Modern polar regions are critical breeding grounds for over 250 species of birds. Some migrate to high latitudes for access to seasonally abundant resources during reproductive periods, whereas others are year-round residents. Despite the major role these birds play in polar ecosystems, we know very little of the origins of the utilization of polar ecosystems for nesting due to the rarity of avialan fossils from high latitudes. The avialan fossil record spans 150 million years, yet evidence for high-latitude bird reproduction extends only to the Eocene La Meseta Formation of Antarctica (56-33.6 Ma). Here, we report a remarkable polar avifauna from the northernmost fossil-bearing Late Cretaceous ecosystem in the world, the Prince Creek Formation of northern Alaska (PCF). The PCF was deposited at 80-85°N paleolatitude, where continuous summer daylight would have lasted nearly six months. It preserves an ancient polar ecosystem including avian and non-avian dinosaurs, mammals, and fishes. The PCF avialan material was found as part of a decade-long microfossil analysis of channel lag deposits. Numerous skeletal elements, representing almost the entire avialan skeleton, constitute one of the most comprehensive and well-preserved Late Cretaceous avifaunas in the world. These fossils share morphological affinities with hesperornithines, ichthyornithines, and crown birds. Further, abundant perinatal fossils represent the youngest-known growth stages of Mesozoic euornithines. This is the oldest direct evidence for polar bird reproduction and demonstrates that multi ---taxic bird nesting has occurred in the High Arctic for at least 73 million years--nearly half the tenure for birds on Earth. Likewise, these fossils demonstrate that this behavior originated in the Mesozoic ancestors of modern birds, millions of years before the radiation of crown group birds following the end-Cretaceous mass extinction.

Volcanic eruptions pose hazards to human lives and livelihoods (Loughlin et al., 2015). To mitigate these hazards, volcano monitoring groups aim to detect signs of unrest and eruption as early as possible. Prior to eruption volcanoes may show various signals of unrest, including: increased surface temperatures, surface deformation, increased seismicity, increased degassing, and more. Here we focus on one approach to monitor volcanic unrest: detecting high-temperature localized volcanic heat emissions, otherwise known as hotspots. The presence of hotspots can indicate subsurface and surface volcanic processes that precede, or coincide with, eruptions. Space-borne infrared sensors can identify hotspots in near-real-time; however, automatic hotspot detection systems are needed to efficiently analyze the large quantities of data produced. While hotspots have been automatically detected for over 20 years with simple thresholding algorithms, new computer vision technologies, such as convolutional neural networks (CNNs), enable improved detection capabilities. Here we introduce HotLINK: the Hotspot Learning and Identification Network, a CNN-based model to detect volcanic hotspots in VIIRS (Visible Infrared Imaging Radiometer Suite) imagery. We find that HotLINK achieves an accuracy of 96% when evaluated on a validation dataset of ~1,700 unseen images from Mount Veniaminof and Mount Cleveland volcanoes, Alaska, and 95% when evaluated on a test dataset of ~3,000 images from six additional Alaska volcanoes (Augustine Volcano, Bogoslof Island, Okmok Caldera, Pavlof Volcano, Redoubt Volcano, Shishaldin Volcano). Additional testing on ~700 labeled MODIS images demonstrates that our model is applicable to this sensor's data as well, achieving an accuracy of 98%. We apply HotLINK to 10 years of VIIRS data and 22 years of MODIS data for the eight aforementioned Alaska volcanoes. From these time series we find that HotLINK accurately characterizes background and eruptive periods, similar to a threshold-based method, MIROVA, but also detects more subtle warming signals, potentially related to volcanic unrest. In particular, analysis of the Mount Veniaminof record demonstrates that HotLINK is able to detect subtle hotspot signals that are coincident with elevated seismicity, potentially indicative of surface heating due to shallow magma intrusion and/or degassing. We identify three advantages to our model over its predecessors: (1) the ability to detect more subtle volcanic hotspots and produce fewer false positives, especially in daytime imagery; (2) the incorporation of probabilistic predictions for each detection that provide a measure of detection confidence; and (3) its transferability to multiple sensors and multiple volcanoes without the need for threshold tuning, suggesting the potential for global application. HotLINK is able to identify eruptions and potentially precursory warming signals in infrared satellite data, making it a valuable tool for monitoring volcanoes and tracking their heat released over time.

