Papers for

climate model developers

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Mesoscale ocean currents linked to errors in El Niño forecasts

Tracing Uncertainties of Extreme El Niño Prediction to Mesoscale Ocean Currents

Abstract: The predictability of the El Niño-Southern Oscillation (ENSO) has historically been linked to large-scale errors in models' initial conditions such as heat content and surface wind stress. Here we demonstrate that these slowly evolving errors may ultimately originate from mesoscale errors in initial oceanic states that are common in current models and ocean observation systems. This hidden source of ENSO forecast uncertainty, in terms of mesoscale ocean-current errors, is convincingly exposed by the conditional nonlinear optimal perturbation (CNOP) method applied to a differentiable ocean AI model. The CNOP-identified most influential current errors are strongly event-dependent. For the 1997/98 El Niño, the initial current errors in the eastern Pacific results in the largest forecast error through locally suppressing original subsurface warming. For the 2015/16 event, the dominate error originates from the western Pacific, which reorganizes upper ocean temperature structure and affects the eastern Pacific through the equatorial waveguide. Further analysis reveals that these current errors perturb mesoscale eddy count and eddy kinetic energy triggering large-scale temperature errors that hinder El Niño development. The timing of the maximum growth rate of the obtained CNOP error is well aligned with the spring predictability barrier of ENSO, and the error structures are reflected in simulated analysis errors caused by unresolved mesoscale currents. These findings extend the potential source of ENSO predictability beyond large-scale thermal signals, reveal for the first time an uncharted route of ENSO forecast failure, and provide an important physical basis for targeted ocean observations and data assimilation in the future.

Fri 13 NovArtificial Intelligence
The gist
El Niño is a climate event that affects weather worldwide, but predicting it accurately is difficult. This paper shows that small-scale ocean current errors, which often go unnoticed, can cause big mistakes when forecasting El Niño. The authors used a special AI-based model to find out where these errors start and how they develop differently for major events. Knowing this helps improve ocean observations and forecasting models in the future.
Open → 2611.17854v1

Physxai detects weather model outputs that break physical laws

PhysXAI: A Theoretical Framework for Diagnosing Physical Hallucinations in Weather Foundation Models

Abstract: Modern machine learning models for global weather forecasting achieve remarkable predictive accuracy, yet they frequently produce states that violate fundamental physical laws, including conservation of mass, hydrostatic balance, or spectral energy scaling. We term these outputs physical hallucinations and propose PhysXAI, a rigorous diagnostic framework that detects, localises, and attributes such inconsistencies. Unlike existing post-hoc metric suites, PhysXAI is a learnable system that explicitly decouples forecast error from physical plausibility. We formalise the problem through a novel joint condition on numerical and physical residuals, and introduce a multi-branch encoder that extracts representations from conservation, dynamical, spectral, cross-variable, and spatiotemporal residuals. A key innovation is a physics-consistency contrastive learning objective, which we theoretically show the latent representation to be approximately isometric to the physical residual space, ensuring that diagnostic decisions remain grounded in physically meaningful deviations. We further derive an adaptive multi-objective loss with gradient-based weighting to balance competing diagnostic objectives. The framework produces spatially resolved hallucination maps, violation-type classifications, and calibrated Bayesian uncertainty, enabling both detection and interpretable attribution of physical inconsistencies. We conduct a comprehensive comparative evaluation of PhysXAI across ERA5, WeatherBench 2, and MERRA-2, using forecasts from GraphCast, Pangu-Weather, and NeuralGCM, with Python-written code and a fully reproducible experimental protocol. Our theoretical analysis and empirical evaluation establish PhysXAI as a principled framework for diagnosing physical hallucinations and advancing trustworthy AI for weather forecasting.

Wed 11 NovMachine Learning
The gist
Weather forecasting AI models can make impressively accurate predictions, but sometimes their forecasts contain impossible results that break basic physics rules like mass conservation. The authors call these mistakes physical hallucinations. They created PhysXAI, a system that spots these errors by comparing forecast outputs against physical laws and highlighting where and why the model went wrong. This tool helps make weather AI forecasts more trustworthy by showing exactly what physics rules were broken and how certain the system is about its diagnosis.
Open → 2611.15053v1

Performance portability challenges with climate model microphysics kernels

Performance Portability Explained: Insights from accelerating ICON microphysics

Abstract: Scientific computing is under increasing pressure to adapt to HPC hardware that is shaped by AI workloads. In these modern ecosystems, extracting reasonable performance without investing major engineering effort is a central issue for domain scientists. Correctness constraints and intricate reference implementations represent further impediments to development. Instead of using a double-precision implementation as a reference, considering the underlying scientific problem directly opens more opportunities for performance improvements when updating and rewriting applications. We take seven different portability frameworks---covering pragma-based, DSL, and data-centric approaches---and evaluate them on a microphysics kernel of the ICON climate model. Our performance benchmarks show that each different HPC system benefits from a different combination of optimizations, so we develop the measure of framework expressivity to gauge how easily each framework can customize portable implementations in this way.

