86 research outputs found
Recommended from our members
Coupled subdaily and multiweek cycles during the lava dome erupt of Soufriere Hills Volcano, Montserrat
No full textObservations of volcanoes extruding andesitic lava to produce lava domes often reveal cyclic behaviour. At Soufriere Hills Volcano, Montserrat, cycles with sub-daily and multi-week periods have been recognised on many occasions. These two types of cycle have been modelled separately as stick-slip magma flow at the junction between a dyke and an overlying cylindrical conduit (Costa et al. 2012), and as the filling and discharge of magma through the elastic-walled dyke (Costa et al., 2007a) respectively. Here, we couple these two models to simulate the behaviour over a period of well-observed multi-week cycles, with accompanying sub-daily cycles, from 13 May to 21 September 1997. The coupled model captures well the asymmetrical first-order behaviour: the first 40% of the multi-week cycle consists of high rates of lava extrusion during short period/high amplitude sub-daily cycles as the dyke reservoir discharges itself. The remainder of the cycle involves increasing pressurization as more magma is stored, and extrusion rate falls, followed by a gradual increase in the period of the sub-daily cycles
A Framework for Probabilistic Multi-Hazard Assessment of Rain-Triggered Lahars Using Bayesian Belief Networks
No full textVolcanic water-sediment flows, commonly known as lahars, can often pose a higher threat to population and infrastructure than primary volcanic hazardous processes such as tephra fallout and Pyroclastic Density Currents (PDCs). Lahars are volcaniclastic flows of water, volcanic debris and entrained sediments that can travel long distances from their source, causing severe damage by impact and burial. Lahars are frequently triggered by intense or prolonged rainfall occurring after explosive eruptions, and their occurrence depends on numerous factors including the spatio-temporal rainfall characteristics, the spatial distribution and hydraulic properties of the tephra deposit, and the pre- and post-eruption topography. Modeling (and forecasting) such a complex system requires the quantification of aleatory variability in the lahar triggering and propagation. To fulfill this goal, we develop a novel framework for probabilistic hazard assessment of lahars within a multi-hazard environment, based on coupling a versatile probabilistic model for lahar triggering (a Bayesian Belief Network: Multihaz) with a dynamic physical model for lahar propagation (LaharFlow). Multihaz allows us to estimate the probability of lahars of different volumes occurring by merging varied information about regional rainfall, scientific knowledge on lahar triggering mechanisms and, crucially, probabilistic assessment of available pyroclastic material from tephra fallout and PDCs. LaharFlow propagates the aleatory variability modeled by Multihaz into hazard footprints of lahars. We apply our framework to Somma-Vesuvius (Italy) because: (1) the volcano is strongly lahar-prone based on its previous activity, (2) there are many possible source areas for lahars, and (3) there is high density of population nearby. Our results indicate that the size of the eruption preceding the lahar occurrence and the spatial distribution of tephra accumulation have a paramount role in the lahar initiation and potential impact. For instance, lahars with initiation volume ≥105 m3 along the volcano flanks are almost 60% probable to occur after large-sized eruptions (~VEI ≥ 5) but 40% after medium-sized eruptions (~VEI4). Some simulated lahars can propagate for 15 km or reach combined flow depths of 2 m and speeds of 5–10 m/s, even over flat terrain. Probabilistic multi-hazard frameworks like the one presented here can be invaluable for volcanic hazard assessment worldwide
Recommended from our members
Time series analysis of lava flux
No full textWe have applied time series analytical techniques to the flux of lava from an extrusive eruption. Tilt data acting as a proxy for flux are used in a case study of the May–August 1997 period of the eruption at Soufrière Hills Volcano, Montserrat. We justify the use of such a proxy by simple calibratory arguments. Three techniques of time series analysis are employed: spectral, spectrogram and wavelet methods. In addition to the well-known ~9-hour periodicity shown by these data, a previously unknown periodic flux variability is revealed by the wavelet analysis as a 3-day cycle of frequency modulation during June–July 1997, though the physical mechanism responsible is not clear. Such time series analysis has potential for other lava flux proxies at other types of volcanoes
The effects of thermomechanical heterogeneities in island arc crust on time-dependent preeruptive stresses and the failure of an andesitic reservoir
No full textUsing ground deformation data from Soufrière Hills volcano (SHV), we present results from numerical modeling of the temperature- and time-dependent stress evolution in a mechanically heterogeneous crust prior to reservoir failure and renewed eruptive activity. The best fit models do not allow us to discriminate between a magmatic plumbing system consisting of either a single vertically elongated reservoir or a series of stacked reservoirs. A prolate reservoir geometry with volumes between 50 and 100 km3, reservoir pressure changes between 4 and 7 MPa, and reservoir volume changes between 0.03 and 0.04 km3 with magma compressibility between 4 × 10−11 and 1 × 10−9 Pa−1 provide plausible thermomechanical model parameters to explain the deformation time series; around an order of magnitude less overpressure than is generally inferred from homogeneous, elastic crustal models. Reservoir failure is predicted to occur at the crest of the reservoir except for reservoirs with highly compressible magma ( Pa) for which subhorizontal sill formation is predicted upon reservoir failure. Introducing a deep-crustal hot zone modulates the partitioning of strains into the hotter underlying crust and results in a further reduction in overpressure estimates to values of around 1–2 MPa upon reservoir failure. Deduced volume fluxes are consistent with constraints from thermal modeling of active subvolcanic systems and imply dynamic failure of a compressible magma mush column feeding eruptions at SHV. Our interpretation of the results is that the combined thermomechanical effects of a deep-crustal hot zone and hot encasing rocks around a midcrustal andesitic reservoir fundamentally alter the time-dependent subsurface stress and strain partitioning upon reservoir priming. These effects substantially influence surface strains recorded by volcano geodetic monitoring
