SWSabrina Wahl

Sabrina Wahl

*This is an algorithmically generated author record based on publicly available information.

Sabrina Wahl has a total of 93 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Leveraging scale dependent accuracy measures for a fair comparison between AI and NWP weather prediction models

    by Britta Seegebrecht, Sabrina Wahl, Stefanie Hollborn, George Craig, Erik Pavel, Wael Almikaeel, Sebastian Buschow, Martin G. Schultz, Christian Lessig, Ilaria Luise, et al.

    Data-driven weather prediction models based on artificial intelligence (AI) have rapidly advanced in recent years and are frequently reported to outperform traditional physics-based numerical weather prediction (NWP) models for selected verification scores. However, optimization with respect to a specific loss function can adversely affect other metrics, potentially leading to unrealistic forecast characteristics, such as overly smooth spatial structures when mean-squared or mean-absolute error–based loss functions are used.In Bonavita & Geer, 2026 an orthogonal decomposition of the Mean Squared Error (MSE) into Information Error and Noise Error is introduced to unravel different strategies for minimizing this commonly used accuracy metric. These insights allow for a more meaningful interpretation of the MSE as accuracy measure.Additionally, a scale dependent analysis of model performance can help to reveal systematic differences between AI and NWP models specifically on smaller scales such as the effective resolution.The combination of both approaches – the decomposition of the scale dependent MSE into Information and Noise Error – has been derived and applied to different AI and NWP models.First results show the potential of this method to disentangle different effects allowing for a fairer, more comprehensive comparison between AI and NWP weather prediction models.The analysis is partly based on forecasts from the Weather Prediction Model Intercomparison Project (WP MIP), which provides a collection of NWP and AI-model forecasts from multiple national weather services and research institutions.The work is conducted within the RAINA project, which aims to develop a foundation model for the atmosphere with a particular focus on reliable, high-resolution forecasts of extreme wind and precipitation events.Bonavita, M. & Geer, A.J. (2026) Forecast verification using information and noise. Quarterly Journal of the Royal Meteorological Society, e70109. Available from: https://doi.org/10.1002/qj.70109

  • Open Access

    Scale-dependent analysis of the accuracy–activity trade-off in AI weather forecasts

    by Britta Seegebrecht, Sabrina Wahl, Stefanie Hollborn, Erik Pavel, Wael Almikaeel, Michael Langguth, Martin Schultz, Christian Lessig, Ilaria Luise, Juergen Gall, et al.

    Data-driven weather prediction models based on artificial intelligence (AI) have rapidly advanced in recent years and are frequently reported to outperform traditional physics-based numerical weather prediction (NWP) models for selected verification scores. However, optimization with respect to a specific loss function can adversely affect other metrics, potentially leading to unrealistic forecast characteristics, such as overly smooth spatial structures when mean-squared or mean-absolute error–based loss functions are used.A robust and meaningful comparison of AI-based and NWP models therefore requires a carefully chosen and diverse set of verification metrics that accounts for potential dependencies. The main focus is placed on the prominent forecast accuracy-activity tradeoff, associated with the double penalty problem of deterministic forecasts. Related questions include: How sensitive is the relationship between accuracy and activity metrics to the choice of verification measure? Are there systematic differences between AI-based and NWP models? What is the impact of the (in)dependence between the AI training loss function and the verification metrics on the assessment of forecast skill?These questions are addressed using both scale-independent and scale-dependent verification metrics, allowing the quantification of forecast performance on individual spatial scales.As a starting point, global deterministic forecasts are considered. The analysis is partly based on forecasts from the Weather Prediction Model Intercomparison Project (WP MIP), which provides a collection of NWP and AI-model forecasts from multiple national weather services and research institutions.The work is conducted within the RAINA project, which aims to develop a foundation model for the atmosphere with a particular focus on reliable, high-resolution forecasts of extreme wind and precipitation events. Consequently, the relation between, e.g., forecast activity and the predictive capability for extreme weather are of special interest.

