We present a deep-learning approach for earthquake detection using waveforms from a seismic array consisting of multiple seismographs. Although automated, deep-learning earthquake detection techniques have recently been developed at the single-station level, they have potential difficulty in reducing false detections owing to the presence of local noise inherent to each station. Here, we propose a deep-learning-based approach to efficiently analyze the waveforms observed by a seismic array, whereby we employ convolutional neural networks in conjunction with graph partitioning to group the waveforms from seismic stations within the array. We then apply the proposed method to waveform data recorded by a dense, local seismic array in the regional seismograph network around the Tokyo metropolitan area, Japan. Our method detects more than $97$ percent of the local seismicity catalogue, with less than $4$ percent false positive rate, based on an optimal threshold value of the output earthquake probability of $0.61$. A comparison with conventional deep-learning-based detectors demonstrates that our method yields fewer false detections for a given true earthquake detection rate. Furthermore, the current method exhibits the robustness to poor-quality data and/or data that are missing at several stations within the array. Synthetic tests demonstrate that the present method has the potential to detect earthquakes even when half of the normally available seismic data are missing. We apply the proposed method to analyze 1-hour-long continuous waveforms and identify new seismic events with extremely low signal-to-noise ratios that are not listed in existing catalogs. (241words)
We use a recently developed spectrally resolved bio-optical module to better represent the interaction between the incoming irradiance and the heat fluxes in the upper ocean within the (pre-)operational physical-biogeochemical model on the North-West European (NWE) Shelf. The module attenuates light based on the simulated biogeochemical tracer concentrations, and thus introduces a two-way coupling between the biogeochemistry and physics. We demonstrate that in the late spring-summer the two-way coupled model heats up the upper oceanic layer, shallows the mixed layer depth and influences the mixing in the upper ocean. The increased heating in the upper oceanic layer reduces the convective mixing and improves by ~5 days the timing of the late phytoplankton bloom of the ecosystem model. This improvement is relatively small compared with the existing model bias in bloom timing, but sufficient to have a visible impact on model skill. We show that the changes to the model temperature and salinity introduced by the module have mixed impact on the physical model skill, but the skill can be improved by assimilating the observations of temperature, salinity and chlorophyll concentrations into the model. However, in the situations where we improved the simulation of temperature, either via the bio-optical module, or via assimilation of temperature and salinity, we have shown that we also improved the simulated oxygen concentration as a result of the changes in the simulated air-sea gas flux. Overall, comparing different 1-year experiments showed that the best model skill is achieved with joint physical-biogeochemical assimilation into the two-way coupled model.
Short term aftershock incompleteness (STAI) can strongly bias any analysis built on the assumption that seismic catalogs have a complete record of events. Despite several attempts to tackle this issue, we are far from trusting any dataset in the immediate future of a large shock occurrence. Here we introduce RESTORE (REal catalogs STOchastic REplenishment), a Python toolbox implementing a stochastic gap-filling method, which automatically detects the STAI gaps and reconstructs the missing events in the space-time-magnitude domain. The algorithm is based on empirical earthquake properties and relies on a minimal number of assumptions about the data. Through a numerical test, we show that RESTORE returns an accurate estimation of the number of missed events and correctly reconstructs their magnitude, location and occurrence time. We also conduct a real-case test, by applying the algorithm to the Mw 6.2 Amatrice aftershocks sequence. The STAI-induced gaps are filled and missed earthquakes are restored in a way which is consistent with data. RESTORE, which is made freely available, is a powerful tool to tackle the STAI issue, and will hopefully help to implement more robust analyses for advancing operational earthquake forecasting and seismic hazard assessment.
Identifying the main environmental drivers of SARS-CoV-2 transmissibility in the population is crucial for understanding current and potential future outbursts of COVID-19 and other infectious diseases. To address this problem, we concentrate on basic reproduction number R0, which is not sensitive to testing coverage and represents transmissibility in an absence of social distancing and in a completely susceptible population. While many variables may potentially influence R0, a high correlation between these variables may obscure the result interpretation. Consequently, we combine Principal Component Analysis with feature selection methods from several regression-based approaches to identify the main demographic and meteorological drivers behind R0. We robustly obtain that country’s wealth/development (GDP per capita or Human Development Index) is by far the most important R0 predictor, probably being a good proxy for the overall contact frequency in a population. This main effect is modulated by built-up area per capita (crowdedness in indoor space), onset of infection (likely related to increased awareness of infection risks), net migration, unhealthy living lifestyle/conditions including pollution, seasonality, and possibly BCG vaccination prevalence. Also, we show that several variables that significantly correlate with transmissibility do not directly influence R0 or affect it differently than suggested by naive analysis.
