Magnetics Information Consortium (MagIC), hosted at http://earthref.org/MagIC, is a database that serves as a Findable, Accessible, Interoperable, Reusable (FAIR) archive for paleomagnetic and rock magnetic data. It has a flexible, comprehensive data model that can accomodate most kinds of paleomagnetic data.The PmagPy software package is a cross-platform and open-source set of tools written in Python for the analysis of paleomagnetic data that serves as one interface to MagIC, accommodating various levels of user expertise. It is available through github.com/PmagPy. Because PmagPy requires installation of Python and the software package, there is a speed bump for many practitioners on beginning to use the software. In order to make the software and MagIC more accessible to the broad spectrum of scientists interested in paleo and rock magnetism, we have prepared a set of Jupyter notebooks, hosted on jupyterhub.earthref.org which serve a set of purposes. 1) There is a complete course in Python for Earth Scientists, and 2) a set of notebooks that introduce PmagPy (importing the software package from the github repository). These notebooks illustrate how to conduct statistical analyses, synthesize create data and create data visualizations of the type that are typically included in papers in the field. The notebooks also demonstrate how to prepare data from the laboratory for the MagIC database. This pathway gives additional tools to researchers so that they can satisfy data archiving requirements from NSF and publishers such as AGU.

In the last 190 years a total of 39 tsunamis affecting Samoa have been recorded. Many of them caused by earthquakes occurring along the Tonga Trench, which is only 150 km away from the islands. In 1917, an earthquake with a magnitude of 8.3 caused a disastrous tsunami in Samoa. Even though it was a major event, historical records are scarce and little is known about the event. In order to overcome this lack of data, this work has modelled the aspects of the 1917 tsunami event, using available historical records. The tsunami model used an earthquake initiation, propagation along the Tonga Trench and generated inundation footprints for the islands of Samoa. Then using this model output, the impact in present-day Samoa was determined to estimate the likely exposure and damage to buildings within the inundation zone. The study identified a number of inconsistencies between the inundation zone and the anecdotal evidence recorded at the time of the event. In addition, discrepancies were identified between the model and the records from tide gauges in Apia harbour at the time. These recorded a fluctuation of the sea five minutes after the 1917 earthquake.

The ocean plays important roles not only in the global climate system but also in the global material cycling through biogeochemical processes, and has close relationships with the daily lives of people through fisheries, marine energy and resources, ocean transportation, marine sports, ocean relaxation activities, and others. The Ocean Literacy movement began in the U.S. in the early 2000s, and has recently become international as shown by the Ocean Literacy Portal website operated by IOC/UNESCO. In Japan, the present national standard course of study at elementary school has not given any room for ocean education while the Oceanographic Society of Japan (JOS) and 34 other academic societies/committees gave a proposal entitled, “On the Setting up of a class named, ‘the role of Ocean’ in the national standard course of study for the 4th grade students in the elementary science education” to the Ministry of Education, Culture, Sports, Science and Technology (MEXT) in April 2016. In the U.S., aiming to serve as a community-based measurement tool that allows the comparison of levels of ocean knowledge across time and location, the International Ocean Literacy Survey (IOLS) project has been started since 2015. Fauville et al. (2018, https://doi.org/10.1080/13504622.2018.1440381) published the results from IOLS Version 2, and the English version IOLS Version 4 was finalized after reviewing the English IOLS Version 3 by IOLS Advisory Board consisting of international marine researchers, communicators, teachers, and psychometricians. The English version of the IOLS Version 4 was translated into 14 languages (Catalan, Chilean, Dutch, Greek, Italian, Japanese, Korean, Polish, Portuguese, Simplified Chinese, Spanish, Tagalog, Thai, and Traditional Chinese) and its field testing was conducted during February and March 2019 with nominal 6945 participants worldwide including nominal 567 participants from Japan. In collaboration with IOLS Version 4 field testing, JOS and the Marine Learning Center took part in correcting the original Japanese draft of IOLS Version 4, sponsored IOLS Version 4 field testing, and made a wide campaign to high-school teachers through sending recommendation letters to academic societies and other ocean-related organizations.

