H22A-01 10:20h
Impacts of Mixed Physics on Ensemble Spread in Warm-Season Rainfall Forecasts
A series of tests have been performed using both the Eta and WRF (Weather Research and Forecasting) models over several different domains in the central United States to better understand the role of mixed physics in generating spread in ensemble forecasts of warm season convective systems. In tests with a 10 km grid spacing version of the Eta model, it was found that the choice of convective parameterization exerted a much stronger impact on the resulting rainfall forecast than a wide range of mesoscale initial perturbations. A four member ensemble consisting of two Eta members, each with a different convective scheme, and two WRF members, also with different convective schemes, was found to be as skillful as an 18 member ensemble using the Eta model alone with most members having different initial conditions. Of particular interest, spread was greater among the two Eta members or two WRF members using different convective schemes than it was between the two different models using the same convective scheme. Additional tests were performed using an 18 member mixed-physics ensemble of 12 km grid spacing WRF simulations run over the International H2O Project domain. In this ensemble, three different convective treatments were used, two different planetary boundary layer schemes, and three different microphysical schemes. Again, the convective treatment had the biggest impact on the forecasts when evaluated subjectively and using spread ratio and correlation coefficient, although a factor separation analysis showed the microphysical scheme choice to exert a big influence on total domain rain volume. Other tests over a larger domain were performed at the WRF Developmental Test Center using 8 and 10 km grid spacing versions of the WRF model with different physics, initial conditions and dynamic cores. These tests suggest that initial conditions play a prominent role in only the first 6-12 hours of the forecast, with the dynamic core choice having its biggest impact after 12 hours, and physical scheme choice having a strong influence throughout the 48 hour simulations.
H22A-02 10:35h
Probabilistic Runoff Forecasting using a Limited-Area Ensemble Prediction System
A high-resolution atmospheric ensemble forecasting system, based on 51 runs of a Limited Area Model (LAM) has been used to make probabilistic runoff forecasts for the Alpine Rhine basin. The operational European Centre for Medium-Range Weather Forecasts Ensemble Prediction System (ECMWF EPS) provides the initial and boundary conditions for the LAM integrations with the Local Model (LM) for a 5 day forecasting period. The LM runs in a horizontal resolution of 0.09 degree (10 km) and provides output with an hourly interval. Output from this model is used to drive a distributed hydrological model with a resolution of 500 m and a time-step of one hour. Runoff generation in the Precipitation Runoff EVApotranspiration Hydrotope (PREVAH) model is based on the HBV-model. The model further contains modules, which calculate snow and glacier melt, after a combined radiation and temperature index approach. The case-study investigated is the spring 1999 flood event, when a combination of snowmelt and heavy precipitation caused severe floods in Central Europe. The area investigated is the upper part of the Rhine catchment (34550 km2) in Central Europe. This river catchment, characterized by highly complex topography, has an altitude range from 262 m up to 4225 m a.s.l. The hydrological model component has been calibrated for the period 1997-1998 using ground observations, and validated for 1999-2002. This study focuses on the feasibility of ensemble prediction data for runoff forecasting and addresses the predictability of this flood event. Forecast uncertainties are investigated and runoff predictions from the deterministic forecast are compared with those obtained from probabilistic atmospheric forecasts. Further analyses include the changes in predictability when using different quantities of representative members.
H22A-03 10:50h
Operational Short-Term Flood Forecasting for Bangladesh: Application of ECMWF Ensemble Precipitation Forecasts
The country of Bangladesh frequently experiences severe catchment-scale flooding from the combined discharges of the Ganges and Brahmaputra rivers. Beginning in 2003, we have been disseminating upper-catchment discharge forecasts for this country to provide advanced warning for evacuation and relief measures. These forecasts are being generated using the European Centre for Medium-Range Weather Forecasting (ECMWF) shortterm ensemble weather forecasts and a combination of distributed and data-based modeling techniques. The forecasts from each of these models are combined using the multi-ensemble technique commonly employed in numerical weather prediction. This leads to a reduction in the overall forecast error and capitalizes on the strengths of each model during different periods of the monsoon season. In addition, the models are combined such that the probabilistic nature of the ensemble precipitation forecasts is retained while being combined with the discharge modeling error to produce true probabilistic forecasts of discharge that are being employed operationally.
http://cfab.eas.gatech.edu/shortterm/
H22A-04 11:05h
OPERATIONAL STREAMFLOW FORECASTS DEVELOPMENT USING GCM PREDICTED PRECIPITATION FIELDS
Monthly updates of streamflow forecasts are required for deriving reservoir operation strategies as well as for quantifying surplus and shortfall for the allocated water contracts. In this study, an operational streamflow forecasts are developed using Atmospheric General Circulation Models (AGCM) predicted precipitation for managing the Angat Reservoir System, Philippines. The methodology employs principal components regression (PCR) for downscaling the AGCM predicted precipitation fields to monthly streamflow forecasts. The performance of this downscaling approach is analyzed with AGCM being forced using the observed sea surface temperature (SST) conditions as well under persisted SST conditions. The ability of downscaled streamflow forecasts in explaining the intraseasonal variability is also explored. Conditional distribution of streamflows obtained from the PCR downscaling approach is also compared with a simple, semi-parametric resampling algorithm that obtains ensembles of streamflow forecasts by identifying similar conditions in that season's climatic predictors state space.
