SMSimon L. L. Michel

Simon L. L. Michel

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Simon L. L. Michel has a total of 39 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Comment on egusphere-2026-5397

    by Simon L. L. Michel, Kristian Strommen, Hannah M. Christensen

    Regional climate projections are usually sharpened by constraining models’ forced responses toward observations, a step that filters out internal variability and works well mainly where the signal-to-noise ratio is high. Here we side-step this limitation with deep learning, treating internal variability not as noise to remove but as part of the future to predict and of the uncertainty to report. A U-Net trained on 290 simulations from 40 CMIP6 Earth System Models maps the 1980 to 2025 evolution of surface air temperature onto decadal-mean monthly climatologies from the 2030s to the 2090s under SSP2-4.5, keeping the forced response and internal variability together. Tested on entirely withheld model families, the forecasts are skilful and empirically calibrated out-of-sample. Within a model's own ensemble, a forecast predicts the future of the specific member it was given more accurately than the futures of that model's other members, a skill that persists to the 2090s, notably over the subpolar North Atlantic, Europe, North America and Southern Ocean. The forecasts also generalise to a next-generation eddy-rich model whose resolved physics lies outside the training distribution, supporting application to real-world data. When applied to observations, the framework reduces uncertainty in global-mean warming similarly to observational constraints, while also providing spatially coherent local predictions. This shows that our deep learning approach can estimate the distribution of future local climate, conditioned on the recent observed trajectory, encompassing uncertainty arising from both forced response and internal variability realisations. This extends observationally constrained projection toward calibrated predictions at the local scale and seasonal resolution where adaptation decisions are made.

  • Open Access

    Comment on egusphere-2026-5397

    by Simon L. L. Michel, Kristian Strommen, Hannah M. Christensen

    Regional climate projections are usually sharpened by constraining models’ forced responses toward observations, a step that filters out internal variability and works well mainly where the signal-to-noise ratio is high. Here we side-step this limitation with deep learning, treating internal variability not as noise to remove but as part of the future to predict and of the uncertainty to report. A U-Net trained on 290 simulations from 40 CMIP6 Earth System Models maps the 1980 to 2025 evolution of surface air temperature onto decadal-mean monthly climatologies from the 2030s to the 2090s under SSP2-4.5, keeping the forced response and internal variability together. Tested on entirely withheld model families, the forecasts are skilful and empirically calibrated out-of-sample. Within a model's own ensemble, a forecast predicts the future of the specific member it was given more accurately than the futures of that model's other members, a skill that persists to the 2090s, notably over the subpolar North Atlantic, Europe, North America and Southern Ocean. The forecasts also generalise to a next-generation eddy-rich model whose resolved physics lies outside the training distribution, supporting application to real-world data. When applied to observations, the framework reduces uncertainty in global-mean warming similarly to observational constraints, while also providing spatially coherent local predictions. This shows that our deep learning approach can estimate the distribution of future local climate, conditioned on the recent observed trajectory, encompassing uncertainty arising from both forced response and internal variability realisations. This extends observationally constrained projection toward calibrated predictions at the local scale and seasonal resolution where adaptation decisions are made.

  • Open Access

    From climate projections to trajectory-conditioned local predictions

    by Simon L. L. Michel, Kristian Strommen, Hannah M. Christensen

    Regional climate projections are usually sharpened by constraining models’ forced responses toward observations, a step that filters out internal variability and works well mainly where the signal-to-noise ratio is high. Here we side-step this limitation with deep learning, treating internal variability not as noise to remove but as part of the future to predict and of the uncertainty to report. A U-Net trained on 290 simulations from 40 CMIP6 Earth System Models maps the 1980 to 2025 evolution of surface air temperature onto decadal-mean monthly climatologies from the 2030s to the 2090s under SSP2-4.5, keeping the forced response and internal variability together. Tested on entirely withheld model families, the forecasts are skilful and empirically calibrated out-of-sample. Within a model's own ensemble, a forecast predicts the future of the specific member it was given more accurately than the futures of that model's other members, a skill that persists to the 2090s, notably over the subpolar North Atlantic, Europe, North America and Southern Ocean. The forecasts also generalise to a next-generation eddy-rich model whose resolved physics lies outside the training distribution, supporting application to real-world data. When applied to observations, the framework reduces uncertainty in global-mean warming similarly to observational constraints, while also providing spatially coherent local predictions. This shows that our deep learning approach can estimate the distribution of future local climate, conditioned on the recent observed trajectory, encompassing uncertainty arising from both forced response and internal variability realisations. This extends observationally constrained projection toward calibrated predictions at the local scale and seasonal resolution where adaptation decisions are made.

