H13E-1616
Seasonal-to-Interannual Predictability of western North American Hydroclimate: The Utility of Multimodel Ensembles
Using ensemble simulations of 20th and 21st century climate from the IPCC AR4 archive, we examine the seasonal to interannual predictability of western North American climate. The utility of multimodel ensembles is assessed within the context of the agreement and spread in conditional hydroclimatic distributions developed on seasonal scales. The impact of projected climatic trajectories (temperature and precipitation) on the nature and frequency of western North American droughts is also examined.
H13E-1617
Multimodel Ensembles of Streamflow Forecasts: Role of Predictor State in Developing Optimal Combinations
Seasonal streamflow forecasts based on climate information are essential for short-term planning and for setting up contingency measures during years of extreme climatic conditions. Recent research shows that operational climate forecasts obtained by combining different General Circulation Models (GCM) have improved predictability/skill in comparison to the predictability that could be obtained from single GCM. In this study, we present a new approach for developing multi-model forecasts that combines streamflow forecasts from various models by evaluating their skill from the predictor state space. Based on this, we show that any systematic errors in model prediction with reference to a particular predictor conditions could be reduced by combining forecasts from multiple models along with climatological ensembles. The methodology is demonstrated through development of multi-model ensembles of streamflow forecasts for the Falls Lake reservoir in Neuse river basin, NC by combining probabilistic streamflow forecasts from two low dimensional statistical models that uses SST conditions in Tropical Pacific, North Atlantic and North Carolina Coast as predictors. Using Rank Probability Score (RPS) for evaluating the predictability of seasonal (July- August-September) streamflow forecasts available each year from the two candidate low dimensional models, the methodology proportionately gives higher representation by drawing increased ensembles for a model that has better predictability under similar predictor conditions. The performance of the multi-model forecasts are compared with the individual model's performance using various performance evaluation measures such as correlation coefficient, root mean square error (RMSE), average Rank Probability Skill Score, average Rank Probability Skill Score (RPSS) and reliability diagrams. By developing multi-model ensembles for leave-one out cross validated forecasts and adaptive forecasts based on the proposed methodology, the study shows that evaluating the model's performance based on the predictor state provides a better alternative in developing multi-model ensembles instead of combining models purely based on their long-term predictability.
H13E-1618
Seasonal River Flow Forecasting Using Multi-model Ensemble Climate Data
Developing skilful seasonal forecasting of river flows is important for many societal applications. Long-lead forecasts have potential to aid water management decision making and preparation for human response to hydrological extremes. The seasonal prediction of river flows has been a topic of increasing interest due to the recent 2004-06 drought and 2007 floods experienced in the UK. We compare the relative skill of predictions of river flow using: (1) a multi-Global Climate Model (GCM) ensemble data set and (2) downscaled multi-GCM data as input to a hydrological model. The period considered is 1980- 2001. The River Dyfi basin in West Wales, UK is the focus of this research. This basin is near natural, hence the climate-flow signal should be clearer. The DEMETER project is the source of the multi-model climate data, and this consists of 7 GCMs each with 9 ensemble members. Hindcasts with lead times up to 6 months are available from 1st February, 1st May, 1st August and 1st November initial conditions. Each hindcast was split into the first 3 and last 3 months, and the subsequent concatenation of the split hindcasts produced 2 time series (total of 7×9×2 ensemble series), which were run through the Probability Distributed Model (PDM). PDM is a lumped rainfall-runoff model that transforms rainfall and potential evaporation data to river flow at the basin outlet. PDM was calibrated with observations from 1980-1990, and then validated from 1991-2001. The coarse resolution of the DEMETER data (standardised to 2.5° × 2.5° resolution) means that the atmospheric motions at sub-grid scales are not captured by the models. The large spatial disparity between the GCM grids and the scale of the study (471.3 km2) lead to underestimation of precipitation by DEMETER models. This difference is addressed through the use of a statistical downscaling tool, the Statistical Downscaling Model (SDSM). The SDSM was calibrated on the ERA-40 re-analysis data set (from the ECMWF), as it provides one of the best estimates of the real atmosphere (a spatial resolution comparable to that of DEMETER models was used for this calibration). Multiple linear regression models (one per month) were used to link DEMETER predictors with basin scale rainfall, and a stochastic weather generator produced downscaled rainfall time series. These new downscaled series are designed to more closely represent catchment rainfall. DEMETER precipitation data and downscaled data are inputted to the PDM to determine their relative river flow modelling skill. Preliminary results show that simulated river flows driven by DEMETER do underestimate the observed flow. The downscaled series improves the hindcast skill, and little reduction in skill is seen when using a longer lead time hindcast. The results drawn from this research will have major implications for assessing (1) the potential skill expected from large scale GCM output, and (2) the relative improvement in skill of using downscaled versus non- downscaled precipitation data. Also, it will be possible to ascertain any degradation in the seasonal hindcast skill when using longer lead times.
