUENTLY, THESE
PERTURBATIONS WILL EVOLVE FOR 6 HOURS WITH THE MODEL DYNAMICS
BEFORE THEY ARE USED IN THE ANALYSIS.  THE MODEL WILL THUS FILTER
SOME OF THE UNBALANCED COMPONENTS THAT ARE PRESENT IN THE
PARAMETERIZED ERROR FIELDS.


IN PREPARATION FOR A FUTURE IMPLEMENTATION OF TIME INTERPOLATION IN
THE ENKF, A DIGITAL FILTER INITIALISATION HAS BEEN INTRODUCED. THIS
FURTHER REDUCES THE IMBALANCE IN THE ENSEMBLE OF 6H FORECASTS
THAT IS USED TO DETERMINE THE DIRECTIONS IN WHICH ANALYSIS INCREMENTS
ARE REQUIRED.

FINALLY, THE 96 ENSEMBLE MEMBERS ARE NOW SPLIT INTO TWO
SUB-ENSEMBLES OF 48 MEMBERS. THIS WAS SET UP SO THAT THE MEMBERS OF
EACH SUB-ENSEMBLE COULD BE ASSIMILATED USING THE INDEPENDENT
STATISTICS OF THE 48 MEMBERS FROM THE OTHER SUB-ENSEMBLE. A
CONFIGURATION WITH FOUR 24 MEMBER SUB-ENSEMBLES WILL BE USED,
PERMITTING STATISTICS TO BE COMPUTED FROM 72 INDEPENDENT MEMBERS.

FORECAST:

THE MEDIUM-RANGE ENSEMBLE FORECASTS WILL NOW BE RUN TWICE A DAY
WITH THE FORECASTS EXTENDING UNTIL DAY 16.

THE CONFIGURATIONS OF THE SEF AND GEM SUB-ENSEMBLES HAVE BEEN
REVISED TO HAVE LESS BIAS IN THE SECOND FORECAST WEEK. TO FACILITATE
THE MAINTENANCE AND TO IMPROVE THE QUALITY OF THE EPS, UPDATED
PHYSICAL PARAMETERIZATIONS WILL BE USED FOR BOTH THE SEF AND THE GEM
SUB-ENSEMBLES. NOTEWORTHY IS THE USE OF THE ISBA (INTERACTIONS
SURFACE BIOSPHERE AND ATMOSPHERE) ALGORITHM FOR HALF OF THE ENSEMBLE
MEMBERS. THE USE OF THIS ALGORITHM SIGNIFICANTLY IMPROVES THE
RELIABILITY AND RESOLUTION OF THE SURFACE TEMPERATURE FORECASTS.

SUMMARY OF THE OBJECTIVE EVALUATIONS

EXTENSIVE TESTING OF THE NEW EPS, IN BOTH SUMMER AND WINTER
CONDITIONS, WAS PERFORMED BY THE R&D TEAM PRIOR TO THE
IMPLEMENTATION OF A PARALLEL RUN STARTING ON SEPTEMBER 2ND, 2005,
ALLOWING ADDITIONAL VERIFICATION OF DIFFERENCES BETWEEN THE
OPERATIONAL EPS AND THE NEW ONE.

OBJECTIVE VERIFICATIONS OF AVERAGE 6-HOUR ENSEMBLE FORECASTS AGAINST
UPPER AIR SOUNDING OBSERVATIONS FOR WIND SHOWED A STATISTICALLY
SIGNIFICANT IMPROVEMENT TO THE RMS ERROR FOR ALL FORECAST PERIODS,.
FOR GEOPOTENTIAL HEIGHT, TEMPERATURE AND DEW POINT DEPRESSION THE
IMPROVEMENTS WERE FAIRLY MINOR. IT MUST BE NOTED, HOWEVER, THAT THE
IMPROVEMENTS WERE RELATIVELY BIGGER DURING THE PARALLEL RUN AND
CLEARLY SIGNIFICANT FOR ALL VARIABLES.

COMPARISON OF THE OPERATIONAL AND THE NEW MEDIUM-RANGE ENSEMBLE
PREDICTIONS SHOWS A SIGNIFICANT REDUCTION IN THE BIASES FOR
TEMPERATURE AND GEOPOTENTIAL HEIGHT FOR ALL FORECAST  PERIODS AND
ALL REGIONS EXCEPT FOR THE ANTARCTIC CONTINENT.

