The skill of seasonal forecasts of midlatitude atmospheric circulation is notoriously intermittent and might be modest on average. The use of seasonal forecasts for real-time applications can therefore benefit greatly from approaches providing an a priori indication of how skillful a single forecast will be at capturing circulation anomalies. This work introduces a methodology to predict the skill of such a single ensemble forecast. It consists of verifying the ensemble mean against each of its individual members before averaging, in order to provide a "perfect model" estimate of forecast skill. This methodology is applied to seasonal forecasts from the Copernicus Climate Change Service (C3S), for a selection of four variables (mean sea-level pressure, geopotential height at 500 and 200 hPa, and streamfunction at 200 hPa) characterizing the atmospheric circulation in three domains covering the Northern Hemisphere midlatitudes: Pacific and North America, North Atlantic, and Asia. Skill is defined as the Anomaly Correlation Coefficient (ACC), which quantifies how well the predicted spatial patterns of atmospheric circulation match the observed ones. Results show that the prediction of forecast skill is quite successful for upper-troposphere variables but is diversely successful in the mid-troposphere and the surface, depending on the region under study. Moreover, the capacity to predict forecast skill exhibits seasonal dependence, with generally better performance in boreal summer and winter and an even greater summer peak for the North Atlantic region. These results are very similar across the six C3S models under consideration and their multi-model combination, which is an indication of robustness.

In 2025, Météo-France implemented a new version of its climate prediction system for seasonal and subseasonal forecasting: System 9. This system belongs to the generation of Météo-France climate prediction systems based on the global coupled climate model CNRM-CM6-1 \citep{Voldoire2019}, developed at the Centre National de Recherches Météorologiques (CNRM). Building on its predecessor System 8, System 9 includes substantial changes in terms of ocean and sea-ice modeling, radiation scheme and initialization procedure. The article documents the configuration of System 9 and provides a comparative evaluation with System 8 in terms of model biases and predictive skill. First, System 9 often exhibits reduced biases, especially for surface temperatures over land and sea. Second, it demonstrates improved predictive skill for key parameters at seasonal timescales, such as Arctic sea-ice area around the September minimum and the Indian Ocean Dipole during its peak season. Third, it is also significantly more skillful at subseasonal timescales in the first forecast weeks, an improvement likely due to the revised initialization procedure. In the meantime, only a limited number of parameters, locations, and seasons show a clear degradation in bias or skill compared to System 8. Overall, the evaluation demonstrates that System 9 outperforms System 8 on several key metrics. This publication serves as the reference for the Météo-France operational seasonal forecasts provided to the Copernicus Climate Change Service (C3S) since 1 May 2025. It also establishes the baseline for the third generation of CNRM subseasonal forecasts feeding the Subseasonal-to-seasonal (S2S) database since 20 November 2025.

Several studies highlight the relevance of considering polar winter stratospheric information such as the occurrence of Sudden Stratospheric Warmings (SSWs) for skillful Subseasonal to Seasonal (S2S) surface climate predictions. However, current S2S forecast systems can only predict these events about two weeks in advance. A potential way of increasing their predictability is to improve the models' representation of the triggering mechanisms of SSWs. Traditional theories indicate that SSWs follow sustained wave dissipation in the stratosphere, but the relative role of tropospheric versus stratospheric conditions in the enhancement of stratospheric wave activity remains unclear. This study aims to quantify the role of the stratospheric state in wave activity preceding SSWs by analyzing three recent SSWs: the boreal SSWs of 2018 and 2019 and the austral minor SSW of 2019, using specific sets of S2S experiments. These ensembles follow the SNAPSI (Stratospheric Nudging And Predictable Surface Impacts) guidelines and include free-evolving atmospheric runs and nudged simulations, where the zonally-symmetric stratospheric state is nudged to either observations of a certain SSW or a climatological state. Our results show that the models struggle to capture the strong enhancement of wave activity preceding the 2018 SSW, limiting predictability beyond 10 d. In contrast, both SSWs of 2019 are better predicted, consistent with a more accurate simulation of the wave activity. Nudging the zonal mean stratospheric state does not drastically influence the upward wave activity flux or tropospheric circulation anomalies prior to these SSWs, but it has some impact on the stratospheric wave activity, although this modulation depends on the event characteristics. The boreal 2019 SSW appears to be primarily driven by tropospheric processes. In contrast, stratospheric contributions may have also played an important role in triggering the boreal 2018 SSW and the austral 2019 SSW. Understanding these variations is key to improving SSW predictability in S2S models.

