EVALUATION OF A GLOBAL PREDICTION SYSTEM: THE MISSISSIPPI RIVER BASIN AS A TEST CASE Nathalie Voisin, Andy W. Wood and Dennis P. Lettenmaier Civil and Environmental Engineering 2008 AMS meeting, New OrleansWeather to Climate Scale Hydrological Forecasting University of Washington, Seattle 22 nd Conference on HydrologyP January 2008 ABSTRACT We describe a prototype system for medium range (up to two week lead) flood prediction intended for use in large rivers, primarily outside the developed world where in situ data are sparse. Our vision is to rely heavily on weather prediction model output, and satellite remote sensing, which will reduce the need for in situ precipitation and other observations in parts of the world where surface networks are sparse, and where a hydrologic forecast capability arguably would have the greatest value. The hydrologic component of the system is the Variable Infiltration Capacity (VIC) macroscale hydrology model. In the prototype, VIC generates hydrologic states for forecast initialization using daily ECMWF analysis wind, surface temperature and 24-hour ECMWF deterministic forecast precipitation, all of which are bias corrected with respect to ECMWF ERA- 40. In hindcast mode, VIC is driven by global ECMWF Ensemble Prediction System 10-day forecasts that are bias corrected with respect to ERA-40 data as well, and spatially disaggregated using two higher spatial resolution satellite products: Global Precipitation Climatology Project (GPCP) 1dd daily precipitation and Tropical Rainfall Measuring System (TRMM) 3B42 precipitation. The prototype system is implemented globally at one-half degree spatial resolution, and tested for selected large river basins during the period. The Mississippi River was selected for evaluation because the in situ data network is relatively dense. It offers a good opportunity to evaluate the global system which is based on much sparser in situ data. In our evaluation, we focus in particular on floods within the Mississippi Basin between 2002 and OBJECTIVES: a) Implement a prototype global flood forecasting scheme b) Evaluate performance using relatively dense observation network over the Mississippi River Basin MEDIUM RANGE LARGE SCALE FLOOD PREDICTION SCHEME Hydrologic model spin up (0.5 degree simulation) Several years back Hydrologic forecast simulation (0.5 degree simulation: stream flow, soil moisture, SWE, runoff ) Nowcasts INITIAL STATE Medium range forecasts ( up to 10 days) ECMWF EPS 50 ensemble members Bias correction at 1 degree, with respect to ECMWF Analysis (Ensures consistency between the spinup and the forecasts) Downscaling from 1 to 0.5 degree Downscaling to 0.5 degree Daily ERA-40 surrogate for near real time analysis fields Atmospheric inputs VIC Hydrology Model Several years back Medium range forecasts CONCLUSIONS -Even after bias correction, the EPS ensemble average is too wet in times of heavy precipitation. This is due to the short period of both the forecasts ( ) and the Analysis ( , 27 day window) used, requiring then the use of extreme value fitting in the procedure. Both the Analysis and the forecast will be bias corrected with respect to the longer ERA-40 dataset in future work. -The simulations using ECMWF Analysis fields is promising, although calibration and/or bias correction remain key issues. Downscaling to 0.5 degree Daily ECMWF Analysis ERA-40 (ending in 2002) and ECMWF Analysis (starting here in 2002) are both used as surrogates for observations. We assume that ERA-40 and ECMWF Analysis are comparable and that ECMWF Analysis can be used in the continuity of ERA-40. We compare daily average precipitation field with TRMM (3B42, i.e. monthly values rescaled to gauges). 4 day flood due to heavy rain, caused $20,000,000 in damages, 100 people displaced (Dartmouth Flood Observatory) April 2003 : Mississippi and Alabama 5 day accumulated precipitation Analysis EPS ens. Avg. (4/6/2003) 5 day accumulated runoff (VIC simulated) Analysis EPS ens. Avg. 9 day flood due to heavy rain, caused $3,000,000 in damages (Dartmouth Flood Observatory). 24 April – 2 May 2004 : Arkansas and Oklahoma 5 day accumulated precipitation Analysis EPS ens. Avg.(4/23/2004) 5 day accumulated runoff (VIC simulated) Analysis EPS ens. Avg. 28 day flood due to “several days of heavy snow, ice and rain on saturated ground followed by warm temperatures ”, 4,000 people displaced (Dartmouth Flood Observatory). 5 January – 1 February 2005 : Ohio River and lower Mississippi 5 day accumulated precipitation Analysis EPS ens. Avg.(1/4/2005) 5 day accumulated runoff (VIC simulated) Analysis EPS ens. Avg. 5 day accumulated change in snow water equivalent (VIC simulated) Analysis EPS ens. Avg 5 day accumulated change in soil moisture (VIC simulated) Analysis EPS ens. Avg. 5 day flood due to heavy rain (Dartmouth Flood Observatory) September 2006: Arkansas and Missouri 5 day accumulated precipitation Analysis EPS ens. Avg. (9/21/2006) 5 day accumulated runoff (VIC simulated) Analysis EPS ens. Avg. Acknowledgements: we wish to acknowledge ECMWF for the access to Analysis and EPS forecasts, and in particular Dr. Florian Pappenberger of ECMWF for his help in retrieving the files. -Calibration of the hydrologic model is still needed before we can evaluate the simulated discharges. -The approach is transferrable worldwide thanks to the global coverage of the ECMWF products. Further work is ongoing in ungauged regions. We compare ERA-40 daily average precipitation fields with GPCP 1dd and gauge based precipitation fields (Adam et al. 2006) TRMM and ECMWF Analysis basin average monthly precipitation.