Fleet Weather Center Norfolk Assessing the value of ensemble guidance for FWC-Norfolk forecasters Presented to the NUOPC Users Workshop This briefing is UNCLASSIFIED / FOUO June 2011 LT Rich Rainer N5 Department Head Forecast Team AGC Brown AG1 Cummings AG2 Nelson AG2 Didier AG3 Dietrich AGAN Gagle Fleet Weather Center - Norfolk AGENDA • Objectives • Challenge • Description of experiment • General results • Data collection • Objective results • Subjective findings • Possible next steps Fleet Weather Center - Norfolk 2 Objectives • In an environment where resources are scarce and billets are being reduced to “optimal manning” levels – How do we maintain forecast quality? – How to we increase forecaster efficiency (reduce the amount of time necessary to make quality forecasts)? Fleet Weather Center - Norfolk 3 Challenge • Fleet Weather Center forecasters have a heavy workload – 2 forecasters – 2 techs – Forecasts for 18 OPAREAS – Forecasts for 20-50 ships – Wind and seas warnings for the • North Atlantic • Mediterranean • Black Sea • South Atlantic Fleet Weather Center - Norfolk 4 Description of experiment • SGOT shadow forecast team – Two forecasters, two assistant forecasters – March 15, 2011 through May 15, 2011 – FOXX1 plus North Atlantic high winds and seas – In addition to traditional guidance products, use NOGAPS ensemble guidance in MetCast/JMV • Mean/spread plots • Probability of exceedance for winds (35kts) and seas (12’ and 18’) – Compete with official forecasts from Watch Floor – Develop best practices for use of ensemble guidance • Chose to use ensemble as primary source of guidance – Validate using satellite, ship, and buoy observations Fleet Weather Center - Norfolk 5 OPAREA Forecasts • FOXX01 – VACAPES • WEAX Parameters - Winds and - Direction - Speed - Seas - Height Fleet Weather Center - Norfolk 6 Example OPAREA forecast UNCLASSIFIED FOXX01 KNGU 041200 MSGID/GENADMIN/FLEWEACEN NORFOLK VA// SUBJ/VACAPES/CHERRY PT/CHARLESTON/JACKSONVILLE/ PORT CANAVERAL/TONGUE OF THE OCEAN OPAREAS FORECAST// POC/COMMAND DUTY OFFICER/-/FLEWEACEN NORFOLK/LOC:NORFOLK VA// TEL:757-444-7750/SECTEL:-0963/EMAIL: CDO.FWC.NRFK.FCT(AT)NAVY.MIL// RMKS/1.HIGH PRESSURE RIDGES IN FROM THE MIDWEST PRODUCING PARTLY CLOUDY SKIES, MODERATE WINDS AND SEAS THROUGHOUT THE AREA. EXPECT WINDS AND SEAS TO DECREASE AS THIS HIGH PRESSURE TRANSITS SOUTHEAST TONIGHT THROUGH FRIDAY AFTERNOON. 2. VACAPES OPAREA 24 HOUR FORECAST COMMENCING 050000Z. SKY, WEATHER: PARTLY CLOUDY. B. VSBY (NM): 7. C. SURFACE WIND (KTS):NORTHWEST 20 TO 25 GUST 30, DECREASING 13 TO 18 GUST 23 BY 05/18Z. D. COMBINED SEAS (FT):COASTAL:NORTHERLY 3 TO 5. SEAWARD:SOUTHERLY NORTHERLY 5 TO 7. E. MAX/MIN TEMPS (F): 68/52. F. SST(F/C): COASTAL: 55/13. VICINITY OF GULF STREAM: 74/22. G. OUTLOOK TO 48 HOURS: WINDS: NORTHWEST 13 TO 18 GUST 23, DECREASING 10 TO 15 GUST 20 BY 06/06Z, BACKING WEST-NORTHWEST BY 06/12Z, BECOMING WEST 5 TO 10 BY 06/18Z. SEAS:COASTAL:NORTHERLY 3 TO 5, ABATING 2 TO 4 BY 06/06Z. SEAWARD: NORTHERLY 5 TO 7, ABATING 4 TO 6 BY 06/18Z. 3.CHERRY PT OPAREA 24 HOUR FORECAST COMMENCING 050000Z. A. SKY, WEATHER: PARTLY CLOUDY. B. VSBY (NM): 7. C. SURFACE WIND (KTS): NORTHWEST 18 TO 23 GUST 