Rates of generation and destruction of the continental crust: implications for continental growth

Abstract

Less than 25% of the volume of the juvenile continental crust preserved today is older than 3 Ga, there are no known rocks older than approximately 4 Ga, and yet a number of recent models of continental growth suggest that at least 60–80% of the present volume of the continental crust had been generated by 3 Ga. Such models require that large volumes of pre-3 Ga crust were destroyed and replaced by younger crust since the late Archaean. To address this issue, we evaluate the influence on the rock record of changing the rates of generation and destruction of the continental crust at different times in Earth's history. We adopted a box model approach in a numerical model constrained by the estimated volumes of continental crust at 3 Ga and the present day, and by the distribution of crust formation ages in the present-day crust. The data generated by the model suggest that new continental crust was generated continuously, but with a marked decrease in the net growth rate at approximately 3 Ga resulting in a temporary reduction in the volume of continental crust at that time. Destruction rates increased dramatically around 3 billion years ago, which may be linked to the widespread development of subduction zones. The volume of continental crust may have exceeded its present value by the mid/late Proterozoic. In this model, about 2.6–2.3 times of the present volume of continental crust has been generated since Earth's formation, and approximately 1.6–1.3 times of this volume has been destroyed and recycled back into the mantle.

This article is part of a discussion meeting issue ‘Earth dynamics and the development of plate tectonics'.

1. The continental growth conundrum

The rates and timings of net addition of newly generated (or juvenile) crust to the continental landmass, commonly referred to as ‘continental growth’, have remained matters of considerable debate. This is, in part, because it is difficult to separate the process of crust generation from the processes of crustal destruction, reworking and preservation from the present rock record (e.g. [1–6]). Less than 25% of the volume of the continents contains rocks with Nd or Hf crust formation ages older than 3 Ga ([1,5,7–11], and light brown dashed curve in figure 1), less than 5% of the exposed surface of the continents is of juvenile rocks older than 3 Ga ([14,15], and dark brown dashed curve in figure 1) and there are no known rocks with crystallization ages older than 4.02 Ga [16]. Juvenile rocks preserved on Earth's surface show peaks and troughs in their age distributions [14,15] that translate into stepped-like curves when plotted as cumulative ‘growth curves’ (dark brown dashed curve in figure 1). These age peaks have been regarded as reflecting episodic pulses of new crust generation during mantle ‘superplume’ events [15,17–20], and/or as resulting from the better preservation potential of some rocks over others through time [3,21,22].

Figure 1.

A selection of recent continental growth models (curves 1–4), which suggest that 60–80% of the present volume of continental crust was established by 3 Ga. These curves are in stark contrast with the cumulative distribution of crust formation ages in the crust preserved today (curves 5 and 6). The gap between curves 1–4 and curves 5–6 implies the destruction of large amounts of ancient continental crust (schematized by the dashed vertical arrows). The rates at which continental crust was destroyed and replaced by younger crust are explored in this contribution. References to curves: (1) Dhuime et al. [2], (2) Dhuime et al. [8], (3) Campbell [12], (4) Pujol et al. [13], (5) Condie & Aster [14], (6) Dhuime et al. [8].

Continental growth curves depict how the volume of continental crust has changed with time. They therefore reflect the balance between the generation and destruction of crust at different times in Earth's history. Some curves are calculated from the cumulative proportion of crust formation ages in the present-day crust (figure 1, dashed curves). Such curves are based on the assumption that the relative proportion of juvenile magmatic rocks of different ages that are exposed on the continents surface [14,15], or the Nd isotope composition of continental sediments with a range of deposition ages [7,23], can be used to estimate the volumes of juvenile continental crust generated at different times in Earth's history. However, as recently discussed in Dhuime et al. [8] (and see also refs [1,24,25]), it seems unlikely that the relative proportions of rocks with Phanerozoic and Precambrian crust formation ages presently preserved in the rock record reflect the relative volumes of Phanerozoic and Precambrian crust that had been generated.

