Bias-corrected Ensemble Research Articles

Assessing the impacts of climate change on streamflow in the upper Omo Gibe basin of Southwest Ethiopia using MIKE NAM

ABSTRACT Climate change affecting the rainfall-runoff processes in the upper Omo Gibe river basin was articulated using bias-corrected ensemble simulations of the CORDEX-Africa under RCP4.5 and RCP8.5 climatic scenarios. CMhyd was used to downscale the ensembles statistically followed by bias-adjustment using distribution mapping. MIKE NAM model was used (1980-2005) to project streamflow for three time periods; the near future (2023-2048), the middle future (2049-2074), and the far future (2075-2100). The average temperature of the basin will rise by 1.59°C and 2.56°C, respectively, under RCP4.5 and RCP8.5, respectively, while, the precipitation for wet and dry seasons decreases by 7.97% and 23.67%, respectively. MIKE NAM has demonstrated a strong correlation between simulated and observed runoff with coefficients of determination (R2) and NSE of more than 0.75. It has been predicted that the mean dry season, wet season, and annual streamflow will decline by 15-27%, 4-13%, and 8-24%, respectively. Intricate study of climate change followed by judicious and timely precautionary measures if not taken today, will decrease in streamflow and put negative consequences on long-term irrigation and hydropower development in the river catchment of upper Omo Gibe basin.

A Multi-Scale Forest Above-Ground Biomass Mapping Approach: Employing a Step-by-Step Spatial Downscaling Method with Bias-Corrected Ensemble Machine Learning

The accurate estimation of forest above-ground biomass (AGB) is vital for monitoring changes in forest carbon sinks. However, the spatial heterogeneity of AGB, coupled with inherent uncertainties, poses challenges in acquiring high-quality AGBs. This study introduced a bias-corrected ensemble machine learning (ML) algorithm for AGB downscaling that integrated a ML for AGB mapping with another for residual mapping. The accuracies of six bias-corrected ensemble ML algorithms were evaluated at resolutions of 0.05°, 0.025°, and 0.01°. Moreover, a step-by-step downscaling (SBSD) method was introduced, utilizing bias-corrected ensemble ML algorithms to downscale AGB from 0.1° to 0.05°, 0.025°, and 0.01° resolutions and was compared with the direct downscaling (DD) at three scales. A comparative analysis was conducted in the Daxing’anling Mountains and Xiaoxing’anling Mountains. AGB and corresponding uncertainty maps at three scales were generated using SBSD. The results showed that the efficacy of the XGBoost-based AGB model combined with the random forest-based residual correction model was superior. Spatial patterns in AGB maps generated by SBSD and DD were found to be similar. Notably, SBSD yielded enhanced accuracy in the Daxing’anling Mountains with complex topography, while both performed comparably in the Xiaoxing’anling Mountains with milder topography, highlighting SBSD’s advantages in high heterogeneity areas.

Assessment of historical and projected changes in extreme temperatures of Balochistan, Pakistan using extreme value theory.

The fundamental consequences of global warming include an upsurge in the intensity and frequency of temperature extremes. This study provides an insight into historical trends and projected changes in extreme temperatures on annual and seasonal scales across "Balochistan, Pakistan". Historical trends are analyzed through the Mann Kendal test, and extreme temperatures (Tmax and Tmin) are evaluated using generalized extreme value (GEV) distribution for historicalperiod (1991-2020) from the observational data and the two projected periods as near-future (2041-2070) and far-future (2071-2100) using a six-member bias-corrected ensemble of regional climate models (RCMs) projections from the coordinate regional downscaling experiment (CORDEX) based on the worst emission scenario (RCP8.5). The evaluation of historical temperature trends suggests that Tmax generally increase on yearly scale and give mixed signals on seasonal scale (winter, spring, summer, and autumn); however, Tmin trends gave mixed signals at both yearly and seasonal scale. Compared to the historical period, the return levels are generally expected to be higher for Tmax and Tmin during the both projection periods in the order as far-future > near-future > historical on yearly and seasonal basis; however, the changes in Tmin are more evident. Station-averaged anomalies of + 1.9°C and + 3.6°C were estimated in 100-year return levels for yearly Tmax for near-future and far-future, respectively, while the anomalies in Tmin were found to be + 3.5°C and + 4.8°C which suggest the intensified heatwaves but milder colder extreme in future. The findings provide guidance on improved quantification of changing frequencies and severity in temperature extremes and the associated impacts.

