1 / 54

MMEFS-XEFS-THORPEX/HYDYO Conf Call #1

MMEFS-XEFS-THORPEX/HYDYO Conf Call #1. May 27, 2008. Objectives. Better-understand short-term needs of the RFCs for MMEFS Discuss the questions and issues raised by the RFCs, and how THORPEX-HYDRO and XEFS may help address them

tatum-sellers
Download Presentation

MMEFS-XEFS-THORPEX/HYDYO Conf Call #1

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. MMEFS-XEFS-THORPEX/HYDYO Conf Call #1 May 27, 2008

  2. Objectives • Better-understand short-term needs of the RFCs for MMEFS • Discuss the questions and issues raised by the RFCs, and how THORPEX-HYDRO and XEFS may help address them • Discuss howewe may develop closer coordination and collaboration so that • MMEFS maximally leverages THORPEX-HYDRO and XEFS • XEFS maximally leverages MMEFS

  3. THORPEX-HYDRO • An OHD-supported, THORPEX-NCEP-OHD joint project in collaboration with HMT, ESRL/GSD, UVA, CPO/CPPA • To produce reliable and skillful ensemble forecasts of hydrometeorological variables at the weather scale, in particular for high-impact events • The targeted users are: • RFCs for operational hydrologic ensemble forecasting using XEFS • Water resources managers and end users for various hydrology and water resources applications

  4. THORPEX-HYDRO: Objectives • Develop and operationally implement real-time bias correction, downscaling and hindcasting techniques that are applicable to global, regional, and climate ensembles • Develop capabilities for routine ingestion, testing, and use in XEFS of NCEP ensembles by OHD and RFCs • Improve land surface modeling capabilities for large-scale hydrologic evaluation of hydrometeorological ensembles at the weather scale and for collaborative research and development of hydrologic ensemble techniques and products that may best utilize them • Enhance capabilities for seamless verification of hydrometeorological and hydrologic ensembles across NCEP, OHD and RFCs For details, see the THORPEX-HYDRO Project Plan (2008)

  5. eXperimental Ensemble Forecast System (XEFS) Forecasters add value Water customers Weather & climate forecasts XEFS will enable seamless hydrologic ensemble prediction from weather to climate scales and translate weather and climate prediction into uncertainty-quantified water information

  6. XEFS/Ensemble Pre-Processor III (EPP3) Short-Range Medium-Range Long-Range From XEFS Design & Gap Analysis Report (NWS 2007)

  7. Focus of today’s call • Reliable and timely acquisition of precip and temp ensembles • Generating precip and temp ensembles that • Are reliable, in the mean and in the spread sense, over the range of MAP & MAT basin scales • Bias correction • Down/up-scaling • Are seamlessly blended from short to medium range • Joining/blending • Best capture collective skill in multimodeldata • Multimodel ensemble • Getting support from and providing feedback to NCEP (NCO, EMC, HPC)

  8. MMEFS ER RFCs 2008 from Joe Ostrowski (2008)

  9. Short-term Hydro Ensembles • MMEFS – Met-Model based Ensemble Forecast System • Initial goal to produce 7 day ESP output based on AWIPS GEFS ensembles • 00Z and 12Z forecast cycles • In production now using initial NERFC method (mix of AWIPS and 64 bit machine)* • *Reservoir operation adjustments needed • Total processing time - ~ 35 minutes • 166 basins • OFS extraction, ESP, ESPADP, and HTML creation on AWIPS • GRASS interpolation, MAP/MAT datacard generation on 64-bit machine

  10. Short-term Hydro Ensembles • Incorporate add’l ensemble model data (outside AWIPS) • NCEP GEFS (21 members) • Canadian model out for 7 days (GEFS grid resolution, 21 members) • SREF out for 84 hours (21 members) • WRF/MM5 data from SUNY-Stonybrook for 48 hours (12 members)

  11. Short-term Hydro Ensembles • Analyze all available data to construct expected value plots of stage, precip, and temps (extract data from “.CS” files, analyze and graph using “R”) • Max 75 members 1st 48 hours • Max 63 members hours 52-84 • Max 42 members hours 90-168

