Apr-S ep R unoff Volum e Forecast C olum bia R iverat the D alles, O R 50 60 70 80 90 1-Jan 1-Feb 1-Mar 1-A pr 1-May forecastdate percent of norm al ESP NW S/NRCS O bs NSIPP NCEP G SM 2003 Summer streamflow volume forecast comparison with NWS / NRCS official forecasts Andy Wood, Alan Hamlet, Seethu Babu, Marketa McGuire and Dennis P. Lettenmaier A seasonal hydrologic forecast system for the western U.S. OVERVIEW We have implemented the Variable Infiltration Capacity (VIC) hydrology model over the western U.S. at 1/8 degree spatial resolution for ensemble hydrologic prediction at lead times of 6 months to 1 year. Real-time hydrologic forecasts are made once monthly using initial conditions simulated with real-time observations of temperature and precipitation, and adjusted via the assimilation of SNOTEL snow water equivalent and, experimentally, MODIS snow-covered area. Benchmark climate forecasts are constructed via the well-known Extended Streamflow Prediction (ESP) method of the National Weather Service. The ESP forecasts are further composited to provide ENSO and PDO-conditioned ensembles. Experimental hydrologic forecasts are also made using climate forecast ensembles derived from the NCEP Global Spectral Model (GSM), the NASA NSIPP-1 model, and the CPC official seasonal outlooks. These are used to drive experimental reservoir forecasts in some locations. Expansion to Current Forecasting System (starting Sept. 2003) References / Acknowledgements Wood, A.W., E.P. Maurer, A. Kumar and D.P. Lettenmaier, 2002. Long Range Experimental Hydrologic Forecasting for the Eastern U.S., J. Geophys. Res., 107(D20). Wood, A.W., A. Kumar and D.P. Lettenmaier, 2004, A retrospective assessment of NCEP GSM-based ensemble hydrologic forecasting in the western U.S., J. Geophys. Res. (in review) Liang, X., D. P. Lettenmaier, E. F. Wood and S. J. Burges, 2004. A Simple Hydrologically Based Model of Land Surface Water and Energy Fluxes for GCMs, J. Geophys. Res., 99(D7). Clark, M., S. Gangopadhyay, L. Hay, B. Rajagopalan and R. Wilby, 2004. The SCHAAKE Shuffle: A method for reconstructing space-time variability in forecasted precipitation and temperature fields, J. Hydrometeorology, 5, 243-262. The authors acknowledge the support of NOAA/OGP, the IRI/ARCS Regional Applications Project, and the NASA Seasonal-to-Interannual Prediction Project (NSIPP). Hydrologic Model (Liang et al., 1994) Hydrologic Forecasting Simulations Climate Model Forecasts A retrospective hydroclimatology is used for: • Statistical bias-correction of climate model ensembles of monthly P, T at climate model scale • Spatial disaggregation to 1/8 degree hydrologic model scale • Temporal disag. from monthly to daily time step • detailed, assessed in Wood et al. (2002, 2004) Our initial forecast domain was the Pacific Northwest. Real-time bi-monthly updates began at the end of December, 2002, and ran through April 2003. NCEP GSM forecasts • T62 (~1.9 degree) resolution • 6 month forecast duration • 20-member ensembles, monthly P, T NSIPP-1 Tier 1 forecasts • 2 x 2.5 degree (lat x lon) resolution • 7 month forecast duration • 9-member ensembles, monthly P, T Review of Pilot Implementation: Columbia River Basin in Winter 2003 Dam Power Plant River/Canal Transfer Eastman, Hensley, & Millerton New Don Pedro & McClure Del ta New Hogan Pardee & Camanche Stanislaus River Tuolumne & Merced Rivers Delta Outflow Mokelumne River Calaveras River S a n J o a q u i n R i v e r New Melones San Luis Trin ity Whiskeyto wn Shast a Oroville (SWP) Folso m Clear Creek American River Feather River Trinity River Sacramento River Dam Power Plant River Transfer Del ta Colorado River San Joaquin River Columbia River Sacramento River UW forecasts halted Jan 15, 2003 Dec 28, 2002 Feb 1, 2003 Apr 1, 2003 Snow Water Equivalent Simulated System Storage (acre-ft) 1) NRCS SNOTEL / EC ASP observed SWE anomalies are interpolated in distance and elevation to hydrologic grid cell elevation bands, and linearly combined with simulated anomalies, to adjust the hydrologic model state at the start of the forecast. 