Grégory Sainton
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Grégory Sainton has a total of 92 co-authors. Here are the top co-authors by number of shared publications:
- Taichi Kawamura
Université Paris Cité Institut de Physique du Globe de Paris CNRS Paris France
9 shared publications - Philippe Lognonné
Université Paris Cité Institut de Physique du Globe de Paris CNRS Paris France
8 shared publications - William Bruce Banerdt
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA
6 shared publications - Domenico Giardini
ETH Zurich
5 shared publications - Brigitte Knapmeyer-Endrun
Bensberg Observatory University of Cologne Bergisch Gladbach Germany
4 shared publications - Eleonore Stutzmann
Université Paris Cité Institut de Physique du Globe de Paris CNRS Paris France
4 shared publications
Latest Publications
- Open Access
- 17 March 2026
- Published by EDP Sciences in Astronomy & Astrophysics
- Vol. 707, A203
YOLO-CIANNA: Galaxy detection with deep learning in radio data
Context . As the scientific exploitation of the Square Kilometre Array (SKA) approaches, there is a need for new advanced data analysis and visualization tools capable of processing large high-dimensional datasets. Aims . In this study, we aim to generalize the YOLO-CIANNA deep learning source detection and characterization method for 3D hyperspectral HI emission cubes. Methods . We present the adaptations we made to the regression-based detection formalism and the construction of an end-to-end 3D convolutional neural network (CNN) backbone. We then describe a processing pipeline for applying the method to simulated 3D HI cubes from the SKA Observatory Science Data Challenge 2 (SDC2) dataset. Results . The YOLO-CIANNA method was originally developed and used by the MINERVA team that won the official SDC2 competition. Despite the public release of the full SDC2 dataset, no published result has yet surpassed MINERVA’s top score. In this paper, we present an updated version of our method that improves our challenge score by 9.5%. The resulting catalog exhibits a high detection purity of 92.3%, best-in-class characterization accuracy, and contains 45% more confirmed sources than concurrent classical detection tools. The method is also computationally efficient, processing the full ~1TB SDC2 data cube in 30 min on a single GPU. Conclusions . These state-of-the-art results highlight the effectiveness of 3D CNN-based detectors for processing large hyperspectral data cubes and represent a promising step toward applying YOLO-CIANNA to observational data from SKA and its precursors.
- 17 March 2026
- Published by EDP Sciences in Astronomy & Astrophysics
- Vol. 707, A203
Open AccessYOLO-CIANNA: Galaxy detection with deep learning in radio data
Context . As the scientific exploitation of the Square Kilometre Array (SKA) approaches, there is a need for new advanced data analysis and visualization tools capable of processing large high-dimensional datasets. Aims . In this study, we aim to generalize the YOLO-CIANNA deep learning source detection and characterization method for 3D hyperspectral HI emission cubes. Methods . We present the adaptations we made to the regression-based detection formalism and the construction of an end-to-end 3D convolutional neural network (CNN) backbone. We then describe a processing pipeline for applying the method to simulated 3D HI cubes from the SKA Observatory Science Data Challenge 2 (SDC2) dataset. Results . The YOLO-CIANNA method was originally developed and used by the MINERVA team that won the official SDC2 competition. Despite the public release of the full SDC2 dataset, no published result has yet surpassed MINERVA’s top score. In this paper, we present an updated version of our method that improves our challenge score by 9.5%. The resulting catalog exhibits a high detection purity of 92.3%, best-in-class characterization accuracy, and contains 45% more confirmed sources than concurrent classical detection tools. The method is also computationally efficient, processing the full ~1TB SDC2 data cube in 30 min on a single GPU. Conclusions . These state-of-the-art results highlight the effectiveness of 3D CNN-based detectors for processing large hyperspectral data cubes and represent a promising step toward applying YOLO-CIANNA to observational data from SKA and its precursors.
