Earthly Machine Learning

Citation: Kaack, L. H., Donti, P. L., Strubell, E., Kamiya, G., Creutzig, F., & Rolnick, D. (2022). Aligning artificial intelligence with climate change mitigation. Nature Climate Change, 12, 518–527. https://doi.org/10.1038/s41558-022-01377-7Main Takeaways:Three Layers of AI's Climate Footprint: The authors propose a framework that splits machine learning's climate impact into three distinct categories — the energy and hardware emissions of computing itself, the immediate effects of specific ML applications, and the broader system-level changes that ML induces across society. The categories that are easiest to measure (like the electricity used to train a model) are likely not the ones with the largest effects, which is why most current discussions of "AI and climate" capture only a sliver of the real picture.Computing Is a Small Slice — For Now: The entire global ICT sector accounts for roughly 1.4% of global greenhouse gas emissions, and AI workloads are only a fraction of that. But the trajectory is steep: at Facebook, ML training compute has been growing about 150% per year and inference compute about 105% per year, far outpacing efficiency gains. Even striking efficiency wins — like Google's TPU being 30–80 times more energy-efficient than contemporary CPUs or GPUs — can be swamped by raw growth in demand.The "Internet of Cows" Problem: ML is a general-purpose tool, which means it's just as good at accelerating oil and gas exploration or scaling up cattle farming (an industry already responsible for about 9% of global emissions) as it is at forecasting solar power or optimizing data center cooling. Whether AI is net-positive or net-negative for the climate is genuinely undetermined, and depends on which applications get funded, deployed, and regulated.System-Level Effects May Dwarf Everything Else: The largest climate impacts of AI may come not from training runs or even individual applications, but from how ML reshapes society — through rebound effects (efficiency gains that drive more consumption), technological lock-in (autonomous cars entrenching private vehicle travel over transit and rail), and ML-powered recommender systems that boost demand for emissions-intensive goods. These effects are the hardest to quantify but potentially the most consequential, and the authors argue they need to be built into climate scenario modeling — something the IEA, EIA, and IPCC's Shared Socioeconomic Pathways largely don't do today.

Machine learning for the physics of climateCitation: Bracco, A., Brajard, J., Dijkstra, H. A., Hassanzadeh, P., Lessig, C., & Monteleoni, C. (2025). Machine learning for the physics of climate. Nature Reviews Physics, 7, 6–20. https://doi.org/10.1038/s42254-024-00776-3Main Takeaways:Breaking the El Niño Spring Barrier: For decades, forecasts of the El Niño Southern Oscillation hit a hard wall at roughly 6 months lead time — a limit known as the spring predictability barrier. Convolutional neural networks trained on a mix of climate model and reanalysis data have shattered this ceiling, delivering skillful forecasts at 17 months out, with newer architectures pushing to 21–24 months. ML models can also now anticipate which type of El Niño will develop (eastern vs. central Pacific), which matters enormously because the two flavors produce very different regional impacts around the world.Weather Forecasting at a Fraction of the Cost: A new generation of ML weather emulators — Pangu-Weather, GraphCast, FourCastNet, FuXi, NeuralGCM — now match or beat the European Centre's flagship physics-based forecasting system on most variables, including hurricane tracks, while running orders of magnitude faster. They achieve this with surprisingly compressed state representations: roughly 10 vertical atmospheric levels and 0.25° horizontal resolution, compared to 100+ levels and 0.1° in conventional models. The catch is that these models can violate basic physics — geostrophic balance, energy conservation, the butterfly effect — which currently blocks naive extension to climate timescales.Hybrid Models Are Eating the Climate Stack: Pure ML works for short-range forecasts, but for climate-length runs the field is converging on hybrid architectures that pair a traditional dynamical core with neural-network parameterizations of sub-grid processes like clouds, turbulence, and gravity waves. Google's NeuralGCM exemplifies the approach and already reduces biases in tropical cyclone frequency and tracks. A telling case study on the quasi-biennial oscillation showed that an offline-trained neural network produced unstable, unphysical results — but retraining just two layers online, coupled to the model, recovered the correct physics. Offline-only or online-only training each fail in characteristic ways; the mix is what works.The Data Wall Is the Real Bottleneck: Climate ML has less than 50 years of dense satellite-era observations to work with, and those observations are heavily biased toward the atmosphere and ocean surface — a single, spatiotemporally correlated realization of one climate. This limits how confidently ML models can extrapolate to warmer, unseen climates, which is exactly what climate projection requires. The path forward involves three parallel bets: hybrid physics-ML models that bake in conservation laws, large-scale "foundation models" for weather and climate trained across simulations and observations together (efforts like ClimaX and AtmoRep are early examples), and rare-event sampling strategies to handle the extremes that matter most for adaptation policy but are by definition underrepresented in any training set.

Citation: Benson, V., Bastos, A., Reimers, C., Winkler, A. J., Yang, F., & Reichstein, M. (2025). Atmospheric transport modeling of CO2 with neural networks. Journal of Advances in Modeling Earth Systems, 17, e2024MS004655. https://doi.org/10.1029/2024MS004655Main Takeaways:A New Benchmark for AI Carbon Tracking: The authors introduce CarbonBench, the first systematic benchmark dataset designed specifically for training and evaluating machine learning emulators of Eulerian atmospheric transport. Built from CarbonTracker CT2022 inversions and ObsPack station observations, it ships at three resolutions (the coarsest being 5.625° × 10 vertical levels × 6h) and is engineered to plug directly into modern deep learning pipelines — opening atmospheric carbon modeling to the broader ML community.SwinTransformer Wins, Decisively: Of the four architectures tested (UNet, GraphCast, SFNO, and SwinTransformer), the SwinTransformer reaches near-perfect emulation with a 90-day R² above 0.99 and stays stable in physically plausible forward runs for over three years — a regime where neural PDE solvers typically blow up. At measurement stations, it actually captures the seasonal cycle in Svalbard better than TM5, the conventional model it was trained to emulate, possibly due to differences in boundary layer transport near the poles.Physics Tricks Were the Unlock: Out of the box, the neural networks were unstable — especially the mesh-based UNet and GraphCast. Two simple physics-aware adjustments fixed this across all four architectures: centering the CO2 input field at each timestep to remove the covariate shift from steadily rising atmospheric CO2 (called CentFlux), and a post-hoc mass fixer that rescales predicted mass to match the surface flux budget. The result is mass conservation with RMSE of just 0.00058 PgC against a total atmospheric carbon mass of ~865 PgC — effectively negligible.Speed Isn't the Selling Point (Yet): Unlike AI weather models, which famously outpace numerical forecasting by orders of magnitude, the SwinTransformer is not significantly faster than TM5 at this resolution — about 1.5 seconds for a 30-day run on an A40 GPU versus a few minutes for TM5 on 24 CPUs. The real promise lies elsewhere: the networks are fully differentiable (useful for inverse modeling of surface fluxes), natively support batched ensembles, and scale better to high resolution where conventional solvers become prohibitively expensive — exactly the regime where current CO2 inversions struggle most.

Citation: Zhu, X. X., Xiong, Z., Wang, Y., Stewart, A. J., Heidler, K., Wang, Y., Yuan, Z., Dujardin, T., Xu, Q., & Shi, Y. (2026). On the foundations of Earth foundation models. Communications Earth & Environment, 7, 103. https://doi.org/10.1038/s43247-025-03127-xMain Takeaways:Current Earth AI Models Are Missing the Point: Researchers have identified eleven features that an ideal Earth foundation model must have — including geolocation awareness, multi-sensor integration, physical consistency, and carbon minimization — yet no existing model comes close to checking all eleven boxes. Most models focus on only one or two features, leaving a major gap between what we have and what we actually need to tackle real-world climate and environmental challenges.The Data Situation Is More Lopsided Than You'd Think: There are now over 1,000 active remote sensing satellites generating nearly 100 petabytes of open satellite data — but labeled datasets used to train AI models account for less than 0.1% of that archive. This massive imbalance is precisely why self-supervised foundation models, which can learn from unlabeled data, are so critical for Earth science going forward.Weather AI Is Already Dramatically More Efficient — But Incomplete: Models like FourCastNet can generate a week-long global weather forecast in under two seconds on a single GPU, using roughly 12,000 times less energy than traditional forecasting systems. Despite this leap in efficiency, major gaps remain: models struggle beyond two-week forecasts, long-term climate projections drift due to incomplete energy balance, and connecting fine-scale satellite imagery with coarse climate models remains largely unsolved.What Comes After the Ideal Model: Once a true Earth foundation model exists, the authors argue the most exciting frontier is using it to build an "Earth Embedding" — a compact, unified representation of our entire planet that researchers worldwide could query without ever touching raw satellite data. Beyond that, challenges like machine unlearning (making models forget sensitive imagery), adversarial defenses, and continual learning as the climate itself changes will define the next generation of Earth AI research.

