Session assets
Orals: Thu, 1 May | Room 0.11/12
Regionalized seasonal forecasts allow improved decision making, particularly when applying the meteorological forecasts to sectors such as agriculture or water management. In regions like the Blue Nile Basin, a transboundary catchment in East Africa, reliable seasonal predictions are crucial for addressing local needs due to the complex topography combined with high dependency on water resources.
In this study, we introduce Seasonal AFNOCast, a Deep Learning (DL) approach designed to bias-correct and downscale global seasonal forecasts (SEAS5). The objective is to provide a computational efficient approach that provides reliable ensembles, realistic and skillful predictions at a daily and monthly scale. The regionalized forecast provides 51 ensemble members with a 215 day forecast horizon of a spatial resolution of approximately 9 km.
Seasonal AFNOCast is a DL network that applies a specific type of transformer, called the Adapted Fourier Neural Operator (AFNO), in combination with an ensemble-member-specific architecture. The network is trained with the ERA5-Land reanalysis product as reference using an ensemble specific loss function. Its performance is evaluated against Bias-Correction and Spatial Disaggregation (BCSD), a well-established statistical baseline method for post-processing global seasonal forecasts. The evaluation includes comprehensive skill metrics such as the continuous ranked probability score (CRPS), normalized rank histograms, and precipitation-specific metrics, along with qualitative analyses.
Our analysis demonstrates that Seasonal AFNOCast delivers skillful regionalized seasonal predictions that are comparable to, and in specific cases outperform, state-of-the-art statistical methods. These findings underscore the potential of DL-based post-processing of seasonal forecasts, particularly in challenging regions like the Blue Nile Basin.
How to cite: Wiegels, R., Polz, J., Glawion, L., Weber, J. N., Schober, T., Lorenz, C., Chwala, C., and Kunstmann, H.: Seasonal AFNOCast: A Deep Learning Approach for Enhanced Regional Seasonal Predictions, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4729, https://doi.org/10.5194/egusphere-egu25-4729, 2025.
Actionable weather information at the subseasonal timescale informs decision-makers in many societally relevant sectors, including energy demand and supply. However, the predictive skill of subseasonal forecasts varies widely: from forecast ‘busts’ with low predictive skill, to windows of opportunity yielding exceptionally skillful forecasts. It is therefore useful to know ahead of time if a given forecast will be skillful enough to form the basis of operational planning: i.e., along with the forecast itself, users wish to have an a priori estimate of the forecast uncertainty. We propose to achieve this with machine learning (ML). In our study, an ML model trained on historical weather data learns to relate the forecast initial conditions to the probabilistic forecast error at subseasonal lead times. As opposed to ensemble forecasting, this is a computationally cheaper approach to estimate the forecast skill. Moreover, explainability techniques allow us to rank the sources of subseasonal predictability in hindcast data by their importance; a first to our knowledge.
Building on studies that examine the link between the forecast skill of the European Centre for Medium-range Weather Forecasts (ECMWF) subseasonal ensemble model, and the atmospheric conditions at forecast initialization time (weather regime, season, phase of the Madden-Julian Oscillation), we propose a decision-tree-based approach to predicting future forecast skill from past observations. Concretely, a gradient boosted decision tree model is trained to predict the Continuous Ranked Probability Score (CRPS) of ECMWF hindcasts at lead times 0-46 days, by leveraging initial conditions (geopotential height, sea surface temperature, zonal wind speed) extracted from the Earth System Reanalysis 5 (ERA5) dataset. The ERA5 data undergo dimensionality reduction (e.g., principal component analysis) before being fed to the ML model, and are supplemented with pre-computed indices like the El Niño-Southern Oscillation Index. Forecast skill is computed for the 500 hPa geopotential height field in the European region with respect to ERA5 ground truth.
The ML model outperforms a climatological baseline (averaged CRPS by calendar date and lead time) at the task of predicting European forecast skill out to week 7. We find the most important predictor of skill to be the strength of the stratospheric polar vortex, in addition to lead time and calendar date. Training separate models by lead time reveals clear differences in feature importance, such that, for example, lead time contributes the most predictability in the first 2 weeks, while the seasonal cycle is a strong predictor in weeks 3-4. Different teleconnections become important at different lead times, but their predictive potential also fluctuates throughout the year. We will provide an in-depth breakdown of the feature importances by lead time and season in our presentation.
In conclusion, machine learning provides a novel way to estimate a priori the forecast skill of numerical weather prediction models. The presented method enables us for the first time to rank the relative contributions of the sources of forecast skill, as deduced from hindcast data, thereby advancing our understanding of subseasonal predictability.
How to cite: Mârza, A.-C., Domeisen, D. I. V., Ramella-Pralungo, L., and Meyer, A.: Unraveling the sources of subseasonal predictability with machine learning, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4253, https://doi.org/10.5194/egusphere-egu25-4253, 2025.
