RESEARCH ARTICLE Compositional Attributes of the Deep Continental Crust

10.1029/2022JB024041
Inferred From Geochemical and Geophysical Data
Key Points:
Laura G. Sammon1 , William F. McDonough1,2 , and Walter D. Mooney3
• W e present a global model for the
composition of the deep continental 1
Department of Geology, University of Maryland, College Park, MD, USA, 2Department of Earth Sciences, Research Center
crust constrained by geochemical and
for Neutrino Science, Tohoku University, Sendai, Japan, 3Earthquake Science Center, United States Geological Survey,
geophysical data
• Crustal SiO2 content decreases with Menlo Park, CA, USA
increasing depth, and compositions
correlate to relative depth rather than
absolute depth Abstract This study provides a global assessment of the abundance of the major oxides in the deep
• Moho heat flux is predicted at
continental crust. The combination of geochemistry and seismology better constrains the composition of the
𝐴𝐴 21.6+16.0 mW/m 2
−5.6
middle and lower continental crust better than either discipline can achieve alone. The inaccessible nature
of the deep crust (typically >15 km) forces reliance on analog samples and modeling results to interpret its
Supporting Information:
bulk composition, evolution, and physical properties. A common practice relates major oxide compositions
Supporting Information may be found in
the online version of this article.
of small- to medium-scale samples (e.g., medium to high metamorphic grade terrains and xenoliths) to large
scale measurements of seismic velocities (Vp, Vs, Vp/Vs) to determine the composition of the deep crust. We
provide a framework for building crustal models with multidisciplinary constraints on composition. We present
Correspondence to:
L. G. Sammon,
a global deep crustal model that documents compositional changes with depth and accounts for uncertainties
[email protected] in Moho depth, temperature, and physical and chemical properties. Our 3D compositional model of the deep
crust uses the USGS Global Seismic Structure Catalog (Mooney, 2015) and a compilation of geochemical
Citation: analyses on amphibolite and granulite facies lithologies (Sammon & McDonough, 2021, https://doi.
Sammon, L. G., McDonough, W. F., & org/10.1029/2021JB022791). We find a SiO2 gradient from 61.2 ± 7.3 to 53.3 ± 4.8 wt.% from the middle to
Mooney, W. D. (2022). Compositional the base of the crust, with the equivalent lithological gradient ranging from quartz monzonite to gabbronorite.
attributes of the deep continental
crust inferred from geochemical
In addition, we calculate trace element abundances as a function of depth from their correlations with major
and geophysical data. Journal of oxides. From here, other lithospheric properties, such as Moho heat𝐴𝐴flux (21.6+16.0
−5.6
mW/m 2), are derived.
Geophysical Research: Solid Earth,
127, e2022JB024041. https://doi.
org/10.1029/2022JB024041
Plain Language Summary Using many different geophysical and geochemical techniques together
helps us understand the composition of the bottom two-thirds of the continental crust. We cannot sample much
Received 14 JAN 2022 of the continental crust directly because of how deep it is. Instead, we rely on rocks that have been brought to
Accepted 15 JUL 2022 the surface and measurements of the speed of seismic waves traveling through the crust in order to determine
what the deepest parts of the crust are made of. Accounting for various factors, such as crust temperature and
Author Contributions: tectonic setting, allows us to create a large-scale model for the composition of the deep crust.
Conceptualization: Laura G. Sammon,
William F. McDonough
Data curation: Laura G. Sammon, 1. Introduction
Walter D. Mooney
Formal analysis: Laura G. Sammon The deepest parts of Earth's crust are widely inaccessible to traditional geochemical sampling and so their compo-
Funding acquisition: William F.
McDonough, Walter D. Mooney sition is poorly understood. Only in areas where eruptions have brought xenoliths to the surface or where tectonic
Investigation: Laura G. Sammon activity has exhumed medium and high grade metamorphic terrains are we able to partially determine the compo-
Methodology: Laura G. Sammon sition of the deep (middle and lower) continental crust. Even so, these ex situ, aged, weathered, and transported
Project Administration: William F.
McDonough rocks may not adequately represent the overall, current composition of the deep crust. Such inaccessibility has
Resources: Walter D. Mooney challenged geochemists for decades (Dumond et al., 2018), leading to competing models for continental crust
Software: Laura G. Sammon and bulk silicate Earth (BSE) compositions, formation, and evolution (e.g., Javoy et al. (2010); McDonough
Supervision: William F. McDonough
Validation: Laura G. Sammon, Walter et al. (2020); Turcotte and Schubert (2014)). Dissonance in the geochemical community stems from known and
D. Mooney unknown unknowns; that is, we are mostly certain of the uncertainties in our geochemical and petrological meas-
urements, but we are uncertain if our samples are truly representative of large swathes of the deep crust or if they
are merely point samples. Xenoliths and terrains are the sum of the processes that form them, which may cause
© 2022. The Authors. them to differ from what is presently at 15–45 km and deeper.
This is an open access article under
the terms of the Creative Commons Seismological techniques, however, provide another piece of the cipher by directly measuring the physical state
Attribution License, which permits use, of large sections of the deep crust. Physical properties (e.g., density, Poisson's ratio, Vp, and Vs) determined from
distribution and reproduction in any
medium, provided the original work is these in situ geophysical experiments can be compared to laboratory experiments on rocks of known composi-
properly cited. tions, particularly medium to high grade metamorphic lithologies (amphibolite and granulite facies rocks) to

SAMMON ET AL. 1 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Visualization: Laura G. Sammon place constraints on estimates of deep crustal composition. Integrating geochemical and geophysical observa-
Writing – original draft: Laura G.
tions, related to each other by empirically (laboratory) derived thermodynamic properties, provides a reinforced,
Sammon
Writing – review & editing: Laura G. clearer, consistent picture of middle and lower crustal composition.
Sammon, William F. McDonough, Walter
D. Mooney This study uses geophysical and geochemical datasets to build a global compositional model of the lower
two-thirds of the continental crust. We generate a composition vs. depth model for the middle and lower conti-
nental crust by applying thermodynamic modeling software to medium and high grade lithologies. We then
compare the thermodynamically-generated seismic velocities to velocities obtained from seismological measure-
ments to produce a jointly constrained geochemical-seismological compositional model.

2. Methods
Our model calculations are conducted in two steps: (a) assembling the data and performing thermodynamic
calculations, and (b) adjusting model parameters to generate deep crustal compositional models with uncertain-
ties. These calculations require seismic velocity depth profiles, Moho depths, and crustal temperature gradients
for areas of interest. Using the thermodynamic modeling software Perple_X (Connolly, 2005), we calculate the
probability that different deep crustal compositions could produce the observed seismic signal. These calcula-
tions are conducted using our modeling software, CrustMaker, which is provided as an electronic supplement.
This software calculates the probability of a given rock composition producing the observed seismic signal by
integrating the area of overlap between the Perple_X-predicted seismic velocity for said sample and an observed
seismic velocity profile.

The calculation adopts a subdivision of the global continental crust into 13 tectonic provinces (Figures 1 and 2)
to speed calculations and extrapolate results to areas with lower data coverage. Individual tile resolution of this
global model is set to 1°latitude × 1°longitude × 3 km depth as a default, but can be changed in the model to suit
user needs. We chose this default resolution for our global model based on the resolution of our crustal catego-
ries (each 1° × 1° of crust was assigned a tectonic province based on models such as CRUST1.0, Litho1.0, and
modifications discussed further in Section 2.1), and the resolution of our crustal thickness and temperature data,
the ramifications of which are discussed further in the Results section. For considering higher resolution, regional
scale data, the same methods can be used. Instead of simplifying the crust into tectonic provinces, calculations are
run for individual seismic velocity profiles, so that if there are, for example, 34 seismic velocity profiles as inputs,
there will be 34 locations for which compositional profiles are generated.

We calculated the overlapping probability between measured seismic velocities and the Perple_X-derived veloc-
ities for amphibolites and granulites equilibrated at middle and lower crustal pressures and temperatures. A
reference average crustal density of 2,900 ± 200 kg/m 3 is used to calculate lithostatic pressure (Wipperfurth
et al. (2020), cf., Christensen and Mooney (1995)). Integrating the area under both curves, the area shown as
magenta in Figure 3, for a sample of composition X yields the total probability of sample X producing the
observed seismic signal. Repeating this technique for a multitude of sample compositions at various depths and
temperatures yields a final Monte Carlo model for a deep crustal composition. Probability distributions are gener-
ated for Vp, Vs, and Vp/Vs and then multiplied together to further constrain the final probability.

