Rubin LSST-France, IJClab Orsay, 25-27/11/2026

25 nov. 2026, 12:00 → 27 nov. 2026, 16:00 Europe/Paris

IJCLab

The next biannual meeting of the Rubin-LSST France community will be held 25-27 November 2026, at IJCLab in Orsay.

Details

• Meeting will begin at 2pm on Wednesday November 25th.
• Meeting ends around 1pm on Friday November 27th.
• The persons of contact for the meeting will be TBD

Social events

• We will have a collaboration meeting dinner on Thursday at "le train bleu" in gare de Lyon !

Information about talks

• Language: Acknowledging that some participants may not understand/speak French, using English is strongly encouraged.
• Speakers: please upload your slides on this website prior to your talk.

Remote attendance

• It will be possible to follow the meeting over Zoom, at least for the plenary sessions. 
• Zoom link for plenary sessions: see email

Code of conduct

French and English versions.

Venue information

• IJCLab: 
  Plenaries will be in Auditorium Joliot-Curie, 100 A, 91400 Orsay, France
  An interactive map of the area with the lunch, dinner, Auditorium and the Metro is available here. 

 

The agendas of the most recent editions of this meeting can be found online:

• May 2026
• November 2025
• June 2025
• November 2024 
• June 2024
• December 2023
• June 2023
• November 2022
• May 2022
• November 2021
• May 2021
• November 2020
• February 2020

mer. 25 novembre

General Updates

1 Welcome

Orateur: Dr Sylvie Dagoret (IJCLab)

2 General news

Orateur: Dr Johan Bregeon (IN2P3 LSPC)

3 EDI

Orateur: Adam Trigui (IP2I)

4 Update on calibration

Orateur: Thierry Souverin (LAPP)

5 Update on computing

Orateur: Fabio Hernandez (CC-IN2P3)

6 Update on Fink

Orateur: Dr Julien Peloton (CNRS-IJCLab)

16:00 Coffee break

Special talks: General talk + invited talks

7 Observation strategy for the LSST Survey

Knowledge of the observation cadence is crucial for tailoring one’s own analysis strategy. The observation strategy is simulated by the Rubin-LSST simulation team (Lynne Jones et al.). The current version is designed for a survey start in late October 2026.
I will present the operational principles of the observation strategy, covering both the angular footprint (DDF, WFD, Mini, and Micro-Surveys) and the temporal aspects, as well as the balance of filters. I will also report on the proposals and decisions regarding recent footprint modifications and potential adjustments to the filter balance in specific regions of the sky.

Orateur: Dr Sylvie Dagoret (IJCLab)

8 Preparing for dark energy with LSST Year 1

With the LSST starting earlier this year, the clock has started ticking before DR1 is available to us for testing dark energy with a variety of probes, mainly galaxies and supernovae. In this talk, I will first highlight our recent results with the full Dark Energy Survey multi-probe analysis, precusor of LSST. I will then go through my research program over the coming years to prepare for testing dynamical dark energy with LSST Year 1 in DESC as I recently joined IN2P3 and LSST-France at LPNHE.

Orateur: Agnès Ferté (LPNHE)

9 Update on Euclid

Orateur: Dr Giulio FABBIAN (Institut d'Astrophysique Spatiale (IAS))

10 Filters measurements and pretty pictures

Orateur: Jérémy Neveu (IJCLab)

jeu. 26 novembre

Science talks: Fink and Skyportal

11 Skyportal

Orateur: Sarah ANTIER (IJCLAB)

12 First Interesting Candidates from Rubin

I will describe some of the most interesting alerts received in the first few months of observation with Rubin and discuss the prospects for SN search in Rubin with Fink.

Orateur: Dr Emille Ishida (CNRS/LPC-Clermont)

13 Optical flares as triggers for blazar flares

Blazars are important emitters of the gamma-ray sky. They are also contributors of the ptical extragalactic sky. An important feature is their dramatic variability (up to two orders of magnitude).

In this presentation, we will discuss the optical-gamma-ray correlation of their variability and how to use the cadence and sensitivity of LSST to trigger Cherenkov telescopes such as the CTAO. We will introduce an early automation of such a trigger system through dedicated alerts filtered through the Fink broker.

