AIFS
Members
First name (team leader)
Jakob
Last name
Schloer
Organisation name
ECMWF
Organisation type
Meteorological Institution
Organisation location
United Kingdom
First name
Steffen
Last name
Tietsche
Organisation name
ECMWF
Organisation type
Meteorological Institution
Organisation location
United Kingdom
First name
Christopher
Last name
Roberts
Organisation name
ECMWF
Organisation type
Meteorological Institution
Organisation location
United Kingdom
First name
Simon
Last name
Lang
Organisation name
ECMWF
Organisation type
Meteorological Institution
Organisation location
United Kingdom
First name
Lorenzo
Last name
Zampieri
Organisation name
ECMWF
Organisation type
Meteorological Institution
Organisation location
United Kingdom
First name
Sara
Last name
Hahner
Organisation name
ECMWF
Organisation type
Meteorological Institution
Organisation location
United Kingdom
First name
Rachel
Last name
Furner
Organisation name
ECMWF
Organisation type
Meteorological Institution
Organisation location
United Kingdom
First name
Mariana
Last name
Clare
Organisation name
ECMWF
Organisation type
Meteorological Institution
Organisation location
United Kingdom
First name
Gareth
Last name
Jones
Organisation name
ECMWF
Organisation type
Meteorological Institution
Organisation location
United Kingdom
Models
Please note that the model information below reflects the latest details provided by the team and does not show previous versions. Only the model summary questionnaires retain a submission history.
Model name
AIFSgaia
Model type
Data-driven
Number of individuals supporting model development:
11-20
Maximum number of Central Processing Units (CPUs) supporting model development or forecast production:
48-1,000
Maximum number of Graphics Processing Units (GPUs) supporting model development or forecast production:
16-64
How would you best classify the IT system used for model development or forecast production:
High-Performance Computing (HPC) Cluster
Model summary questionnaire for model AIFSgaia
The list below shows all model summary questionnaires submitted for this model. Participants may submit an updated questionnaire whenever their methodology changes.
Questionnaires are displayed from the most recent to the earliest, with the latest version opened by default.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
Our forecasts are initialized from the ERA5T reanalysis using the 10 members of the ERA5 Ensemble of Data Assimilations (EDA).
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFSgaia uses two consecutive states (t, t–24 h) from ERA5T and its associated EDA as input. A precise list of input variables is listed in our preprint: https://arxiv.org/abs/2607.05100
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 reanalysis from 1979 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation, cloud cover and top of atmosphere radiotion as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFSgaia follows the architecture introduced in Lang et al. (2024b). It combines a transformer-based graph neural network (GNN) encoder and decoder with a sliding-window transformer processor. Attention is computed along spiral longitudinal bands (Lang et al. 2024a). The processor consists of 16 layers with an embedding dimension of 1024 and 8 attention heads, resulting in approximately 230 million trainable parameters.
Forecast ensembles are generated by conditioning on random noise. For each member, independent noise samples from a standard normal distribution (dimension = latent grid × 4 noise channels) are transformed by a two-layer MLP and injected via conditional layer normalization in each processor layer.
The model is trained to predict the 24-hour difference between t and t+24 h using two consecutive inputs (t, t–24 h). The loss is the fairCRPS computed over 4 ensemble members against the ERA5 target. Training is run for 200k iterations with a batch size of 16, initial learning rate 1e-3, and cosine annealing.
During inference, forecasts are generated autoregressively for 28 steps, producing 200 ensemble members to represent the predictive distribution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
The model is described as AIFS-SUBS-ERA5 in our preprint: https://arxiv.org/abs/2607.05100
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
We withhold 2007–2011 from all training stages and use it as an independent five-year verification set. Five years lies at the lower end of what the S2S community typically uses (around twenty years), but it preserves the bulk of the recent record for training while still providing a meaningful number of independent forecast cases.
Did you face any challenges during model development, and how did you address them?
The main challenges were technical—automating the end-to-end forecast generation and submission pipeline to ensure reproducibility and timely uploads.
Are there any limitations to your current model that you aim to address in future iterations?
AIFS-SUBS forecasts the atmosphere alone. At sub-seasonal lead times, however, much of the predictability arises from the slowly varying boundary conditions such as the land, surface ocean and sea ice. Coupling AIFS-SUBS to land and ocean components is therefore a natural next step (Hahner et al., 2026), and one we expect to be particularly beneficial for surface variables over land and in the tropics.
The 24h time step limits error accumulation over the long autoregressive rollout, but
it comes at a cost: the model sees only instantaneous fields at a single time of day and cannot resolve the diurnal cycle. In future work we aim to stabilise error accumulation so that we can return to a 6h time step, or alternatively predict daily means rather than instantaneous snapshots, either of which would recover subdaily information without sacrificing rollout stability.