- 20 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 64 (01962892) , 1-16
Multiscale Clustering and Source Separation of InSight Mission Seismic Data
Unsupervised source separation involves unraveling an unknown set of source signals recorded through a mixing operator, with limited prior knowledge about the sources, and only access to a dataset of signal mixtures. This problem is inherently ill-posed and is further challenged by the variety of timescales exhibited by sources in time series data from planetary space missions. As such, a systematic multi-scale unsupervised approach is needed to identify and separate sources at different timescales. Existing methods typically rely on a preselected window size that determines their operating timescale, limiting their capacity to handle multi-scale sources. To address this issue, we propose an unsupervised multi-scale clustering and source separation framework by leveraging wavelet scattering spectra that provide a low-dimensional representation of stochastic processes, capable of distinguishing between different non-Gaussian stochastic processes. Nested within this representation space, we develop a factorial variational autoencoder that is trained to probabilistically cluster sources at different timescales. To perform source separation, we use samples from clusters at multiple timescales obtained via the factorial variational autoencoder as prior information and formulate an optimization problem in the wavelet scattering spectra representation space. When applied to the entire seismic dataset recorded during the NASA InSight mission on Mars, containing sources varying greatly in timescale, our approach disentangles such different sources, e.g., minute-long transient one-sided pulses (known as “glitches”) and structured ambient noises resulting from atmospheric activities that typically last for tens of minutes, and provides an opportunity to conduct further investigations into the isolated sources.
- 20 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 64 (01962892) , 1-16
Multiscale Clustering and Source Separation of InSight Mission Seismic Data
Unsupervised source separation involves unraveling an unknown set of source signals recorded through a mixing operator, with limited prior knowledge about the sources, and only access to a dataset of signal mixtures. This problem is inherently ill-posed and is further challenged by the variety of timescales exhibited by sources in time series data from planetary space missions. As such, a systematic multi-scale unsupervised approach is needed to identify and separate sources at different timescales. Existing methods typically rely on a preselected window size that determines their operating timescale, limiting their capacity to handle multi-scale sources. To address this issue, we propose an unsupervised multi-scale clustering and source separation framework by leveraging wavelet scattering spectra that provide a low-dimensional representation of stochastic processes, capable of distinguishing between different non-Gaussian stochastic processes. Nested within this representation space, we develop a factorial variational autoencoder that is trained to probabilistically cluster sources at different timescales. To perform source separation, we use samples from clusters at multiple timescales obtained via the factorial variational autoencoder as prior information and formulate an optimization problem in the wavelet scattering spectra representation space. When applied to the entire seismic dataset recorded during the NASA InSight mission on Mars, containing sources varying greatly in timescale, our approach disentangles such different sources, e.g., minute-long transient one-sided pulses (known as “glitches”) and structured ambient noises resulting from atmospheric activities that typically last for tens of minutes, and provides an opportunity to conduct further investigations into the isolated sources.
- Open Access
- 9 October 2024
- Published by EDP Sciences in Astronomy & Astrophysics
- Vol. 690, A211
YOLO-CIANNA: Galaxy detection with deep learning in radio data
Context. The upcoming Square Kilometer Array (SKA) will set a new standard regarding data volume generated by an astronomical instrument, which is likely to challenge widely adopted data-analysis tools that scale inadequately with the data size.Aims. The aim of this study is to develop a new source detection and characterization method for massive radio astronomical datasets based on modern deep-learning object detection techniques. For this, we seek to identify the specific strengths and weaknesses of this type of approach when applied to astronomical data.Methods. We introduce YOLO-CIANNA, a highly customized deep-learning object detector designed specifically for astronomical datasets. In this paper, we present the method and describe all the elements introduced to address the specific challenges of radio astronomical images. We then demonstrate the capabilities of this method by applying it to simulated 2D continuum images from the SKA observatory Science Data Challenge 1 (SDC1) dataset.Results. Using the SDC1 metric, we improve the challenge-winning score by +139% and the score of the only other post-challenge participation by +61%. Our catalog has a detection purity of 94% while detecting 40–60% more sources than previous top-score results, and exhibits strong characterization accuracy. The trained model can also be forced to reach 99% purity in post-process and still detect 10–30% more sources than the other top-score methods. It is also computationally efficient, with a peak prediction speed of 500 images of 512×512 pixels per second on a single GPU.Conclusions. YOLO-CIANNA achieves state-of-the-art detection and characterization results on the simulated SDC1 dataset and is expected to transfer well to observational data from SKA precursors.
