Ph.D. research


GPS inferred velocity and strain rate fields in eastern Canada

Abstract (proposed)

The Saint Lawrence River valley (SLRV) in eastern Canada is one of the most seismically active regions in eastern North America, which is characterized by many intraplate earthquakes. After its rigid plate rotation, the ongoing glacial isostatic adjustment (GIA) is by far the largest source of geophysical signal in eastern Canada. In this research, the current crustal deformation and strain rate field of this region was studied using more than 14 years (9 years on average) observations of 112 continuously operating GPS stations.
The velocity field was obtained from cleaned position time series of daily GPS solutions by applying a combined model using the weighted least-squares method. We estimated velocity uncertainties by assuming advanced noise models to include the temporal correlation of the position time series. The horizontal velocity field shows the counter-clockwise rotation of the North American plate in the no-net-rotation model with the average of 16.8±0.7 mm/yr constrained to ITRF 2008. The vertical velocity field confirms the GIA-induced uplift all over eastern Canada with the maximum rate of 13.7±1.2 mm/yr and subsidence to the south mainly over north of the United States with a typical rate of −1 to −2 mm/yr and the minimum value of −2.7±1.4 mm/yr.
The noise behavior of the GPS position time series was explored by testing five different noise models of power-law, white, white plus flicker, white plus random-walk, and white plus flicker plus random-walk, using the spectral analysis and the maximum likelihood methods. The results show that combination of white plus flicker noise is the best model for describing the stochastic part of the position time series. Furthermore, amplitudes of all noise models are smallest in the north direction and largest in the vertical direction. While amplitudes of the white noise model are almost equal across the study area, they are prevailed by the flicker and the random-walk noise for all directions. Assuming flicker noise model increases uncertainties of the estimated velocities by a factor of 5–38 compared to the white noise model, while the estimated velocities from all noise models are statistically consistent.
The estimated Euler pole parameters for this region are slightly but significantly different from the overall rotation of the North American plate. This difference potentially reflects local stress in this seismic region, and the difference in intraplate velocities between the two sides of the SLRV accumulates stress in the faults located along the river.
The surface deformation of the region was studied using least-squares collocation. Interpolated intraplate horizontal velocities show a spatially coherent radially outward motion from the centers of maximum uplift to the north and inward motion to the centers of maximum subsidence to the south with a typical velocity of ~1–1.6±0.4 mm/yr. This pattern, however, becomes more complex near the margins of the formerly glaciated areas. Based on their directions, the intraplate horizontal velocities can be divided in three distinct zones. This confirms the existence of three ice domes in the study region before the last glacial maximum. A spatial correlation is observed between areas with higher magnitude of the intraplate horizontal velocity and the seismic zones along the SLRV. The vertical velocities were interpolated to model the ongoing vertical deformation. The model shows maximum uplift rate of 15.6 mm/yr to southeastern of Hudson Bay and a typical subsidence rate of 1–2 mm/yr to the south mainly across the north of the United States. Along the SLRV, horizontal and vertical motions are spatially coherent toward southeast with the typical magnitude of ~1.3 mm/yr relative to North American plate and the average uplift rate of 3.1 mm/yr, respectively. In general, the rate of vertical deformation is typically ~2.4 times larger than the rate of the intraplate horizontal motion in this area.
Results of strain analysis show the present-day straining of eastern Canada in the form of extension to the north (the area under uplift) and shortening to the south (the area under subsidence). On average, rotational rates are at the level of 0.011°/Myr. A NNW-SSE shortening with a typical rate of ~3.6–8.1 nstrain/yr is observed over the Lower Saint Lawrence seismic zone. In the Charlevoix seismic zone, an extension with a typical rate of ~3.0–7.1 nstrain/yr is oriented about ENE-WSW. In the western Quebec seismic zone, the deformation has a shear straining mechanism with a typical shortening rate of ~1.0–5.1 nstrain/yr and extension rate of ~1.6–4.1 nstrain/yr. These results are consistent, to the first order, with GIA models and with the maximum horizontal compressional stress of the World Stress Map resulted from focal mechanism method.

The selected GPS permanent sites in the eastern and center of the North American plate

M.Sc. research


Detection and measurement of land deformations caused by seismic events using InSAR, Sub-pixel correlation, and Inversion techniques


On Monday April 6th, 2009, the city of L'Aquila in the centre of Italy, and the surrounding area was shocked by an earthquake of MW 6.3. The main shock inflicted serious damage to the city and other surrounding towns and villages, and killed several hundred people.
Deformation of the ground surface caused by this earthquake, and source parameters of the causative fault were studied in my M.Sc. research using three different techniques: Synthetic Aperture Radar Interferometry (InSAR), Sub-pixel Correlation Technique (SCT) and Geophysical Inversion problem, with different image data sets of optical and Radar sensors. None of these techniques can offer a complete toolbox to study a seismic event from different points of view, and hence, these three techniques were combined and used together. Results from both optical and radar sensors were compared in each section. The results were also compared with results from other similar studies.
InSAR was applied to ENVISAT ASAR and ALOS PALSAR image data sets, in ascending and descending passes, to reveal deformation along the satellite line of sight direction. The Enhanced Lee filter was applied to ALOS PALSAR interferograms to remove temporal noises. The earthquake affected an ellipse-shaped area of 25 km long and 15 km wide, roughly equal to 300 km2. The initial coseismic surface deformation started from the north of the main event epicentre. Most of the post-seismic relaxation occurred after April 7th, toward the south and south-east of the main shock epicentre and became more homogeneous around the main event. While the results from both the ENVISAT and the ALOS data sets were strongly similar and supportive, the former data set showed more subsidence in the satellite line of sight direction. Maximum 28.1 cm subsidence along the satellite line of sight was observed on the descending interferogram from the ENVISAT ASAR data set, to the north-west of Onna village.
SCT was applied to both the ASTER and the ENVISAT ASAR data sets to discover horizontal displacements of the affected area. Even though the exact amount and pattern of horizontal displacement could not be determined using these data sets, it was shown that the horizontal displacement was not significant (not more than a few centimetres). This result was confirmed by GPS observations. It was also observed that both the optical and SAR data sets were highly sensitive to temporal variations. However, the SAR data set showed higher coherency values.

A geophysical Inversion problem was applied using a two-step procedure (neighbourhood algorithm and linear least square) to two deformation models obtained from InSAR, to find the source parameters of the causative fault and the dislocation model at depth. The neighbourhood algorithm showed almost a pure normal faulting system, with approximately 11.5 km length and 8.0 km depth, strikes 144° from north-west to south-east and dips 54° toward south-west. The linear least square inversion showed a maximum 1.2 m dip slip on the fault plane and predicted a fracture of 2~3 km long, on the ground surface.

Three dimensional dislocation model at depth obtained from the ascending ENVISAT ASAR pair.