2.1. Model Inputs

A global model of Vp, Vs, and Vp/Vs was generated from a compilation of over 5000 (Vp) and 1000 (Vs) 1-D
seismic velocity profiles obtained from the Global Seismic Structure Catalog (GSSC) database (Mooney, 2015).
Both controlled and passive source seismic velocity profiles were included to increase data coverage. We included
profiles with Vp and Vs data that had been sampled at a minimum of 5 depth intervals within the crust. Figure 2
shows our tectonic province designations and the location of each seismic velocity profile used. We used global
Moho depths from Litho1.0, except on the continental margins, where we reference Szwillus et al. (2019) Moho
values. In comparison to Litho1.0, Szwillus et al. (2019) incorporated a larger data set on the continental margins
(∼1600 profiles) and did not average depths across the continent-ocean transition. Global Moho temperatures
were generated from the TC15 global temperature model of Artemieva (2006). We assumed a linear temper-
ature gradient within the continental crust, though we address the contributions from crustal heat production
in a later section of this paper. The Szwillus et al. (2019) and Artemieva (2006) models are based on the same
USGS-maintained database as Mooney (2015). In most areas, the middle and lower crust are not in thermal

SAMMON ET AL. 2 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

equilibrium with the conditions of amphibolite and granulite facies, respec-

tively. Temperature profiles suggest that the deep crust has cooled off since
reaching peak metamorphic grade. To account for this in our composition
calculations, we adopt the assumptions of Hacker et al. (2015), which uses
an equilibration temperature of 500°C for amphibolite facies in the middle
crust and 750°C for the granulite facies in the lower crust. We performed a
temperature correction of 4 × 10 −4 km s −1C −1 (Christensen & Mooney, 1995;
Rudnick & Fountain, 1995) to account for crustal cooling. We also perform a
pressure correction of −2 × 10 −4 km s −1 MPa −1 on both facies (Christensen
& Mooney, 1995; Rudnick & Fountain, 1995).

The foundation of the tectonic provinces chosen for this global model are
the classifications of crust provided by the Crust family of models (Mooney
et al., 1998). To further identify tectonic provinces and group together
geophysically similar crust, we incorporated crustal thickness, seismic
velocity (Vp, Vs), gravity anomaly, sediment thickness, crust elevation, and
surface heat flux observations in a tSNE test (t-distributed stochastic neigh-
bor embedding). Results generally favored grouping the continental crust into
8–12 regimes, mostly matching the designations already given in Crust1.0.
We augmented these regimes with additional groupings, such as “Thinner
Himalyan” crust, when it became clear that the seismic velocity structure
of the perimeter of the Himalayas differed from the thickest Himalayan
crust. Areas with sparse seismic coverage such as central South America and
northern Africa, rely heavily on extrapolation of measurements from similar
tectonic provinces. Average Vp and Vs profiles for most tectonic provinces
were created from a distribution of tens to hundreds of individual measure-
ments. A notable exception is the “Continental Margins” province, which was
represented by >1,600 profiles. Highly localized regimes, such as Andean or
Himalayan crust, tended to have <100 profiles due to the uniqueness of their
crustal profiles.

Figure 1 and Table 1 show the proportion of different crustal provinces by

surface area coverage. These tectonic provinces consider only crust exposed at
the surface, so that provinces such as “Platform” have underlying crystalline
crust that may be Proterozoic or Archean in age. The Proterozoic crust covers
the largest fraction (32%) of the continental crust, followed by continental
margins (17%). The thicknesses of the crust and lithospheric mantle in Table 1
were calculated from the Litho1.0 Moho and lithosphere-asthenosphere
boundary depths. The masses and volumes of the lithospheric mantle are
from Wipperfurth et al. (2020)'s physical properties associated with Litho1.0.

Sammon and McDonough (2021) serves as our geochemical constraint on

the deep (middle and lower) continental crust. That study represents a compi-
lation of major and trace element abundances for amphibolite and granulite
facies rocks. We modeled amphibolite facies lithologies for the middle third
of the crust and granulite facies lithologies for the bottom third, in agreement
with the depth assignment of Rudnick and Gao (2014). We cannot confi-
Figure 1. The weighted area proportion of crustal types, or “tectonic dently determine a priori which portions of the deep crust are more appropri-
provinces,” used for our model as (a) a fraction of total crust and (b) a ately represented by amphibolite versus granulite facies data with our current
fraction of continental crust. Proterozoic crust is most abundant (32% of the model. In theory, one facies-classified lithology would have greater overall
continental crust), followed by continental margins (17%) and Archean crust overlap with the seismic velocity profile(s), thus determining which is the
(10%). Modern and paleo-orogens, including arcs, make up a combined 15%
of the continental crust in our model.
more accurate rock type to use. In practice, however, amphibolite and gran-
ulite facies lithologies of the same SiO2 abundance tend to have similar seis-
mic velocities (see Section 3.1). As such, we have assumed a metamorphic

SAMMON ET AL. 3 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Figure 2. (a) The mapped distribution of our 13 crust types and (b) the seismic velocity profile data distribution from
the USGS Global Seismic Structure Catalog. Data coverage is greatest in the northern hemisphere while places with less
coverage, like Africa and Antarctica, rely more heavily on extrapolation of crust type.

facies switch from amphibolite to granulite at 2/3 the crustal depth. Future investigations should use seismic
anisotropy of the deep crust to further constrain its mineralogy.

Though trace elements do not participate in thermodynamic calculations, we were able to estimate trace element
abundances based on a joint probability analysis with the mineral-forming major oxides. Samples were placed
into bins based on the abundance of the oxide and trace element of interest (e.g., SiO2 and U). Bin width was
selected using Sturge's rule (Nbins = log2(N) + 1). For each major oxide composition bin, there was then a corre-
lated trace element abundance distribution.

SAMMON ET AL. 4 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

2.2. Model Uncertainties

Errors in seismic and/or geochemical inputs will skew results. It is imperative

to understand the uncertainties in the input datasets, if we want an accurate
picture of the uncertainty of our crust compositional models.

The program also will not assess modeling errors stemming from foundational
assumptions about what types of lithologies should be used as geochemical
inputs and the tectonic provinces assigned to the global crust. These two
assumptions are expected to control the systematic error of the model, which
is why we made the program flexible and modular. Our approach facilitates
testing different crustal models and highlights the projected differences in
crust composition.
Figure 3. A conceptual illustration of how overlapping velocity distributions
are used to identify probable crust compositions. The central pink/magenta
The primary sources of model error stem from uncertainty in the assumed
region of the diagram, where the measured seismic velocity distribution crustal temperature gradient and Moho depth. Again, these are param-
(purple) overlaps the Perple_X-generated velocity distribution (tan), are the eters that can be set by the user. For our preferred model, the uncertainty
velocities that are considered the best-fit for the crust. The model records the on Moho depth is on the order of 10% or less in most areas of the global
compositions of the samples that can produce the best-fit velocities. model (Szwillus et al., 2019). The temperature uncertainty is much greater.
Global Moho temperatures are taken from Artemieva (2006), which reports
no uncertainties. Therefore, uncertainty is taken as the standard deviation of
all temperatures found within a given crustal tectonic province (discussed below), and the model runs a Monte
Carlo simulation of at least 10,000 iterations to produce a distribution of Moho depths and temperatures. Future
results could be improved with Moho temperature models that quantify uncertainty more directly.

We have also attempted to mitigate the bias introduced by oversampling of particular geochemical composi-
tions. An over-sampled composition, such as 100 input compositions with nearly identical major oxide content

Table 1
Physical Properties of the Continental and Oceanic Crust and Lithospheric Mantle
Surface area Crust Lithospheric mantle

All Continental
crust crust Density Thickness Total mass Total volume Thickness Total mass Total volume

% % (kg/m ) ± 15%
3
(km) ± 12% (10 kg) ± 19%
21
(10 km ) ± 12%
7 3
(km) ± 12% (10 kg) ± 13%
21
(10 7 km 3) ± 12%
Oceanic 56% – – 10.7 9.5 310 10.7 1.4 42
Continental 7.6% 17% 2930 35.2 4.0 140 84.6 8.4 260
margin
Archean 4.5% 10% 2930 43.8 3.0 100 225 11 350
Proterozoic 14% 32% 2920 44.1 9.4 320 233 24 740
Phanerozoic 5.7% 13% 2920 36.6 3.1 110 172 4.9 150
Platforms 1.8% 4.1% 2940 43.8 1.2 40 147 3.3 100
Extended 3.1% 7.1% 2910 36.5 1.7 58 62.7 2.1 64
Rifted a 0.32% 0.73% 2850 44.0 0.20 7.2 104 0.60 18
Arc 1.5% 3.4% 2910 37.3 0.82 28 72.1 1.8 56
Paleo-orogenic 4.0% 9.0% 2900 44.1 2.6 89 78.6 3.0 90
Tethyan 0.6% 1.3% 2890 48.0 0.41 14 78.1 0.55 17
Andean 0.14% 0.32% 2840 56.3 0.11 4.0 125 0.25 7.5
Thickest crust 0.27% 0.62% 2770 71.6 0.28 10 127 0.35 10
Thin Himalayan 0.22% 0.50% 2740 59.0 0.18 6.7 88.9 0.21 6.3
Total 100% – – 24.1 36 1200 76.1 63 1900
Total CC 44% 100% 2880 41.1 23.2 920 159 61 1860
a
Rifted: includes active, failed, and paleo rifts.

SAMMON ET AL. 5 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

artificially inflates the probability of that composition in our final combined

model. However, we do consider the reporting of compositions to be at least
somewhat reflective of the proportion of rock types present in the deep crust,
that is, if the distribution of reported compositions is bimodal, the rocks
in the deep crust are likely bimodal in composition. Therefore, we only
considered a sample redundant if its oxide content differed from another's
by <3 wt.% (9 major oxides, using the distance between vectors formula
√
𝐴𝐴 d = 𝑥𝑥21 + 𝑥𝑥22 + ⋯ + 𝑥𝑥2𝑛𝑛 , where xn is the difference in wt.% of an oxide
between two samples), and its Perple_X generated values for Vp, Vs, and
Vp/Vs were within uncertainty of each other.

