arXiv subjects

Claudia Hulbert

Publications and source records attributed to Claudia Hulbert.

11 recordsLinked to original sources

arXiv2025

Anomaly detection in satellite imagery through temporal inpainting

Detecting surface changes from satellite imagery is critical for rapid disaster response and environmental monitoring, yet remains challenging due to the complex interplay between atmospheric noise, seasonal variations, and sensor artifacts. Here we show that deep learning can leverage the temporal redundancy of satellite time series to detect anomalies at unprecedented sensitivity, by learning to predict what the surface should look like in the absence of change. We train an inpainting model built upon the SATLAS foundation model to reconstruct the last frame of a Sentinel-2 time series from preceding acquisitions, using globally distributed training data spanning diverse climate zones and land cover types. When applied to regions affected by sudden surface changes, the discrepancy between prediction and observation reveals anomalies that traditional change detection methods miss. We validate our approach on earthquake-triggered surface ruptures from the 2023 Turkey-Syria earthquake sequence, demonstrating detection of a rift feature in Tepehan with higher sensitivity and specificity than temporal median or Reed-Xiaoli anomaly detectors. Our method reaches detection thresholds approximately three times lower than baseline approaches, providing a path towards automated, global-scale monitoring of surface changes from freely available multi-spectral satellite data.

cs.CV↗

arXiv2023

Diffusion Models for Interferometric Satellite Aperture Radar

Probabilistic Diffusion Models (PDMs) have recently emerged as a very promising class of generative models, achieving high performance in natural image generation. However, their performance relative to non-natural images, like radar-based satellite data, remains largely unknown. Generating large amounts of synthetic (and especially labelled) satellite data is crucial to implement deep-learning approaches for the processing and analysis of (interferometric) satellite aperture radar data. Here, we leverage PDMs to generate several radar-based satellite image datasets. We show that PDMs succeed in generating images with complex and realistic structures, but that sampling time remains an issue. Indeed, accelerated sampling strategies, which work well on simple image datasets like MNIST, fail on our radar datasets. We provide a simple and versatile open-source https://github.com/thomaskerdreux/PDM_SAR_InSAR_generation to train, sample and evaluate PDMs using any dataset on a single GPU.

cs.CV↗

arXiv2023

Autonomous Detection of Methane Emissions in Multispectral Satellite Data Using Deep Learning

Methane is one of the most potent greenhouse gases, and its short atmospheric half-life makes it a prime target to rapidly curb global warming. However, current methane emission monitoring techniques primarily rely on approximate emission factors or self-reporting, which have been shown to often dramatically underestimate emissions. Although initially designed to monitor surface properties, satellite multispectral data has recently emerged as a powerful method to analyze atmospheric content. However, the spectral resolution of multispectral instruments is poor, and methane measurements are typically very noisy. Methane data products are also sensitive to absorption by the surface and other atmospheric gases (water vapor in particular) and therefore provide noisy maps of potential methane plumes, that typically require extensive human analysis. Here, we show that the image recognition capabilities of deep learning methods can be leveraged to automatize the detection of methane leaks in Sentinel-2 satellite multispectral data, with dramatically reduced false positive rates compared with state-of-the-art multispectral methane data products, and without the need for a priori knowledge of potential leak sites. Our proposed approach paves the way for the automated, high-definition and high-frequency monitoring of point-source methane emissions across the world.

physics.geo-ph↗

arXiv2020

Tremor Waveform Denoising and Automatic Location with Neural Network Interpretation

Active faults release tectonic stress imposed by plate motion through a spectrum of slip modes, from slow, aseismic slip, to dynamic, seismic events. Slow earthquakes are often associated with tectonic tremor, non-impulsive signals that can easily be buried in seismic noise and go undetected. We present a new methodology aimed at improving the detection and location of tremors hidden within seismic noise. After identifying tremors with a classic convolutional neural network, we rely on neural network attribution to extract core tremor signatures and denoise input waveforms. We then use these cleaned waveforms to locate tremors with standard array-based techniques. We apply this method to the Cascadia subduction zone, where we identify tremor patches consistent with existing catalogs. In particular, we show that the cleaned signals resulting from the neural network attribution analysis correspond to a waveform traveling in the Earth's crust and mantle at wavespeeds consistent with local estimates. This approach allows us to extract small signals hidden within the noise, and therefore to locate more tremors than in existing catalogs.

physics.geo-ph↗

arXiv2020

Autonomous Extraction of Millimeter-scale Deformation in InSAR Time Series Using Deep Learning

Systematic characterization of slip behaviours on active faults is key to unraveling the physics of tectonic faulting and the interplay between slow and fast earthquakes. Interferometric Synthetic Aperture Radar (InSAR), by enabling measurement of ground deformation at a global scale every few days, may hold the key to those interactions. However, atmospheric propagation delays often exceed ground deformation of interest despite state-of-the art processing, and thus InSAR analysis requires expert interpretation and a priori knowledge of fault systems, precluding global investigations of deformation dynamics. Here we show that a deep auto-encoder architecture tailored to untangle ground deformation from noise in InSAR time series autonomously extracts deformation signals, without prior knowledge of a fault's location or slip behaviour. Applied to InSAR data over the North Anatolian Fault, our method reaches 2 mm detection, revealing a slow earthquake twice as extensive as previously recognized. We further explore the generalization of our approach to inflation/deflation-induced deformation, applying the same methodology to the geothermal field of Coso, California.

