OSA3.5 | Forecasting on subseasonal to seasonal to decadal timescales
Forecasting on subseasonal to seasonal to decadal timescales
Conveners: Kristina Fröhlich, Dominik Büeler | Co-conveners: Maria Pyrina, Constantin Ardilouze
Orals Mon2
| Mon, 08 Sep, 11:00–13:00 (CEST)
 
Room E1+E2
Posters P-Tue
| Attendance Tue, 09 Sep, 16:00–17:15 (CEST) | Display Mon, 08 Sep, 08:00–Tue, 09 Sep, 18:00
 
Grand Hall, P31–37
Orals Mon2 |
Mon, 11:00
Posters P-Tue
Tue, 16:00
Climate predictions on timescales of several weeks to months to years are becoming increasingly important for society, particularly in the context of adaptation to climate change. Advancing the quality of these forecasts requires further research on the physical processes acting on these different timescales and on how well prediction models capture these processes, as well as on methods extracting the most skilful information from these model forecasts. While contributions to both topics are welcome, the session will particularly focus on the latter aspect. More specifically, we invite contributions on:
i. advancing the climate forecasts with new initialization and ensemble strategies as well as improved model physics of the earth climate system,
ii. post-processing raw model output (e.g., bias correction, (re)calibration, or downscaling with classic or machine-learning-based statistical methods),
iii. translating physical knowledge on local and remote physical drivers of predictability into tools to detect and indicate “windows of forecast opportunity” (e.g., subsampling or weighting of ensemble members or models),
iv. coupling raw model forecasts to impact models to support early warning systems and adaptation strategies (related to extreme events and hazards in the atmosphere, biosphere, and lithosphere, to health, or to energy).

Orals: Mon, 8 Sep, 11:00–13:00 | Room E1+E2

Chairpersons: Constantin Ardilouze, Kristina Fröhlich, Maria Pyrina
11:00–11:15
|
EMS2025-161
|
Onsite presentation
On the application of generative modelling for seasonal climate predictions 
Lluís Palma, Amanda Duarte, Albert Soret, and Markus Donat

Reliable probabilistic predictions at the seasonal time scale are critical for key societal sectors such as agriculture, energy, and water management. Current operational approaches face significant challenges: General Circulation Models (GCMs) are computationally expensive and often limited by low spatial resolution and model deficiencies. At the same time, traditional statistical methods struggle due to significant modelling assumptions, such as linearity or homoscedasticity. Generative models emerge as a cost-effective, promising alternative, offering the potential to model complex nonlinear climate dynamics inherently probabilistically and at a reduced computational cost. Yet, training these algorithms on the short span of current reanalysis datasets results in almost certain overfitting due to the imbalance between trainable parameters and available training samples.

In this context, the present study compares the effectiveness of different generative methodologies in predicting gridded fields of temperature and rainfall seasonal anomalies. The predictions cover all four seasons and are initialised one month before the start of the season, aligning with most climate services providers. We employ climate model output from CMIP6  and CEMS-lens2 during training and ERA5 reanalysis data during testing to circumvent the short span of current reanalysis and observational datasets. We analyse the method's performance in predicting interannual anomalies beyond the climate change-induced trend. We show that the model's ensemble generation capabilities allow it to provide diverse ensemble members, allowing the derivation of relevant probabilistic information and potentially reliable predictions. While climate change trends dominate the skill of temperature predictions, additional skill over the climatological forecast in regions influenced by known teleconnections is found. We reach similar conclusions based on the validation of precipitation predictions.

This work further demonstrates the effectiveness of training generative models on climate model output for seasonal predictions, providing skilful seasonal climate predictions beyond the induced climate change trend at time scales and lead times relevant for user applications, motivating further research.

How to cite: Palma, L., Duarte, A., Soret, A., and Donat, M.: On the application of generative modelling for seasonal climate predictions, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-161, https://doi.org/10.5194/ems2025-161, 2025.

