The Chesapeake Bay region (defined as longitudes − 78° to -74° and latitudes 36.5° to 40°) experiences the highest rates of relative sea-level rise (RSLR) on the Atlantic Coast. Regional land subsidence influences RSLR, however quantified rates of vertical land motions (VLM) are inconsistent in published solutions. For 5 years from 2019 to 2023, new Global Navigation Satellite System (GNSS) campaign data were collected at over 60 sites across the Chesapeake Bay region annually. These data were processed and combined with continuous GNSS data (120 stations) from the region covering the same time-period using GAMIT-GLOBK to produce 3D velocities and their associated uncertainties. We use the Robust Network Imaging algorithm to interpolate GNSS-derived VLM to produce a new regional VLM solution of the Chesapeake Bay region. We find that land subsidence is ubiquitous throughout the region with rates varying from − 2.97 to -0.40 mm/yr. In major cities across the Chesapeake Bay region, VLM rates are − 1.1 ± 1.6 mm/yr (1-sigma) for Washington DC, -0.8 ± 1.4 mm/yr for Baltimore, MD, -2.4 ± 0.5 mm/yr for Ocean City, MD, and − 2.3 ± 1.0 mm/yr for Hampton, VA. When we compare our VLM rates with a geodetic-based solution from 1974, we observe meaningful shifts in the locations and rates of maximum subsidence. The results of this work underscore that regular monitoring of VLM and can be used to improve projections of relative sea-level changes as well as the associated coastal hazards for communities in the Chesapeake Bay region.

Abstract Shallow creep events provide opportunities to understand the mechanical properties and behavior of faults. However, due to physical limitations observing creep events, the precise spatio‐temporal evolution of slip during creep events is not well understood. In 2023, the Superstition Hills and Imperial faults in California each experienced centimeter‐scale slip events that were captured in unprecedented detail by satellite radar, sub‐daily Global Navigation Satellite Systems, and creepmeters. In both cases, the slip propagated along the fault over 2–3 weeks. The Superstition Hills event propagated bilaterally away from its initiation point at average velocities of ∼9 km/day, but propagation velocities were locally much higher. The ruptures were consistent with slip from tens of meters to ∼2 km depths. These slowly propagating events reveal that the shallow crust of the Imperial Valley does not obey purely velocity‐strengthening or velocity‐weakening rate‐and‐state friction, but instead requires the consideration of fault heterogeneity or fault‐frictional behaviors such as dilatant strengthening.