The internal error contributed by calculational uncertainty is minimal. The

overlap between of seismic velocity measurements and Perple_X-derived
seismic velocities is calculated via trapezoidal numerical integration at inter-
vals determined by the uncertainty in the seismological data. When this inter-
val is too large to use for the integration, the program reduces the interval
by half. The precision errors of Perple_X are generally negligible compared
to the uncertainty on our other inputs (Connolly, 2005). The systematic
Figure 4. Vp as a function of SiO2 wt.% for (a) amphibolite and (b) granulite
uncertainty of Perple_X, however, remains difficult to assess. As the soft-
facies litholoiges at expected deep crustal pressures and temperatures.
Amphibolite results are shown for 450°C and 0.5 GPa. Granulite results are ware is built upon empirical measurements and thermodynamic equations,
shown for 600°C and 0.85 GPa. The color of the data points indicates percent the empirical database for crustal minerals is incomplete and relies on ideal
data point density, with the brighter colors indicating more data points. thermodynamic relationships to compute expected physical properties. The
empirical databases themselves also vary with systematic bias.

2.3. Quality, Expense, and Time: Global Versus Local Models

In numerical modeling, there is often a tradeoff between computation time and model resolution. For a global
view of the continental crust, breadth and total model coverage may be more valuable than high data resolution,
especially if results can be averaged over large areas. A large-scale, globe-encompassing model, however, comes
with the choice of either short computation time and low resolution or longer
computation time and higher resolution. Alternatively, those interested in a
more in-depth analysis of a localized region may be able to accommodate
higher resolution models. We suggest considering the following when deter-
mining whether to use a global or local scale model: data resolution (espe-
cially in seismic velocity profiles), data coverage, and model application.
Those with data resolution on the scale of >0.5° × 0.5° should consider using
the global version of the script. Data sources with higher resolution, such as
that provided by the Earthscope USArray, the AUSArray, or the J-ARRAY,
should use the regional scale model. For the remainder of this study, we will
analyze global model results. Sammon et al. (2020) presents an example of
a local-scale composition analysis using a nascent version of this method.

3. Results
3.1. Empirical Composition-Velocity Trends

For granulite and amphibolite facies lithologies seismic velocities most

strongly correlate with SiO2 content because of its high abundance, as
compared to all other oxides. Perple_X-calculated Vp and Vs values at given
Figure 5. Vs as a function of SiO2 wt.% for (a) amphibolite and (b) granulite
facies litholoiges at expected deep crustal pressures and temperatures, pressure-temperature conditions show a quadratic relationship between
calculated by Perple_X. Amphibolite results are shown for 450°C and 0.5 GPa. SiO2 and velocity (Figures 4 and 5). The coefficients of the quadratic best
Granulite results are shown for 600°C and 0.85 GPa. The color of the data fit curve vary for different pressures and temperatures, and are ultimately
points indicates percent data point density, with the brighter colors indicating correlated to the empirical mineral physics datasets used in the Perple_X
more data points. There is more scatter between SiO2 and Vs than and Vp, but
Gibbs free energy minimization. Amphibolite and granulite facies lithologies
together they can be combined for a tighter constraint on composition than
either compressional or shear velocity alone. span similar Vp and Vs values, though the shapes of their distributions are

SAMMON ET AL. 6 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

marginally different. This is because the velocities of many amphibolite and

granulite facies minerals (e.g., garnet, pyroxene, plagioclase) have similar Vp
(∼7 km/s) and Vs (∼3.6 km/s) values. Despite considerable scatter in the Vs
data, when paired with Vp, a clear trend emerges: increasing SiO2 abundance
leads to decreasing velocities.

Higher Vp values correlate to lower silica content (Figures 6a and 6b). Higher
Vp/Vs ratios also have lower silica content, though for a given SiO2, there is
roughly a 10% spread in Vp/Vs. A slight curve in the amphibolite facies data
becomes more pronounced in the granulites, developing an arcuate shape
in the Vp/Vs vs. Vp plot. The same trends appear when analyzing Vp/Vs
vs. Vs (Figures 6c and 6d), though the data are more acutely curved. For
both amphibolite and granulite lithologies, increasing Vs can lead to either an
increase or a decrease in Vp/Vs ratio. The maximum Vp/Vs for amphibolite
facies lithologies at typical middle crustal P-T conditions, is expected at a Vs
of about 3.5–3.8 km/s, a Vp of 6.5–7 km/s, and SiO2 of 55 wt.%. For gran-
ulite, this maximum is expected at compositions closer to 60–63 wt.% SiO2.
Interestingly, the maximum Vp/Vs in granulite lithologies corresponds to the
lowest Vs rather than the highest Vp, suggesting that Vs variations exert a
stronger control on Vp/Vs ratios.
Figure 6. Perple_X-generated Vp/Vs plotted against SiO2 for (a) amphibolite
and granulite facies lithologies. Amphibolite results are shown for 450°C and 3.2. Deep Crustal Density
0.5 GPa. Granulite results are shown for 650°C and 0.8 GPa. Color indicates
relative data point density. We calculated deep crustal density by tracking the Vp and Vs values from
Perple_X that overlapped with our seismological database back to their
input samples. Then, we report the Perple_X-derived density of those input
samples. We found that the depth-dependent variations of deep crustal densities among the different tectonic
provinces appear more similar when normalized to crustal thickness (Figure 7). Uncertainties in density are ±3%
for each tectonic province, a result which is likely controlled by the ∼3% uncertainty on velocities. Deep crustal
density ranges from 2,700 to 2,780 kg/m 3 at 13 km depth to 3,290–3,340 kg/m 3 at the Moho.

We note that, in order to calculate deep crustal pressure, and thus mineralogy and composition, we already
assumed a bulk crustal density of 2,900 kg/m 3. This initial assumption, though, does not significantly affect
our composition results because there is, at most, a calculated pressure difference of <15% caused by using the
2,900 kg/m 3 a-priori density versus our model-generated density. This <15% pressure difference does not greatly
change the stable mineral assemblages or velocities calculated by Perple_X for the deep crust.

Figure 7. Calculated densities normalized to depth for (a) average continental crust and (b) different tectonic provinces with an imposed lithological transition
(amphibolite to granulite facies) at 2/3 total crust depth. By this method, the bottom ∼20%–30% of the crust can approach or exceed mantle density.

SAMMON ET AL. 7 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

The average densities and total masses of each tectonic province listed in Table 1 are calculated by this study.
The average thicknesses, total surface area, and total volume of each tectonic province are calculated from
Litho1.0's Moho and LAB depths, whose uncertainties are ∼12% (c.f. Pasyanos et al. (2014); Huang et al. (2013);
Dziewonski and Anderson (1981)). Continental Margin Moho depths are an exception and are from Szwillus
et al. (2019) (Section 2.1). The densities of the oceanic crust and lithospheric mantle, not calculated by this
study, were assumed to be 2,900 ± 116 kg/m 3 and 3,400 ± 136 kg/m 3, respectively (Carlson & Herrick, 1990;
Huang et al., 2013). The total volume of continental crust is 9.2 × 10 9 km 3. The continental crust, including
submerged continental margins, comprises 44% of Earth's surface area. Multiplying the density of each tectonic
province by its volume yields the mass of crust predicted in each tectonic province. We assumed one-third of the
crustal column was upper crust with a density of 2,650 kg/m 3 (Christensen & Mooney, 1995). The total mass
of the continental crust, 2.32 × 10 22 kg, is 4% higher than Cogley (1984); Albarède (1998); Rudnick (1995);
Wipperfurth et al. (2020).

3.3. Composition

Our main analysis focuses on SiO2 abundance and its uncertainties because of its strong correlation to seismic
velocities, particularly Vp. The SiO2 content at typical middle and lower crustal depth intervals (Figure 8) is
given in Table 2. All 9 major oxide inputs (SiO2, TiO2, Al2O3, CaO, MgO, FeOT, MnO, K2O, Na2O) can be
found in Table 3 and corresponding maps in Supporting Information S1. We use the notation “Mx%,” where x%
is the percent distance to the Moho (M) from the surface. This notation normalizes comparisons among tectonic
provinces of various thicknesses. We found that when we normalize depth in this manner, the composition versus
depth profiles for each province become more similar. The top of the deep crust starts at an intermediate composi-
tion, globally ranging from 58 to 68 wt.% SiO2, and the composition gradually decreases to 50–55 wt.% SiO2 as it
approaches the Moho. Global scale SiO2 composition of the continental crust mostly decreases (or remains stead-
ily mafic) with increasing depth for all tectonic provinces (Figure 9). Uncertainty in global SiO2 also decreases
with increasing depth due to fewer samples fitting the seismic signal at depth. However, the Andean and Himala-
yan tectonic provinces have larger uncertainties at depth because of the variation in geochemical data fitting the
seismic signal and the sparsity of deep seismological profiles.

The CaO content of the deep crust is also of interest due to its absolute abundance and significance as a contribu-
tor to sedimentary deposits, though only siliciclastic rocks and not carbonates were considered viable deep crustal
components (Hartmann et al., 2012; Wilkinson et al., 2009). In our model, Ca is mostly contained in plagioclase,
pyroxene, and garnet. The CaO abundance tends to increase with depth because of the increasingly mafic nature
of the deep crust, and therefore regions of low SiO2 correlate with regions of high CaO. Globally, the median CaO
at crustal depths of M85% is 9.1 ± 3.1 wt.% (Figure 10).