- 9 October 2024
- Published by EDP Sciences in Astronomy & Astrophysics
- Vol. 690, A211
Open AccessYOLO-CIANNA: Galaxy detection with deep learning in radio data
Context. The upcoming Square Kilometer Array (SKA) will set a new standard regarding data volume generated by an astronomical instrument, which is likely to challenge widely adopted data-analysis tools that scale inadequately with the data size.Aims. The aim of this study is to develop a new source detection and characterization method for massive radio astronomical datasets based on modern deep-learning object detection techniques. For this, we seek to identify the specific strengths and weaknesses of this type of approach when applied to astronomical data.Methods. We introduce YOLO-CIANNA, a highly customized deep-learning object detector designed specifically for astronomical datasets. In this paper, we present the method and describe all the elements introduced to address the specific challenges of radio astronomical images. We then demonstrate the capabilities of this method by applying it to simulated 2D continuum images from the SKA observatory Science Data Challenge 1 (SDC1) dataset.Results. Using the SDC1 metric, we improve the challenge-winning score by +139% and the score of the only other post-challenge participation by +61%. Our catalog has a detection purity of 94% while detecting 40–60% more sources than previous top-score results, and exhibits strong characterization accuracy. The trained model can also be forced to reach 99% purity in post-process and still detect 10–30% more sources than the other top-score methods. It is also computationally efficient, with a peak prediction speed of 500 images of 512×512 pixels per second on a single GPU.Conclusions. YOLO-CIANNA achieves state-of-the-art detection and characterization results on the simulated SDC1 dataset and is expected to transfer well to observational data from SKA precursors.
- Open Access
- 13 January 2024
- Published by Wiley
Possibly seismically triggered avalanches after the S1222a Marsquake and S1000a impact event
A document by Antoine Lucas. Click on the document to view its contents.
- 13 January 2024
- Published by Wiley
Open AccessPossibly seismically triggered avalanches after the S1222a Marsquake and S1000a impact event
A document by Antoine Lucas. Click on the document to view its contents.
- Open Access
- 6 July 2023
- Published by American Geophysical Union (AGU) in Geophysical Research Letters
- Vol. 50 (13)
Evaluation of S1222a Source Parameters Based on Site Effect Simulation and Implication for the Event Origin
Estimating the rupture process of a quake constrains the tectonic setting of its source region. In terrestrial seismology, the rupture process can be described with a classical ω2 model. In this study, we focus on one key source parameter of S1222a, the corner frequency, to investigate its origin. However, estimating the corner frequency of S1222a is complicated by S1222a's non‐classical spectra which deviate from the ω2 model. To explain S1222a's spectra, we take into account the site effect at the InSight landing site. Through numerical simulation, we demonstrate that the spectral fitting is greatly improved after considering the site effect. We further estimated the source spectrum and stress drop of S1222a. The obtained stress drop resembles the values for regional tectonic earthquakes and shallow moonquakes but differs from the Cerberus Fossae marsquakes, which favors a different tectonic origin from the Cerberus Fossae marsquakes for S1222a.
- 6 July 2023
- Published by American Geophysical Union (AGU) in Geophysical Research Letters
- Vol. 50 (13)
Open AccessEvaluation of S1222a Source Parameters Based on Site Effect Simulation and Implication for the Event Origin
Estimating the rupture process of a quake constrains the tectonic setting of its source region. In terrestrial seismology, the rupture process can be described with a classical ω2 model. In this study, we focus on one key source parameter of S1222a, the corner frequency, to investigate its origin. However, estimating the corner frequency of S1222a is complicated by S1222a's non‐classical spectra which deviate from the ω2 model. To explain S1222a's spectra, we take into account the site effect at the InSight landing site. Through numerical simulation, we demonstrate that the spectral fitting is greatly improved after considering the site effect. We further estimated the source spectrum and stress drop of S1222a. The obtained stress drop resembles the values for regional tectonic earthquakes and shallow moonquakes but differs from the Cerberus Fossae marsquakes, which favors a different tectonic origin from the Cerberus Fossae marsquakes for S1222a.
- Open Access
- 22 June 2023
- Published by American Geophysical Union (AGU) in Geophysical Research Letters
- Vol. 50 (12)
Detection of Mars Normal Modes From S1222a Event and Seismic Hum
We present the first detection of normal modes on Mars using the vertical records from InSight's broad‐band seismometer following the marsquake that occurred on sol 1222. The proposed catalog lists 60 potential eigenfrequencies between 3 and 12 mHz. Due to their low signal‐to‐noise ratio, these normal modes were detected using the phasor walkout approach. The normal modes amplitudes are consistent with the upper limit of the S1222a magnitude and with high quality factors. Additionally, we provide the first detection of a Martian hum before the quake for several of these frequencies. The proposed frequencies are at about 1‐sigma of those predicted by published models based on body wave travel time inversions. Our detection of normal modes is the first made on a terrestrial planet other than Earth and opens the way for future interior models that incorporate both normal modes frequencies, surface waves velocities and body wave travel times.