  12. AWIPS GEFS Grid

  13. NCEP GEFS Grid

  14. SREF Grid

  15. SUNY WRF Grid

  16. Critical Elements • Grid to raster conversion • Shedd Grib2LatLon for AWIPS GEFS • MDL’s “degrib” to convert all other GRIB content to shapefiles • GRASS ingest of ascii grid/shapefile for chosen region • GRASS grid-to-raster interpolation • MAP/MAT calculations • GRASS averaging over basins

  17. Critical Elements • Concerns • Appropriate interpolation results • How judged? • Performance throughput • Objective – best MAP/performance ratio, not visual appearance of interpolated grid • Datacard generation performance • OK for MARFC and NERFC (165-185 basins), excessive for OHRFC (~690 basins)

  18. AWIPS GEFS Grid (ThP)

  19. AWIPS GEFS Grid (IDW 12 pt)

  20. AWIPS GEFS Grid (RST)

  21. Preliminary Assessments • Thiessen polygon best on tight grid (speedy) • Inverse distance weighting reasonable with 4 point scheme (over-isolation with more points in scheme, also slower) • Regularized spline with tensioning • Visually appealing, but much slower • Add’l step to eliminate negative precip values needed, adding to step completion time

  22. SUNY WRF Grid Subset (calculations conducted over pink region)

  23. SUNY WRF Grid (IDW 4 pt)

  24. SUNY WRF – Thiessen Polygon

  25. SUNY WRF Grid IDW 4 pt

  26. Preliminary Sensitivity Analysis • SUNY, 12 km resolution, 48 hours (9 time steps), 166 basins • 1 pt ~800 secs • 1pt w/ -n ~1540 secs • 4pt ~2620 secs • 4pt w/ -n ~3250 secs • Max MAP diff among all methods, all MARFC basins analyzed: • .013” out of 1.3” MAP 48 hour total • Conclusion: for 12km resolution simplest method is optimum

  27. MMEFS Status • AWIPS GEFS (12 members) available ~0630 and ~1830Z daily • GRASS and Datacard file generation possible on 64-bit processor • OHRFC experienced 70% time reduction in use of generalized system on 64-bit machine vs AWIPS • Developing generalized system to effectively and efficiently handle all sources of ensemble files • Work both on AWIPS and on AWIPS/64-bit hybrid • Working with NERFC and OHRFC for system improvements and enhancements

  28. MMEFS Status(Preliminary, limited-sample performance) • AWIPS GEFS • 2x daily (00 and 12Z), 336 interpolation steps MAP (12 members x 28 time steps [168 hrs, 6 hour intervals]) • ~1400 secs for GRASS/Datacard on 64-bit • MAP and MAT simultaneously • NCEP GEFS • 4x daily, 588 interpolation steps for MAP (21 members x 28 time steps [168 hrs, 6 hour intervals]) • ~1700 secs for GRASS/Datacard on 64-bit • MAP only (MAT turned off for end of SNOW season) • CMC GEFS • Should be same as NCEP GEFS

  29. MMEFS Status(Preliminary, limited-sample performance) • NCEP SREF • 4x daily, 609 interpolation steps (21 members x 29 time steps [87 hours, 3 hour intervals]) • ~4200 secs for GRASS/Datacard on 64-bit • MAP and MAT simultaneously • SUNY WRF/MM5 • 1x daily (00Z cycle), 216 interpolation steps (12 members x 9 time steps [48 hours – 0 hour included, 6 hour intervals] x 2 pcpn layers [convective and non-convective] per member) • ~800 secs for GRASS/Datacard on 64-bit • Only MAP at this time

  30. MMEFS Status • Next steps • AWIPS-only solution • Port and test 64-bit approach on AWIPS • Conduct performance tests • Integrated AWIPS/64-bit solution • Initiate process on AWIPS • Process GRASS/Datacard steps on 64-bit • Finish processing on AWIPS (ESP, ESPADP and post-processing (summaries, html, etc.) • MARFC Reservoir operation revisions • Additional performance improvements • Datacard generation for OHRFC needs • Additional generalization for portability

