Energetic ash-producing volcanic eruptions are driven by the diffusive and decompressive growth of bubbles (mostly water) during ascent in a magma conduit. The spatial distribution of bubble nucleation sites is one of the primary controls on ash-forming fragmentation. However, the initial formation of bubbles in a supersaturated magma is problematical, especially for homogeneous nucleation. Excessive surface tension pressure should preclude the existence of small bubbles, because exsolved water is driven back into the melt. This is the “tiny bubble paradox.” We suggest that—under special circumstances—the tiny bubble paradox may be circumvented by spinodal decomposition, a process in which uphill diffusion enables spontaneous unmixing of phases to reduce the free energy of the system. As spinodal decomposition progresses, three dimensional, quasi-spherical, zones of water-rich magma develop. These zones are characterized by an increasingly high concentration of dissolved water at the centers and reducing concentration at the margins. Bubbles are born when the concentration of water in the interior of the water-rich zones goes to 100% and the concentration of melt goes to zero. The small, nascent, bubbles that emerge will be buffered from melt by water-rich shells with increasing melt concentration away from newly formed bubbles. This diffuse concentration gradient of water means that there is no surface, per se, for surface tension to arise. This is the crux of the solution of the tiny bubble paradox. Particle morphology may be used to distinguish ash with spinodal origins from ash associated with typical (metastable) bubble nucleation. Spinodal decomposition occurs at a wavelength determined by the pressure, temperature, and viscosity of the magmatic system. This wavelength should create bubbles of uniform size and bubble walls of equal strength in a fragmenting magmatic foam, leading to sharply mono-modal vesicle and ash particle size distributions. Classical bubble nucleation should create more-variable bubble sizes and bubble wall strengths, leading to a broader particle size distribution. Better understanding the mechanism of bubble formation in magmatic systems will, in turn, enable better understanding of hazardous, explosive, eruptions.

Satellite remote sensing often requires a compromise between spatial resolution and spatial coverage for timely and accurate measurements of earth-system processes. But in recent years, increased availability of submeter-scale imagery dramatically altered this balance. Commercial satellite imagery from DigitalGlobe and Planet offer on-demand, very high-resolution panchromatic stereo and multispectral (MS) image collection over snow-covered landscapes, with individual image coverage of up to ~1900 km2. Repeat stereo-derived digital elevation models can be used to accurately estimate snow depth. Integration of contemporaneous ~1–2 m land cover classification maps can provide precise snow-covered area (SCA) products and improved processing, analysis, and interpretation of these snow depth estimates. We are developing machine learning classification algorithms to identify snow, vegetation, water, and exposed rock using varying combinations of available bands (panchromatic, 4/8-band multispectral, SWIR) and band ratios (e.g. NDVI, NDSI) from these products. We present findings for NASA SnowEx campaign sites (Grand Mesa and Senator Beck Basin, CO) and other snow monitoring sites in the Western U.S. using WorldView-3, PlanetScope, and Landsat 8 imagery. Preliminary results show that a tuned random forest algorithm using WorldView-3 MS and SWIR bands yielded the most accurate estimates of SCA of all band combinations and imagery products. With the power to resolve individual trees, these products offer direct measurements of SCA, without the need to account for mixed pixels and fractional SCA as with lower-resolution products. This open-source workflow will be used to process longer time-series and larger areas in a semi-automated fashion, allowing for rapid analysis, increased portability, and broader utility for the community.

