Research Interest

  • Volcanic fast response systems
  • Space- and air-borne radar remote sensing
  • Multitemporal InSAR methods
  • Digital signal processing
  • Inverse theory
  • Numerical and analytical modeling of seismic and aseismic faulting processes,
  • Volcanic source and oil and gas reservoir modeling
  • Hydrological processes
  • Planetary science

Active Projects

  • Multitrack multitemporal InSAR for measuring high-resolution 3D displacement field.
  • 3D phase unwrapping, aplicable to multi-/single-master interferometry
  • Time-dependent model of aseismic slip on the central creeping section of the San Andreas from joint inversion of InSAR time series and repeating earthquakes  
  • Earthquake potential of Mosha fault, northern Iran: evidence from joint inversion of InSAR, GPS and seismic data
  • Time-dependent model of Kilauea volcanic system obtained from InSAR time series
  • Investigating the land subsidence and liquefaction in City of Christchurch using InSAR time series
  • Investigating time dependent seimic hazard associated with fluid injection, Timpson Texas

Completed Projects  

Multitemporal InSAR

1) Advanced multitemporal algorithm with an  improved filtering scheme that combines and inverts a large set of unwrapped interferograms to generate an accurate time series of the surface motion. During various stages of the analysis, this approach applies a variety of sophisticated wavelet based filters to estimate the interferometric phase noise and to reduce the effects of systematic and random artefacts, such as spatially correlated and temporal uncorrelated components of the atmospheric delay, and the digital elevation model and orbital errors. (read more)

2) The Synthetic Aperture Radar (SAR) satellites revisit the overlap zones of their adjacent tracks about twice as frequently than elsewhere. Thus being able to combine the data sets acquired over the overlapped zones, can significantly improve the temporal resolution of the surface deformation time series. However, due to datum and geometric differences between adjacent tracks and environmental artifacts, the task of combining these data sets is not trivial. Here, we present a novel approach that enables us to combine the overlapped segments of SAR images acquired at adjacent tracks and generate an accurate and high spatiotemporal resolution map of the surface deformation field. To this end, new physics-based approaches are developed to unify the datums. The error due to the look angle difference is estimated and removed using a Kalman filter. The error associated with the atmospheric delay is reduced by applying wavelet based filters. The presented approach is tested at Hawaii Island, where tracks 200 and 429 of Envisat satellite overlap over the Kilauea south flank. While initial data sets include 54 and 46 images, respectively, the time series obtained through the method presented here is sampled at 100 time steps. The comparison against the GPS time series shows that the obtained high resolution time series accurately measured the nonlinear displacement field. The dense data series generated using this method enables application of statistical tools, signal processing approaches and time-dependent modeling schemes to the InSAR time series. (read more)

3) Atmospheric delay is one of the major sources of error in repeat pass interferometry. We develop new approaches for correcting the topography-correlated components of this artifact. To this aim we use multiresolution wavelet analysis to identify the components of the unwrapped interferogram that correlate with topography. By using a forward wavelet transform we break down the digital elevation model and the unwrapped interferogram into their building blocks based on their frequency properties. We apply a cross-correlation analysis to identify correlated coefficients that represent the effect of the atmospheric delay. (read more)

4) Interferometric synthetic aperture radar (InSAR) data are often obtained on the basis of repeated satellite acquisitions. Errors in the satellite orbit determination, however, propagate to the data analysis and may even entirely obscure the interpretation. Many approaches have been developed to correct the effect of orbital error, which sometimes may even distort the signal. Phase contributions due to other sources, such as surface deformation, atmospheric delay, DEM error and noise, may reduce the accuracy of the orbital error estimation. Therefore, a more sophisticated approach for estimating the effect of orbital errors is required. In this work, wavelet multi-resolution analysis is employed to distinguish between the effects of orbital errors and other components (e.g., deformation signal).  Next, a robust regression approach is applied to estimate the effect of orbit errors as a ramp. (read more)

Inverse Theory

1) Modern geodetic techniques provide valuable and near real time observations of volcanic activity. Characterizing the source of deformation based on these observations has become of major importance in related monitoring efforts. We investigate random search approaches, such as Simulated Annealing (SA) and Genetic Algorithm (GA), and utilize them in an iterated manner. The iterated approach helps to prevent GA in general and SA in particular from getting trapped in local minima, and it also increases redundancy for exploring the search space. We apply a statistical competency test for estimating the confidence interval of the inversion source parameters, considering their internal interaction through model, the effect of the model deficiency, as well as the observational error. (read more)