Overview Measuring accurately the relative movement of the surface of the Earth is a critical constraint on the slow and broad tectonic loading and unloading to which faults respond, and is one of the few observations of the solid Earth that may be made directly without inference. High-precision geodetic observations, such as from Global Navigation Satellite Systems (GNSS), which includes the Global Positioning System (GPS), and interferometric synthetic aperture radar (InSAR), allow measurement of fault motions between, during and in the aftermath of earthquakes and other related tectonic phenomena, densely in both space and time. The Community Geodetic Model (CGM) provides velocities and time series of observed points on the Earth's surface over Southern California using data from a number of contributing researchers, institutions and analysis centers. The GNSS products provide high temporal resolution (nominally daily measurement points for continuous stations) in three dimensions at specific observation sites and the InSAR products provide high spatial resolution (approximately one point per tens of m on the ground, depending on exact specifications of data and processing). Combined, they provide the ability to study crustal deformation over a wide range of distances and periods. The CGM differs from other SCEC Community Models in that it is constantly extending with time as new data are acquired daily, so it is not static. The CGM version 1 (2016; doi:10.5281/zenodo.4926528) was a collection of time-independent (velocity-only) geodetic products gathered from published papers. The GNSS velocities were then combined and modeled by a Working Group researching methods and contributing interpolated deformation fields. The main goal of the CGM version 2 is to add time-dependent (time series) products to both the GNSS and InSAR products. For the GNSS, this is done by ingesting survey and (mostly) continuous time series from five analysis centers in the U.S.: the Geodetic Facility for the Advancement of Geoscience (GAGE); the Nevada Geodetic Laboratory (NGL) at the University of Nevada, Reno (UNR); the NASA Jet Propulsion Laboratory (JPL) and Scripps Orbital and Permanent Array Center (SOPAC) contributions to the MEaSUREs ESESES project; and the U.S. Geological Survey (USGS). Like the various contributions to the CGMv1 GNSS velocities, these time series are rigorously adjusted to be self-consistent, before a weighted mean is calculated to produce the consensus products. Much of the InSAR contribution is a consensus from research by the SCEC community within the CGM (InSAR) Working Group, whose individual contributions are listed below and in more detail in the README.txt file in the top directory of the archive. The CGMv2 is therefore a "union" or "superset" of survey and continuous GNSS and InSAR time series. Please see https://www.scec.org/research/cgm for more information. Version: CGMv2.0.0 This is the second major release of the CGM (version 2.0.0) and is distributed as a zip-file. See below and in the README.txt file for information about the directory structure and contents of the entire zipped archive. Much of the SCEC5 activity was focused on the assembly of GNSS and InSAR time series for measuring temporally variable motions, expanding the CGMv1 with the time dimension. The CGMv2.0.0 is a time-dependent set of products, consisting of time series and velocities of the Earth's surface measured by GNSS and InSAR. Directory Structure and Contents data/gnss/pos/The CGMv2.0.0 GNSS time series in "pos" format (plain text), relative to various reference frames described below. Header lines in each file provide information about the nominal reference coordinates and data columns. Files named "*.wmrss_*" are the continuous stations (weighted mean with rescaled sigma) and files named "*.final_" are the survey sites. data/gnss/pos/igb14/ The International GNSS Service's (IGS's) revised realization of ITRF2014 data/gnss/pos/nam14/ North America defined by Altamimi et al.'s (2017) ITRF2014 plate motion model data/gnss/pos/pcf14/ The Pacific defined by Altamimi et al.'s (2017) ITRF2014 plate motion model data/gnss/pos/nam17/ North America defined by Kreemer et al. (2018) data/gnss/vel/The CGMv2.0.0 GNSS velocities in a CSV file similar to GAGE's "vel" format (plain text), relative to the same reference frames described above. Header lines in each file provide information about the data columns. data/insar/The CGMv2.0.0 InSAR line-of-sight consensus time series and velocities for four ESA Sentinel-1 tracks (ascending tracks 64 and 166, and descending tracks 71 and 173) over Southern California, in an HDF5 format designed for the CGM. A description of and reader for the HDF5 files may be found here. data/insar/contrib/Individual contributions to the InSAR time series and velocity products, as described below and in more detail in the top-level README.txt file. Contributors The GNSS time series are a weighted mean, after restoration of global scale if processed using Gipsy (JPL, NGL/UNR and USGS) and self-consistent alignment of reference frame, of the following GNSS analysis centers, whose products are publicly available at the embedded hyperlinks: The Geodetic Facility for the Advancement of Geoscience (GAGE) (Herring et al., 2016) The Nevada Geodetic Laboratory at the University of Nevada, Reno (Blewitt et al., 2018) NASA's Jet Propulsion Laboratory contribution to the MEaSUREs ESESES Project SOPAC's contribution to the MEaSUREs ESESES Project The United States Geological Survey (Murray and Svarc, 2017) Zheng-Kang Shen's (UCLA) survey time series Z.-K. Shen processed the raw data from the SCEC survey-mode GPS data archive to provide the corresponding time series and velocities. A. Gonzalez Ortega provided processed time series from CICESE's REGNOM network of continuous GNSS stations. M. Floyd and T. Herring designed the download, alignment and combination of the publicly available continuous GNSS archives, listed above, in various reference frames. Contributions from individuals and institutions within the SCEC community to the CGM (InSAR) products are: K. Wang contributed time series and velocity solutions K. Guns and X. Xu contributed time series and velocity solutions Z. Liu contributed time series and velocity solutions S. Sangha, M. Govorcin and D. Bekaert contributed time series and velocity solutions G. Funning contributed time series and velocity solutions E. Tymofyeyeva calculated the combination of contributed solutions to generate the consensus product K. Materna contributed time series and velocity solutions, and wrote the translation tools for converting to and from HDF5 format, as designed by all InSAR contributors listed immediately above plus M. Floyd Three groups (K. Guns and X. Xu; Z. Liu; and S. Sangha, M. Govorcin and D. Bekaert) independently processed interferograms from common raw datasets using different processing approaches. E. Tymofyeyeva coordinated and led the InSAR Working Group. M. Floyd coordinated and led the wider CGM Working Group. All contributed to the design of the HDF5 format in which the InSAR products are distributed.