The Java margin is part of the 5600 km long Sunda Arc subduction zone that extends from Sumatra to the Lesser Sunda Islands and is dominated by earthquakes with magnitude less than 7. Although several 7.0 ≤ Mw < 8.0 earthquakes have occurred, there has been an absence of Mw>8. Previous earthquake relocation studies have mainly focused on the ml ≥ 4 earthquakes, while seismic tomography studies have mainly focused on the volcanic system in Central and East Java. In this study, we aim to image the entirety of Java margin to investigate segmentation along this margin. We use the arrival time dataset from 2009 – 2018 collected from 44 stations in the BMKG national seismic network to relocate earthquakes and invert for seismic velocity structure along the Java margin using a finite-difference tomography algorithm. A total of 6041 earthquakes, 68250 P- and 22795 S-phases, ml 1.9 – 7.5, were included in the inversion, resulting in 4883 high-quality relocations. The distribution of relocated events shows several isolated clusters of seismicity at the trench, which are distributed nearly vertical, from the near-surface to 80 km depth. Feature with Vp/Vs ~ 1.73, which higher compared to value along the trench, coincides to one of the isolated clusters. Gaps are observed between bands of seismicity at the trench and beneath the forearc region. The seismicity is distributed surrounding or between the high anomalies of residual bathymetry which represent the structure of the subducting slab. Beneath the forearc, bands of seismicity are observed between 30 – 80 km depth and below 100 km depth. Their distribution reveals a steeper slab geometry relative to previously published slab models (Slab 1.0 and Slab 2.0). Several features with Vp/Vs < 1.70 and higher Vp and Vs than the surrounding area coincides with the bands of seismicity observed between 30 – 80 km depth. Shallow structure is also well defined by the earthquake relocations outlining three major faults (Cimandiri, Kencana-Rakutai faults, and an unnamed fault located east of Opak fault). The relocation results and velocity structure show that the distribution of seismicity along the subduction zone is segmented. This segmentation is likely related to the structure of the subducting plate.

In western North America, a recent epidemic of mountain pine beetle (Dendroctonus ponderosae) caused widespread forest mortality. This outbreak was in part due to the changing climate, and damage from pests and diseases is expected to increase in the future. To learn from this event, we used a historical retrospective approach to evaluate business-as-usual and alternative management strategies effects on tree species diversity. The insurance hypothesis proposed that ecosystems with greater species diversity will have greater productivity due to the buffering effect against natural disturbances. Therefore, we hypothesized that management strategies to increase diversity before the beetle outbreak could result in higher growing stocks, harvest rates, and net present value. The assessment was based on simulation modelling of a 1.1 million ha landscape in British Columbia, Canada for 1980–2060. We applied different strategies to affect diversity: harvest more of the most dominant tree species, planting more diverse species, and increase natural regeneration. The most aggressive strategy resulted in higher diversity and growing stocks, higher harvest rates, and higher, more consistent net revenue over time than the business-as-usual strategy. However, the strategy that only employed a diversity of planting negatively affected those indicators. Thus, we have identified limitations to what management strategies may be able to achieve. Sensitivity analyses of species productivity and log price indicated a high level of robustness in the results. Our study showed that reducing forest health risks may be economically viable.

Coupled reservoir flow and geomechanics receives a growing research interests for applications in unconventional oil & gas recovery and geological CO2sequestration. For CO2 sequestration, accurate prediction of the behavior of the injected CO2 is important to the long-term success of the sequestration because even small leakage rates over long time periods can unravel the positive outcomes of net CO2 sequestered. Furthermore, the effect of high rates of injection of CO2 on natural fractures is crucial in determination of earthquake effects. High fidelity multiphysics models are needed to simulate these processes and their interactions accurately and efficiently to ensure the success and safety of the operations. Coupled flow-geomechanics simulations are computationally expensive and most of the computational time is usually spent on geomechanics updates. We present a three-way coupling algorithm for both single phase flow and compositional flow coupled with linear poroelasticity. An error indicator is calculated at each flow time step to determine when displacement must be updated and whether fixed-stress iterative coupling technique is required. Convergence of the three-way coupling is established by extending previous work on a priori analyses of fixed-stress iterative coupling. Numerical results for Mandel’s problem confirm these theoretical results for single phase flow. Numerical results for coupled compositional flow and geomechanics simulations for field-scale CO2sequestration and surfactant-alternating-gas (SAG)-assisted CO2 sequestration achieve a substantial reduction in mechanics update times for 99.4% and 97.5%, respectively.