Assessment on future rainfall variability for adaptive water resource management in Sabah

Climate is changing at an unprecedented rate, and the increasing demands of the growing population will place further pressures on the climate and environment. Water infrastructures such as water treatment plants often face increased stress during extreme weather events such as heavy rainfall or prolonged drought, disrupting societal needs and prompt urgent upgrades to adapt to the changing climate. The use of climate model projections is increasingly adopted in engineering practices for assessing climate risks and impacts to inform adaptive design and strategy of new infrastructures or upgrades. Using a bias-corrected ensemble member of the fifth phase of the Coupled Model Intercomparison Project (CMIP5) tailored for applications in the water sector, we assessed the variability of projected rainfall from 2021 to 2050 under the moderate warming scenario (RCP 4.5). The findings from this study aim to inform adaptive strategies for upgrading 22 water treatment plants across 18 sites in Sabah, Malaysia, where these plants have experienced numerous disruptions due to high turbidity resulting from heavy rainfall in recent years. The average annual rainfall amount ranges from 1630 mm to 4415 mm, with average total monthly rainfall peaks (>460 mm) in June and August at most sites. The observed shift in June and August is a deviation from historical patterns and is also reflected in the maximum consecutive 1- to 5-day rainfall, ranging from 22 mm to 760 mm across all sites, particularly in the interior and west coast districts. This highlights the need for resilient infrastructure upgrade strategy and adaptive water resource management, such as the introduction of off-site river storage to accommodate the anticipated future rainfall.

Development of daily bias-corrected ensemble precipitation estimates over the Upper Indus Basin of the Hindukush-Karakoram-Himalaya

Abstract Accurate precipitation estimates over space and time are critically important, particularly in data-scarce areas, for effective hydrological modeling and efficient regional water resources management. Gridded precipitation datasets are the preeminent alternative in such areas. However, gridded precipitation datasets contain different kinds of uncertainties owing to the retrieval algorithms used in their development. In this study, five precipitation datasets (Tropical Rainfall Measuring Mission (TRMM), Climate Prediction Centre (CPC), APHRODITE, Climate Hazards Group Infra-Red Precipitation with Station data (CHIRPS), and PERSIANN) were evaluated, and an ensemble of daily precipitation datasets from 2001 to 2017 at a resolution of 0.05 degree was created based on three ensemble approaches (Bayesian model ensemble, relative bias-based ensemble, and correlation-based ensemble) over the Upper Indus basin. To improve the accuracy of the ensemble dataset, a linear bias correction technique is applied with respect to gauging precipitation. The accuracy of the bias-corrected ensemble dataset was evaluated using statistical and novelty categorical measures. A reasonable agreement was found between the ensemble and gauge precipitation (Pearson correlation 0.83–0.89 and relative bias 1–8.7 mm/month), while large biases were noted in five precipitation datasets (1.7–53.9 mm/month). The study suggests that utilizing ensemble approaches to gridded precipitation can significantly enhance the accuracy of the estimates compared to relying on a single precipitation dataset.