H22A-05 11:20h
A Coupled Land-Surface-Groundwater Scheme to Study the Influence of Subsurface Heterogeneity on Mass and Energy fluxes at the Land Surface
In this study, we coupled the Community Land Model (CLM) with the fully three-dimensional (3D) variable saturated groundwater flow code Parflow (PF). The coupled scheme, CLM.PF, solves the 2D distributed energy and mass balance equations at the land surface and the 3D Richards equation for a heterogeneous subsurface. The advantages of CLM.PF are quite obvious. Application of this code allows us to assess simplified subsurface parameterizations that are generally used in land-surface models. Also, the impact of the strongly abstracted upper boundary condition (i.e. the land surface) of many widely used groundwater flow codes (e.g. MODFLOW) can be assed using CLM.PF. Here, emphasize is placed on the influence of heterogeneity in the hydraulic properties of the subsurface on the mass and energy fluxes at the land surface. We generate equally likely realizations of the subsurface using transition probability geostatistics and study the ensemble response of the system to spatially uniform atmospheric forcing. Special attention is given to the dependence of land-surface fluxes on the depth and spatial distribution of the free water table, which is not explicitly taken into account by many land-surface schemes. Preliminary results suggest that variable surface fluxes are a direct result of subsurface heterogeneity. The effects of heterogeneity are especially pronounced under dry conditions and heavy precipitation events. The position of the water table naturally depends on the hydraulic properties distribution and plays a key role in the generation of spatially non-uniform surface runoff and evaporation. In the future, CLM.PF will be used in the development of effective or lumped models and in studies of upscaling. This work was conducted under the auspices of the U. S. Department of Energy by the University of California, Lawrence Livermore National Laboratory (LLNL) under contract W-7405-Eng-48 and Lawrence Berkeley National Laboratory (LBNL) under contract DE-AC03-76F00098.
H22A-06 11:35h
A nonparametric weather-state approach for downscaling of multi-site precipitation occurrences
The physical linkages between climate on the large scale and weather on the local scale form the basis of downscaling approaches for assessing the impact of climate variability at point locations. The common approach frequently used for downscaling of precipitation, considers discrete weather classes of the atmospheric patterns and simulates precipitation conditioned on these weather-states. This paper presents the development of a nonparametric weather-state downscaling approach (KNN-W) and its comparison with traditional KNN resampling approach (KNN) and a parametric Non-homogenous Hidden Markov Model (NHMM). The KNN-W defines local scale weather as a function of a weather state that is continuous and auto-regressive in nature and depends on predictor variables representing synoptic atmospheric patterns. Such a formulation offers a simpler alternative to the weather-state based parametric approaches like NHMM. The KNN resampling approach considers a direct probabilistic relationship between the larger scale climatic variables and the local scale weather. On the other hand, the weather-state KNN downscaling approach being structured on continuous weather-state formulation is more opt at representing temporal persistence. A weather-state of KNN-W is defined based on spatial rainfall distribution over the study region. The paper also considers the relative influence of atmospheric circulation variables on the conditional density formulation in the form of influence weights. In the comparison presented here, we applied these downscaling approaches conditional on four atmospheric circulation variables, to estimate precipitation occurrences at a network of 30 raingauge locations around Sydney, Australia. Our results suggest that all the models perform well at representing spatial variations while they lack at representing temporal dependence at scales longer than a few days as exhibited through wet spell length characteristics. The weather-state based KNN approach is more successful in capturing the longer duration as well as extreme rainfall characteristics in comparison to direct KNN approach and NHMM. Local scale features that are difficult to represent through the large scale climate predictors are expectedly not reproduced by any approach.
H22A-07 11:50h
A Bayesian Methodology for Ensemble Forecasting
A Bayesian methodology for probabilistic forecasting of a future river stage time series in terms of an ensemble is presented. The methodology derives from the theory of the Bayesian Forecasting System (BFS). Within the BFS, uncertainty due to future precipitation is quantified in a precipitation uncertainty processor independently of other uncertainties. The other uncertainties are aggregated and quantified in a hydrologic uncertainty processor. Then the precipitation uncertainty and the hydrologic uncertainty are integrated together in an integrator. The resultant probabilistic forecast quantifies the total uncertainty. This paper presents two algorithms for generating an ensemble forecast using the BFS. The first algorithm uses output distributions from the analytic-numerical BFS to recursively generate a river stage time series; this algorithm is suited to headwater basins. The second algorithm implements the precipitation uncertainty processor, hydrologic uncertainty processor, and integrator as Monte Carlo generators, which sequentially process a precipitation time series into a river stage time series; this algorithm is suited to complex river basins. The algorithms are illustrated with numerical examples of Bayesian ensemble forecasts for several different forecasting scenarios. Properties and advantages of the Bayesian forecasts for decision making are highlighted. Sample sizes required for correct representation of uncertainty are examined.
H22A-08 12:05h
Application of Multi-Model Superensemble technique to flood forecasting through distributed hydrologic models
Streamflow forecasts are generally produced through the use of a single hydrologic model. In spite of the existence of a wide range of hydrologic models, it is hard to claim that any single model among them performs better than the rest, for all type of watersheds under all conditions. This is because hydrologic models, lumped or distributed, introduce many assumptions and simplifications in their structure. Since various model structures capture different aspects of the watershed processes, one way of exploiting the strength of different models and compensating for their weaknesses is to obtain consensus predictions by combining their results using model combination techniques such as Multi Model SuperEnsmble (MMSE). MMSE is a special case of ensemble techniques, which consider the model outputs as ensemble members. This study surveys the performance of MMSE for flood forecasting by using the simulation results from various distributed models participated in the Distributed Model Intercomparison Project (DMIP), an international project sponsored by National Weather Service. The key questions addressed in this study are: (1) What is the skill level of the consensus forecast compared to those of individual forecasts? (2) How many models do we need to produce accurate consensus forecasts? (3) Can model combination techniques compensate for the inadequacy of model calibration? Simulations for the Illinois River Basin at Watts from 7 uncalibrated DMIP models are combined and the results are compared to the calibrated model results.