The higher frequency and intensity of sustained heat events have increased the demand for cooling energy across the globe. Current estimates of summer-time energy demand are primarily based on Cooling Degree Days (CDD), representing the number of degrees a day’s average temperature exceeds a predetermined comfort zone temperature. Through a comprehensive analysis of the historical energy demand data across the USA, we show that the commonly used CDD estimates fall significantly short (±25%) of capturing regional thermal comfort levels. Moreover, given the increasingly compelling evidence that air temperature alone is not sufficient for characterizing human thermal comfort, we extend the widely-used CDD calculation to heat index, which accounts for both air temperature and humidity. Our results indicate significant mis-estimation of regional thermal comfort when humidity is ignored. Our findings have significant implications for the security, sustainability, and resilience of the grid under climate change.
While the Madden Julian Oscillation (MJO) is known to influence the midlatitude circulation and its predictability on subseasonal-to-seasonal (S2S) timescales, little is known how this connection may change with anthropogenic warming. This study investigates changes in the causal pathways between the MJO and the North Atlantic Oscillation (NAO) within historical and SSP585 simulations of the CESM2-WACCM coupled climate model. Two data-driven approaches are employed, namely, the STRIPES index and graphical causal models. These approaches collectively indicate that the MJO’s influence on the North Atlantic strengthens in the future, consistent with an extended jet-stream. In addition, the graphical causal models allow us to distinguish the causal pathways associated with the teleconnections. While both a stratospheric and tropospheric pathway connect the MJO to the North Atlantic in CESM2-WACCM, the strengthening of the MJO-NAO causal connection over the 21st century is shown to be due exclusively to teleconnections via the tropospheric pathway.
Groundwater is the largest source of Earth’s liquid freshwater and plays a critical role in global food security. With the rising global demand for drinking water and increased agricultural production, overuse of groundwater resources is a major concern. Because groundwater withdrawals are not monitored in most regions with the highest use, methods are needed to monitor withdrawals at a scale suitable for implementing sustainable management practices. In this study, we combine publicly available datasets into a machine learning framework for estimating groundwater withdrawals over the state of Arizona. This extends a previous study in which we estimated groundwater withdrawals in Kansas, where the climatic conditions and aquifer characteristics are significantly different. Datasets used in our model include energy-balance (SSEBop) and crop coefficient evapotranspiration estimates, precipitation(PRISM), and land-use (USDA-NASS Cropland Data Layer), and a watershed stress metric. Random forests, a widely popular machine learning algorithm, are employed for predicting groundwater withdrawals from 2002-2018 at 5 km spatial resolution. We used in-situ groundwater withdrawals available over the Arizona Active Management Area (AMA) and Irrigation Non-Expansion Area (INA) from 2002-2010 for training and 2011-2018 for validating the model respectively. The results show high training (R2 ≈ 0.98) and good testing (R2 ≈ 0.82) scores with low normalized mean absolute error ≈ 0.28 and root mean square error ≈ 1.28 for the AMA/INA region. Using this method, we are able to spatially extend estimates of groundwater withdrawals to the whole state of Arizona. We also observed that land subsidence in Arizona is predominantly occurring in areas having high yearly groundwater withdrawals of at least 100 mm per unit area. Our model shows promising results in sub-humid and semi-arid (Kansas) and arid regions (Arizona), which proves the robustness and extensibility of our integrated approach combining remote sensing and machine learning into a holistic, automated, and fully-reproducible workflow. The success of this method indicates that it could be extended to areas with more limited groundwater withdrawal data under different climatic conditions and aquifer properties.
Use of Real-Time Control (RTC) technology in Rainwater Harvesting Systems (RWH) can improve performance across water supply, flood protection, and environmental flow provision. Such systems make the most of rainfall forecast information, to release water prior to storm events and thus minimise uncontrolled overflows. To date, most advanced applications have adopted 24-hr forecast information, leaving longer-term forecasts largely untested. In this study, we aimed to predict the performance of four different RTC strategies, based on different forecast lead-time and preferred objectives. RTC systems were predicted to yield comparatively slightly less harvested rainwater than conventional passive systems, but delivered superior performance in terms of flood mitigation and delivery of environmental water for streamflow restoration. More importantly, using a 7-day rainfall forecast, the longest commercially available prediction window, was shown to enhance the ability of RTC in mitigating flood risks and delivering an outflow regime that is close to the natural (reference) streamflow. Such a finding suggests that RTC combined with 7-day forecast can enhance the functionality of rainwater harvesting systems to restore and even mimick the entire natural flow regimes in receiving streams. This also opens up a new opportunity for practitioners to implement smart technology in managing urban stormwater in a range of contexts and for a range of stream health objectives.