Torrential rains, flooding, and storm surges are all considered meteorological and hydrological hazards. These hazards can quickly transition into disasters whenever they begin to effect human life or infrastructure. When these disasters occur, they can lead to devastating outcomes such as damage or complete loss of infrastructure, reduced crop yields and related impacts on food security, as well as risks to human life. For decades, these events have been monitored and tracked primarily from visible/infrared sensors. Unfortunately, flooding and storm surges often occur during extensive weather and cloud cover conditions, which can make monitoring these hazards across large spatial scales difficult if not impossible, especially in poorly developed regions. Synthetic Aperture Radar (SAR), with its capability to penetrate through clouds and monitor during both day and night cycles, may provide substantial improvements to flood hazard management, once robust techniques for the detection of surface water extent were developed. This thesis details the development of HYDRO30, an automatic algorithm to map surface water extent from Sentinel-1 SAR data. At its core the HYDRO30 approach automatically creates surface water extent maps by preforming adaptive thresholding on dual-polarized and radiometrically terrain corrected (RTC) SAR images. We use data from the Sentinel-1 C-band SAR constellation as input, as this constellation provides global access to free and open medium resolution SAR data that comes at a reliable sampling rate of six-to-twelve days. Such free-and-open, regularly sampled data is indispensable for hazard monitoring across regional to continental scales. In this thesis, I will show that the HYDRO30 algorithm achieves promising results in delivering robust and accurate surface water extent maps within an efficient timeline. The work presented in this thesis was developed as part of the HydroSAR project, a NASA-funded effort led by UAF to develop an automatic service for the monitoring of weather-related hazards in the Hindu Kush Himalaya. The thesis will also show an application of the HYDRO30 technology for the monitoring of flood hazards in northern Bangladesh during the 2020 south-Asian monsoon season.

Extratropical cyclones in the Bering Sea have the potential to cause widespread damage to the coastline of western Alaska, a threat that disproportionately affects Indigenous communities that inhabit these remote regions. A paucity of historical and instrumentation data in these regions, combined with the potential for increases in coastal hazards as a result of climate change, hinders coastal resilience at the local level. To address this problem, this study incorporates local and Indigenous Knowledge into a historical assessment of storm impacts in two communities in the Bering Sea, St. Paul Island and Goodnews Bay, to partially fill these gaps and improve coastal hazard mitigation and preparation. We conducted interviews with local and Indigenous Knowledge bearers in both communities, performed topographic field surveys of historical storm heights, and mapped flooding based on surface models generated from Unmanned Aircraft Systems and flood heights. Particularly impactful storms were described in storm histories, which detail information gathered from instrumentation, on-the-ground field visits, and Indigenous Knowledge interviews. Finally, an online Esri Storymap was designed as an outreach product to combine and visualize the various data products from this analysis with narrative on the experience of living through extreme storm events. This thesis integrates Western and Indigenous Knowledge into actionable, community-prioritized data products to support coastal resilience to climate changes.

