1 / 55

METEOROLGICAL FACTS OF LIFE Dealing With Uncertainties in Weather Forecasts

METEOROLGICAL FACTS OF LIFE Dealing With Uncertainties in Weather Forecasts. (Chair, DC-AMS). * Formerly: Program Officer, Office of Naval Research (2002-2006) Research Meteorologist, NWS/NCEP (1975- 2002).

merle
Download Presentation

METEOROLGICAL FACTS OF LIFE Dealing With Uncertainties in Weather Forecasts

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. METEOROLGICAL FACTS OF LIFE Dealing With Uncertainties in Weather Forecasts (Chair, DC-AMS) *Formerly: Program Officer, Office of Naval Research (2002-2006) Research Meteorologist, NWS/NCEP (1975- 2002)

  2. The Butterfly Effect: sensitive dependence on initial conditions in chaos theory where small variations of the initial condition of a nonlinear dynamical system produce increasingly large variations in the long term behavior of the system.  The single flap of a butterfly’s wings change the initial conditions of the system enough to cause large-scale phenomena (hurricanes and such) since any variation in the initial conditions is vastly magnified with each iteration. And every flap of every butterfly wing in the world continually changes those conditions

  3. Lorenz Model More Predictable - Quasi deterministic • Variations in predictability can be illustrated using the Lorenz (1963) model: Predictable at first Initial state Unpredictable Simple non-linear system. Possible atmospheric analogue: Zonal Flow Blocked Flow

  4. BUTTERFLY EFFECT (CHAOS!!) MODERATE CONFIDENCE UNPREDICTABLE HIGH CONFIDENCE LOW CONFIDENCE time

  5. “He believed in the primacy of doubt (uncertainty) not as a blemish upon our ability to know, but as the essence of knowing” Gleick (1992) on Richard Feynman’s philosophy of science.

  6. LIMITS IN WEATHER FORECASTING • Small errors (uncertainties) in initial conditions AND imperfect models can amplify rapidly => Forecasts lose skill with increasing lead time • Initial state is imperfect • Problems with observations and data coverage • Problems with assimilating the data • Imperfect statistical and numerical forecast methods • Random (and systematic) errors • Numerical model is imperfect • Limited resolution • Processes represented in model must be truncated • Spatially • Temporally • Physically • Systematic (and random) errors

  7. When Docs (Weather Forecasters) Are in Doubt**: (Adapted from an article in Newsweek (http://www.newsweek.com/id/78155) by the wife of a doctor concerning the uncertainties doctors face in their profession. With only minor wordsmithing, as highlighted below, we see that our concerns parallel those of other professions, in this case, medicine.) Few people know better than I do (author of article) that doctors (weather forecasters) are imperfect beings: Doctors (weather forecasters) too, are prone to the occasional error; they don't always have all the answers. A few don't like to admit that. They'd rather seem omniscient or at least supremely confident, in line with the old surgical (meteorological) saying, "Sometimes wrong, never in doubt." But lately they're in the minority. More of the doctors (weather forecasters) I encounter now want to talk about their doubts (uncertainties) and mistakes (busted forecasts), not paper over them. They are willing, and even eager, to be seen as human. They're right in tune with popular culture, which has replaced the image of the all-knowinghealer (forecaster) with that of the highly educated guesser. Doctors (forecasters) …. don't always know what's wrong with their patients (users) or what to do in response. They grapple with uncertainty. Increasingly, they're sharing that fact with their patients (users), because that's what patients (users) say they want to hear. …. Two weeks ago I had to look into the black box myself. My mother was in the hospital (on vacation) , and no one knew exactly what was making her sick (what the weather would be). ……The docs (forecasters) did everything by the best-selling books (models) : they admitted their doubts

  8. 00 hour

  9. 5.5 day

  10. 10 day SPAGHETTI CHART

  11. What a mess!!

  12. 10 day WHICH ONE IS CORRECT??