The slope mass rating (SMR) method is universally used for the characterization and classification of rock slopes. SMR is calculated by reducing the value of basic rock mass rating (RMRb) by subtracting three adjustment factors F1, F2, and F3 based on the geometrical relationship between the slope and discontinuity and adding one adjustment factor F4 depending upon the excavation method used. These adjustment factors (F1, F2, and F3) are mathematical functions (continuous/discrete) that require post-processing of field data on a computer for their derivation. Less work has been done to develop the charts for the direct calculation of SMR in the field. In this paper, SMR charts are developed for the onsite classification of rock slopes. With the aid of SMR charts, an engineering geologist can easily assess the onsite SMR class of rock slopes by plotting discontinuity dip amount (plunge amount in case of wedge failure) and strike parallelism between slope dip direction and discontinuity dip direction (slope dip direction and trend direction of intersection line in case of wedge failure). Using SMR charts for any project (open-pit mines, road cut slopes, natural slopes, etc.) onsite suggestions of proper remedial and preventive measures for the rock slopes can be given, which accelerates the overall preliminary slope mass classification process. The proposed SMR charts are straightforward to use and can be adopted as useful tools for the preliminary rock slope stability assessment.

JAMSTEC publicly distributes samples and data obtained by JAMSTEC-AORI-JURCAOS Cooperative Research Cruises [1], as well as those by JAMSTEC private/open-call cruises, through online database (DB) websites, for additional uses with research/educational purposes. These online DBs are, from a long-term perspective, under a process of reorganization and integration: Public-level information systems for rock and sediment core samples were integrated into DARWIN DB [1] in FY2017. As for marine biological samples, whose information has been mainly distributed through “Marine Biological Sample DB [3], a part of information is already shared with DARWIN to utilize its search/map functions. DB integration is in progress for not only public-level but also management-level information system, which are operated by sample holders responsible for storage of physical samples. In FY2018, JSDB (JAMSTEC Sample Management Database) is installed for the management of JAMSTEC rock, sediment core and biological samples, and actual data was transferred into this new system in FY2019. For rock and sediment core samples, information forwarding from management-level system (JSDB) to public-level system (DARWIN) is planned to be established in the near future. JSDB consists of data tables for sample matadata information for onboard sampling activities, sample archive information for status and storage of physical samples, associate data information for various data obtained for samples, literature information for published documents, sample distribution information for service logs for sample requests. User authorization includes group setting, enabling individual sample holders outside Data Management Office to read, create and modify the information in an appropriate manner, although the DB is only accessible from inside JAMSTEC because of security reasons. Associate data information is supposed to include various data types, such as photos/images, text-based descriptions, spread-sheets and URLs etc., and open/close status control and group-setting are available for individual entries. Information for research proposals/projects or other data management issues, which should be associated with multiple samples, can also be managed as associated data information. For the future possibility of involvement with IGSN, sample metadata entries and their vocabulary are reorganized to follow the IGSN Descriptive Metadata [4], as much as possible. Peoples involved in sample and data management tend to spend much effort for dealing with unusual conditions/situations which are not originally expected. Robustness is important for design/operation of information systems, because lack of flexibility and too much complexity of information systems are frequent causes of such kind of problems. [1] https://www.aori.u-tokyo.ac.jp/coop/ [2] http://www.godac.jamstec.go.jp/darwin/ [3] http://www.godac.jamstec.go.jp/bio-sample/ [4] https://www.geosamples.org/metadata