A number of approaches have therefore sought to evaluate the volumes of crust of different ages independent of the volumes preserved at the present day (figure 1, solid lines). These include ‘mantle-derived’ growth curves from Nb/U variations in basalts and komatiites [12,26] (black curve), ‘atmosphere-derived’ growth curves from Ar isotope variations in hydrothermal quartz [13,27] (blue envelope curve) and ‘crust-derived’ growth curves from both Nd isotopes in shales [7,8] (dark red curves) and Hf isotopes in zircon [1,2] (green curve). A key issue is the nature of the geochemical reservoirs sampled by the different approaches, and it is striking, for instance, that the worldwide variation in Nd isotopes in shales over the last 1 Ga does not follow the variation in Hf isotopes in zircon (figure 2), although both are records from the continental crust. It is widely accepted that shales (i.e. fine-grained continental sediments) sample the upper continental crust (e.g. [28,29]), and even though zircon is thought to crystallize preferentially from relatively high silica magmas (e.g. [10,30,31]), the distribution of the median zircon ε_Hf data around chondritic (CHUR) values indicates that the zircon record samples more of the bulk continental crust. The discrepancy between the shale and zircon records over the last billion years has, however, little influence on the shape of the continental growth curves modelled from these two records (figure 1, red curves and green curve), as recently demonstrated by Dhuime et al. [8].

Figure 2.

Variation in the ε_Hf in zircon (primary Y-axis) and in the ε_Nd in shales (secondary Y-axis) as a function of the crystallization age of the zircons (zircon data), or the deposition age of the sediments (shales data). The running median of the data calculated for every 100 Myr time slice is represented by the dots (zircon data) and diamonds (shales data). Zircon database (including zircons from both juvenile and reworked crust) from Roberts & Spencer [11] and shales database from Dhuime et al. [8].

The continental growth curves that are not based on the present-day age distribution of crust formation ages encouragingly yield similar results (figure 1, solid lines), suggesting that at least 60% of the present volume of the continental crust (PVCC) was established by 3 billion years ago. These curves are in contrast with those calculated from the cumulative proportion of crust formation ages in the present-day crust (figure 1, dashed lines), and this, in turn, implies that large volumes of pre-3 Ga crust must have been destroyed and replaced by younger crust since the late Archaean. The break in slope at around 3 Ga in many of the most recent growth models (figure 1, solid lines) marks the transition from relatively rapid (ca 3.4–2.9 km³ yr⁻¹ on average before 3 Ga) to slower (ca 0.9–0.6 km³ yr⁻¹ on average after 3 Ga) rates of continental growth. This is increasingly taken to reflect higher crustal destruction rates as plate tectonics and subduction zones developed [2,8].

We have developed a numerical box model to explore new ways to reconcile models in which more than 60% of the PVCC was present by 3 Ga (e.g. [2,8,12,13,26,32–37]) with the scarcity of rocks older than 3 Ga and the age distributions of the present-day continental record (e.g. [5,7–11,14,15]). Two different end-member curves serve as proxies for the relative volumes of juvenile rocks of different ages preserved in today's crust: (i) Model 1 is based on the surface age distribution curve of rocks with juvenile Nd isotope ratios at their age of crystallization [14]; (ii) Model 2 uses the age distribution curve of juvenile crust modelled from the Nd isotope composition of continental sediments [8]. Unlike recent growth models in which the methodology and/or sampling approaches do not allow cumulative curves of crustal volume to decrease over time (e.g. [1,2,9,12,13,38]), the box model approach let us explore scenarios in which crust destruction exceeded rates of crust generation, i.e. in time frames in which net crustal growth rates were negative (see also [21,32]). We used crust generation rates that vary smoothly through time, in line with temporal changes in mantle temperature [39,40], to show that the present-day age distribution of the juvenile crust [14] can be modelled by a number of key changes in destruction rates at different times in Earth's history. Finally, we show that peaks and troughs in the age distribution of juvenile rocks preserved on the continents surface (Model 1) do not necessarily imply any dramatic change in the rates of crustal generation through time.