Atmospheric carbon monoxide (CO) has been decreasing globally for the last two decades. Recently, positive fire trends in Northern Hemisphere boreal regions may have impacted the decreasing CO. Additionally, time-varying air quality policies will have different impacts on atmospheric composition and related trends. Aerosols are co-emitted with CO from both fires and anthropogenic sources. Consequently, a combined trend analysis of CO and aerosol optical depth (AOD) measurements from space can help elucidate the drivers of regional differences in the CO trend. We use valuable long-term records from two instruments aboard the Terra satellite. Measurements of Pollution in the Troposphere (MOPITT) CO and AOD from the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument are examined hemispherically and in sub-regions to determine trends between 2002 and 2018. The records are further split into two sub-periods in order to examine temporal stability in the trend values. We also assess the CO trends in monthly percentile values to use seasonal information when interpreting trend contributions. Our focus is on four major population centers: Southeast USA, Europe, Northeast China and North India, as well as biomass burning regions in both hemispheres. Our results show that globally, CO declines faster in the first half of the record compared to the second half. Both atmospheric species are important when interpreting trends in the smaller regions. Northern Hemisphere boreal fire regions show a regime-shift in their seasonality for both CO and AOD, which may counteract the downward trend in CO. Anthropogenic regions with minimal air quality management such as North India become more globally relevant as the global CO trend weakens. We also find clear evidence of the atmospheric impact of policy choices. Overall, we observe that local changes in biomass burning and air quality can counteract the global downward trend in CO.

During the last decade, wildfires in the Aburrá Valley watershed, located in northwestern Colombia, have caused significant forest and ecosystem losses, health issues in nearby communities associated with aerosols from biomass burning, and increases in the CO2 emissions. Human activities, along with weather variability, modulate the occurrence of forest fires during the dry seasons, and the efforts to reduce them have shown limited success, highlighting the need for the development of holistic prevention strategies. We implemented a general strategy involving real-time monitoring, modeling, and warning based on a distributed Bayesian model coupled with a distributed hydrological model and a regional weather model (WRF) to estimate wildfire susceptibility in the basin. The model operates with a spatial resolution of 60m and an hourly temporal resolution. The model uses static and time-dependent (dynamic) information. Static variables include land use, urban-rural fringe area, historical fire occurrence, and are updated occasionally. The dynamic variables change at each time step, and they depend on meteorological conditions and include soil moisture, cumulative rainfall during the last ten days, and an estimation of the surface temperature. These variables are obtained from in-situ rain gauges and quantitative precipitation estimation (QPE) techniques using C-band weather radar reflectivity, in-situ pyranometers and automatic weather stations, and output from a distributed hydrological model and WRF-based weather forecasts. The Bayesian model allows the generation of fire susceptibility predictions that help optimize prevention strategies implemented by the fire departments in the region. The model has been evaluated using the location of historical wildfires showing high skill. Along with the model, there are efforts in the region implemented for early-detection, and quantification of forest fires using in-situ and drone-borne thermal and high-definition cameras, a continuous monitoring strategy is established.

The Clouds and the Earth’s Radiant Energy System (CERES) project monitors Earth’s long-term energy balance and produces data products which have improved our understanding of the role clouds and aerosols play in that balance. Cloud property retrievals from the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Visible Infrared Imaging Radiometer Suite (VIIRS) are a major component of many of these data products. It is important for cloud property retrievals to be consistent over the course of the record so that artificial discontinuities are not introduced into the Earth radiation budget record. In practice, the MODIS and VIIRS instruments have different characteristics and different sets of spectral bands so deriving completely consistent cloud properties from the two instruments is a complex task. This paper investigates differences in the cloud properties retrieved from MODIS and VIIRS using the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) instrument as an independent validation source. Particular consideration is given to CALIOP’s sensitivity to optically thin clouds and the effect these clouds have on the retrieved cloud properties, especially cloud thermodynamic phase. Differences in cloud phase and cloud optical depth from MODIS and VIIRS are characterized by different cloud types including multi-layer scenarios. Characterizing these retrieval differences will help understand and mitigate artifacts in the long-term record.

