arXiv subjects

Matthew Willson

Publications and source records attributed to Matthew Willson.

14 recordsLinked to original sources

arXiv2026

WeatherNext 3: Increasing resolution and performance of global weather models with raw observations

State-of-the-art AI weather models have shown impressive medium-range forecast skill and computational efficiency, but suffer two key shortcomings: their forecasts have lower spatial and temporal resolution than the best physics-based models and they are exclusively initialized with and trained on analysis data. As a result, they cannot directly make use of observations, and any biases in the analysis are inherited by the forecast. WeatherNext 3 addresses these shortcomings and establishes a new state-of-the-art for probabilistic medium-range forecasting skill. First, WeatherNext 3 generates new forecasts every hour (rather than every 6 hours like traditional global models) by ingesting low-latency geostationary satellite data. Second, WeatherNext 3's temporal and spatial resolution are on par with physics-based global models, with hourly time steps and 0.1 degree resolution for single-level variables, including solar radiation and cloud cover. Third, WeatherNext 3 moves beyond traditional analysis variables by learning to predict satellite-derived precipitation estimates, as well as tropical cyclone and station observations. Modelling sparse station data allows WeatherNext 3 to make 2m temperature and dewpoint predictions at any location and time, conditioned on local geographical features, with substantially lower error than competing global models, even when evaluated against unseen stations. Together, WeatherNext 3's capabilities move operational AI-based weather forecasting beyond emulating the traditionally distinct stages of data assimilation, forecasting and post-processing, which helps to further push the frontier of performance and granularity for global weather prediction.

cs.LG↗

arXiv2025

Skillful joint probabilistic weather forecasting from marginals

Machine learning (ML)-based weather models have rapidly risen to prominence due to their greater accuracy and speed than traditional forecasts based on numerical weather prediction (NWP), recently outperforming traditional ensembles in global probabilistic weather forecasting. This paper presents FGN, a simple, scalable and flexible modeling approach which significantly outperforms the current state-of-the-art models. FGN generates ensembles via learned model-perturbations with an ensemble of appropriately constrained models. It is trained directly to minimize the continuous rank probability score (CRPS) of per-location forecasts. It produces state-of-the-art ensemble forecasts as measured by a range of deterministic and probabilistic metrics, makes skillful ensemble tropical cyclone track predictions, and captures joint spatial structure despite being trained only on marginals.

cs.LG↗

arXiv2023

GenCast: Diffusion-based ensemble forecasting for medium-range weather

Weather forecasts are fundamentally uncertain, so predicting the range of probable weather scenarios is crucial for important decisions, from warning the public about hazardous weather, to planning renewable energy use. Here, we introduce GenCast, a probabilistic weather model with greater skill and speed than the top operational medium-range weather forecast in the world, the European Centre for Medium-Range Forecasts (ECMWF)'s ensemble forecast, ENS. Unlike traditional approaches, which are based on numerical weather prediction (NWP), GenCast is a machine learning weather prediction (MLWP) method, trained on decades of reanalysis data. GenCast generates an ensemble of stochastic 15-day global forecasts, at 12-hour steps and 0.25 degree latitude-longitude resolution, for over 80 surface and atmospheric variables, in 8 minutes. It has greater skill than ENS on 97.4% of 1320 targets we evaluated, and better predicts extreme weather, tropical cyclones, and wind power production. This work helps open the next chapter in operational weather forecasting, where critical weather-dependent decisions are made with greater accuracy and efficiency.

cs.LG↗

arXiv2023

Neural General Circulation Models for Weather and Climate

General circulation models (GCMs) are the foundation of weather and climate prediction. GCMs are physics-based simulators which combine a numerical solver for large-scale dynamics with tuned representations for small-scale processes such as cloud formation. Recently, machine learning (ML) models trained on reanalysis data achieved comparable or better skill than GCMs for deterministic weather forecasting. However, these models have not demonstrated improved ensemble forecasts, or shown sufficient stability for long-term weather and climate simulations. Here we present the first GCM that combines a differentiable solver for atmospheric dynamics with ML components, and show that it can generate forecasts of deterministic weather, ensemble weather and climate on par with the best ML and physics-based methods. NeuralGCM is competitive with ML models for 1-10 day forecasts, and with the European Centre for Medium-Range Weather Forecasts ensemble prediction for 1-15 day forecasts. With prescribed sea surface temperature, NeuralGCM can accurately track climate metrics such as global mean temperature for multiple decades, and climate forecasts with 140 km resolution exhibit 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. 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.

