Geophysical Research Letters
  • Open Access

Thermal structure and megathrust seismogenic potential of the Makran subduction zone


Corresponding author: G. L. Smith, Ocean and Earth Science, National Oceanography Centre Southampton, University of Southampton, Southampton, SO14 3ZH, U.K. (


[1] The Makran subduction zone experienced a tsunamigenic Mw 8.1 earthquake in 1945 and recent, smaller earthquakes also suggest seismicity on the megathrust; however, its historical record is limited and hazard potential enigmatic. We have developed a 2-D thermal model of the subduction zone. The results are twofold: (1) The thick sediment cover on the incoming plate leads to high (~150°) plate boundary temperatures at the deformation front making the megathrust potentially seismogenic to a shallow depth, and (2) the shallow dip of the subducting plate leads to a wide potential seismogenic zone (up to ~350 km). Combining these results with along strike rupture scenarios indicates that Mw8.7–9.2 earthquakes are possible in the seaward Makran subduction zone. These results have important earthquake and tsunami hazard implications, particularly for the adjacent coastlines of Pakistan, Iran, Oman, and India, as the Makran has not been previously considered a likely candidate for a Mw > 9 earthquake.

1 Introduction

[2] The position of the updip and downdip limits of subduction zone megathrust rupture controls earthquake magnitude and the intensity of shaking and tsunami in coastal regions. These limits are believed to be thermo-petrologically controlled [Hyndman and Wang, 1993], and so can be investigated through studying the thermal structure of a given subduction zone [e.g., Hyndman and Wang, 1993; Hyndman et al., 1995; Klingelhoefer et al., 2010; Oleskevich et al., 1999]. Studies of this nature can be highly informative for seismogenic potential and resultant hazard assessments, especially at margins where background seismicity is low and/or the historical record is limited, such as the Makran.

[3] The Makran subduction zone is located offshore Pakistan and Iran in the Arabian Sea. The Arabian plate is subducting beneath the Eurasian plate at ~4 cm/yr [DeMets et al., 2010]. The age of the incoming Arabian plate is debated due to a lack of identifiable seafloor magnetic anomalies, but has been estimated as either 70–100 Myr [Whitmarsh, 1979] or 50–60 Myr [Mountain and Prell, 1990]. Heat flow data [Hutchison et al., 1981] support the older age, and we use this for our calculations. The incoming plate has a thick overlying sediment section of up to 5–7 km, the majority of which is accreted forming a wide (>200 km) accretionary prism with a low taper of ~4.5° [Smith et al., 2012].

[4] The Makran has low historical seismicity, which has made estimating its seismogenic potential problematic. Compilations of historically reported events in the Makran have suggested a possible recurrence interval of 100–200 years for Mw > 8 earthquakes (Byrne et al., 1992; Heidarzadeh et al., 2008), with the most recent occurring in 1945 (Mw 8.1). This event, modeled as a shallow thrust plate boundary event, caused a tsunami that killed over 4000 people. Its location has been recalculated with significant variations in estimated latitude (~100 km) and depth (~10 km) [Heidarzadeh et al., 2008]. We use the ISC (International Seismological Centre) location (placing it offshore), but it should be noted that a more landward (northerly) epicenter has also been suggested [Byrne et al., 1992]. Subsequent thrust fault seismicity is concentrated between the region of the 1945 event and the coastline, with few events exceeding M6 (Figure 1).

Figure 1.

Location map of the Makran Subduction Zone. Earthquakes from post-1960 (and pre-1960 with assigned magnitudes) from the EHB Catalog [Engdahl et al., 1998] are illustrated by circles. Those from pre-1960 with no assigned magnitude are small black dots. Significant possible plate boundary events with focal mechanisms from Byrne et al. [1992] and the Global CMT Catalog (magnitudes in inset table). Bathymetry is from the GEBCO_08 Grid [Smith and Sandwell, 1997]. Strike lengths of three rupture scenarios for magnitude calculations are indicated by shaded bars. The thermal modeling profile is marked as a black line. Triangles are volcanoes.

[5] A key question is how a M8+ earthquake with a likely plate boundary origin (1945) was generated beneath the Makran outer prism, when the prism is often inferred to be aseismic due to unconsolidated sediment. It is possible that the deepest input sediments, where the plate boundary décollement forms, may be sufficiently lithified to support seismogenic rupture [Fruehn et al., 1997; Smith et al., 2012]. The Makran also has a shallow dipping megathrust (~2°) [Smith et al., 2012], which may increase coupling and facilitate high-magnitude earthquakes [Gutscher and Westbrook, 2009].

