Sulphuric acid-mediated weathering on Taiwan buffers geological atmospheric carbon sinks

The chemical composition of the Gaoping River in Taiwan reflects the weathering of both silicate and carbonate rocks found in its metasedimentary catchment. Major dissolved ion chemistry and radiocarbon signatures of dissolved inorganic carbon (DIC) reveal the importance of pyrite-derived sulphuric acid weathering on silicates and carbonates. Two-thirds of the dissolved load of the Gaoping River derives from sulphuric acid-mediated weathering of rocks within its catchment. This is reflected in the lowest reported signatures DI14C for a small mountainous river (43 to 71 percent modern carbon), with rock-derived carbonate constituting a 14C-free DIC source. Using an inverse modelling approach integrating riverine major dissolved ion chemistry and DI14C, we provide quantitative constraints of mineral weathering pathways and calculate atmospheric CO2 fluxes resulting from the erosion of the Taiwan orogeny over geological timescales. The results reveal that weathering on Taiwan releases 0.31 ± 0.12 MtC/yr, which is offset by burial of terrestrial biospheric organic carbon in offshore sediments. The latter tips the balance with respect to the total CO2 budget of Taiwan such that the overall system acts as a net sink, with 0.24 ± 0.13 MtC/yr of atmospheric CO2 consumed over geological timescales.

The chemical composition of the Gaoping River in Taiwan reflects the weathering of both silicate and carbonate rocks found in its metasedimentary catchment. Major dissolved ion chemistry and radiocarbon signatures of dissolved inorganic carbon (DIC) reveal the importance of pyrite-derived sulphuric acid weathering on silicates and carbonates. Two-thirds of the dissolved load of the Gaoping River derives from sulphuric acid-mediated weathering of rocks within its catchment. This is reflected in the lowest reported signatures DI 14 C for a small mountainous river (43 to 71 percent modern carbon), with rock-derived carbonate constituting a 14 C-free DIC source. Using an inverse modelling approach integrating riverine major dissolved ion chemistry and DI 14 C, we provide quantitative constraints of mineral weathering pathways and calculate atmospheric CO 2 fluxes resulting from the erosion of the Taiwan orogeny over geological timescales. The results reveal that weathering on Taiwan releases 0.31 ± 0.12 MtC/yr, which is offset by burial of terrestrial biospheric organic carbon in offshore sediments. The latter tips the balance with respect to the total CO 2 budget of Taiwan such that the overall system acts as a net sink, with 0.24 ± 0.13 MtC/yr of atmospheric CO 2 consumed over geological timescales.
Taiwan is one of the most rapidly uplifting orogens, with erosion rates in the order of 3-6 mm/yr continuously exposing fresh minerals for chemical weathering 1,2 . Together with volcanic activity, metamorphic degassing, and the organic carbon cycle, chemical weathering of minerals exerts a key control on atmospheric chemistry over geologic timescales 3 . Orogenies sustain high rates of physical erosion and are classically invoked as major CO 2 sinks due to the weathering of silicates by carbonic acid 4,5 . While carbonic acid as a weathering agent is widely considered the most important, recent work has highlighted that sulphuric acid weathering of carbonates plays an important role in catchments that contain abundant pyrite 2,6-9 and acts as a major source of CO 2 to the atmosphere over geological timescales [10][11][12] . Pyrite oxidises to sulphuric acid giving rise to river dissolved sulphate 6,8 following the weathering pathway 10 : The trails of evidence of these weathering reaction pathways (2)(3)(4)(5) lead to unique signatures in the dissolved ion load (see Table 1). The theoretical radiocarbon isotopic composition of DIC (hereafter DI 14 C) arising from these pathways are characteristic for different weathering mechanisms. In reaction (2), carbon is sourced primarily from the atmosphere, which exhibits a modern signature, expressed as 100% modern carbon (pMC).
Reaction (3) does not involve carbon and only adds sulphate to the dissolved river load. In reaction pathway (4), 50% of bicarbonate carbon is sourced from the atmosphere and the other half from the radiocarbon-dead lithosphere, hence characterised by a pMC = 50 "fingerprint", while in the case of (5), bicarbonate is entirely derived from the radiocarbon-dead lithospheric source (i.e., pMC = 0). Radiocarbon has considerable advantages over stable carbon isotopic compositions of DIC by incorporating a correction for fractionation and exhibiting lower end-member uncertainty 13,14 . Additionally, the weathering of silicates and carbonates release characteristic assemblages of major cations (calcium, magnesium, and sodium) to the river dissolved load that are further indicative of the mineral species undergoing chemical weathering 5 . Globally, the reaction pathway of silicates and carbonates dictates the net effect of weathering on atmospheric CO 2 3,4 , with the dissolution of carbonates by sulphuric acid acting as a CO 2 source over geological time scales 10 . Using a novel approach, we quantitatively disentangle the inputs of silicate and carbonate weathering via carbonic and sulphuric acid dissolution by determining DI 14 C and dissolved ion compositions within the Gaoping River catchment of Taiwan, leading to new quantitative estimates on the effect of the Taiwan orogeny on atmospheric chemistry.