Citation: Olivetti, L., & Messori, G. (2025). Whose weather is it? A fairness framework for data-driven weather forecasting. Environmental Research Letters, 20, 121006. https://doi.org/10.1088/1748-9326/ae21f5Main Takeaways:AI Weather Models Aren't Fair to Everyone: The latest generation of AI-powered weather forecasts improves predictions globally — but not equally. Using ECMWF's AIFS model as a case study, the authors show that wealthier and more densely populated areas consistently receive a higher share of forecast improvements compared to poorer and more rural regions, violating basic fairness criteria borrowed from the algorithmic fairness literature.Two Measurable Fairness Tests — Both Failed: The paper proposes two concrete criteria: statistical parity(improvement rates should be similar across income groups) and conditional independence (a region's GDP or population density should not predict whether it benefits from the new model). Across nearly all tested variables and forecast lead times, AIFS fails both tests at the 0.01 significance level — meaning the disparity is not a statistical fluke.Extreme Weather Is Where the Gap Hurts Most: For standard temperature and wind forecasts, gaps between rich and poor regions are modest. But for cold extremes, the fairness gap is especially pronounced — precisely the events where accurate early warnings matter most for vulnerable populations with fewer resources to adapt.Fixing It Is Technically Feasible: Unlike traditional physics-based models, AI weather models offer genuine design levers for fairness. The authors describe two practical approaches: adding penalty terms to the loss function (such as the Hilbert–Schmidt Independence Criterion) to reduce associations with protected variables, and using geographically adaptive weighting that iteratively compensates for emerging performance gaps — without necessarily sacrificing global accuracy.

Citation: Spuler, F. R., Kretschmer, M., Balmaseda, M. A., Kovalchuk, Y., & Shepherd, T. G. (2025). Learning predictable and informative dynamical drivers of extreme precipitation using variational autoencoders. Weather and Climate Dynamics, 6, 995–1014. https://doi.org/10.5194/wcd-6-995-2025Main Takeaways:Innovative Machine Learning Approach: The study introduces the Categorical Mixture Model Variational Autoencoder (CMM-VAE), a novel generative machine learning method designed to identify probabilistic atmospheric circulation regimes by combining targeted dimensionality reduction and probabilistic clustering into a single model.Resolving a Major Forecasting Trade-off: Traditionally, atmospheric regimes are either highly predictable globally but locally uninformative, or highly informative for local impacts but lacking in subseasonal predictability. CMM-VAE resolves this trade-off, successfully identifying patterns that predict local extremes without sacrificing forecast skill at subseasonal lead times.Targeted Application for Moroccan Rainfall: When applied to extreme winter precipitation in Morocco, the CMM-VAE method successfully disentangled a distinct, highly impactful weather pattern—a Scandinavian blocking coupled with a localized cut-off low—that traditional linear clustering methods failed to isolate.Linkages to Global Climate Drivers: The weather regimes identified by the model remain physically interpretable and show clear, predictable teleconnections to large-scale, low-frequency climate drivers, notably the Madden-Julian Oscillation (MJO) and the Stratospheric Polar Vortex (SPV).Enhancing Early Warning Systems: By providing a better representation of regional dynamical drivers, this framework offers significant potential to improve subseasonal-to-seasonal (S2S) forecasts, statistical downscaling, and early-warning systems for severe, localized weather impacts.

Green and intelligent: the role of AI in the climate transitionCitation: Stern, N., Romani, M., Pierfederici, R., Braun, M., Barraclough, D., Lingeswaran, S., Weirich-Benet, E., & Niemann, N. (2025). Green and intelligent: the role of AI in the climate transition. https://doi.org/10.1038/s44168-025-00252-3.Main Takeaways:Five Key Areas for Climate Action: Artificial Intelligence can accelerate the net-zero transition across five primary avenues: transforming complex economic systems, innovating technology discovery and resource efficiency, nudging consumer behavior toward sustainable choices, modeling climate systems for better policy, and managing adaptation and resilience.Significant Emissions Reduction Potential: By applying AI to just three major sectors—power, food (specifically meat and dairy), and mobility (light road vehicles)—global emissions could be reduced by 3.2 to 5.4 GtCO2e annually by 2035.Net-Positive Climate Impact: The emissions savings generated by AI in these three sectors alone would more than offset the projected 0.4 to 1.6 GtCO2e increase in emissions caused by the energy consumption of all global AI activities and data centers.Closing the Emissions Gap: Harnessing AI to improve the efficiency and market adoption of low-carbon solutions could push global progress 36% closer to aligning with an ambitious emissions reduction trajectory by 2035.The Critical Role of Government: Relying solely on market forces to govern AI is risky; an "active state" is essential to direct AI toward public goods, regulate its environmental footprint (like mandating renewable energy for data centers), and ensure equitable deployment so the Global South is not left behind.

Climate Knowledge in Large Language ModelsKuznetsov, I., Grassi, J., Pantiukhin, D., Shapkin, B., Jung, T., & Koldunov, N. (2025). Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research.LLMs have an internal "map" of the climate, but it is fuzzy: Without access to external tools, Large Language Models (LLMs) can recall the general structure of Earth’s climate—correctly identifying that the tropics are warm and high latitudes are cold. However, their specific numeric predictions are often inaccurate, with average errors ranging from 3°C to 6°C compared to historical weather data.Location names matter more than coordinates: The study found that providing geographic context—such as the country, region, or city name—alongside coordinates reduced prediction errors by an average of 27%. This suggests models rely heavily on text associations with place names rather than possessing a precise spatial understanding of latitude and longitude.Performance struggles with altitude and local trends: Models perform significantly worse in mountainous regions, with errors spiking sharply at elevations above 1500 meters. Furthermore, while LLMs can estimate the global average magnitude of warming, they fail to accurately reproduce the specific local patterns of temperature change that are essential for understanding regional climate dynamics.Caution is needed for scientific use: The results highlight that while LLMs encode a static snapshot of climatological averages, they lack true physical understanding and struggle with dynamic trends. Consequently, they should not be relied upon as standalone climate databases; reliable applications require connecting them to external, authoritative data sources.

Artificial Intelligence for Atmospheric Sciences: A Research RoadmapCitation: Zaidan, M. A., Motlagh, N. H., Nurmi, P., Hussein, T., Kulmala, M., Petäjä, T., & Tarkoma, S. (2025). Artificial Intelligence for Atmospheric Sciences: A Research Roadmap.Revolutionizing Environmental Monitoring: The paper illustrates how AI is transforming atmospheric sciences by bridging the gap between computer science and environmental research. It details how AI processes massive datasets generated by diverse sources—including satellite imagery, ground-based research stations, and low-cost IoT sensors—to improve our understanding of air quality, extreme weather events, and climate change.Optimizing Infrastructure and Prediction: Current AI applications are already enhancing operational meteorology and Earth system modeling. By utilizing techniques like deep learning and neural networks, researchers can automate sensor calibration, detect anomalies in real-time, and simulate complex climate scenarios with greater speed and efficiency than traditional physical models allow.A Roadmap for Future Hardware: To handle the escalating demand for data, the authors propose a hardware roadmap that includes self-sustaining and biodegradable sensor networks, CubeSat constellations for high-resolution monitoring, and the adoption of cutting-edge computing paradigms like quantum, neuromorphic, and DNA-based molecular computing.Next-Generation AI Methodologies: The paper argues for the adoption of advanced AI techniques such as Foundation Models and Generative AI (including Digital Twins of Earth) to predict complex atmospheric phenomena. Crucially, it emphasizes the need for Explainable AI (XAI) and Physics-Informed Machine Learning to solve the "black box" problem, ensuring that AI predictions abide by physical laws and are transparent enough for scientists and policymakers to trust.From Data to Action: Beyond observation, the research highlights the shift toward actionable insights. This includes automated feedback loops (such as smart HVAC systems responding to air quality data), the integration of citizen science to augment data collection, and the establishment of robust ethical frameworks to manage data privacy and governance in global monitoring networks.

Differentiable and accelerated spherical harmonic and Wigner transformsMatthew A. Price, Jason D. McEwen*Journal of Computational Physics (2024)** This work introduces novel algorithmic structures for the **accelerated and differentiable computation** of generalized Fourier transforms on the sphere ($S^2$) and the rotation group ($SO(3)$), specifically spherical harmonic and Wigner transforms.* A key component is a **recursive algorithm for Wigner d-functions** designed to be stable to high harmonic degrees and extremely parallelizable, making the algorithms well-suited for high throughput computing on modern hardware accelerators such as GPUs.* The transforms support efficient computation of gradients, which is critical for machine learning and other differentiable programming tasks, achieved through a **hybrid automatic and manual differentiation approach** to avoid the memory overhead associated with full automatic differentiation.* Implemented in the open-source **S2FFT** software code (within the JAX differentiable programming framework), the algorithms support various sampling schemes, including equiangular samplings that admit exact spherical harmonic transforms.* Benchmarking results demonstrate **up to a 400-fold acceleration** compared to alternative C codes, and the transforms exhibit **very close to optimal linear scaling** when distributed over multiple GPUs, yielding an unprecedented effective linear time complexity (O(L)) given sufficient computational resources.

