arXiv Science⌕ Search
Home/Archive/arXiv

arXiv · 2610.17693

Extreme events and statistics in an AI emulator of a global storm-resolving model

Anna Kwa·W. Andre Perkins·Troy Arcomano·Noah D. Brenowitz·Christopher S. Bretherton·Spencer K. Clark·Lucas M. Harris·Jeremy McGibbon·Michael S. Pritchard·Oliver Watt-Meyer
View original source ↗Read source document ↗Export metadata ↓

Abstract

We present HiRO-ACEv2 (High Resolution Output Ai2 Climate Emulator), a two-stage framework for AI emulation of the 3~km resolution X-SHiELD global storm-resolving model (GSRM), and highlight its ability to generate large ensembles of data for studying the characteristics of tropical cyclones and extreme events in high-resolution simulations at an inference computational cost over 300 times lower than that of traditional numerical GSRMs. HiRO-ACEv2 expands upon the precipitation-only HiRO-ACEv1 model to generate coherent precipitation, 10~m winds, and sea level pressure outputs, enabling a broader range of events to be simulated at high-resolution. Fine-resolution predictions are generated by running an autoregressive emulator of coarsened X-SHiELD at 100 km, whose multivariate outputs are regionally downscaled to 3~km resolution via a separately trained diffusion model. Particular importance is placed on capturing the statistics of extreme events during the diffusion downscaling step. Our ability to match the tails of distributions and generate realistic-looking extreme events in both the `perfect prediction' case as well as in free-running emulator rollouts is improved via the use of multiple denoisers at inference time and several updates to the data sampling, noise sampling, and loss weighting during training. HiRO-ACEv2 closely matches the high-resolution reference distributions of precipitation and near-surface winds out to their 99.9999th percentile values, and broadly tracks the reference distribution of tropical cyclone intensities.

Explore related subjects

physics.ao-ph

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Anna Kwa, W. Andre Perkins, Troy Arcomano, Noah D. Brenowitz, Christopher S. Bretherton, Spencer K. Clark, Lucas M. Harris, Jeremy McGibbon, Michael S. Pritchard, Oliver Watt-Meyer. 2026-10-13. Extreme events and statistics in an AI emulator of a global storm-resolving model. https://arxiv.org/abs/2610.17693

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

The hypervariational method: a general, automated technique based on innovations to tune the parameters of the Hybrid covariance matrix

The Hybrid forecast covariance model is an operationally successful way to build a background error covariance matrix. It is constructed as a weighted sum, with scalar weights, of the climatological covariance matrix and a localized ensemble covariance matrix, with a localization matrix typically depending on a localization length. For best performance, the Hybrid needs to be optimized by tuning those weights and, possibly, other parameters, such as the localization length. Here, a systematic and automated method for this tuning operation is introduced. This technique is referred to as hypervariational method, and it is formulated in terms of a cost function, built using a time series of innovations as an essential ingredient. This cost function, which depends on the parameters, is minimized when they have optimal values. Equations for its minimization, based on Newton's method, are derived. Besides the aforementioned, commonly used Hybrid, the hypervariational method is applicable to any known Hybrid (and to any set of parameters on which it depends analytically), including recently developed Hybrids, not necessarily constructed as weighted sums. Here, the hypervariational method is applied to tune the ensemble covariance matrix weight in the seminal-formulation Hybrid (which does not feature localization). This is done in a synthetic experimental setting where a theoretically-derived expression for the optimal value for that weight is known. Moreover, in that experimental setting, a further method, based on a linear fit procedure, to determine that optimal value is possible. Both the hypervariational method and that linear fit procedure are shown to produce an optimal weight value matching (up to a tolerance of $0.1$) the optimal value known from the theory. Furthermore, Newton's minimization in the hypervariational method is shown to converge immediately (at the first recursive step).

physics.ao-ph↗
arXiv2026

Linking Horizontal and Vertical Submesoscale Lagrangian Dynamics South of the Ibiza Channel

Fronts, eddies, jets, and meanders populate the ocean surface and play a key role in organizing transport and mixing across a wide range of scales. From a Lagrangian perspective, frontal structures are conveniently detected as surface filaments that act as barriers to horizontal transport. However, the degree to which these structures maintain coherence below the surface and the relationship between horizontal and vertical motions in their vicinity remain poorly understood. Here we analyze two such filaments in the western Mediterranean Sea, south of the Ibiza Channel, using a high-resolution ocean simulation and Lagrangian diagnostics specifically designed to characterize submesoscale transport. Backward Finite-Size Lyapunov Exponents (FSLE) computed at successive depths reveal nearly vertical, curtain-like structures extending several tens of meters into the water column, with intensity decreasing with depth. Vertical-velocity sections across the filaments show coherent downward motion aligned with the filament cores, while Lagrangian trajectories indicate rapid horizontal convergence of fluid particles accompanied by sustained downward displacement. The barrier effect of Lagrangian Coherent Structures (LCSs) is commonly associated with persistent dynamical features of large-scale flows. Here, however, we demonstrate that FSLE filaments generated by rapidly fluctuating submesoscale dynamics can also act as effective transport barriers. The results demonstrate that the surface signatures of Lagrangian filaments identified through FSLE techniques correspond to three-dimensional transport structures that regulate both lateral aggregation and vertical connectivity in the upper ocean.

physics.ao-ph↗