A Strategy to Use Long-range Climate Forecast Information for Hydrologic Ensemble Prediction

1. A Strategy to Use Long-range Climate Forecast Information for Hydrologic Ensemble Prediction John Schaake, Pedro Restrepo, D.J. Seo, Robert Hartman, Kevin Werner, Limin Wu and Julie Demargne Climate Diagnostics Workshop Boulder October 23-27, 2006

2. From Climate Prediction to Water Management Hydrology and Water Resources Modeling, Data Assimilation Water Management Climate Prediction Water Prediction Hydrologic Ensemble Prediction

3. Elements of a Hydrologic Ensemble Prediction System Weather and Climate Forecasts Single-value and ensemble forecasts Atmospheric Ensemble Pre-Processor Reliable hydrologic inputs Hydrological Models (& Regulation) Land Data Assimilator Hydrologic Ensemble Processor Hydrological Models Obs Ensemble forecasts Ensemble initial conditions Verification Product Generator Forecaster Role Reliable hydrologic products Products and Services

4. PreProcessor Objectives • Add climate forecast capability to EPP • to use CFS forecasts • to use other climate forecast information • Produce ensemble precipitation and temperature forecasts for basins and grids • Develop Community Ensemble Preprocessor (CEPP) as a component of the Community Hydrologic Prediction System (CHPS) • OHD • NCEP • CPPA projects (UW, Princeton, …) • Others (WR, HEPEX, GEWEX/HAP …)

5. CEPP Requirements • Produce inputs for RFC models • 6hr (and 1hr) time steps • Basin and elevation zone segments • Account for time-space scale-dependent uncertainty/variability in both P and T. • Hindcasts required for the period ~1981–present for hydrologic validation for user applications • EPP ensemble P/T climatologies must match RFC P/T calibration climatologies

6. Where are We Now? • Initial EPP • use short-term RFC forecasts • and CPC probability shift products for long-term. • EPP GFS Subsystem • uses RFC forecasts • GFS ensemble mean (out to 14 days) • and climatology for long-term • Temporal scale-dependency of forecast skill is accounted for • Both EPPs • produce basin P and T outputs at 6hr time steps • and are calibrated using RFC basin calibration climatology data.

7. What Do We Know? • Need to pre-process weather and climate forecasts (bias & uncertainty corrections) • Uncertainty is space and time-scale dependent • Weather and Climate forecasts contain information useful for water resources applications • Weather and Climate ensemble re-forecasts are required (>20yrs) for EPP calibration and hydrologic hindcast verification • Major work needed to produce reliable hydrologic ensemble forecasts

8. Present Limitations? • Cannot do hindcasts for monthly probability shifts (no archive) • Seasonal probability shift archive begins only in 1995 (not long enough) • RFC T/P short range forecasts archived only for last few years • GFS re-forecasts are for “old” version of GFS • Existing CFS – GFS gap (weeks 3 – 6) • SREF forecasts are not used • Atmospheric ensemble forecasts do not reliably account for uncertainty

9. Proposed Science Strategy • Future CEPP: • National “coarse” space-time grid operated at NCEP • Local (RFC/WFO) “fine” grid and basin applications • Integrate RFC, HPC, SREF, GFS, CPC and other forecasts into multi-model EPP • Improve ability to do hydrologic hindcasts with “recent” weather and climate forecasts

10. Tasks • Optimize specification of “coarse” and “fine” grids • Test “coarse” vs “fine” grid strategy • Include short range local forecast input in local “fine” grid • Test alternative EPP algorithms • Demonstrate consistency between EPP forecast and RFC analyses climatologies • Demonstrate potential to use EPP gridded analyses for future RFC model calibration • Develop capability to do weather and climate re-forecast cost-benefit analyses

11. ThankYou

12. Figure 13 – Long-term flow forecasts (up to lead-time 6 months) obtained from rainfall forecasts given by the CPTEC AGCM model (a) without statistical correction of rainfall forecasts using cumulative probability distributions, (b) with statistical correction. Forecasts start on 1 October 1997, for the outfall from the Furnas sub-basin. (a) (b) Return

13. Figure 14 – Long-term flow forecasts (up to lead-time 6 months) obtained from rainfall forecasts given by the CPTEC AGCM model Return