  • Open Access

    Observation-based verification of AI weather prediction models: What can we expect?

    by Sabrina Wahl

    Current state-of-the-art artificial-intelligence weather prediction (AI-WP) systems are trained on a large archive of atmospheric reanalysis data. The training objective is to replicate the analysis at a future time step using the previous time steps. Loss functions guide the model to minimize the prediction error on known data. An analysis-based verification of forecasts derived from unseen data will reveal the strength and weaknesses of the AI-WP model in reproducing the statistical and dynamical characteristics of the underlying reanalysis.In contrast, the development and fine-tuning of traditional physics-based numerical weather prediction (NWP) systems relies on verification against observations, with the aim of reducing discrepancies relative to various observational systems. This fundamental difference raises the question of what to expect when applying observation-based verification to AI-WP models that are trained on reanalysis rather than directly on observations.Reanalysis datasets have well-known errors with respect to observations which are documented in literature. Consequently, observation-based verification of AI-WP systems will inherently reflect the observational error characteristics of the reanalysis. Deviations from this expectation are particularly informative: a larger error than that of the reanalysis may indicate deficiencies in emulation, whereas a smaller error raises the question of whether, and from where, additional information beyond the reanalysis has been obtained.To address these questions, we apply the multiple correlation decomposition based on partial correlations introduced by Glowienka-Hense et al. (2020). This method decomposes the explained variance of two different datasets with respect to the same observations into a component of information contained in both datasets (shared explained variance) and the respective added values, i.e., information present in one dataset but not in the other. This decomposition enables quantification of the information transferred from the reanalysis into the forecasts and reveals potential deficiencies, or improvements relative the reanalysis, in the training process. Furthermore, it facilitates comparison of different forecasting systems in terms of there shared and unique information content. The method is demonstrated using 2m-temperature station observations and global deterministic AI-WP and NWP forecasts.Glowienka-Hense et al. (2020): Comparing forecast systems with multiple correlation decomposition based on partial correlation, ASCMO, 6, 103–113, https://doi.org/10.5194/ascmo-6-103-202

  • Open Access

    RAINA - High-resolution nowcasting of precipitation and wind extremes with a foundation model for the atmosphere

    by Erik Pavel, Michael Langguth, Martin G. Schultz, Christian Lessig, Stefanie Hollborn, Jan Keller, Roland Potthast, Britta Seegebrecht, Sabrina Wahl, Juergen Gall, et al.

    Data-driven weather prediction models based on deep learning have been on the rise for several years and have outperformed traditional physics-based numerical models in various benchmark forecasting scores. However, a significant challenge remains: accurately predicting extreme events on a local scale, such as thunderstorms and wind gusts. Previous models struggle in this area, as they were primarily developed for medium-range forecasting and operate at relatively coarse spatio-temporal resolutions. However, the capability of weather models to predict extreme events at a local level is essential for preventing severe consequences for communities, ecosystems, and the financial and material losses they entail. Recently, task-agnostic foundation models, trained on extensive and diverse datasets using self-supervised methods, have demonstrated remarkable skill and robustness, especially in their ability to generalize to rare extreme events. The RAINA project aims to develop a foundation model for the atmosphere, with an emphasis on delivering reliable, high-resolution forecasts of extreme wind and precipitation events. In partnership with the EU Horizon-funded WeatherGenerator project, which aims to create advanced digital twins for Destination Earth, RAINA will extend the pioneering AtmoRep model (Lessig et al., 2023) by employing a multi-modal learning approach. The foundation model seeks to develop a comprehensive, statistically robust, and multi-scale understanding of atmospheric dynamics by incorporating a wide range of meteorological datasets from both models and observations. Innovative deep learning methods, including diffusion models and test-time adaptation, will be investigated to facilitate short-range forecasts of temperature, wind, and precipitation at kilometer-scale resolution over Germany.In a first demonstrator, short-range forecasts are generated using the AtmoRep model and subsequently refined with the CorrDiff downscaling approach (Mardani et al., 2024) that combines a generative diffusion model with a residual approach. This two-step strategy delivers high-resolution forecasts with a maximum lead time of six hours while disentangling uncertainties inherent in the forecasting and downscaling processes, a separation that can enhance training quality when properly applied. By using ERA5 and COSMO REA2 reanalysis data, the approach enhances the precision of high-resolution forecasts over Germany. Initial results from the first demonstrator will be presented in a poster, along with the overall timeline and key milestones of the RAINA project.

  • Open Access

    The German “Hans-Ertel-Centre for Weather Research” (HErZ)

    by Sarah Jones, Mathias W. Rotach, Clemens Simmer, Gerhard Adrian, Maike Ahlgrimm, Sascha-André Albert, George Craig, Hartwig Deneke, Jeannine Ditas, Stephanie Fiedler, et al.