2. Methodology: the box model approach

A numerical box model illustrated schematically in figure 3 was used to address the effects, at each step t_n, of changing the rates of formation and destruction of the continental crust on both the volume and the age distribution of the juvenile continental crust through time. Each step t_n has a duration of 500 Myr. The model starts at t₀ = 4.5 Ga (0% crust), with its first step at t₁ = 4 Ga, its second step at t₂ = 3.5 Ga, and so on until its last step ends up at t_present = 0 Ga. For each step t_n, we have assumed that the continental crust at that time was made of (i) a segment of new crust formed at t_n; and (ii) the crustal segment(s) formed previously at t_n−i and still preserved at t_n. We also assumed that the volume of crust available at each step t_n was controlled by (i) the volume of crust that was present at t_n−1; (ii) the volume of new crust added at t_n and (iii) the volumes of both new crust (age t_n) and pre-existing crustal segments (ages t_n−i) destroyed at t_n. The model is constrained by a PVCC of 7.2 × 10⁹km³ [38,41], 60 to 80% of the PVCC being present at 3 Ga [2,8,12,13,38], the distribution of crust formation ages in the present-day crust after Condie & Aster [14] (Model 1), or after Dhuime et al. [8] (Model 2), a present-day crustal generation rate (CGR) of 3.2 to 4.4 km³ yr⁻¹ [41–44], and a present-day crustal destruction rate (CDR) of 3.2 to 5.5 km³ yr⁻¹ [41–45]. In Model 2, we assumed a value of 6 for the erosion parameter K, following Dhuime et al. [8] (and see discussions on the significance of K and its impact on continental growth curves in refs [7,8,46]). Using a trial-and-error approach, the rates of crust generation and destruction for each crustal segment at each step t_n were adjusted until all the above-mentioned constraints (i.e. PVCC, volume of crust at 3 Ga, present-day age distribution of juvenile crust, present CGR and CDR) were satisfied. As a further constraint, crustal generation rates were assumed to vary smoothly through time, as does the mantle temperature evolution (e.g. [39,40]). Finally, in order to better account for the preferential destruction of younger high-relief crust through erosion processes (e.g. [7]), crustal destruction rates of the younger continental segments formed at t_n−i were, for each step t_n, not allowed to be lower than those of the oldest segments.

Figure 3.

Schematic of the numerical box model used to evaluate the changes in the rates of crust generation and destruction through time. t₁, t₂, … , t_n are the times at which new crustal segments of ages T₁, T₂, … ,T_n were generated. Vol. at t_n: volume of continental crust established at t_n. The present volume of continental crust targeted by the model is 7.2 × 10⁹km³ [38,41].

The key input and output parameters of the box model are summarized in figures 4 and 5, and in electronic supplementary material, table S1. The brown curves represent the distribution of crust formation ages in the present crust targeted by our models: Model 1 targeted curve (figure 4a) is from Condie & Aster [14], and Model 2 targeted curve (figure 5a) is from Dhuime et al. [8]. The red squares represent the calculated present-day distribution of crust formation ages generated by the box model. Crust generation and destruction rates (figures 4b and 5b) were adjusted until the model distribution of today's crust formation ages matches the targeted age distributions within ± 2%. The green curve linking green dots in figures 4c and 5c is the continental growth curve generated by the model.

Figure 4.

Model 1 for the volume and rates of generation and destruction of the continental crust through time, estimated from the box model approach schematized in figure 3. (a) Cumulative curve for the present-day distribution of juvenile crust (brown curve, after [14]), which is the target of the model, and the modelled present age distribution (red squares). The grey box at 3 Ga represents the conditional starting parameter of the model, i.e. 60–80% of the present-day volume of continental crust established by 3 Ga [2,8,12,13]. (b) Rates of generation (blue curve), destruction (red curve) and net growth (green curve) of the continental crust, for every 500 Myr step of the model. The inset shows the smooth evolution of the mantle temperature through time, after Herzberg et al. [39] and Korenaga [40]. (c) Volumes of continental crust calculated for every 500 Myr step of the model (green curve), and estimated changes in the volumes of pre-3 Ga (purple curve) and post-3 Ga (orange curve) continental crust through time.

Figure 5.

Model 2 for the volume and rates of generation and destruction of the continental crust through time, estimated from the box model approach schematized in figure 3. The curves are as for figure 4, with one exception as the targeted present-day age distribution curve (brown curve) is after Dhuime et al. [8].

