Very high spatial resolution (VHSR) commercial satellite imagery affords permafrost scientists the ability to monitor the pan-Arctic system at a fine-scale, enabling detailed monitoring of both the natural and human environments. Geo-AI mapping applications based on the deep learning (DL) convolutional neural network (CNN) have been successful in translating this big imagery resource into Arctic science-ready products. However, many models are computationally intensive due to the constraints of the large geographical extent and complexity of VHSR imagery. In addition, feature recognition is challenged by scarcity of manually-annotated training data and image complexity at fine scales. In this exploratory study, we investigated the ability of a lightweight U-Net DLCNN to efficiently perform semantic segmentation of VHSR commercial satellite imagery with limited training data in automated recognition of human-built infrastructure, including residential, industrial, public, commercial buildings, and roads, in the permafrost affected regions of the Arctic. We conducted a systematic experiment to understand how image augmentation improves the performance of DL-based semantic segmentation of VHSR imagery. Different standard augmentations, including flipping, rotation, and transposition, were applied to input imagery in order to test their impacts on infrastructure recognition and determine the optimal set of augmentations. With a relatively low number of model parameters, limited labelled training data, short training time, and high segmentation accuracy, our findings suggest that overall, the U-Net DLCNN, coupled with image augmentation, could serve as an accurate and efficient method for mapping infrastructure in the Arctic permafrost environment without compromising spatial details and geographical extent.

Water resources are important to both natural ecosystems and human societies. Surface water is the most readily accessible water resource and provides an array of ecosystem services. Water stress, the ratio of water demand to supply, is a global concern as water resources are stressed by changes in climate, land cover, and population size. Understanding current and projected spatial and temporal factors of surface water dynamics is key to better managing our water resources and limiting the effects of water stress. However, few studies estimating changes in surface water account for climate and human drivers synergistically. Therefore, we compared three sets of statistical models using climate only, anthropogenic only, and the combination of climate and anthropogenic explanatory variables to assess the influence of each set of drivers on estimating surface water. We then used the most accurate model, the combination of climate and anthropogenic drivers (-0.17% average watershed mean percent error), with climate and land use projection data to project surface water areas under different climate and land use scenarios. For climate drivers, we used precipitation and temperature data from ensembles of the Inter-Comparison of Coupled Models-Phase 5 (CMIP5) Global Climate Models under three Representative Concentration Pathways (RCPs)–RCP4.5, RCP6.0, and RCP8.5. For anthropogenic drivers, we used three land use/land cover change projections from the U.S. Geological Survey’s FOR-SCE model corresponding to Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions Scenarios (SRES) that have RCP counterparts. Our models suggest an uneven distribution of projected change in surface water area, where watersheds with more natural land cover will experience less change (positive or negative) and watersheds with less natural land cover will experience more change. We also expect to find that, under the business-as-usual scenario, watersheds with greater urbanization will see a reduction in surface water area by 2100. These results highlight our ability to mitigate water stress with land use management and also emphasize the need to account for both climate and anthropogenic drivers when estimating and predicting surface water area.

Conceptual models are among the most frequently used type of models in watershed modelling studies, due to their low computational requirements and ease of interpretation. Model selection requires the comparison of model alternatives, which is complicated when the models present in the literature differ in many uncontrolled aspects, such as conceptualization, implementation, and source code availability. To overcome this limitation, several model-building frameworks have been introduced in the last decade, which facilitate model comparisons by enabling different model alternatives within the same software and numerical architecture. Building on the decennial experience with the development and usage of Superflex, a flexible modeling framework for conceptual model building, so far implemented in FORTRAN language and not available as open source, we propose SuperflexPy, an open source Python framework for building conceptual hydrological models. Compared to other existing models or flexible frameworks, SuperflexPy is designed to be extremely easy to modify or extend, allowing scientists to build models that reflect their processes understanding; thanks to its object-oriented architecture and its complete integration with the Python programming environment, SuperflexPy can be seen as a high-level programming language for constructing hydrological models that are extremely flexible both in terms of elements configuration (i.e. how the elements are connected into a structure) and spatial organization. By design, SuperflexPy is not limited to water quantity but can be extended to simulate transport processes (water quality). In this presentation, we will illustrate the principles behind the design of the framework and showcase some applications.

