ECMWF Ensemble Research Articles

Forecast of precipitation for the 1994 flood in Piedmont: performance of an ensemble system at convection-permitting resolution

The major flood that affected the Piedmont region in Italy in November 1994 is re-forecast after 25 years in ensemble mode at the convection-permitting resolution of 2.2 km using the regional model COSMO. The performance of the probabilistic forecast of precipitation is assessed against rain-gauge observations, also in comparison with the driver system, i.e., the probabilistic re-forecast produced by ECMWF based on the operational IFS (Cycle 46r1) at grid spacings of 18 km. The convection-allowing system dynamically downscales the ECMWF ensemble and includes an explicit treatment of deep convection. Results indicate that both systems can predict up to 4 days in advance the timing and the spatial patterns of the precipitation, although with higher confidence for the convection-resolving system. The benefit of high resolution is shown mainly in the prediction of intense precipitation and in terms of correct amounts and locations, and confidence of occurrence (at day 3, the estimated probability of exceedance of 200 mm was higher than 90% over areas actually hit by such rainfall amounts). Additionally, convection-permitting resolution improves the representation of orographic precipitation, reducing the upwind precipitation displacement typical of coarser models and including the possible development of strong convection episodes embedded in the large-scale-forced orographic rise. For the high-resolution ensemble, the spread indicates large uncertainty at the local scale, mainly in defining the flow tendency to flank or flow over each mountain.

A dipole of tropical cyclone outgoing long‐wave radiation

AbstractLarge‐scale (500 ≤ r ≤ 2,200 km) outgoing long‐wave radiation (OLR) and water vapour (WV) fields are investigated in satellite observations over the Pacific, linked to ERA‐5 reanalysis data and the ECMWF ensemble forecasts of tropical cyclones (TC) globally. A large‐scale OLR dipole pattern of low and high fluxes are found in both the observations and model. As expected, a low OLR region is positioned within the TC circulation, but there is also a robust high OLR region poleward and west of the TC centre. A dry “black hole” on WV grey‐scale imagery occupies the same region of high OLR. Relative to the central low OLR TC signal, the typical dipole magnitude, distance and orientation of the high OLR regions are 230 W·m−2, 1,150 km, and 145° anticlockwise from east. From the reanalysis we find that the interaction between the vortex and the environmental flows produces upper‐level convergence, low‐level divergence and subsidence throughout the troposphere in the region of high OLR. Analysis of the ECMWF model shows that the position of the high OLR region rotates anticlockwise about the TC centre as the TC moves from westward to eastward. Through use of a sub‐ensemble, we test if capturing the high OLR anomaly has a significant relationship with TC track. We apply a perfect model approach and find that sub‐ensembles that are composed of models whose large‐scale OLR fields closely match the target TC also have a better track. This improvement is mostly attributed to the high OLR component of the dipole.

Potential skill of continental-scale, medium-range ensemble streamflow forecasts for flood prediction in South America

Recent scientific and computational developments have pushed hydrological forecasting systems up to continental and global scales. In this study, we evaluate the potential skill of ensemble streamflow forecasts for flood prediction in South America (SA), with a closer look on some of the large basins of the continent. Medium-range precipitation forecasts from the ECMWF Ensemble Prediction System were used to force a continental-scale, fully coupled hydrologic-hydrodynamic model to produce streamflow forecasts up to 15 days in advance. Forecasts were evaluated over a 4-year period against a reference simulation, derived from the hydrological model forced with a gridded precipitation dataset that merges data from multiple sources (MSWEP). To provide a benchmark for assessing the potential skill, forecasts based on the Ensemble Streamflow Prediction (ESP) technique were produced by running the hydrological model with precipitation data resampled from past MSWEP observations. Overall, ECMWF-based forecasts were skillful mainly at the eastern and southeastern regions of the continent (e.g., São Francisco, Parnaíba, Upper Paraná, Tocantins-Araguaia and Uruguay basins), showing little or no added skill in large rivers draining the Andes such as the Magdalena, Orinoco, Madeira (Amazon) and right-bank tributaries of the Paraguay. In the latter basins, ECMWF-based forecasts largely overestimated the exceedance probabilities of critical thresholds due to a positive bias in the predicted precipitation (in relation to MSWEP), visible mostly during the wet season. In addition, forecasts were found to perform better than ESP in terms of both overall quality and discrimination skill at regions characterized by a warm temperate, fully-humid climate. A direct relation between forecast skill and basin size was not always valid, since the former can vary geographically along tributaries. Limitations and future developments needed for continental-scale hydrological forecasting in SA are also discussed.