A new reanalysis of Global Navigation Satellite System (GNSS) data at or near tide gauges worldwide was produced by the University of La Rochelle (ULR) group within the third International GNSS Service (IGS) reprocessing campaign (repro3). The new solution, called ULR-repro3, complies with the IGS standards adopted for repro3, implementing advances in data modelling and corrections since the previous reanalysis campaign and extending the average record length by about 7 years. The results presented here focus on the main products of interest for sea level science: the station position time series and associated velocities on the vertical component at tide gauges. These products are useful to estimate accurate vertical land motion at the coast and supplement data from satellite altimetry or tide gauges for an improved understanding of sea level changes and their impacts along coastal areas. To provide realistic velocity uncertainty estimates, the noise content in the position time series was investigated considering the impact of non-tidal atmospheric loading. Overall, the ULR-repro3 position time series show reduced white noise and power-law amplitudes and lower station velocity uncertainties compared with the previous reanalysis. The products are available via SONEL (https://doi.org/10.26166/sonel_ulr7a; Gravelle et al., 2022).

Several different basis functions have been used to represent the Earth's gravity field in order to generate estimates of mass variations on Earth from the analysis of data of the Gravity Recovery and Climate Experiment ( grace ) and its successor grace Follow‐On missions, including spherical harmonics, mass concentration elements (mascons) and slepian functions. Each approach depends inherently upon accurate modeling of the orbits of the pair of satellites as they revolve around the Earth, so that the observations of inter‐satellite changes in range (or, more specifically, range rate) can be exploited to identify mass variations. We have developed software using a classical orbit modeling approach, mascons and 24‐hr orbit integration, to estimate simultaneously corrections to orbital parameters and the temporal gravity field from grace data. Rather than using the range rate, we use the range acceleration as the inter‐satellite observable as it aids in localizing the mass variations. Level‐1 B range acceleration observations contain high levels of high‐frequency noise that inhibits their usefulness for this purpose. Instead, we generate range acceleration observations by numerical differentiation of the Level‐1B range rate prefit residuals. Simulations show that the gravity signal is not attenuated in this process. Our monthly estimates of mass anomalies from grace data (2003–2016) agree well with previous studies, both spatially and temporally. When converted to spherical harmonics our time series of C 2,0 , derived from grace data alone, are close to the independent estimates from satellite laser ranging, but the overall solution is improved by substituting the SLR C 2,0 .

Interferometry from satellite radar has thrived as a major asset to study surface deformations from earthquakes, volcanoes, aquifers, glaciers, landslides, etc. Most signals recorded in an interferogram have precise enough models to remove them almost completely. Yet, current models still fail to capture the full range and scales of variations of atmospheric perturbations. This work explores the use of an image-to-image translation method, CycleGAN, to learn a function that wraps and improves an imperfect model for atmospheric correction. CycleGAN is a generative adversarial network, in which discriminators determine which images are real, and generators produce images to fool the discriminators. Training the discriminators and generators against each other improves the resulting translation function. We have tested this approach using Sentinel-1A data gathered around Puebla, Mexico, which was the starting point of the 2017 earthquake that devastated Mexico City. CycleGAN can generate visually compelling fringes, including small-scale perturbations that are absent from the atmospheric models, while limiting noisy areas. However, it fails to capture the variations of amplitude behind the fringes, especially at large scale, and the generated interferograms remain too different from the real interferograms. Solving that amplitude issue could create practical applications for a CycleGAN-type method in atmospheric correction or phase unwrapping. The large amount of InSAR data and the continuous progress of deep learning methods provide ample opportunity for improvement.

