arXiv subjects

Anna Kwa

Publications and source records attributed to Anna Kwa.

16 recordsLinked to original sources

arXiv2026

Disentangling the effects of sea surface temperature and CO$_2$ in global machine learned weather-climate emulators

While previous versions of the Ai2 Climate Emulator (ACE) have been trained with CO$_2$ as a forcing, they are only accurate within a narrow range of scenarios, for example climate over the last 80 years forced by observed sea surface temperature (SST), sea ice, and CO$_2$ (AMIP), or equilibrium or near-equilibrium climates with CO$_2$ concentrations ranging from 1x to 4x that of the present day. Attempting to simulate climate forced by AMIP SST perturbed by +4 K or the response to an abrupt quadrupling of CO$_2$, results in unphysical behavior. We attribute this to these models being trained on datasets where the SST and CO$_2$ are correlated, limiting their ability to accurately learn their separate effects. In this study we introduce a new class of "random-CO$_2$" reference simulations where the SST and CO$_2$ are prescribed to vary independently. Trained on a balance of AMIP, equilibrium-climate, and random-CO$_2$ data, and including a total energy conservation constraint for improved interpretability, we present a more data-efficient model that not only accurately emulates its reference model in scenarios in which previous models excelled, but also scenarios like AMIP +4 K and slab-ocean-coupled abrupt 4xCO$_2$ where they did not. Limitations are that it has simplified or prescribed representations of other Earth system components like the ocean, land, and sea ice; does not expose other known climate drivers as forcings; and relies solely on physics-based model output for training data, inheriting the biases relative to observations thereof. Each of these represent opportunities for future work.

physics.ao-ph↗

arXiv2025

HiRO-ACE: Fast and skillful AI emulation and downscaling trained on a 3 km global storm-resolving model

Kilometer-scale simulations of the atmosphere are an important tool for assessing local weather extremes and climate impacts, but computational expense limits their use to small regions, short periods, and limited ensembles. Machine learning offers a pathway to efficiently emulate these high-resolution simulations. Here we introduce HiRO-ACE, a two-stage AI modeling framework combining a stochastic version of the Ai2 Climate Emulator (ACE2S) with diffusion-based downscaling (HiRO) to generate 3 km precipitation fields over arbitrary regions of the globe. Both components are trained on data derived from a decade of atmospheric simulation by X-SHiELD, a 3 km global storm-resolving model. HiRO performs a 32x downscaling--generating 3 km 6-hourly precipitation from coarse 100 km inputs by training on paired high-resolution and coarsened X-SHiELD outputs. ACE2S is a $1^\circ \times 1^\circ$ ($\sim$100 km) stochastic autoregressive global atmosphere emulator that maintains grid-scale precipitation variability consistent with coarsened X-SHiELD, enabling its outputs to be ingested by HiRO without additional tuning. HiRO-ACE reproduces the distribution of extreme precipitation rates through the 99.99th percentile, with time-mean precipitation biases below 10% almost everywhere. The framework generates plausible tropical cyclones, fronts, and convective events from poorly resolved coarse inputs. Its computational efficiency allows generation of 6-hourly high-resolution regional precipitation for decades of simulated climate within a single day using one H100 GPU, while the probabilistic design enables ensemble generation for quantifying uncertainty. This establishes an AI-enabled pathway for affordably leveraging the realism of expensive km-scale simulations to support local climate adaptation planning and extreme event risk assessment.

physics.ao-ph↗

arXiv2025

SamudrACE: Fast and Accurate Coupled Climate Modeling with 3D Ocean and Atmosphere Emulators

Traditional numerical global climate models simulate the full Earth system by exchanging boundary conditions between separate simulators of the atmosphere, ocean, sea ice, land surface, and other geophysical processes. This paradigm allows for distributed development of individual components within a common framework, unified by a coupler that handles translation between realms via spatial or temporal alignment and flux exchange. Following a similar approach adapted for machine learning-based emulators, we present SamudrACE: a coupled global climate model emulator which produces centuries-long simulations at 1-degree horizontal, 6-hourly atmospheric, and 5-daily oceanic resolution, with 145 2D fields spanning 8 atmospheric and 19 oceanic vertical levels, plus sea ice, surface, and top-of-atmosphere variables. SamudrACE is highly stable and has low climate biases comparable to those of its components with prescribed boundary forcing, with realistic variability in coupled climate phenomena such as ENSO that is not possible to simulate in uncoupled mode.

physics.ao-ph↗

arXiv2024

ACE2-SOM: Coupling an ML atmospheric emulator to a slab ocean and learning the sensitivity of climate to changed CO$_2$