Tue 10 NovDistributed, Parallel, and Cluster ComputingPerformanceProgramming Languages
The gist
Scientific computing programs need to run fast on new powerful computers designed for artificial intelligence, but this can be hard without lots of engineering work. The authors studied different programming tools that help make code run efficiently across different machines by focusing on how the scientific problem itself is solved, not just copying old precise code. They tested seven programming approaches on a part of a climate model and found that no single approach works best everywhere. They also created a way to measure how easy it is to adapt these tools for specific hardware, helping developers choose the best method to get good performance.
Open → 2611.14204v1

Energy guided transport improves stable physics informed flow matching

Energy-Guided Transport for Projection-Free Physics-Informed Flow Matching

Abstract: Deep generative models, particularly Flow Matching frameworks, have emerged as powerful tools for synthesizing complex physical systems. However, standard architectures remain purely statistically driven and agnostic to the underlying governing equations, often violating fundamental physical constraints during inference. Existing zero-shot approaches designed to enforce these constraints typically rely on rigid algebraic projections. While effective for sparse boundary conditions, applying such discrete geometric corrections to dense, full-domain PDEs may disrupt the probability path, \rev{potentially deviating from the learned manifold and introducing substantial computational overhead}. To address these challenges, we introduce Energy-Guided Transport, a novel training-free framework that formulates non-linear physical constraints as a smooth potential energy landscape. Rather than interrupting the ODE integration, our method applies a continuous dissipative force directly to the intermediate generative states, steering them toward the physical manifold. By evaluating this force implicitly via reverse-mode automatic differentiation, our framework preserves the natural dynamics of the flow while maintaining an efficient $\mathcal{O}(N)$ algorithmic complexity. We extensively evaluate our framework across various 1D and 2D physical systems, including the chaotic Navier-Stokes equations. Our method achieves stable generation without structural divergence, significantly improves the stability-accuracy trade-off compared to state-of-the-art methods, and substantially accelerates inference with respect to exact projection strategies. Code available at: https://github.com/ispamm/Projection-free-energy-guided-flow-matching.

Wed 4 NovMachine Learning
The gist
Generating correct simulations of physical systems is hard because models often ignore the underlying physics and produce impossible results. The authors introduce a new method that gently nudges the simulation process toward following physical laws without heavy corrections. This approach helps keep simulations stable and more accurate, even for complex flows like chaotic fluid dynamics, and runs faster than previous methods. Their work shows that it is possible to respect physics in generative models while maintaining efficiency.
Open → 2611.06711v1

Physics-informed models improve predictions in multiscale system simulations

Physics-informed traditional and generative Mori-Zwanzig-reduced models for multiscale systems

Abstract: Model reduction via the Mori-Zwanzig (MZ) formalism is well understood for coarse-grained representations constructed from standard basis functions such as Fourier functions. We show that the same formalism can be used to build reduced-order models in bases discovered by physics-informed machine learning. The fluctuation-dissipation theorem (FDT) is used to close the resulting non-Markovian evolution equations for the expansion coefficients of the learned basis. We further introduce a complementary, generative closure model. Relative to the Markovian dynamics alone, both the MZ and generative closures substantially improve predictions of the filtered states.

Tue 3 NovMachine Learning
The gist
Simulating complex systems with many scales is difficult because capturing all details is expensive. The authors show that a mathematical framework called Mori-Zwanzig, traditionally used with simple function bases, can also work with new bases found by machine learning guided by physics. They use physical principles to close the equations so predictions of how systems evolve become more accurate. They also introduce a new generative approach that further improves how these simulations predict behavior over time.
Open → 2611.05211v1

AI coding agent assists porting legacy climate model to modern GPUs

Migrating a Legacy Scientific Application with an AI Coding Agent: NICAM from Fortran to Kokkos

Abstract: Migrating from Fortran to C++/Kokkos broadens access to hardware accelerators through backends such as CUDA and HIP while offering greater flexibility in software design. However, legacy Fortran applications often have large codebases, making manual porting time-consuming and costly. Recent advances in large language models (LLMs) and AI coding agents have made large-scale code migration increasingly feasible. In this study, we ported NICAM (Nonhydrostatic ICosahedral Atmospheric Model), a legacy Fortran climate model accelerated with OpenACC, to C++/Kokkos with the assistance of Claude Code. The agent translated the code directly from Fortran to C++/Kokkos under human review. Approximately 140,000 lines of Fortran were analyzed and ported into 35,000 lines of C++/Kokkos, covering the dynamical core and seven physics packages. A 100-member ensemble comparison with the original Fortran model showed that the port is statistically consistent with the original. Performance was evaluated on Miyabi-G, an NVIDIA GH200 cluster, and Sirius, an AMD MI300A cluster. Compared with the Fortran/OpenACC baseline on GH200, the port achieved application speedups of $1.71\times$ on GH200 and $1.54\times$ on MI300A. The CUDA build achieved a GPU kernel-time speedup of $1.38\times$ on GH200. After normalization by peak memory bandwidth, the HIP build on MI300A achieved $0.96\times$ the kernel performance of the Fortran/OpenACC baseline.

Fri 30 OctDistributed, Parallel, and Cluster Computing
The gist
Older climate modeling software is often written in Fortran, which can limit how well it runs on new, fast computer hardware like GPUs. The authors used an AI coding assistant to help translate a large climate model from Fortran into a newer programming style called C++/Kokkos, which runs better on GPUs. They checked that the results matched the original and found the new version ran up to 1.7 times faster on some GPUs. This shows AI tools can help update scientific codes to use newer hardware more easily.
Open → 2610.41313v1

PDE foundation models show growing prediction errors over long simulations

Beyond Local Accuracy: Benchmarking the Rollout Reliability of PDE Foundation Models

