Automatic detection of methane emissions in multispectral satellite imagery using a vision transformer.

Curbing methane emissions is among the most effective actions that can be taken to slow down global warming. However, monitoring emissions remains challenging, as detection methods have a limited quantification completeness due to trade-offs that have to be made between coverage, resolution, and detection accuracy. Here we show that deep learning can overcome the trade-off in terms of spectral resolution that comes with multi-spectral satellite data, resulting in a methane detection tool with global coverage and high temporal and spatial resolution. We compare our detections with airborne methane measurement campaigns, which suggests that our method can detect methane point sources in Sentinel-2 data down to plumes of 0.01 km2, corresponding to 200 to 300 kg CH4 h-1 sources. Our model shows an order of magnitude improvement over the state-of-the-art, providing a significant step towards the automated, high resolution detection of methane emissions at a global scale, every few days.

Phenotype and outcomes of very early onset and early onset inflammatory bowel diseases in a Montreal pediatric cohort.

Objectives: The incidence of very-early-onset inflammatory bowel disease (VEO-IBD) and early-onset IBD (EO-IBD) is increasing. Here, we report their phenotype and outcomes in a Montreal pediatric cohort. Methods: We analyzed data from patients diagnosed with IBD between January 2014 and December 2018 from the CHU Sainte-Justine. The primary endpoint was to compare the phenotypes of VEO-IBD and EO-IBD. The secondary endpoints involved comparing outcomes and rates of steroid-free clinical remission (SFCR) at 12 (±2) months (m) post-diagnosis and at last follow-up. Results: 28 (14 males) and 67 (34 males) patients were diagnosed with VEO-IBD and EO-IBD, respectively. Crohn's disease (CD) was more prevalent in EO-IBD (64.2% vs. 39.3%), whereas unclassified colitis (IBD-U) was diagnosed in 28.6% of VEO-IBD vs. 10.4% of EO-IBD (p < 0.03). Ulcerative colitis (UC) and IBD-U predominantly presented as pancolitis in both groups (VEO-IBD: 76.5% vs. EO-IBD: 70.8%). Combining all disease subtypes, histological upper GI lesions were found in 57.2% of VEO-IBD vs. 83.6% of EO-IBD (p < 0.009). In each subtype, no differential histological signature (activity, eosinophils, apoptotic bodies, granulomas) was observed between both groups. At 12 m post-diagnosis, 60.8% of VEO-IBD and 62.7% of EO-IBD patients were in SFCR. At a median follow-up of 56 m, SFCR was observed in 85.7% of VEO-IBD vs. 85.0% of EO-IBD patients. Conclusion: The rate of patients in SFCR at 1-year post-diagnosis and at the end of follow-up did not significantly differ between both groups.

Instantaneous tracking of earthquake growth with elastogravity signals.

Rapid and reliable estimation of large earthquake magnitude (above 8) is key to mitigating the risks associated with strong shaking and tsunamis1. Standard early warning systems based on seismic waves fail to rapidly estimate the size of such large earthquakes2-5. Geodesy-based approaches provide better estimations, but are also subject to large uncertainties and latency associated with the slowness of seismic waves. Recently discovered speed-of-light prompt elastogravity signals (PEGS) have raised hopes that these limitations may be overcome6,7, but have not been tested for operational early warning. Here we show that PEGS can be used in real time to track earthquake growth instantaneously after the event reaches a certain magnitude. We develop a deep learning model that leverages the information carried by PEGS recorded by regional broadband seismometers in Japan before the arrival of seismic waves. After training on a database of synthetic waveforms augmented with empirical noise, we show that the algorithm can instantaneously track an earthquake source time function on real data. Our model unlocks 'true real-time' access to the rupture evolution of large earthquakes using a portion of seismograms that is routinely treated as noise, and can be immediately transformative for tsunami early warning.

Autonomous extraction of millimeter-scale deformation in InSAR time series using deep learning.

Systematically characterizing 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.

Laboratory earthquake forecasting: A machine learning competition.