We can also derive the global distribution of a trace element if that trace element has a quantifiable relationship
to one of the thermodynamic components (major oxides) used in our model. We used a geochemical database
of samples with both major and trace element concentrations (Sammon & McDonough, 2021) to generate trace
element maps as a function of major oxide abundances. We used a bivariate probability analysis to generate trace
element distributions from a major oxide abundance, such as SiO2, at a specific depth or location. Although
we suggest using regional analyses for high resolution interpretations of trace element abundances, we present
here global predictions and uncertainties for Sr (Figure 11) and U (Figure 12) content based on their relationships
with CaO and SiO2, respectively, as examples. Global average Sr increases with increasing CaO until plagioclase,
whether due to compositional changes or increased pressures, is no longer the dominant Ca-bearing mineral.
Uncertainties on global U concentration span an order of magnitude because the abundance of U in a given
metamorphic sample set ranges from a few hundreds of ppb to a few ppm. U and SiO2 abundances, however, are
positively correlated, with median U increasing as median SiO2 increases.

4. Discussion
4.1. SiO2 and Overall Deep Crustal Composition

Figure 9 shows steady or decreasing SiO2 with increasing depth in the deep crust. Figure 8 also makes it appar-
ent, though, that the absolute SiO2 at a given depth is not equal across different crustal types. For example,
“Extended” crust appears mafic at 30 km depth while the “Thick Himalayan” crust is felsic at that depth, and

SAMMON ET AL. 8 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Figure 8. Global SiO2 composition at a depth of 30 km shows regional distinctions, whereas measuring composition at
a crustal depth relative to the Moho (M85% notation = 85% of the total crustal depth) produces a view of a deep crust that
is relatively more homogeneous, with SiO2 decreasing gradually with depth. Areas of high projected SiO2 include the
Himalayas, Andes, East African rift, and some continental margins. While the apparently more felsic nature Himalayas and
Andes may be a genuine compositional feature, the high SiO2 in some rifts and continental margins are likely from model
input inaccuracies.

“Proterozoic” crust falls in between. However, a more laterally consistent trend appears when comparing percent
of the crustal column traversed rather than absolute depth (Figure 8). Most regions show a 5–10 wt.% decrease in
median SiO2 through the deep crust regardless of crustal thickness, so that SiO2 decreases more rapidly in areas
of thin crust than in areas of thick crust. We predict the global median SiO2 at 50% above the Moho (or, alterna-
tively, 50% crustal column thickness) to be 61.2 ± 7.3 wt.% SiO2 with CIPW normative mineralogy of <10 wt.%
alkali feldspar and <15 wt.% quartz. The average composition of middle continental crust is therefore expected
to resemble that of a quartz monzonite; the lower crust average composition, with 53.3 ± 4.8 wt.% SiO2 and
9.1 ± 3.1 wt.% CaO, is expected to resemble a gabbronorite. It must be noted, however, that these compositions
can be the result of seismic velocity profiles averaging felsic and mafic endmembers of the deep crust, and need
not be produced by a monzonite or norite.

The processes of magmatic differentiation and density sorting could produce the compositional structure of the
continental crust. The process of crustal genesis leaves mafic, restitic material at the base of the crust. Magmatic

SAMMON ET AL. 9 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Table 2 differentiation can explain the composition correlation with % depth in the
Median SiO2 in wt.% for Different Tectonic Provinces crust on both the thin and the thicker scales. More buoyant, felsic material
ascends to the top of the crust, producing a gradient of SiO2 that scales with
SiO2 at SiO2 at
M50% Uncertainty M85% Uncertainty crustal thickness. Alternatively, the deep crust could be more mafic because
it is simply closer to the mantle and therefore has a greater number of mafic
(∼middle (∼lower
intrusions. Our results do not indicate any need for sharp compositional
crust) ± crust) ±
boundaries in the deep crust. The MX% notation reinforces the importance
Continental margin 61.1 10.2 52.8 2.7 of scaled, relative depth in the crust rather than absolute depth for making
Archean 61.0 7.8 51.9 2.9 compositional comparisons.
Proterozoic 62.0 7.3 52.9 3.7
Two regions that appear conspicuously more felsic than the global deep
Phanerozoic 61.5 9.2 51.8 2.5
crustal median are the Andes and the Thin Himalayan crust (Figure 9). A
Platform 58.9 7.6 51.8 2.7 low temperature gradient could be the cause of the compositional difference
Extended 68.9 7.6 52.7 2.4 between these two tectonic provinces an the global average, but we also
Rifted 66.8 7.3 57.8 6.5 must consider two other possibilities, particularly around the northern and
Arc 68.7 9.8 57.6 6.9 northeastern Tibetan Plateau and Himalayan ramp. The first is that thick,
convergent margins, especially in the Himalayas, might have layers of upper
Paleo-orogenic 63.9 9.2 52.7 3.4
crustal material thrust deeper within the crust, incorporating more felsic
Tethyan 63.9 9.4 52.3 3.1
material into the deep crust. In contrast, underthrust upper crustal material
Andean 59.9 9.1 56.0 6.9 is less likely to appear in the Andes, which is a continent-ocean subduc-
Thick crust 68.2 8.8 58.4 8.6 tion zone. Alternatively, pockets of melt and partially melted material in
Thin Himalayan 70.7 5.9 51.3 2.2 the Andean and Himalayan middle and lower crust could reduce the shear
wave velocity (Caldwell et al., 2009; Hacker et al., 2014; Nelson et al., 1996;
Regis et al., 2016; Schilling & Partzsch, 2001; Schmitz et al., 1997; Searle
et al., 2009). Because our current model does not factor in melt, slower Vs
speeds would be attributed to a more felsic composition instead.

Other anomalous regions in Figure 8, particularly the continental margins of Antarctica, the East African rift
zone, and the Sea of Japan, are likely caused by inaccurate temperature and Moho inputs. The East African
Rift could appear felsic because the model's temperature gradient for that actively rifting region is too low; a
cooler felsic composition can produce the same velocities as a warmer mafic composition. On the other hand,
the highly localized, extremely felsic borders around Antarctica and between Japan and China likely indicate a
misclassification of crust type and/or Moho depth. Oceanic-type crust has been documented in both regions (Cho
et al., 2004; Gohl, 2008; Hirata et al., 1992; McCarthy et al., 2020). Better Moho and temperature resolution of
the ocean-continent transition should increase the accuracy of compositional models in these regions.

Mafic granulite lithologies reach gravitational instability in the lower 10%–

20% of the average crustal column (Jagoutz et al., 2011), surpassing the upper
mantle's density of 3,300–3,400 kg/m 3. Therefore, according to Figure 7,
most of the granulite facies lower crust for continental margins, Andean crust,
Table 3
Middle and Lower Crust Bulk Composition in wt.% Tethyan crust, and Phanerozoic crust should be gravitationally unstable. On
the other hand, most other tectonic provinces would just reach mantle-like
Composition at Composition at
M50% Uncertainty M85% Uncertainty
densities around the Moho depths. Thinner Himalayan type crust has a middle
crustal density ∼9% lower than other provinces, correlating with negative
(∼middle crust) ± (∼lower crust) ± seismic velocity anomalies. Arcs have the next lowest densities on average,
SiO2 61.2 7.30 53.3 4.8 suggesting that a potential denser lower crust beneath some arcs has already
TiO2 0.71 0.40 0.85 0.40 foundered (Jagoutz et al., 2011). Figure 7b shows that accreted arcs of Andean
Al2O3 16.1 1.8 16.8 3.0 type crust in particular display a stark decrease in density interpreted as being
due to delamination of the lowermost crust (Ducea, 2011; Gao et al., 2021;
FeO 6.35 3.43 8.75 2.32
Kay & Kay, 1993).
MnO 0.11 0.06 0.15 0.06
MgO 2.81 2.03 4.93 1.91 Forming continental crust via island arc processes, however, would then
require the deep crust to become denser over time, since most of our tectonic
CaO 4.98 2.79 7.76 3.38
provinces have a lower crust that is denser than arc lower crust. This can be
Na2O 3.79 0.80 2.75 0.95
achieved by cooling the crust, thickening it further, intra-crustal differenti-
K2O 1.47 1.29 0.96 1.09 ation, or by mafic igneous injections into the lower crust. Moreover, if our

SAMMON ET AL. 10 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Moho temperature model is too hot, though, it requires denser, more mafic
lower crust than is actually present to explain the observed Vp and Vs values.
As such, our compositional models are intrinsically tied to Moho temper-
ature, and may be skewed toward mafic granulites. Reducing the assumed
crustal Moho temperatures would bring the estimated average crustal density
closer to arc crust density and increase the average weight percentage of SiO2.

There is a tradeoff between temperature and composition. Vp and Vs both

carry a temperature dependency through their bulk and shear moduli,
so accurate temperature estimates are imperative for modeling the crust;
decreased seismic velocities can be the result of either higher temperatures or
greater SiO2 content. As a reminder, this study implements a linear temper-
ature gradient through the crust from the TC15 global temperature model
(Artemieva, 2006).

