Download
1 / 17
170 likes | 402 Vues
NWS Hydrologic Forecast Verification Team: Status and Discussion. Julie Demargne OHD/HSMB Hydrologic Ensemble Prediction (HEP) group. http://www.nws.noaa.gov/ohd/rfcdev/projects/rfcHVT_chart.html. UCAR. CHPS Verification Service. Verification Statistics & Products. Forecast Products.
E N D
NWS Hydrologic Forecast Verification Team:Status and Discussion Julie Demargne OHD/HSMB Hydrologic Ensemble Prediction (HEP) group http://www.nws.noaa.gov/ohd/rfcdev/projects/rfcHVT_chart.html UCAR
CHPS Verification Service Verification Statistics & Products Forecast Products Verification Component Display/Product Generator IVP FEWS Spatial Display FEWS Time Series Display EVS WR Water Supply Web Graphics Generator Interactive Forecasting System Real-Time Verification(analogs, summary stats…) Databases Archive files External data
RFC Contribution • Input from RFCs is crucialto provide useful verification information to • modelers and forecasters to guide improvements of forecasting system • users to maximize utility of forecasts in their decisions • Key questions: • What verification metrics/products are the most meaningful? • What are the best methods to supply verification information?
1st Team Charter • July 07:1st team charter to propose standard verification strategies (metrics, products, and verification analyses) by Sep. 2009 • August 07:1st RFC verification workshop • November 08:2nd RFC verification workshop • July-August 09: presentation at HIC meeting, feedback from HICs on 2nd team charter draft (5 HICs replied) • Main feedback: include SCHs to develop meaningful products • September 09: final team report on recommended standard verification strategies and future team activities (2nd team charter)
2nd Team Charter • October 09: finalized 2nd team charter for Oct. 2009 - Sep. 2011 • Mission: • test verification standards (RFC case studies) • perform user analysis of products w/ SCHs and OCWWS/HSD and help develop requirements for dissemination • support design/development of CHPS Verification Service • Main deliverables: • report on improved verification standards and case studies done by all RFCs and by OHD/HSMB/HEP • prototype functionalities based on IVP, EVS, and CHPS to generate standard verification products
Status on 2nd Team • 2 meetings in Dec. 09 and Feb. 10: • presentation of OHD verification activities (FY10 plan) • presentation of MBRFC and NCRFC QPF verification case studies • survey on real-time verification (analog query and summary verification displays): feedback from all RFCs
Clarifications on Team Activities • Verification team report listedonly recommendations • Proposed sensitivity analyses for RFCs: • what is the optimized QPF horizon for hydrologic forecasts? • do run-time mods made on the fly improve forecasts? • Each HIC decides how the RFC team member(s) may contribute to the team • Each RFC defines its own verification case study to test proposed verification standards
Recommendations: 1) QPF horizon • Goal: what is the optimized QPF horizon for hydrologic forecasts? • QPF horizon to test • 0 (no QPF), 6-hr, 12-hr, 18-hr, 24-hr, 30-hr, 36-hr, 48-hr, 72-hr • Longer (96-hr, 120-hr…): depends on RFC and location • Model states to use • Similar to operational mods except mods that impact QPF • Metadata to store which mods were used in these runs • What forecast to verify • 6-hr stage forecasts for 7-day window (longer window for slow response basins)
Recommendations: 2) run-time MODs • Goal: do run-time mods made on the fly improve forecasts? • 4 scenarios • Operational forecasts (w/ all mods) • Forecasts w/ best available obs. and fcst. inputs wo/ on-the-fly mods • Forecasts w/ best available obs. inputs (no fcst) w/ all mods • Forecasts w/ best available obs. inputs (no fcst) wo/ on-the-fly mods • What forecast to verify • 6-hr stage forecasts for same window as in operations • Model states • Carryover from 5 days ago (w/ past mods) + a priori mods (known before producing any forecast)
Discussion • Other ways to explore QPF horizon and run-time mods? • Feedback on RFC’s contribution to verification? • How to best coordinate between Verification Team, CAT/CAT2 Teams, Graphics Generator Requirements Team, SCHs and OCWWS/HSD? • Feedback on future verification activities at OHD?
Final team report: recommendations • Key verification metrics for 4 levels of information for single-valued and probabilistic forecasts • Data information (scatter plots, box plots, time series plots) • Summary information (e.g. skill scores) • More detailed information (e.g. measures of reliability, resolution, correlation) • Sophisticated information (e.g. for specific events) • Corresponding verification products
Recommended metrics • 4 different levels of information
Recommended analyses • Analyze any new forecast process with verification • Use different temporal aggregations(e.g. weekly max. flow) • Analyze verification score as a function of lead time; if similar performance, data can be pooled from different lead times • Perform spatial aggregation carefully • Analyze results for each basin and results plotted on spatial maps • Use normalized metrics (e.g. skill scores) • Evaluate forecast performance under different conditions • w/ time conditioning (by month and season) • w/ atmospheric/hydrologic conditioning using probability and absolute thresholds (e.g. 90th percentile vs. Flood Stage) • Report verification scores with sample size(in the future, confidence intervals)
January October April Diagnostic verification products • Example:skill score maps by months Smaller score, better
Analog 3 Analog 2 Observed Live forecast Analog Observed Analog Forecast Real-time verification: analogs • Approach: select analogs from a pre-defined set of historical events and compare with ‘live’ forecast Analog 1 “Live forecast for Flood is likely to be too high”