  • Open Access

    Local temperatures rapidly depart from historical experience beyond 1.5°C

    by Simon Michel, Hannah Christensen

    Local temperatures increasingly exceed values without precedent in the historical record 1,2 , but when these excursions become permanent 3,4 is uncertain due to the underestimated role of internal climate variability 5 . Here we recast the problem in probabilistic terms by asking, for a given global warming level 2,6,7 , how likely is it that annual mean local temperature will remain above its 1850-1999 maximum through 2100, given the preceding climate history. We train a neural network 8 on 408 CMIP6 simulations 9 under SSP2-4.5 10 , the emissions trajectory closest to current policy commitments 11-13 , and obtain calibrated probabilities skilful under out-of-sample validation. Applied to observations, regions where temperatures remain above their historical maximum are confined to just 2.2% of global land area at 1.5°C, but expand rapidly thereafter, reaching nearly half the present-day global population 14 and 35% of aboveground biomass 15 by 2°C, and most of humanity and nearly 80% of aboveground biomass by 3°C. Current policy trajectories 11,13 will therefore commit a rapidly growing share of humanity and Earth’s ecosystems to twenty-first century local temperatures sustained above their historical maxima.

  • Caractérisation des variations climatiques du dernier millénaire à l'aide de nouvelles méthodes statistiques : reconstructions paléoclimatiques et intégration dans un modèle de circulation globale du climat

    by Simon Michel

    La variabilité climatique joue un rôle considérable dans de nombreux domaines de nos vies. L’augmentation globale des températures, forcée par les émissions croissantes de gaz à effet de serre d’origine anthropique, a connu une période de hiatus entre 1998 et 2012. Les causes de cette séquence font l’objet de controverses scientifiques concernant les rôles relatifs de la variabilité intrinsèque au système climatique, et des forçages externes. Afin de résoudre ce type de débat, il est nécessaire de mieux comprendre les modes de variabilité décennaux, issus de l’organisation à grande échelle du système climatique. Cependant, 150 années d’observations instrumentales ne permettent pas d’estimer de manière rigoureuse leurs caractéristiques et donc de comprendre leur dynamique.Cette thèse cherche à parfaire la compréhension de la variabilité climatique multi-décennale en produisant des reconstructions sur le dernier millénaire d’indices ou de grilles climatiques. Pour cela, on se propose d’introduire des méthodes statistiques modernes, dites de machine learning, appliquées à des données paléoclimatiques issues d’archives naturelles (cernes d’arbres, carottes de glace..). La base de données PAGES 2K récemment développée propose plus de 700 enregistrements sur le dernier millénaire et sera abondamment utilisée pour construire les modèles statistiques proposés.Des métriques mathématiques objectives permettent de montrer que l’application d’une méthode non-linéaire telle que la forêt aléatoire produit en général des résultats plus robustes que les méthodes linéaires usuelles. On utilise donc la forêt aléatoire pour reconstruire des variations du mode principal des températures de surface océanique (SST) de l’Atlantique Nord, l’Atlantic Multidecadal Variability (AMV), notamment liée à la circulation océanique de retournement dans ce bassin (AMOC). Cette reconstruction suggère que des changements dans cette circulation, survenus à la fin du XIIe siècle, furent probablement le catalyseur d’une entrée précoce dans le petit âge de glace, une période relativement froide du dernier millénaire. Les importantes activités volcaniques des XIIIe, XVe et XIXe siècles sont cependant identifiées comme les causes principales de la prolongation du petit âge de glace dont le paroxysme se situe au XIXe siècle. De plus, des changements récents dans la variabilité de l’AMV semblent, d’après la théorie des systèmes dynamiques, être un signal d’approche d’un point de bascule de l’AMOC.Une reconstruction des SST sur une grille a également été produite par forêt aléatoire sur tout le millénaire. Cette reconstruction montre que les phases positives de l’AMV auraient tendance à conduire, environ une décennie après, à une phase négative du mode principal des SST du Pacifique, l’Interdecadal Pacific Variability (IPV). Comme les observations montrent que le hiatus de 1998-2012 correspond à une phase négative de l’IPV, ces résultats soutiennent l’hypothèse que l’AMV est l’un des moteurs principaux de telles périodes de hiatus. Dans la lignée des hypothèses de ralentissement de l’AMOC déduites du spectre de l’AMV reconstruite, cette grille suggère que celle-ci atteint aujourd‘hui un niveau bas sans précédent, depuis au moins l’an 850.Enfin, cette même grille de reconstruction a été assimilée par méthode de rappel en SST au sein du modèle de circulation globale IPSL-CM5A2-LR. Bien que la simulation reproduit convenablement les températures et paléo-températures mesurées, la méthode de rappel semble moins efficace que des méthodes d’assimilation plus complexes tel que le filtre de Kalman. De plus, l’AMOC simulée dans cette réanalyse ne correspond pas à son estimation à partir de méthodes statistiques, sans doute du fait que la salinité, est déterminante pour forcer la convection océanique et l’AMOC. Ainsi, assimiler la salinité semble nécessaire pour reproduire correctement les variations d’AMOC sur le dernier millénaire à l’aide d’un modèle.