Downslope Windstorm Forecasting: Easier with a Critical Level, but Still Challenging for High-Resolution Ensembles

Abstract Strong downslope windstorms can cause extensive property damage and extreme wildfire spread, so their accurate prediction is important. Although some early studies suggested high predictability for downslope windstorms, more recent analyses have found limited predictability for such winds. Nevertheless, there is a theoretical basis for expecting higher downslope wind predictability in cases with a mean-state critical level, and this is supported by one previous effort to forecast actual events. To more thoroughly investigate downslope windstorm predictability, we compare archived simulations from the NCAR ensemble, a 10-member mesoscale ensemble run at 3-km horizontal grid spacing over the entire contiguous United States, to observed events at 15 stations in the western United States susceptible to strong downslope winds. We assess predictability in three contexts: the average ensemble spread, which provides an estimate of potential predictability; a forecast evaluation based upon binary-decision criteria, which is representative of operational hazard warnings; and a probabilistic forecast evaluation using the continuous ranked probability score (CRPS), which is a measure of an ensemble’s ability to generate the proper probability distribution for the events under consideration. We do find better predictive skill for the mean-state critical-level regime in comparison to other downslope windstorm–generating mechanisms. Our downslope windstorm warning performance, calculated using binary-decision criteria from the bias-corrected ensemble forecasts, performed slightly worse for no-critical-level events, and slightly better for critical-level events, than National Weather Service high-wind warnings aggregated over all types of high-wind events throughout the United States and annually averaged for each year between 2008 and 2019.

Future projections of wind and solar energy resources over China from regional climate models based on bias correction

In this study, the wind and solar energy resources over China are projected during 2031–2060 under the Representative Concentration Pathway (RCP) 8.5 scenario with the Weather Research and Forecasting (WRF) model and Regional Climate Model version 4 (RegCM4). The bias correction method based on probability distribution is applied to correct the simulated 100-m wind speed and surface down-welling shortwave radiation. The results show that the bias correction method can remarkably improve the skill of the individual simulations and the ensemble mean in capturing the features of wind speed and shortwave radiation from the aspects of spatial distribution, annual cycle, quantile distribution, probability distribution and cumulative probability distribution. The future projections from the bias-corrected ensemble show that both the wind and solar energy resources are abundant over northwestern China, with higher wind power density reaching up to 130 W m−2 in winter and richer surface down-welling shortwave radiation above 240 W m−2 in summer. Generally, the wind and solar energy resources will decrease in the future, by around 6% over the western regions and 4% over southeastern China in winter, respectively. The interannual variability of the two resources is projected to increase over most areas of northern China, indicating a decrease in the stability of resources, while tends to decrease over Southeastern China.

Ensemble Projection of Future Climate and Surface Water Supplies in the North Saskatchewan River Basin above Edmonton, Alberta, Canada

Changes in temperature and precipitation are expected to alter the seasonal distribution of surface water supplies in snowmelt-dominated watersheds. A realistic assessment of future climate change and inter-annual variability is required to meet a growing demand for water supplies in all major use sectors. This study focuses on changes in climate and runoff in the North Saskatchewan River Basin (NSRB) above Edmonton, AB, Canada, using the MESH (Modélisation Environnementale communautaire—Surface Hydrology) model. The bias-corrected ensemble of Canadian Regional Climate Model (CanRCM4) data is used to drive MESH for two 60-year time periods, a historical baseline (1951–2010) and future projection (2041–2100), under Representative Concentration Pathway (RCP) 8.5. The precipitation is projected to increase in every season, there is significant trend in spring (0.62) and fall (0.41) and insignificant in summer (0.008). Winter extreme minimum temperature and summer extreme maximum temperature are increasing by 2–3 °C in the near future and 5–6 °C in the far future. Annual runoff increases by 19% compared to base period. The results reveal long-term hydrological variability enabling water resource managers to better prepare for climate change and extreme events to build more resilient systems for future water demand in the NSRB.

Time-Dependent Downscaling of PM2.5 Predictions from CAMS Air Quality Models to Urban Monitoring Sites in Budapest