physics.geo-ph↗

arXiv2019

A Silent Build-up in Seismic Energy Precedes Slow Slip Failure in the Cascadia Subduction Zone

We report on slow earthquakes in Northern Cascadia, and show that continuous seismic energy in the subduction zone follows specific patterns leading to failure. We rely on machine learning models to map characteristic energy signals from low-amplitude seismic waves to the timing of slow slip events. We find that patterns in seismic energy follow the 14-month slow slip cycle. Our results point towards a recurrent build-up in seismic energy as the fault approaches failure. This behavior shares a striking resemblance with our previous observations from slow slips in the laboratory.

physics.geo-ph↗

arXiv2019

Probing slow earthquakes with deep learning

Slow earthquakes may trigger failure on neighboring locked faults that are stressed enough to break, and slow slip patterns may evolve before a nearby great earthquake. However, even in the clearest cases such as Cascadia, slow earthquakes and associated tremor have only been observed in intermittent and discrete bursts. By training a convolutional neural network to detect known tremor on a single seismic station in Cascadia, we isolate and identify tremor and slip preceding and following known larger slow events. The deep neural network can be used for the detection of quasi-continuous tremor, providing a proxy that quantifies the slow slip rate. Furthermore, the model trained in Cascadia recognizes tremor in other subduction zones and also along the San Andreas Fault at Parkfield, suggesting a universality of waveform characteristics and source processes, as posited from experiments and theory.

physics.geo-ph↗

arXiv2018

Breaking Cascadia's Silence: Machine Learning Reveals the Constant Chatter of the Megathrust

Tectonic faults slip in various manners, ranging from ordinary earthquakes to slow slip events to aseismic fault creep. The frequent occurrence of slow earthquakes and their sensitivity to stress make them a promising probe of the neighboring locked zone where megaquakes take place. This relationship, however, remains poorly understood. We show that the Cascadia megathrust is continuously broadcasting a tremor-like signal that precisely informs of fault displacement rate throughout the slow slip cycle. We posit that this signal provides indirect, real-time access to physical properties of the megathrust and may ultimately reveal a connection between slow slip and megaquakes.

physics.geo-ph↗

arXiv2018

Estimating the Physical State of a Laboratory Slow Slipping Fault from Seismic Signals

Over the last two decades, strain and GPS measurements have shown that slow slip on earthquake faults is a widespread phenomenon. Slow slip is also inferred from correlated small amplitude seismic signals known as nonvolcanic tremor and low frequency earthquakes (LFEs). Slow slip has been reproduced in laboratory and simulation studies, however the fundamental physics of these phenomena and their relationship to dynamic earthquake rupture remains poorly understood. Here we show that, in a laboratory setting, continuous seismic waves are imprinted with fundamental signatures of the fault's physical state. Using machine learning on continuous seismic waves, we can infer several bulk characteristics of the fault (friction, shear displacement, gouge thickness), at any time during the slow slip cycle. This analysis also allows us to infer many properties of the future behavior of the fault, including the time remaining before the next slow slip event. Our work suggests that by applying machine learning approaches to continuous seismic data, new insight into the physics of slow slip could be obtained in Earth.

physics.geo-ph↗

arXiv2018

Estimating Fault Friction from Seismic Signals in the Laboratory

Nearly all aspects of earthquake rupture are controlled by the friction along the fault that progressively increases with tectonic forcing, but in general cannot be directly measured. We show that fault friction can be determined at any time, from the continuous seismic signal. In a classic laboratory experiment of repeating earthquakes, we find that the seismic signal follows a specific pattern with respect to fault friction, allowing us to determine the fault's position within its failure cycle. Using machine learning, we show that instantaneous statistical characteristics of the seismic signal are a fingerprint of the fault zone shear stress and frictional state. Further analysis of this fingerprint leads to a simple equation of state quantitatively relating the seismic signal power and the friction on the fault. These results show that fault zone frictional characteristics and the state of stress in the surroundings of the fault can be inferred from seismic waves, at least in the laboratory.

physics.geo-ph↗

arXiv2017

Machine Learning Predicts Laboratory Earthquakes

Forecasting fault failure is a fundamental but elusive goal in earthquake science. Here we show that by listening to the acoustic signal emitted by a laboratory fault, machine learning can predict the time remaining before it fails with great accuracy. These predictions are based solely on the instantaneous physical characteristics of the acoustical signal, and do not make use of its history. Surprisingly, machine learning identifies a signal emitted from the fault zone previously thought to be low-amplitude noise that enables failure forecasting throughout the laboratory quake cycle. We hypothesize that applying this approach to continuous seismic data may lead to significant advances in identifying currently unknown signals, in providing new insights into fault physics, and in placing bounds on fault failure times.

physics.geo-ph↗