Ocean Reanalyses Workshop of the European Copernicus Marine Service What: Gather together ocean reanalyses users and producers to identify users' needs of ocean reanalyses and design the strategy to improve ocean reanalyses to fulfill users' needs When: 10-12 October 2023 Where: Toulouse, France, and online

Research and development activities around the current Météo-France operational seasonal forecasting system (System 8) are underway to upgrade it to the next version (System 9), along with efforts to improve the initialization of its components. Among these components, sea ice is particularly challenging to initialize. At present, a coupled-nudged initialisation strategy, based on a high-resolution configuration of the CNRM-CM6 climate model, is employed to initialise the System 8, except for the sea-ice. In order to get initial states of sea ice that are consistent with the forecasting model, our procedure consists in making a preliminary continuous run where the ocean and sea ice models are integrated in stand-alone mode, with forcing at the surface from an atmosphere reanalysis.However, in the current operational System 8 – based on the NEMO 3.6 ocean model and the GELATO sea ice model – the initial states of sea ice generated with this procedure are not fully realistic. Results show that the sea ice thickness over the Arctic region in the System 8 initial states is underestimated compared to the reference data. Numerous sensitivity experiments were carried out with the current NEMOv3.6-GELATO system, leading to some minor improvements. Thus, an upgraded version of the ocean model (NEMO version 4.2) coupled to a new sea-ice component (SI3) has been tested (in stand-alone mode, not coupled to the atmosphere) to see if the use of more recent versions of ocean and sea-ice models leads to some improvements in the Arctic sea ice representation. The results are encouraging as the representation of sea ice variables in the Arctic is improved compared to the old version.This incites our team to foresee that System 9 will indeed incorporate the NEMO4.2 and SI3 models, and that the same initialization procedure as before (using these new models) will provide sea-ice initial states closer to those observed.

Research and development activities around the current Météo-France operational seasonal forecasting system (System 8) are underway to upgrade it to the next version (System 9), along with efforts to improve the initialization of its components. Among these components, sea ice is particularly challenging to initialize. At present, a coupled-nudged initialisation strategy, based on a high-resolution configuration of the CNRM-CM6 climate model, is employed to initialise the System 8, except for the sea-ice. In order to get initial states of sea ice that are consistent with the forecasting model, our procedure consists in making a preliminary continuous run where the ocean and sea ice models are integrated in stand-alone mode, with forcing at the surface from an atmosphere reanalysis. However, in the current operational System 8 – based on the NEMO 3.6 ocean model and the GELATO sea ice model – the initial states of sea ice generated with this procedure are not fully realistic. Results show that the sea ice thickness over the Arctic region in the System 8 initial states is underestimated compared to the reference data. Numerous sensitivity experiments were carried out with the current NEMOv3.6-GELATO system, leading to some minor improvements. Thus, an upgraded version of the ocean model (NEMO version 4.2) coupled to a new sea-ice component (SI3) has been tested (in stand-alone mode, not coupled to the atmosphere) to see if the use of more recent versions of ocean and sea-ice models leads to some improvements in the Arctic sea ice representation. The results are encouraging as the representation of sea ice variables in the Arctic is improved compared to the old version. This incites our team to foresee that System 9 will indeed incorporate the NEMO4.2 and SI3 models, and that the same initialization procedure as before (using these new models) will provide sea-ice initial states closer to those observed.