Orateur: Julian Hamo (IJCLab)

14 From Alerts to Orbits: Reconstructing Asteroid Trajectories in the Fink Broker

Since February 2026, the Vera C. Rubin Observatory has been streaming its public alerts to community brokers such as Fink. Many of these alerts come from Solar System objects, which Rubin links on its side through its Solar System Processing pipeline: detections are connected with HelioLinC3D and the resulting trajectories are reported to the Minor Planet Center. By October 2026, this had already produced about 8.1 million observations of more than 300,000 objects, including nearly 27,000 new discoveries, mostly from commissioning and science-validation data. Yet this linking only works when an object is seen on at least three different nights within 14 days, so every detection that falls outside this window is left unlinked.

Fink-FAT is a new Rust engine designed to complement Rubin's linking, not replace it. It aims to recover the trajectories that the official pipeline leaves aside, and it runs directly inside the Fink broker. This placement is a natural fit: a broker's role is to classify alerts, and linking is what makes a consistent classification of asteroid alerts possible.

In the first part, I present the reconstruction method. Alerts from the same night are first grouped into short motion seeds. A topocentric Kalman filter then carries each candidate trajectory across nights, predicting where the object should appear and how uncertain that position is. Rather than committing early to a single link, Fink-FAT keeps several competing hypotheses alive and scores each one on its kinematic consistency. It prunes them as new observations arrive, so ambiguous associations stay open until the data settle them.

In the second part, I introduce Fink-FAT Explorer, a web interface for the reconstructed trajectories. Using results obtained on real Fink alerts, I show how candidates can be inspected through their astrometry and light curves, then validated and exported in ADES format for submission to the Minor Planet Center. The Explorer also offers a REST API for programmatic access. I conclude with the current performance of the pipeline and the challenges of scaling it to the full Rubin alert rate.

Orateur: Roman Le Montagner

10:50 Coffee Break

Science talks: Various topics

15 Hunting for Strongly Lensed Supernovae in the Rubin LSST Era

Strongly lensed supernovae (SNe) offer a unique opportunity to independently measure the Hubble constant (H₀) through time-delay cosmography, providing a valuable probe of the current H₀ tension. Their rapid follow-up also enables unique investigations of early explosion physics and progenitor properties. In this talk, I will present our ongoing efforts to develop deep-learning approaches and coordinated discovery strategies to identify strongly lensed SNe in LSST.

Our first approach leverages known and candidate strong lenses within the Rubin footprint to filter the real-time alert stream. Through the construction of community-driven watchlists, systematic alert monitoring, and human vetting, we aim to identify lensed transient candidates and coordinate their follow-up. I will present recent progress from the joint DESC/SLSC and Euclid project on lensed transients, which aims to establish and maintain shared watchlists and coordinate the monitoring of LSST/ZTF alerts through brokers and dedicated follow-up campaigns.

Our second approach pursues a direct search in the LSST alert stream using deep learning on multi-band image time series. Since individual lensed transients can generate multiple DIAObjects in the LSST Alert Production pipeline, we are developing space-time association methods to group potentially related objects, followed by flexible deep-learning architectures to classify asynchronous, variable-length image sequences and identify lensing signatures despite contamination from unrelated transients. I will present recent developments in these methods and in the simulation frameworks needed to benchmark their performance.

Together, these complementary strategies will provide a robust framework for discovering and characterizing strongly lensed supernovae in the Rubin era.

Orateur: Raoul Canameras (LAM)

16 The Rubin Galaxies Science Collaboration

We will present the Rubin Galaxies Science Collaboration and describe the current activity in the collaboration, with a particular emphasis to the exploitation of the DP2 and the low surface brightness, morphology and photometric redshift challenges.

Orateur: Simona Mei

17 Key Projects: Cosmic Shear

Orateur: Cyrille Doux (LPSC)

18 Key Projects: Clusters

Orateur: Marina Ricci (APC/CNRS)

12:45 Lunch Break

// session

cluster

19 Calibration //

20 Cluster //

21 Fink + Skyportal //

22 Stellar Streams //

Short General Assembly on AI: AI discussion

23 Discussion on AI

16:00 Coffee Break

Science talks: DP2 talks

24 The DP2 data release: a brief description for time domain science

This brief talk will show the time-domain component of the DP2 data release, and summarize some of the dataset types available to the community.

