Multimodel Ensemble Forecasts Research Articles

Hydrological Model Diversity Enhances Streamflow Forecast Skill at Short‐ to Medium‐Range Timescales

AbstractWe investigate the ability of hydrological multimodel ensemble predictions to enhance the skill of streamflow forecasts at short‐ to medium‐range timescales. To generate the multimodel ensembles, we implement a new statistical postprocessor, namely, quantile regression‐Bayesian model averaging (QR‐BMA). Quantile regression‐Bayesian model averaging uses quantile regression to bias correct the ensemble streamflow forecasts from the individual models and Bayesian model averaging to optimally combine their probability density functions. Additionally, we use an information‐theoretic measure, namely, conditional mutual information, to quantify the skill enhancements from the multimodel forecasts. We generate ensemble streamflow forecasts at lead times from 1 to 7 days using three hydrological models: (i) Antecedent Precipitation Index‐Continuous, (ii) Hydrology Laboratory‐Research Distributed Hydrologic Model, and (iii) Weather Research and Forecasting Hydrological modeling system. As forcing to the hydrological models, we use weather ensemble forecasts from the National Centers for Environmental Prediction 11‐member Global Ensemble Forecast System Reforecast version 2. The forecasting experiments are performed for four nested basins of the North Branch Susquehanna River, USA. We find that after bias correcting the streamflow forecasts from each model, their skill performance becomes comparable. We find that the multimodel ensemble forecasts have higher skill than the best single‐model forecasts. Furthermore, the skill enhancements obtained by the multimodel ensemble forecasts are found to be dominated by model diversity, rather than by increased ensemble size alone. This result, obtained using conditional mutual information, indicates that each hydrological model contributes additional information to enhance forecast skill. Overall, our results highlight benefits of hydrological multimodel forecasting for improving streamflow predictions.

Ensemble Sensitivity Analysis of the Blocking System over Russia in Summer 2010

Abstract In summer 2010, the weather conditions in the Euro-Russian sector were affected by a long-lasting atmospheric block that led to a heat wave in Russia and floods in Pakistan. Following previous studies describing the block’s predictability, the present study aims to investigate uncertainties in the upper-level wave pattern and diabatic processes that were responsible for the block’s forecast variability during its onset, mature, and decay phases. With this aim, an ensemble sensitivity analysis (ESA) is performed for three medium-range THORPEX Interactive Grand Global Ensemble multimodel ensemble forecasts, one associated with each phase of the block’s life cycle. The ESA revealed that the block’s predictability was influenced by forecast uncertainties in the general wave pattern and in the vertically integrated water vapor transport (IVT), used here as a proxy for diabatic processes. These uncertainties are associated with spatial shifts and intensity changes of synoptic waves and IVT during the whole life cycle of the block. During the onset phase, specific features include an Atlantic precursor block and the occurrence of several cyclones. During the mature stage, the blocking ridge itself was highly predictable, while forecast uncertainties in the wave pattern and in IVT primarily were associated with uncertainties in the block’s western flank. During the decay phase, the ESA signals were less intense, but the forecast variability significantly depended on the transformation of the block into a high-over-low pattern. It can be concluded that ESA is suitable to investigate the block’s forecast variability in multimodel ensembles.

Short-term Forecasts of TC Tracks over the Western Pacific Using the Multimodel Ensemble Approach

Based on the TC track ensemble forecast data of China Meteorological Administration (CMA), European Centre for Medium-Range Weather Fore-casts (ECMWF), Japan Meteorological Agency (JMA), and United Kingdom Met Office (UKMO) in the TIGGE dataset during the period from 1 May until 31 August 2009, the 24h~72h multimodel ensemble forecasts of the TC tracks over the western Pacific have been conducted by using the multimodel superensemble approaches including the multimodel ensemble mean (EMN), the bias-removed ensemble mean (BREM), and multimodel superensemble in terms of the weighted bias-removed ensemble mean (WEM). A terrifying typhoon Morakot which brought record-breaking rainfalls over Taiwan and caused huge damages in both Taiwan and mainland China was chosen for case study. The results show that forecast skills of the aforementioned four models are quite different for 24h~72h forecasts. The weighted bias-removed ensemble mean and bias-removed ensemble mean reduce the TC track errors considerably. Both of the WEM and BREM techniques show a significant improvement on the forecast skill of TC tracks against the best individual model forecast and the EMN forecast. The multimodel superensemble in terms of the WEM shows the best performance in 24h~72h forecast of TC tracks over the western Pacific. In addition, the multimodel superensemble in terms of the weighted average of different individual model forecasts has better performance than the bias-removed ensemble mean with equal weights of all independent forecasts.