3. Rates of generation, destruction and growth of the continental crust through time

The crust generation and destruction rates calculated from the box model are represented by the blue and red curves respectively in figure 4b (Model 1) and 5b (Model 2). Rates of continental crust generation range 3.0–4.7 km³ yr⁻¹ (Model 1) and 3.1–4.0 km³ yr⁻¹ (Model 2). They broadly follow the evolution of mantle temperature (figures 4b and 5b, inset), increasing between Earth's formation and ca 3.5–3.0 Ga, and then decreasing to the present day. By contrast, crustal destruction rates show much greater variations. They range between 0.1–8.4 km³ yr⁻¹ (Model 1) and 0.1–5.5 km³ yr⁻¹ (Model 2), with a marked peak in the period 3.0–2.5 Ga. This peak was generated by the numerical box model because (a) no restriction was applied for the minimum or the maximum crust destruction rates, and (b) 3.0–2.5 Ga is the period of the maximum difference between the crustal growth curve and the targeted present-day age distribution of the juvenile crust. This peak is followed by lower destruction rates after ca 2.5 Ga (1.7 km³ yr⁻¹ in Model 1 and 1.0 km³ yr⁻¹ in Model 2), with a gradual increase in destruction rates during the Proterozoic. By the Phanerozoic crustal destruction rates (5.3 km³ yr⁻¹ in Model 1 and 5.5 km³ yr⁻¹ in Model 2) exceeded crust formation rates (4.1 km³ yr⁻¹ in Model 1 and 3.5 km³ yr⁻¹ in Model 2).

The changes in the rates of net continental growth through time, calculated from the variations in crust generation and destruction rates, are shown by the green curve in figures 4b (Model 1) and 5b (Model 2). Rates of continental growth are highly variable and range between −3.7 and 4.3 km³ yr⁻¹ in Model 1, and between −2.0 and 3.3 km³ yr⁻¹ in Model 2. From 4.5 to 3.0 Ga, they range from 2.8 to 4.3 km³ yr⁻¹ (Model 1) and 2.8 to 3.3 km³ yr⁻¹ (Model 2). There is a marked reduction in crustal growth rates between 3.0 and 2.5 Ga, with a negative value of −3.7 km³ yr⁻¹ in Model 1 and −1.5 km³ yr⁻¹ in Model 2. From 2.5 Ga until the present day, net growth rates gradually decrease from 3.0–3.2 km³ yr⁻¹ to −1.1 km³ yr⁻¹ (Model 1), and from 3.0 km³ yr⁻¹ to −2.0 km³ yr⁻¹ (Model 2). The cumulative growth curve calculated from the net growth rates generated by the box model is represented by the green curve in figures 4c (Model 1) and 5b (Model 2). This curve differs from recent continental growth models (figure 1) because our box model allows the incorporation of negative crustal growth rates when building a cumulative growth curve.

While the models presented here are not unique (e.g. as evidenced by small differences between Model 1 and Model 2), the approach offers the opportunity of developing more realistic growth curves for the evolution of the continental crust through time. This evolution can be summarized in four main stages: a first stage of rapid growth in the volume of continental crust between ca 4.5 Ga and 3 Ga (70% of the PVCC at 3 Ga in Model 1, and 63% in Model 2); a second stage of crustal destruction and continental shrinking between ca 3 Ga and 2.5 Ga (44% of PVCC at 2.5 Ga in Model 1, and 53% in Model 2); a third stage of crustal growth with gradually decreasing rates of growth between ca 2.5 Ga and 0.5 Ga, at the end of which the volume of continental crust may have exceeded its present volume (108% of the PVCC at 0.5 Ga in Model 1, and 114% in Model 2); a fourth stage between ca 0.5 Ga and present, during which the volume of continental crust has slightly decreased as crustal destruction rates exceeded crustal formation rates (e.g. [44,45]).