Using Artificial Neural Networks for Generating Probabilistic Subseasonal Precipitation Forecasts over California

AbstractForecast skill of numerical weather prediction (NWP) models for precipitation accumulations over California is rather limited at subseasonal time scales, and the low signal-to-noise ratio makes it challenging to extract information that provides reliable probabilistic forecasts. A statistical postprocessing framework is proposed that uses an artificial neural network (ANN) to establish relationships between NWP ensemble forecast and gridded observed 7-day precipitation accumulations, and to model the increase or decrease of the probabilities for different precipitation categories relative to their climatological frequencies. Adding predictors with geographic information and location-specific normalization of forecast information permits the use of a single ANN for the entire forecast domain and thus reduces the risk of overfitting. In addition, a convolutional neural network (CNN) framework is proposed that extends the basic ANN and takes images of large-scale predictors as inputs that inform local increase or decrease of precipitation probabilities relative to climatology. Both methods are demonstrated with ECMWF ensemble reforecasts over California for lead times up to 4 weeks. They compare favorably with a state-of-the-art postprocessing technique developed for medium-range ensemble precipitation forecasts, and their forecast skill relative to climatology is positive everywhere within the domain. The magnitude of skill, however, is low for week-3 and week-4, and suggests that additional sources of predictability need to be explored.

Boreal-winter teleconnections with tropical Indo-Pacific rainfall in HighResMIP historical simulations from the PRIMAVERA project

This study investigates how teleconnections linking tropical rainfall anomalies and wintertime circulation in the northern extra-tropics are represented in historical simulations for the period 1950–2010 run by partners of the EU-funded PRIMAVERA project, following the HighResMIP protocol of CMIP6. The analysis focusses on teleconnections from the western/central Indian Ocean in mid-winter and from the NINO4 region in both the early and the late part of winter; this choice is justified by a substantial change in the relationship between ENSO and the North Atlantic Oscillation (NAO) in the two parts of the season. Model results for both coupled integrations and runs with prescribed sea-surface temperature (SST) are validated against data from the latest ECMWF 20th-century re-analysis, CERA20C. Simulations from six modelling groups are considered, comparing the impact of increasing atmospheric resolution in runs with prescribed SST, and of moving from uncoupled to coupled simulations in the high-resolution version of each model. Single runs were available for each model configurations at the time of writing, with one centre (ECMWF) also providing a 6-member ensemble. Results from this ensemble are compared with those of a 6-member multi-model ensemble (MME) formed by including one simulation from each model. Using only a single historical simulation from each model configuration, it is difficult to detect a consistent change in the fidelity of model-generated teleconnections when either atmospheric resolution is increased or ocean coupling is introduced. However, when simulations from six different models are pooled together in the MME, some improvements in teleconnection patterns can be seen when moving from uncoupled to coupled simulations. For the ECMWF ensemble, improvements in the coupled simulations are only apparent for the late-winter NINO4 teleconnection. While the Indian Ocean teleconnection and the late-winter NINO4 teleconnection appear equally robust in the re-analysis record, the latter is well simulated in the majority of both uncoupled and coupled runs, while the former is reproduced with (generally) much larger errors, and a high degree of variability between individual models and ensemble members. Most of the simulations with prescribed SST fail to produce a realistic estimate of multi-decadal changes between the first and the second part of the 60-year record. This is (at least partially) due to their inability to simulate an Indian Ocean rainfall change which, in observations, has a zonal gradient out of phase with SST changes. In coupled runs, at least one model run with both realistic teleconnections and a good simulation of the inter-decadal pattern of Indian Ocean rainfall also shows a realistic NAO signal in extratropical multi-decadal variability.