Seasonal predictability of early winter (November-December) atmospheric patterns is determined, to a large extent, by the anomalous ocean surface thermal conditions. Globally, sea surface temperatures (SSTs) serve as a key driver of wintertime atmospheric patterns, with their predictive importance varying across different regions and time lags. The extratropical regions present greater challenges for seasonal predictability due to the complexity of their atmospheric processes and interaction of signals from different sources of predictability. Globally, sea surface temperatures (SSTs) serve as a key driver of wintertime atmospheric patterns, with their predictive importance varying across different regions and time lags. In the North Atlantic region, seasonal predictability of early winter (November-December) atmospheric patterns can be determined, to a large extent, by the anomalous ocean surface thermal conditions. Nevertheless, current seasonal prediction systems, which rely significantly on the well known interannual phenomenon known as the El Niño-Southern Oscillation (ENSO), develop large biases in the extratropical SSTs, leading to poor performances in other key variables in those regions, such as the Euro-Atlantic region (EAR). Thus, it is important to develop alternative statistical models to overcome these problems.
This study assesses the predictive capability of global SST anomalies with lead times ranging from 1 to 10 months to forecast November-December sea level pressure (SLP) anomalies. For such purpose, we use three different statistical approaches: a Maximum Covariance Analysis (MCA) to identify dominant patterns of co-variability between SSTs and atmospheric conditions; a neural network-based method (NN) designed to capture non-linear teleconnections; and a hybrid methodology that combines the strengths of the MCA and NN techniques.
Our results highlight regions of high predictive skill across the globe, with a focus on understanding how the different initializations impact the predictability. By comparing traditional statistical methods (MCA) with advanced non-linear approaches (NN and Hybrid), this study provides a comprehensive understanding of global atmospheric predictability during early winter. In particular, significant skill in terms of anomaly correlation coefficient is found for the neural network-based methods in the EAR from 7 to 10 months in advance. Additionally, analysis of the non-stationarity of these teleconnections is found and analyzed throughout the period ranging from 1940 to 2019. Furthermore, the non-stationarity of these teleconnections over the whole period is identified and analysed, detecting windows of opportunity for more accurate seasonal forecasts. The findings aim to improve our understanding of oceanic forced atmospheric teleconnections, not only by establishing windows of opportunity for seasonal forecasts, but also by means of analysing possible drivers of these teleconnections. All of these aid in the development of more accurate and robust prediction models for managing climate-related risks worldwide.
How to cite: Galván Fraile, V., Martín-Rey, M., Polo, I., Rodríguez-Fonseca, B., Alonso Balmaseda, M., Rodríguez-Guisado, E., and Moreno-García, M. N.: Enhancing Seasonal Predictions with Machine Learning: A Global Perspective on SST Influence in Early Winter, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17591, https://doi.org/10.5194/egusphere-egu25-17591, 2025.
At subseasonal to seasonal lead times, the forecast skill of extreme events is known to be intermittent and dependent on specific phenomena or conditions, such as a strong El Niño event or sudden stratospheric warming. These states of enhanced predictability in the climate system are termed windows of forecast opportunity. Although this concept is widely recognised, diagnosing windows of opportunity remains an issue and often relies on evaluating conditional model skill, thereby conflating the window of opportunity with the ability of the model to represent it. Furthermore, identifying suitable representations of the dynamical drivers that provide enhanced predictability of a specific extreme event remains a challenge. Here, we propose an information-theoretic diagnostic of windows of forecast opportunity, which can be evaluated in a causal inference framework based on reanalysis data. We apply this diagnostic to characterise the seasonal modulation of subseasonal teleconnections relevant to weather extremes over Europe. Furthermore, we demonstrate the ability of a novel targeted clustering approach based on variational autoencoders to identify circulation regimes that disentangle the drivers of a specific extreme while maintaining their own predictability and physical teleconnections at S2S lead times. Combining the novel diagnostic with the improved representation of dynamical drivers provides a way forward to addressing the challenge of identifying windows of opportunity at subseasonal to seasonal lead times.
How to cite: Kretschmer, M., Spuler, F., and Shepherd, T.: Subseasonal and seasonal windows of forecast opportunity of extreme European winter weather, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17667, https://doi.org/10.5194/egusphere-egu25-17667, 2025.
Outlooks of drier or wetter condition few weeks ahead has significant societal applications. Skillfully forecasting such conditions can enhance the climate prediction services for the agricultural sectors, provide drought outlooks and identify potential windows of wildfire risk. However, conventional numerical weather prediction (NWP) and emerging artificial intelligence (AI) methods have shown limited skill at this lead time, particularly for water-related variables. Hybrid methods, which blends observational data and numerical model using data-driven approach, have demonstrated potential to improve the skill of the forecast at these timescales. Additionally, identifying flow regime is also a widely used method to provide an outlook of temperature and precipitation in the extended range.
In this study, we combine both hybrid and flow regime approach to predict consecutive days without rain within a week. We first use machine learning to identify large-scale flow regimes that modulates weekly precipitation. Subsequently, a Bayesian Framework is employed to infer the posterior distribution of the predictand. This is done by updating the per-grid prior distribution of the predictand using two likelihood components: one derived from preceding large-scale regimes and another based on instantaneous flow regimes provided by extended-range forecasts from the Integrated Forecasting System (IFS).
How to cite: Chu, H. Y.: Subseasonal Prediction of Consecutive Dry Days in Southern Norway using Flow Regimes within a Bayesian Framework, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13640, https://doi.org/10.5194/egusphere-egu25-13640, 2025.