Orateur: Bruno Sanchez (CPPM - CNRS)

25 Discussion on using CCIN2P3 for contributing to Rubin pipeline improvement and DR1 contribution

The Rubin Observatory software stack and DP2 datasets are now available at CCIN2P3, enabling the execution of the pipeline and processed-data notebooks. Given the anticipated overload of USDF, several of us prefer to work directly at CCIN2P3.
However, useful data remain inaccessible: years ago, I requested the transfer of Auxtel data to CCIN2P3, but this was never completed. As of today, we lack data beyond January 2026, yet we are actively improving the Rubin pipeline. To iterate effectively—running the pipeline repeatedly for testing and refinement—we urgently need access to raw post-DP2 and pre-LSST data, including all Auxtel datasets.
For instance, calibration efforts for DR1 require both high-quality pre-survey data and the full Auxtel dataset to refine our methods. Time is of the essence if we are to contribute meaningfully to DR1. We therefore request the immediate transfer of these datasets to CCIN2P3 to support our technical work.

Orateur: Dr Sylvie Dagoret (IJCLab)

26 How (Not) to Observe Stellar Streams in Rubin Data: The Orphan Stream in DP2

Stellar streams are faint, elongated structures of stars orbiting the Milky Way. We search for the Orphan stream in Rubin DP2 and find no detection. This non-detection places upper limits on its brightness as a function of distance.

Orateur: Matthieu Pelissier (LPSC/UGA)

27 Searching for Ultra-Faint Dwarf Galaxies in the LSST Era

Ultra-faint dwarf galaxies (UFDs) are the oldest, most metal-poor, and most dark-matter-dominated galaxies known, occupying the faintest end of the galaxy luminosity function. This makes them pristine laboratories for testing the nature of dark matter and the physics of the earliest galaxies. In this talk, I will review the census of UFDs around the Milky Way, present two new UFD candidates discovered in Rubin Data Preview 2, and discuss the science that the full LSST will enable.

Orateur: Aashay Pai (University of Chicago)

ven. 27 novembre

Science talks: DP2 variables + Auxtel

28 Towards Rubin SNe Ia Cosmology: Early Lightcurve Extractions from DP2 Data

Precise constraints on dark energy parameters will rely on combining Type Ia supernovae (SNe Ia) from the Rubin Observatory Deep Drilling Fields with external datasets such as ZTF. In this context, I will present a first vanilla version of SNIa candidate lightcurve extraction developed as a preliminary step towards implementing the scene modeling precision photometry method required for Rubin cosmology.

We use DP2 data and alerts obtained from the Fink broker to perform circular aperture photometry directly on the calibrated visit images. We apply quality cuts based on visit summary metrics and calibrate the fluxes using Rubin’s ‘photoCalib’. Multi-band light curves, renormalised on a common baseline, are then processed and fitted using the SALT3 model via ‘sncosmo’, incorporating host galaxy photometric redshift from LSDR8 when available.

We reconstructed clean light curves for SNIa candidates, achieving a photometric dispersion of ~10 mmag on stable reference stars after quality filtering. The aperture photometry approach effectively mitigates astrometric and flux artifacts observed in the forced PSF alert fluxes. Furthermore, preliminary SALT3 fits on multi-band data yield physically consistent stretch and color parameters, highlighting the importance of external redshift constraints to break degeneracies.

Orateur: Garazi GARCÍA PALACIOS (Paris Saclay University - IJCLab)

29 From Andes to Alps: Unveiling the Galaxy with Gaia and LSST

Variable stars, and pulsating variables in particular (e.g., RR Lyrae, Cepheids, or Mira variables), play a key role in different areas of Astrophysics. Many follow well-defined period-luminosity relations, which have allowed them to be extensively used as distance indicators, in the Milky Way's vicinity and beyond.
Variable star research is about to undergo a revolution driven by two large surveys providing time-resolved photometry: the Gaia space mission and Rubin/LSST. Gaia is expected to increase the number of variable stars from 10 million to more than 100 million by the end of 2026. LSST will further increase this number by providing access to the faint variable sky and surveying the variable star populations to larger distances. In addition, Gaia provides time-resolved spectroscopy, allowing the derivation of the stellar chemo-physical parameters.Chemical abundances break degenerated dynamical scenarios with a variety of conserved parameters, playing a key role in merger debris studies.

