EART

H,A

TMO

SPH

ERIC

,A

ND

PLA

NET

ARY

SCIE

NCE

S

Emergence of the Southeast Asian islands as a driverfor Neogene coolingYuem Parka,1 , Pierre Maffrea , Yves Godderisb , Francis A. Macdonaldc, Eliel S. C. Anttilac,and Nicholas L. Swanson-Hysella

aDepartment of Earth and Planetary Science, University of California, Berkeley, CA 94720; bGeosciences Environnement Toulouse, Centre National de laRecherche Scientifique–Universite Paul Sabatier–Institut de Recherche pour le Developpement, Toulouse 31400, France; and cDepartment of Earth Science,University of California, Santa Barbara, CA 93106

Edited by Dennis Kent, Lamont-Doherty Earth Observatory, Palisades, NY, and approved August 25, 2020 (received for review May 29, 2020)

Steep topography, a tropical climate, and mafic lithologies con-tribute to efficient chemical weathering and carbon sequestrationin the Southeast Asian islands. Ongoing arc–continent collisionbetween the Sunda-Banda arc system and Australia has increasedthe area of subaerially exposed land in the region since themid-Miocene. Concurrently, Earth’s climate has cooled since theMiocene Climatic Optimum, leading to growth of the Antarc-tic ice sheet and the onset of Northern Hemisphere glaciation.We seek to evaluate the hypothesis that the emergence ofthe Southeast Asian islands played a significant role in driv-ing this cooling trend through increasing global weatherability.To do so, we have compiled paleoshoreline data and incorpo-rated them into GEOCLIM, which couples a global climate modelto a silicate weathering model with spatially resolved lithol-ogy. We find that without the increase in area of the SoutheastAsian islands over the Neogene, atmospheric pCO2 would havebeen significantly higher than preindustrial values, remainingabove the levels necessary for initiating Northern Hemisphereice sheets.

silicate weathering | weatherability | arc–continent collision |Neogene cooling | Southeast Asian islands

The Southeast Asian islands (SEAIs) have an outsized contri-bution to modern chemical weathering fluxes relative to its

area. The confluence of steep topography, a warm and wet trop-ical climate, and the presence of mafic lithologies result in highfluxes of Ca and Mg cations in the dissolved load and associatedCO2 consumption (1–4). There has been a significant increasein the area of subaerially exposed land within the region sincethe mid-Miocene associated with ongoing arc–continent collisionbetween Australia and the Sunda-Banda arc system (5–7). Con-currently, after the Miocene Climatic Optimum, a cooling trendbegan circa (ca.) 15 million years ago (Ma) and accelerated overthe past 4 million years (m.y.) leading to the development ofNorthern Hemisphere ice sheets (8, 9). Many hypotheses havebeen proposed to explain this cooling trend, including changes inocean/atmosphere circulation (5, 10, 11), a decrease in volcanicdegassing (12), or uplift in the Himalaya (13, 14). Here, we seekto evaluate the hypothesis that emergence of the SEAIs was asignificant factor in driving long-term climatic cooling over theNeogene.

Over geologic time scales, CO2 enters Earth’s ocean–atmosphere system primarily via volcanism and metamorphicdegassing and leaves primarily through the chemical weather-ing of silicate rocks and through organic carbon burial (15).Chemical weathering delivers alkalinity and cations to the ocean,which drives carbon sequestration through carbonate precipi-tation. Steady-state pCO2 is set at the pCO2 level at whichCO2 sinks are equal to sources. As CO2 sinks increase andpCO2 falls, temperature decreases, and the hydrological cycleis weakened, causing the efficiency of the silicate weather-ing sink to decrease until a new steady state is achieved atlower pCO2 (16).

Topography, climate, and lithology all affect chemical weath-ering. High-relief regions generally lack extensive regolith devel-opment and, thus, tend to have reaction-limited weatheringregimes that are more prone to adjust when climate changes(17–19). High physical erosion rates contribute to high chemicalweathering fluxes in these high-relief regions (20). In warm andwet regions, mineral-dissolution kinetics are faster, leading toenhanced chemical weathering (18, 21). Mafic rocks have higherCa and Mg concentrations and dissolution rates than felsic rocksand, thus, have the potential to more efficiently sequester carbonthrough silicate weathering (22). These factors have led to theproposal that arc–continent collisions, which create steep land-scapes that include mafic lithologies, within the tropical rain belthave been important in enhancing global weatherability, lower-ing atmospheric pCO2, and initiating glacial climate over thepast 520 m.y. (7, 23, 24), and perhaps in the Neoproterozoicas well (25).

Quantitatively estimating the magnitude of decrease in steady-state pCO2 associated with the emergence of a region with ahigh carbon-sequestration potential, such as the SEAIs, requiresconstraints on changing tectonic context and accounting of asso-ciated feedbacks. As this region emerges, the total global silicateweathering flux will transiently exceed the volcanic degassingflux, causing pCO2 to initially decline until a new steady stateis established, where the total magnitude of the CO2 sinks is thesame as before the change. However, the sensitivity of the silicateweathering flux in any particular location to this change in pCO2

is variable and dependent on the specific topography, climate,

Significance

The Southeast Asian islands are a modern-day hotspot ofCO2 consumption via silicate weathering. Since ∼15 millionyears ago, these islands have been increasing in size at thesame time that Earth’s climate has been cooling. Here, wetest the hypothesis that this global cooling could have beendriven by tectonic emergence of the Southeast Asian islands.Using a compilation of paleoshorelines, in conjunction witha coupled silicate weathering and climate model, we findthat this emergence is associated with a large decrease inpCO2. Without these changes in tropical island paleogeogra-phy, there would not have been large Northern Hemisphereice sheets as a defining feature of Earth’s climate over the past3 million years.

Author contributions: Y.P., P.M., Y.G., and N.L.S.-H. designed research; Y.P., P.M., Y.G.,F.A.M., E.S.C.A., and N.L.S.-H. performed research; Y.P., P.M., Y.G., and N.L.S.-H. analyzeddata; and Y.P., P.M., Y.G., F.A.M., and N.L.S.-H. wrote the paper.y

The authors declare no competing interest.y

This article is a PNAS Direct Submission.y

Published under the PNAS license.y1 To whom correspondence may be addressed. Email: yuempark@berkeley.edu.y

This article contains supporting information online at https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2011033117/-/DCSupplemental.y

First published September 24, 2020.

www.pnas.org/cgi/doi/10.1073/pnas.2011033117 PNAS | October 13, 2020 | vol. 117 | no. 41 | 25319–25326

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

and lithology at that location. Furthermore, how regional cli-mate responds to this change in pCO2 is itself spatially variable.Therefore, the magnitude of pCO2 change that is required tobalance the total global silicate weathering flux with the volcanicdegassing flux will depend on the specific spatial distribution oftopography, climate, and lithology at the time of emergence. Asa result, any attempt to meaningfully estimate the decrease insteady-state pCO2 associated with the emergence of the SEAIsmust model spatially resolved climatology and silicate weather-ing fluxes in tandem and account for the spatial distribution ofthe factors that affect these interconnected systems.