Are there any other AI/ML model components or innovations that you wish to highlight?
The full model code and utilities are openly available through the anemoi Python packages, developed as part of the broader ECMWF community effort enabling reproducible AI-based weather forecasting.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
The development of AIFSgaia would have not been possible without the great support by the AIFS Team at ECMWF, covering data preparation, code development, computing architecture and deployment.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
Our forecasts are initialized from the ERA5T reanalysis using the 10 members of the ERA5 Ensemble of Data Assimilations (EDA).
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFSgaia uses two consecutive states (t, t–24 h) from ERA5T and its associated EDA as input. Input variables are:
- temperature, humidity, geopotential height and horizonal winds at 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 hPa,
- surface variables: 10m winds, 2m-temperature, mean sea level pressure, total column water vapour,
- forcing variables: orography, land sea mask, insolation.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 reanalysis from 1979 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation, cloud cover, and top of atmosphere radiation as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFSgaia follows the architecture introduced in Lang et al. (2024b). It combines a transformer-based graph neural network (GNN) encoder and decoder with a sliding-window transformer processor. Attention is computed along spiral longitudinal bands (Lang et al. 2024a). The processor consists of 16 layers with an embedding dimension of 1024 and 8 attention heads, resulting in approximately 230 million trainable parameters.
Forecast ensembles are generated by conditioning on random noise. For each member, independent noise samples from a standard normal distribution (dimension = latent grid × 4 noise channels) are transformed by a two-layer MLP and injected via conditional layer normalization in each processor layer.
The model is trained to predict the 24-hour difference between t and t+24 h using two consecutive inputs (t, t–24 h). The loss is the fairCRPS computed over 4 ensemble members against the ERA5 target. Training is run for 200k iterations with a batch size of 16, initial learning rate 1e-3, and cosine annealing.
During inference, forecasts are generated autoregressively for 28 steps, producing 200 ensemble members to represent the predictive distribution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
AIFSgaia is based on Lang et al., AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score (2024b), http://arxiv.org/abs/2412.15832
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
During development, the model was trained on ERA5 data up to 2018 and verified in the period 2020–2024. After fine-tuning on recent years (2018–2024), additional sanity checks were performed for March–May 2025 forecasts.
Did you face any challenges during model development, and how did you address them?
The main challenges were technical—automating the end-to-end forecast generation and submission pipeline to ensure reproducibility and timely uploads.
Are there any limitations to your current model that you aim to address in future iterations?
Current forecasts tend to be overconfident, with ensemble spread smaller than observed variability. Future work will focus on improved uncertainty calibration and probabilistic spread correction.
Are there any other AI/ML model components or innovations that you wish to highlight?
The full model code and utilities are openly available through the anemoi Python packages, developed as part of the broader ECMWF community effort enabling reproducible AI-based weather forecasting.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
The development of AIFSgaia would have not been possible without the great support by the AIFS Team at ECMWF, covering data preparation, code development, computing architecture and deployment.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
Our forecasts are initialized from the ERA5T reanalysis using the 10 members of the ERA5 Ensemble of Data Assimilations (EDA).
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFSgaia uses two consecutive states (t, t–24 h) from ERA5T and its associated EDA as input. Input variables are:
- temperature, humidity, geopotential height and horizonal winds at 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 hPa,
- surface variables: 10m winds, 2m-temperature, mean sea level pressure, total column water vapour,
- forcing variables: orography, land sea mask, insolation.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 reanalysis from 1979 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation, cloud cover, and top of atmosphere radiation as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFSgaia follows the architecture introduced in Lang et al. (2024b). It combines a transformer-based graph neural network (GNN) encoder and decoder with a sliding-window transformer processor. Attention is computed along spiral longitudinal bands (Lang et al. 2024a). The processor consists of 16 layers with an embedding dimension of 1024 and 8 attention heads, resulting in approximately 230 million trainable parameters.
Forecast ensembles are generated by conditioning on random noise. For each member, independent noise samples from a standard normal distribution (dimension = latent grid × 4 noise channels) are transformed by a two-layer MLP and injected via conditional layer normalization in each processor layer.
The model is trained to predict the 24-hour difference between t and t+24 h using two consecutive inputs (t, t–24 h). The loss is the fairCRPS computed over 4 ensemble members against the ERA5 target. Training is run for 200k iterations with a batch size of 16, initial learning rate 1e-3, and cosine annealing.
During inference, forecasts are generated autoregressively for 28 steps, producing 200 ensemble members to represent the predictive distribution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
AIFSgaia is based on Lang et al., AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score (2024b), http://arxiv.org/abs/2412.15832
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
During development, the model was trained on ERA5 data up to 2018 and verified in the period 2020–2024. After fine-tuning on recent years (2018–2024), additional sanity checks were performed for March–May 2025 forecasts.