Convectively coupled tropical waves (CCTWs) are modes of intra-seasonal variability that affect circulation and rainfall in the tropics. Here, we specifically examine their impact on Australian rainfall and circulation, as well as their representation in sub-seasonal-to-seasonal (S2S) models, and their predictability. Our findings reveal that CCTWs with off-equatorial convective centres, such as equatorial Rossby waves (ER), mixed Rossby-gravity waves (MRG), and tropical depression-type waves (TD-type), significantly increase the likelihood of extreme rainfall (above the 90th percentile) during austral summer. Specifically, ER waves enhance the probability by approximately 1.5 to 2.4 times, while MRG and TD-type waves increase it by 1.4 to 1.6 times relative to the seasonal average. These effects are comparable with the Madden-Julian Oscillation (MJO), which increases the probability of extreme rainfall by around 1.3 – 2.7 times compared to the seasonal probability. The increased likelihood of extreme rainfall is attributed to the increase in moisture convergence and advection driven by wave activity. These findings highlight the potential to improve S2S predictions by incorporating CCTWs, thereby increasing the accuracy of extreme event forecasts in tropical Australia.
The representation of CCTWs is then assessed in the operational Australian S2S model, ACCESS-S2, a coupled atmosphere-ocean model. We use a 38-year seasonal hindcast period to evaluate representation of CCTWs. We show that the predictability of ER waves and MJO in the filtered outgoing longwave radiation (OLR) field during austral summer extends out to around 9 and 16 days, respectively (r > 0.5). Meanwhile, other CCTWs have shorter skill forecast periods. Space-time spectral analysis also shows that the representation of CCTWs in the OLR field is underestimated. In particular, the relative OLR spectral power of ER waves, Kelvin waves, and the MJO are 20 – 30% less than observations. Moreover, the MRG waves are nearly non-existent in the model. More skill is identified using the filtered lower-level zonal wind (U850), both in terms of predictability and spectral amplitude. For example, considering the U850 only, predictability extends to around 11 days for ER waves and 18 days for the MJO, and the U850-spectra of ER waves mostly indicates less than 10% difference compared to observations, while Kelvin waves and the MJO show less than around 20% differences. However, the MRG is still non-existent in the U850 field. Further cross-spectral analysis demonstrates that there is a weak convection-circulation coupling bias in the model. Overall, this study highlights the role of CCTWs in driving extreme rainfall in tropical Australia through their coupling with convection and circulation. This also identifies current limitations and emphasizes the need to improve the representation of CCTW variability in S2S models to ultimately enhance extreme rainfall prediction in this region.
How to cite: Muhammad, F. R., Vincent, C., King, A., and Lubis, S. W.: Convectively Coupled Tropical Waves and Their Influence on Rainfall in Tropical Australia: Observations and Predictability, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5407, https://doi.org/10.5194/egusphere-egu25-5407, 2025.
Recently, Richter at al. 2024 investigated the sources of predictability from initializing the ocean, atmosphere and land components and verifying S2S predictions against observations. They find that ocean initialization adds little skill in weeks 4-6 and land initializations deteriorate skill in week 1-2. These results point to possible problems with spin-up and coupled model drift. Here we will revisit these results, but in a perfect modeling framework which eliminates model error. For the perfect model, we find that land initializations do contribute to skill, especially in the summer hemisphere. By studying the evolution of the lead-time dependent bias in the actual and perfect predictions, we attempt to disentangle initialization error from coupled model drift.
Richter, J.H., Glanville, A.A., King, T. et al. Quantifying sources of subseasonal prediction skill in CESM2. npj Clim Atmos Sci 7, 59 (2024). https://doi.org/10.1038/s41612-024-00595-4
How to cite: Jaye, A., Berner, J., Glanville, A. S., and Richter, J. H.: Quantifying sources of subseasonal prediction skill in CESM2 in a perfect modeling framework, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14025, https://doi.org/10.5194/egusphere-egu25-14025, 2025.
Recently, there has been pronoued interest in predictability on the
subseasonal-to-seasonal (S2S) timescale. Skill at this forecast range is
only positive, if the lead-time dependent forecast bias is removed.
Recently, Chapman and Berner, 2024, developed an online bias-correction
from nudging tendencies and saw a bias reduction for surface and
free-atmosphere variables of up to 60% in climate simulations. Here, we
quantify the performance of this model-error scheme against
post-processing in S2S-forecasts. We find that the online bias-correction
reduces the bias, but less so than removing the lead-time dependent bias.
Together, they perform better than reducing the lead-time dependent bias
alone.
How to cite: Berner, J., Jaye, A., and Chapman, W. E.: Benefits of online bias-correction versus postproessing methods, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14345, https://doi.org/10.5194/egusphere-egu25-14345, 2025.
Predicting tailored climate extreme events seamlessly from seasons to multi-annual timescales is one of the challenges in the forecasting community. Novel post processing methodologies are required to address this issue, which is discussed in the present study. A new climate application is designed in co-production framework with the agriculture sector to develop the climate information for them on season to two-years timescale using seasonal to the extended seasonal forecast dataset available from the European Centre for Medium-Range Weather Forecasts (ECMWF) Seasonal Forecasting System (SEAS5) for the period 1981-2022. A temporal merging technique is developed to combine the forecasts on season to multi-annual timescale for the actionable climate information for the Frost risk, which affects the vineyard industry in southwestern Europe, in particular focus over Spain, during Spring (March-April) season. We noted a varying level of forecast skill for the Frequency of Frost Days (FFDs) during spring season (target season) in different start dates from lead month-0 (i.e., March start date) to lead month-23 (i.e., May start date). No Forecast Skill is noted for spring FFDs at lead month-2 (i.e., January start date), while a prominent skill is noted at lead month-11 (i.e., May start date). Temporally merging from lead month-23 to month-0 provide a large ensemble size, which have positive feedback onto the FFD’s forecast skill during the spring season. We also noted most of the forecast skill is mainly modulated by the long-term trends in most of the start dates, except for lead month-11 (May start date), while a combination of long-term trends as well as of internal variability (about 60%) is noted to the forecast skill for the FFDs in spring season. This post-processed seamless climate information will be useful for the local vineyard community to take some preventive measures well in advance from the frost risk, which may help to minimize the losses.