Abstract: PDE foundation models enable data-efficient adaptation but are deployed autoregressively, so short-horizon accuracy need not imply reliable simulation. We ask whether the short-horizon advantage of a foundation-model recipe persists when its predictions are repeatedly fed back as inputs for a long horizon. We define error amplification as G=E(H_f)/E(H_s), which measures full-horizon error relative to short-horizon error. We evaluate seven models on seven two-dimensional time-dependent PDE datasets: scratch-trained FNO and DeepONet; fine-tuned DPOT, GPhyT, Poseidon, and Walrus; and frozen in-context VICON. We report VRMSE, relative L_2, and relative MAE. Six datasets provide paired short- and full-horizon scores, while CE-RM contributes short-rollout sweeps only. At least one fine-tuned foundation model outperforms both scratch baselines in 26 of 30 short-rollout dataset--budget cells under every metric. Nevertheless, full-horizon VRMSE exceeds short-horizon VRMSE in 140 of 144 paired trained configurations, with a median error amplification of $2.04$. The count below VRMSE one falls from 104 to 53, and 12 full rollouts exceed ten although none does at the short horizon. Walrus achieves the lowest full-horizon VRMSE in 16 of 24 cells, yet its median error amplification is $2.70$. Relative L_2 and relative MAE show the same qualitative pattern. These results show that favorable short-horizon rankings can coexist with substantial error amplification. PDE foundation-model benchmarks should therefore report horizon-resolved absolute errors, amplification statistics, and failure incidence rather than a single local forecast score.

Thu 22 OctMachine Learning
The gist
Predicting the future behavior of complex systems using computer models often looks accurate for a short time, but errors can grow when those predictions are used repeatedly to predict further ahead. The authors study different models used for partial differential equations, which describe many physical processes, to see how prediction errors multiply over longer times. They find that models that do well in short tests often have twice as much error when running simulations longer. This means simply measuring short-term accuracy is not enough to tell whether a model will be reliable for long predictions.
Open → 2610.31935v1

Graphcast Mamba improves ten day weather forecasts with memory

GraphCast-Mamba: Learning Non-Markovian Corrections for Weather Forecasting

Abstract: Unresolved atmospheric processes can introduce history dependence beyond the short input windows used by many weather forecasting models. We introduce GraphCast--Mamba, which augments a frozen, pretrained $1^\circ$ GraphCast backbone with a trainable Mamba correction branch combining graph message passing and Mamba layers. The branch retains a temporal memory state across successive six-hour forecast steps, learning corrections informed by temporal context beyond GraphCast's two-frame input. Training is confined to the correction branch, using ERA5 data from 2015--2021. After 2,500 optimizer updates, including warm-up, the selected model achieves substantial improvements relative to the uncorrected GraphCast backbone. It reduces the GraphCast loss (normalized weighted mean-squared error; MSE) by 14.42% when aggregated over the full ten-day forecast and by 21.66% at day 10 (240 h after forecast initialization). At this final forecast lead time, root-mean-square error (RMSE) decreases by 7.09--14.60% across all 11 evaluated variable groups, spanning temperature, wind, pressure, geopotential, specific humidity, and precipitation. These results demonstrate broad improvements in ten-day weather forecasts through a short correction-training schedule while retaining the pretrained backbone.

Tue 20 OctMachine Learning
The gist
Weather forecasts often struggle because they use only recent data and miss longer-term effects. The paper presents GraphCast-Mamba, a system that adds a memory-based correction to an existing forecast model, helping it remember past information better. This addition improves forecast accuracy, especially for longer ten-day predictions across many weather variables. The correction is trained on historical weather data but keeps the original model unchanged, making it an efficient improvement.
Open → 2610.28512v1

Quantum neural networks improve cloud cover predictions using physics knowledge

Physics-anchored residual quantum neural networks for cloud-cover parameterization

Abstract: Compact quantum neural networks have been explored for cloud-cover parameterization, but their predictive value depends on how physical knowledge is incorporated and which classical models provide the reference. We investigate an offline residual formulation that retains a fitted \XR\ cloud-cover estimate and learns a correction to its logit in transformed target space. Using released ICON/DYAMOND-derived data, we compare ten model variants in six- and eight-input settings, with three matched replicates and 10,000 training samples per replicate. All quantum circuits use local Z readout and noiseless exact expectation values. Under a common 30-epoch protocol, residual prediction reduced physical-space test mean squared error relative to direct QNN prediction by 54.3\% and 38.8\% for six-input XYZ and ZZXY circuits, and by 49.2\% and 38.9\% for their eight-input versions. Residual QNNs also outperformed the selected compact residual multilayer perceptron in every paired replicate. A larger, longer-trained Grundner-Q1 adaptation nevertheless achieved lower error: residual Q1 reached mean MSEs of 0.00778 and 0.00791, compared with 0.01124 and 0.01082 for residual XYZ. The results support residual prediction for compact simulated QNNs within the tested budget and establish a clear boundary set by stronger classical references. They do not establish computational quantum advantage or online climate-model skill.