Did you face any challenges during model development, and how did you address them?
The main challenges were technical—automating the end-to-end forecast generation and submission pipeline to ensure reproducibility and timely uploads.
Are there any limitations to your current model that you aim to address in future iterations?
Current forecasts tend to be overconfident, with ensemble spread smaller than observed variability. Future work will focus on improved uncertainty calibration and probabilistic spread correction.
Are there any other AI/ML model components or innovations that you wish to highlight?
The full model code and utilities are openly available through the anemoi Python packages, developed as part of the broader ECMWF community effort enabling reproducible AI-based weather forecasting.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
The development of AIFSgaia would have not been possible without the great support by the AIFS Team at ECMWF, covering data preparation, code development, computing architecture and deployment.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
Our forecasts are initialized from the ERA5T reanalysis using the 10 members of the ERA5 Ensemble of Data Assimilations (EDA).
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFSgaia uses two consecutive states (t, t–24 h) from ERA5T and its associated EDA as input. Input variables are:
- temperature, humidity, geopotential height and horizonal winds at 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 hPa,
- surface variables: 10m winds, 2m-temperature, mean sea level pressure, total column water vapour,
- forcing variables: orography, land sea mask, insolation.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 reanalysis from 1979 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation, cloud cover, and top of atmosphere radiation as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFSgaia follows the architecture introduced in Lang et al. (2024b). It combines a transformer-based graph neural network (GNN) encoder and decoder with a sliding-window transformer processor. Attention is computed along spiral longitudinal bands (Lang et al. 2024a). The processor consists of 16 layers with an embedding dimension of 1024 and 8 attention heads, resulting in approximately 230 million trainable parameters.
Forecast ensembles are generated by conditioning on random noise. For each member, independent noise samples from a standard normal distribution (dimension = latent grid × 4 noise channels) are transformed by a two-layer MLP and injected via conditional layer normalization in each processor layer.
The model is trained to predict the 24-hour difference between t and t+24 h using two consecutive inputs (t, t–24 h). The loss is the fairCRPS computed over 4 ensemble members against the ERA5 target. Training is run for 200k iterations with a batch size of 16, initial learning rate 1e-3, and cosine annealing.
During inference, forecasts are generated autoregressively for 28 steps, producing 200 ensemble members to represent the predictive distribution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
AIFSgaia is based on Lang et al., AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score (2024b), http://arxiv.org/abs/2412.15832
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
During development, the model was trained on ERA5 data up to 2018 and verified in the period 2020–2024. After fine-tuning on recent years (2018–2024), additional sanity checks were performed for March–May 2025 forecasts.
Did you face any challenges during model development, and how did you address them?
The main challenges were technical—automating the end-to-end forecast generation and submission pipeline to ensure reproducibility and timely uploads.
Are there any limitations to your current model that you aim to address in future iterations?
Current forecasts tend to be overconfident, with ensemble spread smaller than observed variability. Future work will focus on improved uncertainty calibration and probabilistic spread correction.
Are there any other AI/ML model components or innovations that you wish to highlight?
The full model code and utilities are openly available through the anemoi Python packages, developed as part of the broader ECMWF community effort enabling reproducible AI-based weather forecasting.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
The development of AIFSgaia would have not been possible without the great support by the AIFS Team at ECMWF, covering data preparation, code development, computing architecture and deployment.
Model name
AIFShera
Model type
Data-driven
Number of individuals supporting model development:
11-20
Maximum number of Central Processing Units (CPUs) supporting model development or forecast production:
48-1,000
Maximum number of Graphics Processing Units (GPUs) supporting model development or forecast production:
4-16
How would you best classify the IT system used for model development or forecast production:
High-Performance Computing (HPC) Cluster
Model summary questionnaire for model AIFShera
The list below shows all model summary questionnaires submitted for this model. Participants may submit an updated questionnaire whenever their methodology changes.
Questionnaires are displayed from the most recent to the earliest, with the latest version opened by default.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
AIFShera is initialized from the ERA5T reanalysis using the 10 members of the ERA5 Ensemble of Data Assimilations (EDA).
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFShera ingests two consecutive 6-hourly states (t, t–6 h) from ERA5T EDA. The input variables are:
- temperature, humidity, geopotential height, and horizontal winds at 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 hPa,
- surface variables: 10m winds, 2m-temperature, mean sea level pressure, total column water vapour,
- forcing variables: orography, land sea mask, isolation.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 reanalysis from 1979 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation, cloud cover, and top-of-atmosphere radiation as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFShera follows the AIFS-CRPS architecture described by Lang et al. (2024b). It combines a transformer-based graph neural network (GNN) encoder and decoder with a sliding-window transformer processor. Attention is computed along spiral longitudinal bands (Lang et al. 2024a). The processor consists of 16 layers with an embedding dimension of 1024 and 8 attention heads, resulting in approximately 230 million trainable parameters.