The generalized solution for the radiance equation is expanded by exploiting multiple hyperspectral image scans acquired by aerial platforms at different viewing angles. A machine learning solution based on convolutional neural networks is used to learn the relationships between the total radiance observed at the sensor, and different atmospheric components of the radiance equation. The goal is to precisely characterize the atmosphere, in order to properly solve the radiance equation, in which atmospheric components constitute important input. Traditionally, these atmospheric components are only estimated from averages of pixels, or assigned using heuristics tables. Compared to traditional image spectroscopy, this expanded radiance equation and machine learning solution integrates quantitative mathematical modeling, multiple scanned hyperspectral images and artificial intelligence. The solution is able to model and predict the transmittance, downwelling and upwelling components of the radiance equation with increased spatial and temporal dimensionality. It’s promising to use different combinations of the multiple scans to parameterize the radiance equation and improve the target detection in varying atmospheric conditions, where current solutions based on a single hyperspectral image normally fail. This works presents initial results of an expanded mathematical solution, along with the results from the convolution networks. Synthetic data were generated using the MODTRAN atmospheric software to simulate different vintage points, atmospheric models, time of the day and year, for an array of specific targets with varying reflectances. More specifically, MODTRAN was used to simulate Longwave Infrared Red between 7.5 and 12 microns with a 17.5 nanometers spectral sensitivity, which correspond to the range and resolution of the Blue Heron Longwave Hyperspectral sensor. Results from the convolutional neural network indicate our machine learning solution is computationally faster than the traditional radiative transfer (RT) model and is able to characterize the impact of varying atmospheric conditions on the at-sensor radiance components.

Warming climate trends will shift the soil thermal gradient and thus impact the soil-nutrient dynamics. This impact will be more pronounced in areas where air temperatures have been significantly less than those of the vadose zone. It has been observed that the soil nitrogen content in the form of nitrate and ammonium has increased in water bodies with an increase in temperature. This observation can be owed to the reason that the denitrification process is significantly affected by warm temperatures. Nitrogen content in the soil thus can be expected to change due to increasing temperatures. The present study attempts to find the effect of air temperature on soil root zone properties (viz. soil temperature and soil moisture content at different depths) consequently affecting soil-nutrient status. A support vector regression (SVR) model was trained to develop a regression between air temperature and soil nitrogen status at different soil moisture contents. The model was applied to simulate the future (i.e. years 2030 and 2050) soil nitrogen content at different depths for different soil temperatures and moisture contents. The effect of increasing temperature on the fate of soil nitrogen was predicted based on the optimal temperature range required for the chemical reactions. The model used the data from experimental watershed Bear Brook Watershed, Maine and soft data from the Global Land Data Assimilation System (GLDAS) for prediction and validation purposes.

In this study, we have applied to northeastern Italy and western Slovenia medium-low seismicity an algorithm for strong aftershock forecasting we originally developed for medium-high seismicity in Italy (Gentili and Di Giovambattista, 2017). This study is possible thanks to the OGS bulletins, an accurate local catalogue, characterized by low completeness magnitude, that has been compiled by the National Institute of Oceanography and Experimental Geophysics, Centre of Seismological Research, since 1977. The method is based on a pattern recognition approach which uses statistical features based on the number of the early aftershocks and on the spatio-temporal evolution of the radiated energy in the first hours/days after the mainshock. The analysis was performed on different time-spans after the mainshock to simulate the increase of information available as time passes during the seismic clusters. Following the approach of Gentili and Giovambattista (2017), we used an operational definition of clusters that defines “mainshock” as the first shock of the cluster over a given threshold. We have adopted this criterion in order to be able to apply the procedure immediately after the occurrence of a shock without waiting to verify if a stronger earthquake followed. If the difference in magnitude between the “mainshock” and the strongest aftershock is lower than 1 the clusters are classified as “Type A”, otherwise as “Type B”,. The Type A and B clusters’ distribution is analyzed also considering a draft subdivision of the region into two sub-regions characterized by different complexity of the clusters (Peresan and Gentili, 2018) Vp/Vs (Bressan et al., 2012) and attenuation characteristcs (Gentili and Gentile, 2015).

A new diver-held magnetometer has been developed that directly senses the total magnetic gradient as a scalar quantity, independent of the orientation of the device. It therefore provides a direct signal if a magnetic (e.g. ferrous) object is in its vicinity, regardless of other ambient geomagnetic effects. The small (7cm diameter x75cm length), battery-powered device is neutrally buoyant, and provides a simple and intuitive audio interface to the diver. The device was field tested in July 2019 during the Lost Ships of Cortés Project offshore Villa Rica, Mexico, an area with strong underlying volcanic geology and large quantities of magnetic sediment that mask buried archaeological materials. Magnetic anomalies were recorded during boat-towed total-field magnetometer surveys, but divers could not identify their source using traditional underwater metal detectors. The new magnetometer decisively and precisely guided divers to an early 16th-century anchor at a range of several meters that was completely buried in magnetite-rich sand with its shallowest point at 50cm below the sea floor.