    Fundamental research addressing the challenges of weather forecasts and climate monitoring is required in order to steadily improve the quality of meteorological and climatological services. However, the potential of transferring results from this research into operations increases significantly if carried out in collaboration with national meteorological and hydrometeorological services (NMHS). Out of this motivation and experience, the German “Hans-Ertel-Centre for Weather Research” (HErZ) has been established in 2011 as a virtual competence centre.HErZ develops and organizes a lasting national consortium along common objectives between the German Meteorological Service (DWD), German universities and scientific institutes to accelerate innovation in DWD’s weather, climate and environmental services. HErZ carries out excellent research, strengthens university education and promotes early career scientists, ensures personnel and strategical development, and facilitates the establishment and success of new cooperation, also outside of Germany. Moreover, HErZ grants access to DWD’s infrastructure, including (global) observation data and prediction models. With the strong support of the German Federal Ministry of Digital and Transport, HErZ was able to overcome the difficulties within initiating a cooperation between partners of different institutional communities and established an unprecedented sustainable and national transdisciplinary network.The current research and objectives of HErZ cover the whole value chain expected from this year’s EMS Annual Meeting’s additional focus on “Connecting Communities”: Bringing together fundamental and applied research, observations with prediction models, research institutions with NHMS, and developers with end-users. This intertwined and progressive structure in science and communication ultimately results in acquiring value for society as a whole.Establishing and managing HErZ and its scientific, strategic and structural goals is challenging. The complex tasks of achieving excellence in science and long-term success require a suitable balance between demands and needs of all institutions involved as well as an independent and strong governance by national and international high-level experts.In the future, HErZ must respond to new societal challenges in a targeted but flexible manner with new topics and competences, national and international partners, and a strengthened dialog with stakeholders and users.With this contribution, we would like to share with the audiencethe establishment, structure and peculiarities of HErZ, the lessons learnt from a transdisciplinary consortium connecting NHMS, universities and research institutes, and our visions towards future seamless science and services.

  • Open Access

    Explaining Heatwaves with Machine Learning

    by Sebastian Buschow, Jan Keller, Sabrina Wahl

    The dynamic and thermodynamic drivers of spatially extended climate extremes are difficult to disentangle due to the earth system’s the high dimensionality, interconnectedness and non-linear relationships between individual processes. One approach in the literature relies on carefully selected case studies and uses dynamical models to obtain physical insights into the development of individual extreme episodes. Other studies focus on the statistical relationship between a class of extreme events and one specific driving mechanism. We aim to complement both of these approaches with a machine learning framework in three steps: firstly, the dimensionality of the predictand and a wide range of potential predictor variables is reduced using an appropriate change of basis functions. Secondly, their relationship is modeled by a statistical learner of intermediate complexity — powerful enough to represent the non-linear relationships but simple enough to allow for fast training and extensive experimentation. Lastly, the contribution of each variable to the overall model performance, and to the representation of individual events is assessed using recently developed methods of explainable machine learning.As a first example application, we model European heatwaves in ERA5 data based on potential explanatory variables including geopotential, sea level pressure, soil moisture and the wind components of the jet stream. The predictors are summarized by classic principal component analysis (PCA); for the heatwave fields we rely on a specialized PCA for binary data. A simple neural network is capable of representing a large part of the variability in the reduced space. With the help of Shapley values, we can then quantitatively asses how much information on heatwaves is contained in each variable, and how individual heatwave events differ in terms of the variables by which the model recognizes them. This type of explanation explicitly allows for predictors with overlapping information and nonlinear interactions.  One advantage of our framework is its ability to represent the impact of all suspected drivers on any arbitrarily defined binary event in a single model.

  • Open Access

    Identifying drivers for heat waves using wavelets and machine learning approaches

    by Sebastian Buschow, Jan Keller, Sabrina Wahl

    The driving mechanisms of extreme heat events are known to live on a range of spatio-temporal scales. The occurrence and severity of a heatwave can be influenced by (a) slow variations in the ocean and sub-surface, (b) planetary tele-connections, (c) variations in the jet-stream and synoptic weather systems, as well as (d) local-scale feedbacks.While important progress has been made on each of these individual contributions, fewer studies have attempted to draw a unified picture including them all. We approach this task with tools from classic statistical modeling, as well as image processing machine learning. With the help of wavelet-transforms, predictor variables can be separated into individual scales. Together with local variables and global principal component time-series, these potential drivers are supplied to a statistical learner with the task of reconstructing the field of heatwave occurrences. Contributions from individual scales can then directly be identified, either via variable selection before or during learning, or by measures of feature importance applied to the trained models.We demonstrate this approach for the case of summer heatwaves in the ERA5 reanalysis. If successful, our  framework can also be transferred to other extreme events such as droughts, cold spells or wind storms.

