1 / 47

A Forecast Evaluation Tool (FET) for CPC Operational Forecasts

A Forecast Evaluation Tool (FET) for CPC Operational Forecasts. Edward O’Lenic Chief, Operations Branch, NOAA-NWS-CPC. Outline. CPC operational products We have discovered users Users are diverse We need to know how users use products We need a new service paradigm

verda
Download Presentation

A Forecast Evaluation Tool (FET) for CPC Operational 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. A Forecast Evaluation Tool (FET) for CPC Operational Forecasts Edward O’Lenic Chief, Operations Branch, NOAA-NWS-CPC

  2. Outline • CPC operational products • We have discovered users • Users are diverse • We need to know how users use products • We need a new service paradigm • FET addresses these needs • Tour of FET • Current status • Implications

  3. Background • People and organizations need planning tools. • For decades, climate forecasts have been sought for use in planning. • The obvious signs of climate change have made this desire more urgent. • The uncertain nature of climate forecasts can lead the uninformed to make poor choices. • Users need an “honest broker” they can trust.

  4. Current Operational Capabilities • CPC products, include extreme events for days 3-14, extended-range forecasts of T, P for week 2, 1-month and 3-month outlooks, and ENSO forecasts. • These products have a scientific basis, have skill, and therefore ought to be useful, but - • We have a less-than adequate understanding of user requirements, or actual practical use. • CPC has neither the staff, nor the expertise to effectively pursue understanding these things.

  5. Challenges from Users Western Governor’s Association (Jones, 2007): - More accurate, finer-resolution long range forecasts - Continued and expanded funding for data collection, monitoring and prediction - Partnerships with federal and state climatologists, RCCs, agricultural extension services, resource management agencies, federal, state and local governments. USDA ARS Grazinglands Research Laboratory (Schneider, 2002) - Fewer “EC” forecasts - Better correspondence between F probability and O frequency - Forecast more useful than climatology - Forecasts of impacts, not meteorological variables

  6. Bridging Research, Operations, Users Know WHO stakeholders are and HOW they USE climate products, (relationships), LEARN from stakeholders WHAT we need to provide (iterative refinement of requirements, relationships), Implement USER-, and SCIENCE-VETTED products, (iterative refinement of requirements, relationships), Operationally support evolving user and producer REQUIREMENTS (extension function, relationships, operations, research) ? Use it if you can Assumed User Needs Basic data and Forecasts GOVT. PROVIDERS Basic Research

  7. Bridging Research, Operations, Users users ? Use it if you can Iterative Use refinement Assumed User Needs Intermediary Applications Products CTB-RISA/PRIVATE/RCC/SC-SBIR, NIDIS Decision-Support Development REQUIREMENTS Iterative Technical Refinement Basic data, Forecasts Transfer User-Vetted products Transfer User-Vetted products GOV PROVIDERS (CPC/EMC) NCEP/NCDC/USGS CTB SUPPOPRT PROVIDERS GOV OPS PRIVATE OPS R2O R2O O2R O2R R2O: CTB O2R: Model Test Facility RESEARCH, MODELING (dyn, stat), OBSERVATIONS Research, Modeling, Obs

  8. Доверяй, но проверяй

  9. Доверяй, но проверяй Trust, but verify

  10. Лроверяй, и Доверяй, Verify, and Trust

  11. Лроверяй, и Доверяй, Verify, and Trust Trust is at the core of a successful product suite. Transparent verification is one way to secure trust.

  12. Лроверяй, и Доверяй, Verify, and Trust Trust is at the core of a successful product suite. Transparent verification is one way to secure trust. VERIFICATION: A MEASURE OF FORECAST QUALITY, SKILL AND VALUE

  13. Types of Verification • Accuracy – e.g., AC, error, #correct, rmse, mae, etc… • Skill – e.g., HSS, RPSS, Brier SS • Bias – forecast too high/low? • Resolution – how well are different events discriminated? • Sharpness – able to predict extreme events?

  14. How to Proceed? Over the last decade, Dr. Holly C. Hartmann, and programmers Ellen Lay and Damian Hammond, of CLIMAS, have developed an interactive, on-line “Forecast Evaluation Tool” (FET) which allows users to evaluate the meaning and skill of CPC 3-Month Outlooks of temperature and precipitation. CPC proposes to: • Make the FET CPC’s outlet to users for forecast skill information, • Become a partner with CLIMAS and others to make the FET a community resource. • Expand the capabilities of FET

  15. Forecast Evaluation Tool: Example of a Means to Address Gaps What FET provides: • User-centric forecast evaluation and data access and display capability. • Leveraging of community software development capabilities. • Opportunity to DISCOVER and collect user requirements.