Mon 19 OctArtificial Intelligence
The gist
Predicting cloud cover accurately is important for weather and climate studies. The authors tested quantum neural networks that use physics-based starting points and learn to correct errors in cloud predictions. They found that these residual quantum models make better predictions than direct quantum models or similar classical methods on their dataset. However, the quantum models still do not outperform the best classical models, and no speed or practical advantage is shown yet.
Open → 2610.26279v1

Neural operators learn best with fractional derivatives for faster pde solutions

Fractional is Better: Learnable Derivative Orders in Neural Operator Learning

Abstract: Neural operators learn mappings between function spaces, enabling fast surrogate solutions to partial differential equations. Despite remarkable architectural diversity, these methods often share a common input representation: raw coordinate-value pairs $(x, u(x))$. We ask whether inputs aligned with PDE differential structure can improve learning. Through Picard iteration on mild solutions, we show that derivatives of the input appear explicitly in the solution operator, suggesting that providing derivative features should reduce the network's implicit differentiation burden. We prove this intuition: providing derivative features improves approximation bounds from $O(W^{-(s-m)/d})$ to $O(W^{-s/d})$, where $W$ is parameter count, $s$ is input regularity, $m$ is the PDE order, and $d$ is spatial dimension. Our central finding, however, is a surprise: the optimal derivative order $β^*$ is strictly less than the PDE order $m$. This gap arises from a bias-variance tradeoff in spectral space that we characterize in closed form. Learning $β$ from data achieves automatic spectral regularization. We introduce $\partial$-NO (del-NO), for derivative-augmented neural operators, an architecture-agnostic augmentation that provides learnable fractional derivative features to any neural operator backbone. Across benchmark problems and architectures, $\partial$-NO consistently improves prediction accuracy, with learned orders that reflect a representation of known physics modulated by noise and finite-sample constraints.

Wed 14 OctMachine Learning
The gist
Solving complex equations called partial differential equations (PDEs) is usually slow. This paper shows that teaching neural networks to use not just raw data points but also fractional derivatives—like partial rates of change—helps the networks learn better and faster. Interestingly, the best amount of derivative information is often less than what the original equation suggests. The authors built a flexible method that adds these learnable derivative features to many neural networks, improving accuracy in tests.
Open → 2610.20363v1

OceanFlow improves probabilistic ocean forecasts with coherent modeling

OceanFlow: Coherent Flow Matching for Probabilistic Ocean Modeling

Abstract: Probabilistic ocean modeling is essential for risk-aware marine operations, climate analysis, and scientific decision-making, where downstream decisions depend on the conditional distribution of future ocean states rather than a single deterministic trajectory. Recent data-driven ocean models have made deterministic global ocean modeling increasingly accurate and efficient, but probabilistic ocean modeling remains underdeveloped. Specifically, perturbation-driven ensembles can create spread without explicitly optimizing the joint distribution of their members, often yielding poor calibration and physical inconsistency. To address these issues, we propose OceanFlow, a coherent flow matching framework for probabilistic global ocean modeling. OceanFlow first establishes a sharp, stable mean anchor for the conditional ensemble. Flow matching then learns a coherent state-conditioned pathwise law that jointly reshapes the residual distribution around this anchor at each transition, preserving the predictive center while refining state-dependent uncertainty. We further introduce a Factorized Dynamics Operator (FDO), which combines factorized transitions with a learned state-conditioned response to parameterize the mean anchor and residual-flow velocity. Extensive experiments show that OceanFlow improves ensemble-mean accuracy, probabilistic reliability, and physical-consistency diagnostics over competitive baselines, supporting coherent flow matching as an effective route toward data-driven probabilistic ocean modeling. Code is available at https://github.com/YuanGao-YG/OceanFlow.

Sun 11 OctMachine LearningArtificial Intelligence
The gist
Accurately predicting the future state of the ocean is important for safe marine operations and climate studies, but existing methods often give uncertain and inconsistent results. The authors introduced OceanFlow, a new technique that creates a reliable average prediction and then carefully models variations around it, maintaining physical realism. This approach leads to better accuracy and more trustworthy predictions about how the ocean might change. It helps planners understand risks by providing a fuller range of possible future ocean conditions, not just a single guess.
Open → 2610.15518v1

Using gradients to speed up inference in costly scientific simulators

Exploiting Gradients in Bayesian Inference of Expensive Simulators

Abstract: Simulators based on differential equations are ubiquitous in science and engineering. They are often used in simulation-based inference to evaluate the posterior distribution of the input parameters based on real-world observations of the simulator outputs. However, inference becomes challenging when individual simulator evaluations are computationally expensive. In such cases, a Bayesian optimization-based active learning approach with Gaussian process surrogate models has been used to maximize the information obtained from a limited simulation budget. Recently, gradients of simulator outputs with respect to input parameters have become increasingly available, yet they are rarely exploited for inference. Even though we only need to learn the simulator input-output relationship, gradient information can provide an additional valuable signal to guide the active learning procedure. This is of particular interest in the case of expensive simulators, when sample efficiency is crucial. In this paper, we demonstrate how incorporating gradient information into the Gaussian process surrogate accelerates Bayesian optimization-based inference under a limited simulation budget. Our results show significant improvement in convergence speed from using gradient information. For reverse-mode differentiation, the inference efficiency gains are maintained when accounting for the additional computational cost. In contrast, for forward-mode differentiation, the inference speed-up does not outweigh the computational costs. These results indicate that gradient-enhanced surrogates are beneficial primarily in problems where the number of parameters exceeds the output dimensionality, where reverse-mode differentiation is efficient.