Table 4 reports compositional models for the middle and lower continental
crust, a practice that is required to make meaningful comparisons to previous
crustal models. While we recognize the assumption of a three-layer crust
as an oversimplification of the diversity of crustal compositions, it is useful
for some calculations to have average composition numbers for the crust.
A few examples include mantle tomography studies which require crustal
correction, crustal corrections for geoneutrino studies; models of Earth's
thermal history; and planetary scale compositional model for comparison
with other rocky bodies. Our middle crustal composition falls between two
possible compositions given by Hacker et al. (2015): the fastest Vp endmem-
ber composition for the middle crust (62.7 wt.% SiO2), and the middle crus-
tal composition expected when the crust takes on a two compositional layer
(upper and lower) structure, instead of three, (57.3 wt.% SiO2). These SiO2
estimates also overlap with the 62 wt.% SiO2 reported by Christensen and
Figure 9. Global SiO2 decreases with increasing depth from the middle to the Mooney (1995) and the 63.5 wt.% SiO2 middle crust reported by Rudnick and
bottom of the continental crust. The middle crust M50% ranges from 60 to 65 Gao (2014). Our proposed lower crust composition is agrees with Rudnick
wt.% SiO2 in most areas and increases at a rate of about 1 wt.% per km until
and Gao (2014) and other mafic models (e.g., Hacker et al. (2015)'s fast Vp
reaching the base of the crust. Uncertainties can be found in Figure S19 in
Supporting Information S1. lower crust; Jagoutz and Schmidt (2012)). Models which predict a more
intermediate-felsic lower crust, such as the North China craton lower crustal
model of Liu et al. (2001) or the higher SiO2, lower Vp options listed by
Hacker et al. (2015), are not consistent with our global average, though isolated regions of more felsic lower crust
may exist.

4.2. CaO and Sr

Bulk CaO concentration increases with depth (Figure 10), but as a component of mafic, siliciclastic rocks, not
carbonate. The lack of carbonate is due, in part, to our imposed amphibolite/granulite grade lithology restric-
tions on possible deep crust composition, but is reinforced by high density and Vp values observed in the deep
crust. Carbonates, with densities of approximately 2,750 kg/m 3 and Vp's of 6.6–6.8 km/s at deep crustal condi-
tions (Christensen & Mooney, 1995), cannot substantially contribute to the observed deep crustal velocities.
Also, there are few carbonate-dominated granulite facies xenoliths and terrains as compared to silicate granulites
(Hartmann et al., 2012; Wilkinson et al., 2009). A comparison of Figures 8b and 10 shows tight correlation glob-
ally between regions of high SiO2 and low CaO (see Figure S26 in Supporting Information S1). Uncertainties in
CaO estimates follow the same trends as SiO2 uncertainties, though the former's uncertainty is roughly 10%–20%
higher. The relationship between Vp or Vs and CaO content has more scatter than with SiO2, which increases
the % uncertainty on CaO estimates.

Sr abundances cannot be directly derived from velocity calculations, but they can be predicted from Sr's geochem-
ical relationship with CaO or Na2O. In amphibolites and granulites, the CaO content tracks predictably with Sr
concentration until CaO abundance reaches 5–6 wt.% (Figure S28 in Supporting Information S1). However, once
CaO exceeds 6 wt.%, Sr abundances plateau and scatter increases. This shape of the Sr versus CaO distribution

SAMMON ET AL. 11 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

correlates with Perple_X-calculated anorthite number. Sr abundance depends

on both CaO and Na2O, and a Sr map can be predicted from either. Figure 11
shows Sr abundances derived from a bivariate probability analysis between
Sr and CaO. The extremely low (dark blue) Sr abundances in thick regions
of the deep crust could be a result of low CaO abundances overall and/or the
production of pyroxene and garnet at the expense of Ca-rich plagioclase. A
subtler drop in Sr is seen in these areas when predicting Sr from Na2O abun-
dance (Figure S33 in Supporting Information S1), possibly because regions
of thick crust maintain relatively constant amounts of Na2O (and Na-rich
plagioclase). While Figure 11 appears to show abrupt changes in median Sr
content, note the 50% uncertainty. The variable Sr median abundances shown
on the map are a result of the scatter in the Sr-CaO relationship. On the
contrary, a Sr abundance map derived from Na2O (Figure S33 in Supporting
Information S1) has lower apparent resolution because of the high percent
uncertainty on Na2O, but less overall uncertainty because of the tighter
correlation between Na2O and Sr. These findings highlight the usefulness
of being able to constrain a trace element from relationships with multiple
major oxides, but also stress the need for a higher resolution understanding of
crustal composition and mineralogy in unique tectonic provinces.

Figure 10. Global CaO abundance and uncertainty at 85% of the total crustal
4.3. Heat Production and Moho Heat Flux
depth. Areas of low CaO correlate to areas of high SiO2. There does not appear Radioactive elements produce heat inside Earth, with K, Th, and U, the heat
to be any correlation between CaO content and uncertainty, with most regions
having 3–4 wt.% uncertainty regardless of CaO abundance.
producing elements (HPEs), accounting for 99.5% of the total radiogenic
energy output (McDonough et al., 2020). The few remaining radioactive
elements in the planet contribute <0.5% to Earth's total heat production. We
can model the abundance and distribution of HPEs in the crust, and Moho
heat flux, with our model and supplementary geochemical relationships.

HPE abundance data are provided in the (Sammon & McDonough, 2021)
geochemical data set. Using a joint probability analysis, we derived U
abundances with uncertainties from SiO2. The middle continental crust
has a median U concentration of 1.3 ± 0.2 ppm and the lower crust has
0.22 ± 0.03 ppm. The uncertainty on these median values is reported as the
standard error of the median. We do this to highlight that this is the confi-
dence in the central value, the median. The full range of possible abundances
spans five orders of magnitude (Figure 12b) Th abundances could also be
generated with a joint Th-SiO2 probability analysis, but we use a constant
Th/Umass ratio of 3.77 ± 0.1 (Sammon & McDonough, 2021; Wipperfurth
et al., 2018) which provides a narrower constraint on Th concentration. K2O
abundance is calculated directly as a thermodynamic component in our
compositional model. We calculate heat production on the same 1°latitude
by 1°longitude cell scale that we use for calculating major element abun-
dances. The concentrations of K, Th, and U in each cell are multiplied by
each cell's density (calculated in previous section) and volume. The resulting
masses of K, Th, and U are then multiplied by the heat produced during their
respective decays (McDonough et al., 2020), resulting in a heat production
value for each cell in units of W/kg. Heat production values can be converted
to W/m 3 via multiplying by cell density once more. Uncertainties are calcu-
lated through standard propagation of errors on concentration and density.
Figure 13 maps the total (depth-integrated) median contribution to surface
heat flux from the upper, middle, and lower continental crust in mW/m 2.

Figure 11. Global Sr abundance and uncertainty derived from a joint We calculated a median heat production of 0.3 nW/kg (0.8 μ W/m 3) for
probability analysis with CaO at 50% of the total crustal depth. the middle third of the crust and 0.05 nW/kg (0.1 μ W/m 3) for the lower

SAMMON ET AL. 12 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Figure 12. Global U abundance derived from a joint probability analysis with SiO2 at 85% of the total crustal depth.
Uncertainties span orders of magnitude because of the range of possible U values, but the global median at this depth is
∼0.2 ppm U. Regions of high SiO2, especially the potentially inaccurate continental margin of Antarctica, correlate with high
U abundances and the highest uncertainties.

third of the crust. Previous studies have also found low heat production in the deep continental crust (Fountain
et al., 1987; Jaupart et al., 2016; Kukkonen et al., 1997). Our model is also consistent with local studies of
HPE analyses of deep crustal xenoliths, such as Gruber et al. (2021); Pinet and Jaupart (1987); and Ashwal
et al. (1987).

The uncertainties on this global model are dominated by uncertainties on U abundances, which span an order of
magnitude. Even so, our uncertainty on the median or central value (standard error of the median) of HPEs or
heat production is well constrained at ± 0.1%. While the range of possible heat production values span an order
of magnitude, the median/average heat production value is tightly constrained. Provinces with high predicted
SiO2 content, such as the Andes, have estimated U content up to four times higher than the global lower crustal
median because of the correlation between high SiO2 samples and high U. We recommend using regional HPE
data for understanding smaller scale variations and reserve this study's results for continent- or global-scale
models.

We took our calculations a step further by estimating Moho heat flux. Once we determined deep crustal heat
production, we calculated a subcontinental Moho heat flux by subtracting continental heat production from
the global surface heat flux of Lucazeau (2019), supplemented with Shen et al. (2020)'s Antarctica surface