While autoregressive machine-learning-based emulators have been trained to produce stable and accurate rollouts in the climate of the present-day and recent past, none so far have been trained to emulate the sensitivity of climate to substantial changes in CO$_2$ or other greenhouse gases. As an initial step we couple the Ai2 Climate Emulator version 2 to a slab ocean model (hereafter ACE2-SOM) and train it on output from a collection of equilibrium-climate physics-based reference simulations with varying levels of CO$_2$. We test it in equilibrium and non-equilibrium climate scenarios with CO$_2$ concentrations seen and unseen in training. ACE2-SOM performs well in equilibrium-climate inference with both in-sample and out-of-sample CO$_2$ concentrations, accurately reproducing the emergent time-mean spatial patterns of surface temperature and precipitation change with CO$_2$ doubling, tripling, or quadrupling. In addition, the vertical profile of atmospheric warming and change in extreme precipitation rates up to the 99.9999th percentile closely agree with the reference model. Non-equilibrium-climate inference is more challenging. With CO$_2$ increasing gradually at a rate of 2% year$^{-1}$, ACE2-SOM can accurately emulate the global annual mean trends of surface and lower-to-middle atmosphere fields but produces unphysical jumps in stratospheric fields. With an abrupt quadrupling of CO$_2$, ML-controlled fields transition unrealistically quickly to the 4xCO$_2$ regime. In doing so they violate global energy conservation and exhibit unphysical sensitivities of and surface and top of atmosphere radiative fluxes to instantaneous changes in CO$_2$. Future emulator development needed to address these issues should improve its generalizability to diverse climate change scenarios.

physics.ao-ph↗

arXiv2024

ACE2: Accurately learning subseasonal to decadal atmospheric variability and forced responses

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{\deg} 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. ACE2 generates emergent phenomena such as tropical cyclones, the Madden Julian Oscillation, and sudden stratospheric warmings. Furthermore, it accurately reproduces the atmospheric response to El Ni\~no variability and global trends of temperature over the past 80 years. However, its sensitivities to separately changing sea surface temperature and carbon dioxide are not entirely realistic.

physics.ao-ph↗

arXiv2022

Improving the predictions of ML-corrected climate models with novelty detection

While previous works have shown that machine learning (ML) can improve the prediction accuracy of coarse-grid climate models, these ML-augmented methods are more vulnerable to irregular inputs than the traditional physics-based models they rely on. Because ML-predicted corrections feed back into the climate model's base physics, the ML-corrected model regularly produces out of sample data, which can cause model instability and frequent crashes. This work shows that adding semi-supervised novelty detection to identify out-of-sample data and disable the ML-correction accordingly stabilizes simulations and sharply improves the quality of predictions. We design an augmented climate model with a one-class support vector machine (OCSVM) novelty detector that provides better temperature and precipitation forecasts in a year-long simulation than either a baseline (no-ML) or a standard ML-corrected run. By improving the accuracy of coarse-grid climate models, this work helps make accurate climate models accessible to researchers without massive computational resources.

physics.ao-ph↗

arXiv2022

Machine-learned climate model corrections from a global storm-resolving model

Due to computational constraints, running global climate models (GCMs) for many years requires a lower spatial grid resolution (${\gtrsim}50$ km) than is optimal for accurately resolving important physical processes. Such processes are approximated in GCMs via subgrid parameterizations, which contribute significantly to the uncertainty in GCM predictions. One approach to improving the accuracy of a coarse-grid global climate model is to add machine-learned state-dependent corrections at each simulation timestep, such that the climate model evolves more like a high-resolution global storm-resolving model (GSRM). We train neural networks to learn the state-dependent temperature, humidity, and radiative flux corrections needed to nudge a 200 km coarse-grid climate model to the evolution of a 3~km fine-grid GSRM. When these corrective ML models are coupled to a year-long coarse-grid climate simulation, the time-mean spatial pattern errors are reduced by 6-25% for land surface temperature and 9-25% for land surface precipitation with respect to a no-ML baseline simulation. The ML-corrected simulations develop other biases in climate and circulation that differ from, but have comparable amplitude to, the baseline simulation.