Score-based diffusion nowcasting of GOES imagery*Randy J. Chase, Katherine Haynes, Lander Ver Hoef, Imme Ebert-Uphoff, a Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, CO, b Electrical and Computer Engineering, Colorado State University, Fort Collins, CO** The research explored score-based diffusion models to perform short-term forecasts (nowcasting) of GOES geostationary infrared satellite imagery (zero to three hours). This newer machine learning methodology combats the issue of **blurry forecasts** often produced by earlier neural network types, enabling the generation of clearer and more realistic-looking forecasts.* The **residual correction diffusion model (CorrDiff)** proved to be the best-performing model, quantitatively outperforming all other tested diffusion models, a traditional Mean Squared Error trained U-Net, and a persistence forecast by one to two kelvin on root mean squared error.* The diffusion models demonstrated sophisticated predictive capabilities, showing the ability to not only advect existing clouds but also to **generate and decay clouds**, including initiating convection, despite being initialized with only the past 20 minutes of satellite imagery.* A key benefit of the diffusion framework is the capacity for **out-of-the-box ensemble generation**, which enhances pixel-based metrics and provides useful uncertainty quantification where the spread of the ensemble generally correlates well to the forecast error.* However, the diffusion models are computationally intensive, with the Diff and CorrDiff models taking approximately five days to train on specialized hardware and about 10 minutes to generate a 10-member, three-hour forecast, compared to just 10 seconds for the baseline U-Net forecast.

FuXi-Ocean: A Global Ocean Forecasting System with Sub-Daily Resolution*Qiusheng Huang, Yuan Niu, Xiaohui Zhong, Anboyu Guo, Lei Chen, Dianjun Zhang, Xuefeng Zhang, Hao Li*---* **First Data-Driven Sub-Daily Global Forecast:** FuXi-Ocean is the first deep learning-based global ocean forecasting model to achieve six-hour temporal resolution at an eddy-resolving 1/12° spatial resolution, with vertical coverage extending up to 1500 meters. This capability addresses a crucial need for high-frequency predictions that traditional numerical models struggle to deliver efficiently.* **Adaptive Temporal Modeling Innovation:** A key component of the model is the **Mixture-of-Time (MoT) module**, which adaptively integrates predictions from multiple temporal contexts based on variable-specific reliability. This mechanism is crucial for accommodating the diverse temporal dynamics of different ocean variables (e.g., fast-changing surface variables vs. slowly evolving deep-ocean processes) and effectively mitigates the accumulation of forecast errors in sequential prediction.* **Superior Performance and Efficiency:** The model demonstrates superior skill in predicting key variables (temperature, salinity, and currents) compared to state-of-the-art operational numerical forecasting systems (like HYCOM, BLK, and FOAM) at sub-daily intervals. Furthermore, it achieves this high performance with remarkable data efficiency, requiring only approximately 9 years of training data and relying solely on ocean variables (T, S, U, V, SSH) as input, without external data dependencies like atmospheric forcing.* **High-Impact Applications:** By providing accurate, high-resolution, sub-daily forecasts, FuXi-Ocean creates critical opportunities for maritime operations, including improved navigation, search and rescue, oil spill trajectory tracking, and enhanced marine resource management, particularly due to its comprehensive vertical coverage (0-1500 m).

Beyond the Training Data: Confidence-Guided Mixing of Parameterizations in a Hybrid AI-Climate Model*By Helge Heuer, Tom Beucler, Mierk Schwabe, Julien Savre, Manuel Schlund, and Veronika Eyring** This paper presents a **successful proof-of-concept for transferring a machine learning (ML) convection parameterization**—trained on the ClimSim dataset—to the ICON-A climate model. The resulting hybrid ML-physics model achieved stable and accurate simulations in long-term AMIP-style runs lasting at least 20 years.* A core innovation is the **confidence-guided mixing scheme**, which allows the Neural Network (NN) to predict its own error. When the NN's predicted confidence is low (e.g., in moist, unstable regimes or high-variability areas), its prediction is mixed with the conventional Tiedtke convection scheme. This mechanism improves reliability, prevents unphysical outputs by detecting potential extrapolation beyond the training domain, and makes the hybrid model tunable against observations.* The scheme's robustness and accuracy were further enhanced through the **use of a physics-informed loss function**—which encourages adherence to conservation laws like enthalpy and mass—and **noise-augmented training**. These techniques mitigate stability issues commonly faced by ML parameterizations and significantly improve physical consistency compared to purely data-driven models.* In evaluation against observational data, several hybrid configurations **outperformed the default Tiedtke scheme**, demonstrating improved precipitation statistics and showing a better representation of global climate variables. The confidence-guided approach demonstrated a fundamental change in the model's behavior, with the ML component contributing approximately 67% of the convective tendencies on average.

Climate in a Bottle: Towards a Generative Foundation Model for the Kilometer-Scale Global Atmosphere(By Noah D. Brenowitz, Tao Ge, Akshay Subramaniam, Peter Manshausen, Aayush Gupta, David M. Hall, Morteza Mardani, Arash Vahdat, Karthik Kashinath, Michael S. Pritchard, NVIDIA* The paper introduces **Climate in a Bottle (cBottle)**, a generative diffusion-based AI framework capable of synthesizing full global atmospheric states at an unprecedented $\mathbf{5 \text{ km resolution}}$ (over 12.5 million pixels per sample). Unlike prevailing auto-regressive paradigms, cBottle samples directly from the full distribution of atmospheric states without requiring a previous time step, thereby avoiding issues like drifts and instabilities inherent to time-stepping models.* cBottle utilizes a **two-stage cascaded diffusion approach**: a global coarse-resolution generator conditioned on minimal climate-controlling inputs (such as monthly sea surface temperature and solar position), followed by a patch-based 16x super-resolution module.* The model demonstrates **foundational versatility** by being trained jointly on multiple data modalities, including ERA5 reanalysis and ICON global cloud-resolving simulations. This enables various zero-shot applications such as climate downscaling, channel infilling for missing or corrupted variables, bias correction between datasets, and translation between these modalities.* cBottle proposes a new form of **interactive climate modeling** through the use of guided diffusion. By training a classifier alongside the generator, users can steer the model to conditionally generate physically plausible **extreme weather events, such as Tropical Cyclones**, at specified locations on demand, circumventing the need to sift through petabytes of output to find rare events.* The model exhibits **high climate faithfulness** across a battery of tests, including reproducing diurnal-to-seasonal scale variability, large-scale modes of variability (like the Northern Annular Mode), and tropical cyclone statistics. Furthermore, it achieves **extreme distillation** by encapsulating massive datasets into a few GB of neural network weights, offering a 256x compression ratio per channel.

Probabilistic measures afford fair comparisons of AIWP and NWP model output (Tilmann Gneiting, Tobias Biegert, Kristof Kraus, Eva-Maria Walz, Alexander I. Jordan, Sebastian Lerch, June 10, 2025)Introduction of a New Fair Comparison Metric: The paper introduces the Potential Continuous Ranked Probability Score (PC), a new measure designed to allow fair and meaningful comparisons between single-valued output from data-driven Artificial Intelligence based Weather Prediction (AIWP) models and physics-based Numerical Weather Prediction (NWP) models. This approach addresses concerns that traditional loss functions (like RMSE) may unfairly favor AIWP models, which often optimize their training using these metrics. Methodology Based on Probabilistic Postprocessing: PC is calculated by applying the same statistical postprocessing technique—specifically Isotonic Distributional Regression (IDR), also known as Easy Uncertainty Quantification (EasyUQ)—to the deterministic output of both AIWP and NWP models. PC is then defined as the mean Continuous Ranked Probability Score (CRPS) of these newly generated probabilistic forecasts. Measure of Potential Skill and Invariance: PC quantifies potential predictive performance. A key property of PC is that it is invariant under strictly increasing transformations of the model output, treating both forecasts equally and facilitating comparisons where the pre-specification of a loss function might otherwise place competitors on unequal footings. AIWP Outperformance and Operational Proxy: When applied to WeatherBench 2 data, the PC measure demonstrated that the data-driven GraphCast model outperforms the leading physics-based ECMWF high-resolution (HRES) model. Furthermore, the PC measure for the HRES model was found to align exceptionally well with the mean CRPS of the operational ECMWF ensemble, confirming that PC serves as a reliable proxy for the performance of real-time operational probabilistic products.

Jigsaw: Training Multi-Billion-Parameter AI Weather Models With Optimized Model ParallelismAuthors: Deifilia Kieckhefen, Markus Götz, Lars H. Heyen, Achim Streit, and Charlotte Debus (Karlsruhe Institute of Technology, Helmholtz AI)The paper introduces WeatherMixer (WM), a multi-layer perceptron (MLP)-based architecture designed for atmospheric forecasting, which serves as a competitive alternative to Transformer-based models. WM's workload scales linearly with input size, addressing the scaling challenges and quadratic computational complexity associated with the self-attention mechanism in Transformers when dealing with gigabyte-sized atmospheric data.• A novel parallelization scheme called Jigsaw parallelism is proposed, combining both domain parallelism and tensor parallelism to efficiently train multi-billion-parameter models. Jigsaw is optimized for large input data by fully sharding the data, model parameters, and optimizer states across devices, eliminating memory redundancy. Jigsaw effectively mitigates hardware bottlenecks, particularly I/O-bandwidth limitations frequently encountered in training large scientific AI models. Due to its partitioned data loading (domain parallelism), the scheme achieves superscalar weak scaling in I/O-bandwidth-limited systems. The method demonstrates excellent scaling behavior on high-performance computing systems, exceeding state-of-the-art performance in strong scaling in computation–communication-limited systems. The training was successfully scaled up to 256 GPUs, reaching peak performances of 9 and 11 PFLOPs.• Beyond hardware efficiency, Jigsaw improves predictive performance: by partitioning the model across more GPUs (model parallelism) instead of relying solely on data parallelism, it naturally enforces smaller global batch sizes, which empirically helps mitigate the problematic large-batch effects observed in AI weather models, leading to lower loss values.