Forecast ensembles are generated by conditioning on random noise. For each member, independent noise samples from a standard normal distribution (dimension = latent grid × 4 noise channels) are transformed by a two-layer MLP and injected via conditional layer normalization in each processor layer.
AIFShera predicts 6-hourly increments using two consecutive inputs (t, t–6 h). The model is trained with the fairCRPS loss over 4 ensemble members against ERA5 targets.
Training is divided into two stages:
1. Pre-training: one-step prediction for 200k iterations (batch size = 16, initial lr = 1e-3, cosine annealing) over 1979–2017.
2. Fine-tuning: progressive autoregressive training (2–12 steps), each trained for 2k iterations on 2017–2024.
At inference, the model performs 112 autoregressive steps (28 days) with 200 ensemble members to sample the predictive distribution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
Apart from minor differences in training schedule, period, and input/output variables, the model is equivalent to Lang et al., 2024b: “AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score” (arXiv:2412.15832).
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
During development, the model was trained on ERA5 data up to 2018 and verified in the period 2020–2024. After fine-tuning on recent years (2018–2024), additional sanity checks were performed for March–May 2025 forecasts.
Did you face any challenges during model development, and how did you address them?
The technical challenges were primarily related to automating forecast generation and submission.
Are there any limitations to your current model that you aim to address in future iterations?
Current forecasts tend to show narrower ensemble spread than observed variability. Future work will focus on improved uncertainty calibration and probabilistic spread correction.
Are there any other AI/ML model components or innovations that you wish to highlight?
Besides its coarser O96 grid resolution, AIFShera closely follows the operational AIFS-ENS medium-range model at ECMWF. It is encouraging to see that a model optimized for medium-range forecasting generalizes well to sub-seasonal timescales.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
AIFShera was developed by the AIFS Team at ECMWF, encompassing efforts in data preparation, model architecture, code development, high-performance computing, and operational deployment.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
AIFShera is initialized from the ERA5T reanalysis using the 10 members of the ERA5 Ensemble of Data Assimilations (EDA).
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFShera ingests two consecutive 6-hourly states (t, t–6 h) from ERA5T EDA. The input variables are:
- temperature, humidity, geopotential height, and horizontal winds at 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 hPa,
- surface variables: 10m winds, 2m-temperature, mean sea level pressure, total column water vapour,
- forcing variables: orography, land sea mask, isolation.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 reanalysis from 1979 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation, cloud cover, and top-of-atmosphere radiation as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFShera follows the AIFS-CRPS architecture described by Lang et al. (2024b). It combines a transformer-based graph neural network (GNN) encoder and decoder with a sliding-window transformer processor. Attention is computed along spiral longitudinal bands (Lang et al. 2024a). The processor consists of 16 layers with an embedding dimension of 1024 and 8 attention heads, resulting in approximately 230 million trainable parameters.
Forecast ensembles are generated by conditioning on random noise. For each member, independent noise samples from a standard normal distribution (dimension = latent grid × 4 noise channels) are transformed by a two-layer MLP and injected via conditional layer normalization in each processor layer.
AIFShera predicts 6-hourly increments using two consecutive inputs (t, t–6 h). The model is trained with the fairCRPS loss over 4 ensemble members against ERA5 targets.
Training is divided into two stages:
1. Pre-training: one-step prediction for 200k iterations (batch size = 16, initial lr = 1e-3, cosine annealing) over 1979–2017.
2. Fine-tuning: progressive autoregressive training (2–12 steps), each trained for 2k iterations on 2017–2024.
At inference, the model performs 112 autoregressive steps (28 days) with 200 ensemble members to sample the predictive distribution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
Apart from minor differences in training schedule, period, and input/output variables, the model is equivalent to Lang et al., 2024b: “AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score” (arXiv:2412.15832).
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
During development, the model was trained on ERA5 data up to 2018 and verified in the period 2020–2024. After fine-tuning on recent years (2018–2024), additional sanity checks were performed for March–May 2025 forecasts.
Did you face any challenges during model development, and how did you address them?
The technical challenges were primarily related to automating forecast generation and submission.
Are there any limitations to your current model that you aim to address in future iterations?
Current forecasts tend to show narrower ensemble spread than observed variability. Future work will focus on improved uncertainty calibration and probabilistic spread correction.
Are there any other AI/ML model components or innovations that you wish to highlight?