2) Modern geodetic methods allow continuous monitoring of deformation fields at volcanoes. The acquired data contributes significantly to the study of the dynamics of magmatic sources prior to, during and after eruptions and intrusions. In addition to advancing monitoring techniques, it is important to develop suitable approaches to deal with deformation time series.Here, we present, test and apply a new approach for time-dependent, nonlinear inversion using a combination of a Genetic Algorithm (GA) and Kalman Filter (KF). The GA is used in the form presented by Shirzaei and Walter, [2009] and KF implementation now allows for the treatment of monitoring data as a full time series, rather than as single time steps. This approach provides a flexible tool for assessing unevenly sampled and heterogeneous time series data and explains the deformation field using time-consistent dislocation sources. (read more)

Volcanic Deformation and Source Modeling

1) The detection and monitoring of gravity-driven volcano deformation is vital for understanding volcanic hazards, such as landslides, lateral blasts and debris avalanches. Although deformation has been detected at several large active volcanoes (e.g., Mt. Etna, Vesuvius, Kilauea), these systems also exhibit persistent magmatic activity, obscuring the gravity-driven signals of ground motion. In this study, we present a first InSAR deformation time series at the dormant Damavand volcano in northern Iran, over the period of 2003 through 2008. The high resolution data show a lateral extension of the volcano at the relative rate of up to ~6 mm/yr accompanied by a subsidence at the rate of up to ~5 mm/yr at the volcano summit. (read more)

2) We apply a time-dependent source modeling for InSAR data available between 1992 and 2008 from the Campi Flegrei volcano in Italy. We obtained multiple episodes of linear velocity for reservoir pressure change associated with parabolic surface deformation at the volcano. This may be interpreted via differential equations as a linear flux to the shallow reservoir and provides new insight into how both the shallow and deep reservoirs communicate beneath Campi Flegrei. (read more)

3) The coupling of Mauna Loa and Kilauea volcanoes, Hawaii, has been debated for the past 100 years. The distinct composition of erupted materials at both volcanoes suggests that they draw on distinct magma reservoir in the mantle. In contrast, statistical analysis of the pattern of historic eruptions implies that Mauna Loa and Kilauea compete for magma supply.  Resolving this discrepancy, we present high-resolution spatiotemporal interferometric deformation maps using a well-populated catalogue of space-borne synthetic aperture radar data over Hawaii Island during 2003-2008. (read more)

4) The ongoing eruption of the Lusi mud volcano in East Java, Indonesia offers the unprecedented opportunity to study a large eruption from its beginning to its eventual end. We use new observations of ground deformation obtained from multitemporal interferometric analysis of L-band synthetic aperture radar data to show that Lusi will stop erupting much sooner than previously anticipated. Using principal component analysis, we find that the rate of ground deformation, and by implication, pressure in the mud source region, has been decaying exponentially with an e-folding time scale of 2.1±0.4 years. We anticipate that discharge will decrease to 10% of the present rate in 5 years. (read more)

5) The Lusi mud eruption, Indonesia, began in May 2006 and continues to the present. Previous analyses of surface deformation data suggested an exponential decay of the pressure in the mud source, but did not constrain the location, geometry and evolution of the possible source(s) of the erupting mud and fluids. To map the surface deformation, we employ multitemporal InSAR and analyze a well-populated data set acquired by ALOS L-band satellite between May 2006 and April 2011. We then apply a time-dependent inverse modeling scheme. Volume changes occur in two regions beneath Lusi, at 0.3-2.0 km and 3.5-4.75 km depth. The cumulative volume change within the shallow source is ~2-3 times larger than that of the deep source. The observation and model suggest that a shallow source plays a key role by supplying the erupting mud, but that additional fluids do ascend from depths >4 km on eruptive timescales. (read more)
6) Volcanoes are often considered as isolated systems, however, evidences increase that adjacent volcanoes are directly coupled or may be closely related to remote triggers. At the dormant but dangerous Italian volcanoes Campi Flegrei and Vesuvius, as well as adjacent  volcano-tectonic systems, all located in the Gulf of Naples with ~2 Mio inhabitants, a new analysis of satellite radar data reveals allied deformation activity. Here we show that during the 16-year records from 1992-2008, identified episodes of deformation occur that are in correlation. Albeit differences in the quantity of deformation, the sign, frequency and rate of pressure changes at reservoirs beneath Campi Flegrei and Vesuvius can be very similar, allowing to infer that pressure changes originating from a magmatic or tectonic source external to the shallow volcano magma plumbing systems is a likely cause. Such a fluidmechanical coupling sheds light on earlier episodes of correlated eruptions and deformations occurring during the historical roman times. (read more)