physics.ao-ph↗

arXiv2022

Emulating Fast Processes in Climate Models

Cloud microphysical parameterizations in atmospheric models describe the formation and evolution of clouds and precipitation, a central weather and climate process. Cloud-associated latent heating is a primary driver of large and small-scale circulations throughout the global atmosphere, and clouds have important interactions with atmospheric radiation. Clouds are ubiquitous, diverse, and can change rapidly. In this work, we build the first emulator of an entire cloud microphysical parameterization, including fast phase changes. The emulator performs well in offline and online (i.e. when coupled to the rest of the atmospheric model) tests, but shows some developing biases in Antarctica. Sensitivity tests demonstrate that these successes require careful modeling of the mixed discrete-continuous output as well as the input-output structure of the underlying code and physical process.

physics.ao-ph↗

arXiv2020

Machine Learning Climate Model Dynamics: Offline versus Online Performance

Climate models are complicated software systems that approximate atmospheric and oceanic fluid mechanics at a coarse spatial resolution. Typical climate forecasts only explicitly resolve processes larger than 100 km and approximate any process occurring below this scale (e.g. thunderstorms) using so-called parametrizations. Machine learning could improve upon the accuracy of some traditional physical parametrizations by learning from so-called global cloud-resolving models. We compare the performance of two machine learning models, random forests (RF) and neural networks (NNs), at parametrizing the aggregate effect of moist physics in a 3 km resolution global simulation with an atmospheric model. The NN outperforms the RF when evaluated offline on a testing dataset. However, when the ML models are coupled to an atmospheric model run at 200 km resolution, the NN-assisted simulation crashes with 7 days, while the RF-assisted simulations remain stable. Both runs produce more accurate weather forecasts than a baseline configuration, but globally averaged climate variables drift over longer timescales.

physics.ao-ph↗

arXiv2018

Reconciling the Diversity and Uniformity of Galactic Rotation Curves with Self-Interacting Dark Matter

Galactic rotation curves exhibit diverse behavior in the inner regions, while obeying an organizing principle, i.e., they can be approximately described by a radial acceleration relation or the Modified Newtonian Dynamics phenomenology. We analyze the rotation curve data from the SPARC sample, and explicitly demonstrate that both the diversity and uniformity are naturally reproduced in a hierarchical structure formation model with the addition of dark matter self-interactions. The required concentrations of the dark matter halos are fully consistent with the concentration-mass relation predicted by the Planck cosmological model. The inferred stellar mass-to-light ($3.6 μm$) ratios scatter around $0.5 M_\odot/L_\odot$, as expected from population synthesis models, leading to a tight radial acceleration relation and baryonic Tully-Fisher relation. The inferred stellar-halo mass relation is consistent with the expectations from abundance matching. These results indicate that the inner dark matter halos of galaxies are thermalized due to the self-interactions of dark matter particles.

astro-ph.GA↗

arXiv2017

What the Milky Way's Dwarfs tell us about the Galactic Center extended excess

The Milky Way's Galactic Center harbors a gamma-ray excess that is a candidate signal of annihilating dark matter. Dwarf galaxies remain predominantly dark in their expected commensurate emission. In this work we quantify the degree of consistency between these two observations through a joint likelihood analysis. In doing so we incorporate Milky Way dark matter halo profile uncertainties, as well as an accounting of diffuse gamma-ray emission uncertainties in dark matter annihilation models for the Galactic Center Extended gamma-ray excess (GCE) detected by the Fermi Gamma-Ray Space Telescope. The preferred range of annihilation rates and masses expands when including these unknowns. Even so, using two recent determinations of the Milky Way halo's local density leave the GCE preferred region of single-channel dark matter annihilation models to be in strong tension with annihilation searches in combined dwarf galaxy analyses. A third, higher Milky Way density determination, alleviates this tension. Our joint likelihood analysis allows us to quantify this inconsistency. We provide a set of tools for testing dark matter annihilation models' consistency within this combined dataset. As an example, we test a representative inverse Compton sourced self-interacting dark matter model, which is consistent with both the GCE and dwarfs.

astro-ph.HE↗

arXiv2017

The Galactic Isotropic $γ$-ray Background and Implications for Dark Matter

We present an analysis of the radial angular profile of the galacto-isotropic (GI) $γ$-ray flux--the statistically uniform flux in circular annuli about the Galactic center. Two different approaches are used to measure the GI flux profile in 85 months of Fermi-LAT data: the BDS statistic method which identifies spatial correlations, and a new Poisson ordered-pixel method which identifies non-Poisson contributions. Both methods produce similar GI flux profiles. The GI flux profile is well-described by an existing model of bremsstrahlung, $π^0$ production, inverse Compton scattering, and the isotropic background. Discrepancies with data in our full-sky model are not present in the GI component, and are therefore due to mis-modeling of the non-GI emission. Dark matter annihilation constraints based solely on the observed GI profile are close to the thermal WIMP cross section below 100 GeV, for fixed models of the dark matter density profile and astrophysical $γ$-ray foregrounds. Refined measurements of the GI profile are expected to improve these constraints by a factor of a few.

