- © 2015 by the Mineralogical Society of America
Intraoceanic volcanic arcs have long been recognized as sites where continental crust is created. Yet, despite their importance to understanding magmatic systems and the evolution of our planet, very little is known about their long-term rates of magma production and crust formation. Constraining both crustal construction and destruction processes at intraoceanic arcs allows for improved estimates of magma production. Our revised magma production rates for active intraoceanic arcs are consistent with those calculated for mid-ocean ridge segments that have slow to moderate spreading rates. This is surprising because magma production at intraoceanic arcs has traditionally been assumed to be significantly less than that at mid-ocean ridges.
Most magmatism observed on Earth is located at mid-ocean ridges and above subduction zones3. At mid-ocean ridges, oceanic plates diverge and the Earth's mantle moves upward. The reduction in the overlying pressure enables mantle melting (a process called “decompression melting”) and leads to magma production and the formation of new oceanic crust. At subduction zones, water and other volatile components stored in the subducting oceanic crust are released at depth due to the increase in pressure and temperature. This water-rich component fluxes the Earth's mantle, thereby lowering its melting point, a process comparable to salt lowering the melting point of snow and ice. This induces melting (referred to as “flux melting”) and produces magma, some of which is erupted at continental and intraoceanic arc volcanoes, but most of the magma never reaches the surface. Instead, it freezes (crystallizes) within the crust to form so-called plutonic rocks. Plutonic and volcanic rocks formed above subduction zones often have chemical characteristics similar to continental crust, and it is a widely held belief that continental crust has been formed by this process throughout Earth's history.
Understanding the rates of global magma generation and crust formation is critical for constraining the evolution of Earth's crust and mantle and for quantifying the overall heat-flow budget of the Earth. For example, magmatic heat loss due to the formation and crystallization of melts that form the oceanic crust contributes ~9% (~4 terawatts) to the total global heat loss of 42–46 terawatts (the other 91% of the Earth's heat loss is thought to occur via conduction through the crust and via mantle convection; Nakagawa and Tackley 2012). However, these estimates generally ignore the role of heat loss related to subduction zone magmatism. Magma production along subduction zones was assumed to be about an order of magnitude less voluminous than that at mid-ocean ridges (Reymer and Schubert 1984; White et al. 2006), therefore contributing only a negligible amount to the global heat-flow budget (about ~1%).
Here, we critically evaluate the above assumption by reviewing recent estimates of magma production and focusing on intraoceanic subduction zones where volcanic arcs form on top of older oceanic crust. The rationale for focusing on intraoceanic arcs is two-fold. First, continental arcs have an additional level of complexity compared to intraoceanic arcs in that they are constructed on preexisting continental crust, thereby making it challenging to decipher what fraction of the crust is composed of old versus new material. Second, continental arcs show pronounced cyclicity: most of the arc crust is built in several short episodes, each of which span 5–20 My. These episodic magmatic “flare-ups” have estimated crust production rates of ~85 km3 km−1 My−1, and are separated by episodes of low magma-production rates of ~20 km3 km−1 My−1, which can last approximately 25–50 My (e.g. Ducea et al. 2015 this issue). The general assumption is that a flare-up is indicative of a period of unusually high magmatic activity, but it is equally possible that the quiescent period is unusually low. Therefore, in order to interpret the periodicity observed in continental arcs, it is essential to quantify the “normal” background magma production rate. This is best done in intraoceanic arcs where periodicity appears to be limited.
We will first critically review the existing data on magma/crust production and propose ways forward to better constrain it. We demonstrate that magma production rates at intraoceanic arcs are comparable to those along mid-ocean ridges and that the high-magma-production periods observed in continental arcs represent the “normal” arc stage. These results have profound implications for understanding Earth's heat loss.
FLAWS IN EXISTING ESTIMATES OF MAGMA PRODUCTION RATES
For mid-ocean ridges, calculating the crust production rate4 is relatively straightforward. The average crust production at mid-ocean ridges is ~18 km3 per year and the global ridge system is 67,338 km long (Bird 2003), which translates to an average crust production rate of ~233–326 km3 km−1 My−1, assuming a 5–7 km thick oceanic crust.