- 22 June 2023
- Published by American Geophysical Union (AGU) in Geophysical Research Letters
- Vol. 50 (12)
Open AccessDetection of Mars Normal Modes From S1222a Event and Seismic Hum
We present the first detection of normal modes on Mars using the vertical records from InSight's broad‐band seismometer following the marsquake that occurred on sol 1222. The proposed catalog lists 60 potential eigenfrequencies between 3 and 12 mHz. Due to their low signal‐to‐noise ratio, these normal modes were detected using the phasor walkout approach. The normal modes amplitudes are consistent with the upper limit of the S1222a magnitude and with high quality factors. Additionally, we provide the first detection of a Martian hum before the quake for several of these frequencies. The proposed frequencies are at about 1‐sigma of those predicted by published models based on body wave travel time inversions. Our detection of normal modes is the first made on a terrestrial planet other than Earth and opens the way for future interior models that incorporate both normal modes frequencies, surface waves velocities and body wave travel times.
- Open Access
- 9 March 2023
- Published by Wiley
Discussion on seismically triggered avalanches on Mars after the s1222a Marsquake
Potential dust avalanches as aftermaths of the seismic event of S1222a detected by the SEIS/InSight seismometer are investigated. Orbital observations emphasize that the area around the estimated location shows rather flat topography with North-South ridge structures. Thermal inertia data attests that the surface is essentially composed of granular materials, including dust. Thus, only a few locations show steep slopes with fragile soil. We investigated the orbital archive and requested targeted locations to assess the avalanche rate in the area and the influence of the M$_{W}^{Ma}$ 4.7, S1222a event. We find an increase in the avalanche rates when thermal inertia is low. We investigate the best location that could explain these avalanche rates from the aftermath ground deformation and discuss the implications in regards to the regional geology.
- 9 March 2023
- Published by Wiley
Open AccessDiscussion on seismically triggered avalanches on Mars after the s1222a Marsquake
Potential dust avalanches as aftermaths of the seismic event of S1222a detected by the SEIS/InSight seismometer are investigated. Orbital observations emphasize that the area around the estimated location shows rather flat topography with North-South ridge structures. Thermal inertia data attests that the surface is essentially composed of granular materials, including dust. Thus, only a few locations show steep slopes with fragile soil. We investigated the orbital archive and requested targeted locations to assess the avalanche rate in the area and the influence of the M$_{W}^{Ma}$ 4.7, S1222a event. We find an increase in the avalanche rates when thermal inertia is low. We investigate the best location that could explain these avalanche rates from the aftermath ground deformation and discuss the implications in regards to the regional geology.
- Open Access
- 11 December 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
FARMYARD: A Generic GPU-based Pipeline for Feature Discovery from Massive Planetary LiDAR Data
In recent decades, with the placement of LiDAR remote sensing instruments in orbit, we now have global coverage of the bare-ground elevation on the Earth, Mars and beyond. Encoded in such planetary LiDAR data are interesting landscape features that promise to further our knowledge of planetary topography. However, discovery of such features entails 3 major challenges: 1) massive data; 2) the need for local multi-scale features; 3) sensitivity to interfering factors. To address these challenges, we propose FARMYARD, a generic pipeline for \underline{F}e\underline{a}ture Discove\underline{r}y Fro\underline{m} Planetar\underline{y} LiD\underline{AR} \underline{D}ata Data. To our knowledge, this is the first time such a pipeline has been proposed, which provides a brand new methodology for comparative studies of planetary topography. Specifically, drawing on the parallel computing power of the Graphics Processing Unit (GPU), we propose a novel pseudo-on-pass sweep (POPS) framework for fast and memory-efficient feature extraction for massive planetary LiDAR data, a two-level division scheme for local regions with support for multi-scale features, and a Domain-Shifted Partition (DSP) scheme for feature evaluation that is robust against interfering factors. To showcase the utility of our FARMYARD pipeline, we deploy it to a real-world research project, which seeks to find topographical signatures of life by discovering features that can potentially distinguish between the Earth and alien worlds with no known life activity. We also highlight the efficiency of our POPS framework with experiments on both synthetic and real data, which can be thousands of times faster than its CPU-based counterpart, including a multi-core parallel solution.
