Predictability of weather and climate

Abstract:

Understanding how fast atmospheric variability shapes slow climate variability and sensitivity is a central challenge in Earth-system science. Recent advances in machine-learned (ML) atmospheric models have demonstrated remarkable skill on weather timescales, but their emergent behaviour in a fully coupled climate system is largely unexplored. We present results from a new hybrid modelling framework that couples a machine-learned atmosphere to a dynamical ocean model. We report on a set of 70-year coupled simulations (1950–2020 historical forcing and fixed-1950s control) in which the ACE2 ML climate emulator is interactively coupled to the NEMO ocean model. These experiments represent, to our knowledge, the first multi-decadal integrations of a machine-learned atmosphere interacting with a full-depth dynamical ocean. We assess the behaviour of the coupled system, with particular focus on low-frequency tropical variability and the climate response to greenhouse-gas forcing. Preliminary results indicate realistic emergent El Nino-like variability and a physically plausible climate sensitivity, suggesting that key atmosphere–ocean feedbacks can be captured within a hybrid ML–dynamical framework. These results evaluate the possible role of entirely machine-learned components in next-generation Earth-system models.

Abstract:

Nudging still is a cost-effective data assimilation technique in coupled climate models, but conventional schemes apply fixed spatial strengths and are less effective in representing heterogeneous ocean processes. An adaptive nudging framework based on a spatially varying gain matrix is proposed to dynamically balance model and observational errors. The method not only preserves the merits of the latitude-dependent nudging approach but also provides a more physically consistent determination of the spatial distribution of nudging coefficients. Implemented in the SPEEDY-NEMO coupled model, the framework is systematically evaluated against the traditional latitude-dependent scheme. Results show that the adaptive approach substantially improves subsurface temperature assimilation, particularly in the Niño3.4 region, the tropical Indian Ocean, North Pacific, North Atlantic, and the northeastern Pacific. In the tropics, the improvement is mainly achieved above and within the thermocline (roughly 100--200 m), where strong vertical stratification and sharp gradients make fixed nudging strengths inadequate:the RMSE decreases by 20% and the correlation with observations increases by 30% compared with the traditional latitude-dependent scheme. By dynamically adjusting the assimilation strength, the adaptive scheme better constrains the thermocline variability and surface-subsurface interactions. In mid- to high-latitude regions, the improvement extends to greater depths, consistent with a deeper thermocline, where oceanic processes dominated by the mixed layer dynamics and convection exhibit large regional biases that require spatially adaptive correction. In addition, compared with the latitude-dependent nudging scheme, the adaptive approach achieves simultaneous corrections of both the systematic bias term and the variance term of temperature deviations, thereby enhancing not only the mean state but also the model’s ability to capture variability. Generally, the root-mean-square errors decrease by 20-30% and the correlation with observations increases around 30-50% by the adaptive scheme. Beyond temperature, improvements are also evident in salinity, currents, and sea surface height anomalies, indicating the broader benefits of the adaptive scheme. These results indicate that spatially adaptive nudging provides a more effective and practical alternative to fixed schemes, offering a solid basis for improving ocean state estimation in coupled models.

Abstract:

As global temperatures rise, extreme weather events are increasingly causing damages across human health, infrastructure, agriculture, and the broader economy. The science of event attribution is evolving to include estimates of economic damages attributable to climate change in addition to physical impacts. A key challenge in this field is to create physically consistent and high-resolution counterfactuals which can be used to estimate to attributable losses.Here, we analyse the precipitation-driven impacts of Storm Irene in August 2011 when it was undergoing extratropical transition in the Northeastern United States. Across the Northeast United States, this storm caused rainfall of up to 180 mm within a few hours, leading to fluvial and pluvial flooding with catastrophic consequences that caused  more than $1.3 billion in property damages in the state of Vermont alone.Our method enables linking economic damages attributable to climate change to meteorological drivers through a direct modelling chain by combining an operational weather forecasting model, hydrodynamic model, and economic damage model.This research underscores the potential of interdisciplinary attribution methodologies to inform climate risk assessments in insurance and provide an evidentiary basis for climate-related liability.

Abstract:

Since 2004, many methods for event attribution have been developed. Early studies showed that attribution statements are sensitive to the framing of research questions but few large comparisons have been undertaken.Here, we firstly motivate the need for multi-method extreme event attribution, highlighting conceptual differences between methods. In a second part, we present a case study of midlatitude storm Babet (2023) to compare three common storyline attribution methods, alongside a severity-based probabilistic method. We discuss three widely relevant questions which highlight the complementarity and the differences between methods: (1) How has climate change impacted the frequency of the event? (2) How has climate change impacted the event severity? (3) Were the dynamics of the event influenced by climate change and if yes, how?We show that methods differ in the extent to which they reproduce observed weather patterns. This influences attribution statements, and can even change the sign of results for events with uncertain climate signals. We argue that limitations and strengths of methods need to be clearly communicated when presenting event attribution reports to ensure findings can be used reliably by a wide range of stakeholders.

Abstract:

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.