H13E-1619
Advantage of Multi-scenario Ensembling for AGCM Seasonal Climate Forecast
The benefit of multi-model ensembling in seasonal forecasting has been increasingly employed in research and operations. In this presentation we examine skill of retrospective forecasts using the ECHAM4.5 atmospheric general circulation model (AGCM) forced with predicted sea surface temperatures (SSTs) from methods of varying complexity. The SST fields are predicted in three ways: persisted observed SST anomalies; empirically predicted SSTs and predicted SSTs from a dynamically coupled ocean-atmosphere model. Here we focus on the skill of seasonal precipitation given its importance in influencing streamflow variability. Analysis of the anomaly correlation skill for precipitation indicates that dynamically predicted SSTs generally improve upon persisted and empirically predicted SSTs when they are used as boundary forcing in the AGCM predictions. The skill differences in these experiments are ascribed to the skill of SST predictions in the tropical ocean basins. The multi-scenario forecast by averaging the results of the three retrospective experiments performs, overall, as well as or better than the best of the three individual experiments in specific seasons and regions. The advantage of multi-scenario forecast manifests both in the deterministic skill measured by the anomaly correlations based on the ensemble mean, and the probabilistic skill measured by the reliability and resolution of the ensemble distribution. In particular, the multi-scenario precipitation forecast for the December-February season demonstrates better skill than the best of the three scenarios over several regions, such as the western US and southeastern South America. These results suggest the potential value in producing super-ensembles spanning different SST prediction scenarios.
H13E-1620
Can multi-model combination really enhance the prediction skill of probabilistic ensemble forecasts?
The success of multi-model ensemble combination has been demonstrated in many studies. However, given that a multi-model contains information of all participating models, including the less skillful ones, the question remains as to why, and under which conditions, a multi-model can outperform the best participating single model. It is the aim of this presentation to resolve this supposed paradox. The study is based on a synthetic forecast generator, allowing the generation of perfectly calibrated single model ensembles of any size and skill. Additionally, the degree of ensemble underdispersion (i.e. overconfidence) can be prescribed. Multi-model ensembles are then constructed from both weighted and unweighted averages of these single model ensembles. Skill is measured with the discrete ranked probability skill score (Weigel et al. 2007a), which is favorable in the context of multi-model studies since it is insensitive to changing ensemble size. Applying this toy model, systematic model-combination experiments are carried out. We evaluate how multi- model performance depends on skill and overconfidence of the participating single models. It turns out that multi- model ensembles can indeed outperform a "best model approach", but only if the single model ensembles are overconfident. The reason is that multi-model combination reduces overconfidence, i.e. that ensemble spread is widened while the average ensemble mean error is reduced. This implies a net gain in prediction skill, because probabilistic skill scores penalize overconfidence. It is under these conditions that even the addition of a consistently poorer model can enhance multi-model skill, as long as the model's poor performance is due to high overconfidence rather than due to low potential predictability. Using real seasonal forecasts from the DEMETER data set, it is shown that the conclusions drawn from the toy model experiments equally hold in a real multi-model ensemble prediction system. References: A.P. Weigel et al., 2007a, Mon. Wea. Rev., 135, 2778-2785 A.P. Weigel at al., 2007b: Can multi-model combination really enhance the prediction skill of probabilistic ensemble forecasts? Quart. J. Roy. Met. Soc. Under review