A multi-threshold technique was used to objectively identify the space-time structure of meso-scale convective systems in the warm pool region using observations of high-resolution Himawari geostationary satellites. Several thresholds of the IR brightness temperature were used to capture the life cycles of cloud clusters with complex behaviors (merging and breaking) and spatial organizations (inclusion, aggregation). According to the inclusions between cloud clusters with different thresholds, cloud clusters were characterized as simple single-core or complex multi-core systems, more likely represented isolated systems or aggregated/merged systems respectively. The mixed depth multi-core clusters existed frequently, even for very cold brightness temperatures (<230K) and more of them were associated with the merging process compared to single core clusters. Thus they were often a mixture of systems at different development levels. A large amount of convective systems experience merging and breaking during their lifetime. Merging and breaking processes were found to result in unrealistic local life stage. These facts questioned the use of any single threshold technique on determining the life cycle of deep convective systems as well as that of their anvil clouds. When the same development height was selected, the multi-core systems systematically produced less warmer clouds around their cold cores than that of single-core systems. Thus per the same cold core area the muti-core systems tended to produce less total cloud coverage and weaker reduction of the TOA OLR. These phenomena were consistent with previous studies suggesting that aggregated convection might resulted in less high cloud coverage and enhanced domain-averaged OLR. The documentation and understanding about the cause of different organizations of cloud clusters and their impacts on clouds and radiation are currently lacking. If these phenomena and effects cannot be effectively documented and stratified in observations, it will be misleading when applying observations as a guide for diagnosing and improving models. This study provided an observational reference to test if current models can realistically reproduce observed organizations of deep convection along with its associated anvil and cirri-form clouds.

Flood Insurance Rate Maps (FIRMs) managed by the Federal Emergency Management Agency (FEMA) have been providing ongoing flood information to most of the communities in the United States over the past half century. However, the uncertainty associated with the modeling of the FIRMs, some of which are created by using a single HEC-RAS one-dimensional (1D) steady flow model, may have adverse effects on the reliability of flood stage and inundation extent. Therefore, a systematic understating of the uncertainty in the modeling process of FIRMs is important and necessary. The Bayesian model averaging (BMA), which is a statistical approach that can combine estimations from multiple models and produce reliable probabilistic predictions, is applied to evaluating the uncertainty associated with the FIRMs. In this study, both the BMA and HBMA approaches are used to quantify the uncertainty within the detailed FEMA models of the Deep River and the Saint Marys River in the state of Indiana based on water stage predictions from 150 HEC-RAS 1D unsteady flow model configurations that incorporate four uncertainty sources including the bridges, channel roughness, floodplain roughness, and upstream flow input. The BMA weight and the variance for each model member are obtained given the ensemble predictions and the observed water stage data in the training period, and then the BMA prediction ability is validated for the observed data from the later period. The results indicate that BMA prediction is more robust than the original FEMA model as well as the ensemble mean. Different types of uncertainty coefficients based on the BMA prediction distribution are also proposed to evaluate the FEMA models. Furthermore, the HBMA framework shows that both the channel roughness and the upstream flow input have a larger impact on prediction variance than bridges, and hence provides some insights for modelers into the relative impact of individual uncertainty sources in the flood modeling process.

Saltwater disposal (SWDs) has been linked to the recent increase of earthquakes in various regions of the United States. In some cases, the strong temporal and spatial associations have provided unequivocal evidences to the scientific community that wastewater injection is one of the dominant causal factors to the onset seismicity. In addition, numerous physical models have suggested that the increase in pore pressure from wastewater injection is capable to induce fault slips, providing further physical evidences. Another growing body of literature sorts to rigorously prove causality with statistical analysis where they propose statistical frameworks with parametric regression models to evaluate whether the observed earthquakes were occurring more often than by random chances and tested the statistical significance of the observed occurrences of earthquake to arrive at causal interpretations. We propose causal inference frameworks with the potential outcomes perspective to explicitly define what we meant by causal effect with mathematical formulations and declare necessary assumptions to ensure consistency between models for model comparison. In particular, we put considerations on two common difficulties in raster-based spatial statistical analysis, the spatial correlation, which can be described by Tobler’s first law of geography where near things are more related than distant things, and interference, a causal inference term, where treatments applied to some spatially indexed units affect the outcomes at other spatially indexed units, mostly due to complex physical processes. The study region, the Fort-Worth Basin of North Central Texas, is discretized into non-overlapping grid blocks. The first proposed workflow adopts a cross-sectional study design on aggregated earthquake catalog and injection data where two statistical methods are employed to test the significance of the causal effect between the presence or absence of saltwater disposals and the number of the earthquakes and to estimate the magnitude of the average causal effect. The second proposed workflow incorporates the temporal domain which holds more scientific interests. Finally, the analysis is repeated for different grid configurations to directly assess the sensitivity of statistical results.