XiChen: An observation-scalable fully AI-driven global weather forecasting system with 4D variational knowledgeAuthors: Wuxin Wang, Weicheng Ni, Lilan Huang, Tao Han, Ben Fei, Shuo Ma, Taikang Yuan, Yanlai Zhao, Kefeng Deng, Xiaoyong Li, Boheng Duan, Lei Bai, Kaijun RenXiChen is the first observation-scalable fully AI-driven global weather forecasting system. Its entire pipeline, from Data Assimilation (DA) to 10-day medium-range forecasting, can be accomplished within only 17 seconds using a single A100 GPU. This speed represents an acceleration exceeding 400-fold compared to the computational time required by operational Numerical Weather Prediction (NWP) systems. The system is architected upon a foundation model that is initially pre-trained for weather forecasting and subsequently fine-tuned to function as both observation operators and DA models. Crucially, the integration of four-dimensional variational (4DVar) knowledge ensures that XiChen’s DA and medium-range forecasting accuracy rivals that of operational NWP systems. XiChen demonstrates high scalability and robustness by employing a cascaded sequential DA framework to effectively assimilate both conventional observations (GDAS prepbufr) and raw satellite observations (AMSU-A and MHS). This design allows for the future integration of new observations simply by fine-tuning the respective observation operators and DA model components, which is critical for operational deployment. In terms of performance, XiChen achieves a skillful weather forecasting lead time exceeding 8.25 days (with ACC of Z500 > 0.6). This result is comparable to the Global Forecasting System (GFS) and substantially surpasses the performance of other end-to-end AI-based global weather forecasting systems, such as Aardvark (less than 8 days) and GraphDOP (about 5 days). A dual DA framework is implemented to operationalize XiChen as a continuous forecasting system. This framework utilizes separate 12-hour and 3-hour Data Assimilation Windows (DAW) to circumvent the multi-hour latency characteristic of high-resolution systems (like IFS HRES), thereby enabling the real-time acquisition of medium-range forecast products.

A data-to-forecast machine learning system for global weather Xiuyu Sun et al. (2025). A data-to-forecast machine learning system for global weather. Nature Communications, https://doi.org/10.1038/s41467-025-62024-1• FuXi Weather is introduced as a groundbreaking end-to-end machine learning system for global weather forecasting. It autonomously performs data assimilation and forecasting in a 6-hour cycle, directly processing raw multi-satellite observations, and notably, it is the first such system to demonstrate continuous cycling operation over a full one-year period.• The system exhibits superior forecast accuracy in observation-sparse regions, outperforming traditional high-resolution forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF HRES) beyond day one in areas like central Africa and northern South America, despite utilizing substantially fewer observations.• Globally, FuXi Weather delivers comparable 10-day forecast performance to ECMWF HRES, generating reliable forecasts at a 0.25° resolution and extending the skillful lead times for a number of key meteorological variables.• FuXi Weather offers a cost-effective and physically consistent alternative to traditional Numerical Weather Prediction (NWP) systems. Its computational efficiency and reduced complexity are valuable for improving operational forecasts and enhancing climate resilience in regions with limited land-based observational infrastructure.• This development challenges the prevailing view that standalone machine learning-based weather forecasting systems are not viable for operational use, demonstrating a significant step forward in the application of AI to real-world weather prediction.

FourCastNet 3: A geometric approach to probabilistic machine-learning weather forecasting at scale Boris Bonev, Thorsten Kurth, Ankur Mahesh, Mauro Bisson, Jean Kossaifi, Karthik Kashinath, Anima Anandkumar, William D. Collins, Michael S. Pritchard, and Alexander Keller• FourCastNet 3 (FCN3) introduces a pioneering geometric machine learning approach for probabilistic ensemble weather forecasting. It is designed to respect spherical geometry and accurately model the spatially correlated probabilistic nature of weather, resulting in stable spectra and realistic dynamics across multiple scales. The architecture is a purely convolutional neural network tailored for spherical geometry.• Achieves superior forecasting accuracy and speed, surpassing leading conventional ensemble models and rivaling the best diffusion-based ML methods. FCN3 produces forecasts 8 to 60 times faster than these approaches; for instance, a 60-day global forecast at 0.25°, 6-hourly resolution is generated in under 4 minutes on a single GPU.• Demonstrates exceptional physical fidelity and long-term stability, maintaining excellent probabilistic calibration and realistic spectra even at extended lead times of up to 60 days. This crucial achievement mitigates issues like blurring and the build-up of small-scale noise, which challenge other machine learning models, paving the way for physically faithful data-driven probabilistic weather models.• Enables scalable and efficient operations through a novel training paradigm that combines model- and data-parallelism, allowing large-scale training on 1024 GPUs and more. All key components, including training and inference code, are fully open-source, providing transparent and reproducible tools for meteorological forecasting and atmospheric science research.

Can AI weather models predict out-of-distribution gray swan tropical cyclones?by Y. Qiang Sun, Pedram Hassanzadeh, Mohsen Zand, Ashesh Chattopadhyay, Jonathan Weare, and Dorian S. AbbotInability to Extrapolate to Gray Swans Globally: AI weather models like FourCastNet struggle to predict "gray swan" tropical cyclones (TCs), which are rare, strong, and absent from training data. When Category 3-5 TCs are entirely removed from the global training dataset, the model cannot extrapolate from weaker storms (Category 1-2) to accurately forecast these stronger, unseen events, often leading to dangerous "false negative" predictions. This limitation persists even if the training data includes strong extratropical cyclones, as their dynamics differ from TCs.Limited Generalization Across Basins for Dynamically Similar Events: Despite the global extrapolation challenge, FourCastNet can demonstrate some ability to generalize learning across tropical basins for dynamically similar strong storms. This means that if the model has seen strong TCs in one ocean basin, it can apply that learned knowledge to forecast similar strong TCs in another basin, even if those specific events were excluded from the training data for that particular region.Lack of Physical Consistency and Masked Performance: Current AI weather models, including FourCastNet, fail to reproduce key physical balances like the gradient-wind balance that TCs obey in real-world data, regardless of whether they were trained on full or reduced datasets. Furthermore, common evaluation metrics (e.g., anomaly correlation coefficient or root-mean-square error) can obscure these critical shortcomings by showing similar overall performance for general weather or less extreme events, highlighting the need for specialized tests for gray swans.Implications and Future Directions: This research suggests that current AI weather models may provide unreliable early warnings for unprecedented extreme weather events, potentially leading to serious societal risks. It also indicates that AI climate emulators might mischaracterize extreme weather statistics for gray swans. The study emphasizes the urgent need for novel learning strategies (such as incorporating physics-based synthetic data or rare-event sampling algorithms) and rigorous testing methodologies to improve and reliably validate AI models for these high-impact, out-of-distribution events.

Probabilistic Emulation of a Global Climate Model with Spherical DYffusionby Salva Rühling Cachay, Brian Henn, Oliver Watt-Meyer, Christopher S. Bretherton, Rose YuThe paper introduces Spherical DYffusion, the first conditional generative model for probabilistic emulation of a realistic global climate model, offering efficient and accurate climate ensemble simulations.It demonstrates that weather forecasting performance is not a strong indicator of long-term climate performance, a crucial insight for developing climate models.Spherical DYffusion significantly reduces climate biases compared to existing baselines like ACE and DYffusion, achieving errors often closer to the reference simulation's "noise floor".The model generates stable, 10-year-long probabilistic predictions with minimal computational overhead, being more than 25 times faster than the physics-based FV3GFS model it emulates, while also reproducing consistent climate variability.

"Do AI models produce better weather forecasts than physics-based models? A quantitative evaluation case study of Storm Ciarán" By Andrew J. Charlton-Perez, Helen F. Dacre, Simon Driscoll, Suzanne L. Gray, Ben Harvey, Natalie J. Harvey, Kieran M. R. Hunt, Robert W. Lee, Ranjini Swaminathan, Remy Vandaele & Ambrogio Volonté. Published in partnership with CECCR at King Abdulaziz University, Nature, DOI: 10.1038/s41612-024-00638-w.Here are the main takeaways from the paper:• AI models (FourCastNet, Pangu-Weather, GraphCast, FourCastNet-v2) demonstrate strong capabilities in capturing large-scale dynamical drivers vital for rapid storm development, such as the storm's position relative to upper-level jets. They also accurately reproduce the larger synoptic-scale structure of cyclones like Storm Ciarán, including the cloud head's position and the warm sector's shape. Despite these strengths, AI models consistently underestimate the peak amplitude of winds, both at the surface and in the free atmosphere, associated with storms. They also struggle to resolve detailed structures crucial for issuing severe weather warnings, such as sharp bent-back warm frontal gradients, and show variable success in capturing warm core seclusion. The underestimation of strong winds is not a consequence of the AI models' output resolution or their training data. This discrepancy persists even when compared against ERA5 (on which these models were trained) and numerical weather prediction (NWP) models of similar resolution, suggesting a more fundamental limitation in their ability to represent intense wind features.The case study of Storm Ciarán highlights the pressing need for a more comprehensive assessment of machine learning weather forecasts. Moving beyond isolated error metrics to evaluate all relevant spatio-temporal features of physical phenomena is essential for identifying specific areas for improvement and fostering rapid advancements in this new and potentially transformative forecasting tool.