Show EMS2025-161 recording (13min) recording
11:15–11:30
|
EMS2025-670
|
Onsite presentation
Towards an improved representation of the land-surface in seasonal forecasts 
Jonathan Day, Frederic Vitart, Patricia de Rosnay, and Tim Stockdale

Recent studies have shown that the treatment of the land surface and its coupling to the atmosphere is a limiting factor in the skill of seasonal forecasts. Over land, predictive skill is limited by errors originating both from the physical representation of the land-surface processes in the model and from inaccuracies in the initial conditions used to initialize these models. These limitations hinder our ability to produce reliable seasonal predictions, which are increasingly important for climate-sensitive sectors such as agriculture and water resource management.

As part of the Copernicus Climate Change Service Evolution (CERISE) project, significant efforts are underway to address these challenges. A key strategy involves the implementation of advanced land-surface data assimilation techniques, which aim to better represent the state of the land surface at the start of a forecast. Currently land-surface initial conditions for the systems contributing to the Copernicus Climate Change Service (C3S) are generated with free-running land-surface models.

Another promising area of development is the inclusion of time-varying vegetation in the forecast systems. Currently, vegetation characteristics such as leaf area index and land-cover type are held fixed throughout the forecast period. However, research suggests that seasonal changes in vegetation may provide an important source of predictability. By allowing vegetation to evolve realistically over time, models may better capture feedback mechanisms between the land surface and atmosphere leading to improved forecast skill.

In this presentation, we will discuss the impacts of these recent developments on seasonal forecast performance and discuss future prospects for further enhancing land-surface representation to improve forecasts.

How to cite: Day, J., Vitart, F., de Rosnay, P., and Stockdale, T.: Towards an improved representation of the land-surface in seasonal forecasts, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-670, https://doi.org/10.5194/ems2025-670, 2025.

Show EMS2025-670 recording (13min) recording
11:30–11:45
|
EMS2025-34
|
Online presentation
Enhancing Weeks 1-2 Forecasts of 2-m Temperature in the Sierra Nevada, California through Analog Ensemble post-processing 
Zhiqi Yang, Weiming Hu, Agniv Sengupta, Luca Delle Monache, Michael DeFlorio, Mohammadvaghef Ghazvinian, Mu Xiao, Ming Pan, Jacob Kollen, Andrew Reising, Angelique Fabbiani-Leon, David Rizzardo, and Julie Kalansky

California relies on Sierra Nevada spring snowmelt for 60% of its water, supporting 23 million residents. Forecasting the Sierra Nevada snowmelt is a critical component of producing water supply forecasts for water managers, which is undertaken by the California Department of Water Resources (DWR) in of state mandated Bulletin-120 forecasts. Accurate forecast the snowmelt relies on subseasonal 2-meter temperature (T2m) forecasts, which improve snowmelt-driven water storage and streamflow predictions, particularly in higher elevations that contribute the largest surface water input, snowpack depth, and runoff efficiency. Current systems like the California-Nevada River Forecast Center's (CNRFC) Hydrologic Ensemble Forecast Service (HEFS) have identified T2m forecasts as a key uncertainty source. However, research on fine-resolution subseasonal temperature forecasting in complex terrain is limited. This study uses the Parameter-elevation Regressions on Independent Slopes Model (PRISM) dataset as ground truth and National Oceanic and Atmospheric Administration (NOAA) Global Ensemble Forecast System (GEFS) reforecasts to apply Analog Ensemble (AnEn) post-processing, producing high-resolution (4-km) daily T2m forecasts for the Sierra Nevada. We find that during the spring snowmelt season (April–July), AnEn reduces T2m forecast root-mean-squared error by 1°C (60% for 1-day leads, 20% for 15-day leads), increases correlation by ~11%, and extends skill by an additional week beyond dynamical benchmarks. Improvements are more pronounced at higher elevations (e.g., 3000–3500 m), with root-mean-squared error reduced by 4°C, correlation rising from 0.1 to 0.9, and skill extended by over two weeks. By enhancing T2m accuracy for Bulletin-120 and CNRFC-HEFS systems, AnEn can boost the precision of snowmelt and streamflow predictions, supporting improved water resource management in a changing climate. Furthermore, we expand the application of AnEn to enhance subseasonal-to-seasonal forecasts of precipitation and T2m over the western U.S., providing valuable insights for widespread water resource and disaster management.