Imaging spectroscopy data is becoming more readily available from different satellite and airborne platforms. As this data becomes more prolific, there is a need for shared data tools and code for wrangling, cleaning, and analyzing it. The geospatial Imaging Spectroscopy Processing Environment on the Cloud (ImgSPEC) pioneers an on-demand science data processing platform with scalable back-end compute. It considers user experience and facilitates open science. ImgSPEC enables users to create data products in areas of interest using default workflows from registered algorithms, while also enabling users to customize scripts and workflows. ImgSPEC seamlessly interfaces with NASA Earthdata Search and tracks appropriate metadata for reproducibility when generating data products to share with others. Users can work in their preferred workspace (e.g., Rstudio, Jupyterlab, or command line) thereby facilitating use of open science software packages and collaborative coding through Git. ImgSPEC leverages existing NASA-funded information technologies such as the hybrid on-premise/cloud science data system (HySDS) and the Multi-mission Algorithm and Analysis Platform (MAAP). It also creates seamless interfaces with NASA-funded ECOSIS – a crowd-sourced spectral database, and ECOSML – a crowd-sourced model database. We demonstrate ImgSPEC on the Terrestrial Ecosystem use case processing through to foliar traits and fractional cover, thus aligning with driving thrusts for the NASA Surface Biology and Geology (SBG) Science and Applications Communities. As this technology is more widely adopted the interface with Amazon Web Services and NASA Earthdata search will enable broader use of more data (publicly available or loaded by the user) across more domains.

Mount Early and Sheridan Bluff (87 degrees south) are the above-ice expression of Earth’s southernmost volcanic field and are isolated by more than 1000 km from any other exposed Cenozoic volcano in Antarctica. These monogenetic, Early Miocene volcanoes are constructed mostly of olivine-phyric basaltic pillow lavas and breccias (Mount Early) and pāhoehoe lavas (Sheridan Bluff) whose differentiation is controlled by the fractional crystallization of olivine with lesser quantities of clinopyroxene, plagioclase and magnetite. Fractional crystallization or contamination by crust cannot account for the coexistence of olivine tholeiite and alkaline compositions but their relationship can be explained by change from higher (5-6%) to lower (1.5-2%) degrees of mantle partial melting, respectively, of a source that is heterogeneous on a small-scale. Both alkaline and subalkaline magma types have geochemical and isotopic signatures that differentiate them from volcanism of the West Antarctic rift system. Data trends in Sr-Nd-Pb isotope space along with major and trace element characteristics indicate mixing of at least two-distinct mantle sources; 1) a relatively depleted component similar to sources for mid-ocean ridge basalt from the extinct Antarctic-Phoenix spreading center, and 2) an enriched component similar to sources for mafic magmas of the Jurassic Karoo‒Ferrar large igneous provinces. The availability of these two mantle source types was facilitated by the detachment, sinking and heating of metasomatized continental lithosphere (enriched source) that released volatiles into the surrounding asthenosphere (depleted source) to promote flux melting. Volcanism triggered by lithospheric detachment is therefore uniquely applied to Mount Early and Sheridan Bluff to explain their isolation and enigmatic tectonic setting but can also account for source heterogeneity and the ephemeral change in degree of mantle partial melting recorded in their mafic compositions.