Besides its coarser O96 grid resolution, AIFShera closely follows the operational AIFS-ENS medium-range model at ECMWF. It is encouraging to see that a model optimized for medium-range forecasting generalizes well to sub-seasonal timescales.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
AIFShera was developed by the AIFS Team at ECMWF, encompassing efforts in data preparation, model architecture, code development, high-performance computing, and operational deployment.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
AIFShera is initialized from the ERA5T reanalysis using the 10 members of the ERA5 Ensemble of Data Assimilations (EDA).
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFShera ingests two consecutive 6-hourly states (t, t–6 h) from ERA5T EDA. The input variables are:
- temperature, humidity, geopotential height, and horizontal winds at 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 hPa,
- surface variables: 10m winds, 2m-temperature, mean sea level pressure, total column water vapour,
- forcing variables: orography, land sea mask, isolation.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 reanalysis from 1979 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation, cloud cover, and top-of-atmosphere radiation as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFShera follows the AIFS-CRPS architecture described by Lang et al. (2024b). It combines a transformer-based graph neural network (GNN) encoder and decoder with a sliding-window transformer processor. Attention is computed along spiral longitudinal bands (Lang et al. 2024a). The processor consists of 16 layers with an embedding dimension of 1024 and 8 attention heads, resulting in approximately 230 million trainable parameters.
Forecast ensembles are generated by conditioning on random noise. For each member, independent noise samples from a standard normal distribution (dimension = latent grid × 4 noise channels) are transformed by a two-layer MLP and injected via conditional layer normalization in each processor layer.
AIFShera predicts 6-hourly increments using two consecutive inputs (t, t–6 h). The model is trained with the fairCRPS loss over 4 ensemble members against ERA5 targets.
Training is divided into two stages:
1. Pre-training: one-step prediction for 200k iterations (batch size = 16, initial lr = 1e-3, cosine annealing) over 1979–2017.
2. Fine-tuning: progressive autoregressive training (2–12 steps), each trained for 2k iterations on 2017–2024.
At inference, the model performs 112 autoregressive steps (28 days) with 200 ensemble members to sample the predictive distribution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
Apart from minor differences in training schedule, period, and input/output variables, the model is equivalent to Lang et al., 2024b: “AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score” (arXiv:2412.15832).
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
During development, the model was trained on ERA5 data up to 2018 and verified in the period 2020–2024. After fine-tuning on recent years (2018–2024), additional sanity checks were performed for March–May 2025 forecasts.
Did you face any challenges during model development, and how did you address them?
The technical challenges were primarily related to automating forecast generation and submission.
Are there any limitations to your current model that you aim to address in future iterations?
Current forecasts tend to show narrower ensemble spread than observed variability. Future work will focus on improved uncertainty calibration and probabilistic spread correction.
Are there any other AI/ML model components or innovations that you wish to highlight?
Besides its coarser O96 grid resolution, AIFShera closely follows the operational AIFS-ENS medium-range model at ECMWF. It is encouraging to see that a model optimized for medium-range forecasting generalizes well to sub-seasonal timescales.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
AIFShera was developed by the AIFS Team at ECMWF, encompassing efforts in data preparation, model architecture, code development, high-performance computing, and operational deployment.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
AIFShera is initialized from the ERA5T reanalysis using the 10 members of the ERA5 Ensemble of Data Assimilations (EDA).
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFShera ingests two consecutive 6-hourly states (t, t–6 h) from ERA5T EDA. The input variables are:
- temperature, humidity, geopotential height, and horizontal winds at 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 hPa,
- surface variables: 10m winds, 2m-temperature, mean sea level pressure, total column water vapour,
- forcing variables: orography, land sea mask, isolation.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 reanalysis from 1979 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation, cloud cover, and top-of-atmosphere radiation as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFShera follows the AIFS-CRPS architecture described by Lang et al. (2024b). It combines a transformer-based graph neural network (GNN) encoder and decoder with a sliding-window transformer processor. Attention is computed along spiral longitudinal bands (Lang et al. 2024a). The processor consists of 16 layers with an embedding dimension of 1024 and 8 attention heads, resulting in approximately 230 million trainable parameters.
Forecast ensembles are generated by conditioning on random noise. For each member, independent noise samples from a standard normal distribution (dimension = latent grid × 4 noise channels) are transformed by a two-layer MLP and injected via conditional layer normalization in each processor layer.
AIFShera predicts 6-hourly increments using two consecutive inputs (t, t–6 h). The model is trained with the fairCRPS loss over 4 ensemble members against ERA5 targets.
Training is divided into two stages:
1. Pre-training: one-step prediction for 200k iterations (batch size = 16, initial lr = 1e-3, cosine annealing) over 1979–2017.