Budapest, the capital of Hungary, has been facing serious air pollution episodes in the heating season similar to other metropolises. In the city a dense urban air quality monitoring network is available; however, air quality prediction is still challenging. For this purpose, 24-h PM2.5 forecasts obtained from seven individual models of the Copernicus Atmosphere Monitoring Service (CAMS) were downscaled by using hourly measurements at six urban monitoring sites in Budapest for the heating season of 2018–2019. A 10-day long training period was applied to fit spatially consistent model weights in a linear combination of CAMS models for each day, and the 10-day additive bias was also corrected. Results were compared to the CAMS ensemble median, the 10-day bias-corrected CAMS ensemble median, and the 24-h persistence. Downscaling reduced the root mean square error (RMSE) by 1.4 µg/m3 for the heating season and by 4.3 µg/m3 for episodes compared to the CAMS ensemble, mainly by eliminating the general underestimation of PM2.5 peaks. As a side-effect, an overestimation was introduced in rapidly clearing conditions. Although the bias-corrected ensemble and model fusion had similar overall performance, the latter was more efficient in episodes. Downscaling of the CAMS models was found to be capable and necessary to capture high wintertime PM2.5 concentrations for the short-range air quality prediction in Budapest.

Rainfall Variability under Present and Future Climate Scenarios Using the Rossby Center Bias-Corrected Regional Climate Model

This study sought to determine the spatial and temporal variability of rainfall under past and future climate scenarios. The data used comprised station-based monthly gridded rainfall data sourced from the Climate Research Unit (CRU) and monthly model outputs from the Fourth Edition of the Rossby Centre (RCA4) Regional Climate Model (RCM), which has scaled-down nine GCMs for Africa. Although the 9 Global Climate Models (GCMs) downscaled by the RCA4 model was not very good at simulating rainfall in Kenya, the ensemble of the 9 models performed better and could be used for further studies. The ensemble of the models was thus bias-corrected using the scaling method to reduce the error; lower values of bias and Normalized Root Mean Square Error (NRMSE) were recorded when compared to the uncorrected models. The bias-corrected ensemble was used to study the spatial and temporal behaviour of rainfall under baseline (1971 to 2000) and future RCP 4.5 and 8.5 scenarios (2021 to 2050). An insignificant trend was noted under the baseline condition during the March-May (MAM) and October-December (OND) rainfall seasons. A positive significant trend at 5% level was noted under RCP 4.5 and 8.5 scenarios in some stations during both MAM and OND seasons. The increase in rainfall was attributed to global warming due to increased anthropogenic emissions of greenhouse gases. Results on the spatial variability of rainfall indicate the spatial extent of rainfall will increase under both RCP 4.5 and RCP 8.5 scenario when compared to the baseline; the increase is higher under the RCP 8.5 scenario. Overall rainfall was found to be highly variable in space and time, there is a need to invest in the early dissemination of weather forecasts to help farmers adequately prepare in case of unfavorable weather. Concerning the expected increase in rainfall in the future, policymakers need to consider the results of this study while preparing mitigation strategies against the effects of changing rainfall patterns.

Using Bayesian Model Averaging to Calibrate Temperature Forecast Ensembles in Hawassa Town, Ethiopia

Bayesian model averaging (BMA) is a statistical way of post-processing forecasts ensembles to create predictive probability density functions (pdfs) for weather quantities. BMA has been proposed as way of correcting underdispersion in ensemble forecasts. The output of ensemble BMA is a weighted average of pdfs centered on the individual bias-corrected ensemble forecast. The BMA weights are posterior probabilities of the models generating the forecast, and show the relative contribution of the component models to the prediction over the training period. This paper focuses on short range forecasting of daily maximum temperature. Forecasting temperature, we have approximated the conditional pdf by normal distribution centered at a linear function of the forecast. Since we are not able to get an ensemble of forecasts for Ethiopia, we have used the previous eight years from 2003 to 2010 data as ensemble and 2011 as observation. The BMA approach yields predictive distribution for temperature thatwere much better calibrated than those based on naive averaging of the forecasts.

