NOCN03 CWAO 101420 GENOT NO. 005 GENOT NO. 006 EST LA VERSION FRANCAISE DU GENOT NO. 005 4D-VAR: A MAJOR IMPROVEMENT TO THE CANADIAN GLOBAL DATA ASSIMILATION SYSTEM AS OF TUESDAY MARCH 15 2005 AT 1200 UTC, THE CANADIAN METEOROLOGICAL CENTRE WILL INTRODUCE A MAJOR CHANGE TO THE GLOBAL DATA ASSIMILATION SYSTEM BY IMPLEMENTING A 4D-VAR ANALYSIS SCHEME. SUMMARY OF CHANGES THE 3D-VAR (3-DIMENSIONAL VARIATIONAL) DATA ASSIMILATION SYSTEM FOR THE GLOBAL FORECASTING SUITE WILL BE EXTENDED TO A 4D-VAR SCHEME. THIS EXTENSION TO INCLUDE TIME AS THE FOURTH DIMENSION OF THE ANALYSIS SYSTEM IMPLIES THAT THE ANALYSIS IS NO LONGER A SIMPLE SNAPSHOT OF THE ATMOSPHERIC CONDITIONS AT THE TIME OF THE ANALYSIS (LIKE IN THE 3D-VAR WHICH DOES THIS FOR THE PRIMARY SYNOPTIC HOURS). INSTEAD, THE 4D-VAR SYSTEM PROVIDES A TIME EVOLUTION OF THE ATMOSPHERIC CONDITIONS DURING THE ASSIMILATION TIME WINDOW (FOR THE CURRENT IMPLEMENTATION, THIS WINDOW WAS CHOSEN TO MAINTAIN COHERENCE WITH THE 3D-VAR, THAT IS 6 HOURS CENTERED ON THE MAIN SYNOPTIC HOURS). THE 4D-VAR ANALYSIS IS DONE BY USING THE GEM FORECAST MODEL ITSELF AS PART OF THE ASSIMILATION PROCESS, AND BY USING OBSERVATIONS THAT ARE WELL DISTRIBUTED DURING THE ANALYSIS TIME WINDOW. THE 4D-VAR THEN COMPUTES THE BEST FIT BETWEEN THESE OBSERVATIONS AND THE MODEL FORECASTS. THE RESULTING ANALYSES SHOW A MUCH IMPROVED DYNAMICAL CONSISTENCY AND ARE ACCOMPANIED BY A SIGNIFICANT IMPROVEMENT IN THE GLOBAL SYSTEM PERFORMANCE. THE TEMPORAL DATA SELECTION HAS BEEN EXTENDED SO THAT THE 4D-VAR WILL MAKE USE OF MUCH MORE DATA THAN THE 3D-VAR, ALTHOUGH WE ARE NOT ADDING ANY NEW TYPE OF DATA WITH THIS IMPLEMENTATION. INDEED, A FAIR AMOUNT OF ASYNOPTIC DATA THAT COULD NOT BE ASSIMILATED PROPERLY BY THE 3D-VAR ANALYSIS SYSTEM WILL NOW BE USED BY THE 4D-VAR (ACTUALLY POTENTIALLY AS MANY MORE OBSERVATIONS AS THERE ARE MODEL TIME STEPS DURING THE ASSIMILATION WINDOW). FOR EXAMPLE, FOR A TYPICAL 6-HOUR ASSIMILATION TIME WINDOW, THE AMOUNT OF AIRCRAFT DATA ASSIMILATED IS INCREASED BY A FACTOR OF 3. IT SHOULD BE NOTED THAT THERE IS NO CHANGE TO THE FORECAST MODEL (TIME STEP, DYNAMICS, PHYSICS NOR RESOLUTION), OR TO THE DATA QUALITY CONTROL PROCESSES. THE 4D-VAR WILL BE USED IN THE GLOBAL ASSIMILATION SYSTEM ONLY. THE REGIONAL SYSTEM WILL CONTINUE TO USE THE 3D-VAR SYSTEM AS IT CURRENTLY DOES. THERE WILL HOWEVER BE SOME IMPACT ON THE REGIONAL SYSTEM PERFORMANCE DUE TO THE USE OF THE GLOBAL ANALYSES AT THE BEGINNING OF THE 12-HOUR REGIONAL SPIN-UP CYCLE. WORK IS UNDERWAY TO DEVELOP A 4D-VAR SCHEME FOR THE REGIONAL SYSTEM BUT A SIGNIFICANT AMOUNT OF WORK REMAINS TO BE DONE BEFORE IT CAN BE IMPLEMENTED. ADDITIONAL TECHNICAL INFORMATION ABOUT THE 4D-VAR AND THE ASSIMILATION PROCEDURES WILL BE FOUND ON THE MSC WEB SITE (HTTP://WWW.SMC-MSC.EC.GC.CA/CMC/OP_SYSTEMS/RECENT_E.HTML). IT IS ALSO