In this project, I study families of deep learning neural networks that are trained on publicly available chest X-ray datasets to identify the best image classification algorithm for automating the diagnosis of respiratory illnesses. Specifically, the learned networks will be used to classify anonymized chest X-ray images to three classes: healthy, COVID-19 and non-COVID pneumonia. As in most real-world applications, publicly available chest X-ray image datasets are not abundant, and ground truth data of COVID-19 diagnosis is especially hard to come by. In this project, the first variable implemented to improve the predictive power of the neural networks is through pretraining on a domain-relevant and much larger than the transfer learning dataset. To address the imbalance within training data, the second variable implemented is to customize the data sampling configuration using the equal-weight-per-epoch method or fixed- fraction-per-batch method. As control for each neural network, pretrained weights learned from the classic ImageNet dataset are used, and no customized training data sampling method is applied. In regard to transfer learning, two scikit-learn functions, average precision and F1 score, are computed during training. Then Precision and Recall are manually calculated based on the confusion matrix for each neural network along with the hyperparameters. The most significant observation is that the Recall metric for the control group is consistently less than 0.6, which is a clear indicator of the underperformance on COVID-19 prediction. The family with significantly higher performance is DenseNet; surprisingly, DenseNet169 has one of the highest Precision and Recall of 0.870 and 0.837. With more than 82 million COVID-19 cases worldwide, the need for efficient, accurate and mass diagnosis of patients is apparent and growing. The utilization of chest X-ray images in medical diagnosis is both a cost-effective and widespread technique for early screening of respiratory illnesses.

Profiling water vapor in the marine boundary layer (MBL) is critical to marine weather prediction, maritime communications, and understanding feedbacks relevant to multi-decadal climate prediction, yet profiling the MBL remotely has proven extraordinarily difficult because of the spatial scales involved and the proximity of the ocean surface. Collocated radio occultation (RO) and nadir passive microwave (MW) soundings can be combined in retrieval to profile water vapor with the vertical resolution of RO and with super-refraction and the wet-dry ambiguity inherent to RO resolved by the MW. We have constructed a retrieval technique that considers collocated RO and MW soundings that yields profiles of water vapor in the MBL with unprecedented precision, accuracy, and vertical resolution. We have also performed RO and MW collocation studies that consider many current RO missions and MW instruments. The joint RO+MW retrieval technique mines the information in MW soundings for an inference of the microwave refractivity in the MBL surface air, removes the biasing effect of super-refraction following the approach of Xie et al. (2006, doi:10.1175/JTECH1996.1), and resolves the wet-dry ambiguity inherent to RO using the MW sounding as a constraint or a weather forecast as a prior in 1DVAR. We constructed a simulation-retrieval demonstration system that uses a multi-phase screen propagator to simulate RO amplitude and phase and the Optimal Spectral Sampler (OSS) to simulate AMSU-A radiances. In its current state, the retrieval technique is capable of producing MBL water vapor profiles with 2% accuracy and 100-meter vertical resolution. Our collocation study shows that existing RO satellites and orbiting MW instruments achieve approximately 1,300 collocations daily, defined with a temporal window of 10 minutes. To facilitate this study, we have formulated a time-dependent rotational transformation that is applied to RO geolocations. It is three orders of magnitude more efficient than a brute force approach to finding collocations and is valid to 4% precision. We have found that the greatest yield for collocations in low latitudes would come from RO satellites that would fly in tandem with the TROPICS MW CubeSats, potentially producing 1,500 daily RO+MW collocations in the Tropics and Subtropics.

Heterogeneous reactions occurring at the surface of atmospheric aerosol particles regulate the production and lifetime of a wide array of atmospheric gases. Aerosol surface area plays a critical role in setting the rate of heterogeneous reactions in atmospheric chemistry. Despite the central role for aerosol surface area, there are few assessments of the accuracy of aerosol surface area concentrations in regional and global models. In this study, we compare aerosol surface area concentrations in the EPA’s CMAQ (Community Multiscale Air Quality) model with commensurate observations from the 2011 NASA flight-based DISCOVER-AQ (Deriving Information on Surface Conditions from COlumn and VERtically Resolved Observations Relevant to Air Quality) campaign. The study region for the 2011 DISCOVER-AQ campaign focused primarily on the Baltimore and Washington, DC region. Dry aerosol surface area was measured aboard the NASA P-3B aircraft using a combination of a Scanning Mobility Particle Sizer (SMPS), Aerodynamic Particle Sizer (APS), and Ultra-High Sensitivity Aerosol Spectrometer (UHSAS). In this study, we focus on the continuous 1s UHSAS measurements in size range of 60-1000nm, as it captures the majority of the dry surface area distribution. Over the course of 13 flight campaigns, we show strong agreement between measured and modeled aerosol number concentration (CMAQ Number/UHSAS Number= 0.88). In contrast, the total surface area showed a larger discrepancy (CMAQ surface area/UHSAS surface area = 0.44). We hypothesize that the emissions and chemistry in CMAQ relating to the production and loss of each moment play a large part in the model/measurement discrepancy. Despite the disagreement, this analysis suggests that modeled aerosol surface area is accurate to within a factor of two, highlighting that uncertainty in the rate of heterogeneous reactions is largely driven by uncertainty in the reactive uptake coefficients. We would like to thank the DISCOVER-AQ NASA Langley Aerosol Research Group Experiment (LARGE) research team, including Richard Moore, Bruce Anderson, Andreas Beyersdorf, Luke Ziemba, Lee Thornhill, and Edward Winstead for the use of their data in this study.

