HAL-INSU
Not a member yet
203706 research outputs found
Sort by
ArchesWeather & ArchesWeatherGen: a deterministic and generative model for efficient ML weather forecasting
No full textInternational audienceWeather forecasting plays a vital role in today's society, from agriculture and logistics to predicting the output of renewable energies, and preparing for extreme weather events. Deep learning weather forecasting models trained with the next state prediction objective on ERA5 have shown great success compared to numerical global circulation models. However, for a wide range of applications, being able to provide representative samples from the distribution of possible future weather states is critical. In this paper, we propose a methodology to leverage deterministic weather models in the design of probabilistic weather models, leading to improved performance and reduced computing costs. We first introduce \textbf{ArchesWeather}, a transformer-based deterministic model that improves upon Pangu-Weather by removing overrestrictive inductive priors. We then design a probabilistic weather model called \textbf{ArchesWeatherGen} based on flow matching, a modern variant of diffusion models, that is trained to project ArchesWeather's predictions to the distribution of ERA5 weather states. ArchesWeatherGen is a true stochastic emulator of ERA5 and surpasses IFS ENS and NeuralGCM on all WeatherBench headline variables (except for NeuralGCM's geopotential). Our work also aims to democratize the use of deterministic and generative machine learning models in weather forecasting research, with academic computing resources. All models are trained at 1.5° resolution, with a training budget of 9 V100 days for ArchesWeather and 45 V100 days for ArchesWeatherGen. For inference, ArchesWeatherGen generates 15-day weather trajectories at a rate of 1 minute per ensemble member on a A100 GPU card. To make our work fully reproducible, our code and models are open source, including the complete pipeline for data preparation, training, and evaluation, at https://github.com/INRIA/geoarches
Discovery of two new millisecond pulsars towards the Galactic bulge
No full textInternational audienceThe mysterious Galactic Center Excess of gamma rays could be explained by a large population of millisecond pulsars hiding in the Galactic bulge, too faint to be detected as individual high-energy point sources by the Fermi Large Area Telescope, as well as too fast and too dispersed to be detected in shallow radio pulsation surveys. Motivated by an innovative candidate selection method, we aim at detecting millisecond pulsars associated with the Galactic Center Excess by carrying deep radio pulsation searches towards promising candidates detected in the inner Galaxy, in X rays by Chandra, and in radio or gamma rays by the Very Large Array or Fermi. We conducted deep radio observation and follow-up campaigns with MeerKAT, the Murriyang and the Green Bank telescopes towards 9 X-ray candidate sources. We here report the detection of two new millisecond pulsars, including a black widow candidate, towards the Galactic bulge: PSRs J1740-2805 and J1740-28. These discoveries double the number of MSPs discovered within the innermost 2 degree from the Galactic center
NICER Magnetar Burst Catalog
No full textInternational audienceIn this paper, we present a comprehensive catalog of short bursts from magnetars based on eight years of NICER observations. A total of 1130 bursts were identified, making this the largest magnetar burst catalog to date. The sample is dominated by SGR 1935+2154, which contributes 76% of all detected bursts. We analyzed burst durations, spectral properties, and their correlations across multiple sources. Bursts from SGR 1935+2154 exhibit significantly longer durations, with a mean of 317 ms, compared to a mean of 23 ms for bursts from other magnetars. Two microsecond-scale bursts were detected for the first time, originating from 1E 1048.1-5937 and CXOU J010043.1-721134. Spectral analysis in the 0.5--8 keV range using both blackbody and power-law models shows that bursts with higher fluences have harder spectra. In contrast, correlations between burst duration and spectral parameters are weak or absent. This catalog provides a valuable dataset for studying magnetar short bursts, enabling future modeling efforts and improving our understanding of the diversity and physical mechanisms of magnetar bursts
A high-dynamic-range view of the growth of structure and the warm/hot Universe
No full textInternational audienceBaryons heat to temperatures above as they accrete onto massive overdensities -- galaxies, groups, clusters, and filaments -- where they ionize and become optically transparent. Deep mm-wave observations such as those with ALMA have begun to probe a handful (4) of massive systems at , while low-resolution mm-wave surveys have detected thousands of objects at arcminute resolution out to . To truly advance the field of the evolution of large-scale structures, mapping the warm/hot distribution of ionized gas out to the redshift of their formation, the ESO community requires a large-aperture single-dish (sub-)mm telescope. This will need to provide several orders of magnitude higher mapping speeds than currently available while preserving the few arcsecond resolution required for imaging the gas and removing contaminating radio and dusty thermal signals across the full (sub-)mm wavelength range