IMPORTANT TO NOTE THAT DUE TO THE LARGER AMOUNT OF COMPUTER TIME REQUIRED TO RUN THE 4D-VAR, THE AVAILABILITY OF THE GLOBAL SUITE OF PRODUCTS WILL BE DELAYED BY ABOUT 20 MINUTES. PERFORMANCE EVALUATION THIS NEW 4D-VAR DATA ASSIMILATION SYSTEM HAS BEEN TESTED ON A LARGE NUMBER OF CASES. DATA ASSIMILATION CYCLES FOR A 2 MONTH WINTER PERIOD (11 DECEMBER 2003 TO 11 FEBRUARY 2004) AND A 2 MONTH SUMMER PERIOD (15 JULY TO 15 SEPTEMBER 2004) HAVE BEEN PERFORMED. THE NEW SYSTEM HAS ALSO BEEN RUNNING IN PARALLEL SINCE EARLY DECEMBER 2004. ANALYSES AND FORECASTS FROM THE PARALLEL GLOBAL SYSTEM HAVE BEEN VERIFIED AGAINST RADIOSONDE DATA OVER SEVERAL OBSERVATIONAL NETWORKS AS WELL AS AGAINST ANALYSES, AND EVALUATED BY CMC OPERATIONAL METEOROLOGISTS. THE IMPACT ON THE REGIONAL SYSTEM HAS BEEN EVALUATED USING SEVERAL CASES AND IN A PARALLEL RUN SINCE MID-JANUARY 2004. OVERALL, THE MASS AND WIND FIELD FORECASTS ARE IMPROVED, MOSTLY FOR THE MEDIUM RANGE FORECASTS. HOWEVER THE IMPACTS ON STATISTICAL GUIDANCE FOR BOTH GLOBAL AND REGIONAL SYSTEMS ARE QUITE SMALL, AND THEREFORE THERE IS LITTLE IMPACT ON SCRIBE. ASSIMILATION AS MENTIONED ABOVE, THE 4D-VAR MAKES A MUCH BETTER USAGE OF ALL AVAILABLE DATA. AS A CONSEQUENCE, THE 4D-VAR ANALYSES DO NOT FIT THE CONVENTIONAL RADIOSONDE DATA AS CLOSELY AS THE 3D-VAR ANALYSES DO. THIS REFLECTS THE MUCH LARGER USE OF ASYNOPTIC AND SATELLITE OBSERVATIONS IN 4D-VAR. IN ADDITION, THE PERCENTAGE OF DATA REJECTED BY THE QUALITY CONTROL (FOR ALL TYPES) IS SUBSTANTIALLY SMALLER IN THE 4D-VAR SYSTEM. THE QUALITY OF THE FIRST GUESS FIELDS IS ALSO SIGNIFICANTLY IMPROVED, AS SHOWN BY A MUCH BETTER FIT OF THE FIRST GUESS TO ALL DATA TYPES, FOR ALL NETWORKS. CMC OPERATIONAL METEOROLOGISTS DID NOTICE SIGNIFICANT DIFFERENCES WITH THE JET STREAM AND MASS FIELD ESPECIALLY OVER THE PACIFIC OCEAN. THE LARGER DIFFERENCES BETWEEN THE 4D-VAR ANALYSIS AND THE CONVENTIONAL RADIOSONDE DATA WERE NOTICED, BUT WERE OFTEN SUBTLE AND DIFFICULT TO SUBJECTIVELY EVALUATE. GLOBAL MODEL FORECASTS AS IT IS OFTEN THE CASE WITH MODIFICATIONS TO THE DATA ASSIMILATION SYSTEM, THE IMPACT ON THE RESULTING MODEL FORECASTS IS A VERY GOOD INDICATION OF THE IMPROVEMENTS IN THE ANALYSES. WITH THE 4D-VAR, THIS IMPACT IS IMPRESSIVE. OBJECTIVE VERIFICATION HAS SHOWN MARKED IMPROVEMENTS IN THE MASS AND WIND FIELDS, OVER ALL REGIONS BUT MORE PARTICULARLY OVER NORTH AMERICA AND THE SOUTHERN HEMISPHERE. THE IMPROVEMENT WAS LARGER AT DAY 3 BUT WAS NOTICEABLE UP TO DAY 10. THE WEST COAST WAS THE AREA SHOWING THE MOST IMPROVEMENTS OVER CANADA. THIS IS AN AREA WHERE THERE ARE LESS CONVENTIONAL OBSERVATIONS UPSTREAM AND WHERE THE BETTER USAGE OF SATELLITE DATA CAN CONTRIBUTE THE MOST. ANOTHER ASPECT NOTED IN THE VERIFICATION IS A REDUCTION IN THE NUMBER OF BUST FORECASTS. VERIFICATION OF QPF INDICATED A SLIGHT A SLIGHT IMPROVEMENT