Detecting moonquakes using convolutional neural networks, a non-local training set, and transfer learning

Civilini, F.; Weber, R. C.; Jiang, Zhuocheng; Phillips, Darci; Pan, W. David

doi:10.1093/gji/ggab083

Cited by 16 publications

(24 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The detection algorithm could be developed for implementation on board missions to optimise data downlink, such as that proposed by Civilini et al (2021). Our proposed algorithm can rank the 34 A. E. Stott et al detection in terms of SNR and so the threshold at which a detection would be downlinked is flexible depending on downlink resources.…”

Section: Discussionmentioning

confidence: 99%

“…In planetary scenarios, machine learning approaches have been applied for event detection on the Moon. Knapmeyer-Endrun & Hammer (2015) and Civilini et al (2021) respectively developed a hidden markov model and convolutional neural network approach to detect patterns related to moonquakes. These are a supervised learning classification style problem.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Machine learning and marsquakes: a tool to predict atmospheric-seismic noise for the NASA InSight mission

Stott

Garcia

Chédozeau

et al. 2022

Geophysical Journal International

View full text Add to dashboard Cite

SUMMARY The SEIS (seismic experiment for the interior structure of Mars) experiment on the NASA InSight mission has catalogued hundreds of marsquakes so far. However, the detectability of these events is controlled by the weather which generates noise on the seismometer. This affects the catalogue on both diurnal and seasonal scales. We propose to use machine learning methods to fit the wind, pressure and temperature data to the seismic energy recorded in the 0.4–1 and 2.2–2.6 Hz bandwidths to examine low- (LF) and high-frequency (HF) seismic event categories respectively. We implement Gaussian process regression and neural network models for this task. This approach provides the relationship between the atmospheric state and seismic energy. The obtained seismic energy estimate is used to calculate signal-to-noise ratios (SNR) of marsquakes for multiple bandwidths. We can then demonstrate the presence of LF energy above the noise level during several events predominantly categorized as HF, suggesting a continuum in event spectra distribution across the marsquake types. We introduce an algorithm to detect marsquakes based on the subtraction of the predicted noise from the observed data. This algorithm finds 39 previously undetected marsquakes, with another 40 possible candidates. Furthermore, an analysis of the detection algorithm’s variable threshold provides an empirical estimate of marsquake detectivity. This suggests that events producing the largest signal on the seismometer would be seen almost all the time, the median size signal event 45–50 per cent of the time and smallest signal events 5−20 per cent of the time.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Machine learning and marsquakes: a tool to predict atmospheric-seismic noise for the NASA InSight mission

Stott

Garcia

Chédozeau

et al. 2022

Geophysical Journal International

View full text Add to dashboard Cite

show abstract

“…Looking forward, researchers should leverage strategies to increase the small amount of training data typically available. For example, adding domain specific data to generalized training data (i.e., transfer learning) has helped detect moonquakes (Civilini et al 2021) and classify volcano seismic events (Titos et al 2020). Integrated seismo-acoustic ML analysis will be particularly beneficial.…”

Section: Instrumentation and Computationmentioning

confidence: 99%

Volcano infrasound: progress and future directions

Watson¹,

Iezzi

Toney

et al. 2022

Bull Volcanol

View full text Add to dashboard Cite

Over the past two decades (2000–2020), volcano infrasound (acoustic waves with frequencies less than 20 Hz propagating in the atmosphere) has evolved from an area of academic research to a useful monitoring tool. As a result, infrasound is routinely used by volcano observatories around the world to detect, locate, and characterize volcanic activity. It is particularly useful in confirming subaerial activity and monitoring remote eruptions, and it has shown promise in forecasting paroxysmal activity at open-vent systems. Fundamental research on volcano infrasound is providing substantial new insights on eruption dynamics and volcanic processes and will continue to do so over the next decade. The increased availability of infrasound sensors will expand observations of varied eruption styles, and the associated increase in data volume will make machine learning workflows more feasible. More sophisticated modeling will be applied to examine infrasound source and propagation effects from local to global distances, leading to improved infrasound-derived estimates of eruption properties. Future work will use infrasound to detect, locate, and characterize moving flows, such as pyroclastic density currents, lahars, rockfalls, lava flows, and avalanches. Infrasound observations will be further integrated with other data streams, such as seismic, ground- and satellite-based thermal and visual imagery, geodetic, lightning, and gas data. The volcano infrasound community should continue efforts to make data and codes accessible and to improve diversity, equity, and inclusion in the field. In summary, the next decade of volcano infrasound research will continue to advance our understanding of complex volcano processes through increased data availability, sensor technologies, enhanced modeling capabilities, and novel data analysis methods that will improve hazard detection and mitigation.