We propose a forecasting tool for precipitation based on analogues of circulation defined from 5-day hindcasts and a stochastic weather generator that we call "HC-SWG." In this study, we aim to improve the forecast of European precipitation for subseasonal lead times (from 2 to 4 weeks) using the HC-SWG. We designed the HC-SWG to generate an ensemble precipitation forecast from the European Centre of Medium-range Weather Forecasts (ECMWF) and Centre National de la Recherche M & eacute;t & eacute;orologique (CNRM) subseasonal-to-seasonal ensemble reforecasts. We define analogues from 5-day ensemble reforecast of Z500 from the ECMWF (11 members) and CNRM (10 members) models. Then, we generate a 100-member ensemble for precipitation over Europe. We evaluate the skill of the ensemble forecast using probabilistic skill scores such as the continuous ranked probability skill score (CRPSS) and receiver operating characteristic curve. We obtain reasonable forecast skill scores within 35 days for different locations in Europe. The CRPSS shows positive improvement with respect to climatology and persistence at the station level. The HC-SWG shows a capacity to distinguish between events and non-events of precipitation within 15 days at the different stations. We compare the HC-SWG forecast with other precipitation forecasts to further confirm the benefits of our method. We found that the HC-SWG shows improvement against the ECMWF precipitation forecast until 25 days. We show the capacity of a statistical system to improve the forecast skill of European precipitation using the analogues of subseasonal-to-seasonal dynamical models. Our results show the high potential of the circulation analogue method to improve ensemble forecasts of European precipitation compared with numerical models. image

Seasonal prediction uses ensemble forecasting to sample the distribution of possible climate outcomes in the upcoming term given the slowly-varying constraints on the atmosphere. However, translating the members’ distribution of a seasonal forecast into meaningful information is a challenge climate services are often faced with. When a large ensemble spread makes the forecast difficult to interpret, highlighting the competing signals from which the uncertainty arises may bear added value to end users. In order to do so, we present an approach to extract alternative seasonal forecast scenarios over Europe (in temperature, precipitation and atmospheric circulation) from ensemble seasonal forecasts. The aim of the scenarios is to refine the ensemble analysis beyond the usual forecast products (e.g ensemble mean, tercile probabilities), and to provide additional guidance for preparation of the seasonal forecast bulletins routinely issued at Météo-France. The seasonal forecast scenarios are determined with a hierarchical clustering of the ensemble members, based on their forecast temperature at 2-m (T2m). The dissimilarity between two members is defined from the spatial correlation between their respective maps of T2m anomalies – relative to model climatology – over a European domain (29.5°W-40.5°E; 30.5°N-70.5°N, land grid points only). The subsequent dissimilarity matrix across the ensemble feeds the clustering algorithm that groups members into clusters eventually defining the scenarios. The seasonal outcomes corresponding to these scenarios are then described through several diagnostics, e.g composites on sensible climate variables (T2m, precipitation), composites on atmospheric circulation variables (Z500, V200), and analysis through modes of variability and weather regimes. In addition, we provide a description of how scenarios diverge in the course of forecast integration and identify teleconnections related to each scenario. Finally, we also assess the skill of the seasonal forecasts assuming that only the subset of members representing the most likely scenario is retained. This methodology has been implemented to the Copernicus Climate Change Services (C3S) real-time seasonal forecasts across the past year for experimental purposes, and it is shown to be a relevant complement for the preparation of the Météo-France operational seasonal bulletins.

Ensemble forecasts of precipitation with sub-seasonal lead times offer useful information for decision makers when they sufficiently sample the possible outcomes of trajectories. In this study, we aim to improve precipitation ensemble forecast systems using a stochastic weather generator (SWG) based on analogs of the atmospheric circulation. This approach is tested for sub-seasonal lead times (from 2 to 4 weeks). The SWG ensemble forecasts yield promising probabilistic skill scores for lead times of 5-10 days for precipitation (Krouma et al, 2022) and for lead times of 40 days for temperature (Yiou and Déandréis, 2019) . In this work, we adapt the parameters of the SWG to optimize the simulation of European precipitations from ensemble dynamical reforecasts of ECMWF and CNRM. We present the HC-SWG forecasting tool (HC refers to Hindcast and SWG to the stochastic weather generator) based on a combination of dynamical and stochastic models.We start by computing analogs of Z500 from the ensemble member reforecast of ECMWF (11 members) and CNRM (10 members). Then, we generate an ensemble of 100 members for precipitation over Europe. We evaluate the ensemble forecast of the HC-SWG using skill scores such as the continuous probabilistic score CRPS and ROC curve.We obtain reasonable forecast skill scores for lead times up to 35 days for different locations in Europe (Madrid, Toulouse, Orly, De Bilt and Berlin). We compare the HC-SWG forecast with other precipitation forecasts to further confirm the benefit of our method. We found that the HC-SWG shows improvement against the ECMWF precipitation forecast until 25 days.