Recent research showed that machine learning, in particular deep learning, can be applied with great success to a multitude of seismological tasks, e.g. phase picking and earthquake localization. One reason is that neural networks can be used as feature extractors, generating generally applicable representations of complex data. We employ a convolutional network to condense earthquake waveforms from a varying set of stations into a high dimensional vector, which we call event embedding. For each event the embedding is calculated from instrument-corrected waveforms beginning at the first P pick and updated continuously with incoming data. We employ event embeddings for real time magnitude estimation, earthquake localization and ground motion prediction, which are central tasks for early warning and for guiding rapid disaster response. We evaluate our model on the IPOC catalog for Northern Chile, containing ∼100,000 events with low uncertainty hypocenters and magnitude estimates. We split the catalog sequentially into a training and a test set, with the 2014 Iquique event (Mw 8.1) and its fore- and aftershocks contained in the test set. Following preliminary results the system achieves a test RMSE of 0.28 magnitude units (m.u.) and 35 km hypocentral distance 1 s after the first P arrival at the nearest station, which improves to 0.17 m.u. and 22 km after 5 s and 0.11 m.u. and 15 km after 25 s. As applications in the hazard domain require proper uncertainty estimates, we propose a probabilistic model using Gaussian mixture density networks. By analyzing the predictions in terms of their calibration, we show that the model exhibits overconfidence i.e. overly optimistic confidence intervals. We show that deep ensembles substantially improve calibration. To assess the limitations of our model and elucidate the pitfalls of machine learning for early warning in general, we conduct an error analysis and discuss mitigation strategies. Despite the size of our catalog, we observe issues with two kinds of data sparsity. First, we observe increased residuals for the source parameters of the largest events, as training data for these events is scarce. Second, similar inaccuracies occur in areas without events of a certain size in the training catalog. We investigate the impact of these limitations on the Iquique fore- and aftershocks.

Inlet change along the Virginia barrier islands has been studied on an individual inlet basis. This study evaluated the ten inlets both individually and as a group. Evaluating the inlets as a group allows the discovery of inlet change patterns among the inlets. Inlet width was measured from satellite and aerial images. Width is the straight-line distance from the northernmost fast land to southernmost fast land. Measurements spanned from July 1999 to April 2018 with a frequency of four to eight width measurements per year. Total width change to 2018 was 2 percent to 357 percent. The largest change was to Great Machipongo Inlet. The inlet widened from approximately 800 meters wide in 1999 to 3645 meters wide in 2018. This widening effectively splits the remaining nine inlets into two groups. Inlet widening patterns of the two groups are high rates for the first three or four inlets and a very low rate for the final inlet of the group. Wind direction change also plays a significant role in net inlet widening. Inlet widening has not abated and wider inlets will allow larger waves to affect environmentally and commercially important areas to the west of the barrier island/inlet system.

