ELSA: Euclid Legacy Science Advanced analysis tools
LightNews LightNews
banner
elsa-euclid.bsky.social
ELSA: Euclid Legacy Science Advanced analysis tools
@elsa-euclid.bsky.social
500 followers 130 following 85 posts
Pushing the boundaries of spectroscopy with the European Space Agency's Euclid space telescope. Account run by @stephenserjeant.bsky.social and the ELSA team https://elsa-euclid.github.io/
Posts Replies Media Videos
elsa-euclid.bsky.social
Many other @ec-euclid.bsky.social papers being released this week, including by Daniela Vergani and team on discoveries of galaxies with unusual emission lines - here's an explainer by @stephenserjeant.bsky.social 🧪🔭
"Euclid's ability to cover such a large area of sky makes it a marvellous place to find rare objects that often could not be discovered in any other way. Euclid also doesn't just take images; its spectroscopic instrument also splits light up like a rainbow to diagnose how redshifted (how far away) each galaxy is and the conditions inside each galaxy. In this paper [Vergani et al] we used Euclid's spectroscopic instrument to unveil a population of rare galaxies with stronger ionizing radiation inside them, perhaps caused by a curious, unusual history of star formation that's assembling the stellar masses of galaxies. Unpicking this will be the subject of a lot of future research, and this is just the tip of the iceberg of the rare object discoveries awaiting us with Euclid."
November 7, 2025 at 5:59 PM
elsa-euclid.bsky.social
Some marvellous @ec-euclid.bsky.social press releases just out! A beautiful image of a shimmering infrared dark cloud observed with Euclid www.esa.int/ESA_Multimed...
Image taken by the Euclid space telescope of stars on a dark background, with wispy clumps in the foreground that comprise an infrared dark cloud, made visible via Euclid's infrared camera.
November 7, 2025 at 5:51 PM
elsa-euclid.bsky.social
Our ELSA meeting conference, after a few days playing with Euclid @ec-euclid.bsky.social spectroscopy. We're enjoying the cognitive dissonance of being in a 18th century building working on a 21st century space telescope 🧪🔭
16 astronomers standing outside an ivy-clad old building that looks like it's out of Downton Abbey. The astronomers are not in period dress but maybe that's an idea for next time. A wide-angle shot of the 16 astronomers standing outside an ivy-clad old building that looks like it's out of Downton Abbey.
January 16, 2025 at 10:19 AM
elsa-euclid.bsky.social
Goldney House picture from www.bristol.ac.uk/venues/meeti... - the room we're in is slightly smaller than this one, but still beautiful
Goldney House Mahogany Parlour, with a beautiful long dining table, a magnificent fireplace, an elegant ceiling, wood panelling on the walls, and a lavish carpet.
January 13, 2025 at 1:24 PM
elsa-euclid.bsky.social
Our annual ELSA meeting has kicked off! Beautiful location in Bristol University’s Goldney House, which looks like it’s straight out of Downtown Abbey
ELSA PI Margherita Talia welcoming everyone at the start of the meeting
January 13, 2025 at 1:18 PM
elsa-euclid.bsky.social
Happy holidays from the @ec-euclid.bsky.social and the Euclid ELSA project
A globular cluster seen by the ESA Euclid space telescope, ringed by a sleigh being pulled by reindeer and Euclid, and by the words "Merry Christmas and a happy new year!"
December 27, 2024 at 8:31 PM
elsa-euclid.bsky.social
Bluesky now has over 10 million users, and this account was number 2,900,618. We’re happy to be part of this community
Celebrating Bluesky having 10 million users, this account was number 2900618
September 17, 2024 at 11:42 AM
elsa-euclid.bsky.social
Marvellous news from the ESA Euclid team announced on 🐦 !
🎉🥳🍾

🧪🔭
Tweet by @ec_euclid: Last Wednesday, we launched our main survey programme, and today, we're proud to share the first data batches have navigated the Euclid pipeline. 🛰️💻 Ready for scientific exploitation, thanks to the SGS team's hard work and dedication.