FOR SURFACE TEMPERATURE, THE NEW FORECAST ENSEMBLE, WHICH BENEFITS
FROM THE ISBA PARAMETERIZATION, BETTER MODELS THE DIURNAL CYCLE.
THIS RESULTS IN MORE RELIABLE AND MORE PRECISE TEMPERATURE FORECASTS.
SIGNIFICANT IMPROVEMENTS IN BRIER SCORE WERE NOTICED FOR THE
PRECIPITATION.   THIS IS DUE TO AN IMPROVED, BUT STILL POOR,
RELIABILITY IN THE PREDICTION OF SMALL PRECIPITATION AMOUNTS.

SUMMARY OF THE SUBJECTIVE EVALUATIONS

THE METEOROLOGISTS OF THE OPERATIONAL ANALYSIS AND PROGNOSIS
DIVISION EVALUATED SUBJECTIVELY THE PARALLEL ENSEMBLE FORECAST
SYSTEM DURING 65 DAYS. THE GEOPOTENTIAL HEIGHT AT 500 HPA, MEAN SEA
LEVEL PRESSURE, AND THE QUANTITATIVE PRECIPITATION FORECAST (QPF)
FIELDS WERE EVALUATED AT FORECAST TIMES OF 48, 72, 96 AND 120 HOURS.
THESE FIELDS WERE EVALUATED ACROSS FOUR REGIONS: ARCTIC, WEST COAST,
CENTER, AND EAST. QPF WAS NOT EVALUATED ACROSS THE ARCTIC DUE TO
LACK OF DATA. 

OVERALL THE EVALUATION INDICATED A SLIGHT ADVANTAGE IN FAVOR OF
THE NEW SYSTEM. THIS ADVANTAGE WAS MORE PRONOUNCED FOR THE 500MB
GEOPOTENTIAL FIELD, ESPECIALLY IN THE EAST BUT ALSO IN THE ARCTIC.
IN SPITE OF THE APPARENT OVERALL NEUTRALITY OF THE SIGNAL,
SIGNIFICANT DIFFERENCES EVEN IN THE 48 HOUR FORECASTS WERE
REGULARLY NOTED BOTH AT THE SURFACE AND ALOFT.   

FOR THE PRECIPITATION FORECAST THE OPERATIONAL SYSTEM WAS
PREFERRED MORE FREQUENTLY THAN THE PARALLEL SYSTEM, ESPECIALLY
ALONG THE WEST COAST. THE LIGHTER AMOUNTS OF PRECIPITATION
FORECASTED BY THE NEW SYSTEM ALONG THE COASTAL MOUNTAINS WERE
GENERALLY JUDGED NOT AS ACCURATE AS IN THE OPERATIONAL SYSTEM.
HOWEVER THE HIGHER AMOUNTS FORECASTED IN THE OPERATIONAL SYSTEM
WERE PRODUCED BY THE BIASED MANABE SCHEME. 

WE NOTED THAT THE INTRODUCTION OF THE ISBA SURFACE SCHEME IN 8
OF THE PARALLEL MEMBERS PRODUCED A DIURNAL VARIATION OF SURFACE
TEMPERATURES WHICH WAS MORE REALISTIC AND ALSO RESULTED IN A
BETTER VERIFICATION OF EXTREME TEMPERATURES OVER THE CONTINENT
(COLD ZONES OVER THE CONTINENT IN PARTICULAR). A MORE REALISTIC
REPRESENTATION OF BAROCLINIC ZONE TEMPERATURE GRADIENTS WAS SEEN,
ESPECIALLY IN SITUATIONS WHERE WE HAD STRONG TEMPERATURE
ADVECTION.

IMPACTS

THE SUITE OF PRODUCTS WILL NOT BE CHANGED IMMEDIATELY, BUT DURING
THE WINTER.

USERS OF ENSEMBLE DATA WILL NOTICE DELAYS IN THE AVAILABILITY OF
00Z PRODUCTS AFTER THE IMPLEMENTATION OF THE NEW SYSTEM IN
OPERATIONS. THE SEF AND GEM MEMBERS RUN FOR ABOUT 15-30 MINUTES
LONGER IN THE ENSEMBLE PARALLEL RUN.  AS WELL, NCEP POST
PROCESSING JOBS TAKE AN ADDITIONAL 10-20 MINUTES TO RUN. ENSEMBLE
COLOR IMAGES SPAGHETTI PLOTS WILL BE AVAILABLE 20-30 MINUTES
LATER WHILE NAEFS GRIB DATA WILL LIKELY BE AVAILABLE 30-50
MINUTES LATER.

THE 12Z PRODUCTION RUN IS A PURE ADDITION. THE 12Z SUITE OF
PRODUCTS,  THE EXTENDED FORECASTS PAST DAY 10, AS WELL AS A
SERIES OF NEW PRODUCTS, WILL BE MADE AVAILABLE ON THE WEB LATER
THIS WINTER.

EVERELL / ADMA / SMA-SMC TORONTO