How to cite: Yang, Z., Hu, W., Sengupta, A., Delle Monache, L., DeFlorio, M., Ghazvinian, M., Xiao, M., Pan, M., Kollen, J., Reising, A., Fabbiani-Leon, A., Rizzardo, D., and Kalansky, J.: Enhancing Weeks 1-2 Forecasts of 2-m Temperature in the Sierra Nevada, California through Analog Ensemble post-processing, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-34, https://doi.org/10.5194/ems2025-34, 2025.

11:45–12:00
|
EMS2025-586
|
Onsite presentation
Towards operational Seasonal Forecasting: statistical downscaling precipitation based on analogue techniques over Iberian Peninsula 
Esteban Rodríguez-Guisado, Marta Domínguez-Alonso, Francisco Javier Pérez Pérez, and Sabela Sanfiz

Downscaling techniques are part of post-processing tools that can improve the potential of seasonal forecast systems for planning in the water resources and agriculture sectors, providing early seasonal anomalies several months in advance (Buontempo et al., 2018). The algorithm presented here is based on analogue synoptic past model circulation variables (predictors) which are linked to the locally observed variable of interest (predictand) through the use of a Euclidean distance metric and multivariate regression. These relationships have been employed to derive, from large-scale atmospheric low-resolution variables (1º x 1º), high-resolution daily precipitation (5km x 5km), covering peninsular Spain and the Balearic Islands. As observations, the AEMET dataset ROCIO_IBEB (Peral et al., 2017) with 5 km of horizontal resolution, has been used.

To assess the performance of the algorithm, the first part of this work presents the evaluation of the ERA5 downscaled precipitation for the period 1997-2016. ROCIO_IBEB was used both to calibrate the method (1981-1996) and to validate the results (1997-2016). Moreover, the analogue techniques have been applied to different realizations of System Models across various seasons. The 20-year hindcast (1997–2016) based on 25 ensemble members has been considered as the model-climatological reference.

The validation results reveal a small bias, low RMSE values and good correspondence of percentiles. For accumulated seasonal precipitation, the following verification scores were computed: forecasted anomaly, lower and upper forecasted probabilities and the ROC area for both the lower and upper terciles. The results demonstrate an enhanced spatial resolution in the probability of occurrence relative to the raw System Models, along with high ROC area values—both spatially and in percentage terms. These findings indicate that, at least during certain seasons and over the Iberian Peninsula, the downscaling algorithm developed by AEMET adds significant value to seasonal forecasts.

How to cite: Rodríguez-Guisado, E., Domínguez-Alonso, M., Pérez Pérez, F. J., and Sanfiz, S.: Towards operational Seasonal Forecasting: statistical downscaling precipitation based on analogue techniques over Iberian Peninsula, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-586, https://doi.org/10.5194/ems2025-586, 2025.

12:00–12:15
|
EMS2025-190
|
Onsite presentation
Balancing Accuracy versus Precision: Enhancing the Usability of Sub-Seasonal Forecasts 
Etienne Dunn-Sigouin, Erik Kolstad, Ole Wulff, Douglas Parker, and Keane Richard

Forecasts are essential for climate adaptation and preparedness, such as in early warning systems and impact models. A key limitation to their practical use is often their coarse spatial grid spacing. However, another less frequently discussed but crucial limitation is that forecasts are often more precise than they are accurate when their grid spacing is finer than the scales they can accurately predict. Here, we adapt the fractions skill score, a metric conventionally used to quantify spatial forecast accuracy by the meteorological community, to help users navigate the trade-off between forecast accuracy versus precision. We demonstrate how this trade-off can be visualized for daily European precipitation, focusing on deterministic predictions of anomalies and probabilistic predictions of extremes, derived from three years of sub-seasonal forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF). Our results show that decreasing precision through spatial aggregation increases forecast accuracy, extends predictable lead times, and enhances the maximum possible accuracy relative to the grid scale, while increased precision diminishes these benefits. Notably, spatial aggregation benefits daily-accumulated forecasts more than weekly-accumulated ones, per unit lead-time.  We demonstrate the practical value of our approach in three examples: communicating early warnings, managing hydropower capacity, and commercial aviation planning—each characterized by distinct user constraints on accuracy, spatial scale, or lead-time. The results suggest a different approach for using forecasts; post-processing forecasts to focus on the most accurate scales rather than the default grid scale, thus offering users more actionable information. XXX XXX XXX XXX XXX XXX XXX XXX XXX XXX