GEOCLIM ModelTo estimate the decrease in steady-state pCO2 associated withthe increase of subaerially exposed land area in the SEAIs,we use the global spatially resolved GEOCLIM model (26).GEOCLIM estimates changes in steady-state pCO2 associatedwith coupled changes in erosion, chemical weathering, and cli-matology by linking a silicate weathering model to climate modelruns at multiple pCO2 levels.

Silicate Weathering Component. The silicate weathering compo-nent of GEOCLIM calculates CO2 consumption resulting fromsilicate weathering for subaerially exposed land. We assumethat Ca and Mg are the only cations that consume CO2 overgeologic time scales, such that each mole of Ca or Mg thatis dissolved by silicate weathering consumes 1 mol of CO2.While reverse weathering is another potential sink for Ca orMg (27), its parameterization is unclear, and it has been inter-preted to be a relatively minor flux in the Cenozoic (28), andwe do not include it in our model. In previous versions of themodel, silicate weathering was a function of temperature andrunoff only, and all bedrock was assigned identical chemical com-positions (26). More recent versions of GEOCLIM implementregolith development and soil shielding (Fig. 1), which intro-

UNWEATHEREDBEDROCK

REGOLITH Pr

Ep

W

χCaMg

z

0

h

undissolvedphases dissolved

solutes

z

proportion of primary phases (x)

0

h

0 1

xs

Fig. 1. A schematic representation of the silicate weathering componentof GEOCLIM in a single profile at steady state. A rock particle leavesthe unweathered bedrock with production rate Pr and transits verticallythrough a regolith of height h. Regolith production and physical erosion(Ep) are equal at steady state. As a particle transits upwards, some frac-tion of the primary phases (x) are chemically weathered (W), with the fluxof dissolved Ca+Mg being W multiplied by the concentration of Ca+Mgin unweathered bedrock (XCaMg). Details of the formulation for the sili-cate weathering component of GEOCLIM can be found in Materials andMethods.

duces a dependence on erosion rate (and, therefore, topographicslope) (29). While this introduction of regolith development intoGEOCLIM is important for assessing the impact of tropical arc–continent collisions on pCO2, the relatively high Ca+Mg con-centration in arc rocks relative to other lithologies must also beconsidered.

We therefore implement variable bedrock Ca+Mg concentra-tion into GEOCLIM (SI Appendix). The spatial distribution oflithologies is sourced from the Global Lithologic Map (GLiM)(30) and is represented by six categories: metamorphic, felsic,intermediate, mafic, carbonate, and siliciclastic sediment. Eachland pixel is assigned these lithologic categories at a resolutionof 0.1◦× 0.1◦. The Ca+Mg concentrations of felsic, interme-diate, and mafic lithologies are assigned based on the meanof data of these lithologic categories compiled in EarthChem(www.earthchem.org/portal). Given that GLiM does not distin-guish ultramafic lithologies, such rocks are grouped with maficrocks. As a result, the Ca+Mg concentration is likely an underes-timate in regions of obducted ophiolites, such that the estimatedeffect of these regions on changing steady-state pCO2 could beconservative (31). The weathering of carbonate does not con-tribute to long-term CO2 consumption, and its Ca+Mg concen-tration is ignored. The Ca+Mg concentrations of metamorphicand siliciclastic sediment lithologies are more difficult to define,since their chemical composition is strongly dependent on pro-tolith composition and, in the case of siliciclastic sediment, thedegree of previous chemical depletion. We explore a range offeasible Ca+Mg concentrations for metamorphic rocks and sili-ciclastic sediment during calibration of the silicate-weatheringcomponent of GEOCLIM.

Calibration. The values of four parameters within the silicate-weathering component that modify the dependence of sili-cate weathering on temperature, runoff, erosion, and regoliththickness are poorly constrained. Rather than prescribing sin-gle values, we selected multiple values for each of these fourparameters along with the Ca+Mg concentration of meta-morphic and siliciclastic lithologies from within reasonableranges (SI Appendix, Table S2). We then permuted all possi-ble combinations of these values for the six parameters, lead-ing to 93,600 unique parameter combinations (i.e., permuta-tions). For each combination, we computed spatially resolvedlong-term CO2 consumption associated with Ca+Mg fluxesusing present-day runoff, temperature, and slope. We sum-computed CO2 consumption over watersheds for which data-constrained estimates are available (1, 32) and then calcu-lated the coefficient of determination (r2) between computedand measured CO2 consumption in each of these water-sheds. After eliminating parameter combinations that resultedin low r2, 573 parameter combinations remained (SI Appendix,Fig. S3). The resulting global CO2 consumption of thesefiltered model runs all overlap with independently derivedestimates of the global CO2 degassing flux (33), as theyshould for a steady-state long-term carbon cycle (SI Appendix,Fig. S3).

Climate Model Component. Having calibrated the silicate weath-ering component of GEOCLIM, we used it to estimate thedecrease in steady-state pCO2 associated with emergence of theSEAIs. For the climate-model component, we used temperatureand runoff from a subset of the Geophysical Fluid Dynam-ics Laboratory (GFDL) CM2.0 experiments (34) (SI Appendix).These experiments are well-suited for this analysis because allnon-CO2 forcings are held constant at values representative ofpreindustrial conditions, allowing the effect of changing pCO2

on climatology to be isolated. Furthermore, the experimentswere run long enough for the final system to approximatesteady state.

25320 | www.pnas.org/cgi/doi/10.1073/pnas.2011033117 Park et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

EART

H,A

TMO

SPH

ERIC

,A

ND

PLA

NET

ARY

SCIE

NCE

S

PaleoshorelinesTo determine the position of paleoshorelines in the SEAIs overthe past 15 m.y., we used terrestrial and marine sedimentarydeposits (Fig. 2; SI Appendix). The paleoshoreline data indicatethat the Sunda-Banda Arc and New Guinea are primarily respon-sible for the increase in area since 15 Ma. Exhumation of themodern Sunda-Banda Arc is the result of ongoing arc–continentcollision with the Australian Plate (37). Most of Sumatra andJava along with the nonvolcanic islands of the Outer Banda Arcwere elevated above sea level after 5 Ma (38). In New Guinea,emergence in the mid-Miocene is associated with collisionbetween the Melanesian Arc and Australia’s distal margin (39),which drove exhumation of the Irian-Marum-April OphioliteBelt. Exhumation accelerated over the past 4 m.y. in the NewGuinea Central Range, due to slab breakoff and buoyant uplift,and in eastern New Guinea due to jamming of the north-dipping subduction zone (39). We also include changes in areasof presently submerged continental shelves, such as the SundaShelf, that were previously exposed (SI Appendix, Fig. S7). Thesetectonic drivers and others throughout the region led to progres-sive emergence over the past 15 m.y. that accelerated following5 Ma (Fig. 2B). This trend mirrors broad cooling over the Neo-

gene that resulted in the initiation of Northern Hemisphere icesheets (Fig. 2C).

