Search USGS⌕ Search

SEARCH · Search USGS

Results for “Advances in Remote Sensing”

Search indexed USGS publications on groundwater, aquifers, geologic maps, mineral resources and earthquakes. Explore source records by subject and place.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 55 records · Page 3Linked to original sources
USGS2017

Fishing activities

Unlike the major anthropogenic changes that terrestrial and coastal habitats underwent during the last centuries such as deforestation, river engineering, agricultural practices or urbanism, those occurring underwater are veiled from our eyes and have continued nearly unnoticed. Only recent advances in remote sensing and deep marine sampling technologies have revealed the extent and magnitude of the anthropogenic impacts to the seafloor. In particular, bottom trawling, a fishing technique consisting of dragging a net and fishing gear over the seafloor to capture bottom-dwelling living resources has gained attention among the scientific community, policy makers and the general public due to its destructive effects on the seabed. Trawling gear produces acute impacts on biota and the physical substratum of the seafloor by disrupting the sediment column structure, overturning boulders, resuspending sediments and imprinting deep scars on muddy bottoms. Also, the repetitive passage of trawling gear over the same areas creates long-lasting, cumulative impacts that modify the cohesiveness and texture of sediments. It can be asserted nowadays that due to its recurrence, mobility and wide geographical extent, industrial trawling has become a major force driving seafloor change and affecting not only its physical integrity on short spatial scales but also imprinting measurable modifications to the geomorphology of entire continental margins.

Book chapter↗
USGS2012

Potential pollutant sources in a Choptank River (USA) subwatershed and the influence of land use and watershed characteristics

Row-crop and poultry production have been implicated as sources of water pollution along the Choptank River, an estuary and tributary of the Chesapeake Bay. This study examined the effects of land use, subwatershed characteristics, and climatic conditions on the water quality parameters of a subwatershed in the Choptank River watershed. The catchments within the subwatershed were defined using advanced remotely-sensed data and current geographic information system processing techniques. Water and sediment samples were collected in May–October 2009 and April–June 2010 under mostly baseflow conditions and analyzed for select bacteria, nitrate-N, ammonium-N, total arsenic, total phosphorus (TP), orthophosphate (ortho-P), and particle-phase phosphorus (PP); n = 96 for all analytes except for arsenic, n = 136, and for bacteria, n = 89 (aqueous) and 62 (sediment). Detections of Enterococci and Escherichia coli concentrations were ubiquitous in this subwatershed and showed no correlation to location or land use, however larger bacterial counts were observed shortly after precipitation. Nitrate-N concentrations were not correlated with agricultural lands, which may reflect the small change in percent agriculture and/or the similarity of agronomic practices and crops produced between catchments. Concentration data suggested that ammonia emission and possible deposition to surface waters occurred and that these processes may be influenced by local agronomic practices and climatic conditions. The negative correlation of PP and arsenic concentrations with percent forest was explained by the stronger signal of the head waters and overland flow of particulate phase analytes versus dissolved phase inputs from groundwater. Service roadways at some poultry production facilities were found to redirect runoff from the facilities to neighboring catchment areas, which affected water quality parameters. Results suggest that in this subwatershed, catchments with poultry production facilities are possible sources for arsenic and PP as compared to catchment areas where these facilities were not present.

Choptank River↗
USGS2024

GNSS reflectometry from low-cost sensors for continuous in situ contemporaneous glacier mass balance and flux divergence

Recent advances in remote sensing have produced global glacier surface elevation change data. Parsing these elevation change signals into contributions from the climate (i.e. climatic mass balance) and glacier dynamics (i.e. flux divergence) is critical to enhance our process-based understanding of glacier change. In this study, we evaluate three approaches for direct, continuous measurements of the climatic mass balance, flux divergence, and elevation change at a site on Gulkana Glacier in Alaska using low-cost GNSS sensors, GNSS interferometric reflectometry (GNSS-IR), banded ablation stakes with time-lapse cameras, and combinations thereof. Cumulative climatic mass balance over the season was 4.85 m and the three approaches were within 0.08 m through early July before the snowpack melted, and within 0.28 m through mid-August. The flux divergence increased from 0.52 ± 0.03 cm d -1 before June 3 to roughly 0.73 cm d -1 after June 27. We demonstrate a single GNSS system fixed atop an ablation stake can measure contemporaneous climatic mass balance, flux divergence, and elevation change based on the antenna’s position and GNSS-IR techniques. The ability of these systems to measure glacier mass balance and flux divergence offers unique opportunities for year-round observations on mountain glaciers in the future.