The National Ecological Observatory Network (NEON) is a long-term ecological observatory focused on collecting and providing open, continental-scale data that characterize and quantify complex and rapidly changing ecological patterns and processes. As part of the broader Observatory design, the NEON Assignable Assets Program provides access to NEON infrastructure enabling Principal Investigator (PI)-driven research. Requests can be from five main areas: Sensor Infrastructure (SI), Observational Sampling Infrastructure (OSI), the Airborne Observation Platform (AOP), the Mobile Deployment Platform (MDP), and Field Site Access and Coordination (FSAC). NEON’s field sampling teams are also available at NEON sites to support specimen collection and observations for specific research needs. The PI or their sponsor is responsible for costs incurred by the NEON program to complete the requested access or collection. In addition, NEON can provide Letters of Support/Collaboration for collaborating with NEON scientists or using NEON data. Sensor Infrastructure (SI): NEON offers access to existing sensor infrastructure (i.e., towers, power, and communications) for installation of physical instrument systems or the addition of new arrays for collecting data from automated sensor suites. Observational Sampling Infrastructure (OSI): PIs can request that NEON Field Ecologists collect data or biological samples at NEON sites. Airborne Observation Platform (AOP): Provides a suite of remote sensing instruments mounted in an aircraft for collecting airborne-based data (discrete and waveform LiDAR, hyperspectral imagery, and RGB camera). Mobile Deployment Platform (MDP): NEON offers a suite of self-contained, mobile arrays of sensors, power systems, and data logging capabilities for capturing atmospheric, soil, and aquatic-based measurements. These are deployable in areas of interest to capture data analogous to ongoing NEON data collection. Field Site Access Coordination (FSAC): Researchers may want to work within or adjacent to NEON sites while not utilizing NEON infrastructure or personnel. Review of these projects ensures no impact to NEON data collection, and provides minimal logistical support for working adjacent to NEON sites. More at https://www.neonscience.org/assignable-assets

Flare suprathermal ions with enhanced 3He and heavy-ion abundances are an essential component of the seed population accelerated by CME-driven shocks in gradual solar energetic particle (GSEP) events. However, the mechanisms through which CME-driven shocks gain access to flare suprathermals and produce spectral and abundance variations in GSEP events remain largely unexplored. We report two recent GSEP events: one observed by Solar Orbiter on 2020 Nov 24 (the first GSEP event on Solar Orbiter) and the other by ACE on 2021 May 29 (the most intense GOES proton event in the present solar cycle). The events were preceded by impulsive SEP (ISEP) events. Abundances and energy spectra are markedly different in the examined events at < 1 MeV/nucleon. For example, in the May event, Fe/O is typical of ISEP events, a factor of 100 to 10 higher than Fe/O in the November event. 3He abundance in the November event is high, typical of ISEP events, while in the May event, it is much lower, though finite. The May event shows a hard 4He spectrum with a power-law index of −1.6, and the November event a soft spectrum with an index of −3.5. The events were associated with halo CMEs with speeds around 900 km/s. The November event was also measured by Parker Solar Probe and the May event by STEREO-A and Solar Orbiter. This paper discusses the origin of vastly different abundances and spectral shapes in terms of variable remnant population from preceding ISEP events. Furthermore, we discuss a possible direct contribution from parent flares.

Regional geodynamics studies of the eastern United States (e.g., Ghosh et al., 2019) have shed light on the force balance controlling intraplate, passive margin seismicity. Recent observations of crustal structure in the Southeastern United States (Chen et al., 2018; Hopper et al., 2016; Parker et al., 2016; Savage et al., 2017; Wagner et al., 2018) now allow us to examine the gravitational component of force balance in more detail. Here, we examine the role of gravitational potential energy (GPE) within the crust and upper mantle in ongoing stress, strain, and tectonism in the Southeastern United States. We first calculate GPE differences using density and crustal thickness data from Crust 1.0 (Laske et al., 2013) which are integrated over the thickness of the lithosphere on a 0.5 by 0.5 degree grid (e.g., Flesch et al., 2001). In further models we modify crustal densities using recent seismic observations (e.g., Parker et al., 2016; Wagner et al., 2018). We compare model results to earthquake epicenters and locations of inherited structures within our study area, including the Reelfoot Rift; New Madrid Seismic Zone; Illinois Basin; New York-Alabama Lineament; sutures between sections of accreted terrane; and a region of high seismic velocity beneath central West Virginia interpreted as delaminated, eclogitized crust (Biryol et al., 2016). Whereas in most locations seismicity is associated with higher magnitudes of GPE, inherited structures may explain regions in which earthquakes occur in areas with lower magnitudes of GPE.