Highstand, drop and stillstand: reconstructing MIS 5.5 sea-level changes in the central Mediterranean
No full textInternational audienc
Relationship between geomorphological characteristics, environmental settings and activity of transitional rock glaciers: Insights from a statistical analysis in the French Alps
No full textInternational audienceAbstract Permafrost creep is manifested by the presence of rock glaciers in mountainous areas, which are climatically driven landforms. Under degrading permafrost conditions, these ice‐rich bodies tend to slow down until deactivation through a transition phase. However, the ongoing processes and their associated geomorphic responses remain are still poorly understood. This study aims to better understand the relationship between their activity, topo‐climate conditions, and associated geomorphic responses of transitional rock glaciers. The activity of 520 landforms in the French Alps was assessed through Differential Interferometry Single Aperture Radar (DInSAR). Kinematic attributes were then correlated with topo‐climatic and geomorphic characteristics using statistical exploration (Multiple Correspondence Analysis, MCA) and modelling (Multinomial/Binomial Logistic Regression, MLR/BLR). Results show that 71% of rock glaciers are stabilized or slow‐moving landforms, while 23% exhibit surface velocities greater than 10 cm/year. Both MCA and MLR/BLR highlight that fast‐moving rock glaciers are strongly correlated with higher latitudes, high elevations, steep slopes and convex morphologies, in contrast to slow‐moving rock glaciers. MLR analysis revealed further differences between slow and fast classes. Rock glaciers with velocities <1 cm/year and 1–10 cm/year are located at lower latitudes and elevations, and in regions with unfavourable permafrost conditions. However, the <1 cm/year class is still found on steep slopes, suggesting that these landforms may not contain enough ice to maintain permafrost creep. Rock glaciers with velocities between 1 and 10 cm/year are more likely on smoother slopes, but they also show high occurrence probabilities at high elevations, indicating dynamic deactivation processes. Finally, the 10–30 cm/year class is slightly more probable under unfavourable permafrost conditions, which may suggest ongoing climatic deactivation. High‐speed ranges were also associated with heterogeneous and small‐moving areas within rock glacier systems, suggesting the presence of restricted permafrost conditions within a deactivating system. This finding raises important questions about spatial transitions and the temporal evolution of such kinematic behaviour
Optimizing electro-oxidation for selective trace micropollutant removal and energy efficiency in secondary effluents
No full textInternational audienceThis study investigated the electro-oxidation (EO) of three priority micropollutants: carbamazepine (CBZ), diuron (DIU), and perfluorooctane sulfonate (PFOS), in secondary-treated wastewater using boron-doped diamond (BDD) anodes. Although BDD anodes generate strong, largely non-selective oxidants; however, their performance at environmentally relevant concentrations, across different classes of micropollutants and under realistic organic-matter loads remains insufficiently characterized, especially for fluorinated and other recalcitrant contaminants. The selected compounds represent persistent pollutants with contrasting physicochemical properties and regulatory relevance (EU 2024/3019), which requires ≥ 80% CBZ between raw wastewater and the treated effluent. A four-factor central composite design (current density 10.7–33.0 mA/cm2, electrolysis time 14–56 min, COD 0.9–29 mg/L, influent concentration 0.3–8.7 µg/L; pH 7–8; flow 40 L/h) was used to quantify the influence of operating conditions on removal efficiency, energy demand and by-product formation. CBZ and DIU were efficiently removed under most conditions (> 98%), while PFOS elimination reached up to 93% under high-time/high-current regimes. At environmentally relevant influent levels (1–2 µg/L), removals of 92% (CBZ), 80% (DIU) and 41% (PFOS) were achieved with a moderate and conceivable energy demand (3.35 kWh/m3). Dissolved organic matter slightly reduced degradation rates but did not prevent effective pollutant targeting. Acute toxicity assays revealed a transient increase occurring alongside the formation of primary transformation products, followed by a decline at extended treatment durations, consistent with the progressive oxidation of toxic intermediates. Overall, the study provides a quantitative, multi-response evaluation of BDD EO under realistic wastewater conditions, clarifying operational limits, energy–performance trade-offs, and transformation-product dynamics relevant to its implementation as an advanced polishing step
Nonphotochemical quenching and beyond: multi-layered photoprotection shapes light-stress tolerance across seasonal niches in Arctic diatoms
