Closing Date: January 6, 2022
This Research Opportunity will be filled depending on the availability of funds. All application materials must be submitted through USAJobs by 11:59 pm, US Eastern Standard Time, on the closing date.
Background: Interferometric Synthetic Aperture Radar (InSAR) is a technique that combines radar images from nearly the same orbit at two different times to form an image of how the ground has deformed over the time spanned. InSAR-capable satellites, which provide synoptic views of volcanoes all over the globe, are growing in number and capabilities, with Cosmo-SkyMed, TerraSAR-X, PAZ, ICEYE, Capella, Sentinel-1, ALOS-2, RADARSAT-2. RADARSAT Constellation Mission, and the future NISAR mission (Jan 2023) providing large volumes of data with which volcano deformation and surface change can be monitored. Unfortunately, data volume is outpacing the ability to process, view, and analyze results. Current processing and analysis methods often require expert supervision, which can limit the utility of InSAR data during rapidly evolving volcanic unrest. This project proposes a collaboration between the USGS Volcano Science Center and Advanced Research Computing Center to automate some InSAR analysis and reduce the human time investment for monitoring volcano deformation.
Artificial intelligence/machine learning (AI/ML), particularly deep learning techniques for computer vision, for image analysis have recently provided a pathway for efficient classification of large volumes of image data into a monitoring framework by letting software filter through the majority of "quiet" data while flagging and classifying changes. The combination of automated processing with AI/ML analysis would feed directly into the fledgling NVEWS Volcano Data Center, serving as an example for how other high-volume satellite datasets, like those tracking ash, gas, and thermal emissions, might best be utilized.
In recent years, AI/ML or deep-learning techniques, have been applied to volcanic to efficiently and automatically classifying large volumes of InSAR data. In particular, deep-learning techniques such as convolutional neural networks (CNNs) have been used to detect, locate, and classify the presence of co-seismic-like surface deformation in an interferogram. Deep-learning techniques have also been used to automate noise removal and signal extraction from InSAR time series, making possible to study of small deformations.
We propose to develop deep learning techniques that can be used to automate the detection of anomalous variations in ground deformation on volcanoes. Once changes are identified, determine if those changes are associated with established patterns of volcanic activity. The technical workflows and tools developed during this activity could be generalized to help automate the processing and interpretation of similar imagery data (e.g., optical satellite, webcam, or thermal imagery) used in volcano monitoring.
High-threat volcanoes in the Western USA are ranked by their eruptive history, observed unrest, hazards, and exposure of people and infrastructure to those hazards. Many high-threat volcanoes are actively deforming, and InSAR has been used effectively to document the deformation. Some examples include Lassen Peak, Long Valley Caldera, Mt. St. Helens, Mauna Loa, Yellowstone, Akutan, and Okmok.
Description of the Research Opportunity: We seek a Mendenhall Fellow to investigate deformation of volcanoes in the USA with novel analysis methods that will improve monitoring capabilities (Natural Hazards Mission Area Priority VH1.2) and better exploit the vast volume of available InSAR data. These data will help inform hazard assessments (VH2.2) and models of subsurface volcanic structure (VH3.4). Available satellite InSAR data covering US volcanoes include images that are freely available from the Sentinel-1 mission, as well as TerraSAR-X, COSMO-SkyMed, and other data that are made available via research agreements and special initiatives (like Geohazards Supersites and Natural Laboratories, which provides comprehensive satellite InSAR coverage of Hawaiian volcanoes). The pending NISAR mission (scheduled for launch in January 2023) will supplement these datasets with large volumes of freely available images using a wavelength that can penetrate most vegetative cover. Additional geodetic data include campaign and continuous GPS/GNSS data both from USGS and other regional GNSS networks (including the Network of the Americas, to which the USGS is a contributor), and in some cases gravity, strainmeter and tiltmeter data. The potential Fellow should have exposure AI/ML analysis methods and their use in image analysis, as well as familiarity with geodetic data (InSAR, and potentially GNSS, tilt, strain, gravity).
Potential Fellows could apply novel analysis techniques to a number of volcanic systems in the USA where InSAR has been documented to effectively observe ongoing deformation. These cases, which provide a range of temporal and spatial scales of deformation (for instance, transient deformation at Kilauea, long-term variable inflation at Mauna Loa, inflation episodes of Okmok, episodic unrest at Long Valley caldera, localized ground motion at Mount St. Helens, and broad surface displacements at Yellowstone), could provide important “training” datasets for the range of deformation styles expected at US volcanoes. Although there are many potential volcanoes to study, we provide descriptive details of two potential areas: Mauna Loa, HI, and Long Valley Caldera, CA.
Unrest at Mauna Loa. Deformation at Mauna Loa volcano has been occurring with changing amplitudes and source locations for the past 25 years. Recent accelerated deformation since 2019, along with increasing seismicity, prompted an elevation of the alert level from Green to Yellow. Frequently changing deformation at Mauna Loa, combined with excellent InSAR coherence over most of the volcano make it an ideal candidate study area for deep-learning techniques.
Episodic inflation at Long Valley Caldera. Episodes of inflation at Long Valley Caldera have been observed since permanent instrumentation was installed in the early 1980’s, and a very long-duration record of multidisciplinary data exists for the caldera. Inflation episodes have occurred rapidly (e.g., 1997) as well as extended for longer periods (e.g. 2011-2021). Previous studies have shown that InSAR coherence is very good in the relatively dry caldera. Multiple signals (multi-year volcanic, annual hydrologic transients, and potential instantaneous tectonic earthquakes) at Long Valley are superimposed at different spatial and temporal scales providing a variety of potential signals for classification by a deep-learning process.
In addition to satellite-based observations, US volcano observatories have extensive networks of ground-based data, including GNSS, tiltmeters, strainmeters, gravity, and data from other disciplines (seismic, gas, geochemistry) that could be incorporated into an AI/ML framework.
Interested applicants are strongly encouraged to contact the Research Advisor(s) early in the application process to discuss project ideas.
Proposed Duty Station: Vancouver, Washington
Areas of PhD: Geophysics, geology, physics, engineering, computer science or applied mathematics with a focus on geodesy or image analysis, or related fields (candidates holding a Ph.D. in other disciplines, but with extensive knowledge and skills relevant to the Research Opportunity may be considered).
(This type of research is performed by those who have backgrounds for the occupations stated above. However, other titles may be applicable depending on the applicant's background, education, and research proposal. The final classification of the position will be made by the Human Resources specialist.)
Human Resources Office Contact: Veronica Guerrero-Nunez, 916-278-9405, firstname.lastname@example.org