How to cite: Abid, M. A., Sarojini, B. B., and Weisheimer, A.: Seamless Climate Information for climate extremes through merging of forecasts across seasonal to multi-annual timescales, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-9459, https://doi.org/10.5194/egusphere-egu25-9459, 2025.
Posters on site: Thu, 1 May, 14:00–15:45 | Hall X5
Based on an elliptic orbit representation of the yearly varying annual cycles of the Northern Hemisphere stratospheric polar vortex (SPV) from 1979 to 2021, we develop a statistical model to predict the parameters of the SPV’s elliptic orbit on a yearly basis. The predictors include indices describing the phase of key climate modes, such as ENSO and the quasi-biennial oscillation (QBO), as well as the initial state of the polar stratosphere, all derived from prior seasons. Our results demonstrate that the predicted annual SPV evolution, initialized on October 1, provides skillful forecasts with anomaly correlation skill exceeding 0.7 throughout the November-to-March period. In particular, our forecasts can accurately predict the timing and magnitude of peak vortex strength, the timing of the final warming, as well as providing insights into the sub-seasonal evolution of the vortex.
How to cite: Secor, M.: Long-Lead Forecasts of the Yearly Varying Annual Evolution of the Northern Hemisphere Stratospheric Polar Vortex, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3789, https://doi.org/10.5194/egusphere-egu25-3789, 2025.
The Madden-Julian Oscillation (MJO) is a critical component of tropical intraseasonal variability, influencing global weather patterns. This study investigates the interaction between the MJO and oceanic waves, specifically Kelvin waves, using indices from Wheeler and Hendon (2004) for the MJO and Rydbeck (2019) for sea surface height (SSH) anomalies. Our methodology involves cross-referencing the phases of the MJO with the phases of the Kelvin Index. We contrast the MJO days with significant oceanic Kelvin wave activity with those when the Kelvin wave signal is weak. By analyzing these intersections, we aim to elucidate the coupling of oceanic waves and the MJO across the three major ocean basins.
Our findings indicate that during significant Kelvin wave activity, there is enhanced convection and more clearly defined oceanic wave structures within the MJO phases in the Pacific basin. This suggests a strong coupling between atmospheric and oceanic processes, where the presence of Kelvin waves can amplify convective activity associated with the MJO. Additionally, we observed that during periods of weak Kelvin wave signals, the MJO tends to be weaker, with more diffuse wave structures. Conversely, in the Atlantic basin, MJO’s impact on ocean Kelvin waves involves episodes of atmospheric convective anomalies over the Amazon, which tend to be related with stronger MJO previous activity in the western Pacific. For the Indian basin, the methodology is able to discriminate the oceanic Kelvin waves triggered by equatorial wind stress anomalies associated with MJO.
We also evaluate 30-yr long simulations from storm-resolving coupled models performed in the framework of EU-NextGEMS project to evaluate their performance in simulating MJO’s footprint on the ocean.
This study provides new insights into the complex dynamics of the MJO and its interaction with oceanic waves, highlighting the importance of considering both atmospheric and oceanic components in understanding tropical variability. The results have significant implications for improving the predictability of the MJO and its associated weather impacts, offering potential advancements in climate modeling and forecasting.
How to cite: Belinchón Martín, F.: The coupling of MJO with oceanic Kelvin waves in the three major oceanic basins, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4325, https://doi.org/10.5194/egusphere-egu25-4325, 2025.
In the tropics, deep convection triggers upper-level quasi-stationary Rossby waves that propagate to higher latitudes and influence local climate patterns. This study examines the teleconnection between the Madden-Julian Oscillation (MJO), the dominant mode of tropical intraseasonal variability, and Antarctica, using daily gridded observational datasets (precipitation from CMAP, 250-hPa geopotential height, 2 m air temperature, and 10 m winds from ERA5, and sea ice concentration from NSIDC) and the Linear Response Theory Method (LRTM) across all southern seasons during 1979-2014. Our results reveal that MJO-driven variations in surface temperature and winds substantially affect Antarctic sea ice concentration throughout the year. In austral summer and autumn, significant sea ice responses are evident in both the eastern (Lazarev Sea to Somov Sea) and western Antarctic sectors (Ross Sea, Amundsen Sea, and Weddell Sea). During summer, the most notable sea ice changes occur in MJO phases 1 and 5, while in autumn, the most potent responses are associated with phases 1–3. Conversely, in winter and spring, the sea ice responses are primarily restricted to the western Antarctic sectors (Ross Sea, Amundsen Sea, Bellingshausen Sea, and Weddell Sea). All MJO phases exert a pronounced influence on sea ice in winter, whereas in spring, phases 1 and 5 dominate. The LRTM effectively elucidates the mechanisms underlying these changes, attributing the observed sea ice variability to wind-driven forcing, thermal advection, or their combined effects.