Alaska↗
USGS2018

Estimating soil respiration in a subalpine landscape using point, terrain, climate and greenness data

Landscape carbon (C) flux estimates are necessary for assessing the ability of terrestrial ecosystems to buffer further increases in anthropogenic carbon dioxide (CO2) emissions. Advances in remote sensing have allowed for coarse-scale estimates of gross primary productivity (GPP) (e.g., MODIS 17), yet efforts to assess spatial patterns in respiration lag behind those of GPP. Here, we demonstrate a method to predict growing season soil respiration at a regional scale in a forested ecosystem. We related field measurements (n=144) of growing season soil respiration across subalpine forests in the Southern Rocky Mountains ecoregion to a suite of biophysical predictors with a Random Forest model (30 m pixel size). We found that Landsat Enhanced Vegetation Index (EVI), growing season AI, temperature, precipitation, elevation, and slope aspect explained spatiotemporal variability in soil respiration. Our model had a psuedo-r2 of 0.45 and root mean squared error (RMSE) of roughly one-quarter of the mean value of respiration. Predicted growing season soil respiration across the region was remarkably consistent across 2004, 2005 and 2006 (150-d averages of 542.8, 544.3, and 536.5 g C m-2, respectively). Yet, we observed substantial variability in spatial patterns of soil respiration predictions that varied between years, suggesting that our method is sensitive to changes in respiration drivers. We compared our estimates to MODIS GPP and nocturnal net ecosystem exchange (NEE) derived from eddy covariance towers as a proxy for ecosystem respiration. Averaged across the predictive region, mean predicted growing season soil respiration was 73% of MODIS GPP, while predicted soil respiration was generally within 20% of nocturnal NEE from eddy covariance towers. This study demonstrated that geospatial and remotely-sensed datasets can be used in a statistical modeling framework to estimate soil respiration at landscape scales.

Colorado, Wyoming↗
USGS2021

Characterization of groundwater recharge and flow in California's San Joaquin Valley from InSAR-observed surface deformation

Surface deformation in California's Central Valley (CV) has long been linked to changes in groundwater storage. Recent advances in remote sensing have enabled the mapping of CV deformation and associated changes in groundwater resources at increasingly higher spatiotemporal resolution. Here, we use interferometric synthetic aperture radar (InSAR) from the Sentinel‐1 missions, augmented by continuous Global Positioning System (cGPS) positioning, to characterize the surface deformation of the San Joaquin Valley (SJV, southern two‐thirds of the CV) for consecutive dry (2016) and wet (2017) water years. We separate trends and seasonal oscillations in deformation time series and interpret them in the context of surface and groundwater hydrology. We find that subsidence rates in 2016 (mean −42.0 mm/yr; peak −345 mm/yr) are twice that in 2017 (mean −20.4 mm/yr; peak −177 mm/yr), consistent with increased groundwater pumping in 2016 to offset the loss of surface‐water deliveries. Locations of greatest subsidence migrated outwards from the valley axis in the wetter 2017 water year, possibly reflecting a surplus of surface‐water supplies in the lowest portions of the SJV. Patterns in the amplitude of seasonal deformation and the timing of peak seasonal uplift reveal entry points and potential pathways for groundwater recharge into the SJV and subsequent groundwater flow within the aquifer. This study provides novel insight into the SJV aquifer system that can be used to constrain groundwater flow and subsidence models, which has relevance to groundwater management in the context of California's 2014 Sustainable Groundwater Management Act (SGMA).