The M-w 6.4 and M-w 7.1 Ridgecrest, California, earthquakes of July 2019 occurred within 34 hr of each other on conjugate strike-slip faults in the Mojave Desert, just north of the central Garlock fault. Here, we present the results of a survey of 18 Global Navigation Satellite Systems (GNSS) sites conducted in the immediate aftermath of the earth- quakes, including five sites that recorded the motion of the second earthquake after having been set up immediately following the first, as well as processed results from continuous GNSS sites throughout the region. Our field work in response to the earthquakes provides additional constraints on the ground displacement due to both earthquakes, complementing data from a spatially sparser network of continuously recording GNSS sites in the area, as well as temporally sparser Interferometric Synthetic Aperture Radar data that were able to capture a combined deformation signal from the two earthquakes.

The Global Positioning System (GPS) has revolutionized the ability to monitor Earth-system processes, including Earth’s water cycle. Several analysis centers process GPS data to estimate ground-antenna positions at daily temporal resolution. Differences in processing strategies can lead to inconsistencies in coordinate-position estimates and therefore influence the analysis of crustal displacement associated with variations in atmospheric and hydrologic mass loading. Here, we compare five GPS data products produced by three processing centers: the Nevada Geodetic Laboratory, Jet Propulsion Laboratory, and UNAVCO Consortium. We find that 5 to 30% of the scatter in residual GPS time series (commonly considered noise) can be explained by atmospheric loading in the contiguous USA and Alaska, but that the percentages vary widely by data product. Positions derived using high-resolution troposphere models (e.g., ECMWF) exhibit significantly lower scatter after correcting for atmospheric loading than positions estimated using constant or slowly varying troposphere models (e.g., GPT2w). The data products also exhibit differences in seasonal deformation (commonly attributed, in large part, to fluctuations in hydrologic mass loading): median vector differences in estimated seasonal amplitude range from 0.4–1.0 mm in the vertical component and 0.1–0.3 mm in the horizontal components, or about 10–40% of the mean amplitudes of seasonal oscillation. Newer products exhibit lower total scatter and stronger correlations than older products. Network-coherent differences in estimates of seasonal deformation reveal reference-frame inconsistencies between data products. We also cross-check two independent models of atmospheric pressure loading: ESMGFZ and LoadDef.

We present fast algorithms for estimating common parameters in geodetic time series based on statistical approaches to assess the impact of temporal correlations. One such assessment is based on the characteristics of the time series residuals averaged over different durations and with the statistical characteristics extrapolated with a first-order Gauss-Markov process to infinite averaging time. This approach circumvents a limitation of spectral methods, which cannot reliably account for the impact of temporal correlations over periods longer than the length of a given time series. The subsequent fast approach is the use of a Kalman filter with process noise values determined from the first-order Gauss-Markov characteristics to estimate all parameters. These methods are particularly useful for assessing long and numerous geodetic time series, which are nowadays ubiquitous, because they are much less computationally intensive than comprehensive methods, such as maximum likelihood estimators. Our approaches are compared to other commonly used programs, such as Hector, to understand the speed and impact of outliers on the algorithms, and to provide advice and suggestions on the uses of such algorithms in operational geodetic processing.

Interferometric synthetic aperture radar (InSAR) has become an essential technique to detect surface variations due to volcanoes, earthquakes, landslides, glaciers, and aquifers. However, Earth's ionosphere, atmosphere, vegetation, surface runoff, etc., introduce noise that requires post-processing to separate its components. This work defines a generator to create interferograms that include each of those components. Our approach leverages deformation models with real data, either directly or through machine learning using geostatistical methods. These methods result from previous developments to more efficiently and better simulate spatial variables and could replace some statistical approaches used in InSAR processing. We illustrate the use of the generator to simulate an artificial interferogram based on the 2015 Illapel earthquake and discuss the improved performance offered by geostatistical approaches compared with classical statistical ones. The generator establishes a tool for multiple applications (1) to evaluate InSAR correction workflows in controlled scenarios with known ground truth; (2) to develop training sets and generative methods for machine learning algorithms; and (3) to educate on InSAR and its principles.