How to cite: Sen, A., Deb, P., Matthews, A., and Joshi, M.: Investigating the role of the Madden-Julian Oscillation in Antarctic sea ice variability, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6102, https://doi.org/10.5194/egusphere-egu25-6102, 2025.
In this study, the Betts-Miller-Janjić (BMJ) convective adjustment scheme in the Weather Research and Forecasting (WRF) model version 4.0 was used to investigate the effect of its α-parameter, which influences the first-guess potential temperature reference profile on the Madden‒Julian oscillation (MJO) propagation and structure. This study diagnosed the MJO active phase composites of the MJO-filtered outgoing longwave radiation (OLR) during the December-to-January (DJF) period of 2006–2016 over the Indian Ocean (IO), Maritime Continent (MC), and western Pacific (WP). The results show that the MJO-filtered OLR intensity, propagation pattern, and MJO classification (standing, jumping, and propagating clusters) are sensitive to the α-value, but the phase speeds of propagating MJOs are not. Overall, with an increasing α-value, the simulated MJO-filtered OLR intensity increases, and the simulated propagation pattern is improved. Results also show that the intensity and propagation pattern of an eastward-propagating MJO are associated with MJO circulation structures and thermodynamic structures. As α increases, the front Walker cell and the low-level easterly anomaly are enhanced, which premoistens the lower troposphere and triggers more active shallow and congestus clouds. The enhanced shallow and congestus convection preconditions the lower to middle troposphere, accelerating the transition from congestus to deep convection, thereby facilitating eastward propagation of the MJO. Therefore, the simulated MJO tends to transfer from standing to eastward propagating as α increases. In summary, increasing the α-value is a possible way to improve the simulation of the structure and propagation of the MJO.
How to cite: Zhu, X., Zhong, Z., and Lu, Ｗ.: Effect of Parameter Variation in the BMJ Scheme on the Simulation of MJO Propagation and Structure, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7776, https://doi.org/10.5194/egusphere-egu25-7776, 2025.
Tropical cyclones (TCs), which often form over the western North Pacific (WNP), have a large socioeconomic impact and result in destructive damage in East Asian countries. Therefore, it is crucial to estimate TCs characteristics and predict TCs using the model. This study analyzed the subseasonal predictability of TC activities over the WNP region from June to September, using 24 years (1993–2016) of 21-member ensemble hindcasts generated by the Global Seasonal Forecast System version 6 (GloSea6). We analyzed TC activities using dynamic genesis potential index (DGPI) developed by Wang and Murakami, and tracking algorithm (i.e., TempestExtremes (TE)). Compared to IBTrACS best track data, these two methods captured TC genesis points well and showed high correlation in TC genesis density. However, particularly in the South China Sea (SCS), a negative bias was observed in TE, while GPI exhibited a positive or zero bias. Despite using the same input data, different results were observed in this region, and we analyzed the reasons for this discrepancy in two parts. First, why does such a bias occur in DGPI? Second, what causes the differences between DGPI and TE? We used ERA5 data to analyze the relative error and bias of DGPI and examined how westerly wind biases in GloSea6 influenced wind shear and omega errors. In conclusion, one of the key reasons for the differences between the two methods was attributed to the wind shear error induced by the westerly wind bias.
How to cite: Kim, E., Kim, T., and Cha, D.-H.: Subseasonal activities of tropical cyclones over the western North Pacific in GloSea6, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7926, https://doi.org/10.5194/egusphere-egu25-7926, 2025.
Despite recent advances, forecasting European weather on a seasonal timescale remains challenging for both numerical and statistical methods. Weather regimes (WRs), which represent recurrent, quasi-stationary, and persistent states of the atmospheric circulation, are well known to exert considerable influence over the European weather, offering a promising window of opportunity for sub-seasonal to seasonal forecasting. However, while much research has focused on the study of the correlation and the impacts of the WRs on the European weather, the estimation of ground-level climate variables, such as temperature and precipitation, from Euro-Atlantic WRs remains largely unexplored and limited to linear methods.
In this study, we present an AI model designed to reconstruct monthly mean anomalies of the European temperature and precipitation based on the dominant WRs. The model can capture and introduce complex non-linearities in the relation between multiple WRs, describing the state of the Euro-Atlantic atmospheric circulation, and the corresponding surface temperature and precipitation anomalies in Europe. The ability to reconstruct anomalies from WRs constitutes only a portion of the overall challenge. Predicting WRs on a monthly timescale is inherently difficult, and such forecasts are inevitably affected by errors, which can propagate and influence the quality of the reconstructed anomalies. In view of future developments, we examine the effect of inaccuracies in the WRs estimation on the anomalies reconstruction, establishing a lower bound on the WRs prediction accuracy required to outperform the ECMWF seasonal forecast system, SEAS5.
The model utilizes the monthly averages of weather regimes (WRs) to reconstruct the monthly averages of two-meter temperature and total precipitation anomalies during winter (DJF) and summer (JJA). ERA5 and NOAA-CIRES-DOE Twentieth Century Reanalysis datasets are used to compute the WRs and train the AI framework. Using ERA5 as the ground truth, the reconstruction performance is assessed through commonly used metrics, including mean squared error (MSE), anomaly correlation coefficient (ACC), and coefficient of efficiency (CE).