The aim of this study is to determine the hydrochemical properties of water resources in the Seyfe Basin and to understand processes involving water chemistry using statistical techniques such as principal component analysis (PCA) and cluster analysis. Seyfe Lake closed basin is located west of Kirsehir (Turkey) in the Central Anatolia. Considering the importance of the Seyfe Lake Basin for its water resources potential and due to its fragile ecosystem, a detailed hydrogeological characterization of the basin has been undertaken. In order to characterize hydrogeological system accurately, water samples were collected from springs, wells, drainage channels during September and December 2019 field campaigns from the Seyfe basin. The temperature, pH, total dissolved solids (TDS) and specific electrical conductance (EC25°C) were measured at the field and spatio-temporal changes of major ions are evaluated. The waters have neutral to slightly alkaline character, and most of the water samples are saturated with calcite, dolomite and aragonite minerals. According to the results obtained from Gibbs, Piper plot, and PCA, the water-rock interaction takes place without any remarkable ion exchange reactions. Water facies identified with the Piper diagram; supports this view. Water types are determined as Ca-HCO3 for springs; Na-HCO3 for wells; Na-SO4 for drainage channel in the dry season and Na-Cl in the rainy season. In Ihe study area; ionic compositions of the springs are not affected by seasonal changes. The high sulfate concentration is probably related to the weathering of evaporitic rocks in the basin. The waters are grouped according to the results, and the difference observed within the groups are related to the lithology of aquifers and flow path of the waters. KEYWORDS: Hydrochemistry, Major ions, Kırsehir, Central Anatolia, Seyfe Closed Basin, Spatio-temporal Changes.

Turbulent kinetic energy dissipation rates in the Indonesian throughflow (ITF) exit passages to the Indian Ocean, the Lesser Sunda waters, are inferred from archived CTD measurements and recent high-resolution time series data sets. Dissipation rates from archived data sets are inferred using an improved Thorpe scale method validated against microstructure measurements. Elevated dissipation rates ~[10-6-10-7] m2s-3 were observed in the straits, where internal tides are generated. Tidal variations seemingly influence the dissipation rates and diffusivities as has been suggested from the yoyo profiling data sets. The spatial pattern of dissipation rates inferred from the high-resolution 3D hydrodynamics model output of Nagai et al (2015) shows a general agreement with the observations in the location of the mixing hot spots and suggests that the M2 internal tide is the dominant factor driving the turbulent kinetic dissipation rates in this region. The bias in the model is possible due to the lack of representation of the ITF and mesoscale circulation in the model.

When the COVID-19 pandemic forced institutions to close their physical spaces, many researchers and professors were required to work from home. Media outlets and individuals wondered if women specifically would be shouldering the childcare and/or homeschooling responsibilities, decreasing the time they had for professional activities. Preliminary analyses show mixed results: some have seen decreased submissions from women (Vincent-Lamarre et al., 2020; Del Boca et al., 2020), with at least one other showing an increase in submissions from women (Dolan & Lawless, 2020); and another suggests the rate of male-athored submissions to preprint archives are growing faster than those from women (Frederickson 2020). AGU has been monitoring our monthly submission data to understand how the pandemic and stay-at home orders are affecting our submissions from specific demographics—based on author gender, age, and country. Our analysis finds that the pandemic so far has not resulted in a decrease in submissions to AGU journals; nor has there so far been a decrease in the participation of women as submitting authors. Instead, we have seen an increase in submissions from most author groups compared to year-to-date numbers in previous years.

