HPC Winter Weather Desk Operations and 2011 and 2012 Winter Weather Experiments Dan Petersen Winter weather focal point with contributions from Keith Brill, David Novak, and Mike Musher

Presentation Goals Recent changes in WWD product suite HPC WWD verification 2011/2012 HPC Winter Weather Experiments

WWD Changes Expand Probabilistic Snow and Ice Forecast Suite Combine HPC deterministic forecast with model/ensemble forecasts to derive forecasts for probabilistic snowfall (1, 2, 4, 6, 8, 12, 18”) and ice (freezing rain 0.01, 0.10, 0.25, 0.5”) accumulations

WWD Changes WWD accumulation graphics on our internal page are now available in 2 GIS formats, KML (for Google Earth) and shapefiles Current forecast and archive of prior forecasts are now available at

WWD Changes 48 hour snowfall and freezing rain accumulation probabilities Snowfall probabilities Freezing rain probabilities

Changes WWD Changes UKMET added to the accumulation blender as a QPF option

HPC Winter Weather Desk Verification

HPC Deterministic Snowfall Forecast East of the Rockies Automated Superensemble: 0.8*(SREF members, GFS, NAM, ECMWF, ECMWF mean,CMC) + 0.2*(GEFS mean) All 4” and 8” metrics were better than last year. All 12” metrics were worse than last year. Day 1 continues to be difficult to improve upon (right)

HPC Deterministic Snowfall Forecast WWD forecasts have a high bias, especially for 12”(ideal bias is 1, less than one would be a low bias) East of the Rockies

Verification of Day 2 forecast for February 2, 2011: Final Human Forecast Chosen Blend Auto SuperensembleObserved

HPC Experimental Probabilistic Snowfall Including human deterministic forecast in calculation of product makes positive improvement at all thresholds (Good!!) Probabilistic skill for thresholds < 6” Difficult to improve upon sample climatology above 6” -HPC is mode -Spread from 21 SREF+GFS+NAM+CMC

HPC Legacy Probabilistic Snowfall Degradation from last year Benchmark likely improved Increased temporal resolution of Snow-liquid Ratios Human have too high of probabilities, especially for larger thresholds Moderate and High risk areas are either too big and/or miss location

Day 1 WWD Verifying frequency 4,8,12” events forecast more frequently than observed, therefore high bias and overconfident forecasts HPC Legacy Probabilistic Snowfall

High bias Low bias Day 3 WWD Verifying frequency 4,8,12” events forecast more frequently than observed, therefore high bias and overconfident forecasts

Verification of HPC low tracks (position at each forecast hour) Lowest errors: Blend of the gfs+ecmwf 36-72hr Among the models: UKMET/ECMWF best >48 h

16 Summary-Verification Winter Weather Desk (WWD) provides skillful deterministic snowfall forecasts WWD categorical probabilistic forecast edits did not improve upon most accumulation thresholds, so fewer edits will be done this year to the automated multi- model/multi-ensemble probabilities Including the human’s deterministic forecast improves the continuous probability forecasts

Freezing rain continuous probability forecasts Expansion of continuous probabilistic snow and ice forecasts to include 48 hour totals (event perspective) UKMET has been added to suite of available models in forecast Heavy snow continuous probability forecasts Summary plans and references QPF archive Gridded continuous probability forecasts can be obtained at

HMT-HPC Winter Weather Experiment 14 participants (WFO, SPC, AWC, HPC, EMC, ESRL, and COMET) Can high-resolution models improve Day 1 forecasts of precipitation type and amount? Can we better quantify and communicate winter weather uncertainty for Day 1-2 forecasts? Jan 10 – Feb 11, 2011 Goals Participants Approach Create snow and ice accumulation forecasts using experimental guidance Write “confidence discussion” Subjectively evaluate experimental guidance

Winter Weather Experiment Jan 10 – Feb 11, 2011 Day 1Day 2 HRW-ARWx HRW-NMMx NMMB nestxx NAM Rime Factorxx ECMWF snowxx SREFxx HPC Superensemble xx Experimental GuidanceActivities 24 hr accumulated snow and ice forecasts for Day 1&2 Forecast Confidence Discussion Model evaluation of precipitation type and amounts Full detailed report with recommendations at:

Winter Weather Experiment Results HRW-NMMObservedHRW-ARWExperimental NAM Can high-resolution models improve Day 1 forecasts of precipitation type and amount? Pros: Improved orographic precipitation, lake effect Visualizing temporal evolutions Providing unique fields (simulated reflectivity). Cons: Overall amounts not superior to operational

Analysis courtesy WFO Tulsa OK New Oklahoma All-Time Record 24-hour Snowfall: 27" in Spavinaw, OK February 9, 2011

Comparison of analyzed 24 hr snowfall vs model 36 hr forecasts from 00z 09 Feb 2011 Analysis and forecast ending time 00z 10 Feb 2011 Observed snowfall12km NAM forecast HPC superensemble ECMWF forecast

Comparison of analyzed 24 hr snowfall vs model 36 hr forecasts from 00z 09 Feb 2011 Observed snowfall WRF NMM forecast New NAM forecast WRF ARW forecast Analysis and forecast ending time 00z 10 Feb 2011

2012 Experiment Plans Experiment Questions What added benefit will storm scale ensembles provide for winter weather forecasting? Is the new SREF system an improvement over the operational system? How can forecasters add value to probabilistic winter weather forecasts? How can ensemble clusters be used in the forecast process? How can awareness of societal impacts improve forecast and decision support services? Jan 9 – Feb

2012 Experiment Plans ProviderDatasetProjectionNotes EMCParallel SREF (20-26 km) 87 hPre-implementation version of the new SREF AFWA10 member 4 km storm scale ensemble 54 hCentral and Eastern U.S. domain NOAA HMT West9 km 9 member ensemble 72 hCalifornia/Nevada domain HPCSuperensemble (SREF+ deterministic runs) 84 hBENCHMARK EMCOperational SREF87 hBENCHMARK Datasets Jan 9 – Feb

Questions or comments? HPC Forecast Operations desk (301)

18z 26 Jan 2011 sea level fronts/pressures 06z 27 Jan 2011 sea level fronts/pressures 00z 27 Jan 2011 sea level fronts/pressures 12z 27 Jan 2011 sea level fronts/pressures

Comparison of analyzed 24 hr snowfall vs model 36 hr forecasts from 12z 27 Jan 2011 Analysis and forecast ending time 00z 28 Jan 2011 Observed snowfall WRF ARW forecast New NAM forecast WRF NMM forecast

Observed snowfall 12km NAM forecast ECMWF forecastHPC superensemble Comparison of analyzed 24 hr snowfall vs model 36 hr forecasts from 12z 27 Jan 2011 Analysis and forecast ending time 00z 28 Jan 2011

27-28 Jan 2011 Event analysis from WFO New York,NY

WFO input used to modify HPC public snow/ice probabilistic forecasts HPC input used to modify grids within GFE to produce local forecast Results in final collaborated forecast HPC/WFO Collaboration via 12 planet, phone, and/or event conference calls. WWD Collaborative Forecast Process

Collaboration Details season summary 224 questions from WFOs via 12Planet 22 telephone calls from WFOs 5 Conference Calls WRSRCRER #12 Planet inquiries # phone calls04216 #conference calls0131