SAMMON ET AL. 13 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Table 4
Continental Crust Composition Estimates
Christensen & Liu et al., 2001 Jagoutz & Rudnick & Gao, 2014 Hacker et al., 2015a Hacker et al., 2015b This
Mooney, 1995 Schmidt, 2012 Study
Middle crust
SiO2 62 – – 63.5 62.7 57.3 61.2
TiO2 – – – 0.69 0.8 0.99 0.71
Al2O3 – – – 15 15.7 16.8 16.1
FeOT – – – 6.02 6.76 8.15 6.35
MnO – – – 0.10 0.13 0.16 0.11
MgO – – – 3.59 3.51 4.46 2.81
CaO – – – 5.25 5.27 6.63 4.98
Na2O – – – 3.39 3.42 3.89 3.79
K2O – – – 2.3 1.6 1.42 1.47
Mg# – – – 51.5 48.1 43.4 44.1
Lower crust
SiO2 52 58.3 52.2 53.4 50.7 57.3 53.3
TiO2 – 0.59 0.78 0.82 1.24 0.99 0.85
Al2O3 – 13.6 18.7 16.9 16.5 16.8 16.8
FeOT – 5.32 8.41 8.57 10.39 8.15 8.75
MnO – 0.08 0.17 0.10 0.19 0.16 0.15
MgO – 9.58 5.86 7.24 7.03 4.46 4.93
CaO – 4.54 10.8 9.59 10.1 6.63 8.06
Na2O – 2.54 2.56 2.65 2.8 3.89 2.75
K2O – 3.23 0.41 0.61 0.79 1.42 0.96
Mg# – 76.2 55.4 60.1 54.7 49.4 50.1
a
Hacker et al. (2015) fast Vp crustal model. Hacker et al. (2015) middle crust composition = lower crust composition model.
b

heat flux. We assumed a Gaschnig et al. (2016) upper crustal heat production of 1.75 μ W/m 3 to complete
the initial calculation. Pairing an upper crustal composition of Gaschnig et al. (2016) with our deep crustal
composition yielded Moho heat fluxes for tectonically stable regions that agree with Jaupart et al. (2007) and
marginally agree with (Hacker et al., 2015)'s models of crust with slow Vp. However, Moho heat flux calcu-
lations depend substantially on the assumed HPE abundance model for the upper crust, which contributes
∼70% of the total crustal heat production in most regions. Using Gaschnig et al. (2016)'s upper crustal U and
Th abundances in low heat flux, cratonic regions, results in roughly 6% (by area) of the continents having a
negative heat flux across the Moho (an unreasonable condition) – or more likely, other factors, such as heat
dissipation through fluid circulation in the near surface, would be needed to explain these low surface heat
flux (e.g., 20–40 mW/m 2) regions. Alternatively, the assumed upper crustal heat production values may need
to be lowered. Most low heat flux areas coincide with stable cratonic lithosphere, where low heat flux and low
heat production is not a new observation (e.g., Nyblade and Pollack (1993); Kukkonen et al. (1997); Jaupart
et al. (2007); Cammarano and Guerri (2017)). Estimates of upper crustal heat production in cratonic regions
range from 0.6 to 1 μ W/m 3 (Gruber et al., 2021; Jaupart et al., 2014, 2016; Mareschal & Jaupart, 2013; Phaneuf
& Mareschal, 2014), so we approximated Archean upper crustal heat production at 0.8 μ W/m 3, which is the
maximum permissible heat production value found by Rudnick and Nyblade (1999) for the Kalahari craton and
also the maximum average crustal heat production expected for crust ≥2.5 Ga (Jaupart et al., 2016). Our final
model for Moho heat flux uses the data and resources listed in Table 5. The median global continental Moho
heat flux, shown in Figure 𝐴𝐴 14, is 21.6+16.0
−5.6
mW/m 2. Figure 15 summarizes the distribution of Moho heat flux
values for each tectonic province.

SAMMON ET AL. 14 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Figure 13. Panels (a–c) show the contribution from each crustal layer (upper, middle, lower) to the total continental surface heat flux (D). Continental crustal heat
production is dominated by upper crustal heat production but with significant contributions from the middle crust.

Table 6 compares our average crustal heat production and Moho heat flux values to the average values calculated
from other studies. We calculated the Moho heat fluxes from the previous studies by taking the average surface
heat flux and subtracting the crust's average radiogenic contribution. For consistency, we recalculated our study's
Moho heat flux in this simplified manner as well. Total crustal HPE concentrations average upper, middle, and
lower crustal abundances. Figure 16 compares the different models of U abundance in the bulk continental crust
from Table 6 to their corresponding Moho heat flux values.
Notably, all models reported in Table 6 assume a similar compositional model for the upper crust (Gaschnig
et al., 2016; Rudnick & Gao, 2014). A 1% change in upper crustal U has eight times the effect on the Moho heat
flux as a 1% change in lower crustal U content. Thus, having an accurate model for the upper crust is of para-
mount importance to determining Moho heat flux. Only after upper crustal heat production is constrained will the
uncertainties on deep crustal heat production impact Moho heat flux calculations.

Table 5
Heat Production Calculation Parameters
Parameter Value
Global surface heat flux Inputs from Lucazeau (2019)
Antarctica surface heat flux Inputs from Shen et al. (2020)
Upper crust heat production 1.75 μW/m 3 (Gaschnig et al., 2016)
Upper crust heat production (cratonic) 0.8 μW/m 3 (see discussion for sources)
Deep crustal densities Inputs from this study

SAMMON ET AL. 15 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Figure 14. Global heat flux across the Moho is calculated by subtracting crustal heat production from measurements of
surface heat flux. The median subcontinental Moho heat 𝐴𝐴
flux is 21.6+16.0
−5.6
mW/m 2 globally. This result is sensitive to the
assumption of a uniform upper crustal heat production of 0.8 μW/m 3 for cratonic and 1.65 μW/m 3 for non-cratonic regions.

Figure 15. The median global Moho heat 𝐴𝐴 flux is 21.6+16.0

−5.6
mW/m 2, shown by the blue shaded area at the bottom of the graph.
Global surface heat 𝐴𝐴
flux is 63.8+17.4
−6.8
mW/m 2
(Lucazeau, 2019), indicated by the red shaded band toward the top. Although
variations exist among the tectonic provinces, they show remarkably consistent central heat flux values. These statistics were
produced by calculating the median heat flux for each latitude × longitude cell. Outliers are not shown on the graphs. The
white line in the middle of each box and whisker plot is the median value. The calculated value written above the box and
whisker plot in blue is for the Moho heat flux and in red for the surface heat flux for each province.

SAMMON ET AL. 16 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Table 6
Heat Production in the Continental Crust
Models of the continental crust

RG13 HKB11 HKB15 low HKB15 high WSM20 This study

K (ppm) 15000 15800 11900 18800 11700 15400
Th (ppm) 5.60 7.30 4.20 5.31 5.64 5.40
U (ppm) 1.30 1.40 1.09 1.33 1.27 1.40
Heat production K (nW/kg) 0.0508 0.0535 0.0403 0.0637 0.0396 0.0522
Heat production Th (nW/kg) 0.147 0.191 0.110 0.139 0.148 0.141
Heat production U (nW/kg) 0.128 0.138 0.107 0.131 0.125 0.138
Total crustal heat production (nW/kg) 0.326 0.383 0.258 0.334 0.312 0.331
Total crustal heat production (mW/m 3) 0.945 1.111 0.748 0.969 0.905 0.960
Crustal radiogenic heat production (TW) 7.56 8.89 5.99 7.75 7.24 7.68
Upper crust contribution to total heat production 59.4% 50.5% 75.0% 57.9% 62.0% 58.4%
Radiogenic contribution to surface heat flux (mW/m 2) 33.7 39.6 26.7 34.6 32.3 34.2
Average crustal Moho heat flux (mW/m 2) 29.3 23.4 36.3 28.4 30.7 28.9
Note. Data sources: RG13 = (Rudnick & Gao, 2014), HKB11 = (Hacker et al., 2011), HKB15 = (Hacker et al., 2015), WSM20 = (Wipperfurth et al., 2020), Šrámek
et al. 13 = (Šrámek et al., 2013). Moho heat flux is calculated by taking the average surface heat flux for the continents 63 mW/m 2 (and the oceans 65 mW/m 2)
(Lucazeau, 2019) and subtracting the continental crust's (and ocean) radiogenic contribution. Upper crust contribution is calculated using the upper crust composition
from Rudnick and Gao (2014) relative to the total crust production; note all models assume this same upper crust composition.

5. Conclusions
We have constructed a global compositional model of the deep continental crust by synthesizing seismic, temper-
ature, heat flux, and geochemical data. We predict deep crustal compositions on the global scale using major
and trace element compositions from amphibolite and granulite facies lithologies, and seismic velocity profiles.
Our proposed global compositional model reconciles the USGS Global Seismic Structure Catalog of crustal
velocities, compositions for thousands of medium and high grade metamorphic rocks, and constraints on Moho
depth (Pasyanos et al., 2014; Szwillus et al., 2019), crust temperature (Artemieva, 2006), and surface heat flux
(Lucazeau, 2019; Shen et al., 2020).

Vp values, and to a lesser extent Vs and Vp/Vs values, correlate with bulk
rock SiO2 content, and SiO2 can be used as a predictor of seismic velocities
if temperature can be estimated accurately. Globally, SiO2 concentrations
tend to decrease with increasing depth, leading to a predominantly mafic
to intermediate-mafic base of the crust. The decreased density and less
mafic nature of the lower crust in younger and tectonically active crust,
such as arcs and active mountain ranges, suggests that those tectonic prov-
inces are hotter than our temperature model predicts, that they have under-
gone lower crustal delamination, or both. Global median SiO2 concentra-
tions for the middle and lower crust are 61.2 ± 7.31 and 53.3 ± 4.8 wt.%,
respectively, though steady composition and velocity gradients in the deep
crust urge us to embrace a less distinctly layered view of the crust. This
mid-to-deep crustal gradient in wt.% SiO2 is the equivalent of a litholog-
ical gradient ranging from quartz monzonite to gabbronorite. We predict
the abundances of multiple major oxides, many of which are correlated to
Figure 16. There is a linear dependency between crustal U concentration
and calculated Moho heat flux. When the different models of Table 6 are
trace element abundances. This correlation allows us to derive expected
compared, differences in each models' predicted Moho heat flux are mostly the heat production in the deep crust. We also predict a Moho heat flux of
result of differences in upper crustal heat producing element concentrations.𝐴𝐴 21.6+16.0
−5.6
mW/m 2.