We attempt to construct a timeline of The Hunga Tonga – Hunga Ha’apai eruption on 15 January 2022 through analyses of seismic, barometric, infrasonic, lightning, and satellite data. Satellite imagery at 04:00 UTC showed no ash in the air, but by 04:10 UTC, a plume had risen to 18 km. Over the next 20 minutes, the plume rose to 58 km. USGS determined that Mw5.8 volcanic earthquake of unknown mechanism had occurred at 04:14:45. Gravity waves were observed in satellite imagery, and barometric and infrasound stations around the world recorded ultra-low frequency pressure variations of more than 100 Pa, inducing ground-coupled airwaves around the globe, and meteo-tsunamis in the Caribbean Sea and Mediterranean Sea. Tsunami waves were recorded in coastal areas around the Pacific Ocean. From record sections, we determined speeds of 3.9 km/s and 299 m/s for the initial seismic and infrasound signals respectively, converging to an eruption onset time of ~0402 UTC ± 1 minute. The global pressure pulse has a speed of ~314 ± 3 m/s, consistent with theoretical models for Lamb waves (Bretherton, 1969), suggesting an origin time of ~0415 ± 2 minutes (consistent with the Mw5.8 volcanic earthquake, and sharp increases in lightning flash rates), and peaking around ~0429 ± 2 minutes. We suggest that Surtseyan volcanic activity commenced at ~04:02, building to a sub-Plinian eruption ~7 minutes later, before a phreato-Plinian eruption commenced at ~04:14. The peak Lamb wave amplitude at the closest station (757 km from HTHH) was 780 Pa. Assuming geometrical spreading like 1/√r (where r is the source-receiver distance), we estimate a lower bound of ~23 kPa for reduced pressure by extrapolation back to 1 km. Adding a near field term that decays like 1/r, we estimate an upper bound of 170 kPa for reduced pressure. Comparison of these values with those from other eruptions (McNutt et al. in this session) suggests the 15 January HTHH eruption was in the VEI 5-6 range.

Seismic activity during the eruption of Soufriere Hills volcano comprised various transient signals, which were classified visually by the Montserrat Volcano Observatory (MVO), considering waveforms recorded at several stations. For 217,290 transients detected on the MVO digital seismic network between 1996/10/21 and 2008/10/16, five main classes have been identified: rockfall (ROC: 58%), hybrid (HYB: 19%), long-period (LPE: 11%), lp-rockfall (LP-ROC: 5.8%), and volcano-tectonic (VT: 3.1%). Temporal trends in the rate and energy release of these different transients (in addition to swarms and tremor) were key to short-term forecasting of eruptive activity. However, visual classification is highly subjective and non-repeatable, and the inconsistency of the catalog is a barrier to research. In a pilot study, we automatically removed waveforms with dropouts, and manually verified transient classifications until we had approximately 100 transients of each class (total 522). We found ~21% of these transients were incorrectly classified at MVO. Our re-labelled dataset was then used as a starting point for supervised learning, using code from http://github.com/malfante/AAA. This code was used by Malfante et al. (2008) to classify 109,609 transients at Ubinas volcano with a 93.5% accuracy. They transformed each waveform into a set of 102 features: 34 features for each of three domains (time, spectral, cepstral). We added 6 frequency features of our own, including band ratios, peak frequency, median frequency, bandwidth, and frequency change. The resulting 108-point vectors of features were then used for modeling. The dataset is randomly divided 50 times into training and testing datasets, to produce a robust model. One model is produced per channel. We use the Random Forest Classifier algorithm from the scikit-learn library. For each waveform, a probability is computed for each class. Initial results are promising. Separate models for 3 channels yield accuracies of 76-80%. If the LP-ROC class is omitted (following Langer et al, 2006), accuracy rises to 82-85%. If only VT and LP classes are considered, accuracy is 96-99%. We intend to expand our labelled dataset to 1000 events, add new features, build models for each channel, and reclassify the catalog of 217,290 transients by a weighted average of probabilities.