physics.ao-ph↗

arXiv2022

GraphCast: Learning skillful medium-range global weather forecasting

Global medium-range weather forecasting is critical to decision-making across many social and economic domains. Traditional numerical weather prediction uses increased compute resources to improve forecast accuracy, but cannot directly use historical weather data to improve the underlying model. We introduce a machine learning-based method called "GraphCast", which can be trained directly from reanalysis data. It predicts hundreds of weather variables, over 10 days at 0.25 degree resolution globally, in under one minute. We show that GraphCast significantly outperforms the most accurate operational deterministic systems on 90% of 1380 verification targets, and its forecasts support better severe event prediction, including tropical cyclones, atmospheric rivers, and extreme temperatures. GraphCast is a key advance in accurate and efficient weather forecasting, and helps realize the promise of machine learning for modeling complex dynamical systems.

cs.LG↗

arXiv2022

Impact of water vapor seeing on mid-infrared high-contrast imaging at ELT scale

The high-speed variability of the local water vapor content in the Earth atmosphere is a significant contributor to ground-based wavefront quality throughout the infrared domain. Unlike dry air, water vapor is highly chromatic, especially in the mid-infrared. This means that adaptive optics correction in the visible or near-infrared domain does not necessarily ensure a high wavefront quality at longer wavelengths. Here, we use literature measurements of water vapor seeing, and more recent infrared interferometric data from the Very Large Telescope Interferometer (VLTI), to evaluate the wavefront quality that will be delivered to the METIS mid-infrared camera and spectrograph for the Extremely Large Telescope (ELT), operating from 3 to 13 {\mu}m, after single-conjugate adaptive optics correction in the near-infrared. We discuss how the additional wavefront error due to water vapor seeing is expected to dominate the wavefront quality budget at N band (8-13 {\mu}m), and therefore to drive the performance of mid-infrared high-contrast imaging modes at ELT scale. Then we present how the METIS team is planning to mitigate the effect of water vapor seeing using focal-plane wavefront sensing techniques, and show with end-to-end simulations by how much the high-contrast imaging performance can be improved.

astro-ph.IM↗

arXiv2021

Skillful Precipitation Nowcasting using Deep Generative Models of Radar

Precipitation nowcasting, the high-resolution forecasting of precipitation up to two hours ahead, supports the real-world socio-economic needs of many sectors reliant on weather-dependent decision-making. State-of-the-art operational nowcasting methods typically advect precipitation fields with radar-based wind estimates, and struggle to capture important non-linear events such as convective initiations. Recently introduced deep learning methods use radar to directly predict future rain rates, free of physical constraints. While they accurately predict low-intensity rainfall, their operational utility is limited because their lack of constraints produces blurry nowcasts at longer lead times, yielding poor performance on more rare medium-to-heavy rain events. To address these challenges, we present a Deep Generative Model for the probabilistic nowcasting of precipitation from radar. Our model produces realistic and spatio-temporally consistent predictions over regions up to 1536 km x 1280 km and with lead times from 5-90 min ahead. In a systematic evaluation by more than fifty expert forecasters from the Met Office, our generative model ranked first for its accuracy and usefulness in 88% of cases against two competitive methods, demonstrating its decision-making value and ability to provide physical insight to real-world experts. When verified quantitatively, these nowcasts are skillful without resorting to blurring. We show that generative nowcasting can provide probabilistic predictions that improve forecast value and support operational utility, and at resolutions and lead times where alternative methods struggle.

cs.LG↗

arXiv2020

A triple star system with a misaligned and warped circumstellar disk shaped by disk tearing

Young stars are surrounded by a circumstellar disk of gas and dust, within which planet formation can occur. Gravitational forces in multiple star systems can disrupt the disk. Theoretical models predict that if the disk is misaligned with the orbital plane of the stars, the disk should warp and break into precessing rings, a phenomenon known as disk tearing. We present observations of the triple star system GWOrionis, finding evidence for disk tearing. Our images show an eccentric ring that is misaligned with the orbital planes and the outer disk. The ring casts shadows on a strongly warped intermediate region of the disk. If planets can form within the warped disk, disk tearing could provide a mechanism for forming wide-separation planets on oblique orbits.