  16. CPC Temperature and Precipitation Outlooks • Probability of three categories • Bottom, middle, top 10 years, 1971-2000 • Maps show probability of likeliest category • Each category has a value at every point • Probabilities sum to 100% • If middle favored, borrow from extremes • “EC” means pr(b,n,a)=33.33, 33.33, 33.33

  17. CPC Temperature and Precipitation Outlooks Pr(b,n,a)= 16.67, 33.33, 50 • Probability of three categories • Bottom, middle, top 10 years, 1971-2000 • Maps show probability of likeliest category • Each category has a value at every point • Probabilities sum to 100% • If middle favored, borrow from extremes • “EC” means pr(b,n,a)=33.33, 33.33, 33.33 Pr(b,n,a)=40, 33.33, 26.67

  18. Downscaling: Probabilities to Temperatures

  19. Downscaling: Probabilities to Amounts

  20. Recent CPC Skill Improvements HSS=% improvement the forecast makes over random forecasts. More non-EC forecasts, and a large HSS are GOOD. Non-EC forecasts O’Lenic et al (2008) compared the HSS and % non-EC of official (OFF) forecasts made in real-time from 1995 through 2004 with forecasts made using an objective consolidation (CON) of the identical four main forecast tools which were used to prepare the real-time forecasts. Period (1995-2004) mean T, P forecast skills for OFF/CON were 22/26 for T, and 8.8/12.1 for P. CON % non-EC is also higher. CPC began using CON as a first guess in 2006.

  21. Top 2 rows: 1995-2004 HSS (lines) of 3-month P Outlooks, Official (OFF) and Consolidation (CON). Colors are the fraction of the time non-EC is predicted (%).Bottom row: Difference, CON-OFF (lines and colors). (See O’Lenic et al, 2008) SPRING SUMMER FALL WINTER FMA, MAM, AMJ MJJ, JJA, JAS ASO, SON, OND NDJ, DJF, JFM HSS HSS OFF HSS HSS OFF OFF OFF HSS CON HSS CON CON HSS HSS CON DIF DIF DIF DIF DIF +20% +18% +8% +16%

  22. Top 2 rows: 1995-2004 HSS (lines) of 3-month T Outlooks, Official (OFF) and Consolidation (CON). Colors are the fraction of the time non-EC is predicted (%).Bottom row: Difference, CON-OFF (lines and colors). SPRING SUMMER FALL WINTER FMA, MAM, AMJ MJJ, JJA, JAS ASO, SON, OND NDJ, DJF, JFM HSS HSS HSS OFF OFF OFF HSS OFF HSS HSS HSS HSS CON CON CON CON DIF DIF DIF DIF DIF +31% +55% +11% +40%

  23. GPRA Score Official Skill Metric:48-Mo. Running Mean U.S. Average T HSS

  24. What is the Forecast Evaluation Tool (FET)? • A web tool that allows users to interact with a database of CPC 3-Month Mean Temperature, Total Precipitation Outlooks, and verifying observations from1995-present, • A tracker of user preferences. • A self-teaching tool to enable a wide range of users to learn what CPC forecasts are, what they mean, and what their implications are for user applications. • A collector of user requirements. • A community tool for CPC, IRI, IPCC, …??? • A laboratory for growing services to users

  25. http://fet.hwr.arizona.edu/ForecastEvaluationTool/

  26. Introduction to the FET

  27. Tutorial

  28. Look at the latest forecasts

  29. Look at the latest forecasts

  30. Interact with observations

  31. Las Vegas Historical La Nina Observed P NEUTRAL EL NINO LA NINA

  32. Central Florida Historical El Nino Observed P EL NINO LA NINA NEUTRAL

  33. Oregon (east) Historical El Nino Observed P PDO + PDO - PDO neut

  34. Verification(regime, score, years)

  35. Verification(regime, score, years)

  36. Frequency of non-EC Forecasts

  37. Probability of Detection

  38. False Alarm Rate

  39. Brier Skill Score

  40. Ranked Probability Skill Score

  41. FUTURE: Expand FET’s Capabilities

  42. CTB User-Centric Forecast Tools Progress • Simple Object Access Protocol (SOAP) Tested • Secured Go-Ahead to Place FET on NWS Web Operations Center (WOC) • Trained CPC Staff in JAVA language • Scheduled Ellen Lay Training session in Nov.

  43. FUTURE of the FET Next 1-4 months: • Finalize and implement FET project plan at CPC. • Ellen Lay (CLIMAS) to train CPC personnel on FET version control and bug tracking at CPC, November 18-21, 2008. • Necessary software (APACHE TOMCAT, JAVA, Desktop View) acquired and installed at CPC. • Forecast, observations datasets in-place at CPC. • FET code ported to CPC, installed, tested. • FET installed to NWS Web Operations Center (WOC) servers

  44. FUTURE of the FET + In partnership with CLIMAS and community we will add: • Other forecasts and organizations • Time and space aggregation options • Significance tests/cautions to users • Requirements requests option • Questions option The stakes are high…..

  45. Source: The Washington Post Outlook Section, July 13, 2008

  46. Summary • Users want partnership, accuracy, specificity, flexibility • “Relationship” is synonymous with “partnership” • TRUST (honest brokerage) is central to these requirements. • Producers must learn WHO users are, HOW they use products and WHAT their evolving requirements are • Need to involve users and producers in iteratively optimizing products • A continuous flow of requirements from users toward research may avert VOD. • Means to fund an ever-expanding, perpetual product suite needed • Stakes are high: 6 of top 20 news/media June 2008 sites were weather-related. 10s-100s of B$ at stake. Climate will only add to this.

More Related