2) spin-up met. data improvements method / MODIS experiments not illustrated Streamflow hydrograph forecasts (example from February 1) Initial hydrologic condition estimates Reservoir system forecast experiments re-evaluating the NASA/NOAA NLDAS 1/8 degree forcing product as a potential real-time forcing in Western U.S. automating nowcast / initial condition simulation to occur on weekly basis adopting selected experimental reservoir system forecasts as routine products comparing nowcasts with retrospective simulations now in progress extending back to 1915 (rather than 1960). Ongoing Work Univ. of Washington Climate Forecasts ESP forecasts • VIC model resolution (1/8 degree) • historical 12-month daily sequences from 1960-99 CPC Official Outlooks Downscaling 1 2 3 Components of Overall Real-time Forecasting Approach Simulated System Storage (acre-ft) min max fcst. ens. mean historical mean Selected Results Primary Upgrades to the forecasting system included: 1) the implementation of a simple method for assimilating snow water equivalent (SWE) observations at the start of the forecast, 2) a modification of the surface forcing estimation immediately prior to the forecast start using a set of real-time index stations We began adapting a set of reservoir system models for the western U.S to produce ensemble forecasts of reservoir system storages, operations and releases. example obs SWE anomalies corresponding SWE adjustment 6-Month Ensemble Forecasts of System Storage for the Columbia River Basin Using VIC Streamflow Forecasts and the ColSim Reservoir Model Initialized by Observed Reservoir Elevations (~ Feb 1, 2001) 5 4 SYNOPSIS: early winter snowpack deficits recovered somewhat, but ultimately led to moderate streamflow deficits in spring and summer. Forecasts posted on web page www.hydro.washington.edu/Lettenmaier/ Projects/fcst/ Feb 1 observed observed N CD C m et. statio n o b s. u p to 2 -4 m on th s fro m cu rren t local scale (1/8 degree) w eatherinputs soil m oisture snow pack H yd rolog ic m odelspin up SN O TEL U p d ate stream flow ,soil m oisture, snow w aterequivalent,runoff 25 th D ay, M onth 0 1-2 years back LDAS/other real-tim e m et. forcin g s for sp in -u p gap H yd ro lo g ic fo recast sim u latio n M onth 6 - 12 IN IT IA L STATE IN IT IA L STATE SNOTEL /M ODIS * Update ensem b le fo recasts ESP traces (40) C PC -based outlook (13) NCEP G SM ensem ble (20) N SIPP-1 ensem ble (9) *experim ental,notyetin real-tim e product • based on 102 Climate Divisions • 13-member ensemble derived from monthly P, T probability distributions CPC Outlooks • Resampling approach (the Schaake Shuffle of Clark et al., 2004) creates ensemble traces. • Spatial/temporal disaggregation as above 1 2 1 2 1 2 Snow Water Equivalent Soil Moisture Streamflow Forecasts Apr-Sep R unoffVolum e Forecast Snake R iver"nearm outh" 50 60 70 80 90 1-Jan 1-Feb 1-Mar 1-A pr 1-May forecastdate percent of norm al ESP NW S/NRCS O bs NSIPP NCEP G SM Apr-Sep R unoffVolum e Forecast Libby ReservoirInflow 50 60 70 80 90 1-Jan 1-Feb 1-Mar 1-A pr 1-May forecastdate percent of norm al ESP NW S/NRCS O bs NSIPP NCEP G SM Apr-Sep R unoffVolum e Forecast Dworshak ReservoirInflow 50 60 70 80 90 100 1-Jan 1-Feb 1-Mar 1-A pr 1-May forecastdate percent of norm al ESP NW S/NRCS O bs NSIPP NCEP G SM Example forecast for 2 (of ~50 active) locations (from March 25 initial conditions) Verification / Comparison with RFC runoff volume forecasts 1 2 In late March, initial conditions (at left) reflect below average moisture (soil and snow) in the mountains of the central and northern. ID, but early snowmelt in the lower areas of eastern WA and OR, and northern CA. These led to below average forecasts of summer streamflow for the Columbia River (location 1), which drains the PNW, and above average flow forecasts for the Sacramento River (location 2), which drains areas of NE California and south central OR. The UW Columbia R. ESP forecast (right) showed in fall 2003 that, due to low soil moisture and below average early snowpack in southern British Columbia, runoff would likely be below average. The NWS forecasts predicted near normal summer runoff until late Feb. and Mar., when dry and warm conditions abruptly reduced western US snowpacks, after which they converged toward the UW forecasts at about 80 percent of long-term average runoff. In N. California, similarly, low fall soil moistures indicated low summer runoff, but then snowpacks grew deeper than normal until their early melt in March. The UW ESP forecasts reflect these conditions. Spatial forecasts related to historical conditions (as anomalies and percentiles w.r.t. 1960-99) Example: NCEP Global Spectral Model (GSM) forecasts (leads 1-3 months shown, for May 25 initial conditions)