astro-ph.SR↗

arXiv2019

Tiny Grains Shining Bright in the Gaps of Herbig Ae Transitional Discs

This work presents a study of two Herbig Ae transitional discs, Oph IRS 48 and HD 169142; which both have reported rings in their dust density distributions. We use Keck-II/NIRC2 adaptive optics imaging observations in the L' filter (3.8 micron) to probe the regions of these discs inwards of ~20AU from the star. We introduce our method for investigating these transitional discs, which takes a forward modelling approach: making a model of the disc (using the Monte Carlo radiative transfer code RADMC), convolving it with point-spread functions of calibrator stars, and comparing the convolved models with the observational data. The disc surface density parameters are explored with a Monte Carlo Markov Chain technique. Our analysis recovers emission from both of the discs interior to the well known optically thick walls, modelled as a ring of emission at ~15AU in Oph IRS 48, and ~7AU for HD 169142, and identifies asymmetries in both discs. Given the brightness of the near-symmetric rings compared to the reported companion candidates, we suggest that the reported companion candidates can be interpreted as slightly asymmetric disc emission or illumination.

astro-ph.EP↗

arXiv2018

Imaging the disc rim and a moving close-in companion candidate in the pre-transitional disc of V1247 Orionis

V1247 Orionis harbours a pre-transitional disc with a partially cleared gap. Earlier interferometric and polarimetric observations revealed strong asymmetries both in the gap region and in the outer disc. The presence of a companion was inferred to explain these asymmetric structures and the ongoing disc clearing. Using an extensive set of multi-wavelength and multi-epoch observations we aimed to identify the origin of the previously detected asymmetries. We have observed V1247 Ori at three epochs spanning $\sim678$ days using sparse aperture masking interferometry with Keck/NIRC2 and VLT/NACO. In addition, we search for signs of accretion through VLT/SPHERE-ZIMPOL spectral differential imaging in H$α$ and R-band continuum. Our SMA sub-millimetre interferometry in 880 $μ$m continuum and in the CO(3-2) line allows us to constrain the orientation and direction of rotation of the outer disc. We find the L'-band emission to be dominated by static features which trace forward-scattered dust emission from the inner edge of the outer disc located to the north-east. In H- and K-band, we see evidence for a companion candidate that moved systematically by 45$^{\circ}$ within the first $\sim$345 days. The separation of the companion candidate is not well constrained, but the observed position angle change is consistent with Keplerian motion of a body located on a 6 au orbit. From the SMA CO moment map, the location of the disc rim, and the detected orbital motion, we deduced the three-dimensional orientation of the disc. We see no indication of accretion in H$α$ and set upper limits for an accreting companion. The measured contrast of the companion candidate in H and K is consistent with an actively accreting protoplanet. Hence, we identify V1247 Ori as a unique laboratory for studying companion-disc interactions and disc clearing.

astro-ph.SR↗

arXiv2018

A multi-instrument and multi-wavelength high angular resolution study of MWC614: quantum heated particles inside the disk cavity

High angular resolution observations of young stellar objects are required to study the inner astronomical units of protoplanetary disks in which the majority of planets form. As they evolve, gaps open up in the inner disk regions and the disks are fully dispersed within ~10 Myrs. MWC 614 is a pre-transitional object with a ~10au radius gap. We present a set of high angular resolution observations of this object including SPHERE/ZIMPOL polarimetric and coronagraphic images in the visible, KECK/NIRC2 near-infrared aperture masking observations and VLTI (AMBER, MIDI, and PIONIER) and CHARA (CLASSIC and CLIMB) long-baseline interferometry at infrared wavelengths. We find that all the observations are compatible with an inclined disk (i ~55deg at a position angle of ~20-30deg). The mid-infrared dataset confirms the disk inner rim to be at 12.3+/-0.4 au from the central star. We determined an upper mass limit of 0.34 Msun for a companion inside the cavity. Within the cavity, the near-infrared emission, usually associated with the dust sublimation region, is unusually extended (~10 au, 30 times larger than the theoretical sublimation radius) and indicates a high dust temperature (T~1800 K). As a possible result of companion-induced dust segregation, quantum heated dust grains could explain the extended near-infrared emission with this high temperature. Our observations confirm the peculiar state of this object where the inner disk has already been accreted onto the star exposing small particles inside the cavity to direct stellar radiation.

astro-ph.SR↗

arXiv2017

Dust-trapping vortices and a potentially planet-triggered spiral wake in the pre-transitional disk of V1247 Orionis