We present a fast model for stratospheric ozone chemistry based on a neural network approach. The model is intended to replace the detailed chemistry schemes of chemistry and transport models (CTMs), general circulation models (GCMs) or Earth system models (ESMs), which are computationally very expensive. The neural network (NN) model estimates the rate of change of ozone in 24 hours at a grid point and is trained on data of the detailed full chemistry model of the ATLAS chemistry and transport model (CTM). The benefit of this surrogate models is a much lower computation time (minutes instead of days) while the same level of accuracy is achieved. This represents a necessary step from understanding the chemistry and building sophisticated CTMs towards the usage of this knowledge in climate models, which is only feasible if much lower computation times can be achieved. Modelling of the Earth system is a complex task and models usually contain a large number of sub-modules and parameterizations. This applies for example to the atmosphere, hydrosphere, solid earth and the ice sheets. Atmospheric chemistry is complex and usually involves dozens of species and hundreds of reactions with a wide range of concentrations and lifetimes. This project concentrates on the estimation of the rate of change of ozone in the extrapolar stratosphere. The dynamics from the polar regions and from other layers of the atmosphere regarding the ozone change are not treated within this work. The ATLAS model is a Lagrangian CTM for stratospheric chemistry. It solves a coupled differential equation system using a stiff solver and a variable time-step. The stratospheric chemistry scheme of ATLAS has 46 active species, 171 reactions and heterogeneous chemistry on polar stratospheric clouds. It is not using the concept of chemical families. The application of the ATLAS CTM has high requirements on computational power. This is the reason why the coupling of full chemical models to climate models is generally not feasible with respect to the computation time of a global climate model. However, the incorporation of detailed chemistry is often desirable, in order to account for various feed-backs between chemistry, atmosphere and ocean. These complex chemical models motivate the formulation of faster but still accurate surrogate models, that are tailored to the coupling into earth climate models. This project builds on the SWIFT project, which has a polar and an extrapolar surrogate model for the stratospheric ozone chemistry. We investigate an alternative approach to the polynomial approach used by extrapolar SWIFT by exploiting the improved approximation capability of NN with respect to nonlinear contexts.
Because cross-polarized radar returns are highly associated with volume scatter, radar polarimetry returns tend to show strong evidence of wildfire scars and recovery in forest and chaparral. We focus on the polarimetry images from UAVSAR line SanAnd_08525, which covers a roughly 20 km wide swath over the Transverse Range including parts of the Santa Monica, San Gabriel and San Bernardino Mountains. We select images from four acquisition dates from October 2009 to September 2020, very roughly four years apart. These are compared to fire perimeters from the national GeoMAC and NIFC databases for years 2003-2020, which shows the areas affected by the major fires (west to east) Springs2013, Woolsey2018, Topanga2005, LaTuna2017, Station2009, BlueCut2016, Pilot2016, Slide2007, Butler2007 and many smaller fires. UAVSAR polarimetry images are shown to be helpful in identifying types and boundaries of fire, fifty-meter scale details of vegetation loss, and variability of vegetation recovery in post-fire years.
This study introduces the results from fitting a Bayesian hierarchical spatiotemporal model to COVID-19 cases and deaths at the county-level in the United States for the year 2020. Two models were created, one for cases and one for deaths, utilizing a scaled Besag, York, Mollié model with Type I spatial-temporal interaction. Each model accounts for 16 social vulnerability variables and 7 environmental measurements as fixed effects. The spatial structure of COVID-19 infections is heavily focused in the southern U.S. and the states of Indiana, Iowa, and New Mexico. The spatial structure of COVID-19 deaths covers less of the same area but also encompasses a cluster in the Northeast. The spatiotemporal trend of the pandemic in the U.S. illustrates a shift out of many of the major metropolitan areas into the U.S. Southeast and Southwest during the summer months and into the upper Midwest beginning in autumn. Analysis of the major social vulnerability predictors of COVID-19 infection and death found that counties with higher percentages of those not having a high school diploma and having minority status to be significant. Age 65 and over was a significant factor in deaths but not in cases. Among the environmental variables, above ground level (AGL) temperature had the strongest effect on relative risk to both cases and deaths. Hot and cold spots of COVID-19 cases and deaths derived from the convolutional spatial effect show that areas with a high probability of above average relative risk have significantly higher SVI composite scores. Hot and cold spot analysis utilizing the spatiotemporal interaction term exemplifies a more complex relationship between social vulnerability, environmental measurements, and cases/deaths.