How to cite: Dunn-Sigouin, E., Kolstad, E., Wulff, O., Parker, D., and Richard, K.: Balancing Accuracy versus Precision: Enhancing the Usability of Sub-Seasonal Forecasts, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-190, https://doi.org/10.5194/ems2025-190, 2025.

Show EMS2025-190 recording (11min) recording
12:15–12:30
|
EMS2025-167
|
Onsite presentation
Skill assessment of seasonal forecasts for crop breeding in the Nordic and Baltic region 
Otto Hyvärinen and Andrea Vajda

In Northern Europe, crop growth conditions are specific due to harsh local climate and day length. The NorBalFoodSec project aims at increasing food security in the Nordic and Baltic regions by increasing knowledge on how to better adapt crop breeding and agricultural production to future climates. As part of the project, tailored seasonal climate forecasts relevant for agri-food production are developed and issued to crop breeders to improve the quality of crop management. The limitations in predictability of key variables for growing season, i.e. temperature and precipitation, are investigated and the findings presented.  We have post-processed and evaluated the skill of temperature and precipitation from SEAS5 seasonal forecast system from ECMWF using the CSTools package for R, which implements most of commonly used methods from the literature. These methods range from the simple bias removal to the ensemble calibration methods that correct the bias, the overall forecast variance and the ensemble spread. In addition, we have explored EMOS (nonhomogeneous regression) that makes it easier to add additional information. For precipitation, downscaling methods from CSTools were also explored.  

The skill of the forecasts varies depending on the season and the temporal and spatial aggregation of the forecast data. The use of different verification measures also leads to different estimates of how long the forecast is still skillful: Simpler, non-proper, measures, such as anomaly correlation, indicate skillful forecasts for up to four months, while stricter, proper, measures, such as CRPS, indicate skillful forecasts for only one to two months. Therefore, engaging in discussions with users is crucial to understand what types of forecasts would be most beneficial for them. As the next step, indicators will be developed for breeders based on these variables, such as the widely used Growing Degree Days (GDD) and novel indicators tailored to specific regional needs. 

How to cite: Hyvärinen, O. and Vajda, A.: Skill assessment of seasonal forecasts for crop breeding in the Nordic and Baltic region, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-167, https://doi.org/10.5194/ems2025-167, 2025.

12:30–12:45
|
EMS2025-217
|
Onsite presentation
An opportunity index for subseasonal prediction 
Dominik Büeler, Maria Pyrina, Adel Imamovic, Christoph Spirig, and Daniela Domeisen

The skill of subseasonal atmospheric forecasts has steadily improved in recent decades. Nevertheless, the operational use of such forecasts is still a major challenge for weather prediction centers and weather-dependent socio-economic sectors. A key reason for this challenge is that often only specifically trained forecasters understand and are able to keep track of the complex variety of so-called "windows of forecast opportunity" (WFOs) – periods during which subseasonal prediction skill is enhanced due to specific states of the atmosphere, the ocean, or the land surface acting as drivers of predictability. Here, we propose a novel method to combine the variety of WFOs into a single daily opportunity index, which can be used operationally like a traffic-light system to anticipate enhanced or reduced subseasonal prediction skill in advance. The opportunity index is a linear combination of the standardized anomalies of different known drivers of predictability at forecast initialization. The value of the index is constructed to increase as more WFOs are simultaneously active. Based on 20 years of subseasonal 2m-temperature anomaly hindcasts for Switzerland during summer, we demonstrate that large values of the opportunity index at forecast initialization are able to predict enhanced skill for weekly, two-weekly, and monthly mean anomalies up to four weeks ahead. Systematic sensitivity testing against overfitting indicates year-to-year variability in the performance of the opportunity index, which is something that might be overcome with training on larger hindcast datasets. Given that subseasonal prediction is particularly challenging for Central Europe and during summer, which is the focus of our study, the principle of such a regionally trained index could advance the operational usability of subseasonal predictions in other regions of Europe and the world throughout the year.