- 11 December 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessFARMYARD: A Generic GPU-based Pipeline for Feature Discovery from Massive Planetary LiDAR Data
In recent decades, with the placement of LiDAR remote sensing instruments in orbit, we now have global coverage of the bare-ground elevation on the Earth, Mars and beyond. Encoded in such planetary LiDAR data are interesting landscape features that promise to further our knowledge of planetary topography. However, discovery of such features entails 3 major challenges: 1) massive data; 2) the need for local multi-scale features; 3) sensitivity to interfering factors. To address these challenges, we propose FARMYARD, a generic pipeline for \underline{F}e\underline{a}ture Discove\underline{r}y Fro\underline{m} Planetar\underline{y} LiD\underline{AR} \underline{D}ata Data. To our knowledge, this is the first time such a pipeline has been proposed, which provides a brand new methodology for comparative studies of planetary topography. Specifically, drawing on the parallel computing power of the Graphics Processing Unit (GPU), we propose a novel pseudo-on-pass sweep (POPS) framework for fast and memory-efficient feature extraction for massive planetary LiDAR data, a two-level division scheme for local regions with support for multi-scale features, and a Domain-Shifted Partition (DSP) scheme for feature evaluation that is robust against interfering factors. To showcase the utility of our FARMYARD pipeline, we deploy it to a real-world research project, which seeks to find topographical signatures of life by discovering features that can potentially distinguish between the Earth and alien worlds with no known life activity. We also highlight the efficiency of our POPS framework with experiments on both synthetic and real data, which can be thousands of times faster than its CPU-based counterpart, including a multi-core parallel solution.
- Open Access
- 23 September 2022
- Published by Copernicus GmbH
Reprocessing Apollo seismic data
The Apollo Passive Seismic Experiments recorded more than 13000 events on the Moon during 1969-1977 [1]. However, not all of these events were used in our previous studies on the Moon due to the degradation of the quality of the recorded data (Figure 1A). To make better use of Apollo seismic data, we propose several new methods to reprocess it. First, since Apollo seismic records sometimes contain invalid data in every station and component (including the short period component), we removed these invalid data from raw Apollo records using time and ground station number. Then we removed several special spikes with the help of amplitude histogram. This enables us to compute the envelopes for a larger number of moonquake events. Acceleration and displacement step responses are used to fit glitches using the Lagrange multiplier method. The ‘glitches’ we term here are artifact one-sided pulses in the raw data. We find that the instrument response parameters vary with time. So, we determined the instrument response parameters with the calibration signal and removed it near the event. Finally, we used the smoothed envelope to detect the single spike (spike with only one point) and removed it when exceeds the envelope. These new processing steps lead to considerably cleaner Apollo data (Figure 1B), yield a more accurate arrival time reading, and result in an envelope that reflects more realistic waveform characteristics. We expect that additional moonquake records will become available with these methods for future research. Further study of the instrument response variations of the Apollo seismometer will also provide a reference for the new lunar seismometers to be deployed in upcoming space missions. Figure 1. An example of a comparison plot before (A) and after (B) Apollo moonquake data processing. Reference: [1] Nunn, C., Garcia, R. F., Nakamura, Y., Marusiak, A. G., Kawamura, T., Sun, D., et al. (2020), Lunar seismology: A data and instrumentation review. Space Science Reviews, 216(5), 89. doi:10.1007/s11214-020-00709-3
- 23 September 2022
- Published by Copernicus GmbH
Open AccessReprocessing Apollo seismic data
The Apollo Passive Seismic Experiments recorded more than 13000 events on the Moon during 1969-1977 [1]. However, not all of these events were used in our previous studies on the Moon due to the degradation of the quality of the recorded data (Figure 1A). To make better use of Apollo seismic data, we propose several new methods to reprocess it. First, since Apollo seismic records sometimes contain invalid data in every station and component (including the short period component), we removed these invalid data from raw Apollo records using time and ground station number. Then we removed several special spikes with the help of amplitude histogram. This enables us to compute the envelopes for a larger number of moonquake events. Acceleration and displacement step responses are used to fit glitches using the Lagrange multiplier method. The ‘glitches’ we term here are artifact one-sided pulses in the raw data. We find that the instrument response parameters vary with time. So, we determined the instrument response parameters with the calibration signal and removed it near the event. Finally, we used the smoothed envelope