Several attempts have been made to calculate crust production rates at intraoceanic arcs by estimating a date of arc inception and using seismic profiles to approximate arc volumes (Table 1; Crisp 1984; Reymer and Schubert 1984; White et al. 2006). These estimates are generally crust production/accumulation rates and not, strictly speaking, magma production rates. These terms should not be interchanged because crust production rates do not take into account the volume of crust that has been removed from the arc since its formation. Early studies by Reymer and Schubert (1984) and Crisp (1984) produced broadly consistent crust production rates for intraoceanic arcs: ~20–40 km3 km−1 My−1. Dimalanta et al. (2002) combined gravity data with seismic data to better estimate the arc volumes of the western Pacific intraoceanic arcs. They concluded that the crust production rates (between 50–95 km3 km−1 My−1) are somewhat higher than previous estimates, but are still significantly less than those of mid-ocean ridges.
For the approach outlined above, a crucial parameter is the crustal thickness of an active intraoceanic arc. As this parameter becomes better constrained, intraoceanic arc crust rate estimates have historically increased with time from the initial 30 km3 km−1 My−1 estimates of Reymer and Schubert (1984) and Crisp (1984). Generally, it is thought that the crust–mantle transition, and thereby the thickness of the crust, coincides with a significant change in seismic-wave speeds, where primary-wave (P-wave) velocities increase from ~7 km s−1 to > 8 km s−1. In recent years, it has become increasingly clear that arc thickness and composition are hard to constrain remotely because the petrologic crust–mantle transition does not correlate with a significant increase in P-wave velocities (Jagoutz and Behn 2013). This significantly hampers the calculation of crust production rates. Moreover, these estimates yield only the time-averaged crust production rates, which are often highly variable within the same arc system. For example, in the Izu–Bonin–Mariana arc, the crust beneath the northern Izu arc is three times as thick (~32 km) as that beneath the central Bonin arc (~10 km) (Kodaira et al. 2007). At a minimum, this corresponds to a factor of three difference in magma production. More importantly, quantifying the magma production rate (instead of crust production rate) is even more complex due to the large number of processes operating at subduction zones, which can add or remove material from an arc and which need to be taken into account (Fig. 1). Of these, there are three main processes. First, rifting: the breaking up of a section of arc crust, which is common in western Pacific intraoceanic arcs and can transport pieces of crust hundreds of kilometers away from where they were emplaced. Second, surficial and subduction erosion: the processes by which arc crust is eroded or scraped away and transported into the mantle on the downgoing plate (von Huene and Scholl 1991). Third, delamination: loss or sinking of dense lower crust back into the mantle. The estimated bulk composition of the crust in intraoceanic arcs differs significantly from that of mantle-derived melts as significant volumes of material are lost due to the three processes described above (Jagoutz and Schmidt 2012). Therefore crust production rates significantly underestimate magma production rates.
In the following section, we briefly summarize existing crust production rate estimates for active and accreted intraoceanic arcs in the light of recent seismic, petrochemical, and geochronologic studies, and then attempt to determine magma production rates for these arcs.
ACTIVE INTRAOCEANIC ARCS
The Tonga–Kermadec arc extends for 2500 km from American Samoa to the North Island of New Zealand (Fig. 2A). West-southwestward convergence of the Pacific Plate beneath the Australian Plate is fast (20–25 cm y−1). Subduction started around 50 Ma and was caused by a subduction polarity-flip and the reactivation of a fossil subduction zone (Whattam et al. 2008).
A detailed seismic tomographic image of the Tonga–Kermadec subduction zone and a derived 2-D velocity model (Contreras-Reyes et al. 2011) indicate that the arc's crustal thickness ranges between 7 and 20 km. The arc is inferred to be predominantly mafic in composition and lacks the P-wave velocities of 6.0 ± 0.4 km s−1 that are typical of continental crust. Crust production rates for the Tonga–Kermadec arc range from 33 km3 km−1 My−1 (Reymer and Schubert 1984) to 56 km3 km−1 My−1 (Dimalanta et al. 2002) (Table 1).
The Alaska–Aleutian arc extends for ~3500 km from central Alaska to the Kamchatka Peninsula along the northern portion of the Pacific–North American plate boundary (Fig. 2B). P-wave velocity models developed from wide-angle seismic studies in the central Aleutian arc (e.g. Holbrook et al. 1999), in combination with rare S-wave arrivals, have been used to infer that the mid-to-lower crust is composed of a mix of mafic and felsic rocks (Shillington et al. 2013). The average crustal thickness is ~38.5 ± 2.9 km (Janiszewski et al. 2013), which is unusually high for an intraoceanic arc.