2. Fine-tuning: progressive autoregressive training (2–12 steps), each trained for 2k iterations on 2017–2024.
At inference, the model performs 112 autoregressive steps (28 days) with 200 ensemble members to sample the predictive distribution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
Apart from minor differences in training schedule, period, and input/output variables, the model is equivalent to Lang et al., 2024b: “AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score” (arXiv:2412.15832).
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
During development, the model was trained on ERA5 data up to 2018 and verified in the period 2020–2024. After fine-tuning on recent years (2018–2024), additional sanity checks were performed for March–May 2025 forecasts.
Did you face any challenges during model development, and how did you address them?
The technical challenges were primarily related to automating forecast generation and submission.
Are there any limitations to your current model that you aim to address in future iterations?
Current forecasts tend to show narrower ensemble spread than observed variability. Future work will focus on improved uncertainty calibration and probabilistic spread correction.
Are there any other AI/ML model components or innovations that you wish to highlight?
Besides its coarser O96 grid resolution, AIFShera closely follows the operational AIFS-ENS medium-range model at ECMWF. It is encouraging to see that a model optimized for medium-range forecasting generalizes well to sub-seasonal timescales.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
AIFShera was developed by the AIFS Team at ECMWF, encompassing efforts in data preparation, model architecture, code development, high-performance computing, and operational deployment.
Model name
AIFSthalassa
Model type
Data-driven
Number of individuals supporting model development:
11-20
Maximum number of Central Processing Units (CPUs) supporting model development or forecast production:
48-1,000
Maximum number of Graphics Processing Units (GPUs) supporting model development or forecast production:
4-16
How would you best classify the IT system used for model development or forecast production:
High-Performance Computing (HPC) Cluster
Model summary questionnaire for model AIFSthalassa
The list below shows all model summary questionnaires submitted for this model. Participants may submit an updated questionnaire whenever their methodology changes.
Questionnaires are displayed from the most recent to the earliest, with the latest version opened by default.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Ensemble-based model, aggregating multiple predictions to assess uncertainty and variability.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
AIFSthalassa is a multi-model mean of AIFSgaia, AIFShera and the previous AIFSthalassa (a version of AIFS-CRPS with surface ocean variables included).
Our models are initialized from ERA5 and ORAS6 reanalysis products.
• The atmospheric variables are initialized from the ERA5T reanalysis using the 10 ensemble members of the ERA5 Ensemble of Data Assimilations (EDA).
• The ocean variables are initialized using the ORAS6 ocean and sea-ice reanalysis. Because ORAS6 becomes available with a latency of about 8–12 days (depending on the alignment between the forecast start time and the 5-day assimilation window of the reanalysis), a separate deterministic machine-learning model was trained to bridge this short gap. This autoregressive model, which uses surface ocean and sea-ice conditions forced by the operational atmospheric analysis, provides the most accurate possible ocean initial conditions during this delay period. Unlike the atmosphere, the ocean initial conditions are identical for all ensemble members.
If any, what data does your model rely on for real-time forecasting purposes?
AIFSthalassa averages the probabilities of three AIFS versions and hence uses their model output.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The underlying models were trained on the ERA5 and ORAS6 reanalyses from 1993 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation and cloud cover as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
The multi-model average is based on three AIFS versions, which are all versions of AIFS-CRPS with
1. AIFShera: AIFS-CRPS architecture with 6h time-step, O96 resolution and atmospheric variables
2. AIFSgaia: AIFS-CRPS architecture with 24h time-step, O96 resolution, and atmospheric + stratospheric variables
3. AIFS-CRPS architecture with 24h time-step, O96 resolution and atmospheric + surface ocean variables
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
Each of the models has a corresponding publication that closely describes it. All models are retrained and might be subject to small deviations from the papers:
1. Lang et al. (2026) – doi:10.1038/s44387-026-00073-7
2. Schloer et al. (2026) – arxiv.org/abs/2607.05100
3. Hahner & Zampieri et al. (2026) - https://arxiv.org/abs/2604.25559
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
Model verification follows the same procedure as for AIFSgaia, using ERA5 data for out-of-sample evaluation.
Did you face any challenges during model development, and how did you address them?
The technical challenges were largely identical to those encountered in AIFSgaia, primarily related to automating forecast generation and submission. .
Are there any limitations to your current model that you aim to address in future iterations?
No answer.
Are there any other AI/ML model components or innovations that you wish to highlight?
No answer.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
AIFSthalassa was developed by the AIFS Team at ECMWF, encompassing efforts in data preparation, model architecture, code development, high-performance computing, and operational deployment.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
Our forecasts are initialized from two distinct reanalysis products.
- The atmospheric variables are initialized from the ERA5T reanalysis using the 10 ensemble members of the ERA5 Ensemble of Data Assimilations (EDA).