Hydrometeorological Ensemble Forecast of a Highly Localized Convective Event in the Mediterranean

The uncertainties that affect hydrometeorological modelling chains can be addressed through ensemble approaches. In this paper, a convection-permitting ensemble system was assessed based on the downscaling of all members of the ECMWF ensemble prediction system through the coupled atmospheric-hydrological WRF-Hydro modelling system. An exemplary highly localized convective event that occurred in a morphologically complex area of the southern Italian coast was selected as a case study, evaluating the performance of the system for two consecutive lead times up to the hydrological forecast on a very small (11.4 km2) catchment. The proposed approach accurately downscales the signal provided by the global model, improving up to almost 200% the quantitative forecast of the accumulated rainfall peak in the area affected by the event and supplying clear information about the forecast uncertainty. Some members of the ensemble simulations provide accurate results up to the hydrological scale over the catchment, with unit peak discharge forecasts up to 3 m3∙s−1∙km−2. Overall, the study highlights that for highly localized convective events in coastal Mediterranean catchments, ensemble approaches should be preferred to a classic single-based simulation approach, because they improve the forecast skills and provide spatially distributed information about the forecast uncertainty, which can be particularly useful for operational purposes.

Predictive skill for atmospheric rivers in the western Iberian Peninsula

Abstract. A large fraction of extreme precipitation and flood events across western Europe are triggered by atmospheric rivers (ARs). The association between ARs and extreme precipitation days over the Iberian Peninsula has been well documented for western river basins. Since ARs are often associated with high impact weather, it is important to study their medium-range predictability. Here we perform such an assessment using the ECMWF ensemble forecasts up to 15 d for events where ARs made landfall in the western Iberian Peninsula during the winters spanning between 2012–2013 and 2015–2016. Vertically integrated horizontal water vapor transport (IVT) and precipitation from the 51 ensemble members of the ECMWF Integrated Forecasting System (IFS) ensemble (ENS) were processed over a domain including western Europe and the contiguous North Atlantic Ocean. Metrics concerning AR location, intensity, and orientation were computed, in order to compare the predictive skill (for different prediction lead times) of IVT and precipitation. We considered several regional boxes over western Iberia, where the presence of ARs is detected in analysis/forecasts, enabling the construction of contingency tables and probabilistic evaluation for further objective verification of forecast accuracy. Our results indicate that the ensemble forecasts have skill in detecting upcoming AR events, which can be particularly useful to better predict potential hydrometeorological extremes. We also characterized how the ENS dispersion and confidence curves change with increasing forecast lead times for each sub-domain. The probabilistic evaluation, using receiver operating characteristic (ROC) analysis, shows that for short lead times precipitation forecasts are more accurate than IVT forecasts, while for longer lead times this reverses (∼10 d). Furthermore, we show that this reversal occurs for shorter lead times in areas where the AR contribution is more relevant for winter precipitation totals (e.g., northwestern Iberia).

Bias Correction of Tropical Cyclone Parameters in the ECMWF Ensemble Prediction System in Australia

Abstract Global ensemble prediction systems have considerable ability to predict tropical cyclone (TC) formation and subsequent evolution. However, because of their relatively coarse resolution, their predictions of intensity and structure are biased. The biases arise mainly from underestimated intensities and enlarged radii, in particular the radius of maximum winds. This paper describes a method to reduce this limitation by bias correcting TCs in the ECMWF Ensemble Prediction System (ECMWF-EPS) for a region northwest of Australia. A bias-corrected TC system will provide more accurate forecasts of TC-generated wind and waves to the oil and gas industry, which operates a large number of offshore facilities in the region. It will also enable improvements in response decisions for weather sensitive operations that affect downtime and safety risks. The bias-correction technique uses a multivariate linear regression method to bias correct storm intensity and structure. Special strategies are used to maintain ensemble spread after bias correction and to predict the radius of maximum winds using a climatological relationship based on wind intensity and storm latitude. The system was trained on the Australian best track TC data and the ECMWF-EPS TC data from two cyclone seasons. The system inserts corrected vortices into the original surface wind and pressure fields, which are then used to estimate wind exceedance probabilities, and to drive a wave model. The bias-corrected system has shown an overall skill improvement over the uncorrected ECMWF-EPS for all TC intensity and structure parameters with the most significant gains for the maximum wind speed prediction. The system has been operational at the Australian Bureau of Meteorology since November 2016.