We analyze the time-series photometry (from the Early Data Preview 2, EDP2) of ~20 million sources with at least 3 observations with a change in magnitude detected by the Rubin pipeline. We calculate well known metrics such as the median, median absolute deviation, inter-quartile range, skewness, and kurtosis for the different bands separately. We also compute more complex metrics like the von Neumann ratio or the "Con" metric (implemented in OGLE), which quantify the way the data is ordered and if the variability displayed by the source is similar to noise or to a smooth variation in brightness.
We crossmatch this sample with catalogues of variable stars available in the literature (Gaia, OGLE, VSX) and compare the distributions of the sample in these metrics with these catalogues. This allows us to identify which metrics more efficiently separate true variables from noisy observations, and where the different variable types differ. We intend to use these metrics to identify and classify variable stars in the Rubin catalogue, and then exploit the stars in common with Gaia DR4 to extend relations calibrated with Gaia observations into the fainter regions observed by Rubin.

Orateur: Nicolás Cristi (Observatoire de la Cote d'Azur)

30 Atmosphere characterization with AuxTel data

A precise characterization of the atmosphere transmission is essential to meet the Rubin Observatory photometric uncertainty requirements, especially for the y-band. The Auxiliary Telescope (Auxtel) of Rubin is a 1.2m telescope equipped with a spectrograph which aims at observing standard star to extract their spectra. The comparison between out-of-atmosphere spectra of those standards and the ones observed with Auxtel gives us information on the different atmospheric parameters contributions. Those atmospheric parameters extraction is done with a forward-modelling procedure called Spectractor. In this talk, I will present the dataset extracted from Auxtel observations between 2022 to 2026, with a particular focus on the precipitable water vapor (PWV). This presentation will detail the repeatability we achieve on PWV measurement, the spatial and temporal variability of water in the atmosphere, and how these variations can impact the photometry of Rubin on the scale of one night or one year.

Orateur: Enya Van den Abeele

31 Rubin Observatory atmospheric transmission: machine learning developments for slitless spectrophotometry

After more than 20 years of development, the Rubin Observatory begins collecting data in 2026. Its unique combination of depth, cadence, and sky coverage will enable an unprecedented number of observations and discoveries over the next decade. In particular, thousands of new supernovae will be detected, together with their associated light curves. However, achieving high-precision flux measurements, and consequently strong constraints on dark energy parameters, requires monitoring the atmospheric transmission along the LSST. Earth's atmosphere components create absorption bands that can vary significantly from night to night, and from season to season. To address this challenge, AuxTel, the Rubin Auxiliary Telescope, was designed to observe reference stars with a spectrograph in view of extracting key atmospheric parameters from their spectra. A crucial step in this process is the extraction of stellar spectra. This extraction is currently performed by Spectractor, a forward modeling tool. However, Spectractor has two main limitations: it relies on an analytical PSF, which is a strong assumption, and its extraction time can be relatively long. So, after presenting the importance of atmosphere in the context of supernovae, I will present an alternative approach based on machine learning (ML), which addresses both limitations. I will show that initial tests with ML models performed on simulated data achieve performance comparable to Spectractor, opening the path for a faster and more accurate reduction of slitless spectrograph data. In the future, this method could be extended to other observatories using slitless spectrophotometry such as StarDICE, Euclid and JWST.

Orateur: Angelo Lamure-Fontanini (IJCLab)

10:50 Coffee Break

Science talks: Galaxies+Clusters+LSS+blending

32 Breaking Through Blending : A Multi-Survey Approach to Dark Energy with LSST, Euclid, and Roman

Surveys like the Vera C. Rubin Observatory's LSST, ESA's Euclid, and NASA's Roman Space Telescope aim to explore dark energy through measurements of weak gravitational lensing and large-scale structure. However, blending, where faint overlapping galaxy sources are misidentified, introduces substantial errors in galaxy shape and flux measurements, especially for deep, ground-based surveys like LSST. This effect poses a major challenge to accurate WL measurements.

In this talk, I will present my results when characterizing blending in LSST commissioning (Data Preview 2) and show the blending entropy for these analyses. I will then talk about my other project which aims to quantify the impact of blending on the 3x2pt cosmological analysis using CosmoDC2.

Orateur: Samuel Mesquita (LPSC - Cosmologie Observationnelle)

33 Deblending real galaxies from EDP2 using DebVAder

In this work, we focus on DebVAder (Arcelin et al, 2021), a python package built on a variational autoencoder ML architecture to deblend and reconstruct isolated galaxies. Its feasibility has been explored but is limited to simulated albeit realistic data.

As a first test on real data, we apply the simulation-trained DebVAder architecture on the LSST EDP2 dataset from the ECDFS field. We assess the ability of the deblender to recover first isolated and then blended sources, comparing against the official object catalogs to assess fidelity. This allows us to both gauge the initial performance of the model and to identify areas where the framework needs improvement. In future work we plan to explore the use of real data to train the model, through transfer learning or by fully re-training the networks and envision using space-based surveys for a joint-survey approach.