The results presented underline the importance of developing reliable WRs forecasting methods alongside reconstruction models to fully realize the potential of WRs-based forecasting systems. Our findings demonstrate that WRs-based anomaly reconstruction powered by AI-tools offers a viable pathway to better understand and predict seasonal variations.
How to cite: Camilletti, A., Tomasi, E., Franch, G., and Cristoforetti, M.: AI-based reconstruction of European temperature and precipitation anomalies from the Euro-Atlantic weather regimes, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-9206, https://doi.org/10.5194/egusphere-egu25-9206, 2025.
Summer heatwaves are a major concern for electricity distribution companies due to the high electrical loads they can place on urban distribution networks. These load peaks, driven by increased cooling demand, pose a serious threat to network infrastructure by accelerating the deterioration of underground components. During the summer, these components are prone to failure, resulting in cascading blackouts across multiple urban areas. In addition to meteorological forecasting of heat waves, it is therefore crucial to accurately estimate the probability that the electrical load in urban areas will exceed pre-defined thresholds.
In this study, temperature outputs from sub-seasonal forecasts are used to derive probabilistic forecasts of the expected electrical load. A machine learning approach is used, focusing on a single grid point representing the urban area of Milan. The chosen algorithm is Random Forest, where the target variable is the daily electrical load in Milan. The period used to train and validate the algorithm ranges from 2013 to 2023, and the predictors include the Degree Days (DD) and the "week of the year", since the electrical load shows strong seasonal variations.
The time series of the daily load in Milan, used to train the model, shows a significant shift from 2020 onwards due to the pandemic and the associated lockdowns, resulting in lower load values on average with respect to the 2013-2019 period. To ensure comparability between the pre-pandemic and the post-pandemic period (2021-2023), the historical series were detrended using a seasonal trend decomposition (STL) based on LOESS (Locally Estimated Scatterplot Smoothing), making the series almost stationary over the period analysed.
With the detrended electricity load time series, two forecasting models, both based on Random Forest, were implemented and tested. The first, called the Ensemble Model, trains the Random Forest with the Degree Days (DD) derived from ERA5 temperatures for 2013-2019 and applies the learned relationship to each of the bias-corrected seasonal S2S ensemble members for 2021-2023 to predict the electric load in the test period. The final load prediction in this case is the ensemble mean load.
The second approach, called the Quantile-Based model, uses the quantiles of the DD distribution derived from the bias-corrected S2S temperatures as predictors, providing greater flexibility for different ensemble configurations (e.g. 50 or 100 members). It is also tailored to specific forecast lead times and includes a simplified version based on the DD median.
The models have been evaluated using both deterministic and probabilistic metrics. The results indicate that while both models provide more reliable load forecasts than climatology, the Quantile-Based model outperforms the Ensemble Model beyond the third forecast week. It provides probability distributions that are more centred on the observed load, thereby improving forecast reliability.
These forecasting methods can help distribution system operators to address critical peak demand issues with preventive or more timely interventions.
How to cite: Bonanno, R. and Collino, E.: Probabilistic Load Forecasting for the City of Milan based on Subseasonal Predictions, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10659, https://doi.org/10.5194/egusphere-egu25-10659, 2025.
Focusing on the intraseasonal variability (ISV) of the Siberian high (SH), this study found that the SH presents two main periods of 10–25 and 25–70 days, which dominated the cold waves in the winter of 1995/96 and 2005/06 over eastern China (EC), respectively. The influence of these two ISV on East Asian climate is reflected in the evaluation of the East Asian winter monsoon and surface air temperature. The southeastward and downward Rossby wave activity indicates that the upper-level Ural anticyclone is the key to the SH ISV. By utilizing a transformed vorticity budget analysis, distinct dynamic processes in 10–25-day and 25–70-day variability of the SH were further revealed. Forcing from the mean flow acts as a guiding role in both 10–25-day and 25–70-day variability that induces the Ural anticyclone to propagate westward and eastward, respectively. Forcing from the ISV flow is similar to that from the mean flow with a smaller intensity. The dynamic synoptic eddy feedback positively contributes to both the 10–25-day and 25–70-day variability. It promotes (restrains) the westward (eastward) propagation of the Ural anticyclone in the 10–25-day (25–70-day) variability, which may be the main reason for these two distinct ISV of the SH.
How to cite: Ren, H. and Zhou, F.: Distinct dynamic processes in 10–25-day and 25–70-day variability of the Siberian High, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11278, https://doi.org/10.5194/egusphere-egu25-11278, 2025.
Estimating power load, a crucial variable, is essential for optimizing power grid management, especially when forecasts are made months in advance. Weather conditions significantly influence power load; for example, high temperatures lead to increased energy demand for cooling during the summer. Utilizing seasonal weather forecasts to predict future power load represents a promising research direction in this field.
This study utilizes the European Centre for Medium-Range Weather Forecasts (ECMWF) SEAS5 model, which is currently one of the most advanced in seasonal forecasting, to predict power load in a large region of Italy. Given the coarse spatial resolution (~30 km) of SEAS5, developing an application that forecasts the monthly aggregated power load for a large region such as the North Italy market area was appropriate. The method involves calculating degree-days from the predicted temperature and other predictors and then employing a multiple linear regression model to estimate the power load.