Intra-continental dip-slip earthquakes often occur in the orogen and rift zones with complex tectonics, providing rare opportunities to illuminate the deformation and evolution of continental structures. However, due to the sparse seismological and geodetic observations, such earthquakes are less studied. Here, we report the fault geometries of two dip-slip earthquakes recently occurred in Southern Tian Shan and the Mongolia-Baikal rift zone revealed by InSAR . The 2020 Mw6.0 Jiashi earthquake occurred in the Keping-tage fold-and-thrust belt in southwest Tian Shan. This region is seismically active, yet most well-recorded earthquakes occurred south of the mountain front. The lack of large earthquakes beneath the belt thus hinders our understanding of the orogenic process to the north. Combining InSAR measurements and relocated aftershocks, we found that a fault model involving a shallow thrust fault and two deeper faults can best reconcile the surface deformation and aftershock distribution. Stress analysis suggests that slips on the shallow fault reactivated the older basement structures at depth. Our results reflect the basement-involved shortening activated by a thin-skinned thrust faulting event with surface deformation, implying a southward orogenic process of the southwest Tian Shan. The 2021 Mw6.7 Lake Hövsgöl earthquake occurred in the Mongolia-Baikal rift zone (MBRZ), which is located in the northern tip of the northern Mongolia, and is bounded by the Tibet Plateau orogenic belt and the Siberian Platform. Using the Bayesian inversion method we derived a fault plane with two slip patches, one is mainly strike slip and the other is mainly normal slip component. The correlation between the observed and predicted displacements by the single fault model is 97.46%. Coulomb stress analysis shows that the 2021 event has a triggering effect on the western segment of the Tunka Fault to the north, where no large earthquakes have occurred since the 1905 M8+ earthquake, raising the potential for seismic risk in this region.

The discipline of informatics, generically cast as the science and engineering of information system within a socio-technical framework, originating in the middle of last century has undergone generational adaptations as computer hardware, networks and software have evolved. Within the “eScience” era of the last two decades, discipline-specific fields of informatics have flourished, such as geoinformatics, mineral informatics and many more. In fact even in geosciences, there may be few fields of study that have not added an informatics sub-field. Over the same time, efforts at systematizing the common (or core, i.e. discipline neutral) aspects of informatics have been successful: use cases, human-centered design, iterative approaches, information models and more are some of the key elements. However new pressures are being placed on functional and non-functional requirements of information systems: with the now somewhat routine underlying data that are high dimensional, heterogeneous, sparse and with uncertain quality. However, demands have arisen from renewed attention to hub/ server/ cloud-based provision of the application of machine learning, neural networks and artificial intelligence in general. Those methods as implemented in software libraries producing results to be assessed and interpreted (often leading to decisions made) by “humans-in-the-loop”. Informatics, revisited is a possible answer. This presentation features some history of informatics, recent disruptions that need to be addressed, and offers ideas for new directions, with the goal of advancing geoscience in general.

VISAGE (Visualization for Integrated Satellite, Airborne, and Ground-based data Exploration) aims to provide visualization and analytic capabilities for diverse datasets in an interactive user interface. Proof-of-concept use cases are centered around the Global Precipitation Measurement (GPM) mission’s Ground Validation (GV) program, which provides a wealth of intensive, coincident observations of atmospheric phenomena from a wide variety of ground-based, airborne and satellite instruments. These data have diverse temporal and spatial scales, variables, and data formats and organization. Key technical challenges include: 3D data rendering and visualization of multiple diverse datasets on a web-based platform 3D data interrogation via map user interface Temporal alignment of data with diverse time scales and resolutions Computations on data fields across instruments and platforms To address these issues, the VISAGE project is working with cloud-native serverless technologies to render data as 3D Tile Point Clouds for display in the Cesium geospatial 3D global mapping platform.

The response of atmospheric variability to the Tibetan Plateau (TP) snow cover (TPSC) at interannual to decadal time-scales has been extensively investigated. However, the response of the atmosphere to the subseasonal variability of the TPSC has been largely ignored. Here, we found that the subseasonal variability of the TPSC revealed by daily data can serve as an indicator of East Asian atmospheric circulation on medium-range time-scales (approximately 3–7 days in advance). The subseasonal variability of the TPSC influences the upper-level regional atmospheric circulations, especially the East Asia upper-level westerly jet stream (EAJS), during wintertime. The variability of the TPSC modulates the land surface thermal conditions over the TP, which acts as an elevated cooling source in the middle troposphere during wintertime. The upper-level geopotential height is affected by the TPSC and then the zonal wind. As a result, the EAJS is influenced.