No full textInternational audienceThroughout their species seasonal succession, diatoms of the Arctic Ocean experience a radical habitat transformation, from surviving the dimly lit winter within sea-ice or in the water column, to rapid growth under increasing irradiances, forming massive spring blooms beneath melting ice and later in open waters. Therefore, their evolutionary path has been moulded by the opposing challenges of maximizing light capture part of the year while maintaining highly efficient photoprotection capacities to limit photodamage upon bursts of supra-optimal illumination. Two main photoprotection mechanisms exist in diatoms i) nonphotochemical quenching (NPQ) supported primarily by the xanthophyll cycle (XC) and stress-related Lhcx antenna proteins and, ii) a rapid repair cycle of photosystem (PS) II core protein, PsbA, upon photodamage. Previous studies suggest that freezing temperatures slow protein turnover and favour photoprotection strategies that rely primarily upon XC-NPQ in polar taxa. We aim to revisit this hypothesis by dissecting the high-light response of five Arctic diatom species that dominate contrasting ecological niches: sea-ice, marginal ice-zone and open waters. We exposed each species to a high-light stress and subsequent recovery period under low light, with and without, inhibitors of XC-NPQ (dithiothreitol) or of plastid protein translation (lincomycin), blocking de novo replacement of PsbA. We confirmed the crucial role of XC-NPQ in protecting PSII but also report unexpected observations that challenge our current understanding of psychrophile species response to light stress. First, the impact of lincomycin on PSII photoinhibition was stronger than that of DTT, despite PsbA turnover being undetectable by immunoblots in most cases. Second, while our data support planktonic species showing better tolerance to high light than sympagic species, we found unsuspected diversity in photoprotection strategies. We hypothesize that these differences support a gradient from conservative strategies, possibly optimized for survival in the extreme sea-ice habitat of sympagic species, to productivity-oriented strategies in open water planktonic species dominating during the bloom period. In the transforming, brighter, Arctic Ocean, the adaptedness of this community-wide strategy scheme could be undermined, shaking up the historical dominance of certain diatom taxa
Euclid preparation. Galaxy 2-point correlation function modelling in redshift space
No full textInternational audienceThe Euclid satellite will measure spectroscopic redshifts for tens of millions of emission-line galaxies. In the context of Stage-IV surveys, the 3-dimensional clustering of galaxies plays a key role in providing cosmological constraints. In this paper, we conduct a model comparison for the multipole moments of the galaxy 2-point correlation function (2PCF) in redshift space. We test state-of-the-art models, in particular the effective field theory of large-scale structure (EFT), one based on the velocity difference generating function (VDG), and different variants of Lagrangian perturbation theory (LPT) models, such as convolutional LPT (CLPT) and its effective-field-theory extension (CLEFT). We analyse the first three even multipoles of the 2PCF in the Flagship 1 simulation, which consists of four snapshots at . We study both template-fitting and full-shape approaches and find that with the template-fitting approach, only the VDG model is able to reach a minimum fitting scale of at without biasing the recovered parameters. Indeed, the EFT model becomes inaccurate already at . Conversely, in the full-shape analysis, the CLEFT and VDG models perform similarly well, but only the CLEFT model can reach while the VDG model is unbiased down to at the lowest redshift. Overall, in order to achieve the accuracy required by Euclid, non-perturbative modelling such as in the VDG or CLEFT models should be considered. At , the CLPT model is sufficient to describe the data with high figure of merit. This comparison selects baseline models that perform best in ideal conditions and sets the stage for an optimal analysis of Euclid data in configuration space
The MeerKLASS L-band On-the-Fly Continuum Survey: Data Release 1
No full textInternational audienceThe MeerKAT Large Area Synoptic Survey (MeerKLASS) collaboration has acquired multiple passes of L-band (856-1712 MHz) scanning observations over a 268 deg sky region. This scanning enables efficient, large-area sky surveys by continuously scanning the MeerKAT array back and forth at fixed elevation while recording data at 2 sec intervals, progressively covering the survey region as the Earth rotates. We employ a novel on-the-fly (OTF) interferometric imaging technique to construct continuum images and catalogs from 16 hours of scan data. These data products, constituting the first MeerKLASS L-band data release (DR1), consist of high-fidelity radio continuum images and a catalogue of 34,874 radio sources detected with a SNR9. The resulting Stokes I images achieve a median noise level of 33 Jy beam and a median angular resolution of approximately . Cross-comparisons with previous surveys confirm the consistency of our flux density scale within 5% and astrometric precision within . Additionally, flux densities measured across the seven sub-bands enable in-band spectral-index estimates for the detected sources, providing insights into their physical properties and the broader source population. We compute the differential source counts, finding good agreement with existing measurements and validating our end-to-end processing. This data release demonstrates the effectiveness of scanning surveys when combined with OTF interferometric imaging. Commensal intensity mapping and interferometric imaging offers a dramatic enhancement of survey science per invested hour of observations and could therefore be an appealing option for next generation facilities like SKA-Mid