Volcanic gas emissions are challenging to quantify. Achieving high confidence in gas composition, column concentrations, and emission rates acquired using remote sensing techniques is thought to require optimal atmospheric conditions. These conditions are often not met, creating a reluctance to preform measurements under non-ideal atmospheric conditions with inherent uncertainty about how useful those measurements may be. In the case of volcanic eruptions, the hazardous nature of the volcanic plume creates an environment where it is often not safe to collect measurements. This dissertation presents three projects which aim to constrain the quantity of two specific volcanic gases, mercury (Hg) and sulfur dioxide (SO₂), released under non-ideal measurement conditions. Specifically, chapter 2 aims to constrain Hg emission during volcanic eruptions, chapter 3 aims to characterize the uncertainty in SO₂ emission rates acquired under specific non-ideal atmospheric conditions, and chapter 4 aims to improve constraints on plume altitude for scanning remote sensing measurements of SO₂ emission rates acquired from a single instrument. Ash is a potential sink of volcanically-sourced atmospheric mercury, and the concentration of particle-bound Hg may provide constraints on Hg emissions during eruptions. In Chapter 2, the Hg concentrations in 227 bulk ash samples from the Mt. Spurr (1992), Redoubt (2009), and Augustine (2006) volcanic eruptions are examined to investigate large-scale spatial, temporal, and volcanic-source trends. No significant difference in Hg concentrations is found in bulk ash by distance from the eruption source or for discrete eruptive events at each volcano, suggesting that in-plume reactions converting gaseous Hg⁰ to adsorbed Hg²⁺ are happening on timescales shorter or longer than considered in this study (minutes to hours) and any additional in-plume controls may be masked by intra-volcanic sample variability. A significant difference is found in Hg concentration in ash among volcanic sources, which indicates that specific volcanoes may emit comparatively high or low quantities of Hg. These findings allow for the calculation of minimum, first-order estimates of volcanic Hg emissions during eruption in combination with total mass estimates of ashfall deposits. Mt. Spurr is found to be a high Hg emitting volcano such that its 1992 particulate Hg emissions likely contributed substantially to the global eruptive volcanic Hg budget for that year. Based on this study, previous approaches that used long-term Hg/SO₂ mass ratios to estimate eruptive total Hg under-account for Hg emitted in explosive events, and global volcanogenic Total Hg estimates need revisiting. A large source of error in SO₂ emission rates derived from mobile differential optical absorption spectroscopy (DOAS) is the uncertainty in atmospheric light paths between the scattered sunlight and the instrument, particularly under non-ideal atmospheric conditions such as the presence of clouds beneath the volcanic plume. In Chapter 3, numerical simulations using the McArtim model are used to examine the radiative transfer associated with zenith-facing mobile DOAS traverses for scenarios where there is a cloud layer between the instrument and the volcanic plume. In total, 217 permutations of atmospheric optical conditions are considered, allowing for the determination of errors associated with atmospheric scattering. Objective criteria are also developed for selecting SO₂ baselines and plume limits for each simulated traverse. This study then applies models to a real-world dataset from the 2021 Cumbre Vieja eruption to explore the effects of ground-level haze on a measured SO₂ column densities for the volcanic plume. All modeling results find large modifications in the shape of the analyzed plume SO₂ column density versus distance curve, even under scenarios with translucent clouds. Despite modification of the plume shape, the presence of a low cloud or haze layer is typically not a large source of error in determination of the total SO₂ quantity measured over the entirety of the traverse, which suggests that fairly accurate SO₂ emission rate measurements can be obtained even under non-ideal atmospheric measurement conditions. The real-world dataset from Cumbre Vieja is found to be best explained by a layer of ground-level laze containing SO₂ and a volcanic plume located between 2 - 4 km altitude. A large source of uncertainty in SO₂ emission rates derived from scanning DOAS instruments is the cross-sectional area of the detection, which is determined from the vertical and horizontal distance of the plume from the instrument. In Chapter 4, a novel method is employed to estimate plume altitude based on modeled wind speed data and validated against available webcam imagery at Cleveland Volcano in the Aleutian Islands, Alaska. This estimated plume altitude is used to calculate SO2 emission rates from single-station campaign scanning DOAS measurements at Cleveland Volcano, Gareloi Volcano, and Korovin Volcano (Alaska) in 2019, where the instrument was deployed for several days at each site. This method is also applied to a long-term dataset of scanning SO2 measurements acquired from a permanent scanning DOAS instrument installed at Cleveland Volcano September 2022 - June 2023. It is found that the method of estimating plume altitude in the long-term dataset produces a lower emission rate and a smaller sample variance than assuming a fixed summit plume altitude. The remaining variance in the data is then interpreted to represent variability in SO₂ emissions during times of relative quiescence at each studied volcano.