The monthly aggregated power load for North Italy is estimated using seasonal forecast data from the ECMWF SEAS5 model at a 0.25° resolution, covering the period from July 2017 to June 2024. The ECMWF SEAS5 system has been providing operational forecasts since 2017, and forecasts are made for horizons ranging from 2 to 7 months ahead. The earlier period (1993-2016) is used for bias correction of the SEAS5 forecasts by comparing them with the ERA5 reanalysis dataset.
Measured load data are retrieved from the European Network of Transmission System Operators for Electricity (ENTSO-E) portal. The data from 2020 are excluded, as they are considered an anomaly, to avoid negatively impacting the training of the forecasting system.
Daily forecast data, predicted 2 to 7 months in advance, are used to calculate degree days and other predictors, which are then translated into predicted power load on a seasonal scale by the multi-linear regression. While daily forecasts at the seasonal scale typically exhibit very low or no skill, we managed to retain some skill by aggregating them over a one-month period. Specifically, this application used forecasts with daily time resolution to estimate monthly cumulative degree days derived from the SEAS5 model data.
The meteorological variables considered include daily maximum and minimum temperatures as well as daily cumulative solar irradiance, spatially aggregated for the area of interest (Northern Italy). To calculate Heating Degree Days (HDD) and Cooling Degree Days (CDD), thresholds of 18°C and 21°C, respectively, were used, reflecting the characteristics of the selected region.
The load forecasting system was evaluated using commonly used metrics, including the mean absolute percentage error (MAPE), mean error, and correlation. The system demonstrates highly promising results, proving to be more skillful up to 7 months ahead compared to climatology and persistency approaches. In these alternative methods, mean meteorological data are used as predictors instead of SEAS5 data (climatology), or the previous year's monthly load observations are directly used as load predictions (persistency).
How to cite: Sperati, S. and Alessandrini, S.: Predicting the Power Load of a Market Area in Italy at the Seasonal Scale, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15260, https://doi.org/10.5194/egusphere-egu25-15260, 2025.
Accurate weather and climate predictions are crucial for agriculture, water management, and disaster preparedness across Africa. However, several studies have highlighted the need to improve rainfall prediction at short-range, medium-range, sub-seasonal, and seasonal timescales. The inability of numerical weather prediction models to reliably capture probabilities of near-normal rainfall, coupled with their overconfidence, poses a significant challenge for many operational weather and climate prediction centers in Africa.
To address these challenges, we developed a machine learning (ML)-based framework for March–May (MAM) seasonal rainfall forecasting within East Africa Region, utilizing Random Forest (RF) and Extreme Gradient Boosting (XGB) models. These models leverage key climatic indicators, including the Indian Ocean Dipole (IOD), Mozambique Channel Trough (MOZ), and Oceanic Niño Index (ONI), computed as lagged indices (December–January) to capture antecedent conditions driving seasonal rainfall. About fifteen climate drivers computed from winds, soil moisture and sea surface temperatures were used as inputs for the machine learning models outputting MAM seasonal total rainfall.
Feature selection using mutual information scoring identified predictors with the strongest relationships to rainfall variability. Separate ML models were developed for each IGAD country to account for the spatial heterogeneity of climatic drivers, ensuring localized precision. A fair forecast performance of the RF and XGB models was achieved so far and also offering advantages in handling complex non-linear relationships.
This study demonstrates the potential of integrating ML with traditional forecasting methods to address the limitations of current model products, providing improved predictions to inform disaster risk reduction and climate adaptation strategies. By advancing the understanding of rainfall drivers, this work supports actionable decision-making for climate resilience within the East Africa region.
How to cite: Chinyoka, S., Steeneveld, G.-J., Vila-Guerau de Arellano, J., Gudoshava, M., Seid Endris, H., and Atheru, Z.: Advancing March-May seasonal rainfall prediction for the East Africa region through Machine Learning to facilitate agriculture management, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21268, https://doi.org/10.5194/egusphere-egu25-21268, 2025.
The Madden-Julian Oscillation (MJO) is the most dominate mode of subseasonal-to-seasonal (S2S) variability, which bridges global weather and climate (Zhang, 2013). The MJO has been recognized as a source of predictability of the global weather on the S2S time scales and can influence onset of the El Nino (e.g., Kerns and Chen, 2021, Jauregui and Chen 2024a, 2024b). We developed a new capability by tracking the multiscale systems like the MJO, atmospheric rivers (ARs), jet stream, the ITCZ, easterly waves, tropical cyclones (TCs), and mesoscale convective systems (MCSs) using Multiscale Objects-Tracking and AI Climate Modeling for Extremes (Mosaic4E). One of the unique capabilities of Mosaic4E is the MJO Large-scale Precipitation Tracking (LPT, Kerns and Chen 2016, 2022) that can identify the MJO large-scale convective heating over the Northern Hemisphere (NH) and Southern Hemisphere (SH) can be used to study teleconnection patterns that are fundamental to extreme rainfall, heat waves and drought, which is not possible with the traditional MJO RMM index. When the MJO convection/precipitation is in the NH, it has a direct impact on the blocking patterns influencing the heatwaves and flooding and drought events. The MJO influence on the global high-impact weather involving heavy rainfall and flooding such as tropical cyclones (TCs) and the atmospheric rivers (ARs) are investigated using Mosaci4E. It is found that the number of TC activities increases 50-100% when MJO LPTs ended up over NH than when is in SH. Similar results are found for the MJO LPTs impacts on the ARs and extreme rainfall over the CONUS. The MJO-LPT represents the S2S time scale bridging the weather and climate and is a key for better understanding and predicting extreme events. The multiscale tracking capability will be enhanced by AI/ML tools in Mosaic4E for identifying, understanding, and predicting the extreme events. Mosaic4E is developed and tested using satellite and in situ observations as well as the ERA5 reanalysis data from 1979-2024. Mosaic4E has shown high skills in coastal flooding over the US and is currently tested globally. The satellite observation and reanalysis data are used to evaluate global weather and climate models. Results show that most models overproduce precipitation over land in non-LPTs and underestimate large-scale precipitation over the oceans compared with the observations. For example, the MJO contributes up to 40-50% of the observed annual precipitation over the Indio-Pacific warm pool region, which are usually much less in the models because of models’ inability to represent the MJO dynamics. Furthermore, the spatial variability of precipitation ENSO is more pronounced in the observations than models.