Thu 8 OctMachine Learning
The gist
Simulators help study complex systems but can be slow to run, making it hard to figure out the best input settings. The paper shows how using extra information about how the simulator’s outputs change with inputs (called gradients) can make this process faster. They combine gradients with a method that predicts outputs to choose smarter inputs for testing. This approach works best when the problem has many inputs but fewer outputs, especially when using a technique called reverse-mode differentiation.
Open → 2610.12076v1

Deep learning advances weather forecasts and aids climate predictions

Artificial intelligence pathways from weather to climate

Abstract: Deep learning has made rapid advances in weather forecasting: autoregressive models trained on atmospheric reanalyses now rival dynamical models across nowcasting, medium-range, and subseasonal-to-seasonal lead times, producing well-calibrated ensemble forecasts at reduced cost. We review these advances and consider their extension to climate horizons, where the challenge shifts from initial-condition skill to producing reliable statistical responses under altered forcings. AI-powered climate prediction systems must produce credible forced responses to drivers (e.g., greenhouse gases, land-use change) typically outside the observed record. We propose two minimum requirements for AI in climate modeling: (i) external forcing agents must enter explicitly enough to support interventions in which they vary independently; and (ii) robustness must be stress-tested in out-of-distribution regimes, including extremes and counterfactual trajectories. Using leading AI autoregressive emulators and hybrid physics-AI models, we identify development and coupling challenges. Comparing the reported throughput of these models with that of GPU-ported dynamical models highlights how AI can reduce time-to-solution by advancing only the target variables at the required resolution and using longer time steps, rather than integrating a full high-frequency, multivariate state. Diverse AI downscaling strategies can partially substitute for explicit fine-scale resolution, paving the way toward inexpensive local hazard assessment across prediction horizons.

Wed 7 OctArtificial IntelligenceMachine Learning
The gist
Forecasting the weather has improved a lot thanks to deep learning, which is a type of artificial intelligence (AI). This paper looks at how these AI methods can be stretched to predict climate changes, which is harder because it requires understanding how the climate reacts to things like greenhouse gases over a long time. The authors suggest key ways to build climate AI models that can handle new, extreme situations and different environmental factors clearly. They also explain how AI can speed up climate models by focusing on important features and running faster than traditional physics-based models while still giving detailed local predictions.
Open → 2610.09770v1

Climatebench v2 enables fair evaluation of climate model predictions

ClimateBench v2.0: Probabilistic Climate Model Benchmarking

Abstract: We present ClimateBench v2, a standardized protocol for evaluating climate models on diagnostics expected to be informative for their skill in projecting mid-century regional temperature and precipitation changes. The protocol is designed to evaluate any physics-based, data-driven, or hybrid climate model on equal footing using a common set of observational and out-of-distribution tests. We define three tiers of evaluation. Tier I establishes physical credibility through entry-ticket tests of energy conservation, coupled (co-)variability, and basic forced responses. Tier II scores models against post-2015 observations of surface temperature, precipitation, radiative fluxes, sea ice, and key modes of variability using fair CRPS as the primary probabilistic score, complemented by distributional and ensemble-consistency diagnostics. Tier III tests out-of-distribution generalization through paleoclimate simulations spanning the Last Interglacial, Last Glacial Maximum, and Mid-Holocene, and through perfect-model experiments in which data-driven models must predict the future climate of existing Earth system models from historical data alone. We reserve all observational data after 2015 for testing, and submissions must include multiple ensemble members to enable probabilistic evaluation. This reservation exploits a new opportunity provided by the decade of observations accumulated since the end of the CMIP6 historical experiment, which constitutes an out-of-sample record of forced climate change (and internal variability) for the current generation of models, and we quantify, in an idealized setting, the information it carries about mid-century warming. We provide the evaluation code, observational reference datasets, and perfect-model training data as an open benchmark to drive measurable progress in climate projection across all modeling approaches.

Sat 3 OctMachine Learning
The gist
Predicting how Earth’s climate will change is important but challenging, especially when comparing different types of models. The authors present ClimateBench v2, a set of tests designed to check how well various climate models predict temperature and rainfall changes. These tests include checking basic physical rules, matching recent observations, and seeing how models perform on past climates and unknown scenarios. This helps ensure models are accurate and reliable for future climate forecasting.
Open → 2610.04558v1

Soft mixtures improve long-horizon forecasting of physical processes

COSMOS: Soft Mechanism Mixtures with Verifiable Routing for Long-Horizon PDE Forecasting

Abstract: Neural operators offer an efficient alternative to classical PDE solvers, but most learn a monolithic map per equation family and discretization. Real systems are compositional: transport, diffusion, wave, and reaction processes can act simultaneously. Existing mixture-of-experts operators typically use sparse top-$K$ routing, although concurrent physics is naturally a blend rather than a discrete choice. We propose COSMOS (Cooperative Operator Specialists with Mechanism-level Operator Soft-routing), a soft mechanism-mixture neural operator. Four process-biased specialists remain active at every step and are continuously mixed by a learned gate, with their features fused by a small network. Specialists share a coarse latent grid, while a zero-initialized full-resolution residual restores detail lost through the bottleneck. We also introduce an operator-splitting compositional benchmark with known mixture weights $w^\star$ per trajectory. Against family-tuned FNO under 20-step rollouts, an initial 3-seed evaluation suggested gains on diffusion--reaction, parity on Navier--Stokes, and weaker shallow-water performance. An 11-seed audit showed that the diffusion--reaction gain was unstable, motivating caution in small-seed rollout comparisons and precluding a reliable accuracy-win claim on these families. Ablations show that uniform routing or removing the specialist mixture substantially degrades stable-regime rollouts. On the labeled benchmark, dense soft routing yields $2.2\times$ lower error than hard top-1 routing at identical fusion; using generator weights $w^\star$ at inference further lowers rollout error to $0.034$, diagnosing limitations of the learned gate. However, routing labels do not align with the specialists' intended mechanisms: COSMOS supports compositional accuracy, not mechanism identity.