4. Geological implications

The processes involved in the development of the peaks and troughs in the age distributions of juvenile rocks, and more generally in the age distribution of zircons, have remained controversial for a number of decades (e.g. see recent reviews in [4,38,47]). Dramatic, episodic variations in crustal growth rates in relation to changes in mantle convection and the formation of ‘superplumes’ have been invoked to account for age peaks widely observed in the preserved rock record. Although our model does not rule out an episodic Earth evolution (e.g. [18]), it offers new insights into alternative models in which continental crust was continuously extracted from the mantle, and the age distribution of today's juvenile rock record is at least partly explained by changes in the rates of destruction of continental crust. This, in turn, relates to the development of subduction zones in a planet evolving from a ‘single/stagnant lid’ to a regime in which plate tectonics dominates (Cawood et al. [48] and references therein).

Recent studies have suggested that ca 3 Ga marked the transition between two different types of continental crust. New continental crust generated before 3 Ga was on average mafic, dense and relatively thin (less than 20 km) [49], and the upper crust preserved from that time was also relatively mafic [50]. By contrast, continental crust that formed after 3 Ga gradually became more intermediate in composition, increasingly buoyant and thicker [49]. The volumes of pre-3 Ga and post-3 Ga continental crust available through time calculated with our box model are shown in figures 4c (Model 1) and 5c (Model 2), and these volumes are represented by the purple dashed curve and red dashed curve, respectively. These highlight that the reduction in estimated crustal volumes at the end of the Archaean reflects the destruction of largely mafic continental crust, and the initiation of the generation of continental crust with more immediate compositions.

In the context of the inferred onset of widespread subduction at around 3 Ga (e.g. [2,8,24,51]), the high peak of destruction rate predicted by the box model for this period is strikingly consistent with the rapid recycling of mafic/dense pre-3 Ga crust back into the mantle through subduction. Dense, predominantly mafic and relatively thin pre-3 Ga continental crust reached a volume of 60–70% of the PVCC at 3 Ga (figures 4c and 5c), but has been almost completely destroyed since then. Only small amounts of that crust preferentially survived after a billion years (i.e. after 2 Ga) of crustal evolution and recycling (figures 4a and 5a, brown curve). High 3.0–2.5 Ga destruction rates are also consistent with the age distributions of zircons from sediments with a range of deposition ages, because ages greater than 3 Ga are poorly preserved in sediments younger than 2.5 Ga [52,53]. The gradual increase in destruction rates from the Archaean–Proterozoic transition predicted by the model can be accommodated by a change in subduction dynamics as the Earth became cooler [54–57] and/or by the emergence of a thicker, buoyant, higher relief and therefore more prone to erosion, new continental crust [31,49,50,58–60].

The growth of ‘modern’, increasingly thicker, more differentiated and buoyant post-3 Ga continental crust is represented by the orange dashed curve in figures 4c (Model 1) and 5c (Model 2). The generation of that crust, dominantly through subduction [49], is likely to be associated with the emergence of the continents from 3 Ga [49,50,58,59,61]. The shape of the growth curve of the post-3 Ga crust is similar to that of the juvenile crust thickness curve of Dhuime et al. [49], as both curves show a gradual increase from the Mesoarchaean to the Meso/Neoproterozoic followed by a decrease towards the present. This latter feature in the crustal thickness curve was interpreted by Dhuime et al. [49] as crustal destruction exceeding crustal generation rates, consistent with Phanerozoic rates [41–45], and it is independently validated by both Model 1 and Model 2 (figures 4b and 5b).

Finally, the data generated by our box model imply that about 2.6 times (Model 1), or 2.3 times (Model 2), of the PVCC has been generated since Earth's formation, and thus ca 1.6 times (Model 1), or 1.3 times (Model 2), of the PVCC has been destroyed and recycled back into the mantle since the onset of plate tectonics. This opens new perspectives for models of mantle evolution and mantle–crust interaction through time, including testing the scenarios offered by our box model by modelling of radiogenic isotope systematics in the mantle–crust system.

Data accessibility

This article has no additional data.

Authors' contributions

B.D. and C.J.H. designed the study and wrote the paper. B.D. and H.D. processed the data and designed the figures, and P.A.C. contributed to linking the box model approach with geodynamical processes.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by the Natural Environment Research Council (NERC grant NE/K008862/1 to B.D.), the Leverhulme Trust (grant RPG-2015-422 to B.D. and C.J.H., and Emeritus Fellowship EM-2017-047∖4 to C.J.H.), and by the Australian Research Council (grant FL160100168 to P.A.C.).