Aseismic Faulting Processes

1) Spatial and temporal variations of aseismic fault creep represent important factors in realistic estimation of seismic hazard due to their influence on the size and recurrence interval of large earthquakes along partially coupled faults. To solve for a time-dependent model of creep on the Hayward fault, we invert 18 years of surface deformation data (1992 - 2010), obtained by interferometric processing of 52 and 50 synthetic aperture radar (SAR) images acquired by the ERS1/2 and ENVISAT satellites, respectively, and surface creep data obtained at more than 25 alinement and creepmeter stations. (read more)

2) We show a rare example of aseismic response of a creeping fault to the earthquake cycle of a nearby megathrust. Interferometric synthetic aperture radar (InSAR) is used to detect and analyze shallow creep of two crustal faults at Mejilones Peninsula, Northern Chile, located in the hanging wall of the 2007 Mw7.7 Tocopilla subduction earthquake. (read more)

3) The Hilina Fault System (HFS) is located on the south flank of Kilauea volcano and is thought to represent the surface expression of an unstable edifice sector that is active during seismic events such as the 1975 Kalapana earthquake. Despite its potential for hazardous landsliding and associated tsunamis, no fault activity has yet been detected by means of modern geodetic methods since the 1975 earthquake. Using wavelet transforms in a statistical framework, we jointly analyze InSAR and continuous GPS deformation data from 2003 to 2010 to resolve a subtle deformation signal about the HFS normal fault scarps. (read more)

4) Availability of dense continuous GPS and seismic monitoring networks provides a unique opportunity to study a variety of time-dependent processes associated with the 11 March 2011 Tohoku earthquake (Mw 9.0), such as afterslip deformation and postseismic relaxation. To this end we establish a time-dependent inversion scheme as a combination of L1-Norm minimization and Kalman filter. This framework allows inverting the time series of the surface deformation obtained from GPS networks constrained with direct observations of the fault slip obtained from repeating earthquakes. (read more)

5) Rising and falling slip rates on the creeping section of the San Andreas Fault have been inferred from variations of recurrence intervals of characteristically repeating micro-earthquakes, but this observation has not previously been confirmed using modern geodetic data. Here, we report on observations of this ‘pulsing’ slip obtained from advanced multi-temporal Interferometric Synthetic Aperture Radar (InSAR) data, confirmed using continuous GPS sites of the Plate Boundary Observatory. The surface deformation time series show a strong correlation to the previously documented slip rate variations derived from repeating earthquakes on the fault interface, at various spatial and temporal scales. Time series and spectral analyses of repeating earthquake and InSAR data reveal a quasi-periodic pulsing with a roughly 2-year period along some sections of the fault, with the earthquakes on the fault interface lagging behind the far-field deformation by about six months. This suggests a temporal delay between the pulsing crustal strain generated by deep-seated shear and the time-variable slip on the shallow fault interface, and that at least in some places this process may be cyclical. There also exist potential impacts for time-dependent seismic hazard forecasting in California and, as it becomes better validated in the richly-instrumented natural laboratory of the central San Andreas Fault, the process used here will be even more helpful in characterizing hazard and fault zone rheology in areas without California’s geodetic infrastructure. (read more)
6) The Central segment of San Andreas Fault (CSAF) is characterized by a nearly continuous right-lateral aseismic slip. However, observations of the creep rate obtained using small Characteristically Repeating Earthquakes (CREs) show pulses of creep along the CSAF, which may indicate spatially and temporally variable seismic hazard along the CSAF. Therefore, the goal of this study is to obtain a high resolution time-dependent model of creep along the CSAF to examine this hypothesis. To this end, we apply a time-dependent creep modeling approach, which combines InSAR surface deformation time series and observations of fault creep obtained from CREs. The SAR dataset includes C-band scenes acquired by the ERS-2 and Envisat satellites between 2003 and 2011. The resulting creep rate distribution implies a peak rate up to 32 mm/yr along the central part of the CSAF. Afterslip due to the 2004 Parkfield earthquake on the southeastern segment of the CSAF is also manifest in the model and there is clear evidence of creep pulsing along the strike and depth of the CSAF. The estimated annual rate of slip deficit accumulation is equivalent to a magnitude 5.6-5.7 earthquake. Taking advantage of the time-dependence of our model, we also refine the scaling relationship, which associates the released seismic moment due to a CRE event with the amount of creep on the fault, surrounding the CRE patches. This study provides the first kinematic model of creep pulsing, constrained using geodetic and seismic data, which can enhance time-dependent seismic hazard maps and improve earthquake operational forecast models.