California↗
USGS2025

Effect of land cover type on 3D deformation recovery from synthetically deformed high-resolution satellite optical imagery

The limits of detection for earthquake surface deformation in the spatial domain have improved with advances in remote sensing imagery data availability, resolution, and analysis. Sub-pixel correlation and digital elevation model (DEM) differencing from sub-meter, earthquake-spanning satellite optical imagery has enhanced surface rupture mapping and deformation measurements. However, knowledge of measurement accuracy and uncertainty is limited. To address this, we construct orthophotos and digital elevation models (DEMs) from repeat high resolution (∼0.5 m) satellite optical imagery along two sections of the Garlock fault, California with clear fault geomorphology and differing land cover. We deform later sets of DEMs and images with synthetic earthquakes containing both diffuse and discrete horizontal and vertical displacements. Sub-pixel image correlation and DEM differencing demonstrate how vegetation degrades recovered displacement accuracy. In barren land cover, horizontal displacements are detectable to an expected ∼1/10th-pixel size. With shrubs, trees, and grass, detectable displacements increase to >1/2-pixel size, and filtering results by correlation score and using elevation values as input rather than image values improves accuracy. Vertical displacement detection thresholds remain lower in vegetation, at >1-pixel size. Higher slope angles degrade displacement recovery, worsened by vegetation. Diminishing seasonal separation improves accuracy over vegetated regions, though not to the level achieved in barren environments. These results will inform research and operational efforts on the utility of high resolution satellite optical imagery for detecting deformation in varied land cover. Furthermore, they reveal where alternative measurements, such as from LiDAR or radar interferometry, are required to mitigate the effects of vegetation and capture fine-scale crustal deformation.

California↗
USGS2023

Data integration reveals dynamic and systematic patterns of breeding habitat use by a threatened shorebird

Incorporating species distributions into conservation planning has traditionally involved long-term representations of habitat use where temporal variation is averaged to reveal habitats that are most suitable across time. Advances in remote sensing and analytical tools have allowed for the integration of dynamic processes into species distribution modeling. Our objective was to develop a spatiotemporal model of breeding habitat use for a federally threatened shorebird (piping plover, Charadrius melodus ). Piping plovers are an ideal candidate species for dynamic habitat models because they depend on habitat created and maintained by variable hydrological processes and disturbance. We integrated a 20-year (2000–2019) nesting dataset with volunteer-collected sightings (eBird) using point process modeling. Our analysis incorporated spatiotemporal autocorrelation, differential observation processes within data streams, and dynamic environmental covariates. We evaluated the transferability of this model in space and time and the contribution of the eBird dataset. eBird data provided more complete spatial coverage in our study system than nest monitoring data. Patterns of observed breeding density depended on both dynamic (e.g., surface water levels) and long-term (e.g., proximity to permanent wetland basins) environmental processes. Our study provides a framework for quantifying dynamic spatiotemporal patterns of breeding density. This assessment can be iteratively updated with additional data to improve conservation and management efforts, because reducing temporal variability to average patterns of use may cause a loss in precision for such actions.

Montana, North Dakota, South Dakota↗
USGS2026

Reassessing water availability in the Nile River Basin using satellite-based irrigation water use accounting

Existing estimates of irrigation areas and water use in the Nile River Basin vary widely and are based on inconsistent data and methodologies. This study leverages advances in remote sensing to provide an up-to-date and consistent basin-wide reassessment of water use in the Nile River Basin. Land cover and actual evapotranspiration data from 2013 to 2022 were used to quantify irrigated areas, irrigation water use, and the average naturalized water yield. Across the basin, 7.1 million hectares of irrigated land consume an average of 72.5 ± 2.6 billion cubic meters (BCM) of blue water annually. Egypt and Sudan together account for 92% of the irrigated area and 97% of the irrigation water use. The basin’s average naturalized yield, evaluated at Dongola station, is determined to be between 108 and 114.4 ± 3.2 BCM. Although conservative, this updated analysis indicates the basin’s average naturalized yield is 14%–20% higher than the previous naturalized yield estimate (95 BCM) and 29%–36% higher than the commonly cited observed flow of 84 BCM, which is frequently treated as the total basin-wide available water. Importantly, the study underscores that the 84 BCM represents flow at the Aswan gauging station after upstream consumptive uses, not the total water available in the basin under naturalized conditions. This distinction is critical for accurate water accounting, planning, and governance in this water-scarce basin. This study demonstrates the value of open-source remote sensing resources in data-scarce regions while emphasizing the need for region specific validation and bias correction for improved accuracy.