astro-ph.HE↗

arXiv2017

Hidden Sector Hydrogen as Dark Matter: Small-scale Structure Formation Predictions and the Importance of Hyperfine Interactions

We study the atomic physics and the astrophysical implications of a model in which the dark matter is the analog of hydrogen in a secluded sector. The self interactions between dark matter particles include both elastic scatterings as well as inelastic processes due to a hyperfine transition. The self-interaction cross sections are computed by numerically solving the coupled Schrödinger equations for this system. We show that these self interactions exhibit the right velocity dependence to explain the low dark matter density cores seen in small galaxies while being consistent with all constraints from observations of clusters of galaxies. For a viable solution, the dark hydrogen mass has to be in 10--100 GeV range and the dark fine-structure constant has to be larger than 0.02. Precisely for this range of parameters, we show that significant cooling losses may occur due to inelastic excitations to the hyperfine state and subsequent decays, with implications for the evolution of low-mass halos and the early growth of supermassive black holes. Cooling from excitations to higher $n$ levels of dark hydrogen and subsequent decays is possible at the cluster scale, with a strong dependence on halo mass. Finally, we show that the minimum halo mass is in the range of $10^{3.5}$ to $10^7 M_\odot$ for the viable regions of parameter space, significantly larger than the typical predictions for weakly-interacting dark matter models. This pattern of observables in cosmological structure formation is unique to this model, making it possible to rule in or rule out hidden sector hydrogen as a viable dark matter model.

hep-ph↗

arXiv2016

Investigating the Uniformity of the Excess Gamma rays towards the Galactic Center Region

We perform a composite likelihood analysis of subdivided regions within the central $26^\circ\times20^\circ$ of the Milky Way, with the aim of characterizing the spectrum of the gamma-ray galactic center excess in regions of varying galactocentric distance. Outside of the innermost few degrees, we find that the radial profile of the excess is background-model dependent and poorly constrained. The spectrum of the excess emission is observed to extend upwards of 10 GeV outside $\sim5^\circ$ in radius, but cuts off steeply between 10--20 GeV only in the innermost few degrees. If interpreted as a real feature of the excess, this radial variation in the spectrum has important implications for both astrophysical and dark matter interpretations of the galactic center excess. Single-component dark matter annihilation models face challenges in reproducing this variation; on the other hand, a population of unresolved millisecond pulsars contributing both prompt and secondary inverse Compton emission may be able to explain the spectrum as well as its spatial dependency. We show that the expected differences in the photon-count distributions of a smooth dark matter annihilation signal and an unresolved point source population are an order of magnitude smaller than the fluctuations in residuals after fitting the data, which implies that mismodeling is an important systematic effect in point source analyses aimed at resolving the gamma-ray excess.

astro-ph.HE↗

arXiv2016

Lepton-Flavor Violating Mediators

We present a framework where dark matter interacts with the Standard Model through a light, spin-0 mediator that couples chirally to pairs of different-flavor leptons. This flavor violating final state weakens bounds on new physics coupled to leptons from terrestrial experiments and cosmic-ray measurements. As an example, we apply this framework to construct a model for the Fermi-LAT excess of GeV $γ$-rays from the galactic center. We comment on the viability of this portal for self-interacting dark matter explanations of small scale structure anomalies and embeddings in flavor models. Models of this type are shown to be compatible with the muon anomalous magnetic moment anomaly. We review current experimental constraints and identify possible future theoretical and experimental directions.

hep-ph↗

arXiv2015

Discovery of a New Galactic Center Excess Consistent with Upscattered Starlight

We present a new extended gamma ray excess detected with the Fermi Satellite Large Area Telescope toward the Galactic Center that traces the morphology of infrared starlight emission. Combined with its measured spectrum, this new extended source is approximately consistent with inverse Compton emission from a high-energy electron-positron population with energies up to about 10 GeV. Previously detected emissions tracing the 20 cm radio, interpreted as bremsstrahlung radiation, and the Galactic Center Extended emission tracing a spherical distribution and peaking at 2 GeV, are also detected. We show that the inverse Compton and bremsstrahlung emissions are likely due to the same source of electrons and positrons. All three extended emissions may be explained within the framework of a model where the dark matter annihilates to leptons or a model with unresolved millisecond pulsars in the Galactic Center.

astro-ph.HE↗