---

Arcelin, B., Doux, C., Aubourg, E., Roucelle, C., & LSST Dark Energy Science Collaboration (2021). Deblending galaxies with variational autoencoders: A joint multiband, multi-instrument approach. MNRAS, 500(1), 531-547.

Orateur: Lia Lammert (APC / CNRS)

34 Extracting cosmological parameters from multi-wavelength cross-correlations

Galaxy clustering and weak lensing on small scales hold much of the cosmological information in the coming surveys of large scale structure. They also carry the imprint of baryons: gas expelled from haloes by feedback suppresses the matter power spectrum at the level of the statistical errors, and this effect is degenerate with . The same gas is observed directly, in X-rays, through the thermal Sunyaev–Zel'dovich effect and in 21-cm emission, and galaxies and active nuclei trace the haloes that hold it. We use a differentiable halo model that describes galaxies, AGN, hot and neutral gas and matter within one population of haloes and one baryon budget. With it we predict every auto- and cross-spectrum among these tracers, together with their joint covariance. Because each prediction is differentiable with respect to cosmological and astrophysical parameters, we compute Fisher forecasts by automatic differentiation. We find that the gain comes from constraining feedback with the data instead of marginalising over it.

Orateur: johan comparat

35 Multi-modal analysis of photometry and spectroscopy with variational auto-encoders

Recent advances in wide-area astronomical imaging surveys, such as DES, HSC-SSP, KiDS, and the current LSST, have enabled systematic cosmological studies based on large samples of galaxies and their observed properties. At the same time, ongoing and planned spectroscopic surveys, including DESI, 4MOST, and MOONS, are providing extensive redshift measurements. These datasets, when used alone, already allow detailed studies of the large-scale matter distribution in the Universe and improve constraints on cosmological parameters. Combining photometric and spectroscopic information has the potential to significantly strengthen cosmological analyses, but remains methodologically challenging due to the heterogeneous nature and differing statistical properties of the data. In this work, we present a multi-modal framework for the joint analysis of spectroscopic and photometric datasets. Our approach maps both data types into a shared latent space that can be used to infer galaxy properties. Specifically, we use variational autoencoders to construct a latent space based on spectral templates corresponding to different star-formation histories, and design encoders and decoders to transform photometric observations into this space and extract constraints on stellar population parameters. After the training on mock data only, we validated our method using photometric data from the HSC-SSP survey combined with spectroscopic redshift measurements from the DESI survey. Our results demonstrate that a latent space informed by spectroscopy enables precise analysis of photometric data, even with smaller neural networks, highlighting the crucial role of the spectroscopic modality in the construction of the latent space.

Orateur: Dr Kirill Grishin (LUX, Observatoire de Paris)

36 Deep Learning Detection of Galaxy Clusters with YOLOv10: From SDSS to Rubin/LSST

Galaxy clusters are powerful and key cosmological probes that trace the evolution of large-scale structures of the universe and inform models of cosmic evolution, but their identification in the increasingly large imaging datasets produced by current and upcoming surveys requires detection methods that are both accurate and scalable. The Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) will provide deep, wide-field optical imaging over an unprecedented survey area, enabling the detection of thousands of galaxy clusters up to high redshifts and motivating the development of new approaches for identifying galaxy clusters directly from astronomical images.

Building upon the YOLO-CL framework, originally based on YOLOv3, I have developed an updated galaxy-cluster detection pipeline using the more recent YOLOv10 object-detection architecture. The model detects and localizes galaxy clusters directly in multiband optical images, without requiring a preliminary galaxy catalog. Applied to Sloan Digital Sky Survey (SDSS) imaging using redMaPPer clusters as reference detections, the method achieves high completeness and purity, demonstrating the effectiveness of modern object-detection architectures for this task.

In parallel, I am investigating the incorporation of astrophysical prior information through class-activation-map-based auxiliary supervision, encouraging the network to focus on physically relevant galaxy populations and providing greater interpretability of its detections.

This effort aims to contribute a scalable and interpretable, high-fidelity cluster detection pipeline for next-generation cosmological analyses and to assess its robustness as we move from controlled cluster samples toward realistic next-generation survey data.

Orateur: Vinh Phat Tran (APC/UPCité)