We used GEOCLIM to estimate pCO2 associated with thereconstructed subaerial extent of the SEAIs at ca. 15, 10, and5 Ma (“paleo-SEAIs” scenarios; Fig. 3). Because we used a cli-mate model forced with modern geography, the position of thetectonic blocks remains fixed. Although there has been motionof these tectonic blocks since 15 Ma, they have remained withintropical latitudes, such that this fixed scenario is a good approx-imation of the paleogeography (SI Appendix, Fig. S10). Wealso tested an end-member scenario, in which all islands asso-ciated with arc–continent collision in the region were removed(“removed SEAIs” scenario; Fig. 3).

pCO2 EstimatesUsing the 573 unique parameter combinations, the paleo-SEAIsscenarios resulted in 526 to 678 parts per million (ppm) for15 Ma, 457 to 516 ppm for 10 Ma, and 391 to 434 ppmpCO2 for 5 Ma (Fig. 3). These results indicate a progressivedecrease in pCO2 over the Neogene associated with the emer-gence of the SEAIs and suggest that, without this emergence,preindustrial pCO2 would have been ∼526 to 678 ppm. These

A

B

C

Fig. 2. The emergence of the SEAIs (also referred to as the Maritime Continent in the climate-science literature) from the mid-Miocene to present. Pastshorelines at 5, 10, and 15 Ma are shown in A, with associated land area summarized in B. A significant increase in area over the past 5 m.y. is coincidentwith cooling and the onset of Northern Hemisphere glaciation, as reflected in the benthic oxygen isotope record (35) shown in C.

Park et al. PNAS | October 13, 2020 | vol. 117 | no. 41 | 25321

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

Fig. 3. Steady-state pCO2 estimates from GEOCLIM for the various scenarios discussed in the text. For each of the seven scenarios, each point representsan estimate from one of the 573 unique parameter combinations that most closely matched estimates of present-day CO2 consumption in 80 watershedsaround the world (SI Appendix). The box encloses the middle 50% of the pCO2 estimates (i.e., the interquartile range), and the notch represents the medianwith its 95% CI. The whiskers extend to the 2.5th and 97.5th percentile values. Glaciation thresholds (36) are shown on the x axis.

paleo-SEAIs scenarios do not account for Neogene changes out-side of the SEAIs (e.g., changes in ocean/atmosphere circulation,volcanic degassing, and weathering fluxes elsewhere on Earth,discussed in Alternative Mechanisms for Neogene Cooling). There-fore, these results are not estimating pCO2 at 15 Ma, but, rather,are quantifying pCO2 change associated with emergence of theSEAIs.

Proxy-based estimates of the magnitude and trajectory ofpCO2 change from the Miocene to the Pliocene are variablebetween techniques and associated assumptions underlying theirinterpretation (SI Appendix, Fig. S11). The pCO2 values from the5 Ma paleo-SEAIs scenario overlap with many proxy-based esti-mates (40), as well as values that emerge from approaches thatassimilate climate and ice-sheet model output with benthic δ18Odata (41, 42). The modeled pCO2 values for 15 Ma resemble thehigher end of proxy-based pCO2 estimates for the early to mid-Miocene, indicating that the increase in subaerially exposed landarea and tectonic topography of the SEAIs is sufficient to explainlong-term cooling of Earth’s climate over the Neogene. ThepCO2 threshold for Antarctic glaciation is estimated to be ∼750ppm, with that for Northern Hemisphere glaciation being signifi-cantly lower at ∼280 ppm (36). These modeled values of decreas-ing pCO2 associated with emergence of the SEAIs are, therefore,consistent with the record of Neogene climate with Miocene icesheets on Antarctica (43), followed by Northern Hemisphere icesheets developing in the Pliocene (44), as pCO2 subsequentlydecreased.

The results of our paleo-SEAIs scenarios highlight the impor-tance of the combination of topography, runoff, and lithology insetting Earth’s climate state. To independently explore the effectof the modern-day surface exposure of lower-relief basaltic lavason steady-state pCO2 (45), we replaced mafic volcanics asso-ciated with the Deccan Traps, Ethiopian Traps, and ColumbiaRiver Basalts with the Ca+Mg concentration of bulk continentalcrust in GEOCLIM (Fig. 3). The resulting pCO2 is ∼300 to 500ppm, indicating that the presence of mafic rocks in these igneousprovinces affects steady-state pCO2, as has been suggested to beimportant for Paleogene cooling (45). However, the higher 526-to 678-ppm values for the 15 Ma paleo-SEAIs scenario illustratethat higher relief and a wet tropical climate significantly increasethe efficiency of CO2 consumption, especially when paired with

high Ca+Mg lithologies. As such, arc–continent collisions in thetropics are likely more important for driving long-term changesin pCO2 than the eruption of flood basalts (7, 46).

Previous work has estimated that the decrease in pCO2 since5 Ma associated with the emergence of the SEAIs and enhancedsilicate weathering is ∼19 ppm (5), in which case their emergencewould be a relatively minor contributor to Neogene cooling. This19-ppm estimate was obtained by using an equation that assumesa direct linear relationship between mean global temperatureand changes in weathering-rate-weighted land area, scaled bya factor that is intended to account for the influence of bothrunoff and temperature. pCO2 was then estimated from the cal-culated temperature by using a simple energy-balance equation.However, the relationship between mean global temperature(or pCO2) and weathering-rate-weighted land area is not linear.Furthermore, this simple linear relationship ignores spatial vari-ability in topography and climatology and only crudely accountsfor spatial variability in lithology. In fact, the 19-ppm estimateis closer in magnitude to the decrease in pCO2 that we estimateif mafic volcanics associated with the Deccan Traps (a relativelyflat area outside of the warm and wet tropics) are replaced withthe Ca+Mg concentration of bulk continental crust (22 to 70ppm; Fig. 3). The significant difference in steady-state pCO2

estimated between the removed Deccan Traps scenario and thepaleo-SEAIs scenarios (Fig. 3) demonstrates that consideringchanges in the spatial distribution of lithologies alone is notadequate for estimating changes in steady-state pCO2. Instead,spatially varying topography and climatology significantly mod-ulates silicate weathering rates and must be accounted forwhen estimating pCO2 change associated with paleogeographicchange.