- The ocean variables are initialized using the ORAS6 ocean and sea-ice reanalysis. Because ORAS6 becomes available with a latency of about 8–12 days (depending on the alignment between the forecast start time and the 5-day assimilation window of the reanalysis), a separate deterministic machine-learning model was trained to bridge this short gap. This autoregressive model, which uses surface ocean and sea-ice conditions forced by the operational atmospheric analysis, provides the most accurate possible ocean initial conditions during this delay period. Unlike the atmosphere, the ocean's initial conditions are identical for all ensemble members.
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFSthalassa uses two consecutive states (t, t–24 h), similar to AIFSgaia and AIFShera. As described above, the atmospheric input comes from ERA5T and its associated EDA, while the ocean and sea ice input comes from an ORAS6-like model. Input variables are:
- Temperature, humidity, geopotential height, and horizontal winds at 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 hPa.
- Surface variables: 10m winds, 2m-temperature, mean sea level pressure, total column water vapour.
- Ocean variables: sea surface temperature, sea surface height, sea surface salinity, and sea surface velocity (zonal and meridional components).
- Forcing variables: orography, land-sea mask, insolation.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 and ORAS6 reanalyses from 1993 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation and cloud cover as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFSTalassa follows the AIFS-CRPS architecture described by Lang et al. (2024b) and summarized in the AIFSgaia documentation. The main differences with AIFSgaia are:
• The embedding dimension of the processor is 1536 for approximately 500M trainable model parameters.
• Training is run for 300k iterations.
In AIFSthalassa, there is no explicit coupling between the ocean and atmospheric components. The fields are trained jointly in a single model with a common loss, at the same spatiotemporal resolution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
AIFSthalassa is based on Lang et al., AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score (2024b), http://arxiv.org/abs/2412.15832. The specific treatment of the ocean variables is not yet described in a peer-reviewed publication.
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
Model verification follows the same procedure as for AIFSgaia, using ERA5 data for out-of-sample evaluation and one March-May of this year for sanity checks after fine-tuning.
Did you face any challenges during model development, and how did you address them?
The technical challenges were largely identical to those encountered in AIFSgaia, primarily related to automating forecast generation and submission. A forced ML ocean model has been developed to obtain the updated ocean initial conditions that enable us to run the forecast system in real time.
Are there any limitations to your current model that you aim to address in future iterations?
Like AIFSgaia, AIFSthalassa tends to produce an ensemble spread narrower than the observed variability. Future work will focus on improving ensemble calibration.
Are there any other AI/ML model components or innovations that you wish to highlight?
AIFSthalassa is the first model of the AIFS family that explores a prognostic representation of the surface ocean and sea ice as sources of S2S predictability.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
AIFSthalassa was developed by the AIFS Team at ECMWF, encompassing efforts in data preparation, model architecture, code development, high-performance computing, and operational deployment.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
Our forecasts are initialized from two distinct reanalysis products.
- The atmospheric variables are initialized from the ERA5T reanalysis using the 10 ensemble members of the ERA5 Ensemble of Data Assimilations (EDA).
- The ocean variables are initialized using the ORAS6 ocean and sea-ice reanalysis. Because ORAS6 becomes available with a latency of about 8–12 days (depending on the alignment between the forecast start time and the 5-day assimilation window of the reanalysis), a separate deterministic machine-learning model was trained to bridge this short gap. This autoregressive model, which uses surface ocean and sea-ice conditions forced by the operational atmospheric analysis, provides the most accurate possible ocean initial conditions during this delay period. Unlike the atmosphere, the ocean's initial conditions are identical for all ensemble members.
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFSthalassa uses two consecutive states (t, t–24 h), similar to AIFSgaia and AIFShera. As described above, the atmospheric input comes from ERA5T and its associated EDA, while the ocean and sea ice input comes from an ORAS6-like model. Input variables are:
- Temperature, humidity, geopotential height, and horizontal winds at 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 hPa.
- Surface variables: 10m winds, 2m-temperature, mean sea level pressure, total column water vapour.
- Ocean variables: sea surface temperature, sea surface height, sea surface salinity, and sea surface velocity (zonal and meridional components).
- Forcing variables: orography, land-sea mask, insolation.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 and ORAS6 reanalyses from 1993 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation and cloud cover as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFSTalassa follows the AIFS-CRPS architecture described by Lang et al. (2024b) and summarized in the AIFSgaia documentation. The main differences with AIFSgaia are:
• The embedding dimension of the processor is 1536 for approximately 500M trainable model parameters.
• Training is run for 300k iterations.
In AIFSthalassa, there is no explicit coupling between the ocean and atmospheric components. The fields are trained jointly in a single model with a common loss, at the same spatiotemporal resolution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
AIFSthalassa is based on Lang et al., AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score (2024b), http://arxiv.org/abs/2412.15832. The specific treatment of the ocean variables is not yet described in a peer-reviewed publication.