The large-scale electric field in the inner magnetosphere is a key driver of many processes and the dynamics of magnetospheric plasmas. During geomagnetic storms, the enhanced convection electric field is responsible for eroding the plasmasphere and for moving the inner edge of the plasma sheet earthward. In this presentation, we show the preliminary results of an examination of the distribution and variations of the inner magnetospheric quasi-static electric field as measured by the Electric Field and Waves (EFW) instruments onboard the twin spacecraft of the Van Allen Probes mission. We investigate the role that the electric field plays in plasmasphere erosion and plasma sheet access to the inner magnetosphere by analyzing the electric field measurements in conjunction with cold plasma density and plasma sheet particle flux measurements. Since the coupling between plasma populations in the magnetosphere is inherently related to the electric field, we expect that the combined measurements of the electric field and plasmas will enhance our understanding of the physical processes that drive the magnetospheric dynamics.

HiRISE-based mapping reveals five landform assemblages in western Jezero crater, each defined by a landform association interpretable using Earth-based landsystem models and well-understood Earth analogues. 1) The northwestern assemblage hosts boulder hills rimmed by lobate ridges, mounds and mesas on a valley floor, and valley-bounding ridges superposed by striations (= parallel boulder-bearing ridges and grooves). 2) A trough zone hosts variously shaped depressions, intra-trough islands, linear and curvilinear boulder ridges, and highland strips topped by striated surfaces and rimmed by boulder-bearing ridges. 3) The steep-sided fan-shaped plateau (“western Jezero delta”) hosts mesas, highland-rim boulder ridges, depressions, linear and curvilinear ridges, and a plain superposed by radially trending striations. 4) The crater-margin assemblage hosts a steep-sided ridged and pitted hummocky terrain, a terrace-like capping surface, and mounds surrounded by radially trending boulder ridges. 5) The crater-floor assemblage hosts a polished and striated terrain that displays fold-like and streamlined ridges, hummocky landforms dominated by quasi-circular depressions with raised rims, mesas exposing fold-thrust strata, flat-topped steep-sided ridges with U-shaped map traces, polygonal-grooved plains, and unconsolidated boulder mounds and ridges. Although any aforementioned landform unit could be explained by multiple formative mechanisms, the spatiotemporal relationships mapped in this study within and among the assemblages place stringent constraints for any self-consistent interpretation. A model capable of explaining the mapping results involves northeast-flowing glaciation, ice-sheet collapse with ice-fracture patterns controlling the formation of polygonal grooves via crevasse filling and ice pressing, and minor aeolian modification. In the model, the plateau and crater-margin assemblages were formed by ice-walled subglacial deposition, the trough zone by subglacial flooding, the northwestern and basin-floor assemblages by glacial deformation and deposition, circular depressions with raised rims by melt out and down pressing of spherical dead-ice blocks (i.e., thermal karsts and kettle holes), mesas by kame formation, and striations by glacial fluting.

Reliability of future global warming projections depends on how well climate models reproduce the observed climate change over the twentieth century. In this regard, deviations of the model simulated climate change from observations, such as a recent “pause” in global warming, have received considerable attention. Such decadal mismatches between model simulated and observed climate trends are common throughout the twentieth century, and their causes are still poorly understood. Here we show that the discrepancies between the observed and simulated climate variability on decadal and longer time scale have a coherent structure suggestive of a pronounced global multidecadal oscillation. Surface temperature anomalies associated with this variability originate in the North Atlantic and spread out to the Pacific and Southern oceans and Antarctica, with Arctic following suit in about 25–35 years. While climate models exhibit various levels of decadal climate variability and some regional similarities to observations, noneof the model simulations considered match the observed signal in terms of its magnitude, spatial patterns and their sequential time development. These results highlight a substantial degree of uncertainty in our interpretation of the observed climate change using current generation of climate models.