How to cite: Büeler, D., Pyrina, M., Imamovic, A., Spirig, C., and Domeisen, D.: An opportunity index for subseasonal prediction, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-217, https://doi.org/10.5194/ems2025-217, 2025.

12:45–13:00
|
EMS2025-56
|
Onsite presentation
Lessons Learned from the Co-Development and Integration of a Subseasonal Forecast into the Yr weather service 
Lonneke van Bijsterveldt, Erik Kolstad, Ivar Seierstad, Thomas Nipen, Anders Sivle, and Jørn Kristiansen
Subseasonal forecasting, which bridges the gap between short-term weather forecasts and seasonal outlooks, covers lead times of 2–6 weeks. Traditionally, these forecasts are presented as anomalies over broad areas due to decreasing skill with lead time. However, this approach significantly limits their usability for laypeople, who are used to highly localized forecasts and often struggle with interpreting anomalies, especially without knowledge of the area's climatology for that time of year.

We therefore want to share the development and implementation process of a localized subseasonal (3-week) forecast, presenting actual weather variables, integrated into the existing Yr weather service. Yr is a weather app and website (www.yr.no) and is a collaboration between the Norwegian Meteorological Institute and the Norwegian Broadcasting Corporation (NRK). The subseasonal forecast has been operational since January 2024 and is available for the Nordic countries and the Baltic states.

During the development process, we prioritized usability over maximizing forecast skill by engaging in co-production with existing Yr users through diary studies, user tests, and feedback forms. This approach provided valuable insights into presenting subseasonal forecasts that users find understandable and actionable. User feedback suggests that despite the inevitable decline in skill by week 3, localized forecasts presenting actual weather variable values—rather than aggregated anomalies—still provide significant value, as long as forecast uncertainty is clearly communicated.

Creating a new forecast product through co-production is time-intensive and highly interactive, involving multiple iterations to test different methods for data dissemination. This process requires continual adjustments to both the design and forecast parameters. We will share our lessons learned by presenting not only the final forecast product but also the intermediate versions that were tested and deemed unfit based on user feedback.

Lastly, we will discuss the integration of the subseasonal forecast into the Yr weather service, with a primary focus on obtaining consistency with existing forecast products (e.g. the 10-day forecast). Ensuring coherence across products is essential, as user feedback highlighted that most users base their decisions on combining information from different forecast products on Yr.

How to cite: van Bijsterveldt, L., Kolstad, E., Seierstad, I., Nipen, T., Sivle, A., and Kristiansen, J.: Lessons Learned from the Co-Development and Integration of a Subseasonal Forecast into the Yr weather service, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-56, https://doi.org/10.5194/ems2025-56, 2025.

Posters: Tue, 9 Sep, 16:00–17:15 | Grand Hall

Display time: Mon, 8 Sep, 08:00–Tue, 9 Sep, 18:00
Chairpersons: Dominik Büeler, Kristina Fröhlich, Constantin Ardilouze
P31
|
EMS2025-136
Statistical Postprocessing of Long-Range Air Temperature Forecasts in the Czech Republic Using Neural Networks 
Stanislava Kliegrová, Ladislav Metelka, Jana Solanská, and Petr Štěpánek

Long-range weather forecasts represent an important tool for planning across various sectors, from agriculture to energy. Their use on a global scale continues to grow, supported by the improving quality of numerical models. However, under the conditions of Central Europe—particularly in the Czech Republic—their reliability and applicability face a number of challenges. This contribution focuses on the potential of statistical postprocessing of long-range air temperature forecasts in the Czech Republic.