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
Model verification follows the same procedure as for AIFSgaia, using ERA5 data for out-of-sample evaluation and one March-May of this year for sanity checks after fine-tuning.
Did you face any challenges during model development, and how did you address them?
The technical challenges were largely identical to those encountered in AIFSgaia, primarily related to automating forecast generation and submission. A forced ML ocean model has been developed to obtain the updated ocean initial conditions that enable us to run the forecast system in real time.
Are there any limitations to your current model that you aim to address in future iterations?
Like AIFSgaia, AIFSthalassa tends to produce an ensemble spread narrower than the observed variability. Future work will focus on improving ensemble calibration.
Are there any other AI/ML model components or innovations that you wish to highlight?
AIFSthalassa is the first model of the AIFS family that explores a prognostic representation of the surface ocean and sea ice as sources of S2S predictability.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
AIFSthalassa was developed by the AIFS Team at ECMWF, encompassing efforts in data preparation, model architecture, code development, high-performance computing, and operational deployment.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Machine learning-based weather prediction.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
Our forecasts are initialized from two distinct reanalysis products.
- The atmospheric variables are initialized from the ERA5T reanalysis using the 10 ensemble members of the ERA5 Ensemble of Data Assimilations (EDA).
- The ocean variables are initialized using the ORAS6 ocean and sea-ice reanalysis. Because ORAS6 becomes available with a latency of about 8–12 days (depending on the alignment between the forecast start time and the 5-day assimilation window of the reanalysis), a separate deterministic machine-learning model was trained to bridge this short gap. This autoregressive model, which uses surface ocean and sea-ice conditions forced by the operational atmospheric analysis, provides the most accurate possible ocean initial conditions during this delay period. Unlike the atmosphere, the ocean's initial conditions are identical for all ensemble members.
If any, what data does your model rely on for real-time forecasting purposes?
For real-time forecasting, AIFSthalassa uses two consecutive states (t, t–24 h), similar to AIFSgaia and AIFShera. As described above, the atmospheric input comes from ERA5T and its associated EDA, while the ocean and sea ice input comes from an ORAS6-like model. Input variables are:
- Temperature, humidity, geopotential height, and horizontal winds at 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50 hPa.
- Surface variables: 10m winds, 2m-temperature, mean sea level pressure, total column water vapour.
- Ocean variables: sea surface temperature, sea surface height, sea surface salinity, and sea surface velocity (zonal and meridional components).
- Forcing variables: orography, land-sea mask, insolation.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
The model was trained on the ERA5 and ORAS6 reanalyses from 1993 – 2024 with 6-hourly temporal resolution on an o96 grid (~1°). In addition to the input variables, the model predicts precipitation and cloud cover as diagnostic outputs.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
AIFSTalassa follows the AIFS-CRPS architecture described by Lang et al. (2024b) and summarized in the AIFSgaia documentation. The main differences with AIFSgaia are:
• The embedding dimension of the processor is 1536 for approximately 500M trainable model parameters.
• Training is run for 300k iterations.
In AIFSthalassa, there is no explicit coupling between the ocean and atmospheric components. The fields are trained jointly in a single model with a common loss, at the same spatiotemporal resolution.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
AIFSthalassa is based on Lang et al., AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score (2024b), http://arxiv.org/abs/2412.15832. The specific treatment of the ocean variables is not yet described in a peer-reviewed publication.
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
Model verification follows the same procedure as for AIFSgaia, using ERA5 data for out-of-sample evaluation and one March-May of this year for sanity checks after fine-tuning.
Did you face any challenges during model development, and how did you address them?
The technical challenges were largely identical to those encountered in AIFSgaia, primarily related to automating forecast generation and submission. A forced ML ocean model has been developed to obtain the updated ocean initial conditions that enable us to run the forecast system in real time.
Are there any limitations to your current model that you aim to address in future iterations?
Like AIFSgaia, AIFSthalassa tends to produce an ensemble spread narrower than the observed variability. Future work will focus on improving ensemble calibration.
Are there any other AI/ML model components or innovations that you wish to highlight?
AIFSthalassa is the first model of the AIFS family that explores a prognostic representation of the surface ocean and sea ice as sources of S2S predictability.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
AIFSthalassa was developed by the AIFS Team at ECMWF, encompassing efforts in data preparation, model architecture, code development, high-performance computing, and operational deployment.
Submitted forecast data in previous period(s)
Please note: Submitted forecast data is only publicly available once the evaluation of a full competitive period has been completed. See the competition's full detailed schedule with submitted data publication dates for each period here.
Access forecasts data