🧠 Abstract:Climate change is increasing the frequency and severity of disasters, demanding more effective Early Warning Systems (EWS). While current systems face hurdles in forecasting, communication, and decision-making, this episode examines how integrated Artificial Intelligence (AI) can revolutionize risk detection and response.📌 Bullet points summary:Current EWS struggle with forecasting accuracy, impact prediction across diverse contexts, and effective communication with affected communities.Integrated AI and Foundation Models (FMs) enhance EWS by improving forecast precision, offering impact-specific alerts, and utilizing diverse data sources—from weather to social media.Foundation Models for geospatial and meteorological data, combined with natural language processing, pave the way for user-adaptive, intuitive warning systems, including chatbots and realistic visualizations.Ensuring equity and effectiveness in AI-driven EWS requires addressing data bias, robustness, ownership issues, and power dynamics—guided by FATES principles and supported by open-source tools, global cooperation, and digital inclusivity.💡 The Big Idea:Integrated AI holds the key to transforming climate early warning—from hazard alerts to adaptive, inclusive, and impact-driven systems that empower communities worldwide.📖 Citation:Reichstein, Markus, et al. "Early warning of complex climate risk with integrated artificial intelligence." Nature Communications 16.1 (2025): 2564. https://doi.org/10.1038/s41467-025-57640-w

🧠 Abstract:Machine Learning (ML) is increasingly influential in weather and climate prediction. Recent advances have led to fully data-driven ML models that often claim to outperform traditional physics-based systems. This episode evaluates forecasts from three leading ML models—Pangu-Weather, FourCastNet, and GraphCast—focusing on their accuracy and physical realism.📌 Bullet points summary:ML models like Pangu-Weather, FourCastNet, and GraphCast fail to capture sub-synoptic and mesoscale phenomena with adequate fidelity, producing forecasts that become overly smooth over time.Their energy spectra diverge significantly from traditional models and reanalysis data, leading to poor representation of features below 300–400 km scales.They lack accurate representation of key physical balances in the atmosphere, such as geostrophic wind balance and the divergent-rotational wind ratio, affecting the realism of weather diagnostics.Though computationally efficient and strong in certain metrics, these models should be seen as forecast refiners rather than full-fledged atmospheric simulators or "digital twins," as they still rely heavily on traditional models for training and input.💡 The Big Idea:While ML models mark a significant advancement, their current limitations highlight the indispensable role of physical principles and traditional modeling in weather prediction.📖 Citation:Bonavita, Massimo. "On some limitations of current machine learning weather prediction models." Geophysical Research Letters 51.12 (2024): e2023GL107377. https://doi.org/10.1029/2023GL107377

🌍 Abstract:Artificial intelligence (AI) is transforming Earth system science, especially in modeling and understanding extreme weather and climate events. This episode explores how AI tackles the challenges of analyzing rare, high-impact phenomena using limited, noisy data—and the push to make AI models more transparent, interpretable, and actionable.📌 Bullet points summary:🌪️ AI is revolutionizing how we model, detect, and forecast extreme climate events like floods, droughts, wildfires, and heatwaves, and plays a growing role in attribution and risk assessment.⚠️ Key challenges include limited data, lack of annotations, and the complexity of defining extremes, all of which demand robust, flexible AI approaches that perform well under novel conditions.🧠 Trustworthy AI is critical for safety-related decisions, requiring transparency, interpretability (XAI), causal inference, and uncertainty quantification.📢 The “last mile” focuses on operational use and risk communication, ensuring AI outputs are accessible, fair, and actionable in early warning systems and public alerts.🤝 Cross-disciplinary collaboration is vital—linking AI developers, climate scientists, field experts, and policymakers to build practical and ethical AI tools that serve real-world needs.💡 Big idea:AI holds powerful promise for extreme climate analysis—but only if it's built to be trustworthy, explainable, and operationally useful in the face of uncertainty.📚 Citation:Camps-Valls, Gustau, et al. "Artificial intelligence for modeling and understanding extreme weather and climate events." Nature Communications 16.1 (2025): 1919.https://doi.org/10.1038/s41467-025-56573-8

🎙️ Abstract:Recent progress in data-driven weather forecasting has surpassed traditional physics-based systems. Yet, the common use of mean squared error (MSE) loss functions introduces a “double penalty,” smoothing out fine-scale structures. This episode discusses a simple, parameter-free fix to this issue by modifying the loss to disentangle decorrelation errors from spectral amplitude errors.🌪️ Data-driven weather models like GraphCast often produce overly smooth outputs due to MSE loss, limiting resolution and underestimating extremes.⚙️ The proposed Adjusted Mean Squared Error (AMSE) loss function addresses this by separating decorrelation and amplitude errors, improving spectrum fidelity.📈 Fine-tuning GraphCast with AMSE boosts resolution dramatically (from 1,250km to 160km), enhances ensemble spread, and sharpens forecasts of cyclones and surface winds.🔬 This shows deterministic forecasts can remain sharp and realistic without explicitly modeling ensemble uncertainty.Redefining the loss function in data-driven weather forecasting can drastically sharpen predictions and enhance realism—without adding complexity or parameters.📚 Citation:https://doi.org/10.48550/arXiv.2501.19374🔍 Bullet points summary:💡 Big idea:

🌍 Abstract:Projecting climate change is a generalization problem: we extrapolate the recent past using physical models across past, present, and future climates. Current climate models require representations of processes that occur at scales smaller than the model grid size, which remain the main source of projection uncertainty. Recent machine learning (ML) algorithms offer promise for improving these process representations but often extrapolate poorly outside their training climates. To bridge this gap, the authors propose a “climate-invariant” ML framework, incorporating knowledge of climate processes into ML algorithms, and show that this approach enhances generalization across different climate regimes.📌 Key Points:Highlights how ML models in climate science struggle to generalize beyond their training data, limiting their utility in future climate projections.Introduces a "climate-invariant" ML framework, embedding physical climate process knowledge into ML models through feature transformations of input and output data.Demonstrates that neural networks with climate-invariant design generalize better across diverse climate conditions in three atmospheric models, outperforming raw-data ML approaches.Utilizes explainable AI methods to show that climate-informed mappings learned by neural networks are more spatially local, improving both interpretability and data efficiency.💡 The Big Idea:Combining machine learning with physical insights through a climate-invariant approach enables models that not only learn from data but also respect the underlying physics—paving the way for more reliable and generalizable climate projections.📖 Citation:Beucler, Tom, et al. "Climate-invariant machine learning." Science Advances 10.6 (2024): eadj7250. DOI: 10.1126/sciadv.adj7250

🎙️ Episode 25: ClimaX: A foundation model for weather and climateDOI: https://doi.org/10.48550/arXiv.2301.10343🌀 Abstract:Most cutting-edge approaches for weather and climate modeling rely on physics-informed numerical models to simulate the atmosphere's complex dynamics. These methods, while accurate, are often computationally demanding, especially at high spatial and temporal resolutions. In contrast, recent machine learning methods seek to learn data-driven mappings directly from curated climate datasets but often lack flexibility and generalization. ClimaX introduces a versatile and generalizable deep learning model for weather and climate science, capable of learning from diverse, heterogeneous datasets that cover various variables, time spans, and physical contexts.📌 Bullet points summary:ClimaX is a flexible foundation model for weather and climate, overcoming the rigidity of physics-based models and the narrow focus of traditional ML approaches by training on heterogeneous datasets.The model utilizes Transformer-based architecture with novel variable tokenization and aggregation mechanisms, allowing it to handle diverse climate data efficiently.Pre-trained via a self-supervised randomized forecasting objective on CMIP6-derived datasets, ClimaX learns intricate inter-variable relationships, enhancing its adaptability to various forecasting tasks.Demonstrates strong, often state-of-the-art performance across tasks like multi-scale weather forecasting, climate projections (ClimateBench), and downscaling — sometimes outperforming even operational systems like IFS.The study highlights ClimaX's scalability, showing performance gains with more pretraining data and higher resolutions, underscoring its potential for future developments with increased data and compute resources.💡 Big idea:ClimaX represents a shift toward foundation models in climate science, offering a single, adaptable architecture capable of generalizing across a wide array of weather and climate modeling tasks — setting the stage for more efficient, data-driven climate research.📖 Citation:Nguyen, Tung, et al. "Climax: A foundation model for weather and climate." arXiv preprint arXiv:2301.10343 (2023).