The current Météo-France seasonal prediction system (MF System 8) has 25 members for hindcast from 1993 to 2016 and 51 members for real-time forecast. In order to investigate the benefits of increasing the ensemble size within our Copernicus Climate Change Services (C3S) seasonal prediction contract, we extended the system 8 hindcast to 51 members for the four main start dates (February, May, August, November). We compare the forecast skill between the official 25-member hindcast and the 51-member extended hindcast. We focus on the European region at the lead-time 1 for the next trimester. To describe the performance of forecasts, we use correlation and relative operating characteristic (ROC) on the mean 2-meter temperature (T2M) and the mean precipitation (RR) averaged over Europe. Similarly, we also evaluate the forecast of modes of variability (East Atlantic, North Atlantic Oscillation, and Scandinavian Blocking) which impact the European climate. The scores with 51 members are similar and not necessarily better than with 25 members. Moreover, we use 1000 random draws of 25 members out of 51 to determine the uncertainty of the official forecast scores. These scores can be at the edge of the confidence interval, while the 51-member scores are close to the median of the 1000 random draws. We did the same analysis for different regions that there is less uncertainty on the scores in the Tropics (e.g. Northeast Brazil) than in the mid-latitudes (e.g. Europe). These results suggest that it is not necessary to increase the ensemble size for verification of the seasonal forecasts beyond the available 25 members.

This study proposes an objective methodology to highlight windows of opportunity in a numerical subseasonal forecasting system. The methodology is based on a contingency table and is applied to the prediction of heavy tropical precipitation by the European Centre for Medium‐range Weather Forecasts (ECMWF) subseasonal‐to‐seasonal (S2S) reforecasts in the November‐to‐April season, in relation with the Madden–Julian oscillation (MJO). As a slowly propagating signal of enhanced convection, the MJO may indicate favorable conditions for heavy precipitation a few weeks ahead in some tropical areas. The combined knowledge of these climatological impacts and the current phase of the MJO at initialization defines observation‐based “climatological windows of opportunity.” We then investigate whether the ECMWF S2S forecasts are indeed more performant when there is increased likelihood of heavy rainfall, that is, whether the model converts “climatological windows of opportunity” into “model windows of opportunity.” Our results show that, by Week 2, this is only verified for a limited number of tropical areas, mostly located in the western Pacific and Africa. Meanwhile, failures to seize the opportunities lie in misplaced MJO impacts, signal loss, or too many false alarms.

This study proposes an objective methodology to highlight windows of opportunity related to a precursor phenomenon in a numerical subseasonal forecasting system. The methodology is based on a contingency table and is illustrated with the relationship between the Madden-Julian oscillation (MJO) and heavy rainfall in the tropical band. As a slowly propagating signal of enhanced convection, the MJO may indicate favorable conditions for heavy precipitation a few weeks ahead in some tropical areas. The combined knowledge of these climatological impacts and the current phase of the MJO at initialization defines observation-based "climatological windows of opportunity". In a second step, we analyze whether S2S forecasts are indeed more performant when there is increased climatological likelihood of heavy rainfall, i.e whether the forecasts convert "climatological windows of opportunity" into "model windows of opportunity". The methodology is implemented to the prediction of the upper quintile of weekly precipitation in 20 years of ECMWF S2S reforecasts in the November-to-April season. The ability of the ECMWF forecasts to convert periods with more predictable events into periods of actual forecast skill is only verified for a limited number of small areas, while failures to seize the opportunities lie in misplaced MJO impacts, signal loss or too many false alarms.