Multi-model Ensemble Forecast of <italic>PM</italic><sub>2.5</sub> Concentration Based on the Improved Wavelet Neural Networks

Multi-model Ensemble Forecast of <italic>PM</italic><sub>2.5</sub> Concentration Based on the Improved Wavelet Neural Networks

North Western Pacific Tropical Cyclone Ensemble Forecast Project

ABSTRACT The WMO/TCP and WWRP launched the North Western Pacific Tropical Cyclone Ensemble Forecast Project (NWP-TCEFP) in 2009 to explore the utility of ensemble forecasts, including multi-model ensemble forecasts of TCs, and to promote such products for operational TC forecasting. Operational global medium-range ensembles, which have been exchanged in real-time in a CXML format under the initiative of the WMO GIFS-TIGGE Working Group, have been used to create ensemble products of TC tracks that were used by the Typhoon Committee Members and forecasters participating in the SWFDP in Southeast Asia through a password-protected website developed and maintained by the JMA. In some cases many or all of the ensembles have simultaneously predicted small or large ensemble spreads in TC tracks. The implication is that multi-model ensemble products provide forecasters with additional information on forecast certainty or uncertainty and thus increase the level of confidence in the forecasts. Another important outcome of the project was the responses to surveys conducted by the WMO/TCP and WWRP and also by the WMO GIFS-TIGGE Working Group. The responses confirmed that ESCAP/WMO Typhoon Committee Members have routinely accessed the website and have recognized the usefulness of the ensemble products available on the website for operational TC forecasting.

Seasonal forecasting of western North Pacific tropical cyclone frequency using the North American multi-model ensemble

This study focuses on the seasonal forecasting of the June–November frequency of the western North Pacific tropical cyclones (WNP TCF) by combining observations and the North-American multi-model ensemble (NMME) forecasts. We model the interannual variability in WNP TCF using Poisson regression with two sea surface temperature (SST)-based predictors, the first empirical orthogonal function of SST in the Pacific meridional mode (PMM) region (“PSPMM”) after linearly removing the impacts of the El Nino Southern Oscillation, and the SST anomalies averaged over the North Atlantic Ocean (“NASST”). The Poisson regression model trained with the observed WNP TCF and the two predictors exhibits a high skill (correlation coefficient of 0.73 and root mean square error equal to 3.2 TC/year) for the 1965–2016 period. Using SST forecasts by eight models from the NMME as predictors, we can forecast the WNP TCF with promising skill in terms of correlation coefficient (0.63) and root mean square error (3.27 TC/year) for forecasts initialized in June. This study highlights the crucial role played by PMM and NASST in modulating WNP TCF, and suggests the use of PSPMM and NASST as potentially valuable predictors for WNP TCF.

Improving Multi-Model Ensemble Forecasts of Tropical Cyclone Intensity Using Bayesian Model Averaging