Time-dependent Hazard Assessment

1) The probability of large seismic events on a particular fault segment may vary due to external stress changes imparted by nearby deformation events, including other earthquakes and aseismic processes, such as fault creep and postseismic relaxation. The Hayward fault (HF), undergoing both seismic and aseismic fault slip, provides a unique opportunity to study the mutual relation of seismic and aseismic processes on a fault system. We use surface deformation data obtained from InSAR (interferometric synthetic aperture radar), creepmeters and alinement arrays, together with constraints provided by repeating earthquakes to investigate the kinematics of fault creep on the northern HF and its relation to two seismic clusters in October 2011 and March 2012, and an Mw4.2 event in July 2007. Recurrences of nearby repeating earthquakes show that these episodes involved both seismic and aseismic slip. We model the stress changes due to fault creep and the recent seismic activity on the locked central asperity of the HF, which is believed to be the rupture zone of past and future M~7 earthquakes. We estimate that the 1-day probability of a large event on the HF only increased by up to 0.18% and 0.05% due to the static stress increase and stressing rate change by the 2011 and 2012 clusters. For the July 2007 south Oakland event (Mw4.2) the estimated increase of short-term probabilities is 50%, highlighting the importance of short-term probability changes due to transient stress changes. (read more)

Hydrological Processes

1) The effects of land subsidence pose a significant hazard to the environment and infrastructure in the arid, alluvial basins of Phoenix, Arizona. Improving our understanding of the source and mechanisms of subsidence is important for planning and risk management. Here, we employ multitemporal interferometric analysis of large SAR datasets acquired by ERS and Envisat satellites to investigate ground deformation. The ERS datasets from 1992-1996 and Envisat, 2003-2010, are used to generate LOS time series and velocities in both the ascending and descending tracks. The general deformation pattern is consistent among datasets and is characterized by three zones of subsidence and a broad zone of uplift. The multi-track Envisat LOS time series of surface deformation are inverted to obtain spatiotemporal maps of the vertical and horizontal deformation fields. We use observation wells to provide an in situ, independent dataset of hydraulic head levels. Then we analyze vertical InSAR and hydraulic head level time series using continuous wavelet transform to separate periodic signal components and the long-term trend. The isolated signal components are used to estimate the elastic storage coefficient, the inelastic skeletal storage coefficient, and compaction time constants. Together these parameters describe the storage response of an aquifer system to changes in hydraulic head and surface elevation. Understanding aquifer parameters is useful for the ongoing management of groundwater resources. (read more)

Induced Seismicity

1) US sates, such as Texas and Oklahoma, producing high-volume of unconventional oil and gas, are facing sharp increase in seismicity. The observation of the surface deformation and physical models to unequivocally link the seismicity and waste water injection are scarce. Here, we find that the waste water injection in eastern Texas causes uplift, detectable using radar interferometric data. Combining the uplift and injection data through a poroelastic model allows resolving complex distribution of hydraulic conductivity and pore pressure in the crust. We find that the ~5-years pore pressure increase is capable of triggering the 17 May 2012, Mw 4.8 earthquake, the largest event recorded instrumentally in east Texas. This study shows that surface deformation data are vital in order to constrain the spatiotemporal variations of the stress field in the vicinity of injection sites. (read more)