28, DECREASING 15 TO 20 BY 05/06Z, DECREASING 13 TO 18 BY 05/18Z. COMBINED SEAS (FT):COASTAL: NORTH 2 TO 4. SEAWARD:NORTH 5 TO 7, ABATING 4 TO 6 BY 05/18Z. E. MAX/MIN TEMPS (F): 72/54. F. SST(F/C): COASTAL: 56/13. VICINITY OF GULF STREAM: 72/22. G. OUTLOOK TO 48 HOURS: WINDS:NORTHWEST 13 TO 18, DECREASING 8 TO 13 BY 06/06Z, BECOMING WEST 5 TO 10 BY 06/12Z, BACKING SOUTHWEST BY 06/18Z. SEAS:COASTAL: NORTH 2 TO 4, BECOMING CONFUSED BY 06/12Z. SEAWARD: NORTH 4 TO 6, BECOMING NORTHWEST 3 TO 5, BY 06/12Z. 4. CHARLESTON OPAREA 24 HOUR FORECAST COMMENCING 050000Z. A. SKY, WEATHER: PARTLY CLOUDY. B. VSBY (NM): 7. C. SURFACE WIND (KTS):NORTHWEST 15 TO 20, BECOMING NORTH 10 TO 15 BY 05/18Z. Fleet Weather Center - Norfolk 7 OPAREA validation C. SURFACE WIND (KTS):NORTHWEST 20 TO 25 GUST 30, DECREASING 13 TO 18 GUST 23 BY 05/18Z. D. COMBINED SEAS (FT):COASTAL:NORTHERLY 3 TO 5. • Collect all available observations in OPAREA • Compute average observation value • Calculate distance outside of forecast range • Example: Average observation was 7ft seas, forecast error was +2ft. Fleet Weather Center - Norfolk 8 Example wind and seas warning HIGHLIGHTED CONTOURS: WINDS >35KTS & >50KTS; Fleet Weather - Norfolk SEASCenter >12FT & >18FT SGOT Shadow in Black Watch Floor in Yellow 9 General results • VACAPES forecasts made with the help of ensemble guidance outperformed official VACAPES forecasts • No quantitative assessment of high winds and seas forecasts (yet), but sense is that they are comparable and quicker to generate • Uncertainty information is welcome, but slows down the forecast process when large uncertainty is present – Need for products that clarify the “form” of the uncertainty and the reason for the uncertainty Fleet Weather Center - Norfolk 10 Data collection Fleet Weather Center - Norfolk 11 Objective results (24hr-48hr) VACAPES wind direction Bias RMSE % outside threshold Prob. guidance 48.7deg 16.3 (39) Det. guidance 57.1deg 23.8 (57) 14.8% 31.6% Percent improvement VACAPES wind speed Bias RMSE % outside threshold Prob. guidance 2.3kts 4.6kts 6.9 (16) Det. guidance 3.1kts 5.4kts 10.8 (25) 25% 13.8% 36% Percent improvement VACAPES wave height Bias RMSE % outside threshold Prob. guidance -0.02ft 0.57ft 0.42 (1) Det. guidance 0.25ft 1.06ft 7.59 (18) Percent improvement 93.3% 46.7% 94.4% Fleet Weather Center - Norfolk 12 Objective results (48hr-72hr) VACAPES wind direction Bias RMSE % outside threshold Prob. guidance 54.4deg 24.1 (54) Det. guidance 61.2deg 26.3 (59) 11.2% 8.5% Percent improvement VACAPES wind speed Bias RMSE % outside threshold Prob. guidance 2.0kts 4.9kts 8.9 (20) Det. guidance 2.6kts 4.9kts 9.4 (21) 21% 1% 4.8% Percent improvement VACAPES wave height Bias RMSE % outside threshold -0.004ft 0.64ft 1.5 (3) Det. guidance 0.33ft 1.5ft 4.6 (14) Percent improvement 98.7% 48.4% 68.2% Prob. guidance Fleet Weather Center - Norfolk 13 24hr wind direction errors Wind direction, 24hr forecast error (deg) 200 180 160 Direction error (deg) 140 120 100 Shadow Watch Floor 80 60 40 20 0 Fleet Weather Center - Norfolk 14 Wind direction errors Wind direction threshold exceedance (%) 80 70 60 50 40 30 20 10 0 50 Shadow Shadow Watch Floor Exceedance (%) RMSE (deg) Wind direction RMSE (deg) Watch Floor 40 30 20 10 0 24 30 36 42 48 54 Fleet Weather Center - Norfolk 60 66 24 30 36 42 48 54 60 66 15 24hr wind speed errors Wind speed, 24hr forecast error (kts) 16 14 12 Speed error (kts) 10 8 6 Shadow Watch Floor 4 2 0 -2 -4 Fleet Weather Center - Norfolk 16 Wind speed errors Bias (kts) Wind speed bias (kts) 4 3.5 3 2.5 2 1.5 1 0.5 0 Shadow Watch Floor 24 30 36 42 48 Watch Floor 30 36 42 48 54 60 Fleet Weather Center - Norfolk 66 Exceedance (%) RMSE (kts) Shadow 24 60 66 Wind speed threshold exceedance (%) Wind speed RMSE (kts) 7 6 5 4 3 2 1 0 54 16 14 12 10 8 6 4 2 0 Shadow Watch Floor 24 30 36 42 48 54 60 66 17 24hr wave height forecast errors Wave height, 24hr forecast error (ft) 3 2 Height error (ft) 1 0 Shadow -1 Watch floor -2 -3 -4 Fleet Weather Center - Norfolk 18 Wave height forecast errors Wave height bias (ft) 0.6 Shadow 0.5 Watch Floor Bias (ft) 0.4 0.3 0.2 0.1 0 -0.1 -0.2 24 30 36 42 48 54 1.6 Shadow 9 Shadow Watch Floor 8 Watch Floor Percent exceeded RMSE (ft) 1.4 1.2 1 0.8 0.6 0.4 7 6 5 4 3 2 0.2 1 0 0 24 30 36 42 48 54 66 Wave height threshold exceedance (%) Wave height RMSE (ft) 1.8 60 60 Fleet Weather Center - Norfolk 66 24 30 36 42 48 54 60 66 19 Subjective wind/sea warnings findings • Shadow wind and seas warnings seem comparable to Watch Floor products. – Data exists for objective validation • Shadow wind and seas production was faster. Fleet Weather Center - Norfolk 20 Utility of ensemble guidance • Used ensemble as primary source of guidance. Used deterministic to validate. – Not obviously best way forward • Ensemble is great when uncertainty is low (typically associated with high pressure). • Ensemble slows the forecast process when uncertainty is high (especially for fast moving and deep low pressure systems). • Found it useful to apply color bar lower thresholds to cut down on overwhelming data. Fleet Weather Center - Norfolk 21 Needs • Ensembles raise questions. We need new products to tell us why the uncertainty looks the way it does – Sensitivity products • Need quick, detailed information about form of uncertainty. – E.g. plumes, skewness. • Guidance on how to define relevant uncertainty thresholds. • Uncertainty parameters that support all relevant mission areas. Fleet Weather Center - Norfolk 22 Personal takeaways… • How I would use this on the watch floor tomorrow: – Probability products for winds and seas – Use deterministic models as primary form of guidance and ensemble as means of identifying where additional attention is necessary • But… – Really want information about the form of the uncertainty and information about why the uncertainty is there Fleet Weather Center - Norfolk 23 Possible next steps • NPS thesis topics – Validate winds/seas – Conditional verification – Compare with objective guidance • Automated winds and seas warnings • Possible additional experiments – Alter CONOPS so that deterministic model is primary source of guidance and use ensemble secondarily – Inclusion of GFS ensemble – Experiment with products that communicate form of uncertainty and sensitivity information – Write 120hr forecast products Fleet Weather Center - Norfolk 24