Earthquake prediction, the long-sought holy grail of earthquake science, continues to confound Earth scientists. Could we make advances by crowdsourcing, drawing from the vast knowledge and creativity of the machine learning (ML) community? We used Google's ML competition platform, Kaggle, to engage the worldwide ML community with a competition to develop and improve data analysis approaches on a forecasting problem that uses laboratory earthquake data. The competitors were tasked with predicting the time remaining before the next earthquake of successive laboratory quake events, based on only a small portion of the laboratory seismic data. The more than 4,500 participating teams created and shared more than 400 computer programs in openly accessible notebooks. Complementing the now well-known features of seismic data that map to fault criticality in the laboratory, the winning teams employed unexpected strategies based on rescaling failure times as a fraction of the seismic cycle and comparing input distribution of training and testing data. In addition to yielding scientific insights into fault processes in the laboratory and their relation with the evolution of the statistical properties of the associated seismic data, the competition serves as a pedagogical tool for teaching ML in geophysics. The approach may provide a model for other competitions in geosciences or other domains of study to help engage the ML community on problems of significance.

An exponential build-up in seismic energy suggests a months-long nucleation of slow slip in Cascadia.

Slow slip events result from the spontaneous weakening of the subduction megathrust and bear strong resemblance to earthquakes, only slower. This resemblance allows us to study fundamental aspects of nucleation that remain elusive for classic, fast earthquakes. We rely on machine learning algorithms to infer slow slip timing from statistics of seismic waveforms. We find that patterns in seismic power follow the 14-month slow slip cycle in Cascadia, arguing in favor of the predictability of slow slip rupture. Here, we show that seismic power exponentially increases as the slowly slipping portion of the subduction zone approaches failure, a behavior that shares a striking similarity with the increase in acoustic power observed prior to laboratory slow slip events. Our results suggest that the nucleation phase of Cascadia slow slip events may last from several weeks up to several months.

Probing Slow Earthquakes With Deep Learning.

Slow earthquakes may trigger failure on neighboring locked faults that are stressed sufficiently 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.

Prevalence and Risk Factors for Symptoms of Methotrexate Intolerance in Pediatric Inflammatory Bowel Disease.

BACKGROUND: Methotrexate (MTX) intolerance is defined as gastrointestinal and behavioral symptoms occurring before or after MTX administration that may lead to treatment discontinuation. The aim of this study was to determine prevalence of MTX intolerance in pediatric inflammatory bowel disease (IBD) using the Methotrexate Intolerance Severity Score developed in rheumatology and to identify risk factors for MTX intolerance. METHODS: Patients with pediatric IBD followed in the IBD clinic of Sainte Justine Hospital who had received MTX for IBD between 2004 and 2016 and were still actively on MTX were invited to fill out the Methotrexate Intolerance Severity Score questionnaire. A cutoff score of ≥6 points was used to define MTX intolerance, with at least one point for anticipatory, associative or behavioral items. RESULTS: Among 102 pediatric patients with IBD, 32 (31%) patients reported symptoms of MTX intolerance. Using a multivariable logistic regression model, factors that were associated with having symptoms of MTX intolerance were female sex (odds ratio 4.31 [95% confidence interval, 1.37-13.60], P = 0.01), receiving a dose of MTX higher than 20 mg/wk at the time of the questionnaire (odds ratio 4.06 [95% confidence interval, 1.30-12.70], P = 0.02), and having active disease according to Physician's Global Assessment (odds ratio 3.44 [95% confidence interval, 1.15-10.26], P = 0.03). Prophylactic prescription of antiemetics and folic acid did not prevent symptoms of MTX intolerance. CONCLUSIONS: Symptoms of MTX intolerance are frequent in pediatric IBD. The Methotrexate Intolerance Severity Score questionnaire could help better recognition of these symptoms. Identification of risk factors could have important implications for the success of treatment.

Optimisation of GaN LEDs and the reduction of efficiency droop using active machine learning.

A fundamental challenge in the design of LEDs is to maximise electro-luminescence efficiency at high current densities. We simulate GaN-based LED structures that delay the onset of efficiency droop by spreading carrier concentrations evenly across the active region. Statistical analysis and machine learning effectively guide the selection of the next LED structure to be examined based upon its expected efficiency as well as model uncertainty. This active learning strategy rapidly constructs a model that predicts Poisson-Schrödinger simulations of devices, and that simultaneously produces structures with higher simulated efficiencies.

Nanocathodoluminescence Reveals Mitigation of the Stark Shift in InGaN Quantum Wells by Si Doping.

Nanocathodoluminescence reveals the spectral properties of individual InGaN quantum wells in high efficiency light emitting diodes. We observe a variation in the emission wavelength of each quantum well, in correlation with the Si dopant concentration in the quantum barriers. This is reproduced by band profile simulations, which reveal the reduction of the Stark shift in the quantum wells by Si doping. We demonstrate nanocathodoluminescence is a powerful technique to optimize doping in optoelectronic devices.