🎙️ Episode 24: AI-empowered Next-Generation Multiscale Climate Modelling for Mitigation and Adaptation🔗 DOI: https://doi.org/10.1038/s41561-024-01527-w🌐 AbstractDespite decades of progress, Earth system models (ESMs) still face significant gaps in accuracy and uncertainty, largely due to challenges in representing small-scale or poorly understood processes. This episode explores a transformative vision for next-generation climate modeling—one that embeds AI across multiple scales to enhance resolution, improve model fidelity, and better inform climate mitigation and adaptation strategies.📌 Bullet points summaryExisting ESMs struggle with inaccuracies in climate projections due to subgrid-scale and unknown process limitations.A new approach is proposed that blends AI with multiscale modeling, combining fine-resolution simulations with coarser hybrid models that capture key Earth system feedbacks.This strategy is built on four pillars:Higher resolution via advanced computingPhysics-aware machine learning to enhance hybrid modelsSystematic use of Earth observations to constrain modelsModernized scientific infrastructure to operationalize insightsAims to deliver faster, more actionable climate data to support urgent policy needs for both mitigation and adaptation.Envisions hybrid ESMs and interactive Earth digital twins, where AI helps simulate processes more realistically and supports climate decision-making at scale.💡 The Big IdeaIntegrating AI into climate models across scales is not just an upgrade—it’s a shift towards smarter, faster, and more adaptive climate science, essential for responding to the climate crisis with precision and urgency.📖 CitationEyring, Veronika, et al. "AI-empowered next-generation multiscale climate modelling for mitigation and adaptation." Nature Geoscience 17.10 (2024): 963–971.

🎙️ Episode 23: FourCastNet – Accelerating Global High-Resolution Weather Forecasting Using Adaptive Fourier Neural Operators🔗 DOI: https://doi.org/10.1145/3592979.3593412🌍 AbstractAs climate change intensifies extreme weather events, traditional numerical weather prediction (NWP) struggles to keep pace due to computational limits. This episode explores FourCastNet, a deep learning Earth system emulator that delivers high-resolution, medium-range global forecasts at unprecedented speed—up to five orders of magnitude faster than NWP—while maintaining near state-of-the-art accuracy.📌 Bullet points summaryFourCastNet outpaces traditional NWP with forecasts that are not only faster by several magnitudes but also comparably accurate, thanks to its data-driven deep learning approach.Powered by Adaptive Fourier Neural Operators (AFNO), the model efficiently handles high-resolution data, leveraging spectral convolutions, model/data parallelism, and performance optimizations like CUDA graphs and JIT compilation.Scales excellently across supercomputers such as Selene, Perlmutter, and JUWELS Booster, reaching 140.8 petaFLOPS and enabling rapid training and large-scale ensemble forecasts.Addresses long-standing challenges in weather and climate modeling, including limits in resolution, complexity, and throughput, paving the way for emulating fine-scale Earth system processes.Enables "Interactivity at Scale"—supporting digital Earth twins and empowering users to explore future climate scenarios interactively, aiding science, policy, and public understanding.💡 The Big IdeaFourCastNet revolutionizes weather forecasting by merging the power of deep learning and spectral methods, unlocking interactive, ultra-fast, and high-fidelity Earth system simulations for a changing world.📖 CitationKurth, Thorsten, et al. "Fourcastnet: Accelerating global high-resolution weather forecasting using adaptive fourier neural operators." Proceedings of the Platform for Advanced Scientific Computing Conference. 2023.

🎙️ Episode 22: Knowledge-guided machine learning can improve carbon cycle quantification in agroecosystems🔗 DOI: https://doi.org/10.1038/s41467-023-43860-5🧠 AbstractImproving the accuracy and scalability of carbon cycle quantification in agroecosystems is essential for climate mitigation and sustainable agriculture. This episode discusses a new Knowledge-Guided Machine Learning (KGML) framework that integrates process-based models, high-resolution remote sensing, and machine learning to address key limitations in conventional approaches.📌 Bullet points summaryIntroduces KGML-ag-Carbon, a hybrid model combining process-based simulation (ecosys), remote sensing, and ML to improve carbon cycle modeling in agroecosystems.Outperforms traditional models in capturing spatial and temporal carbon dynamics across the U.S. Corn Belt, especially under data-scarce conditions.Delivers high-resolution (250m daily) estimates for critical carbon metrics such as GPP, Ra, Rh, NEE, and crop yield, with field-level precision.Benefits from pre-training with synthetic data, remote sensing assimilation, and a hierarchical architecture with knowledge-guided loss functions for better accuracy and interpretability.Shows promise for broader applications including nutrient cycle modeling, large-scale carbon assessment, and scenario testing under various management and climate conditions.💡 The Big IdeaKGML-ag-Carbon represents a leap in modeling agroecosystem carbon cycles, blending scientific knowledge with data-driven insights to unlock precision and scalability in climate-smart agriculture.📖 CitationLiu, Licheng, et al. "Knowledge-guided machine learning can improve carbon cycle quantification in agroecosystems." Nature Communications 15.1 (2024): 357.

🎙️ Episode 21 — AtmoRep: A Stochastic Model of Atmospheric Dynamics Using Large-Scale Representation LearningThis week, we explore AtmoRep, a novel task-independent AI model for simulating atmospheric dynamics. Built on large-scale representation learning and trained on ERA5 reanalysis data, AtmoRep delivers strong performance across a variety of tasks—without needing task-specific training.🔍 Highlights from the episode:Introduction to AtmoRep, a stochastic computer model leveraging AI to simulate the atmosphere.Zero-shot capabilities for nowcasting, temporal interpolation, model correction, and generating counterfactuals.Outperforms or matches state-of-the-art models like Pangu-Weather and even ECMWF's IFS at short forecast horizons.Fine-tuning with additional data, like radar observations, enhances performance—especially for precipitation forecasts.Offers a computationally efficient alternative to traditional numerical models, with potential for broader scientific and societal applications.📚 Read the paper: https://doi.org/10.48550/arXiv.2308.13280✍️ Citation:Lessig, Christian, et al. "AtmoRep: A stochastic model of atmosphere dynamics using large scale representation learning." arXiv:2308.13280 (2023)

🎙️ Episode 20: Finding the Right XAI Method—Evaluating Explainable AI in Climate Science🔗 DOI: https://doi.org/10.48550/arXiv.2303.00652🧩 AbstractExplainable AI (XAI) methods are increasingly used in climate science, but the lack of ground truth explanations makes it difficult to evaluate and compare them effectively. This episode dives into a new framework for systematically evaluating XAI methods based on key properties tailored to climate research needs.📌 Bullet points summaryIntroduces XAI evaluation for climate science, offering a structured approach to assess and compare explanation methods using key desirable properties.Identifies five critical properties for XAI in this context: robustness, faithfulness, randomization, complexity, and localization.Evaluation shows that different XAI methods perform differently across these properties, with performance also depending on model architecture.Salience methods often score well on faithfulness and complexity but lower on randomization.Sensitivity methods typically do better on randomization but at the expense of other properties.Proposes a framework to guide method selection: assess the importance of each property for the research task, compute skill scores for available methods, and rank or combine methods accordingly.Highlights the role of benchmark datasets and evaluation metrics in supporting transparent and context-specific XAI adoption in climate science.💡 The Big IdeaThis work empowers climate researchers to make informed, task-specific choices in explainable AI, turning a fragmented XAI landscape into a guided, comparative process rooted in scientific needs.📖 CitationBommer, Philine Lou, et al. "Finding the right XAI method—A guide for the evaluation and ranking of explainable AI methods in climate science." Artificial Intelligence for the Earth Systems 3.3 (2024): e230074.

🎧 Abstract:Weather forecasting is essential for both science and society. This episode explores a breakthrough in medium-range global weather forecasting using artificial intelligence. The researchers introduce Pangu-Weather, an AI-powered system that leverages 3D deep networks with Earth-specific priors and a hierarchical temporal aggregation strategy to significantly enhance forecast accuracy and reduce error accumulation over time.📌 Bullet points summary:Pangu-Weather applies 3D deep learning with Earth-specific priors for accurate medium-range global weather forecasts.It utilizes a hierarchical temporal aggregation strategy to minimize accumulation errors.Outperforms ECMWF’s operational Integrated Forecasting System (IFS) in deterministic forecasting and tropical cyclone tracking.Achieves over 10,000× faster performance than IFS, enabling efficient large-member ensemble forecasts.Though trained on reanalysis data and limited in variable scope, Pangu-Weather presents a promising hybrid approach combining AI and traditional numerical weather prediction (NWP).💡 The Big Idea:AI is reshaping how we predict the weather. With Pangu-Weather, deep learning meets atmospheric science—delivering faster, more accurate forecasts that could redefine the future of meteorology.📚 Citation:Bi, K., Xie, L., Zhang, H. et al. Accurate medium-range global weather forecasting with 3D neural networks. Nature 619, 533–538 (2023). https://doi.org/10.1038/s41586-023-06185-3

🎧 Abstract:In this episode, we dive into GraphDOP, a novel data-driven forecasting system developed by ECMWF. Unlike traditional models, GraphDOP learns directly from Earth System observations—without relying on physics-based reanalysis. By capturing relationships between satellite and conventional observations, it builds a latent representation of Earth’s dynamic systems and delivers accurate weather forecasts up to five days ahead.📌 Bullet points summary:GraphDOP is developed by ECMWF and operates purely on observational data, without physics-based (re)analysis or feedback.Produces skillful forecasts for surface and upper-air parameters up to five days into the future.Competes with ECMWF’s IFS for two-metre temperature (t2m), outperforming it in the Tropics at 5-day lead times.Can generate forecasts at any time and location—even where observational data is sparse—without using gridded ERA5 fields for training.Combines data from various instruments to create accurate joint forecasts of surface and tropospheric temperatures in the Tropics.Learns observation relationships that generalize well to data-sparse regions, with upper-level wind forecasts aligning closely with ERA5 even in low-coverage areas.💡 The Big Idea:GraphDOP reimagines weather forecasting by proving that pure observational data—when paired with intelligent modeling—can rival and even surpass traditional, physics-based systems in both speed and accuracy.📚 Citation:Alexe, Mihai, et al. "GraphDOP: Towards skilful data-driven medium-range weather forecasts learnt and initialised directly from observations." arXiv preprint arXiv:2412.15687 (2024). https://doi.org/10.48550/arXiv.2412.15687