An important caveat for these estimates of pCO2 is that ourmodeling is determining the climatology in the GFDL CM2.0model at which steady state is achieved—a climatology that hasan associated pCO2 value in the model. However, climate mod-els are variable in their response to changes in pCO2. One wayto summarize this variability is through the equilibrium climate-sensitivity value—the steady-state change in global mean surfaceair temperature associated with a doubling of pCO2. A rangeof 1.5 to 4.5 ◦C per pCO2 doubling was proposed in the land-mark Charney report (47), and this range was considered to

25322 | www.pnas.org/cgi/doi/10.1073/pnas.2011033117 Park et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

EART

H,A

TMO

SPH

ERIC

,A

ND

PLA

NET

ARY

SCIE

NCE

S

be the credible interval (>66% likelihood) in the last Intergov-ernmental Panel on Climate Change report (48). Integratingconstraints both from understanding of climate feedback pro-cesses and the climate record, a recent comprehensive reviewestimates the 66% probability range of climate sensitivity to be2.6 to 3.9 ◦C per pCO2 doubling, with a 5 to 95% range of 2.3 to4.7 ◦C per pCO2 doubling (49). The equilibrium climate sensitiv-ity associated with the CM2.0 climate models is 2.9 ◦C per pCO2

doubling, which falls within these ranges, although these rangesremain broad. An alternative way to consider the results fromour analysis would be that an estimate of 572 ppm (2× preindus-trial pCO2) for the 15 Ma paleo-SEAIs scenario is implying thatEarth would be ∼ 2.9 ◦C warmer. If Earth’s climate sensitivity isat the higher end of the probable range and higher than in theCM2.0 model, as it is in some climate models, this same amountof Neogene cooling resulting from the emergence of the SEAIscould have been driven by a smaller change in pCO2.

Alternative Mechanisms for Neogene CoolingOcean/Atmosphere Circulation. Some hypotheses to explain ice-sheet growth over the Neogene invoke changes in ocean/atmosphere circulation, including further climatic isolation ofAntarctica due to strengthening of the circumpolar current (11);increased atmospheric moisture in the Northern Hemispheredue to intensified thermohaline circulation following PanamaIsthmus emergence (10); and cooling of North America resultingfrom a strengthened Walker Circulation associated with emer-gence of the SEAIs (5). Such changes in ocean/atmospherecirculation are likely to modulate pCO2 thresholds for glacial ini-tiation and ice-sheet growth (36). However, the prolonged timescale of the cooling trend since 15 Ma (Fig. 2C) is most readilyattributable to decreasing pCO2 associated with evolving geo-logical sources and sinks of carbon, modulated by the silicateweathering feedback (16, 50–53).

Volcanic Degassing. A decrease in volcanic degassing (12) has alsobeen proposed as a driver for Neogene cooling. However, proxy-based estimates of the evolution of volcanic degassing fluxesthroughout the Neogene are inconsistent with each other, suchthat not even the sign of the change in volcanic degassing overthe past ∼15 m.y. is without ambiguity (26). For example, ithas both been estimated that the volcanic degassing flux was∼25% lower (54) and ∼10% higher (55) at 15 Ma relative to thepresent day.

Our model framework provides an opportunity to estimate thedecrease in volcanic degassing flux necessary to achieve the samechange in pCO2 predicted for the increase in global weatherabil-ity associated with the emergence of the SEAIs over the past15 m.y. If we use the parameter combination that had the high-est r2 between computed and measured CO2 consumption inwatersheds around the world during calibration (Calibration andSI Appendix, Fig. S4), GEOCLIM estimates a preindustrial vol-canic degassing flux of 4.1×1012 mol/y to balance the silicateweathering flux at 286 ppm pCO2. If we then assume that thisvolcanic degassing flux did not change over the past 15 m.y., thenGEOCLIM estimates that the increase in global weatherabilityassociated with the emergence of the SEAIs led to a changein pCO2 of ∼280 ppm (“increase in weatherability only” sce-nario in Fig. 4). If we, instead, assume that global weatherabilitydid not change over the past 15 m.y., then we estimate that thevolcanic degassing flux needs to have been ∼13% greater at15 Ma relative to the preindustrial to drive the same ∼280 ppmchange in pCO2 (“decrease in degassing only” scenario in Fig. 4).This ∼13% value is higher than 10%, the highest current esti-mate for the volcanic degassing flux at 15 Ma relative to thepresent day (55).

However, changes in the volcanic degassing flux would havemodulated changes in pCO2 associated with changes in global

A1

A2

mor

ew

eath

erab

le

increase inweatherability

only

decrease indegassing only

increase inweatherabilityand degassing

568 ppm

B

400 ppm

Fig. 4. Weatherability curves for the modern and “paleo-SEAIs” scenariosshown in Fig. 3. Lower expands the lower pCO2 range (x axis) of Upper.Details on how these curves were generated are described in Materials andMethods. Each of the four curves represent a different tectonic boundarycondition (i.e., the reconstructed paleoshorelines of the SEAIs; Fig. 2A) and,therefore, a different global weatherability. The curves show the resultingpCO2 for a given volcanic degassing flux such that the input flux is balancedby the silicate weathering output flux. Point B represents the preindustrial,in which pCO2 is 286 ppm. The arrow from point A1 to B represents the“increase in weatherability only” scenario, in which global weatherabilityincreases as the SEAIs emerge, but the volcanic degassing flux does notchange over the past 15 m.y. In this scenario, the pCO2 decreases from thevalue dictated by the 15 Ma paleo-SEAIs weatherability curve (568 ppm).The arrow from point A2 to B instead represents the “decrease in degassingonly” scenario, in which global weatherability remains the same as thepreindustrial, but the same change in pCO2 as the “increase in weather-ability only” scenario is achieved by decreasing the volcanic degassing fluxfrom a value∼13% greater than the preindustrial. The arrow from Point A3

to B represents the “increase in weatherability and degassing” scenario, inwhich a change in pCO2 from 400 to 286 ppm is achieved by increasingboth global weatherability from our 15-Ma tectonic boundary conditionand the volcanic degassing flux from a value ∼7% smaller than thepreindustrial flux.

weatherability. For example, some proxy-based approaches, aswell as some model-data assimilation approaches, estimatedthat mid-Miocene pCO2 was lower than 568 ppm (SI Appendix,Fig. S11). Take a scenario in which pCO2 was 400 ppm at15 Ma. If we assume that the pCO2 decrease to the preindus-trial value of 286 ppm was driven by the increase in globalweatherability associated with emergence of the SEAIs in con-junction with an increase in volcanic degassing, which coun-teracts cooling by increasing the flux of CO2 to the atmo-sphere (“increase in weatherability and degassing” scenario inFig. 4), the volcanic degassing flux would have had to have been∼7% smaller than the preindustrial. More robust constraints onpCO2 (SI Appendix, Fig. S11) and/or volcanic degassing ratesover the past 20 m.y. are needed to constrain which of the“increase in weatherability only” or “increase in weatherability

Park et al. PNAS | October 13, 2020 | vol. 117 | no. 41 | 25323

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

and degassing” scenarios (Fig. 4) is more representative of themechanisms driving Neogene cooling.