  31. Using Atmospheric Model Ensembles to generate Short Term Hydrologic Ensembles A joint project of the 3 Eastern Region RFCs From Rob Shedd (2008)

  32. Project Origins • Contingency forecasts are time consuming to generate and rarely accurate • Longer term ensembles (30-90 day AHPS plots) do not meet the needs of most users • Assist in Flood Outlook Product production

  33. Question • Can we use atmospheric model ensembles to derive 5-7 day probabilistic hydrologic forecast?

  34. Proposal • Initially, use the GFS Ensemble precip and surface temperature individual members that are available on AWIPS as input traces for driving hydrologic model • GFS selected because: • Individual members available on AWIPS, not just mean/spread • Long enough time horizon available to provide enhanced information

  35. Issues and Questions • Will the 12 members provide a valid statistical sample? • Will the use of the GFS ensemble provide sufficient spread of the ensembles? • Is the grid resolution adequate to provide sufficient resolution of the basin average precip? • Does this input provide an unbiased data set relative to our calibrated hydrologic models?

  36. Final Question • Will this approach provide better hydrologic forecast information for days 3-7 than we are able to provide currently?

  37. Procedure • De-grib individual member, time step, data type • Use Grass to convert point to MAP/MAT • Post MAP/MAT to database • Generate Datacard files • Run ESP

  38. Text Products ... GFS Ensemble Run : 04 / 09 / 2008 00 Z ACTION MINOR MODERATE MAJOR Stg Pct Stg Pct Stg Pct Stg Pct COHN6 17.0 25% 20.0 <5% 21.0 <5% 22.0 <5% EAGN6 9.0 16% 11.0 <5% 12.0 <5% 16.0 <5% FLVC3 6.0 15% 7.0 <5% 10.0 <5% 15.0 <5% FTEN6 24.0 95% 26.0 95% 27.0 95% 29.0 74% GAYC3 7.0 15% 8.0 <5% 10.0 <5% 15.0 <5% HDYN6 9.0 95% 14.0 95% 17.0 81% 19.0 61% HOPN6 6.0 95% 7.0 95% 9.0 27% 10.0 <5% KASN6 5.0 95% 6.0 95% 7.0 77% 8.0 30% LTLN6 13.0 95% 15.0 58% 17.0 23% 18.0 14% MCKN6 10.0 95% 12.0 43% 13.0 23% 15.0 <5% MRNN6 18.0 <5% 20.0 <5% 22.0 <5% 24.0 <5% MTRN6 8.0 <5% 11.0 <5% 15.0 <5% 18.0 <5% NCKN6 8.0 95% 10.0 84% 11.0 64% 12.0 25% PTVN6 9.0 <5% 12.0 <5% 14.0 <5% 16.0 <5% ROSN6 14.0 <5% 18.0 <5% 21.0 <5% 23.0 <5% RVRN6 6.0 95% 7.0 95% 8.5 95% 9.0 95%

  39. Graphical Products – MAP / MAT

  40. Graphical Products - Stage

  41. Expected Value Plots

  42. So, what happened?

  43. Issuance • Products are issued typically about 0830 and 2100 UTC • At NERFC takes roughly 65-75 minutes to run • Products are issued internally only • Text product available on AWIPS • Graphics on internal page: ftp://ftp.werh.noaa.gov/share/nerfc/gens/gens.html

  44. Ensemble Precip – March 8

  45. GFS40 Precip – March 8

  46. GFS Ensemble Grid

  47. Next Steps • Investigate use of SREF and other ensembles • FTP access to full SREF data at NCEP • MARFC working with SUNY-SB for WRF/MM5 data • MARFC working with CTP SOO to gather NCEP GEFS and CMC data • Effectively combine information where appropriate • Create expected value plots for all ensemble results for each time period • Create traces and exceedance information for common periods • Utilize "run_espdata" script and locally developed software • MARFC is working with Stony Brook on WRF • MARFC running on 64-bit machine; OHRFC moving there • Verification analysis • Identify means of improving performance • And someday … XEFS

  48. OHRFC From Tom Adams

More Related