February 21, 2024 at 10:30 AM
elsa-euclid.bsky.social
This week is also the #Euclid Galaxy/AGN Evolution meeting in Bologna, where our PI Margherita presented the project. Audience Q: how to engage volunteers on the hard problem of slitless spectra, eg Galaxy Nurseries was tricky for volunteers? 🧪🔭
Margherita, standing in front of her summary slide showing our ELSA conference photo
February 15, 2024 at 2:19 PM
elsa-euclid.bsky.social
Belatedly, here are a few key slides from the Dissemination, Engagement and Citizen Science talk. Besides showing our social media comms channels, we explained why we need & use crowdsourced data mining (citizen science)
Title slide: Work package 5: Dissemination, Engagement and Citizen Science. Slide saying: CITIZEN SCIENCE IS NOT OUTREACH Provocative question: What would you like 100000 people to do for you?
February 15, 2024 at 8:49 AM
elsa-euclid.bsky.social
Some catastrophic photo-z failures in template fitting but these are reduced with ML. Also tested on a COSMOS-like reference sample. Lots of multi-wavelength data for EDF from MeerKAT/LOFAR(2.0) to VISTA,HSC,Spitzer can help with deblending - could use GANs?
For most of the 15000 deg' of the EWS, photo-¡s recovery will be optimal for cosmology. Things are more of a mixed bag for stellar masses and SFRs; to a certain degree, we can confide in recovered stellar masses, not SFRs, at least for the whole catalog of detected sources. Better confidence could be put in bright sources whose photo- is almost perfectly recovered (or with spec-z available), but even in the most ideal scenario, those nine photometric bands yield at least 8% catastrophic outliers in SFR. For the EDE, the recovered SFRs still look worrisome, but the fields will undoubtedly rely on a wealth of ancillary or forthcoming multiwavelength observations in those domains where we can obtain better constraints (i.e., sub-mm, radio) Deblending to Euclid resolution has to be done. Within ELSA, we want to test the performance of super-resolution GANs to achieve this task, at least at the Proof of Concept level. Improvements to the EWS - outlier analysis
February 13, 2024 at 11:02 AM
elsa-euclid.bsky.social
Another application of supervised machine learning is to use human classifications as labels, already successfully demonstrated with clumpy galaxies. Also will deploy unsupervised learning (eg VAE) for photometry and specta, using clustering in latent space to classify and find rare objects
Supervised learning: going beyond Zoobot to get detailed galaxy morphologies, segment galaxy substructure. Targets for spectral extraction? Flexible object detection model built on Zoobot backbone. Zoobot will train on Euclid data. Test case: clump detection. Slide shows virtuous circle between citizen science, faster R-CNN model, then Mask R-CNN for masks, then feeding back into citizen science.
February 13, 2024 at 9:46 AM
elsa-euclid.bsky.social
Day 2 of our ELSA Euclid kick-off meeting! First up is Sotiria Fotopoulou on our Machine Learning work package. They will use the AstronomicAL visualisation tool. Euclid data are BIG: 1B galaxies, 100M spectra - use supervised learning to get physical properties, and deblend overlapping spectra 🧪🔭
Diagram showing how the Euclid data feeds into ELSA, into the 1d and 2d spectra work packages, and then into machine learning, which in turn links to citizen science. The project outputs are legacy data products and public analysis tools, and there is an overarching management work package. Slide showing the AstronomicAL visualisation tool, which is good for heterogeneous data, data labelling, active learning, tables and web services. The goal is to connect to the Euclid archive, 1d and 2d Euclid spectra visualisation, and holistic data exploration.
February 13, 2024 at 9:21 AM
elsa-euclid.bsky.social
Obligatory conference photos from our kick off meeting today!