Himalayan Uplift. Marine 87Sr/86Sr has overall been increasingsince ca. 35 Ma (56). The traditional explanation for this trendis that it reflects increased weathering of radiogenic (i.e., high87Sr/86Sr) silicate rocks (13, 57). Associated with this expla-nation is the proposal that increasing weathering of radiogenicsilicate rocks in the Himalayas was the primary driver of Neogenecooling (58). It could be argued that increasing marine 87Sr/86Sris inconsistent with the hypothesis that increasing weatheringof juvenile (i.e., low 87Sr/86Sr) silicate rocks in the SEAIs wasan important driver of Neogene cooling. However, the globallyaveraged ratio of silicate weathering fluxes from radiogenic cra-tonic rocks versus juvenile arc lithologies can be at least partiallydecoupled from marine 87Sr/86Sr via the regional weatheringof isotopically unique lithologies. For example, in addition tohighly radiogenic granites and gneisses (57), unusually radio-genic carbonates are abundant in Himalayan strata, and it isestimated that ∼75% of Sr coming from the Himalayas canbe attributed to carbonate rather than silicate weathering (59–61). As such, there are challenges in interpreting the marine87Sr/86Sr record as a direct proxy for silicate weathering fluxes.Nevertheless, steadily increasing marine 87Sr/86Sr is interruptedca. 16 Ma, when the slope of the 87Sr/86Sr curve decreases (56).This decrease in slope has been attributed to coincident exhuma-tion of relatively low 87Sr/86Sr Outer Lesser Himalaya carbon-ates (62, 63), but could also be at least partially driven by theemergence of low 87Sr/86Sr lithologies in the SEAIs during arc–continent collision. Increasing seawater Mg/Ca since ca. 15 Ma(64) is consistent with an increasing proportion of the global sil-icate weathering flux being derived from mafic and ultramaficsources.

Himalayan uplift would have affected geological carbon sinks,either via increased weathering of silicate rocks (58) or enhancedburial of organic matter in the Bengal Fan (14). Increased weath-ering of the emerging SEAIs would have occurred in tandem withsuch changes in the Himalayas, such that the effects of thesepaleogeographic changes on geochemical proxy records, likemarine 87Sr/86Sr, become difficult to disentangle. In addition,given the large uncertainty associated with changes in regionalclimatology across Asia due to Himalayan orogeny, developingquantitative estimates of the evolution of global silicate weath-ering fluxes associated with Himalayan orogeny remains a majorchallenge.

The Geologic Carbon CycleIf geological carbon sources remain approximately constant,global alkalinity delivery from silicate weathering needs to beapproximately constant as well to keep the long-term carboncycle in steady state (16). Enhanced silicate weathering in aregion such as the SEAIs is compensated by a decrease insilicate weathering elsewhere. Global alkalinity delivery from sil-icate weathering does not change, but occurs more efficientlyand, thereby, at lower pCO2. Given that carbonate weatheringis disconnected from the long-term carbon-cycle mass balance,changes in carbonate accumulation through time (65) could bedriven by changes in carbonate weathering.

The long-term carbon-cycle mass balance can be perturbedvia mechanisms that are disconnected from changes in volcanicdegassing and silicate weathering rates. For example, sulfide oxi-dation coupled to carbonate dissolution could act as a sourceof CO2 on million year time scales (66). Similarly, the weath-ering of sedimentary organic matter could serve as a source ofCO2 (67). On the other hand, enhanced burial of organic mat-ter enabled by higher sediment and nutrient delivery could bean important sink of CO2, as has been suggested in the Ben-gal Fan (14) and Taiwan (68). The fluxes of CO2 represented

by these processes are not accounted for in our model frame-work and could have been affected by emergence of the SEAIsand/or Himalayan orogeny. pCO2 changes that result from theseprocesses would be superimposed on pCO2 changes associatedwith evolving silicate weathering fluxes. However, our coupledweathering-climate model indicates that the pCO2 change asso-ciated with increased global weatherability driven by emergenceof the SEAIs is sufficient to explain the majority of Neogenecooling (Fig. 3). Without this emergence, pCO2 would haveremained above the level necessary for the growth of NorthernHemisphere ice sheets.

ConclusionsCoupled geological constraints and modeling experimentsdemonstrate that the SEAIs have been a growing hot spotfor carbon sequestration due to silicate weathering from theMiocene to present. Changes in volcanic degassing and paleo-geography elsewhere on Earth, particularly in the Himalayas andCentral America, would have also affected geological carbonsources and sinks. Yet, not only does the history of emergenceof the SEAIs coincide with Neogene cooling and the onset ofNorthern Hemisphere glaciation, but our coupled weathering-climate model also indicates that the associated steady-statepCO2 change is sufficient to explain much of this cooling. Theseresults highlight that the Earth’s climate state is particularlysensitive to changes in tropical geography.

Materials and MethodsGEOCLIM Silicate Weathering Component. The silicate weathering compo-nent of the GEOCLIM model has been modified from the published version(20). The component implements the model of Gabet and Mudd (2009) (17)for the development of a chemically weathered profile. We refer to thischemically weathered profile as regolith, where the base of the regolithis unweathered bedrock. In the model of Gabet and Mudd (2009), mate-rial enters the regolith and leaves either as a solute through chemicalweathering of the material during its travel from the bedrock toward thesurface, or as a physically weathered particle once it reaches the top. Weused the DynSoil implementation of the Gabet and Mudd (2009) model,which integrates chemical weathering within the regolith (18). The tran-sient time-varying version of this regolith model is described by threeequations:

dh

dt= Pr − Ep, [1]

∂x

∂t=−Pr

∂x

∂z−Kτσx, [2]

∂τ

∂t=−Pr

∂τ

∂z+ 1. [3]

Eq. 1 is a statement of material conservation, where h is the total heightof the regolith (m), t is the model time (y), Pr is the regolith productionrate (m/y), and Ep is the physical erosion rate (m/y). Eq. 2 describes how theresidual fraction of weatherable phases (x, unitless) changes as a functionof time (t, y) and depth (z, m). Kτσ is the dissolution rate constant, whichdepends on the local climate (captured by K, y−1−σ) and the time that agiven rock particle has spent in the regolith (τ , y) to some power σ (unitless)which implements a time dependence. Eq. 3 describes how the time that agiven rock particle has spent in the regolith changes as time in the modelprogresses.

