A Post-Processor for Hydrologic Ensemble Forecasts John Schaake, 1 Robert Hartman, 2 James Brown, 1 D.J. Seo 1, and Satish Regonda 1 1. NOAA/NWS Office of Hydrologic Development 2. NOAA/NWS California-Nevada River Forecast Center Presentation to European Geosciences Union April 17, 2008 Vienna
Elements of a Hydrologic Ensemble Prediction System Ensemble Pre- Processor Parametric Uncertainty Processor Data Assimilator Ensemble Post- Processor Hydrology & Water Resources Ensemble Product Generator Hydrology & Water Resources Models QPF, QTFQPE, QTE, Soil Moisture Streamflow Ensemble Verification System Fig 1 Ensemble Product Post-Processor
CNRFC Ensemble Prototype Smith River Salmon RiverMad River Navarro River American River (11 basins) Van Duzen River
NFDC1 – March day Post-Processor Calibration Analysis of Historical Model Simulation Results
NFDC1 – March day GFS-Based Hydrologic Ensemble Forecasts Ensemble Mean vs ObservedCumulative Rank Histograms
NFDC1 – March 15 Forecasts Cumulative Rank Histograms for Different Forecast Products
LAMC1 (Lake Mendocino, CA) Russian River Basin
l Total Area 3465 km2. l Elevation 17m m. l 2 Flood Control Reservoirs l 3 Local Areas. l 3 Official Flood Forecast Points. l Floods Nearly Every Year. l 3 Major Floods in Past 40 Years. Russian River
LAMC1 – Schematic of Possible Post Processor Applications Basin Model Of Natural Flow Post-Processor To Adjust to Observed Inflow Reservoir Operations Model Post-Processor To Adjust to Observed Outflow Gaged Outflow COE Estimated Inflow Diversion from Eel Basin Estimated Natural Flow
Full Natural Flow – March 15 Analysis of Historical Model Simulation Results
Full Natural Flow to Inflow – March 15 Analysis of Historical Model Simulation Results
Climatologies of Measured Inflow and Modeled Natural Flow (December – June)
Full Natural Inflow to Resevoir Outflow - March 15 Analysis of Historical Model Simulation Results
GLDA3 (Lake Powell Inflow) EPG Post-Processor Calibration Results
June Calibration – Lake Powell Analysis of Historical Model Simulation Results
July Calibration – Lake Powell Analysis of Historical Model Simulation Results
Some Challenges Alternative ways to evaluate Post-Processor integral equation to relax bivariate normality assumption that I used to get started? Can we adjust individual ESP traces (preserving temporal scale-dependent uncertainty) by using a cascade approach to apply multiple window applications of the product-based postprocessor? Multi-model applications?
Thank You