ELSA consortium meeting conference photo from a rooftop in Bologna ELSA consortium meeting conference photo from a rooftop in Bologna
February 12, 2024 at 9:09 PM
elsa-euclid.bsky.social
Now Ben Granett is presenting ELSA WP3 on advanced 2D reconstructed spectra in Euclid, using forward modelling. Very nice proof of concept with a simulated strong gravitational lens. Other applications include galaxy kinematics
Task 3.1 - spatially resolved spectroscopy Spectral data cubes for selected sources. Task 3.2 - Inference of galaxy properties with forward modelling Forward modeling pipeline for dispersed images; Stellar population synthesis and extinction modelling; Spectroscopic verification of gravitational lens systems. Task 3.3 - 2D spectra data mine Tools to visualise Euclid spectroscopic data products; Tools to visualise and identify artefacts in NISP data; Machine learning classification methods to identify sources such as gravitational lens systems. Science drivers The Euclid Deep Fields will survey 50 sqr deg. Grism spectroscopy will give spectral information for every pixel (7 billion). Spatially-resolved spectroscopy allows us to: Probe galaxy metallicity gradients; Study star-formation histories of clumps and substructures, Measure galaxy kinematics; Identify rare galaxies and lens systems spectroscopically; Technical drivers: The spectroscopic reduction by the Science Ground Segment (SGS) does not support spatially-resolved spectroscopic analyses. Tools are not available to visualize Euclid spectroscopic products. Data cube reconstruction of an Einstein ring, showing a roughly 1 arcsecond Einstein radius, two dispersed observations, and a deconvolved map showing emission line contours.
February 12, 2024 at 4:22 PM
elsa-euclid.bsky.social
Salvatore Quai is next, with a science talk on spectral stacking using MAMBO simulations (based on Millenium dark matter haloes and the Empirical Galaxy Generator), with a powerful stackSpec python routine. Nice showcase of simulated SFR-Mstar relation
Tasks 2.1 in ELSA-WP2 Task 2.1: Spectral stacking (UNIBO, INAF) (1) Develop a high-performant spectral stacking code specifically suited for slitless Euclid spectra, handling thousands to millions spectra [...] using photometric redshift and specific machine learning techniques (e.g., Self Organizing Maps); this will be done in collaboration with WP4. (2) Develop specific codes for spectral analysis (lines and continuum properties, breaks and spectral indices) on stacked spectra, adapted to the higher S/N that will be achieved through stacking. (3) Produce stacked spectra with high-S/N covering different regions of the parameter space, in order to derive galaxy physical properties (e.g. age, gas and stellar metallicity, AGN contribution, SFR, dust content, gas ionisation state) and spectral features measurements catalogues. (4) Methodology to use stacked spectra to derive purity and completeness for a given galaxy population (5) Tools to visualise the Euclid spectra Slide shows detailed flowchart. Tool is versatile, allows for handling various normalisations, resampling, sigma clipping, bootstraping etc. Tool is user friendly: a GUI or Jupyter notebook will assist users in creating their own stacking configurations. Plot of simulated star formation rate - stellar mass relation, with some some simulated spectral stacks showing clear features for both starbust-mode and passive galaxies.
February 12, 2024 at 2:39 PM
elsa-euclid.bsky.social
After lunch and more policy chats, we're on to Lucia Pozzetti talking about WP2, on advanced analysis of 1d spectra. 30 million galaxies with spectra and spectro-z in the wide survey alone! ELSA extracting physical parameters and making spectral stacks.
Description and tasks of WP2 This WP is subdivided into two tasks: Task 2.1: Spectral stacking (UNIBO, INAF) (1) Develop a high-performant spectral stacking code, handling thousands to millions spectra (2) Produce stacked spectra with high-S/N covering different regions of the parameter space (3) Develop specific codes for spectral analysis (lines and continuum properties, breaks and spectral indices) (4) Develop a methodology to use stacked spectra to derive purity and completeness (5) Develop tools to visualise the Euclid spectroscopy data products Task 2.2: Spectro-photometric fitting and theoretical models (INAF, UNIBO) (1) Adapt existing codes and develop the pipeline to perform spectrophotometric fitting using both photometry, spectral features and full spectra, including AGN contribution. (2) Update state-of-the-art theoretical models and use them to produce simulated spectra to be used to interpret the results
February 12, 2024 at 1:48 PM