Cet article est un condensé de la thèse soutenue par l’auteur le 6 novembre 2020 et récompensée par le prix Gérard Beltrando de l’Association Internationale de Climatologie au titre de l’année 2021. Cette thèse a été préparée au Centre National de Recherches Météorologiques (UMR 3589, Météo-France & CNRS) entre 2017 et 2020 sous la direction de Lauriane Batté et Michel Déqué.

Major disruptions of the winter season, high-latitude stratospheric polar vortices can result in stratospheric anomalies that persist for months. These sudden stratospheric warming events are recognized as an important potential source of forecast skill for surface climate on subseasonal to seasonal timescales. Realizing this skill in operational subseasonal forecast models remains a challenge, as models must capture both the evolution of the stratospheric polar vortices in addition to their coupling to the troposphere. The processes involved in this coupling remain a topic of open research. We present here the Stratospheric Nudging And Predictable Surface Impacts (SNAPSI) project. SNAPSI is a new model intercomparison protocol designed to study the role of the Arctic and Antarctic stratospheric polar vortex disturbances for surface predictability in subseasonal to seasonal forecast models. Based on a set of controlled, subseasonal ensemble forecasts of three recent events, the protocol aims to address four main scientific goals. First, to quantify the impact of improved stratospheric forecasts on near-surface forecast skill. Second, to attribute specific extreme events to stratospheric variability. Third, to assess the mechanisms by which the stratosphere influences the troposphere in the forecast models. Fourth, to investigate the wave processes that lead to the stratospheric anomalies themselves. Although not a primary focus, the experiments are furthermore expected to shed light on coupling between the tropical stratosphere and troposphere. The output requested will allow for a more detailed, process-based community analysis than has been possible with existing databases of subseasonal forecasts.

Although there is an increasing interest in precipitation information at the subseasonal timescales in a wide range of sectors, the use of subseasonal precipitation forecasts from general circulation models is often impaired by poor reliability and low forecast skill. One crucial step to improve forecast quality is statistical correction and post-processing, which is particularly important for a parameterized variable like rainfall. This study introduces and evaluates a statistical-dynamical post-processing scheme, based on a Bayesian framework, that aims at providing more skillful and more reliable subseasonal forecasts of weekly precipitation. On the one hand, this method relies on the statistical relationship between observed and dynamically-forecast precipitation, that is determined in a set of reforecasts and depends on the lead time. On the other hand, it also takes advantage of the climatological impacts of large-scale drivers affecting rainfall, that are generally better represented by numerical models than rainfall itself. These two aspects of the method are respectively called calibration and bridging. This statistical-dynamical prediction scheme is illustrated with an application to the austral summer precipitation in the southwest tropical Pacific, using the Météo-France and ECMWF reforecasts in the Subseasonal-to-seasonal (S2S) database. Indices representing El Niño Southern Oscillation and the Madden-Julian Oscillation – the major sources of predictability in the area – are used for bridging. Probabilistic forecasts of heavy rainfall spells are evaluated in terms of discrimination (ROC skill score) and reliability, which are both improved by the Bayesian method at all lead times (from week 1 to week 4). Additional results show that the calibration part of the method, using forecast precipitation as a predictor, is necessary to enhance forecast skill. The bridging part also provides additional discrimination skill, that is mostly due to the ENSO-related information.

Issuing skillful forecasts beyond the typical horizon of weather predictability remains a challenge actively addressed by the scientific community. This study evaluates winter subseasonal reforecasts delivered by the CNRM and ECMWF dynamical systems and identifies that the level of skill for predicting temperature in Europe varies fairly consistently in both systems. In particular, forecasts initialized during positive NAO phases tend to be more skillful over Europe at week three in both systems. Composite analyses performed in an atmospheric reanalysis, a long-term climate simulation and both forecast systems unveil very similar temperature and sea-level pressure patterns 3 weeks after NAO+ conditions. Furthermore, regressing these fields onto the 3-week previous NAO index in a reanalysis shows consistent patterns over Europe but also eastern North America, thereby revealing a lagged teleconnection, either related to the persistence or recurrence of the NAO+ weather regime. Since this feature is well captured by forecast systems, this is a key mechanism for determining a priori confidence in the skill of wintertime subseasonal forecasts over Europe and North America.