IN THE 4D-VAR BEYOND 48 HOURS. VERIFICATION OF WEATHER ELEMENT FORECASTS, INCLUDING THOSE USED BY SCRIBE, WAS ALSO CARRIED OUT BUT DID NOT SHOW ANY SIGNIFICANT DIFFERENCE UP TO DAY 5. CMC OPERATIONAL METEOROLOGISTS COMPARED THE TWO MODELS UP TO 120 HOUR FORECAST DURING THE PARALLEL RUN PERIOD. OVERALL, THE 500MB HEIGHT AND MEAN SEA LEVEL PRESSURE WERE FOUND EQUIVALENT IN NEARLY 60 PERCENT OF CASES AT 48-HOUR WHILE IN NEARLY 40 PERCENT OF CASES AT 120 HOUR FORECAST. THE DIFFERENCES BETWEEN THE TWO FORECAST SYSTEMS WERE WERE GROWING SIGNIFICANTLY BEYOND 48-HOUR FORECAST EXCEPT OVER THE ARCTIC REGION. THE LARGEST IMPROVEMENTS WERE FOUND OVER THE PACIFIC AND DECREASING TOWARDS THE ATLANTIC. THE 4D-VAR SYSTEM WAS PREFERRED 2 TIMES OUT OF 3 CASES SHOWING DIFFERENCES OVER THE PACIFIC REGION (INCLUDING WATERS) WHILE IT HAD ONLY AN EDGE OVER THE 3D-VAR OVER THE ATLANTIC REGION (INCLUDING WATERS). THIS SMALL ADVANTAGE OF THE 4D-VAR IS HOWEVER MORE SYSTEMATIC BEYOND 72 HOURS OVER THE ATLANTIC. FOR 24-HOUR PRECIPITATION, THE 4D-VAR SYSTEM WAS MUCH PREFERRED OVER THE PACIFIC REGION ALREADY AT DAY-2, AND ESPECIALLY AT DAY-3. RESULTS ARE MIXED AT DAY 2 ELSEWHERE, BUT CLEAR IMPROVEMENTS WERE FOUND WITH THE 4D-VAR SYSTEM BEYOND 72 HOURS OVER CENTRAL NORTH AMERICA AS WELL AS OVER THE ATLANTIC. OVERALL, THE CMC OPERATIONAL METEOROLOGISTS JUDGED THE PARALLEL CONFIGURATION MUCH BETTER. REGIONAL MODEL FORECASTS ALTHOUGH THE REGIONAL SPIN-UP SYSTEM STILL USES THE 3D-VAR, THERE IS NONETHELESS SOME IMPACT AS THE SPIN-UP DOES MAKE USE OF THE IMPROVED GLOBAL MODEL FIELDS AT THE BEGINNING OF THE 12 HOUR SPIN-UP CYCLE. VERIFICATION DURING THE PARALLEL RUN HAS SHOWN A SLIGHT GAIN, MOSTLY FOR 24 AND 48 HOUR FORECASTS OF WIND AND MASS FIELDS. THE IMPACT IS LARGER OVER WESTERN NORTH AMERICA BUT CAN ALSO BE SEEN OVER THE ENTIRE NORTH-AMERICAN NETWORK. VERIFICATION OF QPF INDICATED A VERY SLIGHT IMPROVEMENT FOR CLASSES BELOW 20 MM. VERIFICATION OF WEATHER ELEMENTS USED BY SCRIBE HAS SHOWN A SLIGHT IMPROVEMENT FOR THE PARALLEL, MOSTLY FOR WIND, POP AND MAX TEMPERATURE. THE COMPARATIVE EVALUATION INDICATES VERY SMALL IMPROVEMENTS TO THE REGIONAL FORECAST SYSTEM FOR ALL REGIONS. OVERALL THE TWO SYSTEMS WERE RATED EQUIVALENT IN 3 CASES OUT OF 4. THE MOST SIGNIFICANT SIGNIFICANT IMPROVEMENT IS NOTED OVER CENTRAL CONTINENT WITH THE MEAN MEAN SEA LEVEL PRESSURE ESPECIALLY AT 48 HOUR. AGAIN THE IMPROVEMENT IS FOUND OVER THE CENTRAL CONTINENT AND WEST BUT MIXED RESULTS ARE OBTAINED OVER THE EAST COAST WHEN CONSIDERING THE MASS FIELD. FOR 24-HOUR PRECIPITATION, THE DIFFERENCES WERE GENERALLY VERY SMALL, BUT IMPROVEMENTS WERE NOTED FOR DAY-1 OVER THE WEST COAST WHILE, SURPRISINGLY, THE PARALLEL SYSTEM DAY-2 FORECAST WAS PREFERRED IN 20 PERCENT OF CASES AGAINST 10 PERCENT OVER THE EAST COAST. EVERELL / ADMA / SMA-SMC TORONTO