The full-featured and CoreTrustSeal certified long term archiving service LTA WDCC (World Data Centre for Climate) at DKRZ (German Climate Computing Center, Hamburg) offers long term preservation for datasets relevant for climate and Earth System research. The WDCC collects, stores, and disseminates Earth System data with a focus on climate simulation data and climate related data products. It has established itself as a staple infrastructure for the global climate modelling research community. Data preservation in LTA WDCC is preceded by a thorough technical quality control and provides intense data curation for storage periods longer than 10 years. During the preservation period, long term findability, searchability and reusability of the data are ensured. Long term findability of the curated data is enabled through the possibility of assigning DataCite DOI’s to archived datasets. The data undergo additional quality checks before being eligible for DOI assignment. This process is performed in close collaboration with the data providers. The focus of these quality checks is to ensure the unambigous (inter-)disciplinary reusability of the preserved datasets and includes checking for proper documentation, adherence to domain specific (meta)data standards, uncertainty analysis and cross-referencing. Only then can a high level of reusability of the data be achieved, justifying the involved effort. The perceived need for research data repositories to comply with the 2016-published FAIR Guiding Principles has motivated us to perform an even-handed and systematic self assessment of LTA WDCC FAIRness. Due to lack of a standardised evaluation framework, this assessment reflects our specific, albeit objective, interpretation of the principles. Our assessment, published on the DKRZ webpages, shows that the native philosophy behind DKRZ’s LTA WDCC service – especially the focus on reusability – reflects the FAIR Guiding Principles by design and even goes beyond them by ensuring very long-term (>10 years) preservation and therefore reusability of archived data.

Many high value data products made available from agencies such as NASA, NOAA, and USGS are broadcast in the clear directly from the satellite. Traditional access patterns involve downlinking the data to an agency ground segment, processing by a science team, and subsequent product distribution through a program such as the NOAA Big Data Initiative or a NASA Distributed Active Archive Center (DAAC). This talk demonstrates the ability to acquire, process, and analyze open broadcast science data entirely in the cloud. Leveraging the new AWS Elastic Ground Station service, we demonstrate selecting and scheduling satellite acquisitions for open science data then leveraging the scalable nature of the cloud to execute a full processing stack. This talk demonstrates the ability to acquire L0 data, leveraging open data processing algorithms from agencies like NASA and NOAA, and a highly scalable, cost effective, and secure processing architecture to produce near real-time data. By taking advantage of direct satellite acquisition and cloud scale processing, we demonstrate reduced latency from acquisition to data availability for scenarios like disaster response, tipping and cueing, and location monitoring. The deployed cloud stack leverages open source components and demonstrates ingest and archival of L0 data, processing to L1 products and above, horizontally scaling data processing algorithms, metadata generation, and search and discovery via APIs and a simple serverless UI.

Seismic time-lapse studies are generally concerned with variations in a specific target zone, situated inside an otherwise static medium. In seismic monitoring the entire reflection response at the surface needs to be remodeled for every change in the target zone. Ideally, however, only the response of the target zone is remodeled, which is then inserted into the stationary response of the surrounding medium. Such an insertion is relatively straightforward when solely primary reflections are considered, but more complicated when internal multiples are introduced. The data-driven Marchenko method retrieves the responses to virtual sources inside the subsurface, and fully accounts for all orders of multiple scattering. Wapenaar and Staring (2018), therefore, use the method to predict the reflection response at the Earth’s surface for different target-zone scenarios, from a single reflection response at the surface and a model of the changing target zone. This is achieved in two steps. First, the responses of the static overand underburden are resolved from the reflection response. Second, the newly modeled response from the target zone is inserted. While this methodology was originally only tested for laterally invariant media, we show how it can be extended for laterally inhomogeneous media. A 2D model is considered to compare the performance of the Marchenko-based replacement with a remodeled response of the entire medium. The results show that the method accurately captures the different arrivals in the reflection response, including the multiples, aside from minor differences in amplitude. We conclude that the method can be used to more efficiently model reflection responses of variable target zones, therefore, enabling more effective modeling for time-lapse monitoring.