1.1 Constraints on Earthquake Rupture Area

[6] The shallowest portion of the plate interface (upper 5–10 km) is traditionally thought to slip in a velocity-strengthening fashion [Byrne et al., 1988; Hyndman et al., 1997; Saffer et al., 2012; Wang and Hu, 2006]. This view has been challenged by earthquakes such as 2004 Sumatra-Andaman and 2011 Tohoku-Oki, with both showing evidence for locally shallow rupture [Chlieh et al., 2007; Fujiwara et al., 2011; Gulick et al., 2011]. It has been suggested that, although large magnitude earthquakes may not initiate in this shallow region, ruptures which initiate deeper on the plate boundary may propagate to shallow depths [Hu and Wang, 2008]. The resulting wide and shallow rupture increases the tsunamigenic hazard associated with the earthquake. The downdip limit of very large ruptures has been proposed to occur at the intersection with the forearc mantle if hydrated mantle material is present [Peacock and Hyndman, 1999]. If the downgoing plate is very warm, the limit may occur at a shallower depth where the plate boundary reaches temperatures of ~350°C to ~450°C [Hyndman and Peacock, 2003; Oleskevich et al., 1999].

[7] The along-strike rupture length may be influenced by the subduction of topographic basement highs such as oceanic ridges and seamount chains. Basement features have been proposed to either prevent rupture propagation or to initiate rupture [Wang and Bilek, 2011]. The Makran may be seismically segmented along its length into a western and an eastern segment, distinguished by contrasting levels of seismicity (lower in the west), and separated by the Sistan Suture Zone, an onshore strike-slip feature of the continental Eurasian Plate. In addition to this large-scale potential segmentation, the subduction of the Little Murray Ridge in the eastern Makran (Figure 1) may influence rupture propagation. These features will be included in our magnitude calculations to provide different rupture length scenarios.

2 Thermal Modeling Method

[8] To estimate the thermal structure of the Makran subduction zone, we construct a steady-state two-dimensional finite element model using the PGCtherm2D code developed by one of the authors (JH) and previously used by, e.g., Wada and Wang [2009]. Where we developed the model transect (62.9°E), the average convergence velocity is 4 cm/yr, and the sediment thickness is 7 km, of which 5 km is accreted and 2 km underthrust [Smith et al., 2012]. The boundary conditions of our model are 0°C at the top surface and 1450°C at the base of the subducting slab. Because the thermal structure of the subduction zone is controlled mainly by the advection of the temperature profile of the incoming plate, where to place the base of the slab is unimportant [Wang et al., 1995]. The seaward side of the model is defined by a geotherm generated with the SEDTEM 1-D finite element code of Wang and Davis [1992] for the cooling oceanic lithosphere and overlying sediment. As sedimentation rates in the deep Makran section remain poorly constrained, low (~0.1 mm/yr), constant sedimentation rates were used to generate the 7 km sediment section based on the average basement age, while matching the observed average heat flow in the incoming section [Hutchison et al., 1981]. Oceanic plate age was taken as 85 Myr (midpoint of 70–100 Myr: estimate of Hutchison et al. [1981]), but the thermal state of the oceanic plate is similar for this entire age range [Stein and Stein, 1992]. The continental side is defined by a continental geotherm to generate the global average back-arc heat flow value of 80 mW/m2 [Currie and Hyndman, 2006] in the absence of direct observations.

[9] The subducting plate geometry is constructed from combined 2-D seismic reflection lines [Smith et al., 2012], intraslab seismicity (ISC/EHB Catalog: though seismicity is not extensive), and an assumed depth of 100 km for the slab beneath the volcanic arc based on the global average [Wada and Wang, 2009] (Figure 2a). The estimated errors in the depth of the plate boundary range from ±500 m in the outer prism to ±10 km at 100 km depth due to the low level of intraslab seismicity. This construction produces shallowly-dipping plate geometry similar to that predicted by previous studies [Bijwaard et al., 1998; Byrne et al., 1992]. The forearc Moho is assumed to occur at 30–35 km depth in the absence of observational constraints. The main material parameters are described in Table 1. We use a thermal conductivity in the upper plate increasing linearly from 1.2 Wm−1K−1 in the outer prism [Hutchison et al., 1981] to 2.5 Wm−1K−1 at 200 km along the profile. This is to account for the low thermal conductivity of the high-porosity outer prism sediments, relative to the more consolidated forearc rocks. Published values for local density, specific heat, or radiogenic heat production have not been identified, so we have adopted typical values. Viscous mantle-wedge flow is modeled exactly as in Wada and Wang [2009] using the dislocation-creep rheology; however, the wedge flow is unimportant for our study of the shallow thermal regime.