The net weathering rate in the regolith column (W , m/y) can then becalculated with:

W =

∫ h

0Kτσx dz. [4]

The regolith production rate can be expressed as the product of the optimalproduction rate (P0) and a soil-production function (f(h)):

Pr = P0 f(h), [5]

P0 = krp q e− Ea

R

(1T − 1

T0

), [6]

25324 | www.pnas.org/cgi/doi/10.1073/pnas.2011033117 Park et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

EART

H,A

TMO

SPH

ERIC

,A

ND

PLA

NET

ARY

SCIE

NCE

S

f(h) = e−hd0 . [7]

P0 is the “optimal” regolith production rate (m/y), which is defined to be theregolith production rate when there is no overlying regolith. In Eq. 6, wherekrp is a proportionality constant (unitless), q is the runoff (m/y), Ea is the activa-tion energy (J/K/mol), R is the ideal gas constant (J/mol), T is the temperature(K), and T0 is the reference temperature (K), we parameterize the optimalregolith production rate (69). f(h) is the soil-production function (unitless),which describes how regolith production decreases as the thickness of theregolith increases. It takes an exponential form, as is commonly applied in theliterature (17). In Eq. 7, d0 is a reference to regolith thickness (m) (70).

Our implementation of the erosion rate is parameterized based on runoffand slope (s; m/m):

Ep = ke qm sn. [8]

ke is a proportionality constant [(m/y)1−m] and m and n are adjustable expo-nents that are kept as 0.5 and 1 (29). This formulation is directly inspired bythe stream power law (71). This formulation and these exponent values aresupported by compilations, but variability in the proportionality constant isdifficult to capture at a global scale (72).

The K in the dissolution rate constant in Eq. 2 describes the dependenceof the chemical weathering on climate:

K = kd

(1− e−kw q

)e− Ea

R

(1T − 1

T0

). [9]

Eq. 9 is an empirical simplification of mineral-dissolution rates derived fromkinetic theory and laboratory experiments (18), where kd is a proportionalityconstant that modifies the dependence of dissolution rate on runoff andtemperature (y−1−σ), and kw is a proportionality constant that modifiesthe dependence of dissolution rate on runoff (y/m).

In this study, we are interested in obtaining the steady-state solutionrather than the transient time-varying solution. The steady-state solution forDynSoil can be calculated analytically by setting the time derivatives equalto zero, resulting in the following set of equations:

h = max(

0, d0 log(

P0

Ep

)), [10]

x(z) = exp(−

K

σ+ 1

(z

Ep

)σ+1)

, [11]

W = Ep(1− x(h)) = Ep(1− xs). [12]

x(z) is the abundance profile of primary phases inside the regolith, vary-ing with height upward from the base of the regolith, as shown inFig. 1. Setting z equal to the regolith thickness (h) gives xs, which isthe proportion of primary phases remaining at the top of the regolithcolumn.

Weatherability Curves. To create the 15-Ma paleo-SEAIs curve shown inFig. 4, we used the reconstructed paleoshorelines of the SEAIs at 15 Ma(Fig. 2A). We then selected the parameter combination that had the high-est r2 between computed and measured CO2 consumption in watershedsaround the world during calibration (Calibration and SI Appendix, Fig. S4)and fixed pCO2 at the three pCO2 levels at which the GFDL CM2.0 climatemodel experiments were computed (SI Appendix). We then ran GEOCLIMat each of these pCO2 levels until steady state was achieved (i.e., untilthe volcanic degassing flux was equal to the silicate weathering flux). Wethen repeated this process for the 10- and 5-Ma paleo-SEAIs paleoshorelinesand the present-day shorelines to generate the three other weather-ability curves. Each estimated pCO2 in Fig. 3 is the result of underlyingweatherability curves that change with the different chemical weatheringparameters.

SI Appendix. A detailed description of the implementation of lithology intothe silicate weathering component of GEOCLIM, the calibration of the sili-cate weathering component of GEOCLIM, the GFDL CM2.0 climate model,and the paleoshoreline reconstructions can be found in SI Appendix.

Data Availability. The code for the GEOCLIM model used in this study can befound at GitHub, https://github.com/piermafrost/GEOCLIM-dynsoil-steady-state/releases/tag/v1.0. The code that generated the inputs and analyzedthe output of the GEOCLIM model, as well as these input and outputfiles, can be found at GitHub, https://github.com/Swanson-Hysell-Group/2020 Southeast Asian Islands, and Zenodo, https://doi.org/10.5281/zenodo.4021653.

ACKNOWLEDGMENTS. Collaborative research between N.L.S.-H. and Y.G.was initially supported by a grant from the France-Berkeley Fund. Projectresearch was supported by NSF Frontier Research in Earth Sciences Grants1926001 and 1925990. We thank Alec Brenner, Sam Lo Bianco, Mariana Lin,and Judy Pu for their data compilation contributions to the paleoshorelinereconstructions.

1. J. Gaillardet, B. Dupre, P. Louvat, C. J. Allegre, Global silicate weathering and CO2consumption rates deduced from the chemistry of large rivers. Chem. Geol. 159, 3–30(1999).

2. J. Hartmann, N. Jansen, H. H. Durr, S. Kempe, P. Kohler, Global CO2-consumption bychemical weathering: What is the contribution of highly active weathering regions?Glob. Planet. Change 69, 185–194 (2009).

3. J. D. Milliman, K. L. Farnsworth, River Discharge to the Coastal Ocean: A GlobalSynthesis (Cambridge University Press, Cambridge, UK, 2013).

4. J. Hartmann, N. Moosdorf, R. Lauerwald, M. Hinderer, A. Joshua West, Global chemi-cal weathering and associated P-release—The role of lithology, temperature and soilproperties. Chem. Geol. 363, 145–163 (2014).

5. P. Molnar, T. W. Cronin, Growth of the Maritime Continent and its possiblecontribution to recurring ice ages. Paleoceanography 30, 196–225 (2015).

6. R. Hall, Southeast Asia: New views of the geology of the Malay Archipelago. Annu.Rev. Earth Planet. Sci. 45, 331–358 (2017).