Based on an across-arc seismic reflection and refraction survey, Holbrook et al. (1999) determined the volume of material beneath the arc, assumed 55 Ma as the age of arc formation, and calculated a magma production rate of 82 km3 km−1 My−1. Jicha et al. (2006) revised this production rate to 152–182 km3 km−1 My−1 based on a 46 Ma age of arc inception and taking into account material lost over the last 46 My via subduction and glacial erosion (Table 1).
Izu–Bonin–Mariana (IBM) Arc
The Izu–Bonin–Mariana (IBM) arc is located in the western Pacific and extends from Japan to Guam (Fig. 2C). It has been extensively studied over the last decade via dredging, manned submersible sampling, and drilling because its fore-arc preserves widespread exposures of lavas that were generated during the initial stages of subduction. U–Pb zircon and 40Ar/39Ar ages from fore-arc gabbros and basalts spanning the entire 3000 km of the arc date back to 52 Ma (Ishizuka et al. 2011). Despite the wealth of recent petrologic, seismic, and geochronologic data acquired for the IBM arc, there have been no recent attempts to constrain the rates of magma and crust production.
Lesser Antilles Arc
Subduction of the South American plate beneath the eastern margin of the Caribbean plate has resulted in the creation of the 850 km long Lesser Antilles volcanic intraoceanic arc (Fig. 2D). Four features make the Lesser Antilles arc distinctive. First, the arc bifurcates north of the island of Martinique, with the active volcanoes forming the inner arc to the west and inactive, limestone-capped islands forming the outer arc to the east. Second, the convergence rate (~2 cm y−1) is very slow (Wadge 1984). Third, unlike the western Pacific arcs, which have very little sediment accreted to them, the accretionary prism in the Lesser Antilles arc is extremely thick (up to 20 km near Barbados). Fourth, most of the arc may be underlain by a Cretaceous island arc and/or a segment of the Caribbean plateau.
Wide-angle seismic data from a 280 km across-arc transect indicates that the arc has a total average thickness of ~24 km with little variation along strike (Kopp et al. 2011). There is no evidence for an ultramafic cumulate layer like the one that is inferred to exist at the base of the IBM and Aleutian arcs. The only estimate long-term magma production rate for the Lesser Antilles arc is that of Reymer and Schubert (1984) (69 km3 km−1 My−1) (Table 1).
We recognize that there are other active intraoceanic arcs (e.g. Kuriles, South Sandwich), but knowledge of these arcs is generally poor and, thus, fruitful comparisons with the aforementioned active arcs cannot be made.
ACCRETED INTRAOCEANIC ARCS
There are several fossil intraoceanic arcs that have been accreted to continents. These accreted arcs preserve a nearly continuous rock record from the upper mantle to the sediments overlying the volcanic sequence and serve as a physical analog for understanding the structure and processes that are inferred to have taken place in active intraoceanic arcs.
Talkeetna Arc (Alaska, USA)
The Talkeetna arc was active during most of the Jurassic Period (200–153 Ma) and was accreted to south-central Alaska by the end of the Jurassic. In the Chugach Mountains, a crustal cross section is exposed from the petrologic Moho (crust–mantle boundary) through a sequence of lower-crustal gabbro norites to mid-crustal intermediate and felsic plutonic bodies to a 7 km thick section of submarine and subaerial volcanic and volcaniclastic deposits (Hacker et al. 2008). Only 15–30% of the plutonic section survives today. Even though the geochemical, petrologic, and isotopic characteristics of the Talkeetna arc are now fairly well constrained, the crust production rates have yet to be established.
Kohistan Arc (Northern Pakistan)
The Kohistan arc in northern Pakistan is a Cretaceous island arc that was wedged between the Indian and Asian plates during collision around 50 Ma. Unlike the Talkeetna arc, which is partially missing, the entire crustal section of the Kohistan intraoceanic arc is preserved (e.g. Jagoutz and Behn 2013), including a thick section of ultramafic rocks at its base. Subduction-related magmatic activity in the mafic lower crust and mid- to upper-crust intrusive and volcanic rock spans 120 to 50 Ma (Bouilhol et al. 2013). The crust production rate of the Kohistan arc is estimated to have been ~48–72 km3 km−1 My−1 (Jagoutz and Schmidt 2012). Jagoutz and Schmidt (2013), using an approach similar to that outlined in the next section, estimated a significantly higher magma production rate of 220–290 km3 km−1 My−1.