Sat 3 OctMachine LearningArtificial Intelligence
The gist
Simulating physical processes like heat spreading, waves, or fluid flow often involves solving complex equations. The authors propose a new method called COSMOS that mixes multiple specialized neural modules softly instead of choosing a single one, better reflecting how real-world processes blend together. This approach improves prediction stability and accuracy over time for certain physical systems, though the improvements vary by problem. The method also includes tests to verify how well the model understands the underlying physics, revealing that while it predicts well, the internal experts do not always align exactly with specific physical mechanisms.
Open → 2610.04427v1

Kernel ridge regression models improve dynamical system predictions

Learning Closure of Dynamical Systems with Kernel Ridge Regression

Abstract: We develop a closure modeling framework for identifying missing components of dynamical systems using Kernel Ridge Regression (KRR). The framework addresses two classes of closure problems: difference-equation closures arising in ODE and PDE settings, and algebraic closures arising from moment closure in kinetic equations. For the first class, we derive an error bound in an ODE setting that quantifies contributions from time integration, approximation of unresolved scales, and interpolation required to couple unresolved-scale effects to the resolved solver. Numerical experiments on the Lorenz-63 system and the Kuramoto-Sivashinsky equation demonstrate accurate long-horizon predictions and substantial improvements over an LSTM-based closure model. For the second class, we consider moment closure for a one-dimensional kinetic equation by modeling discrepancies between kinetic and macroscopic fluxes as a function of the resolved macroscopic variables. We compare global KRR models based on PCA coordinates with spatially local models. While the global model performs well for unimodal initial conditions, its accuracy deteriorates for bimodal initial conditions. Spatially local models with appropriate modeling inputs improve robustness and achieve higher predictive accuracy.

Thu 1 OctMachine Learning
The gist
Predicting the future behavior of complex systems can be tricky when some parts of the system are unknown or too small to track. The authors use a mathematical tool called Kernel Ridge Regression to learn and fix these missing parts. They show their method works well on classic systems that scientists study, and it does better than some other machine learning methods. This helps make longer and more accurate predictions for both general and specialized types of systems.
Open → 2610.02564v1

AI models emulate sudden shifts in polar vortex weather patterns

AI Emulation of Stochastic Sudden Stratospheric Warming with Interpretable Latent Structure

Abstract: Rare weather regime transitions pose a challenge for data-driven modeling due to class imbalance. In this study, we develop a probabilistic deep learning emulator for a prototypical system with regime transitions, the stochastic Holton--Mass model of stratospheric variability, and analyze the structure of its learned latent space. The Holton--Mass model exhibits two metastable regimes, a strong and a weak polar vortex, maintained by nonlinear wave--mean flow interactions, with weak stochastic forcing intermittently triggering rare transitions between these regimes that qualitatively represent SSW events. We employ a ResNet-inspired Conditional Variational Autoencoder with six-layer encoder and decoder layers and explicit current-state conditioning to model the distribution of the system's state at the next time step (one day). The emulator accurately reproduces short-term dynamics, steady-state probability distributions, regime persistence statistics, rare transition rates, the transition committor function, and the transition expected lead time of the physical model. Beyond emulation fidelity, we interrogate the learned latent representation to understand how the model internalizes the underlying metastable structure of the dynamics. Principal Component Analysis of the 32-dimensional latent space reveals a clear and unsupervised separation into four physically interpretable clusters corresponding to strong versus weak vortex regimes and stable versus transition-prone configurations. Such emergent regime separation in latent space is hard to identify for deep generative models applied to high-dimensional stochastic systems. Our results show that carefully designed probabilistic emulators can uncover physically meaningful manifolds governing extreme-event dynamics, potentially aiding the development of improved operational advanced warning systems.

Thu 1 OctMachine Learning
The gist
Sudden Stratospheric Warming (SSW) events are rare shifts in the Earth's polar vortex that impact weather but are hard to predict due to their rarity. The authors created an AI model that can imitate these sudden weather changes in a simplified system, learning to capture both common and rare weather states. Their AI not only predicts short-term weather patterns but also organizes its internal understanding into clear groups that reflect underlying physical weather conditions. This approach could help improve forecasting systems by better spotting when rare events are likely to happen.
Open → 2610.02069v1

Proper scoring rules improve weather forecasts under uncertainty

Proper Scoring Rule-based Diffusion for Probabilistic Weather Forecasting

Abstract: Recent probabilistic weather forecasters train stochastic predictors with the continuous ranked probability score (CRPS) to generate each ensemble member in a single forward pass. These models learn the predictive distribution from the forecast context alone, which becomes difficult at longer forecast horizons where uncertainty is high. To learn the predictive distribution more effectively, we introduce auxiliary conditional denoising tasks that predict the same future state from the context and its corrupted version, which provides partial future information that can reduce prediction ambiguity. Building on distributional diffusion models, we learn the conditional distributions of these tasks with a single stochastic predictor by minimizing a proper scoring rule across noise levels. At inference, the predictor can still generate each ensemble member in a single forward pass at the fully corrupted endpoint. Standard CRPS training is recovered as the endpoint-only special case of our formulation, so our framework extends existing CRPS-based forecasters with only additional conditioning inputs. Controlled experiments show that the auxiliary tasks improve one-step forecasting across architectures, with larger gains at longer forecast horizons. The gains extend to high-dimensional global weather forecasting under both training from scratch and fine-tuning, along with improved calibration and potential benefits for generalization under distribution shift.