We study the impact of estimating offsets within GNSS position time series on the uncertainties of site velocity estimates. Site velocity uncertainties are estimated by assuming different observational noise types, including white noise, fractal white noise, flicker noise, first-order Gauss-Markov noise, random walk, fractal random walk, and mixture of flicker noise, and white noise. For each noise type, we assess the sensitivity of the velocity uncertainty to offsets numerically by computing the inflation rate of the uncertainty as the number of offset parameters increases. We show that the sensitivity reduces when the noise becomes more temporally correlated. For the mixture of flicker noise plus white noise, which has been concluded as an appropriate noise model for the majority of current GNSS position time series, the rate uncertainty increases by about 11% as the number of offset parameters doubles from 1 to 2 provided that the white noise does not dominate the flicker noise. Based on the analysis of the Plate Boundary Observatory position time series, we find that adding one additional offset parameter would increase the median velocity standard deviation by 14%, consistent with the result of the simulation study. Such increase is much smaller than the conclusion of a 40% by a previous study. Therefore, given the noise characteristics of current GNSS site position estimates, position offsets have limited impact on uncertainties of site velocity estimates. This result implies that improving instrumentation, even though may introduce position offsets, is the best approach for improving reference frame stability.

Interferometric synthetic aperture radar (InSAR) measures surface deformation from repeated passes of satellites or aircraft and has become an important tool to study geophysical phenomena such as earthquakes. However, InSAR data analysis is challenging due to atmospheric water vapor that can mimic the effects of Earth deformation and thus lead to wrong interpretations. We present preliminary results on how to differentiate between tropospheric effects and surface deformation from earthquakes using a convolutional neural network. As earthquake training sets are sparse, our approach leverages transfer learning techniques for tropospheric patterns from areas where deformations are known to be mostly absent over short time periods, and classifies specific areas of interferograms to reflect regions that are dominated by deformation or tropospheric noise. The applicability of the training set to a new area may depend on the similarity of the two climates. Examples of tropospheric delays are shown from interferograms constructed from Sentinel-1 data over part of southern California with short temporal baselines, and examples of deformation are taken from interferograms generated using Okada models. Our classifier is tested on data from the 2018 Oaxaca earthquake in Mexico from Sentinel-1. This work is a step towards using neural networks for a fine-granular tile-based validation of interferograms and automatically removing unwanted effects from InSAR signals, as well as towards enhancing the agility of disaster response programs. The open source code is available in the PyInSAR package on GitHub under the MIT license. We acknowledge support from NASA AIST80NSSC17K0125 (PI Pankratius) and NSF ACI1442997 (PI Pankratius).

Thesis: S.M. in Geophysics, Massachusetts Institute of Technology, Department of Earth, Atmospheric, and Planetary Sciences, May, 2020

The response of a 413-m-tall building to the 12 November 2017 M-w 7.3 earthquake 642 km from the building is measured with a Global Positioning System (GPS) receiver located near the top of the building and operating with a 1 Hz sampling rate. Nearby GPS and seismic stations measure the ground motion near the building. The ground motions have amplitudes of similar to 40 mm, while the top of the building moves by up to 160 mm. The building motion continues with levels greater than the noise level of the GPS measurement for about 15 min after the earthquake. After the ground-motion excitation ends, the building motion decays with a time constant of similar to 2 min and the beat between the two lowest frequency modes of deformation of the building can be seen. There are two large amplitude peaks in the building motion with magnitudes of 120 and 160 mm. The timing of the peaks is consistent with ground excitation in an 8.3-6.5-s-period (120-180 mHz) band, which covers the 7.25 and 5.81 s periods (138 and 172 mHz frequencies) of the fundamental modes of the building. The ground motions in this band show two large pulses of the excitation, which have timing consistent with the large amplitude building signals. The response of the top of the building is amplified by an order magnitude over the ground motions in this band. There is no apparent permanent displacement of the top of the tower.