Figure 2.

(a) Results of thermal modeling. Plate interface and forearc Moho indicated by red lines. For simplicity, the plate interface extends to the seaward model boundary to avoid dealing with complications of sediment deformation at the deformation front. Circles indicate seismicity as illustrated on Figure 1. Depths of earthquakes prior to 1980 have very large errors which likely cause the disparity between seismicity and the plotted plate boundary in the seaward region. (b) Comparison of model-predicted heat flow (with varying degrees of frictional heating from Figure 3) with published data located within 50 km of the thermal profile. μ′ = Effective Coefficient of Friction. Data from Hutchison et al. [1981] collected using a thermistor probe attached to a gravity corer. Data from Minshull and White [1989] calculated from BSR depth in reflection data. Heat probe data from Kaul et al. [2000] are fairly consistent with other datasets. Their BSR-derived values may vary from the Minshull and White [1989] results due to differing calculation methods.

Table 1. Table of Input Parameters to the 2-D Numerical Model
  1. Further details of reasoning behind the choice of values are discussed in the text.

Incoming oceanic plate age (Myr)85 (Midpoint of 70–100)[Mountain and Prell, 1990; Hutchison et al., 1981]
Plate convergence rate (mm/yr)4[DeMets et al., 2010]
Incoming sediment thickness (km)7 (5 accreted, 2 subducted)[Smith et al., 2012]
Thermal conductivity of upper plate (Wm−1 K−1)1.2–2.9 Continental crust[Hutchison et al., 1981; Kaul et al., 2000; Wada and Wang, 2009]
2.9 Oceanic Crust
3.1 Mantle
Density (kg/m3)2750 – Continental Crust[Wada and Wang, 2009]
3300 – Mantle and Oceanic Crust
Specific Heat (J/kg K−1)1250 (constant value)[Wada and Wang, 2009]
Heat Production (μWm−3 )0.02 – Mantle and Oceanic Crust[Wada and Wang, 2009]
0.4 to 0.7 – Continental Crust
Effective coefficient of friction (μ′)0.03[Wada and Wang, 2009]

[10] Frictional heating along the plate interface is calculated using the static friction law. The effective coefficient of friction (Table 1) includes the effect of pore fluid pressure. For example, if a true friction coefficient of 0.6 is assumed (similar to Byerlee's law), then the effective coefficient (μ′) of 0.03 corresponds to a pore fluid pressure 95% of lithostatic. At greater depth, the frictional heating changes to viscous shear heating along the interface with rapidly diminishing intensity, assigned using the method of Wada and Wang [2009]. For the continental crust-slab contact with very low effective friction and high strain rate, the transition occurs at temperatures much higher than 350°C. For the shallow part of the mantle wedge-slab contact, because of the presence of talc and other hydrous minerals, the interface becomes extremely weak [Peacock and Hyndman, 1999]; and therefore, we assume the frictional heating changes to shear heating at the depth of the slab-Moho intersect.

[11] The model results are compared with observed heat flow values, both bottom-simulating reflectors (BSR)-derived values and probe measurements (Figure 2b) [Hutchison et al., 1981; Kaul et al., 2000; Minshull and White, 1989]. The differences between these datasets are likely due to differing data collection and processing methods. We favor the direct heat flow measurements of Kaul et al. [2000] and Hutchison et al., 1981, and the BSR-derived values of Minshull and White [1989] to constrain our model as they show good mutual agreement. The scatter within individual datasets may be due to localized fluid flow within the upper sediments and the effects of sedimentation/erosion in the prism. These values for comparison with our model are quite variable, but our model fits within the range of observed data. Sensitivity tests on heat production, thermal conductivity, and sedimentation rate indicate that varying these values has little to no effect on the location of the updip limit (located at 150°) and causes a minimal variation (<5 km) in the placement of the downdip limit (located at the Moho intersection as a minimum or at 350°C as a maximum landward position). Varying these parameters affects the heat flow profile produced by the model by up to 10%.