🎧 Abstract:In this episode, we explore DiffDA, a novel data assimilation approach for weather forecasting and climate modeling. Built on the foundations of denoising diffusion models, DiffDA uses the pretrained GraphCast neural network to assimilate atmospheric variables from predicted states and sparse observations—providing a data-driven pathway to generate accurate initial conditions for forecasts.📌 Bullet points summary:Introduces DiffDA, a machine learning-based data assimilation method that leverages predicted states and sparse observations.Utilizes the pretrained GraphCast weather model, repurposed as a denoising diffusion model.Employs a two-phase conditioning strategy: on predicted states (training/inference) and sparse observations (inference only).Capable of generating assimilated global atmospheric data at 0.25° resolution.Demonstrates that initial conditions created via DiffDA retain forecast quality with a lead time degradation of at most 24 hours compared to top-tier assimilation systems.Enables autoregressive reanalysis dataset generation without full observation availability.💡 The Big Idea:DiffDA represents a step forward in data assimilation—merging the strengths of diffusion models and machine learning to produce accurate, observation-consistent initial conditions for future-focused forecasting.📚 Citation:Huang, Langwen, et al. "Diffda: a diffusion model for weather-scale data assimilation." arXiv preprint arXiv:2401.05932 (2024). https://doi.org/10.48550/arXiv.2401.059327

🎙️ Abstract:Embedding physical constraints as inductive priors is key in AI weather forecasting models. Locality—a common prior—relies on local neural interactions like 3D convolutions or attention. ARCHESWEATHER challenges this norm by introducing global vertical interactions, improving efficiency without sacrificing accuracy.📌 Bullet points summary:ARCHESWEATHER is a lightweight, efficient AI model trained at 1.5º resolution with minimal compute (a few GPU-days), offering low-cost inference and strong performance.The Cross-Level Attention (CLA) mechanism enables vertical atmospheric feature interactions, replacing 3D local attention with 2D horizontal attention and column-wise CLA in a 3D Swin U-Net with Earth-specific biases.Ensemble versions (MX4 and LX2) outperform or match IFS HRES and NeuralGCM in RMSE for 1–3 day forecasts on upper-air variables; it gains edge on wind variables at longer lead times.Fine-tuning on post-2007 ERA5 data yields modest gains, pointing to distributional shifts in the dataset.A convolutional head with bilinear upsampling avoids checkerboard artifacts, offering cleaner projections. The code is open-source.💡 Big Idea:ARCHESWEATHER shows that global vertical interactions via cross-level attention can outperform traditional locality-based models, paving a path toward more efficient, physically grounded weather forecasting systems.📚 Citation:Mukkavilli, S. Karthik, et al. "Ai foundation models for weather and climate: Applications, design, and implementation." arXiv preprint arXiv:2309.10808 (2023). DOI: 10.48550/arXiv.2405.14527

🌍 Abstract:Surface-level weather is what matters most to the public—but it's also where feedback loops and complex interactions dominate. Land Surface Models (LSMs) capture these dynamics. Coupled with atmospheric models, they help forecast water, carbon, and energy fluxes. This study explores machine learning emulators as fast, accurate alternatives for ecLand, the ECMWF’s land surface scheme.⚡ Bullet points summary:Three machine learning models—LSTM, XGB, and MLP—were evaluated as statistical emulators for ECLand to enable faster experimentation in land surface forecasting.All models showed strong accuracy, but LSTM excelled in long-range continental forecasts, XGB was robust across tasks, and MLP balanced accuracy and ease of use.Emulators offered significant runtime savings over traditional numerical models, boosting potential for quicker simulations and integration into data assimilation pipelines.Model strengths varied by scale and variable: XGB led in European soil water predictions, MLP scored highest in global accuracy, and LSTM improved snow cover forecasts on a continental scale.The study provides a clear comparison of model trade-offs, helping guide the choice of emulator based on accuracy needs, compute budget, and geographic focus.💡 Big Idea:Machine learning emulators can dramatically speed up land surface forecasting without compromising accuracy—empowering faster, more adaptable weather research and operations.📚 Citation:Wesselkamp, M., et al. Advances in Land Surface Model-based Forecasting: A Comparison of LSTM, Gradient Boosting, and Feedforward Neural Networks as Prognostic State Emulators in a Case Study with ECLand, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2024-2081, 2024.

DOI:https://doi.org/10.48550/arXiv.2411.11268Abstract:Existing machine learning models of weather variability are not formulated to enable assessment of their response to varying external boundary conditions such as sea surface temperature and greenhouse gases. Here we present ACE2 (Ai2 Climate Emulator version 2) and its application to reproducing atmospheric variability over the past 80 years on timescales from days to decades. ACE2 is a 450M-parameter autoregressive machine learning emulator, operating with 6-hour temporal resolution, 1° horizontal resolution and eight vertical layers. It exactly conserves global dry air mass and moisture and can be stepped forward stably for arbitrarily many steps with a throughput of about 1500 simulated years per wall clock day....Citation:Watt-Meyer, Oliver, et al. "ACE2: Accurately learning subseasonal to decadal atmospheric variability and forced responses." arXiv preprint arXiv:2411.11268 (2024).

DOI:https://doi.org/10.48550/arXiv.2405.13063 Abstract:Reliable forecasts of the Earth system are crucial for human progress and safety from natural disasters. Artificial intelligence offers substantial potential to improve prediction accuracy and computational efficiency in this field, however this remains underexplored in many domains. Here we introduce Aurora, a large-scale foundation model for the Earth system trained on over a million hours of diverse data... Citation: Bodnar, Cristian, et al. "Aurora: A foundation model of the atmosphere." arXiv preprint arXiv:2405.13063 (2024).

Abstract: Existing ML-based atmospheric models are not suitable for climate prediction, which requires long-term stability and physical consistency. We present ACE (AI2 Climate Emulator), a 200M-parameter, autoregressive machine learning emulator of an existing comprehensive 100-km resolution global atmospheric model. The formulation of ACE allows evaluation of physical laws such as the conservation of mass and moisture... Citation: Watt-Meyer, Oliver, et al. "ACE: A fast, skillful learned global atmospheric model for climate prediction." arXiv preprint arXiv:2310.02074 (2023). DOI:https://doi.org/10.48550/arXiv.2310.02074

DOI:https://doi.org/10.48550/arXiv.2308.15560Abstract: WeatherBench 2 is an update to the global, medium-range (1-14 day) weather forecasting benchmark proposed by Rasp et al. (2020), designed with the aim to accelerate progress in data-driven weather modeling. WeatherBench 2 consists of an open-source evaluation framework, publicly available training, ground truth and baseline data as well as a continuously updated website with the latest metrics and state-of-the-art model... Citation: Rasp, Stephan, et al. "WeatherBench 2: A benchmark for the next generation of data‐driven global weather models." Journal of Advances in Modeling Earth Systems 16.6 (2024): e2023MS004019.

Abstract: Ensemble forecasting is crucial for improving weather predictions, especially for forecasts of extreme events. Constructing an ensemble prediction system (EPS) based on conventional NWP models is highly computationally expensive. ML models have emerged as valuable tools for deterministic weather forecasts, providing forecasts with significantly reduced computational requirements and even surpassing the forecast performance of traditional NWP models. However, challenges arise when applying ML models to ensemble forecasting. Recent ML models, such as GenCast and SEEDS model, rely on the ERA5 EDA or operational NWP ensemble members for forecast generation. Their spatial resolution is also considered too coarse for many applications. To overcome these limitations, we introduce FuXi-ENS, an advanced ML model designed to deliver 6-hourly global ensemble weather forecasts up to 15 days. This model runs at a significantly increased spatial resolution of 0.25\textdegree, incorporating 5 atmospheric variables at 13 pressure levels, along with 13 surface variables. By leveraging the inherent probabilistic nature of Variational AutoEncoder (VAE), FuXi-ENS optimizes a loss function that combines the CRPS and the KL divergence between the predicted and target distribution, facilitating the incorporation of flow-dependent perturbations in both initial conditions and forecast. This innovative approach makes FuXi-ENS an advancement over the traditional ones that use L1 loss combined with the KL loss in standard VAE models for ensemble weather forecasting. Results demonstrate that FuXi-ENS outperforms ensemble forecasts from the ECMWF, a world leading NWP model, in the CRPS of 98.1% of 360 variable and forecast lead time combinations. This achievement underscores the potential of the FuXi-ENS model to enhance ensemble weather forecasts, offering a promising direction for further development in this field. Zhong, Xiaohui, et al. "FuXi-ENS: A machine learning model for medium-range ensemble weather forecasting." arXiv preprint arXiv:2405.05925 (2024). arXiv:2405.05925