Tue 29 SeptMachine Learning
The gist
Making accurate weather predictions is hard, especially for times far in the future when there's a lot of uncertainty. The authors present a new way to help weather models learn by giving them extra clues about what the future might look like, even if those clues are a bit noisy. This helps the model know what to expect better and reduces confusion when making predictions. Their method builds on existing approaches and can generate multiple possible weather outcomes quickly, improving forecast accuracy and reliability.
Open → 2609.38632v1

New framework reveals hidden flaws in weather model dynamics

Evaluating Dynamical Fidelity through Predictive Structure in Physical Representations

Abstract: Machine-learning models for physical systems are currently evaluated primarily through errors between predicted and reference states and, increasingly, through tests of physical consistency. These metrics assess whether predictions are accurate and satisfy selected physical requirements, but provide limited insight into whether learned trajectories reproduce the underlying dynamics. Domain experts examine such relationships through physical representations that expose relevant processes, interactions, and responses, but these analyses are often separated from typical machine-learning evaluation. We introduce a practical framework for evaluating dynamical fidelity through predictive structure in physical representation spaces. Experts define the representations, while reference trajectories determine which relationships are predictive and retained as evaluation tests. We demonstrate the approach in atmospheric forecasting using ERA5 representations of planetary-wave activity and Northern Annular Mode evolution, and evaluate Pangu-Weather, GraphCast, and FengWu. The models exhibit distinct departures from reference predictive structure that are not reflected by conventional forecast errors. The framework thereby turns domain-expert representations into systematic tests of learned physical dynamics without prescribing the relationships in advance.

Mon 28 SeptMachine Learning
The gist
Weather prediction models are usually judged by how close their forecasts match real weather and if they obey known physical rules. But these checks don't always show whether the models truly capture how weather changes over time. The authors created a method that uses expert-chosen ways of representing weather processes to test if models follow real dynamic patterns. They tested this on three advanced weather models and found differences that normal error tests missed. This helps experts better understand where models fail and how to improve them.
Open → 2609.34627v1

Extreme event prediction improves with local instability sensing

Mechanism-Aware Ensemble Conditioning for Data-Limited Emulation of Extreme Events

Abstract: Extreme events in chaotic systems are difficult to learn from short trajectories because they are controlled by transient finite-time instability rather than by frequently observed bulk dynamics. We propose a mechanism-aware conditioning plug-in framework that turns a nudged coarse ensemble into a non-intrusive sensor of local instability geometry. In the small-noise regime, the ensemble covariance aggregates the same finite-time deformation kernels that govern local instability, providing a Jacobian-free proxy for the local amplification structure around a synchronized coarse trajectory. A small FiLM module injects statistics of this ensemble geometry into an otherwise unchanged backbone while leaving the coarse simulator unchanged. We demonstrate this interface in two distinct pipelines: a Transformer-style residual-attention corrector for a controlled low-dimensional chaotic system and a probabilistic recurrent STORN corrector for topographic two-layer quasi-geostrophic (QG) flow. In the low-dimensional benchmark, ensemble covariance directions co-activate with OTD modes and FiLM conditioning improves 99th-percentile exceedance-frequency errors over an identical no-context Transformer baseline. In QG, a fixed ensemble-conditioned FiLM-STORN model trained on only \(50\) time units substantially improves long-horizon rare-event statistics in the data-limited regime, including density-tail errors, exceedance frequencies, and spatial exceedance-area distributions relative to an unconditioned STORN trained on the same data; on averaged high-threshold exceedance diagnostics, it also outperforms the baseline STORN trained with $20$ times more high-resolution data. These results show that local instability geometry is not merely interpretable post hoc, but an actionable conditioning signal for data-efficient rare-event emulation.

Fri 25 SeptMachine Learning
The gist
Predicting rare, extreme events in chaotic systems is tough because these events depend on short-term instabilities that are not common in usual data. The authors developed a method that uses a group of slightly different simulations to detect these local instabilities without complex calculations. By feeding this information into machine learning models, the method improves the prediction of rare events, even with limited training data. They tested their approach on both simple chaotic systems and more complex fluid models, showing better accuracy compared to traditional methods.
Open → 2609.30746v1

Contrastive learning clarifies cloud model differences and observations

Understanding Perturbed Parameter Ensemble Sensitivities Using A Contrastive Learning Approach

Abstract: Perturbed parameter ensembles (PPEs) reveal how physics parameters affect climate simulations, but interpreting parameter sensitivities across multivariate, spatially structured outputs remains challenging, particularly when calibrating models against observations. We develop an explainable contrastive learning model that maps 5 monthly cloud and radiation fields into a shared representation space. We train the model on the fields of two 100-member Community Atmosphere Model version 6 (CAM6) PPEs, spanning 34 parameters, that only differ in the warm rain microphysics scheme: KK2000, the default bulk microphysics scheme, and TAU-ML, a neural network emulator of a bin microphysics scheme. The learned representations separates two PPEs with over 94\% linear classification accuracy while preserving the seasonal variability and ensemble spread due to parameter perturbations. In the shared representation space, the representations of satellite observations occupy the same low-dimensional manifold as the PPEs but are displaced from them most strongly during boreal spring and autumn. TAU-ML PPE has a lower distance to observations compared to KK2000 in the representation space. Integrated Gradients attributions highlights the contributions in subtropical low-cloud regions, Northern and Southern Hemisphere storm track regions, and tropical convection regions to differences between PPEs and observations. Regional attributions correlate most strongly with parameters associated with cloud microphysics, boundary layer turbulence, and deep convection. These results demonstrate that explainable representations of climate fields can attribute model differences to specific variables, regions, seasons, and physical parameters.

