H23C-1139 1340h
Streamflow Simulations for the Mississippi River Basin Based on Ensemble Regional Climate Model Simulations
Ensemble simulations provide a useful tool for studying uncertainties in climate projections and for deriving probabilistic information from deterministic forecasts. Although a number of studies have examined variability within climate models, fewer have quantified the extent to which variability and uncertainty in climate simulations then propagates through impacts models. Here we evaluate the variability in simulated streamflow that result from taking the streamflow model's inputs from different members of an ensemble of simulations by a decadal-scale nested regional climate model. The regional climate model, RegCM3, simulated a domain covering the continental U.S. and most of Mexico for the period 1986-2003 using initial and lateral boundary conditions from the NCEP-DOE Reanalysis 2. Three RegCM3 realizations were created, each initialized one month apart but otherwise identical in configuration so that their collective behavior provides a measure of internal variability of the climate model. RegCM3 output for daily precipitation, temperature, and radiation were then used as input to the Soil and Water Assessment Tool (SWAT) over the upper Mississippi River basin. Seasonal and interannual variability of SWAT-predicted streamflow indicate that the internal variability of the RegCM3 climate model carries through to produce spread in simulated streamflow from SWAT.
H23C-1140 1340h
Sampling Uncertainties for Ensemble Forecast Verification Measures
Verification of forecasts is an essential first step for their operational use in decision-making. Forecast verification is carried out using a verification data set, which contains a record of forecasts and subsequent observations. A comparison of the forecasts with the observations is then made to assess forecast quality. For some forecasting systems, the archive of operational forecasts may be sufficient to make a comparison. For others, it may be necessary to reconstruct forecasts for the past, where a longer record of observations is available (a technique also known as hindcasting). Regardless of the approach, forecast quality measures evaluated during forecast verification are {\it sample estimates}, and are often based on relatively small sample sizes. For instance, with long-range ensemble streamflow forecasts, there is only one forecast made each year (for a given lead time and forecast period). Since most flow records have about 50 years or fewer of observations, even a reconstructed forecast sample is severely limited. In this presentation, we examine the sampling uncertainty of distributions-oriented (DO) forecast quality measures for probability forecasts from ensemble forecasting systems. With the DO verification approach, the correspondence between forecasts and observations is modeled explicitly by their joint probability distribution. Aspects of forecast quality of interest in verification are derived from the joint distribution model. Sampling theory is used to develop exact or approximate bias and standard error estimates for derived forecast quality measures. We also explore various statistical modeling approaches, which are needed to completely model the joint distribution, to estimate biases and standard errors for certain measures. The uncertainty estimators are evaluated for several prototype forecasts using Monte Carlo simulation. The estimators are then applied to long-range streamflow forecasts from the National Weather Service's (NWS) Advanced Hydrologic Prediction System (AHPS) for the Des Moines River basin. The results illustrate how sampling uncertainty affects inferences on forecast quality for probability distribution forecasts from ensemble systems.
http://www.iihr.uiowa.edu/~verification/
H23C-1141 1340h
Toward an Ensemble Streamflow Forecast Over the Entire France
Since the year 2003, the French National Weather Service (Meteo-France) uses an operationnal real-time system that provides a daily monitoring of the water budget, streamflows and aquifer levels over the entire France : the SAFRAN-ISBA-MODCOU (SIM) system. This coupled model is composed of the ISBA surface scheme and of the distributed hydrological model MODCOU. The system is used in a forced mode, with the atmospheric forcing derived from observations through the use of the SAFRAN analysis system. Such a system has been validated over 3 large french basins~: the Rhone, the Adour-Garonne and the Seine basins. It was shown that the system satisfactorily reproduces the water and energy budgets, as well as the observed streamflows, aquifer levels and snow-packs. In particular, the main long-duration floods of the Seine are well simulated. The SIM system is also used for streamflow forecasting. As a first step, experiments of determinist forecasts have been performed over the Rhone basin, using 2- and 3-day quantitive precipitation forecast. The encouraging results showed the potential of SIM for flood forecasting. As a next step, an ensemble streamflow prediction system is now being built. The forecasts from the Ensemble Prediction System of the ECMWF are used to force the system. The initial conditions of soil moisture, aquifer levels, etc. are given by the operationnal run of SIM, and the results are analysed for each forecast day. This system is expected to give 10-day forecasts of the streamflow of the main french rivers with a measure of the associated confidence, which is greatly valuable for flood warning and water management.
