Mapping and modelling the spatial variation in strain accumulation along the North Anatolian Fault

Since 1900, earthquakes worldwide have been responsible for over 2 million fatalities and caused nearly $2 trillion of economic damage. Accurate assessment of earthquake hazard is therefore critical for nations in seismically active regions. For a complete understanding of seismic hazard, the temporal pattern of strain accumulation, which will eventually be released in earthquakes, needs to be understood. But earthquakes typically occur every few hundred to few thousand years on any individual fault, and our observations of deformation usually only cover time periods of a decade or less. For this reason, our knowledge of the temporal variation in strain accumulation rate is limited to insights gleaned from kinematic models of the earthquake cycle that use measurements of present-day strain to infer the behaviour on long time scales. Previous studies have attempted to address this issue by combining data from multiple faults with geological estimates of long-term strain rates. In this thesis I propose a different approach, which is to observe deformation at multiple stages of the earthquake cycle for a single fault with segments that that have failed at different times. In the last century the North Anatolian Fault (NAF) in Turkey has accommodated 12 large earthquakes (Mw >6.5) with a dominant westward progression in seismicity. If we assume that each of these fault segments are at a different stage of the earthquake cycle then this provides a unique opportunity to study the variation in along-strike surface deformation, which can be equated to variation of deformation in time. In this thesis I use Interferometric Synthetic Aperture Radar (InSAR) and Global Navigation Satellite System (GNSS) observations to examine the spatial distribution of strain along the NAF. InSAR is an attractive technique to study surface displacements at a much higher spatial resolution (providing a measurement every 30 m) compared to established GNSS measurements, with station separations between 10 km to 100 km in Turkey. I specifically address a key technical challenge that limits the wide uptake of InSAR: phase unwrapping, the process of recovering continuous phase values from phase data that are measured modulo 2π radians. I develop a new unwrapping procedure for small baseline InSAR measurements that iteratively unwraps InSAR phase. For each iteration, this method identifies pixels unwrapped correctly in the previous iteration and applies a high cost to changing the phase difference between these pixels in the next iteration. In this way, the iterative unwrapping method uses the error-free pixels as a guide to unwrap the regions that contained unwrapping errors in previous iterations. I combine measurements of InSAR line-of-sight displacements with published GNSS velocities to show that an ∼80 km section of the NAF that ruptured in the 1999 Izmit earthquake (Mw 7.4) is creeping at a steady rate of ∼5 mm/yr with a maximum rate of 11 ± 2 mm/yr near the city of Izmit within the observation period 2002-2010. I show that in terms of the moment budget and seismic hazard the effect of the shallow, aseismic slip in the past decade is small compared to that from plate loading. Projecting the shallow creep displacement rates late into the earthquake cycle does not produce enough slip to account for the 2-3 m shallow coseismic slip deficit observed in the Izmit earthquake. Therefore, distributed inelastic deformation in the uppermost few kilometers of the crust or slip transients during the interseismic period are likely to be important mechanisms for generating the shallow slip deficit. I used similar techniques to confirm that a ∼130 km section of the central NAF near the town of Ismetpasa, is also undergoing aseismic creep at a steady rate of 8±2 mm/yr. Using simple elastic dislocation models to fit fault perpendicular velocities I show that there is an eastward decreasing fault slip rate in this region from ∼32 mm/yr to ∼21 mm/yr over a distance of about 200 km. The cause of this decrease remains unclear, but it could be due to postseismic effects from the 1999 Izmit and Duzce earthquakes and/or long-term influence from the 1943 (Mw 7.4) and 1944 (Mw 7.5) earthquakes. Finally, I combine line-of-sight displacements from 23 InSAR tracks to produce the first high resolution horizontal velocity field for the entire continental expression of the NAF (∼1000 km). I show that the strain rate does not vary significantly along the fault, and since each segment of the NAF is at a different stage of the earthquake cycle, the strain rate is invariant with respect to the time since the last earthquake. This observation is inconsistent with viscoelastic coupling models of the earthquake cycle, which predict a decreasing strain rate with time after an earthquake. My observations imply that strain accumulation reaches a steady-state fairly rapidly after an earthquake (<7-10 years) after which strain is localised on a narrow shear zone centred on the fault and does not vary with time. A time-invariant strain rate is consistent with a strong lower crust in the region away from the fault with a viscosity ≥1020 Pas. My results imply that short term snapshots of the present-day strain accumulation (as long as it is after the postseismic period) are representative of the entire earthquake cycle, and therefore geodetic estimates of the strain rate can be used to estimate the total strain accumulation since the last earthquake on a fault, and be used as a proxy for future seismic hazard assessment. The techniques I developed to explore the spatial and temporal pattern of aseismic fault creep and long-term strain accumulation along the NAF are general and can be ap- plied to all strike-slip faults globally. The archived ERS-1/2 and Envisat satellite data are an extremely valuable resource that can and should be used to extend InSAR time series measurements back to the early 1990s. Together with the new Sentinel-1 data sets, this provides an unprecedented opportunity to explore tectonic deformation over several decades and on continental scales. Despite the availability of numerous correction techniques (in this thesis I use global weather models to calculate the atmospheric contribution), atmospheric delays remain the major challenge to exploiting Sentinel-1 data for global strain mapping, the mitigation of these delays are an important goal for the InSAR community.