Thu 24 SeptArtificial Intelligence
The gist
Climate models have many settings that affect their weather predictions, but it is hard to understand how these settings influence complex weather features. The authors used a machine learning tool to create simple summaries of many climate patterns, which helps to compare two versions of a climate model and real satellite data. Their approach shows which regions, seasons, and weather variables differ between models and observations, linking these to specific model settings. This helps scientists see where models match reality and where they need improvement.
Open → 2609.30420v1

Hybrid AI models improve weather forecasts using learned memory variables

Learning Prognostic Variables for AI Convective Parameterizations via Symbolic Distillation

Abstract: Hybrid AI-physics climate modeling aims to improve coarse (~100km-resolution) Earth system models by learning to parameterize subgrid processes from high-fidelity data. However, this so far mostly involves local-in-time, diagnostic parameterizations, in which the subgrid state depends only on the current coarse state with no memory of previous states, which is unrealistic for processes such as convection that have intrinsic persistence. To address this, we enhance local-in-time parameterizations by learning prognostic variables that compactly carry important, additional past information where no explicit sub-grid information is available. First we compress past information into a low-dimensional latent space using an autoencoder, which then informs a neural network trained to parameterize targeted subgrid-scale processes. We then replace the autoencoder with symbolic equations that govern the time evolution of the latent variables, yielding additional prognostic memory variables that can be integrated alongside the resolved atmospheric state. We evaluate this approach on two systems: the Lorenz-96 model (online) and surface precipitation from high-resolution atmospheric simulations (offline). A forced multivariate linear ordinary differential equation recovers most of the added value achieved by the autoencoder-based approach in both experiments. Benchmarked against diagnostic parameterizations without memory, our memory-informed approach improves climate statistics and temporal structure, including a realistic diurnal cycle of tropical land precipitation.

Mon 21 SeptMachine Learning
The gist
Weather and climate models often miss important slower processes because they only look at the current state and ignore past information. This paper shows how to add a kind of memory to AI models that predict weather patterns by learning compact variables that capture past information. They use math equations to describe how these memory variables change over time, making forecasts more realistic. Their method improves predictions of rain patterns and the daily cycle of tropical weather compared to models without memory.
Open → 2609.24882v1

Machine learning weather models struggle with energy flow and error growth patterns

Butterfly Effect and the Kinetic Energy Cascade in Probabilistic Machine Learning Weather Prediction Models

Abstract: This study analyses kinetic energy (KE) spectra, difference kinetic energy (DKE) spectra, and signatures of KE transfer across spatial scales in four state-of-the-art probabilistic machine learning weather prediction (MLWP) models: NeuralGCM-ENS, FourCastNet 3, AIFS-ENS, and GenCast. Results are compared with those from the physics-based numerical weather prediction model IFS-ENS. While NeuralGCM-ENS successfully reproduces the expected upscale transfer of KE, noise injection at its encoder stage underestimates mesoscale KE. Conversely, AIFS-ENS, GenCast, and FourCastNet 3 produce realistic KE spectral magnitudes but do not capture the expected upscale transfer of KE. In particular, AIFS-ENS and GenCast, which employ spatially uncorrelated stochastic perturbations, exhibit enhanced accumulation of KE at high wavenumbers. All examined models exhibit upscale error growth, reflected by the progressive shift of the DKE spectral peak toward larger wavelengths over time. However, the MLWP models struggle to reproduce the rapid initial growth of ensemble spread at small spatial scales associated with the butterfly effect. The results show that MLWP models can misrepresent the known scale transfer of kinetic energy despite producing skilful weather forecasts.

Wed 16 SeptMachine Learning
The gist
The study looks at how well several advanced machine learning weather models handle the flow of energy at different scales in the atmosphere and how errors spread out over time. The authors find that while some models mimic certain energy patterns well, others do not capture important ways energy moves between scales. All models have trouble showing the quick early spread of small errors known as the butterfly effect. This means that even though the machine learning models can predict weather well, they don't fully represent how energy and errors behave in the real atmosphere.
Open → 2609.18489v1

Physics-informed model predicts Mars nightside atmospheric gases reliably

Physics-Informed Multi-Task Surrogate Model for the Martian Nightside Thermosphere

Abstract: Modeling the Martian nightside thermosphere remains challenging due to sparse in situ sampling and strong coupling among transport, magnetic, and seasonal processes. Purely data-driven models can produce non-physical artifacts, such as density inversions, in poorly sampled altitude regimes. We present a multi-task physics-informed neural network that simultaneously predicts the base-10 logarithmic densities of four neutral species (O, CO$_2$, N$_2$, and Ar) using more than a decade of MAVEN/NGIMS observations (MY 32-38, 2014-2025). A shared backbone learns a common representation of the nightside thermospheric state and branches into species-specific output heads. A weak monotonicity prior is incorporated via automatic differentiation by penalizing positive vertical gradients in logarithmic density. Experiments using an orbit-disjoint train/validation/test split show that physics-informed regularization substantially reduces non-physical inversions while preserving predictive skill and slightly improving it in the best-performing configuration, as measured by RMSE, MAE, and $R^2$. The resulting model provides a computationally efficient surrogate for nightside thermospheric reconstruction with improved vertical consistency.

Wed 9 SeptMachine Learning
The gist
Understanding the atmosphere on Mars's night side is hard because there aren’t many direct measurements and many processes are linked. Purely data-based methods sometimes make unrealistic predictions like densities that don’t decrease with altitude. The authors created a neural network that predicts the amounts of four gases at once, using over ten years of spacecraft data. They included physics rules to stop impossible results and found it improved the model’s realism without losing accuracy. This model can quickly recreate the state of Mars’s nightside upper atmosphere with better vertical consistency.
Open → 2609.10077v1