H23C-1142 1340h
Generation of Ensembles of Spatial Hydrologic Fields
In ensemble prediction applications, if only deterministic forecasts are available (e.g. rainfall, temperature), an ensemble of forecasts often needs to be generated from the deterministic forecast to properly quantify the forecast uncertainty. The generation is usually done by Monte Carlo simulation and the distributional parameters needed are obtained by comparing the (deterministic) forecasts to observations in the historical records. John Schaake (NOAA/NWS) has developed procedures for generating at a point, or for an areal average forecast ensembles conditioned on the deterministic forecast. For spatial hydrologic fields like deterministic quantitative precipitation forecasts (QPF), or even temperature, the generation becomes significantly more complicated due the need to preserve the spatial correlation structure within each generated ensemble field. Spatial generation techniques like Turning Bands may fail to account for the statistical conditioning between the forecasts and observations, and therefore may not properly represent the forecast uncertainty across the ensembles. The ensemble generating procedure we describe here is utilizes the conditional generation approach within a high-dimensional joint distribution that represents the forecasted and observed hydrologic variables at all locations. In the implementation of the ensemble generation procedures, a transformation of the sample distribution to the standard multi-normal and back is used to deal with the difficulties that arise from the highly irregular marginal distribution of observed hydrologic variables (e.g. rainfall). This approach can also be viewed as a Bayesian estimation of future observations given the (deterministic) forecast. Results will be presented comparing the spatial correlations within the generated ensembles, and the variability across ensembles, with observations to determine whether the uncertainty in the hydrologic variables is preserved.
H23C-1143 1340h
Ensemble Land Surface Modeling Using Coarse Satellite-Based Precipitation Forcing
Precipitation is the key forcing variable for land surface hydrologic processes and is largely responsible for variability in soil moisture and surface flux fields. The measurement of precipitation over large scales is difficult due to its inherent spatial and temporal intermittency and the lack of sufficiently dense ground-based monitoring networks in many regions of the globe. Methods for remotely sensing precipitation are now generally available, but provide estimates that are spatially coarse (e.g. tens to hundreds of kilometers), aggregated in time (daily to monthly), and have complex error structures (positive non-detection probability, non-zero false alarm rate, large uncertainty, etc.). Due to the nonlinearity of surface hydrologic processes, these products cannot be directly used in modeling studies. Furthermore, for accurate hydrologic state and flux predictions it is crucial that uncertainty in these estimates be properly incorporated into modeling frameworks. In this paper we present an ensemble forecasting framework that contains a spatio-temporal disaggregation scheme for remotely sensed Global Precipitation Climatology Project-1 degree daily (GPCP-1DD) precipitation product. A detailed study of the error characteristics of the GPCP 1DD precipitation forcing is undertaken and then incorporated in the framework. This approach significantly enhances the utility of the remote sensing product for hydrologic applications. The spatio-temporal disaggregation scheme takes advantage of the ensemble nature of the system by introducing precipitation realizations that are conditioned on the remote sensing data. The ability to use coarse precipitation observations is tested in experiments using data from the SGP97 field experiment. This approach not only captures the large scale spatial variability in precipitation contained in the remote sensing observations, but introduces a more realistic error structure in the precipitation forcing that accounts for errors in storm magnitudes, arrivals, and spatial structure. Results from tests using the remotely sensed precipitation show improvement in both soil moisture and land surface flux estimates over those using sparse ground-based precipitation. Furthermore the general ensemble framework is easily adapted to assimilate other observations (e.g. microwave radiobrightness) using the Ensemble Kalman Filter (EnKF). The combined ensemble data assimilation framework allows for the incorporation of information contained in coarse remote sensing observations of fluxes (precipitation) and states (soil moisture) to better capture the response of the land surface.