GLACIAL ISOSTATIC ADJUSTMENT
5.1 Introduction
C h a p t e r 5
GEODYNAMIC CONTROLS ON INTRAPLATE SEISMICITY IN
to loads viscoelastically (Cathles, 1975; Peltier, 1974), meaning that even after the ice load is gone, there is a time delay of the uplift of the Earth to its pre-glaciation level. GIA continues in the present day, with GPS observable maximum uplift rates of >10 mm/yr (Sella et al., 2007; Lidberg et al., 2010) near the former centers of the Laurentide ice-sheet in North America and the Fennoscandian ice-sheet in northern Europe. During the last glacial maximum, the Laurentide ice sheet covered most of North America, with the thickest ice cover (as much as 3-5 km) positioned at Hudson Bay (Peltier et al., 2015; Lambeck et al., 2017), the site of the largest modern day rebound rates. The southern extent of the former ice-margin during the LGM reached as far south as southern Illinois and as far east as offshore Maine and Nova Scotia (Dyke et al., 2002; Peltier et al., 2015; Dalton et al., 2020). The ice-sheet melted rapidly between 10-12 ka, and deglaciation was nearly complete byβΌ7-8 ka (Dalton et al., 2020).
During glaciation, the weight of the ice sheet creates an additional vertical stress in the lithosphere, and the flexure of the lithosphere under the load generates horizontal bending stresses. These additional vertical and horizontal stresses increase all three principal stresses. In terms of Mohr Coulomb failure theory (Appendix C.3) for a compressive background stress regime, this stress increase pushes the Mohr circle in the positive direction along the normal stress axis, away from the failure threshold.
In other words, during glaciation, the weight of the ice-sheet acts to stabilize faults and suppress fault movement (Johnston, 1987; Wu and Hasegawa, 1996; Steffen et al., 2014a). When the ice-sheet melts, the vertical stress from the load decreases, but the GIA induced horizontal stresses do not decrease as quickly because of the viscoelastic nature of the lithosphere and the upward migration of stress from mantle relaxation. Thus, after deglaciation, the vertical stress disappears but the horizontal stress remains, which increases the differential stress, both expanding the Mohr circle radius and shifting it back in the negative direction along the normal stress axis, bringing it closer to the failure criterion (Steffen et al., 2014a).
The magnitudes of glacial rebound stresses are also dependent on the wavelength of the load relative to the thickness of the lithosphere and are largest for ice sheets with diameters on the order of a couple hundred km rather than continental-scale ice-sheets like the Laurentide (Johnston et al., 1998). However, while the Laurentide itself was of large spatial extent, lobes extending from its southern margin, such as those that covered New York and Massachusetts or extended south of the Great Lakes (Dalton et al., 2020), were of much smaller wavelength and thus may have been more
effective at perturbing stress magnitudes in areas of interest regarding the patterns of intraplate seismicity. Nevertheless, the largest confirmed glacially induced fault offset, about 100m, is located in the Canadian Arctic at the Laurentian margin (Dyke et al., 1991). Most glacially reactivated faults, however, are located in the northern Lapland Province of Sweden, Finland, and Norway at the ancient margins of the much smaller Fennoscandian ice sheet (Steffen et al., 2021), consistent with the findings of Johnston et al. (1998). As with many of the faults in eastern North America, especially within the Canadian seismic zones, glacially triggered faults are often steeply dipping normal faults reactivated with a thrust sense of motion, but reactivation under GIA stress perturbations is likely only possible for a coefficient of friction less than 0.4 (Steffen et al., 2014a).
Using a spherical, self-gravitating viscoelastic Earth model, Wu and Johnston (2000) found that GIA is capable of triggering paleo-earthquakes within the Charlevoix Seismic Zone and even in the Wabash Valley, Indiana north of the NMSZ. Within the NMSZ, they find that faults do pass the instability threshold within the last 200 years, consistent with the timing of the 1811-1812 New Madrid earthquakes.
However, the magnitude by which the stresses exceed that threshold are very small and are therefore believed unlikely to have triggered the M 7-8 earthquakes of that sequence. This is due to the fact that GIA induced stresses decay rapidly away from the former ice margin (Wu and Johnston, 2000) but also with time after initial deglaciation, as faults are likely to be most unstable immediately following the removal of the load (Steffen et al., 2020; Wu and Hasegawa, 1996), around 8-9 ka. Nevertheless, paleoseismicity in the NMSZ records a Holocene slip rate four orders of magnitude greater than that in the Cretaceous or earlier in the Cenozoic and a slip rate as high as 4.4 mm/yr over the last 2400 years alone, which may be tied to crustal motion from Laurentide deglaciation (Van Arsdale, 2000). Likewise, paleoseismological and geodetic evidence suggestive of recurrence times anywhere between 400 β 1100 years in combination with the lack of accumulated deformation on faults suggests places like the NMSZ became active recently and that the seismic zone itself is only a couple tens of thousands of years old (Schweig and Ellis, 1994), suggesting New Madrid seismicity could be a glacially controlled phenomenon.