The distribution of fold related fractures and other mesoscopic structures in asymmetrically folded Mississippian to Pennsylvanian Lisbume Group carbonates gives clues concerning the mechanism of folding. Since fracture sets pre-date and post-date folding, it is important, but sometimes difficult, to determine which fracture sets are related to folding. Higher density of fold related fractures and dissolution cleavage in the hinges than limbs of two folds in the study area is evidence for fixed hinge detachment folding. However, geometric modeling of box shaped folds in the study area suggests that some folds may have formed by either detachment folding or trishear fault propagation folding. Formulaic modeling of fracture density in a stratigraphic section using stratigraphic attributes such as lithology, bed thickness, and chert content predicts general trends in fracture density, but other factors such as slip along bed contacts may obscure the relationship between fracture density, lithology and bed thickness.

Earth processes, such as large landslides and volcanic eruptions, occur globally and can be hazardous to life and property. Geophysics -- the quantitative study of Earth processes and properties -- is used to monitor and rapidly respond to these geohazards. In particular, seismoacoustics, which is the joint study of seismic waves in the solid Earth and acoustic waves in Earth's atmosphere, has been proven effective for a variety of geophysical monitoring tasks. Typically, the acoustic waves studied are infrasonic: They have frequencies less than 20 hertz, which is below the threshold of human hearing. In this dissertation, we use seismic and acoustic waves and techniques to characterize geohazards, and we examine the propagation of the waves themselves to better understand how seismoacoustic energy is transformed on its path from a given source to the measurement location. Chapter 1 provides a broad overview of seismoacoustics tailored to this dissertation. In Chapter 2, we use seismic and acoustic waves to reconstruct the dynamics of two very large, and highly similar, ice and rock avalanches occurring in 2016 and 2019 on Iliamna Volcano (Alaska). We determine their trajectories using seismic data from distant stations, demonstrating the feasibility of remote seismic landslide characterization. Chapter 3 details the application of machine learning to a rich volcano infrasound dataset consisting of thousands of explosions recorded at Yasur Volcano (Vanuatu) over six days in 2016. We automatically generate a labeled catalog of infrasound waveforms associated to two different locations in Yasur's summit crater, and use this catalog to test different strategies for transforming the waveforms prior to classification model input. In Chapter 4, we use the coupling of atmospheric waves into the Earth to leverage a dense network of about 900 seismometers around Mount Saint Helens volcano (Washington state) as a quasi-infrasound network. We use buried explosions from a 2014 experiment as sources of infrasound. The dense spatial wavefield measurements permit detailed examination of the effects of wind and topography on infrasound propagation. Finally, in Chapter 5 we conclude with some discussion of future work and additional seismoacoustic topics.

Exploring and expanding our understanding of the planets (i.e., planetary science) encompasses a vast array of topics and disciplines. This dissertation concentrates on the surficial processes of and examination of terrestrial planets, primarily via the study of volcanism and impact cratering. The first project starts with an exploration of NIR remote sensing techniques as applied to Venus. This work found that NIR remote sensing at the clement conditions just beneath the cloud deck provide vastly improved imaging capability. This improved visibility is most notable for the tesserae, regions of Venus of great interest to the scientific community. Radar imagery and derived data products were then used to survey 21 mid-sized volcanoes on the surface of Venus. Similar to volcanoes at larger diameters, the midsized volcanoes of Venus are significantly flatter than those on other terrestrial bodies. Several of these volcanoes also show deformation that requires a negligibly thin lithosphere some time after the emplacement of the construct. The third project then evaluates the hazards involved with safely placing a lander on the Venusian tesserae and examines potential methods by which to detect and then avoid these hazards. Safely placing a suite of scientific instruments on tesserae is necessary to answer long-standing questions about Venus. Current technologies put relevant hazards at the edge of detection (i.e., zero fault tolerance) and can execute divert maneuvers of only a few tens of meters. Investment in hazard detection and avoidance technologies is necessary to bring safety margins to acceptable levels; data from future missions - while helpful - will be insufficient to select safe landing zones prior to launch. Oblique impact cratering is a ubiquitous event (approximately half of all impacts are at 45 or less). Our poor understanding of this process leaves a significant amount of information buried and waiting to be uncovered. Low-velocity oblique impact experiments were conducted at John's Hopkins University's Applied Physics Laboratory Planetary Impact Lab to better understand the oblique impact process and prepare for high velocity experiments at similar impact angles. These experiments also sought to understand the effect of target tilt, which is currently necessary at existing experimental facilities in order to simulate changes in impact angle smaller than 15°. These experiments show that target tilt significantly amplifies oblique characteristics (e.g., aspect ratio, butterfly ejecta). The time-delayed and spatially offset transference of energy from the impactor to the target is important in determining the excavation process and final crater morphology and ejecta distribution.