  • Open Access

    Short- to long-range climate forecasts with deep learning

    by Simon Michel, Kristian Strommen, Hannah Christensen

    Uncertainty in projections of future regional climate change remains large, driven by structural differences among Earth System Models and the influence of internal climate variability. Existing uncertainty-reduction approaches, including emergent constraints and Bayesian variants, primarily focus on forced climate responses derived from simple aggregate metrics, thereby requiring strong assumptions and exploiting only low-dimensional climate information. Here we propose a data-driven deep-learning framework that directly forecasts spatially and monthly resolved decadal mean climatologies of surface temperature anomalies from the 2030s to the 2090s, using only recent monthly trajectories spanning 1980-2025. The training ensemble contains 265 historical+SSP2-4.5 simulations, distributed across 40 ESMs from 25 different families (i.e., modelling centers) over which the cross validation is performed. The architecture couples pluri-annual to multi-decadal temporal convolutions with a spatial U-Net encoder-decoder and is evaluated on CMIP6 simulations using a leave-one-model-family-out cross-validation (LOMFO-CV) design to ensure generalisation across separately developed ESMs. Predictive uncertainty is quantified via LOMFO-CV errors, yielding conservative and reliable ranges that incorporate irreducible internal variability and systematic model shifts.To further evaluate the predictive capacity beyond the CMIP6 distribution, we evaluated the network on historical+SSP2-4.5 simulations from a recent HadGEM3-GC5 model hierarchy developed within the European Eddy-Rich ESMs (EERIE) project, the European contribution to HighResMIP2 for CMIP7. In particular, the eddy-rich GC5-HH configuration explicitly simulates mesoscale ocean dynamics that are absent in CMIP6-type models, providing a rigorous test of generalisation to richer and more realistic physical representations. Despite these substantial differences, the network successfully reproduces warming trajectories and future climate patterns for all three model configurations (GC5-LL, GC5-MM, GC5-HH), with forecast errors largely contained within empirically calibrated uncertainty bounds from the LOMFO-CV, both globally and locally. These results, notably for GC5-HH and its more realistic physics, strengthens confidence in the applicability of the framework to real-world data.When applied to observations, the extracted end-of-century global-mean surface temperature and its uncertainty range are consistent with prior estimates from Bayesian frameworks. At local scales, the network reduces uncertainty by 40% (2030s) to 30% (2090s) on average, and by up to 75% in some regions for all future decades. Importantly, these uncertainty estimates account not only for uncertainty in the forced response (as emergent constraint methods do), but also for errors associated with predicting different realisations of internal variability, providing a physically meaningful reduction of local and global climate uncertainty.

  • Open Access

    Deep learning based reconstructions of the Atlantic meridional overturning circulation confirm twenty-first century decline

    by Simon L L Michel, Henk A Dijkstra, Francesco Guardamagna, Valérian Jacques-Dumas, René M van Westen, Anna S von der Heydt

    Gaining knowledge of the past and present variations of the Atlantic meridional overturning circulation (AMOC) is crucial for the development of accurate future climate projections. The short range covered by direct AMOC observations, inconsistent paleoclimate records, and scattered hydrographic data are insufficient to realistically reconstruct the AMOC strength since 1900. An AMOC proxy index based on sea surface temperatures suggests that the AMOC has declined by 15% since the late 19th century but this index received extensive scientific criticism. Here, we use a deep learning algorithm and climate model simulations to accurately reconstruct the AMOC strength between 20° N and 60° N since 1900. In contrast with the existing indices, our reconstructions are well in agreement with AMOC strength variations simulated by climate models and direct observations at 26.5° N. Our novel set of AMOC reconstructions contribute to a larger confidence in 21st century AMOC decline projections from climate models.