This paper proposes a method for multi-model ensemble forecasting based on Bayesian model averaging (BMA), aiming to improve the accuracy of tropical cyclone (TC) intensity forecasts, especially forecasts of minimum surface pressure at the cyclone center (Pmin). The multi-model ensemble comprises three operational forecast models: the Global Forecast System (GFS) of NCEP, the Hurricane Weather Research and Forecasting (HWRF) models of NCEP, and the Integrated Forecasting System (IFS) of ECMWF. The mean of a predictive distribution is taken as the BMA forecast. In this investigation, bias correction of the minimum surface pressure was applied at each forecast lead time, and the distribution (or probability density function, PDF) of Pmin was used and transformed. Based on summer season forecasts for three years, we found that the intensity errors in TC forecast from the three models varied significantly. The HWRF had a much smaller intensity error for short lead-time forecasts. To demonstrate the proposed methodology, cross validation was implemented to ensure more efficient use of the sample data and more reliable testing. Comparative analysis shows that BMA for this three-model ensemble, after bias correction and distribution transformation, provided more accurate forecasts than did the best of the ensemble members (HWRF), with a 5%–7% decrease in root-mean-square error on average. BMA also outperformed the multi-model ensemble, and it produced “predictive variance” that represented the forecast uncertainty of the member models. In a word, the BMA method used in the multi-model ensemble forecasting was successful in TC intensity forecasts, and it has the potential to be applied to routine operational forecasting.

Predictable and Unpredictable Components of the Summer East Asia–Pacific Teleconnection Pattern

The East Asia–Pacific (EAP) teleconnection pattern is the dominant mode of circulation variability during boreal summer over the western North Pacific and East Asia, extending from the tropics to high latitudes. However, much of this pattern is absent in multi-model ensemble mean forecasts, characterized by very weak circulation anomalies in the mid and high latitudes. This study focuses on the absence of the EAP pattern in the extratropics, using state-of-the-art coupled seasonal forecast systems. The results indicate that the extratropical circulation is much less predictable, and lies in the large spread among different ensemble members, implying a large contribution from atmospheric internal variability. However, the tropical–mid-latitude teleconnections are also relatively weaker in models than observations, which also contributes to the failure of prediction of the extratropical circulation. Further results indicate that the extratropical EAP pattern varies closely with the anomalous surface temperatures in eastern Russia, which also show low predictability. This unpredictable circulation–surface temperature connection associated with the EAP pattern can also modulate the East Asian rainband.

Inter-comparison between retrospective ensemble streamflow forecasts using meteorological inputs from ECMWF and NOAA/ESRL in the Hudson River sub-basins during Hurricane Irene (2011)

Abstract The objective of this work was to evaluate the benefits of using multi-model meteorological ensembles in representing the uncertainty of hydrologic forecasts. An inter-comparison experiment was performed using meteorological inputs from different models corresponding to Hurricane Irene (2011), over three sub-basins of the Hudson River basin. The ensemble-based precipitation inputs were used as forcing in a hydrological model to retrospectively forecast hourly streamflow, with a 96-hour lead time. The inputs consisted of 73 ensemble members, namely one high-resolution ECMWF deterministic member, 51 ECMWF members and 21 NOAA/ESRL (GEFS Reforecasts v2) members. The precipitation inputs were resampled to a common grid using the bilinear resampling method that was selected upon analysing different resampling methods. The results show the advantages of forcing hydrologic forecasting systems with multi-model ensemble forecasts over using deterministic and single model ensemble forecasts. The work showed that using the median of all 73 ensemble streamflow forecasts relatively improved the Nash–Sutcliffe Efficiency and lowered the biases across the examined sub-basins, compared with using the ensemble median from an individual model. This research contributes to the growing literature that demonstrates the promising capabilities of multi-model systems to better describe the uncertainty in streamflow predictions.

Medium-range reference evapotranspiration forecasts for the contiguous United States based on multi-model numerical weather predictions