Storms and a changing climate can cause disasters for coastal communities. This is particularly felt in the Arctic, where temperatures are rising faster than the global average and sea ice decline has led to a longer open-water period when storms can make landfall. Communities in western and northern Alaska have experienced decades of frequent flooding and erosion. Climate change is a significant contributor today and for the future, but the current prevalence of disaster loss is more closely related to the continued development within hazard-prone areas. This dissertation maps areas prone to erosion and flooding to provide decision-making products for adaptation planning. Long-term erosion rates are projected 60 years into the future to compute the time until infrastructure is undermined. Flood histories are quantified to map hazard areas and improve storm impact forecasts. Record flood levels are estimated to support floodplain mapping. These studies are primarily exposure analyses that compare existing hazards to existing infrastructure to identify at-risk and safe areas around communities. However, as climate change progresses, existing hazards will change in unanticipated ways. The erosion history of Port Heiden tells an important story about how an unexpected change in coastal barrier islands led to rapid erosion and forced the community to relocate inland. This dissertation fills in the long-standing baseline data gaps about hazards in Arctic Alaska coastal communities. These results will help communities develop outside of hazard prone areas. However, the next iterations of hazard analyses will need to build off this baseline and carefully apply climate-informed approaches to predict changing hazards. Moreover, risk assessments must address community priorities and be improved by incorporating more of the values of mixed subsistence economies common to this region. Through community collaboration, hazard exposure analyses like this dissertation, and improving geophysical models, a roadmap is being built for communities to navigate towards a safer future.

Wildfires are a natural and essential part of Alaska ecosystems, but excessive wildfires pose a risk to the ecosystem's health and diversity, as well as to human life and property. To manage wildfires effectively, vegetation/fuel maps play a critical role in identifying high-risk areas and allocating resources for prevention, suppression, and recovery efforts. Furthermore, vegetation/fuel maps are an important input for fire behavior models, along with weather and topography data. By predicting fire behavior, such as spread rate, intensity, and direction, fuel models allow fire managers to make informed decisions about wildfire suppression, management, and prevention. Traditionally used vegetation/fuel maps in Alaska are inadequate due to a lack of detailed information since they are primarily generated using coarser resolution (30m) multispectral data. Hyperspectral remote sensing offers an efficient approach for better characterization of forest vegetation due to the narrow bandwidth and finer spatial resolution. However, the high cost associated with data acquisition remains a significant challenge to the widespread application of hyperspectral data. The aim of this research is to create accurate and detailed vegetation maps and upscale them for the boreal region of Alaska. The study involves hyperspectral data simulation using Airborne Visible InfraRed Imaging Spectrometer - Next Generation (AVIRIS-NG) data and publicly available Sentinel-2 multispectral data, ground spectra convolved to Sentinel-2 and AVIRIS-NG using the spectral response function of each sensor. Simulated data captured the minute details found in the real AVIRIS-NG data and were classified to map vegetation. Using the ground data from Bonanza Creek Long-Term Ecological Research sites, we compared the new maps with the two existing map products (the LANDFIRE's Existing Vegetation Type (EVT) and Alaska Vegetation and Wetland Composite). The maps generated using simulated data showed an improvement of 33% in accuracy and are more detailed than existing map products. In addition to fuel maps, we performed sub-pixel level mapping to generate a needleleaf fraction map, which serves fire management needs since needleleaf species are highly flammable. However, validating the sub-pixel product was challenging. To overcome this, we devised a novel validation method incorporating high-resolution airborne hyperspectral data (1m) and ground data. The study addresses the limitations of traditional fuel/vegetation maps by providing a more detailed and accurate representation of vegetation/fuel in Alaska. The methods and findings advance fuel and vegetation mapping research in Alaska and offer a novel pathway to generate detailed fuel maps for boreal Alaska to aid wildfire management.