show abstract

“…Angel et al 1990;Odewahn et al 1992;Snider et al 2001;Pearson et al 2018;Hon et al 2018;Ni et al 2021), geophysics and planetary science (e.g. Rigol-Sanchez et al 2003;Ross et al 2018;Civilini et al 2021), or asteroid studies (e.g. Howell et al 1994;Misra & Bus 2008;Lieu et al 2019;Wallace et al 2021;Penttilä et al 2021;Penttilä et al 2022) because they combine a high degree of flexibility with efficient processing of new observations.…”

Section: Introductionmentioning

confidence: 99%

Neural network for determining an asteroid mineral composition from reflectance spectra

Korda

Penttilä

Klami

et al. 2023

A&A

View full text Add to dashboard Cite

Context. Chemical and mineral compositions of asteroids reflect the formation and history of our Solar System. This knowledge is also important for planetary defence and in-space resource utilisation. In the next years, space missions will generate extensive spectral datasets from asteroids or planets with spectra that will need to be processed in real time. Aims. We aim to develop a fast and robust neural-network-based method for deriving the mineral modal and chemical compositions of silicate materials from their visible and near-infrared spectra. The method should be able to process raw spectra without significant pre-processing. Methods. We designed a convolutional neural network with two hidden layers for the analysis of the spectra, and trained it using labelled reflectance spectra. For the training, we used a dataset that consisted of reflectance spectra of real silicate samples stored in the RELAB and C-Tape databases, namely olivine, orthopyroxene, clinopyroxene, their mixtures, and olivine-pyroxene-rich meteorites.Results. We used the model on two datasets. First, we evaluated the model reliability on a test dataset where we compared the model classification with known compositional reference values. The individual classification results are mostly within 10 percentage-point intervals around the correct values. Second, we classified the reflectance spectra of S-complex (Q-type and V-type, also including A-type) asteroids with known Bus-DeMeo taxonomy classes. The predicted mineral chemical composition of S-type and Q-type asteroids agree with the chemical composition of ordinary chondrites. The modal abundances of V-type and A-type asteroids show a dominant contribution of orthopyroxene and olivine, respectively. Additionally, our predictions of the mineral modal composition of S-type and Q-type asteroids show an apparent depletion of olivine related to the attenuation of its diagnostic absorptions with space weathering. This trend is consistent with previous results of the slower pyroxene response to space weathering relative to olivine. Conclusions. The neural network trained with real silicate samples and their mixtures is applicable for a quantitative mineral evaluation of spectra of asteroids that are rich in dry silicates. The modal abundances and mineral chemistry of common silicates (olivine and pyroxene) can be derived with an accuracy better than 10 percentage points. The classification is fast and has a relatively small computer-memory footprint. Therefore, our code is suitable for processing large spectral datasets in real time.

show abstract

Detecting moonquakes using convolutional neural networks, a non-local training set, and transfer learning

Cited by 16 publications

References 47 publications

Machine learning and marsquakes: a tool to predict atmospheric-seismic noise for the NASA InSight mission

Machine learning and marsquakes: a tool to predict atmospheric-seismic noise for the NASA InSight mission

Volcano infrasound: progress and future directions

Neural network for determining an asteroid mineral composition from reflectance spectra

Contact Info

Product

Resources

About