Reference evapotranspiration (ET0) plays a fundamental role in agronomic, forestry, and water resources management. Estimating and forecasting ET0 have long been recognized as a major challenge for researchers and practitioners in these communities. This work explored the potential of multiple leading numerical weather predictions (NWPs) for estimating and forecasting summer ET0 at 101 U.S. Regional Climate Reference Network stations over nine climate regions across the contiguous United States (CONUS). Three leading global NWP model forecasts from THORPEX Interactive Grand Global Ensemble (TIGGE) dataset were used in this study, including the single model ensemble forecasts from the European Centre for Medium-Range Weather Forecasts (EC), the National Centers for Environmental Prediction Global Forecast System (NCEP), and the United Kingdom Meteorological Office forecasts (MO), as well as multi-model ensemble forecasts from the combinations of these NWP models. A regression calibration was employed to bias correct the ET0 forecasts. Impact of individual forecast variables on ET0 forecasts were also evaluated. The results showed that the EC forecasts provided the least error and highest skill and reliability, followed by the MO and NCEP forecasts. The multi-model ensembles constructed from the combination of EC and MO forecasts provided slightly better performance than the single model EC forecasts. The regression process greatly improved ET0 forecast performances, particularly for the regions involving stations near the coast, or with a complex orography. The performance of EC forecasts was only slightly influenced by the size of the ensemble members, particularly at short lead times. Even with less ensemble members, EC still performed better than the other two NWPs. Errors in the radiation forecasts, followed by those in the wind, had the most detrimental effects on the ET0 forecast performances.

Subseasonal Forecasting with an Icosahedral, Vertically Quasi-Lagrangian Coupled Model. Part I: Model Overview and Evaluation of Systematic Errors

Abstract The atmospheric hydrostatic Flow-Following Icosahedral Model (FIM), developed for medium-range weather prediction, provides a unique three-dimensional grid structure—a quasi-uniform icosahedral horizontal grid and an adaptive quasi-Lagrangian vertical coordinate. To extend the FIM framework to subseasonal time scales, an icosahedral-grid rendition of the Hybrid Coordinate Ocean Model (iHYCOM) was developed and coupled to FIM. By sharing a common horizontal mesh, air–sea fluxes between the two models are conserved locally and globally. Both models use similar adaptive hybrid vertical coordinates. Another unique aspect of the coupled model (referred to as FIM–iHYCOM) is the use of the Grell–Freitas scale-aware convective scheme in the atmosphere. A multiyear retrospective study is necessary to demonstrate the potential usefulness and allow for immediate bias correction of a subseasonal prediction model. In these two articles, results are shown based on a 16-yr period of hindcasts from FIM–iHYCOM, which has been providing real-time forecasts out to a lead time of 4 weeks for NOAA’s Subseasonal Experiment (SubX) starting July 2017. Part I provides an overview of FIM–iHYCOM and compares its systematic errors at subseasonal time scales to those of NOAA’s operational Climate Forecast System version 2 (CFSv2). Part II uses bias-corrected hindcasts to assess both deterministic and probabilistic subseasonal skill of FIM–iHYCOM. FIM–iHYCOM has smaller biases than CFSv2 for some fields (including precipitation) and comparable biases for other fields (including sea surface temperature). FIM–iHYCOM also has less drift in bias between weeks 1 and 4 than CFSv2. The unique grid structure and physics suite of FIM–iHYCOM is expected to add diversity to multimodel ensemble forecasts at subseasonal time scales in SubX.

Using a multi-model ensemble forecasting approach to identify key marine protected areas for seabirds in the Portuguese coast

Marine Protected Areas (MPAs) have been established across all marine environments, though their coherence and effectiveness in protecting umbrella species remains unclear. We used a multi-model ensemble forecasting approach, on 8 years of at-sea censuses of 30 seabird species to identify candidate MPAs in the Portuguese coast, prioritizing important areas for their conservation based on their occurrence and distribution. We overlapped the outputs generated by the Ensemble Ecological Niche Models (EENMs) with layers representing important environmental stressors (fishing intensity, ship density and oil pollution risk), and calculated loss in conservation value using them as cost layers. Three key marine areas were identified along the Portuguese coast: For breeders, there was a key marine area encompassing the Tagus and Sado estuaries and Berlengas archipelago; for non-breeders and migratory species two important areas were identified in the Northern and Southern coast. The key marine area identified in the Northern coast is characterized by high productivity and biodiversity, and can be affected by oil pollution from the refineries and the intensive ship traffic in this area. Also, the area identified in the Southern coast of Portugal for migratory seabirds overlaps extensively with areas of high fishing activity. Our results show that the Important Bird and Biodiversity Areas (IBAs) established along the Portuguese coast protect more than a third of the areas that we prioritized for breeding species and the official MPAs near 65% of the same areas. In contrast, current IBAs and national legislation protect less than 4% of the coastal areas that we prioritize for non-breeding species in this study. Our study, combining multi-species distribution with environmental constraints induced by human activities, allowed us to assess the coherence of the Portuguese marine planning and identify candidate areas to join the Portuguese network of marine protected areas. Our results, employing data from annual at-sea surveys together with the human stressors known to affect the Portuguese coast, proved to be an extremely useful strategy to identify spatial conservation areas along the Portuguese coast as well as to access the adequacy and consistency of those areas. Despite the constraints of this demanding approach, we are confident that our study provides a reliable strategy to inform marine conservation efforts and management planning in similar coastal environments elsewhere, characterized by strong coastal upwelling movements.