Dynamical forecasts use full three-dimensional climate models to simulate potential changes in the atmosphere and oceans over the coming months based on current conditions. Ensembles of simulations provide probabilistic weather scenarios that indicate the likelihood of a given period being wetter, drier, warmer, or colder compared to the seasonal average. The added value of various postprocessing approaches for seasonal forecasts remains a topic of ongoing debate.

This work focuses on statistical postprocessing based on empirical relationships derived between a locally observed predictand of interest (in this case, air temperature) and one or more suitable model predictors from global seasonal forecasting systems. The study analyses the seasonal forecast systems available in the Copernicus Climate Change Service (C3S) archive, which provide near-surface air temperature data at 1° × 1° spatial resolution. It examines the statistical postprocessing of air temperature forecasts for the Czech Republic using four weather forecast systems: the European Centre for Medium-Range Weather Forecasts (ECMWF), Météo-France (MF), Deutscher Wetterdienst (DWD), and Centro Euro-Mediterraneo sui Cambiamenti Climatici (CMCC).

The analysis covers the period 1993–2016, which represents the longest hindcast period common to all systems, and the domain of the Czech Republic in Central Europe (49–51°N, 12–19°E). For statistical postprocessing using a neural network method implemented in STATISTICA software, air temperature and sea level pressure data from global forecast models were used as predictors. The reference data used in this study are gridded station-based observational air temperature datasets. The forecast performance is evaluated across three temperature categories: above normal, normal, and below normal.

How to cite: Kliegrová, S., Metelka, L., Solanská, J., and Štěpánek, P.: Statistical Postprocessing of Long-Range Air Temperature Forecasts in the Czech Republic Using Neural Networks, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-136, https://doi.org/10.5194/ems2025-136, 2025.

P33
|
EMS2025-548
Subseasonal prediction of two extreme precipitation events over Italy 
Daniele Mastrangelo, Paolo Ghinassi, and Silvio Davolio

The Mediterranean basin is often affected by potentially disruptive precipitation events. Particularly during late summer and autumn, a slowly moving deep trough represents the typical synoptic configuration that, causing a continuous inflow of humid southerly currents on the exposed slopes, yields intense and persistent rainfall. Recent studies of extreme precipitation and flood events affecting the Alpine area in northern Italy have revealed that, besides the local contribution due to evaporation from the Mediterranean Sea, a relevant amount of moisture may move from remote areas towards the Mediterranean within narrow and long corridors, i.e. atmospheric rivers.  

Subseasonal forecasting of precipitation extremes is less skillful than other meteorological extremes, however the occurrence of large-scale factors may favor a successful forecast beyond lead week 2. In this work, we test this hypothesis evaluating the predictive ability of the ECMWF and CNR-ISAC operational subseasonal forecasting systems for two extreme precipitation events occurred over Italy in November 2016 and October 2018, the latter also known as the Vaia storm.  

The two forecasting systems predicted positive precipitation anomalies, although underestimated, and higher probabilities for the upper tercile of the model climatological distribution. The analysis of the ensemble members shows that, by reproducing the low/high dipole of geopotential anomaly over the Mediterranean basin on week 3 (days 15–21), a few members are responsible for the precipitation forecast. The higher-skill members are then used to identify the dynamical processes providing enhanced predictability for the selected precipitation events. In both cases, the circulation evolving between North America and North Atlantic in the previous 2 weeks favored the occurrence of an atmospheric river entering the Mediterranean area, a feature more relevant for the Vaia event. 

Although limited to a couple of case studies, this work confirms the possibility to predict precipitation extremes beyond week 2, and the benefit of exploring the whole forecast ensemble to take advantage of the potential useful information provided by a few higher-skill ensemble members. 

The financial support from Next Generation EU, Mission 4, Component 1, CUP B53D23007490006, project ”ARMEX” is acknowledged.