Hydrologic variability can present severe financial challenges for organizations that rely on water for the provision of services, such as water utilities and hydropower producers. While recent decades have seen rapid growth in decision-support innovations aimed at helping utilities manage hydrologic uncertainty for multiple objectives, support for managing the related financial risks remains limited. However, the mathematical similarities between multi-objective reservoir control and financial risk management suggest that the two problems can be approached in a similar manner. This paper demonstrates the utility of Evolutionary Multi-Objective Direct Policy Search (EMODPS) for developing adaptive financial risk management policies in the context of hydropower production in a snow-dominated region. These policies dynamically balance a portfolio, consisting of snowpack-based financial hedging contracts, cash reserves, and debt, based on evolving system conditions. Performance is quantified based on four conflicting objectives, representing the classic tradeoff between “risk” and “return” in addition to decision-makers’ unique preferences towards different risk management instruments. The dynamic policies identified here significantly outperform static management formulations that are more typically employed for financial risk applications in the water resources literature. Additionally, this paper combines visual analytics and information theoretic sensitivity analysis to help decision-makers better understand how different candidate policies achieve their comparative advantages through differences in how they adapt to real-time information. The methodology developed in this paper should be applicable to any organization subject to financial risk stemming from hydrology or other environmental variables (e.g., wind speed, insolation), including electric utilities, water utilities, agricultural producers, and renewable energy developers.
The steepness of the beach face is a fundamental parameter for coastal morphodynamic research. Despite its importance, it remains extremely difficult to obtain reliable estimates of the beach-face slope over large spatial scales (1000’s of km of coastline). In this letter, a novel approach to estimate this slope from time-series of satellite-derived shoreline positions is presented. This new technique uses a frequency-domain analysis to find the optimum slope that minimises high-frequency tidal fluctuations relative to lower-frequency erosion/accretion signals. A detailed assessment of this new approach at 8 locations spanning a range of tidal regimes, wave climates and sediment grain sizes shows strong agreement (R = 0.9) with field measurements. The automated technique is then applied across 1000’s of beaches in eastern Australia and California USA, revealing similar regional-scale distributions along these two contrasting coastlines and highlights the potential for new global-scale insight to beach-face slope spatial distribution, variability and trends.
The sustainable management of groundwater demands a faithful characterization of the subsurface. This, in turn, requires information which is generally not readily available. To bridge the gap between data need and availability, numerical models are often used to synthesize plausible scenarios not only from direct information but also additional, indirect data. Unfortunately, the resulting system characterizations will rarely be unique. This poses a challenge for practical parameter inference: Computational limitations often force modelers to resort to methods based on questionable assumptions of Gaussianity, which do not reproduce important facets of ambiguity such as Pareto fronts or multi-modality. In search of a remedy, an alternative could be found in Stein Variational Gradient Descent, a recent development in the field of statistics. This ensemble-based method iteratively transforms a set of arbitrary particles into samples of a potentially non-Gaussian posterior, provided the latter is sufficiently smooth. A prerequisite for this method is knowledge of the Jacobian, which is usually exceptionally expensive to evaluate. To address this issue, we propose an ensemble-based, localized approximation of the Jacobian. We demonstrate the performance of the resulting algorithm in two cases: a simple, bimodal synthetic scenario, and a complex numerical model based on a real-world, pre-alpine catchment. Promising results in both cases - even when the ensemble size is smaller than the number of parameters - suggest that Stein Variational Gradient Descent can be a valuable addition to hydrogeological parameter inference.