Bivariate Gaussian models for wind vectors in a distributional regression framework

Abstract. A new probabilistic post-processing method for wind vectors is presented in a distributional regression framework employing the bivariate Gaussian distribution. In contrast to previous studies, all parameters of the distribution are simultaneously modeled, namely the location and scale parameters for both wind components and also the correlation coefficient between them employing flexible regression splines. To capture a possible mismatch between the predicted and observed wind direction, ensemble forecasts of both wind components are included using flexible two-dimensional smooth functions. This encompasses a smooth rotation of the wind direction conditional on the season and the forecasted ensemble wind direction. The performance of the new method is tested for stations located in plains, in mountain foreland, and within an alpine valley, employing ECMWF ensemble forecasts as explanatory variables for all distribution parameters. The rotation-allowing model shows distinct improvements in terms of predictive skill for all sites compared to a baseline model that post-processes each wind component separately. Moreover, different correlation specifications are tested, and small improvements compared to the model setup with no estimated correlation could be found for stations located in alpine valleys.

Constrained Quantile Regression Splines for Ensemble Postprocessing

Abstract Statistical postprocessing of ensemble forecasts is widely applied to make reliable probabilistic weather forecasts. Motivated by the fact that nature imposes few restrictions on the shape of forecast distributions, a flexible quantile regression method based on constrained spline functions (CQRS) is proposed and tested on ECMWF Ensemble Prediction System (ENS) wind speed forecasting data at 125 stations in Norway. First, it is demonstrated that constraining quantile functions to be monotone and bounded is preferable. Second, combining an ensemble quantile with the ensemble mean proved to be a good covariate for the respective quantile. Third, CQRS only needs to be applied to about 10 equidistant quantiles, while those between can be obtained by interpolation. A comparison of CQRS versus a mixture model of truncated and lognormal distributions showed slight overall improvements in quantile score (less than 1%), reliability, and to some extent also sharpness. For strong wind speed forecasts the quantile score was improved by up to 4.5% depending on lead time.

Investigating the Factors That Contribute to African Easterly Wave Intensity Forecast Uncertainty in the ECMWF Ensemble Prediction System

Abstract Although there have been numerous studies documenting the processes/environments that lead to the intensification of African easterly waves (AEWs), only a few of these studies investigated the effect of those processes or the environment on the predictability of AEWs. Here, the large-scale modulation of AEW intensity predictability is evaluated using the 51-member ECMWF ensemble prediction system (EPS) during an active AEW period (July–September 2011–13). Forecasts are stratified based on the 72-h AEW intensity standard deviation (SD) to evaluate hypotheses for how different processes contribute to large forecast SD. While large and small SD forecasts are associated with similar baroclinic and barotropic energy conversions, forecasts with large SD are characterized by higher relative humidity values downstream of the AEW trough. These areas of higher humidity are also associated with higher precipitation and precipitation SD, suggesting that uncertainty associated with diabatic processes could be linked with large AEW intensity SD. Although water vapor is a strong function of longitude and phase of convectively coupled equatorial waves, the cases with large and small SD are characterized by similar longitude and wave phase, suggesting that AEWs occurring in certain locations or convectively coupled equatorial wave phases are not more or less predictable.

Monitoring trends in ensemble forecast performance focusing on surface variables and high‐impact events

This paper discusses methodological aspects of a monitoring process which focuses simultaneously on ensemble forecasts, surface variables, and high‐impact events. Which score(s) is (are) suitable for this task is a central question but not the only one to be answered. Here, we investigate the properties of the Brier score, logarithmic score, and diagonal elementary score in the context of forecast performance monitoring as well as the impact of methodological choices such as the event threshold definition, the reference forecast, and the role assigned to representativeness errors. A consistent picture of the verification process is eventually drawn where the design of the event climatology plays a key role. This study is illustrated by verification results for three surface variables (24 hr precipitation, 10 m wind speed, and 2 m temperature) over 15 years of operational ECMWF ensemble forecasting activities. Results are also compared with a current ECMWF headline score: the relative operating characteristic skill score for the Extreme Forecast Index.

Selective ensemble mean technique for severe European windstorms

We show that ensemble forecasts of extreme European windstorms can be improved up to a lead time of 36–48 hr by sub‐selecting ensemble members based on their performance at very short lead times (up to 12 hr). This applies to both the ensemble mean position of the cyclone centre and the ensemble windstorm footprint over the continent. A number of ensemble forecasts, including those from the ECMWF Ensemble Prediction System, are initialised every 12 hr and disseminated several hours after initialisation; therefore our approach has the potential to provide improved forecasts in an operational context. The analysis is performed on GEFS reforecast data.