Modification of a linear regression-based multi-model super-ensemble technique and its application in forecasting of wave height during extreme weather conditions

ABSTRACTIn this study, we focus on the improvement of wave forecast of the Indian coastal region using a multi-model ensemble technique. Generally, a number of wave forecast are available for the same region from different wave models. The main objective of this study is to merge the wave forecasts available at Indian National Centre for Ocean Information Services from different wave models to obtain an improved wave forecast using a multi-model super-ensemble method [Krishnamurti et al. 1999. Improved weather and seasonal climate forecasts from multi-model super-ensemble. Science. 285:1548–1550] during extreme weather conditions and to modify Krishnamurthy’s techniques and validate with observations for a better prediction. Here, Multi-grid WAVEWATCH III, Simulating WAves Nearshore and MIKE 21 Spectral Waves are used for the generation of wave forecast. We propose a modification of Krishnamurthy’s linear regression-based ensemble model. By using both of these ensemble techniques, we perform a multi-model ensemble forecasting of significant wave height up to 24-h lead time in the Indian Ocean for three different cyclones (Nilofar, Hudhud and Phailin) and during the southwest monsoon. A comparison of ensemble predictions and individual model predictions with the actual observations showed generally satisfactory performance of the chosen tools. At the time of severe cyclones such as Hudhud and Phailin, our modified technique shows significantly better prediction than the linear regression-based ensemble technique.

Verification of Multimodel Ensemble Forecasts of North Atlantic Tropical Cyclones

Abstract North Atlantic tropical cyclone (TC) forecasts from four ensemble prediction systems (EPSs) are verified using the National Hurricane Center’s (NHC) best tracks for the 2008–15 seasons. The 1–5-day forecasts are evaluated for the 21-member National Centers for Environmental Prediction (NCEP) Global Ensemble Forecast System (GEFS), the 23-member UKMO ensemble (UKMET), and the 51-member European Centre for Medium-Range Weather Forecasts (ECMWF) ensemble, as well as a combination of these ensembles [Multimodel Global (MMG)]. Several deterministic models are also evaluated, such as the Global Forecast System (GFSdet), Hurricane Weather Research and Forecasting Model (HWRF), the deterministic ECMWF model (ECdet), and the Geophysical Fluid Dynamical Laboratory model (GFDL).The ECdet track errors are the smallest on average at all lead times, but are not significantly different from the GEFS and ECMWF ensemble means. All models have a slow bias (90–240 km) in the along-track direction by 120 h, while there is little bias in the cross-track direction. Much of this slow bias is attributed to TCs undergoing extratropical transition (ET). All EPSs are underdispersed in the along-track direction, while the ECMWF is slightly overdispersed in the cross-track direction. The MMG and ECMWF track forecasts have more probabilistic skill than the ECdet and comparable skill to the NHC climatology-based cone forecast. TC intensity errors for the HWRF and GFDL are lower than the coarser models within the first 24 h, but are comparable to the ECdet at longer lead times. The ECMWF and MMG have comparable or better probabilistic intensity forecasts than the ECdet, while the GEFS’s weak bias limits its skill.