Nile River Basin↗
USGS2026

Perspectives and insights gained from applying landscape ecological approaches to the Upper Mississippi riverscape

Advances in remote sensing, spatial data collection, hydraulic and geospatial models, and the maturation of landscape ecology during the early 2000s spurred the concept of rivers as landscapes or ‘riverscapes’. Since that time, the Upper Mississippi River Restoration (UMRR) Program - one of the world’s longest running large-river ecosystem restoration and monitoring programs - has applied landscape ecological principles to the Upper Mississippi River System (UMRS). I examine spatial patterns of aquatic nutrient concentrations, submersed aquatic vegetation, and fish and mussel communities, finding that within-river patterns are often patchy in nature, likely resulting from how nutrient processes and community traits interact with spatial variability in hydraulic exchange patterns. Local-scale patchy spatial patterns in the UMRS are embedded in system-scale longitudinal gradients and a hierarchical view of the river has provided a foundation for strategic river restoration planning. Key Words: Aquatic Habitat, hydraulic connectivity, patch, restoration, riverscape

BioScience↗
USGS2022

Climate change risks to global forest health: Emergence of unexpected events of elevated tree mortality worldwide

Recent observations of elevated tree mortality following climate extremes, like heat and drought, raise concerns about climate change risks to global forest health. We currently lack both sufficient data and understanding to identify whether these observations represent a global trend towards increasing tree mortality. Here, we document events of sudden and unexpected elevated tree mortality following heat and drought events in ecosystems that previously were considered tolerant or not at risk of exposure. These events underscore the fact that climate change may affect forests with unexpected force in the future. We use the events as examples to highlight current difficulties and challenges for realistically predicting such tree mortality events and the uncertainties about future forest condition. Advances in remote sensing technology and greater availably of high-resolution data, from both field assessments and from satellites, are needed to improve both understanding and prediction of forest responses to future climate change.

California, New Mexico↗
USGS2025

Artificial neural network multilayer perceptron models to classify California’s crops using Harmonized Landsat Sentinel (HLS) data

Advances in remote sensing and machine learning are enhancing cropland classification, vital for global food and water security. We used multispectral Harmonized Landsat 8 Sentinel-2 (HLS) 30-m data in an artificial neural network (ANN) multi-layer perceptron (MLP) model to classify five crop classes (cotton, alfalfa, tree crops, grapes, and others) in California's Central Valley. The ANN MLP model, trained on 2021 data from the United States Department of Agriculture's Cropland Data Layer, was validated by classifying crops for an independent year, 2022. Across the five crop classes, the overall accuracy was 74%. Producer's and user's accuracies ranged from 65% to 87%, with cotton achieving the highest accuracies. The study highlights the potential of using deep learning with HLS time series data for accurate global crop classification.