Development of Verification Methodology for Extreme Weather Forecasts

AbstractIn 2006, the statistical postprocessing of the National Centers for Environmental Prediction (NCEP) Global Ensemble Forecast System (GEFS) and North American Ensemble Forecast System (NAEFS) was implemented to enhance probabilistic guidance. Anomaly forecasting (ANF) is one of the NAEFS products, generated from bias-corrected ensemble forecasts and reanalysis climatology. The extreme forecast index (EFI), based on a raw ensemble forecast and model-based climatology, is another way to build an extreme weather forecast. In this work, the ANF and EFI algorithms are applied to extreme cold temperature and extreme precipitation forecasts during the winter of 2013/14. A highly correlated relationship between the ANF and EFI allows the determination of two sets of thresholds to identify extreme cold and extreme precipitation events for the two algorithms. An EFI of −0.78 (0.687) is approximately equivalent to a −2σ (0.95) ANF for the extreme cold event (extreme precipitation) forecast. The performances of the two algorithms in forecasting extreme cold events are verified against analysis for different model versions, reference climatology, and forecasts. The verification results during the winter of 2013/14 indicate that ANF forecasts more extreme cold events with a slightly higher skill than EFI. The bias-corrected forecast performs much better than the raw forecast. The current upgrade of the GEFS has a beneficial effect on the extreme cold weather forecast. Using the NCEP Climate Forecast System Reanalysis and Reforecast (CFSRR) as a climate reference gives a slightly better score than the 40-yr reanalysis. The verification methodology is also extended to an extreme precipitation case, showing a broad potential use in the future.

Bias-Corrected Short-Range Ensemble Forecasts for Near-Surface Variables during the Summer Season of 2010 in Northern China

AbstractA running mean bias (RMB) correction approach was applied to the forecasts of near-surface variables in a seasonal short-range ensemble forecasting experiment with 57 consecutive cases during summer 2010 in the northern China region. To determine a proper training window length for calculating RMB, window lengths from 2 to 20 days were evaluated, and 16 days was taken as an optimal window length, since it receives most of the benefit from extending the window length. The raw and 16-day RMB corrected ensembles were then evaluated for their ensemble mean forecast skills. The results show that the raw ensemble has obvious bias in all near-surface variables. The RMB correction can remove the bias reasonably well, and generate an unbiased ensemble. The bias correction not only reduces the ensemble mean forecast error, but also results in a better spreaderror relationship. Moreover, two methods for computing calibrated probabilistic forecast (PF) were also evaluated through the 57 case dates: 1) using t...

Precipitation and Temperature Forecast Performance at the Weather Prediction Center

Abstract The role of the human forecaster in improving upon the accuracy of numerical weather prediction is explored using multiyear verification of human-generated short-range precipitation forecasts and medium-range maximum temperature forecasts from the Weather Prediction Center (WPC). Results show that human-generated forecasts improve over raw deterministic model guidance. Over the past two decades, WPC human forecasters achieved a 20%–40% improvement over the North American Mesoscale (NAM) model and the Global Forecast System (GFS) for the 1 in. (25.4 mm) (24 h)−1 threshold for day 1 precipitation forecasts, with a smaller, but statistically significant, 5%–15% improvement over the deterministic ECMWF model. Medium-range maximum temperature forecasts also exhibit statistically significant improvement over GFS model output statistics (MOS), and the improvement has been increasing over the past 5 yr. The quality added by humans for forecasts of high-impact events varies by element and forecast projection, with generally large improvements when the forecaster makes changes ≥8°F (4.4°C) to MOS temperatures. Human improvement over guidance for extreme rainfall events [3 in. (76.2 mm) (24 h)−1] is largest in the short-range forecast. However, human-generated forecasts failed to outperform the most skillful downscaled, bias-corrected ensemble guidance for precipitation and maximum temperature available near the same time as the human-modified forecasts. Thus, as additional downscaled and bias-corrected sensible weather element guidance becomes operationally available, and with the support of near-real-time verification, forecaster training, and tools to guide forecaster interventions, a key test is whether forecasters can learn to make statistically significant improvements over the most skillful of this guidance. Such a test can inform to what degree, and just how quickly, the role of the forecaster changes.