7. F. A. Macdonald, N. L. Swanson-Hysell, Y. Park, L. Lisiecki, O. Jagoutz, Arc-continentcollisions in the tropics set Earth’s climate state. Science 364, 181–184 (2019).

8. N. J. Shackleton et al., Oxygen isotope calibration of the onset of ice-rafting andhistory of glaciation in the North Atlantic region. Nature 307, 620–623 (1984).

9. J. Zachos, M. Pagani, L. Sloan, E. Thomas, K. Billups, Trends, rhythms, and aberrationsin global climate 65 Ma to present. Science 292, 686–693 (2001).

10. G. H. Haug, R. Tiedemann, Effect of the formation of the Isthmus of Panama onAtlantic Ocean thermohaline circulation. Nature 393, 673–676 (1998).

11. A. E. Shevenell, Middle Miocene Southern Ocean cooling and Antarctic cryosphereexpansion. Science 305, 1766–1770 (2004).

12. R. A. Berner, A. C. Lasaga, R. M. Garrels, The carbonate-silicate geochemical cycle andits effect on atmospheric carbon dioxide over the past 100 million years. Am. J. Sci.283, 641–683 (1983).

13. M. E. Raymo, W. F. Ruddiman, P. N. Froelich, Influence of late Cenozoic mountainbuilding on ocean geochemical cycles. Geology 16, 649–653 (1988).

14. V. Galy et al., Efficient organic carbon burial in the Bengal fan sustained by theHimalayan erosional system. Nature 450, 407–410 (2007).

15. L. R. Kump, S. L. Brantley, M. A. Arthur, Chemical weathering, atmospheric CO2, andclimate. Annu. Rev. Earth Planet. Sci. 28, 611–667 (2000).

16. L. R. Kump, M. A. Arthur, “Global chemical erosion during the Cenozoic: Weather-ability balances the budgets” in Tectonic Uplift and Climate Change, W. F. Ruddiman,Ed. (Springer US, Boston, MA, 1997), pp. 399–426.

17. E. J. Gabet, S. M. Mudd, A theoretical model coupling chemical weathering rates withdenudation rates. Geology 37, 151–154 (2009).

18. A. Joshua West, Thickness of the chemical weathering zone and implications for ero-sional and climatic drivers of weathering and for carbon-cycle feedbacks. Geology 40,811–814 (2012).

19. K. Maher, C. P. Chamberlain, Hydrologic regulation of chemical weathering and thegeologic carbon cycle. Science 343, 1502–1504 (2014).

20. Y. Godderis et al., Onset and ending of the late Palaeozoic ice age triggered bytectonically paced rock weathering. Nat. Geosci. 10, 382–386 (2017).

21. A. C. Lasaga, J. M. Soler, J. Ganor, T. E. Burch, K. L. Nagy, Chemical weatheringrate laws and global geochemical cycles. Geochim. Cosmochim. Acta 58, 2361–2386(1994).

22. C. Dessert, B. Dupre, J. Gaillardet, L. M. Francois, C. J. Allegre, Basalt weathering lawsand the impact of basalt weathering on the global carbon cycle. Chem. Geol. 202,257–273 (2003).

23. J. Oliver, F. A. Macdonald, L. Royden, Low-latitude arc-continent collision as a driverfor global cooling. Proc. Natl. Acad. Sci. U.S.A. 113, 4935–4940 (2016).

24. N. L. Swanson-Hysell, F. A. Macdonald, Tropical weathering of the Taconic orogeny asa driver for Ordovician cooling. Geology 45, 719–722 (2017).

25. Y. Park et al., The lead-up to the Sturtian Snowball Earth: Neoproterozoicchemostratigraphy time-calibrated by the Tambien Group of Ethiopia. GSA Bull. 132,1119–1149 (2020).

26. Y. Godderis, Y. Donnadieu, A sink- or a source-driven carbon cycle at the geo-logical timescale? Relative importance of palaeogeography versus solid Earthdegassing rate in the Phanerozoic climatic evolution. Geol. Mag. 156, 355–365(2017).

27. P. Michalopoulos, R. C. Aller, Rapid clay mineral formation in Amazon Delta sed-iments: Reverse weathering and oceanic elemental cycles. Science 270, 614–617(1995).

28. T. T. Isson, N. J. Planavsky, Reverse weathering as a long-term stabilizer of marine pHand planetary climate. Nature 560, 471–475 (2018).

29. P. Maffre et al., Mountain ranges, climate and weathering. Do orogens strengthenor weaken the silicate weathering carbon sink? Earth Planet. Sci. Lett. 493, 174–185(2018).

30. J. Hartmann, N. Moosdorf, The new global lithological map database GLiM: A rep-resentation of rock properties at the Earth surface. Geochem. Geophys. Geosyst. 13,1–37 (2012).

Park et al. PNAS | October 13, 2020 | vol. 117 | no. 41 | 25325

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

31. H. H. Schopka, L. A. Derry, C. A. Arcilla, Chemical weathering, river geochemistry andatmospheric carbon fluxes from volcanic and ultramafic regions on Luzon Island, thePhilippines. Geochem. Cosmochim. Acta 75, 978–1002 (2011).

32. J.-S. Moquet et al., Temporal variability and annual budget of inorganic dissolvedmatter in Andean Pacific Rivers located along a climate gradient from northernEcuador to southern Peru. C. R. Geosci. 350, 76–87 (2018).

33. T. Gerlach, Volcanic versus anthropogenic carbon dioxide. Trans. Am. Geophys. Union92, 201–202 (2011).

34. T. L. Delworth et al., GFDL’s CM2 global coupled climate models. Part I: Formulationand simulation characteristics. J. Clim., 19, 643–674 (2006).

35. J. C. Zachos, G. R. Dickens, R. E. Zeebe, An early Cenozoic perspective on greenhousewarming and carbon-cycle dynamics. Nature 451, 279–283 (2008).

36. R. M. DeConto et al., Thresholds for Cenozoic bipolar glaciation. Nature 455, 652–656(2008).

37. R. Harris, Rise and fall of the Eastern Great Indonesian arc recorded by the assembly,dispersion and accretion of the Banda Terrane, Timor. Gondwana Res. 10, 207–231(2006).

38. R. Hall, The palaeogeography of Sundaland and Wallacea since the Late Jurassic. J.Limnol. 72, 1–17 (2013).

39. M. Cloos et al., “Collisional delamination in New Guinea: The geotectonics of sub-ducting slab breakoff” in Geological Society of America Special Papers (GeologicalSociety of America, Boulder, CO, 2005), vol. 400, pp. 1–51.