MAGMA PRODUCTION RATES: WAYS FORWARD
As noted by Arculus (2004), the dynamic nature of a subduction zone, in which the arc is losing material simultaneously with growth, makes it difficult to determine what the magma production rate is or has been. This is an important point because most of the existing crust accumulation or magma production rate estimates for intraoceanic arcs do not take into account the material that has been removed from the arc since its inception. The total arc magma production rate (Jarc) is the sum of the volume of arc crust produced (Vcrust) and lost (Vlost) divided by the age of the arc (Fig. 1):
Crust can be lost via any number of processes, including rifting (Vrift), subduction erosion (Vsub), delamination (Vdelam), and surficial erosion (Vsurf). Low density sediments, however, might be relaminated to the base of the arc crust (Vrelam) and so counterbalance the surficial and subduction erosion:
We will now review each one of the processes contributing to Vlost.
For intraoceanic arcs such as the Izu–Bonin–Mariana system, which has experienced periodic rifting since the Oligocene, the volume of the remnant arcs (Vrift) must be taken into account. For other arcs that are mostly intact, such as the Aleutians, the Vrift term is null.
Surficial Erosion, Subduction Erosion, and Relamination
Von Huene and Scholl (1991) proposed that large volumes of sediment and fore-arc crust are being subducted at many of the Earth's subduction zones, and they suggested that arcs can be subdivided into two groups: accretionary (Vsub = 0) and erosive (Vsub > 0). Erosive margins (e.g. IBM, Tonga–Kermadec) are characterized by a fast subduction rate (>6.0 mm y−1), thin sedimentary trench fill (<1.0 km thick), and abundant subduction erosion (Vsub = 20–76 km3 km−1 My−1; Clift et al. 2009). Fore-arc crust may also be eroded in accretionary subduction zones, but it is difficult to quantify how much is being accreted versus recycled into the mantle (von Huene and Scholl 1991).
In addition to accounting for the volume of crust removed from the base of the arc via subduction erosion or delamination (see below), before surface erosion (Vsurf) must also be estimated. Unfortunately, the number of weathering and erosion studies in intraoceanic arcs is limited. For the Aleutian arc, Jicha et al. (2006) suggested that 6–9 km of vertical thickness of the subaerial Finger Bay pluton (2 × 2 km in outcrop) has been removed over the past 35 My, which corresponds to a maximum surface erosion rate of ~1 km3 km−1 My−1. This erosion rate estimate was extrapolated to the entire arc, though this may not be representative of the surface erosion rate throughout the Aleutians. However, most intraoceanic arcs are located in areas that receive significant precipitation, and it has been shown that these sites are, in turn, characterized by the highest erosion rates on Earth (up to 1.5 km3 km−1 My−1) (Gaillard et al. 2011).
Additionally, it has been proposed that some of this recycled material is less dense than the overlying mantle wedge and, therefore, might rise as diapirs and be relaminated at the base of the arc (Vrelam; Hacker et al. 2011).
The observations and mass balance calculations of Jagoutz and Schmidt (2013) indicate that the chemical and petrologic composition of the middle to lower crust in intraoceanic arcs is often not in equilibrium with the upper mantle. They suggest that foundering or delamination of dense, unstable mafic and ultramafic cumulates near the Moho is a volumetrically important and underappreciated process. They also suggest that the principal formation mechanism for evolved melts in intraoceanic arcs is fractionation and that the volume lost to delamination (Vdelam) is approximately twice that of the remaining arc crust (Vcrust). In the Kohistan arc, the combination of fractionation of primitive, high-Mg basalt and delamination of ultramafic cumulates reproduces the intermediate-composition plutons found in the middle crust (Jagoutz and Schmidt 2013).