Spread Calibration of Ensemble MOS Forecasts

Abstract Ensemble forecasting systems often contain systematic biases and spread deficiencies that can be corrected by statistical postprocessing. This study presents an improvement to an ensemble statistical postprocessing technique, called ensemble kernel density model output statistics (EKDMOS). EKDMOS uses model output statistics (MOS) equations and spread–skill relationships to generate calibrated probabilistic forecasts. The MOS equations are multiple linear regression equations developed by relating observations to ensemble mean-based predictors. The spread–skill relationships are one-term linear regression equations that predict the expected accuracy of the ensemble mean given the ensemble spread. To generate an EKDMOS forecast, the MOS equations are applied to each ensemble member. Kernel density fitting is used to create a probability density function (PDF) from the ensemble MOS forecasts. The PDF spread is adjusted to match the spread predicted by the spread–skill relationship, producing a calibrated forecast. The improved EKDMOS technique was used to produce probabilistic 2-m temperature forecasts from the North American Ensemble Forecast System (NAEFS) over the period 1 October 2007–31 March 2010. The results were compared with an earlier spread adjustment technique, as well as forecasts generated by rank sorting the bias-corrected ensemble members. Compared to the other techniques, the new EKDMOS forecasts were more reliable, had a better calibrated spread–error relationship, and showed increased day-to-day spread variability.

Bias Correction for Global Ensemble Forecast

Abstract The main task of this study is to introduce a statistical postprocessing algorithm to reduce the bias in the National Centers for Environmental Prediction (NCEP) and Meteorological Service of Canada (MSC) ensemble forecasts before they are merged to form a joint ensemble within the North American Ensemble Forecast System (NAEFS). This statistical postprocessing method applies a Kalman filter type algorithm to accumulate the decaying averaging bias and produces bias-corrected ensembles for 35 variables. NCEP implemented this bias-correction technique in 2006. NAEFS is a joint operational multimodel ensemble forecast system that combines NCEP and MSC ensemble forecasts after bias correction. According to operational statistical verification, both the NCEP and MSC bias-corrected ensemble forecast products are enhanced significantly. In addition to the operational calibration technique, three other experiments were designed to assess and mitigate ensemble biases on the model grid: a decaying averaging bias calibration method with short samples, a climate mean bias calibration method, and a bias calibration method using dependent data. Preliminary results show that the decaying averaging method works well for the first few days. After removing the decaying averaging bias, the calibrated NCEP operational ensemble has improved probabilistic performance for all measures until day 5. The reforecast ensembles from the Earth System Research Laboratory’s Physical Sciences Division with and without the climate mean bias correction were also examined. A comparison between the operational and the bias-corrected reforecast ensembles shows that the climate mean bias correction can add value, especially for week-2 probability forecasts.

Comparison of ensemble methods for summer-time numerical weather prediction over East Asia

Summer-time short- to medium-range predictability of precipitation, 500-hPa geopotential height, and wind fields over East Asia are investigated by comparing three ensemble forecast configurations: multi-analysis, multi-convection, and multi-model. These three systems are used in this study in order to assess initial condition uncertainties, model uncertainties, and a combination of initial condition and model uncertainties in an ensemble forecast approach. Each system has a set of six members. Ensemble forecast skill is verified in both deterministic and probabilistic senses using the European Center for Medium-range Weather Forecasting analyses and the Tropical Rainfall Measuring Mission Microwave Imager 2A12 rain estimates. The multi-model configuration, which considers both the initial condition and model uncertainties to predict weather phenomena over East Asia, is an optimal set of ensemble members. The bias-corrected ensemble and the superensemble (SE) show similar predictability, but slightly better skill is obtained from the SE forecasts.