Subseasonal Predictability of Boreal Summer Monsoon Rainfall from Ensemble Forecasts

Subseasonal forecast skill over the broadly defined North American (NAM), West African (WAM) and Asian (AM) summer monsoon regions is investigated using three Ensemble Prediction Systems (EPS) at sub-monthly lead times. Extended Logistic Regression (ELR) is used to produce probabilistic forecasts of weekly and week 3-4 averages of precipitation with starts in May-Aug, over the 1999-2010 period. The ELR tercile category probabilities for each model gridpoint are then averaged together with equal weight. The resulting Multi-Model Ensemble (MME) forecasts exhibit good reliability, but have generally low sharpness for forecasts beyond one week; Multi-model ensembling largely removes negative values of the Ranked Probability Skill Score (RPSS) seen in individual forecasts, and broadly improves the skill obtained in any of the three individual models except for the AM. The MME week 3-4 forecasts have generally higher RPSS and comparable reliability over all monsoon regions, compared to week 3 or week 4 forecast separately. Skill is higher during La Nina compared to El Nino and ENSO-neutral conditions over the 1999-2010 period, especially for the NAM. Regionally averaged RPSS is significantly correlated with the Maden-Julian Oscillation (MJO) for the AM and WAM. Our results indicate potential for skillful predictions at subseasonal time-scales over the three summer monsoon regions of the Northern Hemisphere.

Multimodel Ensembling of Subseasonal Precipitation Forecasts over North America

Probabilistic forecasts of weekly and week 3–4 averages of precipitation are constructed using extended logistic regression (ELR) applied to three models (ECMWF, NCEP, and CMA) from the Subseasonal-to-Seasonal (S2S) project. Individual and multimodel ensemble (MME) forecasts are verified over the common period 1999–2010. The regression parameters are fitted separately at each grid point and lead time for the three ensemble prediction system (EPS) reforecasts with starts during January–March and July–September. The ELR produces tercile category probabilities for each model that are then averaged with equal weighting. The resulting MME forecasts are characterized by good reliability but low sharpness. A clear benefit of multimodel ensembling is to largely remove negative skill scores present in individual forecasts. The forecast skill of weekly averages is higher in winter than summer and decreases with lead time, with steep decreases after one and two weeks. Week 3–4 forecasts have more skill along the U.S. East Coast and the southwestern United States in winter, as well as over west/central U.S. regions and the intra-American sea/east Pacific during summer. Skill is also enhanced when the regression parameters are fit using spatially smoothed observations and forecasts. The skill of week 3–4 precipitation outlooks has a modest, but statistically significant, relation with ENSO and the MJO, particularly in winter over the southwestern United States.

Combination of Multimodel Probabilistic Forecasts Using an Optimal Weighting System

Abstract In this study, an optimal weighting system is developed that combines multiple seasonal probabilistic forecasts in the North American Multimodel Ensemble (NMME). The system is applied to predict temperature and precipitation over the North American continent, and the analysis is conducted using the 1982–2010 hindcasts from eight NMME models, including the CFSv2, CanCM3, CanCM4, GFDL CM2.1, Forecast-Oriented Low Ocean Resolution (FLOR), GEOS5, CCSM4, and CESM models, with weights determined by minimizing the Brier score using ridge regression. Strategies to improve the performance of ridge regression are explored, such as eliminating a priori models with negative skill and increasing the effective sample size by pooling information from neighboring grids. A set of constraints is put in place to confine the weights within a reasonable range or restrict the weights from departing wildly from equal weights. So when the predictor–predictand relationship is weak, the multimodel ensemble forecast returns to an equal-weight combination. The new weighting system improves the predictive skill from the baseline, equally weighted forecasts. All models contribute to the weighted forecasts differently based upon location and forecast start and lead times. The amount of improvement varies across space and corresponds to the average model elimination percentage. The areas with higher elimination rates tend to show larger improvement in cross-validated verification scores. Some local improvements can be as large as 0.6 in temporal probability anomaly correlation (TPAC). On average, the results are about 0.02–0.05 in TPAC for temperature probabilistic forecasts and 0.03–0.05 for precipitation probabilistic forecasts over North America. The skill improvement is generally greater for precipitation probabilistic forecasts than for temperature probabilistic forecasts.