to detect the single spike (spike with only one point) and removed it when exceeds the envelope. These new processing steps lead to considerably cleaner Apollo data (Figure 1B), yield a more accurate arrival time reading, and result in an envelope that reflects more realistic waveform characteristics. We expect that additional moonquake records will become available with these methods for future research. Further study of the instrument response variations of the Apollo seismometer will also provide a reference for the new lunar seismometers to be deployed in upcoming space missions. Figure 1. An example of a comparison plot before (A) and after (B) Apollo moonquake data processing. Reference: [1] Nunn, C., Garcia, R. F., Nakamura, Y., Marusiak, A. G., Kawamura, T., Sun, D., et al. (2020), Lunar seismology: A data and instrumentation review. Space Science Reviews, 216(5), 89. doi:10.1007/s11214-020-00709-3
- 9 November 2021
- Published by Seismological Society of America (SSA) in Bulletin of the Seismological Society of America
- Vol. 111 (6) , 2964-2981
Anatomy of Continuous Mars SEIS and Pressure Data from Unsupervised Learning
The seismic noise recorded by the Interior Exploration using Seismic Investigations, Geodesy, and Heat Transport (InSight) seismometer (Seismic Experiment for Interior Structure [SEIS]) has a strong daily quasi-periodicity and numerous transient microevents, associated mostly with an active Martian environment with wind bursts, pressure drops, in addition to thermally induced lander and instrument cracks. That noise is far from the Earth’s microseismic noise. Quantifying the importance of nonstochasticity and identifying these microevents is mandatory for improving continuous data quality and noise analysis techniques, including autocorrelation. Cataloging these events has so far been made with specific algorithms and operator’s visual inspection. We investigate here the continuous data with an unsupervised deep-learning approach built on a deep scattering network. This leads to the successful detection and clustering of these microevents as well as better determination of daily cycles associated with changes in the intensity and color of the background noise. We first provide a description of our approach, and then present the learned clusters followed by a study of their origin and associated physical phenomena. We show that the clustering is robust over several Martian days, showing distinct types of glitches that repeat at a rate of several tens per sol with stable time differences. We show that the clustering and detection efficiency for pressure drops and glitches is comparable to or better than manual or targeted detection techniques proposed to date, noticeably with an unsupervised approach. Finally, we discuss the origin of other clusters found, especially glitch sequences with stable time offsets that might generate artifacts in autocorrelation analyses. We conclude with presenting the potential of unsupervised learning for long-term space mission operations, in particular, for geophysical and environmental observatories.
- 9 November 2021
- Published by Seismological Society of America (SSA) in Bulletin of the Seismological Society of America
- Vol. 111 (6) , 2964-2981
Anatomy of Continuous Mars SEIS and Pressure Data from Unsupervised Learning
The seismic noise recorded by the Interior Exploration using Seismic Investigations, Geodesy, and Heat Transport (InSight) seismometer (Seismic Experiment for Interior Structure [SEIS]) has a strong daily quasi-periodicity and numerous transient microevents, associated mostly with an active Martian environment with wind bursts, pressure drops, in addition to thermally induced lander and instrument cracks. That noise is far from the Earth’s microseismic noise. Quantifying the importance of nonstochasticity and identifying these microevents is mandatory for improving continuous data quality and noise analysis techniques, including autocorrelation. Cataloging these events has so far been made with specific algorithms and operator’s visual inspection. We investigate here the continuous data with an unsupervised deep-learning approach built on a deep scattering network. This leads to the successful detection and clustering of these microevents as well as better determination of daily cycles associated with changes in the intensity and color of the background noise. We first provide a description of our approach, and then present the learned clusters followed by a study of their origin and associated physical phenomena. We show that the clustering is robust over several Martian days, showing distinct types of glitches that repeat at a rate of several tens per sol with stable time differences. We show that the clustering and detection efficiency for pressure drops and glitches is comparable to or better than manual or targeted detection techniques proposed to date, noticeably with an unsupervised approach. Finally, we discuss the origin of other clusters found, especially glitch sequences with stable time offsets that might generate artifacts in autocorrelation analyses. We conclude with presenting the potential of unsupervised learning for long-term space mission operations, in particular, for geophysical and environmental observatories.