A Comparison between Raw Ensemble Output, (Modified) Bayesian Model Averaging, and Extended Logistic Regression Using ECMWF Ensemble Precipitation Reforecasts

Abstract Using a 20-yr ECMWF ensemble reforecast dataset of total precipitation and a 20-yr dataset of a dense precipitation observation network in the Netherlands, a comparison is made between the raw ensemble output, Bayesian model averaging (BMA), and extended logistic regression (LR). A previous study indicated that BMA and conventional LR are successful in calibrating multimodel ensemble forecasts of precipitation for a single forecast projection. However, a more elaborate comparison between these methods has not yet been made. This study compares the raw ensemble output, BMA, and extended LR for single-model ensemble reforecasts of precipitation; namely, from the ECMWF ensemble prediction system (EPS). The raw EPS output turns out to be generally well calibrated up to 6 forecast days, if compared to the area-mean 24-h precipitation sum. Surprisingly, BMA is less skillful than the raw EPS output from forecast day 3 onward. This is due to the bias correction in BMA, which applies model output statistics to individual ensemble members. As a result, the spread of the bias-corrected ensemble members is decreased, especially for the longer forecast projections. Here, an additive bias correction is applied instead and the equation for the probability of precipitation in BMA is also changed. These modifications to BMA are referred to as “modified BMA” and lead to a significant improvement in the skill of BMA for the longer projections. If the area-maximum 24-h precipitation sum is used as a predictand, both modified BMA and extended LR improve the raw EPS output significantly for the first 5 forecast days. However, the difference in skill between modified BMA and extended LR does not seem to be statistically significant. Yet, extended LR might be preferred, because incorporating predictors that are different from the predictand is straightforward, in contrast to BMA.

Calibrating Multimodel Forecast Ensembles with Exchangeable and Missing Members Using Bayesian Model Averaging

Abstract Bayesian model averaging (BMA) is a statistical postprocessing technique that generates calibrated and sharp predictive probability density functions (PDFs) from forecast ensembles. It represents the predictive PDF as a weighted average of PDFs centered on the bias-corrected ensemble members, where the weights reflect the relative skill of the individual members over a training period. This work adapts the BMA approach to situations that arise frequently in practice; namely, when one or more of the member forecasts are exchangeable, and when there are missing ensemble members. Exchangeable members differ in random perturbations only, such as the members of bred ensembles, singular vector ensembles, or ensemble Kalman filter systems. Accounting for exchangeability simplifies the BMA approach, in that the BMA weights and the parameters of the component PDFs can be assumed to be equal within each exchangeable group. With these adaptations, BMA can be applied to postprocess multimodel ensembles of any composition. In experiments with surface temperature and quantitative precipitation forecasts from the University of Washington mesoscale ensemble and ensemble Kalman filter systems over the Pacific Northwest, the proposed extensions yield good results. The BMA method is robust to exchangeability assumptions, and the BMA postprocessed combined ensemble shows better verification results than any of the individual, raw, or BMA postprocessed ensemble systems. These results suggest that statistically postprocessed multimodel ensembles can outperform individual ensemble systems, even in cases in which one of the constituent systems is superior to the others.

Bias-Corrected Short-Range Ensemble Forecasts of Near-Surface Variables during the 2005/06 Cool Season

Abstract A postprocessing method initially developed to improve near-surface forecasts from a summertime multimodel short-range ensemble forecasting system is evaluated during the cool season of 2005/06. The method, known as the bias-corrected ensemble (BCE) approach, uses the past complete 12 days of model forecasts and surface observations to remove the mean bias of near-surface variables from each ensemble member for each station location and forecast time. In addition, two other performance-based weighted-average BCE schemes, the exponential smoothing method BCE and the minimum variance estimate BCE, are implemented and evaluated. Values of root-mean-squared error from the 2-m temperature and dewpoint temperature forecasts indicate that the BCE approach outperforms the routinely available Global Forecast System (GFS) model output statistics (MOS) forecasts during the cool season by 9% and 8%, respectively. In contrast, the GFS MOS provides more accurate forecasts of 10-m wind speed than any of the BCE methods. The performance-weighted BCE schemes yield no significant improvement in forecast accuracy for 2-m temperature and 2-m dewpoint temperature when compared with the original BCE, although the weighted BCE schemes are found to improve the forecast accuracy of the 10-m wind speed. The probabilistic forecast guidance provided by the BCE system is found to be more reliable than the raw ensemble forecasts. These results parallel those obtained during the summers of 2002–04 and indicate that the BCE method is a promising and inexpensive statistical postprocessing scheme that could be used in all seasons.