  13. SCREAMING MESSAGE: Weather Forecasts Will ALWAYS Be Coupled With Varying Degrees of Uncertainty Varying EFFECT!!!

  14. FORECAST PROCESS IS INHERENTLY STOCHASTIC (PROBABILISTIC) IN NATURE INCORPORATING UNCERTAINTIES INTO FORECASTS ENHANCES THEIR VALUE • NEED TO DO SO IN READILY COMPREHENSIBLE AND RELEVANT TERMS

  15. YOUR FORECAST HAS A 30% CHANCE OF BEING 70% CORRECT

  16. FORECAST PROCESS IS INHERENTLY STOCHASTIC (PROBABILISTIC) IN NATURE INCORPORATING UNCERTAINTIES INTO FORECASTS ENHANCES THEIR VALUE • NEED TO DO SO IN READILY COMPREHENSIBLE AND RELEVANT TERMS • EDUCATION OF PROVIDERS AND USERS • ESSENTIAL

  17. An analysis produced to run an NWP model is somewhere in a cloud of likely states. Any point in the cloud is equally likely to be the truth. 48h forecast T T T T T 24h forecast 12h forecast 36h forecast 12h verification 24h verification The true state of the atmosphere exists as a single point in phase space that we never know exactly. 36h verification 48h verification P H A S S P E A C E Deterministic Forecasting Limitations Error AC = 0 Nonlinear error growth and model deficiencies drive apart the forecast and true trajectories (i.e., Chaos Theory) A point in phase space completely describes an instantaneous state of the atmosphere. (pres, temp, etc. at all points at one time.)

  18. LIMITS OF PREDICTABILITY: Deterministic Forecast Forecast Uncertainty Initial Condition Uncertainty X Analysis Climatology Verif time PREDICTAILITY LIMIT WHEN ERROR VARIANCE  CLIMO VARIANCE

  19. Uncertainties arise also because models are imperfect!! Same Deterministic Model with Different Convection Schemes Results In Different Precipitation Forecasts

  20. INFO on DISTRIBUTION of SCENARIOS: - how many scenarios - how likely is each - how sharply defined is each Initial uncertainty = distribution of possible initial states Forecast uncertainty = distribution of possible forecast states Single deterministic forecast Ensemble members verification

  21. F1 F2 F3 F4 . . FN T = chance of critical value being exceeded  aid in making decisions, risk analysis Fcr Models 100% P(F) T 0% Q (e.g., T< 0o, Sn>4”) F1, F2 … FN = predicted variables of interest, for example, wind speed Fcr = user-specified “critical value”

  22. SSSS Decision-making with probabilities B Rational decision-making depends on the user’s sensitivity – illustrate with how we respond to low probabilities: • 5% risk that a plane will crash - would you board it? • 5% risk of rain – would you play golf? Decisions must be based on user’s Cost/Loss ratio • users with low C/L should protect at low probabilities

  23. USER REQUIREMENTS:PROBABILISTIC FORECAST INFORMATION IS CRITICAL

  24. GENERAL IDEA!!!: JAN 9, 2002: “SURPRISE” ICE STORM ISSUE: No advance warning in actuality How to deal with uncertainty - 50/50 chance of freezing rain during morning rush ~ 12hr in advance CONSIDER: Cost of not taking action (staging sand trucks) and event happens (as was the case THIS time) versus Taking action and event doesn’t occur

  25. Prob of Frzng rain: 15 hr fcst from 21Z 01/08/02, vt 12Z 01/09/02 Prob of Frzng rain: 18hr fcst from 21Z 01/08/02, vt 15Z 01/09/02

  26. The Associated Press – Jan 9, 2002 "Surprise Icy Conditions Claim Teenager's Life" Washington Post Wednesday, January 9, 2002 "This morning's weather evolved without warning. ...." Weather forecasters said … had little advance notice of the rain … first indications of the light rain appeared on radar about 4AM..” (just before morning rush) WJLA Viewer Opinions and Feedback “I think its a real shame the NWS can't prepare us for this weather. We seem to have spent thousands of $$$ on this fancy weather reporting equipment so they can forecast weather and guess what??

  27. Jan 25-26, 2000 DC “Surprise” Snowstorm

  28. MAJOR SNOWSTORM AMBUSHES WASHINGTON Not Good- especially when effecting DC (just after announce-ment of new Super Computer by NWSHQ

  29. Wash Post Interview – Jan 26, 2000 Warren Washington (NCAR) said, “It (computer forecast) wasn’t as accurate as it should have been, but forecasters can’t always account for every ‘chaos aspect’, little changes in weather that can cause major shifts. The public should not expect a perfect forecast” Reporter: “ OK, OK, WHAT THE HELL should we expect when they (NWS) just paid $35 million for the (new computer) system” “In my next life I want to come back as a weather- man. That way I can be dead wrong 80% of the time and not get fired” “Models? Next time, read pig entrails.” Tony Kornheiser (Wash Post) What’s wrong with the forecast models? (Earth Observatory) “Uccellini admits that there was some uncertainty in the minds of meteorologists as to what path the storm would take. He says the NWS could have emphasized that uncertainty more in its forecasts. Forecasters know the models aren’t perfect. There should be some way to present alternate scenarios without following the mainstream (models) and creating a panic (i.e., ‘crying wolf’)."  Impetus to SREF and Winter Weather Exp