California↗
USGS2011

Mapping and monitoring Louisiana's mangroves in the aftermath of the 2010 Gulf of Mexico Oil spill

Information regarding the present condition, historical status, and dynamics of mangrove forests is needed to study the impacts of the Gulf of Mexico oil spill and other stressors affecting mangrove ecosystems. Such information is unavailable for Louisiana at sufficient spatial and thematic detail. We prepared mangrove forest distribution maps of Louisiana (prior to the oil spill) at 1 m and 30 m spatial resolution using aerial photographs and Landsat satellite data, respectively. Image classification was performed using a decision-tree classification approach. We also prepared land-cover change pairs for 1983, 1984, and every 2 y from 1984 to 2010 depicting “ecosystem shifts” (e.g., expansion, retraction, and disappearance). This new spatiotemporal information could be used to assess short-term and long-term impacts of the oil spill on mangroves. Finally, we propose an operational methodology based on remote sensing (Landsat, Advanced Spaceborne Thermal Emission and Reflection Radiometer [ASTER], hyperspectral, light detection and ranging [LIDAR], aerial photographs, and field inventory data) to monitor the existing and emerging mangrove areas and their disturbance and regrowth patterns. Several parameters such as spatial distribution, ecosystem shifts, species composition, and tree height/biomass could be measured to assess the impact of the oil spill and mangrove recovery and restoration. Future research priorities will be to quantify the impacts and recovery of mangroves considering multiple stressors and perturbations, including oil spill, winter freeze, sea-level rise, land subsidence, and land-use/land-cover change for the entire Gulf Coast.

Journal of Coastal Research↗
USGS2022

A sagebrush conservation design to proactively restore America’s sagebrush biome

A working group of experts with diverse professional backgrounds and disciplinary expertise was assembled to conceptualize a spatially explicit conservation design to support and inform the Sagebrush Conservation Strategy Part 2. The goal was to leverage recent advancements in remotely sensed landcover products to develop spatially and temporally explicit maps of sagebrush rangeland condition and landscape threats. In addition, the group sought to provide a common basis for understanding the state of sagebrush rangelands through time. First, the study team developed a spatially explicit model to assess geographic patterns in sagebrush ecological integrity and used this model to identify core sagebrush areas (CSAs), growth opportunity areas (GOAs), and other rangeland areas (ORAs) across the biome. Among the identified rangelands, 33.4 million acres were classified as CSAs; 84.3 million acres as GOAs; and 127.2 million acres as ORAs as of 2020. Second, the team sought to demonstrate the ecological relevance of the identified CSAs and GOAs by comparing these data with independent datasets for sagebrush obligate species of conservation concern. Geographical patterns in sagebrush ecological integrity were strongly associated with the occurrence of high-priority species and also displayed clear links to population performance for greater sage-grouse. Third, the team parsed out the type, location, and acres of primary threats within the different categories (CSAs, GOAs, and ORAs) to help focus active management by identifying places where multiagency and organization efforts can protect CSAs and GOAs that have higher levels of integrity with lower cumulative threats. The assessment of the condition of the sagebrush biome (that is, the location, amount, and conservation status) indicated that complex ecosystem function problems are driving ~73 percent of the demonstrated threats within the CSAs and GOAs (rather than point-source problems, such as human development). Fourth, the team developed trend estimates for the identified CSAs and GOAs and three selected primary threats (invasive annual grasses, conifer encroachment, and human modification) to the sagebrush biome from 2001 to 2020. Results showed that an average of 1.3 million acres per year have transitioned to ORAs at an annual rate of −1.34 percent. Fifth, the team developed an approach to integrate climate change effects into the threat-based landscape conservation design and conducted an initial assessment on the magnitude of near-term climate effects in the context of observed historical trends. The team’s analysis suggests that climate change alone is unlikely to be the dominant threat to sagebrush ecological integrity in the next few decades, although interactions of climate with wildfire and invasive annual grasses may be an important threat, especially in the longer term. A spatial overlap analysis was performed and highlighted 45.8 million acres of shared priorities among existing conservation frameworks to help anchor and guide collaborative landscape-scale conservation of areas that still have no to low threats. This information is critical to provide context for decisions about the volume and nature of conservation actions and funding requirements.

western United States↗
USGS2014

High spatial resolution WorldView-2 imagery for mapping NDVI and its relationship to temporal urban landscape evapotranspiration factors