Windmill Islands flying bird nesting areas from field work by Jan van Franeker
No full textProgress Code: completedStatement: Each feature has a Qinfo value which can be used to access data quality information (see provided URL).This GIS dataset contains flying bird data from field work in the Windmill Islands by Jan van Franeker at Ardery Island and Odbert Island. Polygon data represents nesting areas. Point data represents nest locations on Ardery Island
Recommended from our members
Lava dome growth and mass wasting measured by a time series of ground-based radar and seismicity observations
No full textExogenous growth of Peléean lava domes involves the addition of lava from a central summit vent and mass wasting on the flanks as rockfalls and pyroclastic flows. These processes were investigated at the Soufrière Hills Volcano, Montserrat, between 30 March and 10 April 2006, using a ground-based imaging millimeter-wave radar, AVTIS, to measure the shape of the dome and talus surface and rockfall seismicity combined with camera observations to infer pyroclastic flow deposit volumes. The topographic evolution of the lava dome was recorded in a time series of radar range and intensity measurements from a distance of 6 km, recording a southeastward shift in the locus of talus deposition with time, and an average height increase for the talus surface of about 2 m a day. The AVTIS measurements show an acceleration in lava extrusion rate on 5 April, with a 2-day lag in the equivalent change in the rockfall seismicity record. The dense rock equivalent volumetric budget of lava added and dispersed, including the respective proportions of the total for each component, was calculated using: (1) AVTIS range and intensity measurements of the change in summit lava (∼1.5 × 106 m3, 22%), (2) AVTIS range measurements to measure the talus growth (∼3.9 × 106 m3, 57%), and (3) rockfall seismicity to measure the pyroclastic flow deposit volumes (∼1.4 × 106 m3, 21%), which gives an overall dense rock equivalent extrusion rate of about 7 m3·s−1. These figures demonstrate how efficient nonexplosive lava dome growth can be in generating large volumes of primary clastic deposits, a process that, by reducing the proportion of erupted lava stored in the summit region, will reduce the likelihood of large hazardous pyroclastic flows
Bayesian deep learning for spatial interpolation in the presence of auxiliary information
No full textEarth scientists increasingly deal with ‘big data’. For spatial interpolation tasks, variants of kriging have long been regarded as the established geostatistical methods. However, kriging and its variants (such as regression kriging, in which auxiliary variables or derivatives of these are included as covariates) are relatively restrictive models and lack capabilities provided by deep neural networks. Principal among these is feature learning: the ability to learn filters to recognise task-relevant patterns in gridded data such as images. Here, we demonstrate the power of feature learning in a geostatistical context by showing how deep neural networks can automatically learn the complex high-order patterns by which point-sampled target variables relate to gridded auxiliary variables (such as those provided by remote sensing) and in doing so produce detailed maps. In order to cater for the needs of decision makers who require well-calibrated probabilities, we also demonstrate how both aleatoric and epistemic uncertainty can be quantified in our deep learning approach via a Bayesian approximation known as Monte Carlo dropout. In our example, we produce a national-scale probabilistic geochemical map from point-sampled observations with auxiliary data provided by a terrain elevation grid. By combining location information with automatically learned terrain derivatives, our deep learning approach achieves an excellent coefficient of determination (R2 = 0.74) and near-perfect probabilistic calibration on held-out test data. Our results indicate the suitability of Bayesian deep learning and its feature-learning capabilities for large-scale geostatistical applications where uncertainty matters
A framework for probabilistic weather forecast post-processing across models and lead times using machine learning
No full textForecasting the weather is an increasingly data-intensive exercise. Numerical weather prediction (NWP) models are becoming more complex, with higher resolutions, and there are increasing numbers of different models in operation. While the forecasting skill of NWP models continues to improve, the number and complexity of these models poses a new challenge for the operational meteorologist: how should the information from all available models, each with their own unique biases and limitations, be combined in order to provide stakeholders with well-calibrated probabilistic forecasts to use in decision making? In this paper, we use a road surface temperature example to demonstrate a three-stage framework that uses machine learning to bridge the gap between sets of separate forecasts from NWP models and the ‘ideal’ forecast for decision support: probabilities of future weather outcomes. First, we use quantile regression forests to learn the error profile of each numerical model, and use these to apply empirically derived probability distributions to forecasts. Second, we combine these probabilistic forecasts using quantile averaging. Third, we interpolate between the aggregate quantiles in order to generate a full predictive distribution, which we demonstrate has properties suitable for decision support. Our results suggest that this approach provides an effective and operationally viable framework for the cohesive post-processing of weather forecasts across multiple models and lead times to produce a well-calibrated probabilistic output
- …