14. Some EPP ParametersNorth Fork American River Observed POP Observed CAVG Forecast CAVG Forecast POP Return

15. GFS Precipitation Forecast Verification North Fork American River Raw Forecasts Ensemble Mean Bias Day of Year Day of Year CRPSS Forecast Period Forecast Period Return

16. Figure 6 – Evaluation of the quality of rainfall forecasts given by the 40km-grid ETA model, for the whole of the Rio Grande basin, and for some sub-basins: (a) correlation coefficient as a function of lead-time; (b) correlation coefficient as a function of length of period over which rainfall is accumulated. (a) (b) Return

17. Correlation CoefficientGFS Precipitation Forecast vs ObservationNorth Fork American River Forecast Uncertainty Depends on both Lead-Time and Aggregation- Time Aggregate periods Return

18. Effect of Spatial Scale on 24hr Forecast Skill (July – 5 locations) Return

19. Example: SST Forecast • Seasonal SST forecast from ECMWF DEMETER project • 7 climate models • 6 months forecast starting August • 9 ensembles from each model • 20 years (1980-1999) • RMS error of all SST forecast initiated at August • Forecast over Equatorial Pacific • Multi-model posterior always has the smallest RMS error There is great potential to apply the approach for seasonal hydrologic forecasting. Return

20. TMX/TMN Calibration Data RFC-specific Utility Stations used for MAT Analysis MAT Time Series Identifiers Temperature Normals MAT Area Locations MAP Time Series Identifiers MAP Area Locations 6hr RFC QPF Verifications Files and Operational MAPs Parameter Estimator 6hr MAP Calibration Files Raw Historical Data Files CFS Ensemble Forecast (Application Under Construction) CFS Ensemble Forecast (Application Under Construction) Verification Results National Weather Service Hydrologic Ensemble Pre-Processor (EPP) GFS Subsystem J. Schaake, R. Hartman, J. Demargne, L. Wu, M. Mullusky, E. Welles, H. Herr, D. J. Seo, and P. Restrepo Average observed values of 6-hour precipitation corresponding to RFC and GFS forecasts Continuous Rank Probability Skill Score (CRPSS) for 6-hour precipitation forecasts Historical 6hr RFC Operational MAP Observed Data Files Operational RFC QPF Files Pre-ProcessorPrecipitation Algorithms RFC single-value forecasts (mm) Operational GFS Files Historical 6hr RFC Operational MAP QPF Data files Operational Forecast files Operational Ensemble MAPs Historical GFS Ensemble Mean Precipitation Forecasts (mm) Ensemble forecasts based on RFC single-value forecasts Precipitation Parameters Ensemble Generator 6hr MAP Unformatted Calibration Data Files MAP Conversion Utility Hindcast Ensemble MAPs GFS Processor Historical Data Sets Average forecast values of 6-hour precipitation for RFC and GFS GFS single-value forecasts (mm) Precipitation Verification Index Files Data Analysis Ensemble forecasts based on GFS single-value forecasts (mm) Verification Results Precipitation Ensemble Pre-Processor Statistics Average observed values of daily minimum temperature corresponding to RFC and GFS forecasts Continuous Rank Probability Skill Score (CRPSS) for daily minimum temperature forecasts RFC Historical Operational Observed TMX/TMN Operational RFC QTF Files Pre-Processor Temperature Algorithms RFC single-value forecasts RFC TMX/TMN Forecast Verification Files and Operational Observed TMX/TMN RFC-specific Utility Operational GFS Files RFC Historical Operational Forecast TMX/TMN MAT Analysis Operational Forecast Files MOS TMX/TMN Forecast Verification Files MOS Historical Forecast TMX/TMN MOS Utility Ensemble forecasts based on RFC single-value forecasts Operational Ensemble MATs Temperature Parameters Parameter Estimator Ensemble Generator Hindcast Ensemble MATs Average forecast values of daily minimum temperature for RFC and GFS 6hr MAT Calibration Files 6hr to TMX/TMN GFS single-value forecasts Historical GFS Ensemble Mean Temperature Forecasts Raw Historical Data Files Temperature Verification Index Files Ensemble forecasts based on GFS single-value forecasts Unformatted TMX/TMN Calibration Data Files Conversion Utility GFS Processor Historical Data Sets Temperature Ensemble Pre-Processor TMX: maximum temperature TMN: minimum temperature Return