Evapotranspiration estimation has benefitted from recent advances in remote sensing and GIS techniques particularly in agricultural applications rather than urban environments. This paper explores the relationship between urban vegetation evapotranspiration (ET) and vegetation indices derived from newly-developed high spatial resolution WorldView-2 imagery. The study site was Veale Gardens in Adelaide, Australia. Image processing was applied on five images captured from February 2012 to February 2013 using ERDAS Imagine. From 64 possible two band combinations of WorldView-2, the most reliable one (with the maximum median differences) was selected. Normalized Difference Vegetation Index (NDVI) values were derived for each category of landscape cover, namely trees, shrubs, turf grasses, impervious pavements, and water bodies. Urban landscape evapotranspiration rates for Veale Gardens were estimated through field monitoring using observational-based landscape coefficients. The relationships between remotely sensed NDVIs for the entire Veale Gardens and for individual NDVIs of different vegetation covers were compared with field measured urban landscape evapotranspiration rates. The water stress conditions experienced in January 2013 decreased the correlation between ET and NDVI with the highest relationship of ET-Landscape NDVI (Landscape Normalized Difference Vegetation Index) for shrubs (r 2 = 0.66) and trees (r 2 = 0.63). However, when the January data was excluded, there was a significant correlation between ET and NDVI. The highest correlation for ET-Landscape NDVI was found for the entire Veale Gardens regardless of vegetation type (r 2 = 0.95, p > 0.05) and the lowest one was for turf (r 2 = 0.88, p > 0.05). In support of the feasibility of ET estimation by WV2 over a longer period, an algorithm recently developed that estimates evapotranspiration rates based on the Enhanced Vegetation Index (EVI) from MODIS was employed. The results revealed a significant positive relationship between ETMODIS and ETWV2 (r 2 = 0.9857, p > 0.05). This indicates that the relationship between NDVI using high resolution WorldView-2 imagery and ground-based validation approaches could provide an effective predictive tool for determining ET rates from unstressed mixed urban landscape plantings.

Adelaide↗
USGS2014

Global land cover mapping: a review and uncertainty analysis

Given the advances in remotely sensed imagery and associated technologies, several global land cover maps have been produced in recent times including IGBP DISCover, UMD Land Cover, Global Land Cover 2000 and GlobCover 2009. However, the utility of these maps for specific applications has often been hampered due to considerable amounts of uncertainties and inconsistencies. A thorough review of these global land cover projects including evaluating the sources of error and uncertainty is prudent and enlightening. Therefore, this paper describes our work in which we compared, summarized and conducted an uncertainty analysis of the four global land cover mapping projects using an error budget approach. The results showed that the classification scheme and the validation methodology had the highest error contribution and implementation priority. A comparison of the classification schemes showed that there are many inconsistencies between the definitions of the map classes. This is especially true for the mixed type classes for which thresholds vary for the attributes/discriminators used in the classification process. Examination of these four global mapping projects provided quite a few important lessons for the future global mapping projects including the need for clear and uniform definitions of the classification scheme and an efficient, practical, and valid design of the accuracy assessment.

Remote Sensing↗
USGS2023

Automated mapping of culverts, bridges, and dams

Accurate maps of built structures around stream channels, such as dams, culverts, and bridges, are vital in monitoring infrastructure, risk management, and hydrologic modeling. Hydrologic modeling is essential for research and decisionmaking related to infrastructure and development planning, emergency management, ecology, and developing hydrographic data. Technological advances in remote sensing afford increasingly fine-scale elevation data, such as the U.S. Geological Survey 1-meter digital elevation models (DEMs), that can accurately model the Earth’s surface characteristics and related hydrologic dynamics. A long-standing challenge in flow modeling is the presence of built structures in an elevation model that resist flow in a way that does not reflect actual dynamics, such as culverts, bridges, and dams. This challenge is exacerbated in fine-scale elevation data as more built structures are resolved. Here we present a test of the extensibility of a culvert and dam detection workflow, culvert-net (CN). CN was developed using a large dataset of field-validated culverts, bridges, and dam locations for Alexander County, North Carolina, USA, supplemented by manual review and identification of additional features. In this workflow, the CN model is tested on a new study area in western Michigan, USA, where culverts and associated hydrography have recently been manually compiled.