SAMMON ET AL. 17 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Data Availability Statement

Data and modeling software can be accessed at https://doi.org/10.5281/zenodo.6967908.

Acknowledgments References
We gratefully acknowledge the support by
NSF grants EAR1650365 and 2050374 to Albarède, F. (1998). The growth of continental crust. Tectonophysics, 296(1–2), 1–14. https://doi.org/10.1016/s0040-1951(98)00133-4
WFM and support from the United States Artemieva, I. M. (2006). Global 1°×1° thermal model TC1 for the continental lithosphere: Implications for lithosphere secular evolution. Tectono-
Geological Survey Earthquake Hazards physics, 416(1–4), 245–277. https://doi.org/10.1016/j.tecto.2005.11.022
Program to W. D. Mooney. We would Ashwal, L., Morgan, P., Kelley, S., & Percival, J. (1987). Heat production in an Archean crustal profile and implications for heat flow and mobi-
like to thank our insightful and thorough lization of heat-producing elements. Earth and Planetary Science Letters, 85(4), 439–450. https://doi.org/10.1016/0012-821x(87)90139-7
reviewers, Derrick Hasterok and Joshua Caldwell, W. B., Klemperer, S. L., Rai, S. S., & Lawrence, J. F. (2009). Partial melt in the upper-middle crust of the northwest Himalaya revealed
Garber for their suggestions and improve- by Rayleigh wave dispersion. Tectonophysics, 477(1–2), 58–65. https://doi.org/10.1016/j.tecto.2009.01.013
ments on this manuscript. We also thank Cammarano, F., & Guerri, M. (2017). Global thermal models of the lithosphere. Geophysical Journal International, 210(1), 56–72. https://doi.
Wolfgang Szwillus for his insights on heat org/10.1093/gji/ggx144
flux modeling. Any use of trade, firm, or Carlson, R. L., & Herrick, C. N. (1990). Densities and porosities in the oceanic crust and their variations with depth and age. Journal of Geophys-
product names is for descriptive purposes ical Research, 95(B6), 9153–9170. https://doi.org/10.1029/jb095ib06p09153
only and does not imply endorsement by Cho, H.-M., Kim, H.-J., Jou, H.-T., Hong, J.-K., & Baag, C.-E. (2004). Transition from rifted continental to oceanic crust at the southeastern
the U.S. Government Korean margin in the east sea (Japan sea). Geophysical Research Letters, 31(7), L07606. https://doi.org/10.1029/2003gl019107
Christensen, N. I., & Mooney, W. D. (1995). Seismic velocity structure and composition of the continental crust: A global view. Journal of
Geophysical Research, 100(B6), 9761–9788. https://doi.org/10.1029/95jb00259
Cogley, J. G. (1984). Continental margins and the extent and number of the continents. Reviews of Geophysics, 22(2), 101–122. https://doi.
org/10.1029/rg022i002p00101
Connolly, J. A. D. (2005). Computation of phase equilibria by linear programming: A tool for geodynamic modeling and its application to subduc-
tion zone decarbonation. Earth and Planetary Science Letters, 236(1), 524–541. https://doi.org/10.1016/j.epsl.2005.04.033
Ducea, M. N. (2011). Fingerprinting orogenic delamination. Geology, 39(2), 191–192. https://doi.org/10.1130/focus022011.1
Dumond, G., Williams, M. L., & Regan, S. P. (2018). The Athabasca granulite terrane and evidence for dynamic behavior of lower continental
crust. Annual Review of Earth and Planetary Sciences, 46(1), 353–386. https://doi.org/10.1146/annurev-earth-063016-020625
Dziewonski, A. M., & Anderson, D. L. (1981). Preliminary reference Earth model. Physics of the Earth and Planetary Interiors, 25(4), 297–356.
https://doi.org/10.1016/0031-9201(81)90046-7
Fountain, D. M., Furlong, K. P., & Salisbury, M. H. (1987). A heat production model of a shield area and its implications for the heat flow-heat
production relationship. Geophysical Research Letters, 14(3), 283–286. https://doi.org/10.1029/gl014i003p00283
Gao, Y., Tilmann, F., van Herwaarden, D.-P., Thrastarson, S., Fichtner, A., Heit, B., et al. (2021). Full waveform inversion beneath the central
Andes: Insight into the dehydration of the Nazca slab and delamination of the back-arc lithosphere. Journal of Geophysical Research: Solid
Earth, 126(7), e2021JB021984. https://doi.org/10.1029/2021JB021984
Gaschnig, R. M., Rudnick, R. L., McDonough, W. F., Kaufman, A. J., Valley, J. W., Hu, Z., et al. (2016). Compositional evolution of the upper
continental crust through time, as constrained by ancient glacial diamictites. Geochimica et Cosmochimica Acta, 186, 316–343. https://doi.
org/10.1016/j.gca.2016.03.020
Gohl, K. (2008). Antarctica's continent-ocean transitions: Consequences for tectonic reconstructions. In Antarctica: A keystone in a changing
world: Proceedings of the 10th international symposium on antarctic earth sciences (pp. 29–38).
Gruber, B., Chacko, T., Pearson, D. G., Currie, C., & Menzies, A. (2021). Heat production and Moho temperatures in cratonic crust: Evidence
from lower crustal xenoliths from the slave craton. Lithos, 380, 105889. https://doi.org/10.1016/j.lithos.2020.105889
Hacker, B. R., Kelemen, P. B., & Behn, M. D. (2011). Differentiation of the continental crust by relamination. Earth and Planetary Science
Letters, 307(3–4), 501–516. https://doi.org/10.1016/j.epsl.2011.05.024
Hacker, B. R., Kelemen, P. B., & Behn, M. D. (2015). Continental lower crust. Annual Review of Earth and Planetary Sciences, 43(1), 167–205.
https://doi.org/10.1146/annurev-earth-050212-124117
Hacker, B. R., Ritzwoller, M. H., & Xie, J. (2014). Partially melted, mica-bearing crust in Central Tibet. Tectonics, 33(7), 1408–1424. https://
doi.org/10.1002/2014tc003545
Hartmann, J., Dürr, H. H., Moosdorf, N., Meybeck, M., & Kempe, S. (2012). The geochemical composition of the terrestrial surface (without
soils) and comparison with the upper continental crust. International Journal of Earth Sciences, 101(1), 365–376. https://doi.org/10.1007/
s00531-010-0635-x
Hirata, N., Karp, B. Y., Yamaguchi, T., Kanazawa, T., Suyehiro, K., Kasahara, J., et al. (1992). Oceanic crust in the Japan basin of the Japan sea
by the 1990 Japan-USSR Expedition. Geophysical Research Letters, 19(20), 2027–2030. https://doi.org/10.1029/92gl02094
Huang, Y., Chubakov, V., Mantovani, F., Rudnick, R. L., & McDonough, W. F. (2013). A reference Earth model for the heat-producing elements
and associated geoneutrino flux. Geochemistry, Geophysics, Geosystems, 14(6), 2003–2029. https://doi.org/10.1002/ggge.20129
Jagoutz, O., Müntener, O., Schmidt, M. W., & Burg, J.-P. (2011). The roles of flux-and decompression melting and their respective fractionation
lines for continental crust formation: Evidence from the Kohistan arc. Earth and Planetary Science Letters, 303(1–2), 25–36. https://doi.
org/10.1016/j.epsl.2010.12.017
Jagoutz, O., & Schmidt, M. W. (2012). The formation and bulk composition of modern juvenile continental crust: The Kohistan arc. Chemical
Geology, 298, 79–96. https://doi.org/10.1016/j.chemgeo.2011.10.022
Jaupart, C., Mareschal, J., & Watts, A. (2007). Heat flow and thermal structure of the lithosphere. Treatise on geophysics, 6, 217–252. https://doi.
org/10.1016/b978-044452748-6.00104-8
Jaupart, C., Mareschal, J.-C., Bouquerel, H., & Phaneuf, C. (2014). The building and stabilization of an Archean Craton in the Supe-
rior Province, Canada, from a heat flow perspective. Journal of Geophysical Research: Solid Earth, 119(12), 9130–9155. https://doi.
org/10.1002/2014jb011018
Jaupart, C., Mareschal, J.-C., & Iarotsky, L. (2016). Radiogenic heat production in the continental crust. Lithos, 262, 398–427. https://doi.
org/10.1016/j.lithos.2016.07.017
Javoy, M., Kaminski, E., Guyot, F., Andrault, D., Sanloup, C., Moreira, M., et al. (2010). The chemical composition of the Earth: Enstatite chon-
drite models. Earth and Planetary Science Letters, 293(3), 259–268. https://doi.org/10.1016/j.epsl.2010.02.033