How to cite: Chen, S., Kerns, B., Mazza, E., and Jauregui, Y.: Global Impacts of MJO Large-Scale Precipitation on Tropical Cyclones, Atmospheric Rivers, and Extreme Rainfall, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21170, https://doi.org/10.5194/egusphere-egu25-21170, 2025.
Posters virtual: Tue, 29 Apr, 14:00–15:45 | vPoster spot 5
EGU25-56 | Posters virtual | VPS2
Links between the Indian Ocean Dipole and Persistent Dry Spells in the Eastern Mediterranean WinterTue, 29 Apr, 14:00–15:45 (CEST) vPoster spot 5 | vP5.6
Persistent Dry Spells (PDS) during winter in the eastern Mediterranean are crucial to understanding the regional challenges of water resources and mitigating agricultural and economic impacts. Winter dry spells significantly affect ecosystem stability, public health, and socioeconomic conditions in a region susceptible to climate variability. Therefore, extending the forecast horizon of these extreme weather events to subseasonal time scales is a key challenge. With this aim, we examine the covariability of the sea surface temperature of the Indian Ocean and Persistent Dry Spells during winter over the eastern Mediterranean. The positive Indian Ocean Dipole (IOD) phase alters global circulation patterns, notably increasing the geopotential height at 500 hPa and the sea-level pressure over western Russia, eastern Europe, and the eastern Mediterranean during PDS events. Concurrently, the positive IOD phase enhances moisture fluxes and decreases sea level pressure and geopotential height at 500 hPa in the Western Mediterranean, suggesting increased cyclonic activity in that region. This type of activity probably influences the formation of PDS in the eastern Mediterranean through latent heating and the formation of ridges downstream of the cyclones. The baroclinic, subtropical, and polar regimes are large-scale synoptic regimes alternately prevailing during PDS events. Changes due to the DMI phase are not identical under these regimes and sometimes have opposite trends. The baroclinic regime is the most frequent regime during PDS events. Consequently, the average changes in pressure intensity during PDS events strongly resemble those during baroclinic days. Positive DMI case studies exemplify the effect of these large-scale regimes. We provide evidence for a link between the positive phase of IOD in December and the frequency of longer (> 15 days) PDS events. The normalized frequencies of persistent 15-20-day events under the positive dipole mode index (DMI) are ~ 2% higher than the frequency of negative DMI. The frequencies of 6-7 day events are ~20% lower. Finally, we emphasize the sensitivity of persistent dry spells during winter to event definition, the chosen precipitation data source, and threshold definitions for climate indices. These considerations are essential for improving the accuracy of regional weather and climate predictions, further enhancing our understanding of the climatic impacts of IOD and other teleconnection patterns in the eastern Mediterranean and worldwide.
How to cite: Berkovic, S. and Hochman, A.: Links between the Indian Ocean Dipole and Persistent Dry Spells in the Eastern Mediterranean Winter, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-56, https://doi.org/10.5194/egusphere-egu25-56, 2025.
EGU25-19687 | ECS | Posters virtual | VPS2
Verification of weather variables linked to Dengue incidence inthe sub‐seasonal scale in VietnamTue, 29 Apr, 14:00–15:45 (CEST) | vP5.7
Dengue fever outbreaks impose a severe healthcare burden in Vietnam, therefore the development of an early Dengue warning system is key to improve public health planning and mitigate the future burden produced by this disease. This study assessed the ECMWF ensemble re-forecast skill for relative humidity, temperature and precipitation, which are key factors for vector-borne disease transmission in Vietnam between 1-4 weeks in advance. We focused the analysis on the rainy season (May-October) using ERA5 reanalysis as a reference dataset. Re-forecast data was pre-processed using a quantile mapping technique to reduce the bias between re-forecast and observations. Results showed that corrected re-forecasts of weekly mean temperature, relative humidity and accumulated precipitation are skilful up to 2-3 weeks in advance and rank histograms verified the forecast reliability. Nonetheless the model is less skillful for the region of South Vietnam and seems to struggle at predicting extremely high/low values of temperature, relative humidity and precipitation. Results from this study demonstrate that ECMWF ensemble forecasts are suitable to use as inputs for a dengue early warning system up to 14-21 days in advance
How to cite: Perez, I., Sparrow, S., Weisheimer, A., Wright, M., and Main, L.: Verification of weather variables linked to Dengue incidence inthe sub‐seasonal scale in Vietnam, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19687, https://doi.org/10.5194/egusphere-egu25-19687, 2025.