[12] In thermal models of the shallow part of subduction zones, the greatest source of error is frictional heating along the interface. It overshadows errors in all the other parameters in Table 1. To illustrate model uncertainties, we show interface temperatures predicted with larger (μ′ = 0.06) or no (μ′ = 0) frictional heating (Figure 3). With large frictional heating, the interface may reach 350°C just shallower than the Moho intersect. With no frictional heating, the interface stays cool until warmed by mantle wedge flow, although this is considered to be an unlikely scenario based on the results of previous studies [e.g., Wada and Wang, 2009]. As explained in the preceding section, we assume frictional heat changes to shear heating at the Moho intersect. Having a deeper Moho thus makes the transition deeper; however, with our preferred μ′ = 0.03, the Moho intersect is still much cooler than 350°C (Figure 3). Additional heat flow measurements from the accretionary prism, particularly further landward, would help to constrain the appropriate level of frictional heating.

Figure 3.

Plate boundary temperatures with distance and sensitivity tests results. The peaks in the lines at ~300 km indicate where frictional heating changes to shear heating at the intersection with the Moho.

3 Results

[13] The high plate boundary temperatures (~150°C) at the deformation front (Figure 2a) due to the thick sediment section result in no thermally-predicted aseismic zone beneath the prism. However, the outermost 60 km of the accretionary prism appears to be devoid of significant observed seismicity (Figure 1), so we incorporate this as a possible aseismic zone when calculating minimum-width potential rupture scenarios (Table 2). We do not use the 1945 rupture when defining the seismogenic zone position because of its large location error, and instead we use the cluster of more recent and better located events slightly to the north. The shallow dip of the plate boundary, coupled with the shallow temperature contours, means that a slight (±1–2 km) vertical shift in plate boundary position could lead to significant (±10–20 km) lateral changes in the position of the 150°C contour intersection; however, the thermal structure at the deformation front does suggest temperatures of 150°C at the plate boundary.

Table 2. Potential Magnitude Calculations
ScenarioLength (km)Width (km)Rupture Area (m2)Mo (Seismic Moment) (N m)Mw
  1. Estimated potential magnitudes generated by different rupture scenarios (does not account for partial/heterogeneous rupture). 10 m of coseismic slip is used. Maximum potential rupture width is taken from the deformation front to the 350°C contour. Minimum width is taken from the limit of significant offshore seismicity (~60 km landward of the deformation front) to the forearc Moho (30 km)—subducting plate intersection.

Scenario 1: Full length of subduction zone8002103551.68E + 112.84E + 115.04E + 228.52E + 229.079.22
Scenario 2: Eastern half of Makran, east of Sistan Suture Zone4002103558.4E + 101.42E + 112.52E + 224.26E + 228.879.02
Scenario 3: Sistan Suture Zone to Little Murray Ridge2202103554.62E + 107.81E + 101.386E + 222.343E + 228.698.85

[14] The plate boundary does not reach 350°C until 350 km landward (60 km depth) of the deformation front. The 450°C contour is reached at 370 km (75 km depth). The plate boundary reaches our assumed forearc mantle 90 km seaward of the 350°C contour (at 260 km), and so this may represent the downdip limit if hydrated mantle material is present. It should however be noted that there is currently no evidence to support or refute the presence of hydrated mantle in the Makran, and the position of the forearc Moho is poorly constrained.

4 Implications for Earthquake and Tsunami Hazards

[15] We combine the thermo-petrologically-predicted limits with three likely rupture length scenarios (Figure 1) to calculate the maximum likely earthquake magnitude using the equations of Hanks and Kanamori [1979] (Table 2).

display math

[16] The shear modulus used was 30 GPa [Hanks and Kanamori, 1979]. The slip used to calculate seismic moment (MO) was 10 m. Reducing the slip to 7 m, as modeled by Byrne et al. [1992] for the 1945 event, results in a magnitude reduction of ~Mw 0.1. The minimum rupture width (across strike) used was 210 km, measured from the limit of significant offshore seismicity (therefore, incorporating a ~60 km updip aseismic zone) to the plate boundary intersection with the forearc Moho at 30 km depth (a more conservative value to account for uncertainty in Moho depth). The maximum potential rupture width of 355 km was measured from the deformation front to the 350°C contour. Using the 450°C contour intersection would increase the potential magnitudes by up to a further Mw 0.1. Using the youngest suggested oceanic plate age (~50 Ma) moves the downdip limit seaward by ~50 km, with an associated <Mw 0.1 decrease in magnitude. The updip limit is unaffected. We used the plate boundary temperatures calculated for the μ′ = 0.03 case for these calculations. Using a higher or lower degree of frictional heating would shift the seismogenic zone seaward or landward, respectively, with the most significant potential change being the introduction of a ~40 km updip aseismic zone in the μ′ = 0 case. However, this scenario is already encompassed by our minimum width scenario.