40. G. L. Foster, D. L. Royer, D. J. Lunt, Future climate forcing potentially withoutprecedent in the last 420 million years. Nat. Commun. 8, 14845 (2017).

41. R. S. W. van de Wal, B. de Boer, L. J. Lourens, P. Kohler, R. Bintanja, Reconstruction ofa continuous high-resolution CO2 record over the past 20 million years. Clim. Past 7,1459–1469 (2011).

42. C. J. Berends, B. de Boer, R. S. W. van de Wal, Reconstructing the evolution of icesheets, sea level and atmospheric co2 during the past 3.6 million years. Clim. PastDiscuss., 10.5194/cp-2020-52 (2020).

43. D. E. Sugden et al., Preservation of Miocene glacier ice in East Antarctica. Nature 376,412–414 (1995).

44. G. H. Haug et al., North Pacific seasonality and the glaciation of North America 2.7million years ago. Nature 433, 821–825 (2005).

45. D. V. Kent, G. Muttoni, Modulation of Late Cretaceous and Cenozoic climate byvariable drawdown of atmospheric pCO2 from weathering of basaltic provinces oncontinents drifting through the equatorial humid belt. Clim. Past 9, 525–546 (2013).

46. Y. Park, N. Swanson-Hysell, F. A. Macdonald, L. Lisiecki, Evaluating the relationshipbetween the area and latitude of large igneous provinces and Earth’s long-termclimate state. EarthArXiv, https://doi.org/10.31223/osf.io/p9ndf (10 June 2019).

47. J. G. Charney et al., Carbon Dioxide and Climate: A Scientific Assessment (NationalAcademy of Sciences, Washington, DC, 1979).

48. T. F. Stocker et al., Climate Change 2013: The Physical Science Basis. Contribution ofWorking Group I to the Fifth Assessment Report of the Intergovernmental Panel onClimate Change (Cambridge University Press, Cambridge, UK, 2013).

49. S. Sherwood et al., An assessment of Earth’s climate sensitivity using multiple lines ofevidence. Rev. Geophys., 10.1029/2019RG000678 (2020).

50. J. C. G. Walker, P. B. Hays, J. F. Kasting, A negative feedback mechanism for the long-term stabilization of Earth’s surface temperature. J. Geophys. Res. 86, 9776–9782(1981).

51. M. E. Raymo, Geochemical evidence supporting T. C. Chamberlin’s theory ofglaciation. Geology 19, 344–347 (1991).

52. R. A. Berner, K. Caldeira, The need for mass balance and feedback in the geochemicalcarbon cycle. Geology 25, 955–956 (1997).

53. R. A. Berner, GEOCARB III: A revised model of atmospheric CO2 over Phanerozoictime. Am. J. Sci., 301, 182–204 (2001).

54. J.-P. Cogne, E. Humler, Trends and rhythms in global seafloor generation rate.Geochem. Geophys. Geosyst. 7, 1–17 (2006).

55. D. G. Van Der Meer et al., Plate tectonic controls on atmospheric CO2 levels since theTriassic. Proc. Natl. Acad. Sci. U.S.A. 111, 4380–4385 (2014).

56. J. M. McArthur, R. J. Howarth, G. A. Shields, “Strontium isotope stratigraphy” inThe Geologic Time Scale, F. M. Gradstein, J. G. Ogg, M. D. Schmitz, G. M. Ogg, Eds.(Elsevier, 2012), pp. 127–144.

57. J. M. Edmond, Himalayan tectonics, weathering processes, and the strontium isotoperecord in marine limestones. Science 258, 1594–1597 (1992).

58. M. E. Raymo, W. F. Ruddiman, Tectonic forcing of late Cenozoic climate. Nature 359,117–122 (1992).

59. A. D. Jacobson, J. D. Blum, L. M. Walter, Reconciling the elemental and Sr iso-tope composition of Himalayan weathering fluxes: Insights from the carbon-ate geochemistry of stream waters. Geochem. Cosmochim. Acta 66, 3417–3429(2002).

60. J. Quade, N. English, P. G. DeCelles, Silicate versus carbonate weathering in theHimalaya: A comparison of the Arun and Seti River watersheds. Chem. Geol. 202,275–296 (2003).

61. L. Oliver et al., Silicate weathering rates decoupled from the 87Sr/87Sr ratio of thedissolved load during Himalayan erosion. Chem. Geol. 201, 119–139 (2003).

62. P. M. Myrow et al., Neogene marine isotopic evolution and the erosion of LesserHimalayan strata: Implications for Cenozoic tectonic history. Earth Planet. Sci. Lett.417, 142–150 (2015).

63. C. L. Colleps et al., Zircon (U-Th)/He thermochronometric constraints on Himalayanthrust belt exhumation, bedrock weathering, and Cenozoic seawater chemistry.Geochem. Geophys. Geosyst. 19, 257–271 (2018).

64. J. A. Higgins, D. P. Schrag, Records of Neogene seawater chemistry and diagenesisin deep-sea carbonate sediments and pore fluids. Earth Planet. Sci. Lett. 357-358,386–396 (2012).

65. W. Si, Y. Rosenthal, Reduced continental weathering and marine calcification linkedto late Neogene decline in atmospheric CO2. Nat. Geosci. 12, 833–838 (2019).

66. M. A. Torres, A. Joshua West, G. Li, Sulphide oxidation and carbonate dissolution as asource of CO2 over geological timescales. Nature 507, 346–349 (2014).

67. R. G. Hilton, J. Gaillardet, D. Calmels, J.-L. Birck, Geological respiration of a moun-tain belt revealed by the trace element rhenium. Earth Planet. Sci. Lett. 403, 27–36(2014).

68. S.-J. Kao et al., Preservation of terrestrial organic carbon in marine sediments off-shore Taiwan: Mountain building and atmospheric carbon dioxide sequestration.Earth Surf. Dyn. 2, 127–139 (2014).

69. S. Carretier, Y. Godderis, T. Delannoy, D. Rouby, Mean bedrock-to-saprolite conversionand erosion rates during mountain growth and decline. Geomorphology 209, 39–52(2014).

70. A. M. Heimsath, W. E. Dietrich, K. Nishiizumi, R. C. Finkel, The soil production functionand landscape equilibrium. Nature 388, 358–361 (1997).

71. P. Davy, A. Crave, Upscaling local-scale transport processes in large-scale reliefdynamics. Phys. Chem. Earth Solid Earth Geodes. 25, 533–541 (2000).

72. D. Lague, The stream power river incision model: Evidence, theory and beyond. EarthSurf. Process. Landforms 39, 38–61 (2014).

25326 | www.pnas.org/cgi/doi/10.1073/pnas.2011033117 Park et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1