How to cite: Mastrangelo, D., Ghinassi, P., and Davolio, S.: Subseasonal prediction of two extreme precipitation events over Italy, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-548, https://doi.org/10.5194/ems2025-548, 2025.

P34
|
EMS2025-516
Subsampling a multimodel seasonal forecast ensemble in the Mediterranean region 
Francisco Javier Pérez Pérez and Esteban Rodríguez Guisado

Regional Climate Outlook Forums gather experts in climate prediction from operational centers, sharing knowledge and bridging the gap between research, operational centers and users, with the aim to produce seasonal forecast outlooks. Traditionally, the production of information followed a subjective approach of analysis and discussion. Although this procedure adds value from expert knowledge, the process is difficult to trace and reproduce, and not suitable for generating products and applications for decision making. Identifying this issue, WMO encourages Regional Climate Centers and RCOFs to develop objective seasonal forecast procedures. As a step towards that goal in the context of the Mediterranean Climate Outlook Forum (MedCOF), we explored ways of developing an objective approach that adds value to raw model forecasts in the Mediterranean region.
Some works have shown potential improvements in skill by selecting specific members of a multimodel ensemble. Here, we propose a methodology based on comparing the trajectory of ensemble members with the observations, retaining those members whose trajectories were the closest to observed values. For that, we used lead 2 and lead 3 forecasts, which in some cases (particularly for Dec-Feb winter) showed comparable skill to lead 1 forecasts. Then, we compared the state of each ensemble member with the latest ERA5 observations available before the period to be forecast. We kept those members who were closer to observed values and calculated the forecast for months 2-4 (lead 2) or 3-5 (lead 3) of the model run. Comparisons were based on precipitation and sea level pressure observations, but other sources of information could be used, depending on relevant elements for predictability in a particular area. We found local skill improvements by using this method.

How to cite: Pérez Pérez, F. J. and Rodríguez Guisado, E.: Subsampling a multimodel seasonal forecast ensemble in the Mediterranean region, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-516, https://doi.org/10.5194/ems2025-516, 2025.

P35
|
EMS2025-198
ICON-XPP in the CERISE project: towards an LDAS-including seasonal prediction System 
Nils Noll, Vanya Romanova, Kristina Fröhlich, and Martin Lange

The Copernicus Climate Change Service Evolution (CERISE) project, as an EU-HORIZON project, aims to enhance the quality of the C3S (Copernicus Climate Change Service) reanalysis and seasonal forecast portfolio, with a focus on land-atmosphere coupling. CERISE will develop new and innovative ensemble-based coupled land-atmosphere data assimilation approaches and land surface initialisation techniques to pave the way for the next generations of the C3S reanalysis and seasonal prediction systems. Deutscher Wetterdienst is developing its land data assimilation for the initialisation of seasonal forecasts with ICON-XPP. ICON-XPP (ICON eXtended Predictions and Projections) is not only the state-of-art climate modelling system in Germany but also an effort to unify knowledge and experiences from many institutes in one model system that can deliver seamless weather and climate simulations. The aim is to include snow analysis, soil moisture analysis and leaf area index assimilation into our climate forecast data assimilation system. Here we present results from the intermediate step of assimilating snow depth and the impact on multi-year historical forecasts for the period of 1993-2022. The system for generating initial conditions consists at this stage of an Ensemble Kalman Filter for the ocean data assimilation, nudging for the atmosphere and a 2DVAR scheme for the snow. The impact of the assimilation frequency of snow on the historical forecasts will be discussed. Further, we show our efforts in extending our land data assimilation system with soil moisture analysis and leaf area index assimilation based on an Extended Kalman Filter and we show the results of our first sensitivity experiments.

How to cite: Noll, N., Romanova, V., Fröhlich, K., and Lange, M.: ICON-XPP in the CERISE project: towards an LDAS-including seasonal prediction System, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-198, https://doi.org/10.5194/ems2025-198, 2025.