However, paleoseismic evidence suggestive of an elevated slip rate on the Reelfoot Fault during the last two major earthquake cycles (Van Arsdale, 2000) is also consistent with idea that the New Madrid sequence was a rare one-off event, and that modern seismicity may be a transient clustering of mainshock earthquake
activity in response to viscoelastic relaxation after those major earthquakes (Kenner and Segall, 2000; Kenner and Simons, 2005). Such a prolonged viscoelastic decay of seismicity rate beyond what is expected for a typical aftershock sequence is not unheard of for large earthquakes, such as, for example, the elevated seismicity rates observed after the 2010 El-Mayor Cucapah event in Mexico (Gualandi et al.
(2020), e.g., Chapter 3), and long aftershock sequences are more typical of intraplate environments (Stein and Liu, 2009). Nevertheless, such transient seismicity would likely require relaxation of a local lithospheric weak zone that cyclically transfers coseismic stress to the upper crust and triggers slip on faults until the weak zone fully relaxes (Kenner and Segall, 2000). Thus, the presence of a pre-existing weak zone remains a key ingredient to the generation of intraplate earthquakes. This hypothesis begs the question, however, of what loaded the weak zone in the first place. One may consider a mechanism whereby external forcing by GIA gradually loads the weak zone since deglaciation, eventually leading to a large earthquake later on that then continues to load the crust locally due to cyclic stress transfer (Brandes et al., 2015). Such a delay in seismicity is also possible considering that intraplate seismic zones tend to stay in a stress shadow for 100s-1000s of years after large events, owing to the difficulty of full stress restoration due to the strength of the ambient crust and low regional strain rates (Stein, 2007; Stein and Liu, 2009).
Lithospheric weak zones in combination with glacial unloading have also been pro- posed to explain both the magnitude and orientation of perturbations to the observed stress field. Focal mechanism stress inversion shows some aulacogens in eastern North America exhibit rotational deviation from the regional NE-SW maximum horizontal compressive stress (ππ» π ππ₯) direction (Hurd and Zoback, 2012; Mazzotti and Townend, 2010). In the Central Virginia, Lower St. Lawrence, and Charlevoix Seismic Zones, there is as much as a 30-50Β° statistically significant clockwise ro- tation of the seismically derived ππ» π ππ₯ direction relative to the regional borehole derivedππ» π ππ₯ direction (Mazzotti and Townend, 2010), and depth-dependent stress rotations of up to 40-60Β°are observed in the Charlevoix seismic zone (Verdecchia et al., 2022). It has been argued that the consistency of these rotations across seis- mic zones separated by 1000s of kilometers requires a common mechanism, which suggests a long-wavelength source of loading like GIA. GIA stresses show a strong spatial coherence over 1000s of kilometers but are typically only on the order of 10s of MPa (Wu and Hasegawa, 1996; Wu et al., 2021). The observed stress rotations, on the other hand, require perturbations at mid-seismogenic depths on the order of at least 160-250 MPa (Mazzotti and Townend, 2010). However, the presence of a
low viscosity lithospheric weak zone may amplify GIA stress magnitudes by a factor of 5-10 compared to homogeneous lithosphere (Grollimund and Zoback, 2001; Wu and Mazzotti, 2007) and could produce clockwise rotations of post-glacial rebound stresses in the crust above the weak zone (Wu and Mazzotti, 2007). Inclusion of such weak zones has also yielded better fits to both modeled stress and strain data in the St. Lawrence River Valley (Mazzotti et al., 2005).
We explore the hypothesis that glacial isostatic adjustment promotes intraplate seis- micity in eastern North America via perturbation to the intraplate stress field and reactivation of pre-existing faults. We develop high resolution global models of the solid earth response to glacial loading and unloading with CitcomSVE (Zhong et al., 2022), a spherical finite-element viscoelastic GIA code that implements the sea level equation and ice-loading history of ICE-6G (Peltier et al., 2015) for a fully 3D or 1D Earth viscosity structure. The models use a viscosity structure based on the seismically and geologically constrained thermal structure implemented in the mantle flow models with CitcomS (see Chapter 4). Like in the CitcomS models, we include local-scale, low-viscosity lithospheric weak zones at the locations of the geologically mapped aulacogens (Whitmeyer and Karlstrom, 2007) and other tectonically inherited structures. We calculate the stress tensor, with which we com- pute the ππ» π ππ₯ direction, the deviatoric stress magnitude, and the Coulomb stress on known faults. We compare our results for the present day to stresses of the World Stress Map (Heidbach et al., 2018) and to those obtained from mantle flow models using the same Earth structure (Chapter 4).