Conference Paper↗
USGS2018

Detecting snow depth change in avalanche path starting zones using uninhabited aerial systems and structure from motion photogrammetry

Understanding snow depth distribution and change is useful for avalanche forecasting and mitigation, runoff forecasting, and infrastructure planning. Advances in remote sensing are improving the ability to collect snow depth measurements. The development of structure from motion (SfM), a photogrammetry technique, combined with the use of uninhabited aerial systems (UASs) allows for high resolution mapping of snow depth over complex terrain. The primary objective of this study was to determine the feasibility and efficacy of SfM to examine snow depth distribution and variability in complex terrain such as avalanche path starting zones at multiple times during the season. We used a 3DR Solo quadcopter UAS equipped with a Ricoh GR II camera at 90 m above ground level to acquire images of one avalanche starting zone in northwest Montana, USA. We also placed 4 to 13 ground control points (GCPs) around the area of interest to avoid traveling in steep, avalanche terrain. Ground control measurements resulted in 5 to10 cm horizontal accuracy and 5 to 15 cm vertical accuracy for 90 to 95 % of the collected points (a minimum of 100 points collected at each GCP). In-situ measurements of snow depth difference between sampling days ranged from 20 to 60 cm. We processed the images to create point clouds and digital surface models (DSMs). The resolution of the resultant DSMs was approximately 5 cm. Preliminary DSM and point cloud differencing efforts suggest relative change detection of snow depth at 5 to 15 cm resolution. The use of these relatively low cost and easily accessible methods of snow depth data collection will enhance accuracy of snow depth change estimates in starting zones and can be used to inform avalanche forecasting and mitigation efforts.

Conference Paper↗
USGS2021

Bayesian updating of seismic ground failure estimates via causal graphical models and satellite imagery

Earthquake-induced secondary ground failure hazards, such as liquefaction and landslides, result in catastrophic building and infrastructure damage as well as human fatalities. To facilitate emergency responses and mitigate losses, the U.S. Geological Survey provides a rapid hazard estimation system for earthquake-triggered landslides and liquefaction using geospatial susceptibility proxies and ShakeMap ground motion estimates. However, the resolution and accuracy of these models are often limited by coarse-granularity and large uncertainties of available geospatial features provided at a regional scale. Recently, with the advancement of remote sensing technologies, synthetic aperture radar (SAR) images are captured and analyzed to obtain a rapid estimate of earthquake-induced correlation changes between pre- and post-event images. These correlation changes indicate ground failures and building damage t, showing the potential to provide supplementary information for rapid hazard and loss estimation. However, the exact causes of changes in satellite images are not directly ascertained by the DPM alone. For example, changes could be due to building damage, landslides, liquefaction, noise or any combination thereof. More importantly, the occurrence and intensity of landslides, liquefaction, and building damages are spatially correlated, which makes it yet more challenging to distinguish the sources of any such changes. In this study, we develop a generalized causal graph-based Bayesian Network that models the physical interdependencies between geospatial features, seismic ground failures and building damage, as well as DPMs. Geospatial features provide physical insights for estimating ground failure occurrence while DPMs contain event-specific surface change observations. This physics-informed causal graph incorporate these variables with complex physical relationships in one holistic Bayesian updating scheme to effectively fuse information from both geospatial models and remote sensing data. This framework is scalable and flexible enough to deal with highly complex multi-hazard combinations. We then develop a stochastic variational inference algorithm to jointly update the intractable posterior probabilities of unobserved landslides, liquefaction, and building damage at different locations efficiently. In addition, a local graphical model pruning algorithm is presented to reduce the computational cost of large-scale seismic ground failure estimation. We apply this framework to September 2018 Hokkaido Iburi-Tobu, Japan (M6.6) earthquake and January 2020 Southwest Puerto Rico (M6.4) earthquake to evaluate the performance of our algorithm

Conference Paper↗