AbstractMultimodel ensemble (MME) reforecasts of rainfall at subseasonal time scales in the southwest tropical Pacific are constructed using six models (BoM, CMA, ECCC, ECMWF, Météo‐France, and UKMO) from the Subseasonal‐to‐Seasonal (S2S) database by member pooling. These reforecasts are verified at each grid point of the 110°E to 200°E; 30°S to 0° domain for the 1996–2013 DJF period. The evaluation is based on correlation and on the ROC skill score of the upper quintile of precipitation for both weekly targets and Weeks 3–4 outlook. Confirming previous results at the seasonal time scales, the MME reaches the highest skill and also improves the reliability of probabilistic forecasts. However, an equivalent ensemble size comparison between the MME and the individual models shows that the better performance of the MME compared to the best individual models is significantly related to the larger ensemble size of the MME. Forecast skill is then explained in light of potential sources of predictability by evaluating the performance of the models depending on the initial ENSO and MJO state. While the role of ENSO on predictability is quite consistent with its related rainfall anomalies, the role of the MJO is more ambiguous and strongly depends on the location: An initialization in active MJO conditions does not necessarily imply better forecasts. This influence of ENSO and the MJO on predictability does not change when switching from individual models to the MME.

Subseasonal forecasts are based on coupled general circulation models that often have a good representation of large-scale climate drivers affecting rainfall. Yet, they have more difficulty in providing accurate precipitation forecasts. This study proposes a statistical-dynamical post-processing scheme based on a bayesian framework to improve the quality of subseasonal forecasts of weekly precipitation. The method takes advantage of dynamically-forecast precipitation (calibration) and large-scale climate features (bridging) to enhance forecast skill through a statistical model. It is applied to the austral summer precipitation reforecasts in the southwest tropical Pacific, using the Météo-France and ECMWF reforecasts in the Subseasonal-to-seasonal (S2S) database. The large-scale predictors used for bridging are climate indices related to El Niño Southern Oscillation and the Madden–Julian Oscillation, that are the major sources of predictability in the area. Skill is assessed with a Mean Square Skill Score for deterministic forecasts, while probabilistic forecasts of heavy rainfall spells are evaluated in terms of discrimination (ROC skill score) and reliability. This bayesian method leads to a significant improvement of all metrics used to assess probabilistic forecasts at all lead times (from week 1 to week 4). In the case of the Météo-France S2S system, it also leads to strong error reduction. Further investigation shows that the calibration part of the method, using forecast precipitation as a predictor, is necessary to achieve any improvement. The bridging part, and particularly the ENSO-related information, also provides additional discrimination skill, while the MJO-related information is not really useful beyond week 2 over the region of interest.

The relationship between the large-scale intraseasonal variability, synoptic wind regimes, and the local daily variability of precipitation over the main island of New Caledonia (southwest tropical Pacific) is investigated with a focus on the austral summer wet season (November–April). The average diurnal cycle of precipitation over the island is characterized by a sharp afternoon maximum around 1600 local time, with significant differences between the windward east coast, the leeward west coast, and the mountain range. The afternoon peak is related to the afternoon sea-breeze circulation and to the diurnal cycle of convection over land. In general, its magnitude follows the same evolution as the daily mean. In agreement with past studies, a clear modulation of the Madden–Julian oscillation (MJO) on both the diurnal cycle of precipitation and the probability of occurrence of four robust wind regimes can be identified in the New Caledonia region during the wet season. From the evidence that there is a qualitative correspondence between the effects of both the MJO phases and the wind regimes on features in the diurnal cycle of precipitation, a simple model is proposed to inspect the MJO forcing mediated by wind regimes on the diurnal variability of rain. The complete decomposition of the MJO impact shows that the modulation of diurnal cycle by the MJO relies on complex interactions between the MJO and synoptic winds that involve both large-scale MJO convective anomalies and MJO-induced modification of wind patterns.