SAMMON ET AL. 18 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Kay, R. W., & Kay, S. M. (1993). Delamination and delamination magmatism. Tectonophysics, 219(1–3), 177–189. https://doi.
org/10.1016/0040-1951(93)90295-u
Kukkonen, I. T., Golovanova, I. V., Khachay, Y. V., Druzhinin, V., Kosarev, A. M., & Schapov, V. A. (1997). Low geothermal heat flow
of the Urals fold belt—Implication of low heat production, fluid circulation or palaeoclimate? Tectonophysics, 276(1–4), 63–85.
https://doi.org/10.1016/s0040-1951(97)00048-6
Liu, Y.-S., Gao, S., Jin, S.-Y., Hu, S.-H., Sun, M., Zhao, Z.-B., & Feng, J.-L. (2001). Geochemistry of lower crustal xenoliths from Neogene Hann-
uoba Basalt, North China craton: Implications for petrogenesis and lower crustal composition. Geochimica et Cosmochimica Acta, 65(15),
2589–2604. https://doi.org/10.1016/s0016-7037(01)00609-3
Lucazeau, F. (2019). Analysis and mapping of an updated terrestrial heat flow data set. Geochemistry, Geophysics, Geosystems, 20(8), 4001–
4024. https://doi.org/10.1029/2019gc008389
Mareschal, J.-C., & Jaupart, C. (2013). Radiogenic heat production, thermal regime and evolution of continental crust. Tectonophysics, 609,
524–534. https://doi.org/10.1016/j.tecto.2012.12.001
McCarthy, A., Falloon, T., Sauermilch, I., Whittaker, J., Niida, K., & Green, D. (2020). Revisiting the Australian-Antarctic ocean-continent tran-
sition zone using petrological and geophysical characterization of exhumed subcontinental mantle. Geochemistry, Geophysics, Geosystems,
21(7), e2020GC009040. https://doi.org/10.1029/2020gc009040
McDonough, W. F., Šrámek, O., & Wipperfurth, S. A. (2020). Radiogenic power and geoneutrino luminosity of the Earth and other terrestrial
bodies through time. Geochemistry, Geophysics, Geosystems, 21(7), e2019GC008865. https://doi.org/10.1029/2019gc008865
Mooney, W. D. (2015). Crust and lithospheric structure–global crustal structure. In B. Romanowicz & A. Dziewonski (Eds.), (Vol. 1). Elsevier.
Mooney, W. D., Laske, G., & Masters, T. G. (1998). CRUST 5.1: A global crustal model at 5°×5°. Journal of Geophysical Research, 103(B1),
727–747. https://doi.org/10.1029/97jb02122
Nelson, K. D., Zhao, W., Brown, L., Kuo, J., Che, J., Liu, X., et al. (1996). Partially molten middle crust beneath southern Tibet: Synthesis of
project INDEPTH results. Science, 274(5293), 1684–1688. https://doi.org/10.1126/science.274.5293.1684
Nyblade, A. A., & Pollack, H. N. (1993). A global analysis of heat flow from precambrian terrains: Implications for the thermal structure of
Archean and Proterozoic lithosphere. Journal of Geophysical Research, 98(B7), 12207–12218. https://doi.org/10.1029/93jb00521
Pasyanos, M. E., Masters, T. G., Laske, G., & Ma, Z. (2014). LITHO1.0: An updated crust and lithospheric model of the Earth. Journal of
Geophysical Research: Solid Earth, 119(3), 2153–2173. https://doi.org/10.1002/2013JB010626
Phaneuf, C., & Mareschal, J.-C. (2014). Estimating concentrations of heat producing elements in the crust near the Sudbury Neutrino Observa-
tory, Ontario, Canada. Tectonophysics, 622, 135–144. https://doi.org/10.1016/j.tecto.2014.03.001
Pinet, C., & Jaupart, C. (1987). The vertical distribution of radiogenic heat production in the Precambrian crust of Norway and Sweden: Geother-
mal implications. Geophysical Research Letters, 14(3), 260–263. https://doi.org/10.1029/gl014i003p00260
Regis, D., Warren, C., Mottram, C. M., & Roberts, N. M. (2016). Using monazite and zircon petrochronology to constrain the P–T–t evolution of
the middle crust in the Bhutan Himalaya. Journal of Metamorphic Geology, 34(6), 617–639. https://doi.org/10.1111/jmg.12196
Rudnick, R. L. (1995). Making continental crust. Nature, 378(6557), 571–578. https://doi.org/10.1038/378571a0
Rudnick, R. L., & Fountain, D. M. (1995). Nature and composition of the continental crust: A lower crustal perspective. Reviews of Geophysics,
33(3), 267–309. https://doi.org/10.1029/95rg01302
Rudnick, R. L., & Gao, S. (2014). Composition of the continental crust. In H. D. Holland & K. K. Turekian (Eds.), Treatise on Geochemistry (2nd
edn., pp. 1–51). Elsevier. https://doi.org/10.1016/B978-0-08-095975-7.00301-6
Rudnick, R. L., & Nyblade, A. A. (1999). The thickness and heat production of Archean lithosphere: Constraints from xenolith thermobarometry
and surface heat flow. Mantle petrology: field observations and high pressure experimentation: a tribute to Francis R(Joe) Boyd, 6, 3–12.
Sammon, L. G., Gao, C., & McDonough, W. F. (2020). Lower crustal composition in the southwestern United States. Journal of Geophysical
Research: Solid Earth, 125(3), e2019JB019011. https://doi.org/10.1029/2019jb019011
Sammon, L. G., & McDonough, W. F. (2021). A geochemical review of amphibolite, granulite, and eclogite facies lithologies: Perspectives on
the deep continental crust. Journal of Geophysical Research: Solid Earth, 126(12), e2021JB022791. https://doi.org/10.1029/2021jb022791
Schilling, F. R., & Partzsch, G. M. (2001). Quantifying partial melt fraction in the crust beneath the central Andes and the Tibetan plateau. Physics
and Chemistry of the Earth - Part A: Solid Earth and Geodesy, 26(4–5), 239–246. https://doi.org/10.1016/s1464-1895(01)00051-5
Schmitz, M., Heinsohn, W.-D., & Schilling, F. (1997). Seismic, gravity and petrological evidence for partial melt beneath the thickened central
Andean crust (21°–23°S). Tectonophysics, 270(3–4), 313–326. https://doi.org/10.1016/s0040-1951(96)00217-x
Searle, M., Cottle, J., Streule, M., & Waters, D. (2009). Crustal melt granites and migmatites along the Himalaya: Melt source, segregation,
transport and granite emplacement mechanisms. Earth and Environmental Science Transactions of the Royal Society of Edinburgh, 100(1–2),
219–233. https://doi.org/10.1017/s175569100901617x
Shen, W., Wiens, D. A., Lloyd, A. J., & Nyblade, A. A. (2020). A geothermal heat flux map of Antarctica empirically constrained by seismic
structure. Geophysical Research Letters, 47(14), e2020GL086955. https://doi.org/10.1029/2020gl086955
Šrámek, O., McDonough, W. F., Kite, E. S., Lekić, V., Dye, S. T., & Zhong, S. (2013). Geophysical and geochemical constraints on geoneutrino
fluxes from Earth's mantle. Earth and Planetary Science Letters, 361, 356–366. https://doi.org/10.1016/j.epsl.2012.11.001
Szwillus, W., Afonso, J. C., Ebbing, J., & Mooney, W. D. (2019). Global crustal thickness and velocity structure from geostatistical analysis of
seismic data. Journal of Geophysical Research: Solid Earth, 124(2), 1626–1652. https://doi.org/10.1029/2018jb016593
Turcotte, D., & Schubert, G. (2014). Geodynamics (3rd edn.). Cambridge University Press.
Wilkinson, B. H., McElroy, B. J., Kesler, S. E., Peters, S. E., & Rothman, E. D. (2009). Global geologic maps are tectonic speedometers—Rates
of rock cycling from area-age frequencies. The Geological Society of America Bulletin, 121(5–6), 760–779. https://doi.org/10.1130/b26457.1
Wipperfurth, S. A., Guo, M., Šrámek, O., & McDonough, W. F. (2018). Earth's chondritic Th/U: Negligible fractionation during accretion, core
formation, and crust–mantle differentiation. Earth and Planetary Science Letters, 498, 196–202. https://doi.org/10.1016/j.epsl.2018.06.029
Wipperfurth, S. A., Šrámek, O., & McDonough, W. F. (2020). Reference models for lithospheric geoneutrino signal. Journal of Geophysical
Research: Solid Earth, 125(2), e2019JB018433. https://doi.org/10.1029/2019jb018433

References From the Supporting Information

Hacker, B. R., & Abers, G. A. (2004). Subduction Factory 3: An Excel worksheet and macro for calculating the densities, seismic wave speeds,
and H2O contents of minerals and rocks at pressure and temperature. Geochemistry, Geophysics, Geosystems, 5(1), Q01005. https://doi.
org/10.1029/2003GC000614

SAMMON ET AL. 19 of 20
 21699356, 2022, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JB024041, Wiley Online Library on [06/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Solid Earth 10.1029/2022JB024041

Holland, T. J. B., & Powell, R. (2004). An internally consistent thermodynamic data set for phases of petrological interest. Journal of Metamor-
phic Geology, 16(3), 309–343. https://doi.org/10.1111/j.1525-1314.1998.00140.x
Ries, J., Bettadpur, S., Eanes, R., Kang, Z., Ko, U.-d., McCullough, C., et al. (2016). The development and evaluation of the global gravity model
GGM05 (Tech. Rep.). Report number: CSR-16-02 NASA. Center for Space Research,
Semprich, J., & Simon, N. S. C. (2014). Inhibited eclogitization and consequences for geophysical rock properties and delamination models:
Constraints from cratonic lower crustal xenoliths. Gondwana Research, 25(2), 668–684. https://doi.org/10.1016/j.gr.2012.08.018
Stixrude, L., & Lithgow-Bertelloni, C. (2011). Thermodynamics of mantle minerals-ii. phase equilibria. Geophysical Journal International,
184(3), 1180–1213. https://doi.org/10.1111/j.1365-246x.2010.04890.x

SAMMON ET AL. 20 of 20