  • Open Access

    Evaluation of wind speed estimates in reanalyses for wind energy applications

    by Sebastian Brune, Jan D. Keller, Sabrina Wahl

    A correct spatio-temporal representation of retrospective wind speed estimates is of large interest for the wind energy sector. In this respect, reanalyses provide an invaluable source of information. However, the quality of the various reanalysis estimates for wind speed are difficult to assess. Therefore, this study compares wind measurements at hub heights from 14 locations in Central Europe with two global (ERA5, MERRA-2) and one regional reanalysis (COSMO-REA6). Employing metrics such as bias, RMSE and correlation, we evaluate the performance of the reanalyses with respect to (a) the local surface characteristics (offshore, flat onshore, hilly onshore), (b) various height levels (60 to 200 m) and (c) the diurnal cycle. As expected, we find that the reanalyses show the smallest errors to observations at offshore sites. Over land, MERRA-2 generally overestimates wind speeds, while COSMO-REA6 and ERA5 represent the average wind speed more realistically. At sites with flat terrain, ERA5 correlates better with observations than COSMO-REA6. In contrast, COSMO-REA6 performs slightly better over hilly terrain, which can be explained by the higher horizontal resolution. In terms of diurnal variation, ERA5 outperforms both other reanalyses. While the overestimation of MERRA-2 is consistent throughout the day, COSMO-REA6 significantly underestimates wind speed at night over flat and hilly terrain due to a misrepresentation of nightly low level jets and mountain and valley breezes. Regarding the representation of downtime of wind turbines due to low/high wind speeds, we find that MERRA-2 is consistently underperforming with respect to the other reanalyses. Here COSMO-REA6 performs better over the ocean, while ERA5 shows the best results over land.


  • Open Access

    A novel approach to surface reanalysis

    by Sabrina Wahl, Clarissa Figura, Jan D. Keller

    Reanalysis is a procedure to merge numerical model integrations and observations to obtain a synergetic representation of the past climatological state of a system, e.g., of the atmosphere. An alternative to running a full reanalysis scheme is a so-called surface reanalysis. Here, an existing reanalysis is used as prior information (for the near-surface state). This first guess is then corrected in a data assimilation step preferrably by applying observations not used in the original assimilation. In such a scheme, an additional downscaling is often performed to enhance the spatial representation of the surface reanalysis.We present here the development of a new approach aiming to establish such a data set based on the COSMO-REA6 regional reanalysis of the Hans-Ertel-Centre and Deutscher Wetterdienst (DWD). The data assimilation step is based on the operational Local Ensemble Transform Kalman Filter (LETKF) of DWD. While the data assimilation is often performed univariately in such surface reanalysis schemes, here we apply it to various parameters at once thus conserving the covariances among the parameters and allowing for a consistent multivariate utilization of the data. Further, this reanalysis will not be restricted to the ground level and near-surface parameters. Instead, it will be extended to the lower part of the boundary layer aiming at an improved representation of wind speeds in wind turbine hub heights especially relevant for renewable energy applications. The envisaged resolution is approximately 1km and therefore enables an enhanced representation of spatial variability and heterogeneity on small scales. In addition, the LETKF is an ensemble-based data assimilation scheme which also provides uncertainty estimates through an ensemble of the re-analyzed parameters which can also be used as input for downstream applications.

  • Open Access

    Evaluation of multi-parameter dependencies in weather and climate simulations

    by Sabrina Wahl, Jan D. Keller

    Weather and climate simulations based on numerical models provide 4-dimensional reconstructions of multiple meteorological parameters describing the atmopsheric state. Yet, the vast majority of evaluation studies focus on the evaluation of single parameters in time and space without looking at the statistical multivariate dependence between the parameters. This, however, is necessary especially with respect to specific events where two or more parameters are involved, i.e., so called compound events. Previous studies have investigated the representation of natural hazards such as wildfires, heat stress, droughts by evaluating corresponding indices based on two or more parameters. Thereby the evaluation process stays in a single parameter framework with well established verification methods at hand. In this work, we present a more sophisticated and generalized approach to investigate physical dependencies between parameters by employing copula theory. With this method, we aim at evaluating the multivariate statistical dependence between two parameters (i.e. the copula) separately from their marginal distributions. This separation enables a more detailed investigation of compound indices (CI) based on the involved parameters. The differences in CI derived from model simulations and observations can now be related to deficiencies of the numerical model due to (i) a misrepresentation of the marginal distributions of the contributing variables, (ii) a misrepresentation of the statistical dependence between the parameters (the copula), (iii) or both. While the method is applicable to all combinations of two parameters, we will present the results of a specific joint copula-based evaluation of temperature and humidity which are the basis for natural hazards mentioned above.