For over a century, the enigmatic Carolina bays have captivated geologists and spurred contentious debate on their origin. These circular to ovate and shallow (median diameter of 222 m, median depth of 2.17 m, median area of 26,249 sq. m) depressions span the Atlantic Coastal Plain (ACP) from northern Florida to southern New Jersey, with total counts ranging between 10,000 and 500,000. Using 1 meter gridded, 1.7 km by 1.7 km LiDAR digital elevation models (DEMs) of Delaware as training images, a convolutional neural network (CNN) was trained to detect Carolina bays. With such a large population size and with such uncertainty around the actual population size, mapping the Carolina bays is a problem that requires an automated detection scheme. Manual detection of bays from LiDAR across the entire Atlantic Coastal Plain would be extremely time intensive and prone to human annotation errors. Using Faster R-CNN within the TensorFlow Python library, a network was trained on 978 LiDAR images for 24 hours (42,450 iterations) on an Intel Core i7-4790K CPU at 4.00 GHz. This network automatically detects bays from LiDAR images with a bounding box and a confidence level. These bounding boxes can then be used to subset and then analyze regions of the DEM for statistics on the bays’ three-dimensional shape. Extending this algorithm to DEMs from other areas of the ACP will provide a better understanding of the bays’ geographic distribution as well as any differences in morphology between different geographic regions. This method for detecting geomorphic features is a highly efficient process that will provide better means for mapping various types of abundant geomorphic features in the future.

Many studies have shown that the characteristics of lightning such as size and peak current differ between ocean and continental thunderstorms. In general, the lightning in oceanic thunderstorms are larger and have higher peak current than in continental thunderstorms. The reason for these differences have been stipulated to be related to differences in thermal properties and aerosol concentration, however, there is still disagreement over which is the dominant mechanism. In this study, we focus on how thunderstorm trends, as opposed to lightning trends (i.e., flash density), are affected by these mechanisms. We develop a lightning clustering algorithm that takes individual lightning strokes and creates thunderstorms based on their spatiotemporal proximity. We use lightning data from the Earth Networks Total Lightning Network and compare storms from the Eastern U.S.A. to storms off that coast, where the detection efficiency of the network is still good. Once these thunderstorms are obtained, we can split them into ocean and land thunderstorms and compare various characteristics (size, duration, flash rate, polarity and IC/CG ratio, etc.) to determine if any differences stand out. In this presentation, we will discuss the clustering algorithm used, analyze the results of the study, and discuss implications.

As a multi-year coding in the classroom initiative, the Earth Science Information Partners’ education committee and community members have been exploring how to integrate coding skills and computational thinking within an Earth science context. This presentation describes a Jupyter notebook project that resulted in a professional development Data to Action Education Workshop during the ESIP 2019 Summer Meeting in Tacoma, Washington. The main goal of the workshop was to provide K-12 teacher participants the opportunity to learn how modify and run code that would display and analyze data about hurricanes in the Atlantic basin using a Jupyter Notebook. The topic was chosen for two specific reasons. Middle- and high- school Earth Science curricula include a learning unit about weather, particularly hurricanes, as a topic of instruction thus the content could be a natural fit into most classrooms. Secondly, employing a familiar content area in a coding-dedicated workshop meant that teachers could focus on the code itself and bring their knowledge about hurricanes to create new questions to explore. Workshop participants included in-service teachers currently instructing in middle and/or high school geoscience or technology courses, most of whom had very little or no background in programming. Teachers became familiar with the Jupyter environment, gained experience in interpreting coding cell content, and built confidence in manipulating code and parameters to develop graphical outputs. Through merging coding practice applied to learning instructional content, teachers were able to experiment and succeed (or fail) in a safe environment, practicing the skills their students need to learn. By encouraging teachers to learn how to code, we help them to foster their students curiosity to be creators, more than just consumers, of the technology around us and develop skills paramount to be career- and college-ready. We discuss the pedagogical and technical approaches and lessons learned through the design, development, implementation, and evaluation phases of this project.