  • Open Access

    Unravelling the role of increased model resolution on surface temperature fields using explainable AI

    by Simon Michel, Kristian Strommen, Hannah Christensen

    Reducing climate model biases is crucial for decreasing uncertainties in future climate projections. Despite recent efforts, improvements between the latest generations of Earth System Models (ESMs) have been modest, primarily due to the continued reliance on subgrid-scale parametrizations. These parametrizations are necessary because the model resolutions in CMIP6 are too coarse to explicitly simulate too small-scale processes such as ocean mesoscale eddies and deep atmospheric convection, which significantly influence regional and global climate patterns. Recent advances in computational power have enabled higher-resolution models, allowing for some of these processes to be simulated explicitly, reducing the need for parametrization. Here, we combine a convolutional neural network (CNN) classifier and explainable AI (XAI) to investigate the role of increased resolution in simulating winter surface temperature fields. The CNN is used to classify ESMs with varying resolutions based on snapshots of their surface temperature fields, while the XAI approach explains which regions and features the CNN relies on to make these distinctions, providing deeper insights into ESM performance. Results indicate that models with similar ocean grids are more frequently confused by the CNN than those from similar modeling centers, emphasizing the crucial role of ocean resolution, particularly the presence of mesoscale eddies, in shaping climate simulations. Although the analysis is restricted to surface air temperature, the XAI approach offers a more nuanced understanding of model differences compared to traditional bias analyses. This methodology can be extended to other climate variables and ESM features, offering a powerful tool for enhancing model intercomparison and evaluating ESM performance.

  • Open Access

    A past and present perspective on the European summer vapor pressure deficit

    by Viorica Nagavciuc, Simon L. L. Michel, Daniel F. Balting, Gerhard Helle, Mandy Freund, Gerhard H. Schleser, David N. Steger, Gerrit Lohmann, Monica Ionita

    The response of evapotranspiration to anthropogenic warming is of critical importance for the water and carbon cycle. Contradictory conclusions about evapotranspiration changes are caused primarily by their brevity in time and sparsity in space, as well as the strong influence of internal variability. Here, we present the first gridded reconstruction of the summer (June, July, and August) vapor pressure deficit (VPD) for the past 4 centuries at the European level. This gridded reconstruction is based on 26 European tree ring oxygen isotope records and is obtained using a random forest approach. According to validation scores obtained with the Nash–Sutcliffe model efficiency, our reconstruction is robust over large parts of Europe since 1600, in particular for the westernmost and northernmost regions, where most tree ring records are located. Based on our reconstruction, we show that from the mid-1700s a trend towards higher summer VPD occurred in central Europe and the Mediterranean region that is related to a simultaneous increase in temperature and decrease in precipitation. This increasing summer VPD trend continues throughout the observational period and in recent times. Moreover, our summer VPD reconstruction helps to visualize the local and regional impacts of the current climate change, as well as to minimize statistical uncertainties of historical VPD variability. This paper provides also new insights into the relationship between summer VPD and large-scale atmospheric circulation, and we show that summer VPD has two preferred modes of variability, namely a NW–SE dipole-like mode and a N–S dipole-like mode. Furthermore, the interdisciplinary use of the data should be emphasized, as summer VPD is a crucial parameter for many climatological feedback processes in the Earth's surface system. The reconstructed summer VPD gridded data over the last 400 years are available at the following link: https://doi.org/10.5281/zenodo.5958836 (Balting et al., 2022).


  • Open Access

    Deep learning reconstruction of Atlantic Meridional Overturning Circulation strength validates ongoig twenty-first century decline.

    by Simon Michel, Henk Dijkstra, Francesco Guardamagna, Valérian Jacques-Dumas, René van Westen, Anna von der Heydt

    The response of the Atlantic Meridional Overturning Circulation (AMOC) – an important ocean system transporting heat and salt across the Atlantic Ocean1,2 – to anthropogenic greenhouse gas emissions is crucial in projections of future climate change3. For Earth System Model (ESM) validation, the short range covered by AMOC observations4-8, inconsistent paleoclimate records9,10, and scattered hydrographic data11 are insufficient. A well-established sea surface temperature (SST) proxy index of the AMOC strength suggests that the AMOC has decreased substantially since the late ninetieth century12 but has received extensive scientific criticism13-21. Here, we use a deep learning method to provide a novel SST-based reconstruction of the twentieth century AMOC strength and show that it has declined by -0.151 Sv/decade since 1900, with the largest contribution in recent decades. In contrast to the earlier SST-based AMOC reconstruction12, the new one is in good agreement with direct 21st century AMOC measurements4 and historical ESM results22. In this way, the new AMOC reconstruction allows for new avenues for ESM calibration3,19 and contributes to a larger confidence in 21st century AMOC decline projections23.