Relationship between the Track and Structural Evolution of Hurricane Sandy (2012) Using a Regional Ensemble

Abstract An ensemble of 72 Weather Research and Forecasting (WRF) Model simulations is evaluated to examine the relationship between the track of Hurricane Sandy (2012) and its structural evolution. Initial and boundary conditions are obtained from ECMWF and GEFS ensemble forecasts initialized at 0000 UTC 25 October. The 5-day WRF simulations are initialized at 0000 UTC 27 October, 48 h into the global model forecasts. Tracks and cyclone phase space (CPS) paths from the 72 simulations are partitioned into 6 clusters using regression mixture models; results from the 4 most populous track clusters are examined. The four analyzed clusters vary in mean landfall location from southern New Jersey to Maine. Extratropical transition timing is the clearest difference among clusters; more eastward clusters show later Sandy–midlatitude trough interaction, warm seclusion formation, and extratropical transition completion. However, the intercluster variability is much smaller when examined relative to the landfall time of each simulation. In each cluster, a short-lived warm seclusion forms and contracts through landfall while lower-tropospheric potential vorticity concentrates at small radii. Despite the large-scale similarity among the clusters, relevant intercluster differences in landfall-relative extratropical transition are observed. In the easternmost cluster the Sandy–trough interaction is least intense and the warm seclusion decays the most by landfall. In the second most eastward cluster Sandy retains the most intact warm seclusion at landfall because of a slightly later (relative to landfall) and weaker trough interaction compared to the two most westward clusters. Nevertheless, the remarkably similar large-scale evolution of Sandy among the four clusters indicates the high predictability of Sandy’s warm seclusion extratropical transition before landfall.

The ECMWF ensemble prediction system: Looking back (more than) 25 years and projecting forward 25 years

This paper has been written to mark 25 years of operational medium‐range ensemble forecasting. The origins of the ECMWF Ensemble Prediction System are outlined, including the development of the precursor real‐time Met Office monthly ensemble forecast system. In particular, the reasons for the development of singular vectors and stochastic physics – particular features of the ECMWF Ensemble Prediction System ‐ are discussed. The author speculates about the development and use of ensemble prediction in the next 25 years.

The diagonal score: Definition, properties, and interpretations

The diagonal score is proposed as a new scoring rule for the assessment of univariate probabilistic forecasts. First, a general framework for the definition of proper user‐oriented scores is set up using the concept of elementary scoring rules. Building on this fundamental notion, a complete overview of forecast skill is provided by a new verification tool. The forecast skill card introduced displays the forecast value as a function of two decision parameters, a probability level and an event threshold, simultaneously. The diagonal score emerges as a summary score focusing on the diagonal of the skill card, that is, by fixing the relationship between the event base rate and forecast probability level. The properties of the new score, as well as interpretations from the perspectives of users and developers, are discussed based on synthetic datasets and 2 m temperature forecasts of the operational ECMWF ensemble prediction system.

An evaluation of the convection‐permitting ensemble COSMO‐E for three contrasting precipitation events in Switzerland

Specific aspects of the performance of COSMO‐E, a newly operational convection‐permitting ensemble prediction system run by the Swiss national weather service, are evaluated for three contrasting precipitation events and compared with its driving model, the ECMWF ensemble (EC ENS). The events include locally triggered air‐mass convection on four consecutive days (CONV), a complex flood‐producing rainfall episode (LSF1), and a summertime cold‐frontal precipitation event (LSF2). Investigating the precipitation forecasts reveals that COSMO‐E outperforms EC ENS in all cases in terms of precipitation pattern, whereas both ensembles fail in predicting the correct domain‐averaged precipitation for LSF1 and struggle with the correct timing of the daily precipitation onset for CONV. For LSF2, EC ENS produces an almost perfect time series of domain‐averaged precipitation, but this seemingly excellent forecast fails in predicting the spatial distribution. Spread–error relationships also yield a rather complex picture. The higher resolution of COSMO‐E leads to increased spread and reduced underdispersion for near‐surface variables. However, in the free troposphere the two ensembles perform similarly, and the agreement of spread and error varies between cases, variables and levels. For both events with large‐scale advection, underdispersion occurs for mid‐tropospheric relative humidity near fronts, whose propagation is overconfident in EC ENS, while for the convective event COSMO‐E clearly shows larger spread than EC ENS. Because frontal propagation is largely determined by lateral boundary conditions, this underdispersion also occurs for COSMO‐E. In summary, this study confirms the benefit of a convection‐permitting model for many aspects of ensemble forecasting, and illustrates the challenge of robustly assessing the quality of an ensemble forecast for hydrometeorologically relevant events.