  30. Bottom Line “Deterministic forecasting is not healthy”! CAUTION Louis U.

  31. EVEN EXPERTS CAN HAVE A BAD DAY

  32. SCHEMATIC: BASIC PROBLEM WITH FORECASTS Figure shows the difference between the storm’s observed behavior and the forecast made 24 hours before. The storm traveled much closer to shore than was predicted, and dropped a lot more snow.

  33. Ensemble provides a clear “heads up” on morning of 24th for the possibility of a major snow event, especially when considered in context of independent information from satellite imagery and radar that suggested storm track closer to coast and precip further inland than available operational models were indicating

  34. Wide range of solutions (CTL vs Best) in precip and Storm Track (next) => Deterministic (yes/no forecast) very risky!

  35. TO THE VIDEO TAPE, PLEASE !!! Bob Ryan BOB’S ORIGINAL BROADCAST EVENING BEFORE JAN 25 “SURPRISE” SNOWSTORM, FOLLOWED BY HYPOTHETICAL BROADCAST HE MIGHT HAVE GIVEN INCORPORATING INFORMATION ON UNCERTAINTIES

  36. CASES Dec 29-30, 2000 DC surprise NON Snowstorm (The Millennium Snowstorm elsewhere)

  37. NWS FOR DC/BALT9PM DEC 28 (Thurs): …WINTER STORM WATCH FOR FRIDAY NIGHT THROUGH SATURDAY... SNOW WILL BEGIN LATE FRIDAY EVENING AND INTENSIFY TOWARD DAYBREAK. SNOWFALL MAY BE HEAVY AT TIMES SATURDAY MORNING INTO THE AFTERNOON. BRISK NORTH WINDS WILL CAUSE BLOWING AND DRIFTING OF SNOW. SNOWFALL ACCUMULATIONS OF 4 TO 8 INCHES ARE POSSIBLE. 10PM DEC 29 (Friday): …WINTER STORM WARNING FOR LATE TONIGHT THROUGH SAT EVENING ….5-10” EXPECTED FROM THIS STORM…

  38. Dec 29: "It's going to be ugly," … Air Lines spokesman "We're going to be taking down a significant portion of our schedule throughout the Northeast." because of predictions of heavy snowfall CNN: “In Washington, more than 300 snowplows and salt trucks were ready to go into action Saturday, according to city officials.” ___________________ Dec 30 “ …Washington got no snow and flights were expected to resume at 6 p.m., an airline spokeswoman said CNN: “The storm unexpectedly spared Washington, hit Philadelphia less hard than expected …. Forecasters said.

  39. WHAT DC MISSED!!

  40. WHAT DC GOT!

  41. SREF 24hr spaghetti from 12Z 29 Dec for .50” 12 hr precip ending 12 GMT 30 Ensemble indicated a 30-40% chance of signif- icant snow; thus, in this case SREF gave a “heads up” (60%) for the chance of no snowstorm

  42. KEY POINTS January 24/25, 2000 DC Snowstorm: Ensembles gave “heads up” for snowstorm in face of deterministic model and official forecasts of no snow December 30, 2000 DC Non-Snowstorm: Ensemble gave “heads up” of no snow in face of deterministic model (Eta) predicted and official forecasts of snowstorm

  43. PROBABILITY CHARTS SREF Percentage of members with QPF > .25”/24h 010519/0000V63 SREFX-CMB; 24HR PQPF OF .25”

  44. SREF Combined or Joint Probability Probability of convection in high CAPE, high shear environment (favorable for supercells) Pr [MUCAPE > 2000 J/kg] X Pr [ESHR > 40 kts] X Pr [C03I > 0.01”]

  45. Risk Identification: Early warnings (2) Strike probability (within 65 nm) of Typhoon Rusa over the next 120 hours. Starting time of the forecast is 27 August 2002 12 UTC. Full dots give the observed position over the period 27 August to 1 September 2002 GLOBAL DATA PROCESSING AND FORECASTING SYSTEMS

More Related