The radial drift problem constitutes one of the most fundamental problems in planet formation theory, as it predicts particles to drift into the star before they are able to grow to planetesimal size. Dust-trapping vortices have been proposed as a possible solution to this problem, as they might be able to trap particles over millions of years, allowing them to grow beyond the radial drift barrier. Here, we present ALMA 0.04"-resolution imaging of the pre-transitional disk of V1247 Orionis that reveals an asymmetric ring as well as a sharply-confined crescent structure, resembling morphologies seen in theoretical models of vortex formation. The asymmetric ring (at 0.17"=54 au separation from the star) and the crescent (at 0.38"=120 au) seem smoothly connected through a one-armed spiral arm structure that has been found previously in scattered light. We propose a physical scenario with a planet orbiting at $\sim0.3$"$\approx$100 au, where the one-armed spiral arm detected in polarised light traces the accretion stream feeding the protoplanet. The dynamical influence of the planet clears the gap between the ring and the crescent and triggers two vortices that trap mm-sized particles, namely the crescent and the bright asymmetry seen in the ring. We conducted dedicated hydrodynamics simulations of a disk with an embedded planet, which results in similar spiral-arm morphologies as seen in our scattered light images. At the position of the spiral wake and the crescent we also observe $^{12}$CO (3-2) and H$^{12}$CO$^{+}$ (4-3) excess line emission, likely tracing the increased scale-height in these disk regions.

astro-ph.SR↗

arXiv2016

TYC 8241 2652 1 and the case of the disappearing disk: no smoking gun yet

TYC 8241 2652 1 is a young star that showed a strong mid-infrared (mid-IR, 8-25 mu) excess in all observations before 2008 consistent with a dusty disk. Between 2008 and 2010 the mid-IR luminosity of this system dropped dramatically by at least a factor of 30 suggesting a loss of dust mass of an order of magnitude or more. We aim to constrain possible models including removal of disk material by stellar activity processes, the presence of a binary companion, or other explanations suggested in the literature. We present new X-ray observations, optical spectroscopy, near-IR interferometry, and mid-IR photometry of this system to constrain its parameters and further explore the cause of the dust mass loss. In X-rays TYC 8241 2652 1 has all properties expected from a young star: Its luminosity is in the saturation regime and the abundance pattern shows enhancement of O/Fe. The photospheric Ha line is filled with a weak emission feature, indicating chromospheric activity consistent with the observed level of coronal emission. Interferometry does not detect a companion and sets upper limits on the companion mass of 0.2, 0.35, 0.1 and 0.05 M_sun at projected physical separations of 0.1-4 AU,4-5 AU, 5-10 AU, and 10-30 AU, respectively (assuming a distance of 120.9 pc). Our mid-IR measurements, the first of the system since 2012, are consistent with the depleted dust level seen after 2009. The new data confirms that stellar activity is unlikely to destroy the dust in the disk and shows that scenarios where either TYC 8241 2652 1 heats the disk of a binary companion or a potential companion heats the disk of TYC 8241 2652 1 are unlikely.

astro-ph.SR↗

arXiv2015

Observational calibration of the projection factor of Cepheids I. The Type II Cepheid kappa Pavonis

The distances of pulsating stars, in particular Cepheids, are commonly measured using the parallax of pulsation technique. The different versions of this technique combine measurements of the linear diameter variation (from spectroscopy) and the angular diameter variation (from photometry or interferometry) amplitudes, to retrieve the distance in a quasi-geometrical way. However, the linear diameter amplitude is directly proportional to the projection factor (hereafter p-factor), which is used to convert spectroscopic radial velocities (i.e., disk integrated) into pulsating (i.e., photospheric) velocities. The value of the p-factor and its possible dependence on the pulsation period are still widely debated. Our goal is to measure an observational value of the p-factor of the type-II Cepheid kappa Pavonis, whose parallax was measured with an accuracy of 5% using HST/FGS. We used this parallax as a starting point to derive the p-factor of kappa Pav, using the SPIPS technique, which is a robust version of the parallax-of-pulsation method that employs radial velocity, interferometric and photometric data. We applied this technique to a combination of new VLTI/PIONIER optical interferometric angular diameters, new CORALIE and HARPS radial velocities, as well as multi-colour photometry and radial velocities from the literature. We obtain a value of p = 1.26 +/- 0.07 for the p-factor of kappa Pav. This result agrees with several of the recently derived Period-p-factor relationships from the literature, as well as previous observational determinations for Cepheids. Individual estimates of the p-factor are fundamental to calibrating the parallax of pulsation distances of Cepheids. Together with previous observational estimates, the projection factor we obtain points to a weak dependence of the p-factor on period.

astro-ph.IM↗