African Easterly Wave Forecast Verification and Its Relation to Convective Errors within the ECMWF Ensemble Prediction System

Abstract African easterly waves (AEWs) are the primary synoptic-scale weather feature found in sub-Saharan Africa during boreal summer, yet there have been few studies documenting the performance of operational ensemble prediction systems (EPSs) for these phenomena. Here, AEW forecasts in the 51-member ECMWF EPS are validated against an average of four operational analyses during two periods of enhanced AEW activity (July–September 2007–09 and 2011–13). During 2007–09, AEW position forecasts were mainly underdispersive and characterized by a slow bias, while intensity forecasts were characterized by an overintensification bias, yet the ensemble-mean errors generally matched the forecast uncertainty. Although 2011–13 position forecasts were still underdispersive with a slow bias, the ensemble-mean error is smaller than for 2007–09. In addition, the 2011–13 intensity forecasts were overdispersive and had a negligible intensity bias. Forecasts from 2007 to 2009 were characterized by higher precipitation in the AEW trough center and high correlations between divergence errors and intensity errors, suggesting the intensity bias is associated with errors in convection. By contrast, forecasts from 2011 to 2013 have smaller precipitation biases than those from 2007 to 2009 and exhibit a weaker correlation between divergence errors and intensity errors, suggesting a weaker connection between AEW forecast errors and convective errors.

Revisiting the synoptic-scale predictability of severe European winter storms using ECMWF ensemble reforecasts

Abstract. New insights into the synoptic-scale predictability of 25 severe European winter storms of the 1995–2015 period are obtained using the homogeneous ensemble reforecast dataset from the European Centre for Medium-Range Weather Forecasts. The predictability of the storms is assessed with different metrics including (a) the track and intensity to investigate the storms' dynamics and (b) the Storm Severity Index to estimate the impact of the associated wind gusts. The storms are well predicted by the whole ensemble up to 2–4 days ahead. At longer lead times, the number of members predicting the observed storms decreases and the ensemble average is not clearly defined for the track and intensity. The Extreme Forecast Index and Shift of Tails are therefore computed from the deviation of the ensemble from the model climate. Based on these indices, the model has some skill in forecasting the area covered by extreme wind gusts up to 10 days, which indicates a clear potential for early warnings. However, large variability is found between the individual storms. The poor predictability of outliers appears related to their physical characteristics such as explosive intensification or small size. Longer datasets with more cases would be needed to further substantiate these points.

Simulations of the Asian summer monsoon in the sub‐seasonal to seasonal prediction project (S2S) database

Several monsoon indices have been applied to multiple models from the sub‐seasonal to seasonal (S2S) prediction project database during the period May to October 1999–2010 to assess their ability to simulate the Asian monsoon. The bivariate anomaly correlation (BAC) of the Boreal Summer IntraSeasonal Oscillation (BSISO) index suggests that the operational models can predict the BSISO1 and BSISO2 events up to 6–24.5 and 6.5–14 days in advance respectively, although the models tend to underestimate the amplitude of BSISO as the lead time increases. For the strong BSISO events, BSISO1 (BSISO2) display lower skill mostly in phases 3–5 for all the models, suggesting that the BSISO1 (BSISO2) is not easy to predict when it is located over India and the Maritime Continent (South China Sea and Bay of Bengal). On the other hand, the higher skills appear in different phases for different models. For instance, the limit of predictive skill of strong BSISO1 and BSISO2 events in phases 6–7 for the ECMWF ensemble forecast could exceed 30 and 28 days, respectively. The comparisons of the BSISO life cycle among the ECMWF, NCEP and CMA models also indicate that the ECMWF model can better predict the evolution of strong BSISO events. Predictions of additional monsoon circulation indices including the Webster–Yang index (WY), Indian Summer Monsoon index (ISM), South Asian Monsoon index (SAM), and SouthEast Asian Monsoon index (SEAM) in the S2S models have statistically significant skill over the corresponding monsoon regions up to 9–31, 3–17, 7–13 and 7–14 days, respectively. However, the significant skill of summer monsoon precipitation over the SAM and SEAM regions varies significantly among the models, with the skill ranging from 2 days to 2 weeks lead time.