An Alternative Multi-Model Ensemble Forecast for Tropical Cyclone Tracks in the Western North Pacific

This study introduces an unequally weighted technique for Multi-model Ensemble (MME) forecasting for western North Pacific Tropical Cyclone (TC) tracks. Weights are calculated by partial least square regression, and members are selected by paired t-test. The performances for shorter forecast time ranges, such as 24, 48 and 72 h, are examined in order to improve the MME model, in which the weights for members are equally assigned. For longer forecast time ranges, such as 96 and 120 h, weights for MME members are thought to be less reliable, since the modeling is more likely to be influenced by the climate variability in the data period. A combination of both techniques for the shorter and the longer forecast time ranges is suggested as an alternative MME forecast procedure in operational meteorological agencies.

Weighting of NMME temperature and precipitation forecasts across Europe

Multi-model ensemble forecasts are obtained by weighting multiple General Circulation Model (GCM) outputs to heighten forecast skill and reduce uncertainties. The North American Multi-Model Ensemble (NMME) project facilitates the development of such multi-model forecasting schemes by providing publicly-available hindcasts and forecasts online. Here, temperature and precipitation forecasts are enhanced by leveraging the strengths of eight NMME GCMs (CCSM3, CCSM4, CanCM3, CanCM4, CFSv2, GEOS5, GFDL2.1, and FLORb01) across all forecast months and lead times, for four broad climatic European regions: Temperate, Mediterranean, Humid-Continental and Subarctic-Polar. We compare five different approaches to multi-model weighting based on the equally weighted eight single-model ensembles (EW-8), Bayesian updating (BU) of the eight single-model ensembles (BU-8), BU of the 94 model members (BU-94), BU of the principal components of the eight single-model ensembles (BU-PCA-8) and BU of the principal components of the 94 model members (BU-PCA-94). We assess the forecasting skill of these five multi-models and evaluate their ability to predict some of the costliest historical droughts and floods in recent decades. Results indicate that the simplest approach based on EW-8 preserves model skill, but has considerable biases. The BU and BU-PCA approaches reduce the unconditional biases and negative skill in the forecasts considerably, but they can also sometimes diminish the positive skill in the original forecasts. The BU-PCA models tend to produce lower conditional biases than the BU models and have more homogeneous skill than the other multi-models, but with some loss of skill. The use of 94 NMME model members does not present significant benefits over the use of the 8 single model ensembles. These findings may provide valuable insights for the development of skillful, operational multi-model forecasting systems.

Spatio‐temporal characteristics of Indonesian drought related to El Niño events and its predictability using the multi‐model ensemble

ABSTRACTSpatio‐temporal drought characteristics related to weak, moderate, and strong El Niño events in the Indonesian region were investigated. The empirical orthogonal function (EOF) patterns of normalized monthly precipitation and Standardized Precipitation Index (SPI) for the period 1950–2010 were analysed; the influence of El Niño events are better represented by SPI. The effects of El Niño events on meteorological drought were investigated based on a composite analysis. The effects of El Niño were more obvious during June–July–August and September–October–November, while they were not significant in March–April–May especially during weak and moderate El Niño events. Spatial distributions of affected areas varied by season, intensity of El Niño events, and local factors. During moderate and strong El Niño events, meteorological drought was more common in southern Sumatra, Java, Bali, Nusa Tenggara, southeastern Kalimantan, Sulawesi, Maluku, and Papua. Based on the spatial and temporal variability of droughts, the monthly or seasonal meteorological drought outlook can be improved by using SPI focusing more on the regions and seasons that are affected by El Niño events. The multi‐model ensemble (MME) forecast of SPI performed better for the target months of September to November even at medium lead times. The months are consistent with the period of serious El Niño effects on drought, implying that serious drought conditions during the most harshly affected season can be predicted using the MME forecast.