Journals, funding agencies, and researchers are more frequently expecting manuscripts to include links to shared research data. Effective data sharing requires that data be findable, accessible, interoperable, and reusable (FAIR), and is thus predicated on establishing a common understanding on how to communicate: data exchange standards, common data formats, controlled vocabularies, and a communal data repository. When conducting research, we still communicate in shorthand that is effective for everyone on the team who understands our context, but is lost when data is shared in the absence of that context. “Water temperature” means only one thing to my research team, yet can mean dozens of things outside of that context. Data sharing is thus an exercise in sharing not just the data, which is typically readily available, but also the context of that data, which requires additional effort. This effort is one of the barriers to sharing data. We’ll describe an alternative model for accepting data to a repository: the immediate ingestion of data regardless of its metadata quality, then behavioural nudges and crowd-sourcing features that ensure this data meets appropriate standards prior to publication. We’ll show a work-in-progress prototype software tool that supports this alternative model, capable of accepting and standardizing a research data set to use CF conventions and ISO 8601 dates.
Understanding future land-use related water demand is important for planners and resource managers in identifying potential shortages and crafting mitigation strategies. This is especially the case for regions dependent on limited local groundwater supplies. For the groundwater dependent Central Coast of California, we developed two scenarios of future land use and water demand based on sampling from a historic land change record: a business-as-usual scenario (BAU; 1992–2016) and a recent-modern scenario (RM; 2002–2016). We modeled the scenarios in the stochastic, empirically based, spatially explicit LUCAS state-and-transition simulation model at a high resolution (270-m) for the years 2001-2100 across 10 Monte Carlo simulations, applying current land zoning restrictions. Under the BAU scenario, regional water demand increased by an estimated ~ 222.7 Mm by 2100, driven by the continuation of perennial cropland expansion as well as higher than modern urbanization rates. Since 2000, mandates have been in place restricting new development unless adequate water resources could be identified. Despite these restrictions, water demand dramatically increased in the RM scenario by 310.6 Mm by century’s end, driven by the projected continuation of dramatic orchard and vineyard expansion trends. Overall, increased perennial cropland leads to a near doubling to tripling perennial water demand by 2100. Our scenario projections can provide water managers and policy makers with information on diverging land use and water use futures based on observed land change and water use trends, helping better inform land and resource management decisions.
Publicly accessible data has been used to construct a county-scale supply chain model of United States gasoline consumption and quantify the scope 3 CO2; emissions from gasoline consumption. Our model tracks the movement of refined fuels from county of refinement to county of blending and eventually to county of consumption via multiple infrastructure networks – pipelines, tankers, trains, and trucks. Where quantities of the fuel moved across different linkages and different transportation modes are known, they are used as is. However, for the vast majority of the country, the exact quantities of fuel moved between county of refining and county of blending or county of blending and county of consumption, as well as the mode of transportation, is not known with certainty. Linear optimization is used to model those links with constraints related to total supply and demand at lower spatial resolutions (State-level and Petroleum Administration for Defense (PAD) Districts). This is the first real attempt at a spatially-resolved scope 3 style CO2 emissions data product specific to United States gasoline consumption. This model can improve understanding of the complex liquid fuel supply chain, and has significant implications for local policy. With a complete model of scope 3 CO2 emissions, it is also possible to analyze how the differences between scope 1 and scope 3 emissions vary across the country. Finally, this model lays the foundation to model the evolution of the U.S. gasoline supply chain – its dependencies, critical linkages, and pinch points – and the evolution of scope 1 and scope 3 CO2 emissions using the full extent of available public data.
The most dynamic electromagnetic energy and momentum exchange processes between the upper atmosphere and the magnetosphere take place in the polar ionosphere, as evidenced by the aurora. Accurate specification of the constantly changing conditions of high-latitude ionospheric electrodynamics has been of paramount interest to the geospace science community. In response this community’s need for research tools to combine heterogeneous observational data from distributed arrays of small ground-based instrumentation operated by individual investigators with global geospace data sets, an open-source Python software and associated web-applications for Assimilative Mapping of Geospace Observations (AMGeO) are being developed and deployed (https://amgeo.colorado.edu). AMGeO provides a coherent, simultaneous and inter-hemispheric picture of global ionospheric electrodynamics by optimally combining diverse geospace observational data in a manner consistent with first-principles and with rigorous consideration of the uncertainty associated with each observation. In order to engage the geospace community in the collaborative geospace system science campaigns and a science-driven process of data product validation, AMGeO software is designed to be transparent, expandable, and interoperable with established geospace community data resources and standards. This paper presents an overview of the AMGeO software development and deployment plans as part of a new NSF EarthCube project that has started in September 2019.