Abstract : Fourier Neural Operators (FNOs) have proven to be an efficient and effective method for resolution-independent operator learning in a broad variety of application areas across scientific machine learning. A key reason for their success is their ability to accurately model long-range dependencies in spatio-temporal data by learning global convolutions in a computationally efficient manner. To this end, FNOs rely on the discrete Fourier transform (DFT), however, DFTs cause visual and spectral artifacts as well as pronounced dissipation when learning operators in spherical coordinates since they incorrectly assume a flat geometry. To overcome this limitation, we generalize FNOs on the sphere, introducing Spherical FNOs (SFNOs) for learning operators on spherical geometries. We apply SFNOs to forecasting atmospheric dynamics, and demonstrate stable auto\-regressive rollouts for a year of simulated time (1,460 steps), while retaining physically plausible dynamics. The SFNO has important implications for machine learning-based simulation of climate dynamics that could eventually help accelerate our response to climate change. Citation: Bonev, Boris, et al. "Spherical fourier neural operators: Learning stable dynamics on the sphere." International conference on machine learning. PMLR, 2023 DOI: https://doi.org/10.48550/arXiv.2306.03838

Abstract: Artificial intelligence (AI) can be used to improve performance across a wide range of Earth system prediction tasks. As with any application of AI, it is important for AI to be developed in an ethical and responsible manner to minimize bias and other effects. In this work, we extend our previous work demonstrating how AI can go wrong with weather and climate applications by presenting a categorization of bias for AI in the Earth sciences. This categorization can assist AI developers to identify potential biases that can affect their model throughout the AI development life cycle. We highlight examples from a variety of Earth system prediction tasks of each category of bias.Citation: McGovern, Amy, et al. "Identifying and Categorizing Bias in AI/ML for Earth Sciences." Bulletin of the American Meteorological Society 105.3 (2024): E567-E583.DOI: https://doi.org/10.1175/BAMS-D-23-0196.1

Abstract: Weather forecasting is critical for a range of human activities including transportation, agriculture, industry, as well as the safety of the general public. Machine learning models have the potential to transform the complex weather prediction pipeline, but current approaches still rely on numerical weather prediction (NWP) systems, limiting forecast speed and accuracy. Here we demonstrate that a machine learning model can replace the entire operational NWP pipeline. Aardvark Weather, an end-to-end data-driven weather prediction system, ingests raw observations and outputs global gridded forecasts and local station forecasts. Further, it can be optimised end-to-end to maximise performance over quantities of interest. Global forecasts outperform an operational NWP baseline for multiple variables and lead times. Local station forecasts are skillful up to ten days lead time and achieve comparable and often lower errors than a post-processed global NWP baseline and a state-of-the-art end-to-end forecasting system with input from human forecasters. These forecasts are produced with a remarkably simple neural process model using just 8% of the input data and three orders of magnitude less compute than existing NWP and hybrid AI-NWP methods. We anticipate that Aardvark Weather will be the starting point for a new generation of end-to-end machine learning models for medium-range forecasting that will reduce computational costs by orders of magnitude and enable the rapid and cheap creation of bespoke models for users in a variety of fields, including for the developing world where state-of-the-art local models are not currently available.Citation: Vaughan, Anna, et al. "Aardvark weather: end-to-end data-driven weather forecasting." arXiv preprint arXiv:2404.00411 (2024).DOI: https://doi.org/10.48550/arXiv.2404.00411

Abstract: Triggered by the realization that AI emulators can rival the performance of traditional numerical weather prediction models running on HPC systems, there is now an increasing number of large AI models that address use cases such as forecasting, downscaling, or nowcasting. While the parallel developments in the AI literature focus on foundation models -- models that can be effectively tuned to address multiple, different use cases -- the developments on the weather and climate side largely focus on single-use cases with particular emphasis on mid-range forecasting. We close this gap by introducing Prithvi WxC, a 2.3 billion parameter foundation model developed using 160 variables from the Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2). Prithvi WxC employs an encoder-decoder-based architecture, incorporating concepts from various recent transformer models to effectively capture both regional and global dependencies in the input data. The model has been designed to accommodate large token counts to model weather phenomena in different topologies at fine resolutions. Furthermore, it is trained with a mixed objective that combines the paradigms of masked reconstruction with forecasting. We test the model on a set of challenging downstream tasks namely: Autoregressive rollout forecasting, Downscaling, Gravity wave flux parameterization, and Extreme events estimation. The pretrained model with 2.3 billion parameters, along with the associated fine-tuning workflows, has been publicly released as an open-source contribution via Hugging Face.Citation: Schmude, Johannes, et al. "Prithvi wxc: Foundation model for weather and climate." arXiv preprint arXiv:2409.13598 (2024).DOI:https://doi.org/10.48550/arXiv.2409.13598

Abstract: General circulation models (GCMs) are the foundation of weather and climate prediction1,2. GCMs are physics-based simulators that combine a numerical solver for large-scale dynamics with tuned representations for small-scale processes such as cloud formation. Recently, machine-learning models trained on reanalysis data have achieved comparable or better skill than GCMs for deterministic weather forecasting3,4. However, these models have not demonstrated improved ensemble forecasts, or shown sufficient stability for long-term weather and climate simulations. Here we present a GCM that combines a differentiable solver for atmospheric dynamics with machine-learning components and show that it can generate forecasts of deterministic weather, ensemble weather and climate on par with the best machine-learning and physics-based methods. NeuralGCM is competitive with machine-learning models for one- to ten-day forecasts, and with the European Centre for Medium-Range Weather Forecasts ensemble prediction for one- to fifteen-day forecasts. With prescribed sea surface temperature, NeuralGCM can accurately track climate metrics for multiple decades, and climate forecasts with 140-kilometre resolution show emergent phenomena such as realistic frequency and trajectories of tropical cyclones. For both weather and climate, our approach offers orders of magnitude computational savings over conventional GCMs, although our model does not extrapolate to substantially different future climates. Our results show that end-to-end deep learning is compatible with tasks performed by conventional GCMs and can enhance the large-scale physical simulations that are essential for understanding and predicting the Earth system.Citation: Kochkov, D., Yuval, J., Langmore, I. et al. Neural general circulation models for weather and climate. Nature 632, 1060–1066 (2024). https://doi.org/10.1038/s41586-024-07744-y.DOI:https://doi.org/10.1038/s41586-024-07744-y

Abstract: Nonlinear dynamical systems exposed to changing forcing values can exhibit catastrophic transitions between distinct states. The phenomenon of critical slowing down can help anticipate such transitions if caused by a bifurcation and if the change in forcing is slow compared with the system’s internal timescale. However, in many real-world situations, these assumptions are not met and transitions can be triggered because the forcing exceeds a critical rate. For instance, the rapid pace of anthropogenic climate change compared with the internal timescales of key Earth system components, like polar ice sheets or the Atlantic Meridional Overturning Circulation, poses significant risk of rate-induced tipping. Moreover, random perturbations may cause some trajectories to cross an unstable boundary whereas others do not—even under the same forcing. Critical-slowing-down-based indicators generally cannot distinguish these cases of noise-induced tipping from no tipping. This severely limits our ability to assess the tipping risks and to predict individual trajectories. To address this, we make the first attempt to develop a deep learning framework predicting the transition probabilities of dynamical systems ahead of rate-induced transitions. Our method issues early warnings, as demonstrated on three prototypical systems for rate-induced tipping subjected to time-varying equilibrium drift and noise perturbations. Exploiting explainable artificial intelligence methods, our framework captures the fingerprints for the early detection of rate-induced tipping, even with long lead times. Our findings demonstrate the predictability of rate-induced and noise-induced tipping, advancing our ability to determine safe operating spaces for a broader class of dynamical systems than possible so far. Citation: Huang, Y., Bathiany, S., Ashwin, P. et al. Deep learning for predicting rate-induced tipping. Nat Mach Intell 6, 1556–1565 (2024). https://doi.org/10.1038/s42256-024-00937-0. DOI:https://doi.org/10.1038/s42256-024-00937-0

Abstract: Machine learning-based weather forecasting models have quickly emerged as a promising methodology for accurate medium-range global weather forecasting. Here, we introduce the Artificial Intelligence Forecasting System (AIFS), a data driven forecast model developed by the European Centre for Medium-Range Weather Forecasts (ECMWF). AIFS is based on a graph neural network (GNN) encoder and decoder, and a sliding window transformer processor, and is trained on ECMWF's ERA5 re-analysis and ECMWF's operational numerical weather prediction (NWP) analyses. It has a flexible and modular design and supports several levels of parallelism to enable training on high-resolution input data. AIFS forecast skill is assessed by comparing its forecasts to NWP analyses and direct observational data. We show that AIFS produces highly skilled forecasts for upper-air variables, surface weather parameters and tropical cyclone tracks. AIFS is run four times daily alongside ECMWF's physics-based NWP model and forecasts are available to the public under ECMWF's open data policy.Citation: Lang, Simon, et al. "AIFS-ECMWF's data-driven forecasting system." arXiv preprint arXiv:2406.01465 (2024).DOI:https://doi.org/10.48550/arXiv.2406.01465