Providing environmental flows is challenging in the middle portion of the Rio Grande/Bravo Basin (between Elephant Butte Reservoir in New Mexico and Presidio, Texas) where water demand has continued to increase over time despite limited river water and dropping groundwater levels. Riparian ecosystems in this agriculture-dominated desert environment will likely become more vulnerable as competition over scarce water increases in the face of growing demand and dwindling supply. Little to no water is allocated to riparian ecosystems unless water or water rights are purchased or transferred to sustain those systems. Ongoing debates about providing environmental flows for riparian forest galleries in this water-scarce region have not been backed by quantitative modeling results of potential impacts on surface water and groundwater availability. We quantify water requirements to provide a menu of options for environmental flow allocation to establish cottonwood forest galleries. We apply hydrologic modeling under a projected warm-dry future to determine the frequency of river water availability for providing minimum environmental flows, and to evaluate water budget tradeoffs associated with environmental flow allocations in this region. Results inform water resources management decisions that support riparian habitats.

In dry regions, millions of people depend on freshwater provided by the mountain cryosphere. Its likely depletion would make productive land-use management and access to water supply an even more urgent priority. Therefore water-security-oriented policies increasingly rely on solid information feedbacks for projections provided by Earth Sciences. Nevertheless, this type of research still has a lot to understand regarding headwater catchment hydrology, the top global “water towers.” For example, there are many theoretical and logistical uncertainties: “data deserts” in isolated areas, outdated legislation, or scarce research funding. Yet, one more important issue to highlight is the evolving nature of hydric resources, particularly where baselines have a large uncertainty and supply to many as in dry regions in the Andes or the Himalayas. The main concern here is the legislative inadequacy for evolving hydric resources as their baselines change. For example, groundwater within transboundary or paleowater aquifers could have unaccounted climate-sensitive recharge sources (e.g., permafrost thaw). Hence, the specific way of legislating mountain groundwater could turn ambiguous and useless. By reviewing particular legislation and landing the discussion on study cases in mountainous areas, we commit to showing the inadequacy of current legislation on hydric-potential evolution. Overall, water-security-oriented legislation will not assess and protect headwater catchments within the spectrum of different recharge processes throughout different hydroclimatic zones. First, the “evolving value” of specific catchments changes the nominal priority and purpose for protection. Secondly, a consistent failure to assess incommensurable (latent), climate-sensitive fractions of water supply structure is also found. Therefore, the policy recommendation is to use a hydric scale absorbing all nested processes necessary for hydric supply to persist, requiring defining a lifespan for legislation.

Recently, statistical machine learning and deep learning methods have been widely explored for corn yield prediction. Though successful, machine learning models generated within a specific spatial domain often lose their validity when directly applied to new regions. To address this issue, we designed an unsupervised adaptive domain adversarial neural network (ADANN). Specifically, through domain adversarial training, the ADANN model reduced the impact of domain shift by projecting data from different domains into the same subspace. Also, the ADANN model was designed to be trained in an adaptive way, which guaranteed the model can learn the domain-invariant features and perform accurate yield prediction simultaneously. Informative variables including time-series vegetation indices and sequential weather observations were first collected from multiple data sources and aggregated to the county level. Then, we trained the ADANN model with the extracted features and corresponding reported county-level corn yield from the U.S. Department of Agriculture (USDA). Finally, the trained model was evaluated in four testing years 2016-2019. The U.S. corn belt was used as the study area and counties under study were grouped into two diverse ecological regions. The experimental results showed that the developed ADANN model had better performance than three other state-of-the-art machine learning models in both local experiments (train and test in the same region) and transfer experiments (train and test in different regions). As the first study using adversarial learning for crop yield prediction, this research demonstrates a novel solution for improving model transferability on crop yield prediction.