Unfortunately, many of these processes are poorly constrained or entirely unconstrained. Nevertheless, we can exploit the fact that some of the processes resulting in loss of material also produce a net change in composition of the bulk arc crust. A simple mass balance determines the concentration of an element i in the bulk arc crust: where Ciarc, Cimantle melt, and Cilost represent the concentration of an element of interest (e.g. SiO2) in the bulk arc crust, the mantle derived melts and in the material lost, respectively, and x indicates the relative mass proportions. To outline the importance of these different mechanisms for arc magma production rates, we used seismic velocity measurements to roughly constrain the bulk compositions of different segments of the arc crust. Specifically, we employed the relationship between vP/vS and vP (where vP = P-wave velocity; vS = S-wave velocity) and chemical composition, as calculated by Jagoutz and Behn (2013). Given the uncertainty in relating seismic velocities to crustal composition, we roughly divided arc crust into two compositional groups based on the seismic properties: basaltic (e.g. SiO2 < 52 wt%) and basaltic–andesitic (SiO2 > 52 wt%) (Table 1). Based on the results of Jagoutz and Schmidt (2013), we estimated that the Vdelam is about equal to the total crust produced (Vcrust + Vrift + Vsurf + Vsub) for basaltic–andesitic bulk composition, whereas in basaltic arcs the Vdelam is 1/3 of the total crust produced (see footnote, Table 1). We also made the assumption that the material removed by rifting and by surficial and subduction erosion is comparable in composition to that of the bulk arc crust. Due to the complete lack of constraints on Vrelam, we simply ignored this process. We also decided to ignore the potential volume of preexisting oceanic crust because in exposed arc terranes (e.g. Talkeetna and Kohistan arcs) as well as in the IBM fore-arc (Reagan et al. 2010), old oceanic basement has not been observed. This might relate either to how subduction initiates and new suprasubduction zone crust is formed or to the preexisting oceanic crust being completely assimilated and intruded so that it is unrecognizable. Clearly, a better understanding the mechanisms of subduction initiation and the fate of a possible preexisting oceanic crust in arcs might alter the results presented here.
Using the parameters outlined above, the total arc magma production rates for intraoceanic arcs can be constrained better. Our magma production rate estimates for active intraoceanic arcs vary significantly (Fig. 3).
The lowest magma production rate is for the Tonga–Kermadec arc (157 km3 km−1 My−1) and the highest is for the Izu arc (220 km3 km−1 My−1). Intermediate values come out for the Aleutians (211 km3 km−1 My−1), Bonin (160 km3 km−1 My−1), Mariana (209 km3 km−1 My−1), and Lesser Antilles (162 km3 km−1 My−1) (Fig. 3; Table 1). Similarly, Jagoutz and Schmidt (2013) calculate Vlost and Vdelam and suggest that the rate of magma production in the Kohistan arc was ~220–290 km3 km−1 My−1.
SIGNIFICANCE OF REVISED MAGMA PRODUCTION RATES
Our results indicate that magma production in intraoceanic arcs has been profoundly underestimated because reworking processes have not been properly accounted for. The crust production rate for the Sierra Nevada during the flare-up periods is not unusual and is comparable to crust production rates observed in intraoceanic arcs (Fig. 3). The unusual stages are the intermittent quiescent periods. Similar quiescent periods are currently observed in the Andean margin between the northern, central, and southern volcanic zones where magmatism has essentially ceased due to flat-slab subduction (Stern 2004).
Our reanalysis further suggests that despite the difference in tectonic setting, the rates at which magma is produced at intraoceanic arcs (157–290 km3 km−1 My−1) and mid-ocean ridges (~280 km3 km−1 My−1) is similar (Fig. 3). The contribution from intraoceanic arcs to the global magma generation budget and Earth's loss of internal heat has been underappreciated and needs to be reevaluated. As the total length of subduction zones on Earth (51,310 km) is about 75% of the total length of ridges on Earth (67,338 km; Bird 2003), our revised magma production rates of ~157–290 km3 km−1 My−1 indicate that the contribution of heat loss from subduction zone magmatism should be on the order of 1.2–3.3 terawatts. Accordingly, the heat loss due to magmatism along ridges and subduction zones on Earth is, collectively, ~5.2–7.3 terawatts (~13–17% of the total heat loss). Future studies addressing the role of melt production and crystallization at both divergent and convergent plate boundaries are needed to better constrain the thermal evolution of the Earth.
We thank Mihai Ducea and Scott Paterson for the opportunity to contribute to this issue of Elements. Constructive comments of Richard Arculus, Blair Schoene, and editor John Valley were extremely helpful in shaping this paper. 4 Long-term crust addition and magma production rates for arcs and ridges are commonly expressed in units of volume (km3) per km of arc length per unit of time, which is typically expressed in 106 years or My (km3 km−1 My−1).
↵3 A third location of magmatism is intraplate volcanism (such as Hawaii), which is not discussed in this paper.
↵4 Long-term crust addition and magma production rates for arcs and ridges are commonly expressed in units of volume (km3) per km of arc length per unit of time, which is typically expressed in 106 years or My (km3 km−1 My−1).