P36
|
EMS2025-672
Soil-moisture-atmosphere coupling hotspots and their representation in seasonal forecasts of boreal summer 
Jonathan Day, Tim Stockdale, Frederic Vitart, Patricia de Rosnay, Constantin Ardilouze, Daniele Peano, Kristina Fröhlich, and Martin Andrews

This study examines the spatial distribution and characteristics of soil-moisture–atmosphere coupling "hotspots" across the Northern Hemisphere during the boreal summer months, and evaluates how well these features are represented in dynamical seasonal forecasting systems. Specifically, the study utilizes hindcasts from the Copernicus Climate Change Service (C3S) multi-model seasonal forecast ensemble to investigate the role of soil moisture anomalies in modulating atmospheric conditions, with a focus on their impact on temperature and precipitation predictability.

The analysis reveals that regions with strong soil-moisture atmosphere coupling such as parts of the western United States, southern Europe, and Eurasia exhibit substantial potential for improving seasonal climate predictions, particularly for surface temperature and rainfall. These regions act as “memory reservoirs,” where soil moisture anomalies can influence atmospheric conditions weeks to months ahead, offering a window of opportunity for enhancing forecast skill.

However, the study also identifies key limitations that currently hamper forecast reliability. Notably, there is considerable uncertainty in the initialization of soil moisture states due to limitations in observational data and inconsistencies across land surface models. In addition, estimates of soil moisture persistence timescales vary significantly between models, impacting the realism of the simulated land–atmosphere interactions.

Furthermore, while some regions demonstrate realistic coupling, others including large areas of North America, Eastern Europe, and Northern India, show evidence of excessive coupling strength, which lead to systematic biases in seasonal temperature forecasts as well as errors in the sign of atmospheric anomaly forecasts. The findings highlight the importance of improving soil moisture initialization and coupling parameterizations to enhance the reliability of seasonal climate forecasts.

How to cite: Day, J., Stockdale, T., Vitart, F., de Rosnay, P., Ardilouze, C., Peano, D., Fröhlich, K., and Andrews, M.: Soil-moisture-atmosphere coupling hotspots and their representation in seasonal forecasts of boreal summer, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-672, https://doi.org/10.5194/ems2025-672, 2025.

P37
|
EMS2025-203
The bias correction of the outputs from RCMs: Comparison of advanced Delta change method with other approaches 
Marta Martinkova and Martin Hanel

As the outputs from regional climate models (RCMs) have to be adjusted and further downscaled to be used, e.g., in hydrology impact studies, the two fundamental approaches are usually implemented. These two approaches differ in how the bias of a climate output is dealt with. The first approach (delta change method) gets the information on the climate signal from a comparison of the model control and future periods. Such information (change factor) is then applied to modify the observational data. The second possibility is to get information on model bias by comparing observational data and model outputs for the control period. This approach is called bias correction, and it uses the information on the bias (correction factor) of the model output to adjust the outputs from the climate model for a future period. Both approaches are based on assumptions that cannot be verified: the stationarity of bias (bias correction) and/or independence of the changes on bias (delta change method).

Here, we present the results of the application of different types of Bias correction and Delta change methods on precipitation data outputs from the model ALADIN-CLIMATE/CZ (CNRM-ESM2-1) for the area of the domain of the model implemented in the project PERUN (roughly the area of the Czech Republic, PERUN domain) in daily time step. The grid of the PERUN domain contains more than 29K points with a spatial resolution of 2.3 km. The applied delta change method is the advanced delta change method. In contrast with the classical delta change method, which reflects only the changes in mean of observations and the model outputs, the advanced delta change method considers changes in means and variability. Along with the advanced delta change method, we tested the performance of several types of bias correction methods (methods based on statistical distribution, parametric transformations, and non-parametric empirical quantiles method). The climate model outputs differ significantly from the precipitation observations, mainly in the mountainous areas. All the tested methods bring the climate model outputs towards the precipitation observations. However, the use of a specific method depends on the specifics of the given impact study.

How to cite: Martinkova, M. and Hanel, M.: The bias correction of the outputs from RCMs: Comparison of advanced Delta change method with other approaches, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-203, https://doi.org/10.5194/ems2025-203, 2025.