[17] The longest calculated rupture would encompass the full length of the subduction zone (~800 km). There is no historical record of such a rupture occurring (although this record is not extensive or precise), and it is possible that the Makran predominantly ruptures in shorter segments. However, evidence from other subduction zones indicates that megathrust earthquakes can rupture multiple segments, therefore this scenario is included. The second scenario (400 km) encompasses the Pakistan section of the Makran, east of the Sistan Suture Zone. This half of the subduction zone has significantly higher background seismicity than the western half and includes the 1945 rupture. The final scenario (220 km) accounts for the Little Murray Ridge (Figure 1) potentially impeding eastward propagation of a rupture. The different rupture scenarios produce earthquakes with a potential magnitude range of Mw 8.7–9.2 (Table 2). The uncertainties discussed above generate a cumulative ±5% estimated error in these magnitude values.

5 Discussion and Conclusions

[18] The potential earthquake magnitudes of Mw 8.7–9.2 based on thermal modeling of the Makran are larger than any recorded historical event. However, paleoseismological evidence from margins such as Cascadia [e.g. Jacoby et al., 1997] has shown that recurrence intervals can exceed a short historical record. Considering the Makran convergence rate (4 cm/yr), a recurrence interval of 250 years would be required to accumulate the 10 m slip used in this study.

[19] The Makran displays a wide potential seismogenic zone most similar to that of the Alaskan margin (~250 km) [Gutscher and Peacock, 2003; Oleskevich et al., 1999]. However, at the Alaskan margin, the updip limit occurs 50–100 km landward of the trench because the thinner sediment cover results in a much cooler décollement. The potential for shallow rupture due to the occurrence of thick sediments is similar to, but more extreme than, that predicted by thermal modeling of the N. Sumatra region of the 2004 earthquake [Klingelhoefer et al., 2010] where a shallow updip limit is also attributed to thick sediment cover (up to 4–5 km) on the 60 Ma subducting plate, with resultant high temperature and strength basal prism sediments. Our rupture scenarios give the maximum potential rupture extent, but the plate boundary may rupture in a partial or heterogeneous manner in a given earthquake. We currently have no information regarding the degree of coupling between the plates.

[20] Following the 1945 earthquake, tsunami runups of up to 15 m were reported at Pasni, 1.4 m at Karachi, and 2 m at Mumbai [Heidarzadeh et al., 2008]. Previous tsunami modeling of the near-field effects of a rupture east of the Sistan Suture Zone (Rupture Scenario 2, Table 2) produced wave heights of up to 9 m, and up to 15 m for a rupture of the entire subduction zone (Rupture Scenario 1, Table 2) [Heidarzadeh et al., 2009]. These models used rupture widths of 100–150 km and fault displacements of 13–25 m (therefore, comparable Mo to our rupture scenarios). Far-field modeling of similar scenarios highlights the potential hazard to Western India, the Maldives, and the Seychelles from a Makran megathrust event [Okal and Synolakis, 2008].

[21] Assuming worst case scenarios, our thermal modeling indicates that the Makran has a wide potential seismogenic zone and may be capable of generating a very significant (>Mw 8.5) tsunamigenic earthquake. The thick sediment cover (up to 7 km) on the incoming plate leads to high (~150°) plate boundary temperatures at the deformation front making the megathrust potentially seismogenic to shallow depths. The shallow dip of the subducting plate and the lack of significant along-strike rupture barriers lead to a potential seismogenic zone of up to ~350 km in width and 800 km in length, generating large potential earthquake magnitudes. These results illustrate the need for careful analysis of the risks posed by the Makran subduction zone to the neighboring coastlines of Pakistan, Iran, Oman, and India and the need for further paleoseismological investigations. This study may have important global implications for the hazard potential associated with other high sediment input subduction zones, with significant accretionary prisms.

6 Acknowledgments

[22] The authors would like to thank Ikuko Wada for her assistance in developing the numerical model. We thank the Natural Environment Research Council-NERC (Grant No. NE/H524922/1) for studentship support. We thank Harold Tobin and an anonymous reviewer for their helpful comments. This is also Geological Survey of Canada contribution 20120387.