Study Area and Methods
The Gaoping River covers a length of 170 km and drains 3257 km 2 in southwest Taiwan and is the island's second largest river as measured by sediment discharge 6,15 . Nearly half of its catchment is situated above 1000 m elevation, reaching a maximum of 3997 masl 15 . Within the catchment, (meta) sedimentary rocks ranging from shales to conglomerates of ages spanning from the Mesozoic to Pleistocene are exposed (see appendix for details and references). Due to the monsoonal climate, >90% of river discharge takes place in the flood season (focused in June to October) 15 .
Surface water samples were collected from the Gaoping River catchment during the dry season in February 2017 and 2018 as well as during the wet season in June and October 2017 (Fig. 1). Filtered water samples were enclosed in 12 ml glass exetainer vials without head-space and pre-poisoned with 12 µl of dried HgCl 2 saturated solution. One and six millilitre aliquots were purged with He, acidified with 150 µl 85% H 3 PO 4 , and measured for their DI 13 C and DI 14 C isotope compositions, respectively, using online CO 2 sparging-mass spectrometry setups described elsewhere 16,17 . Radiocarbon values are reported in units pMC 18 . Concentrations of major cations and anions were measured by ion chromatography with details reported in the appendix. Bicarbonate concentrations where calculated by charge balance, following previous Taiwanese river studies 2,8,12,19 . Inputs from rainwater were removed by assuming all chloride is sourced from the atmosphere and by subtracting ions using ratios typical for rainwater (Ca/Na = 0.023; Mg/Na = 0.11; Cl/Na = 1.15; S/Na = 0.06; HCO 3 /Na = 0.004) 5,7 .
Using a constrained linear least-squares approach with the Matlab solver lsqlin, contributions from weathering reactions (2)(3)(4)(5) are assessed from the mineral unit-normalised stoichiometries (see definition below), DI 14 C fingerprints, and ionic ratios characteristic for different weathering reactions listed in Table 1. Here we use α to denote the relative contributions stemming from the carbonic acid weathering of silicates (α Silicate,H2CO3 ) and carbonates (α Carbonate,H2CO3 ) and sulphuric acid weathering of silicates (α Silicate,H2SO4 ) and carbonates (α Carbonate,H2SO4 ). The model output is bound by three equality constraints: (1) the sum of the weathering contributions α i for the mineral-normalised ion concentrations for the four reactions equals 100%, (2) measured and modelled mineral unit-normalised SO 4 must be equal, and (3) the measured and modelled DI 14 C must be equal. Additional constraints are included to find the best least-squares fit between modelled and measured dissolved major ion www.nature.com/scientificreports www.nature.com/scientificreports/ composition based on ratios between calcium, magnesium, and sodium and the sum of sulphate and bicarbonate, which are indicative of different weathering pathways from silicates and carbonates (see Table 1 and appendix for equations). Here, we define a "mineral unit" as (Ca, Mg, 2Na, 2 K)SiO 3 for silicates and (Ca,Mg,2Na,2 K) CO 3 for carbonates. Within each silicate and carbonate mineral unit, calcium is interchangeable with charge balance equivalent amounts of magnesium, sodium, and potassium. For silicates, the relative molar abundances of these cations are typically 0.35 ± 0.25 for Ca/Na and 0.25 ± 0.2 for Mg/Na. For carbonates, Ca/Na and Mg/Na are 60 ± 30 and 30 ± 15, respectively 5,7 , with the uncertainties taken to represent 2-σ uncertainty in the model. Potassium is also present in carbonate mineral lattices in minor quantities (Ca/K ≈ 250-350) 20 , besides its presence in micas and potassic feldspars. Based on these ratios and the ideal mineral unit definition introduced here, the average formula for carbonates and for silicates is Ca 0.66 Mg 0.33 Na 0.011 K 0.0055 CO 3 and Ca 0.30 Mg 0.21 Na 0.85 K 0.14 SiO 3 , respectively. Model output uncertainty was quantified by applying a Monte Carlo approach (10,000 iterations) propagating analytical uncertainties from measured ionic ratios (5% relative 1-σ uncertainty assigned), measured DI 14 C uncertainty (≈1% 1-σ uncertainty based on analytical uncertainty), uncertainty in the cation composition of the weathered minerals, and uncertainty in the radiocarbon signature of the carbonic acid weathering agent. In the case of the latter, an end member signature of 100 ± 5 (1-σ uncertainty) pMC was assigned, with the uncertainty stemming from possible contributions of respired soil organic carbon 21,22 or degraded kerogen, which attenuate the atmospheric carbon signature with the addition of bomb and aged carbon isotopic signatures. Used equations are reported in the appendix. www.nature.com/scientificreports www.nature.com/scientificreports/

Results and Discussion
Over the sampled time interval, pH averaged 7.9 ± 0.2 (n = 13) and bicarbonate represented the major anion (2.3 ± 0.6 mmol/l) followed by sulphate (0.9 ± 0.4 mmol/l) and chloride (0.2 ± 0.3 mmol/l) (supplemental Table 1). For the cations, calcium was the most abundant (1.4 ± 0.4 mmol/l n = 13) followed by magnesium and sodium (both 0.5 ± 0.2 mmol/l), and potassium (0.2 ± 0.2 mmol/l). Carbon isotopic compositions were centred at −5.3 ± 1.3‰ for DI 13 C (n = 13), while DI 14 C values ranged from 43 to 68 pMC (average 54 ± 8 pMC n = 20) (supplemental Table 1). One sample contained 0.02 mmol/l of ammonium that was presumably sourced from anthropogenic inputs and was also characterized by the highest nitrate concentrations suggesting that nitrification induced mineral weathering was ongoing 23,24 . As this additional weathering pathway could not be deconvolved using the approaches devised here, this sample (2018FebAR) was not considered further in this study.
Taiwanese bedrock primarily consists of (metamorphosed) sandstones, siltstones, and claystones, up to a metamorphic grade of amphibolite facies with no reports of evaporitic deposits occurring on the island 25 (see also Supplemental Table 2). In agreement with these geological observations, sulphur isotopes in sulphate 24,26 demonstrate that pyrite oxidation is the predominant source of dissolved sulphate in Taiwanese rivers 6,24,26 . Following the theoretical outline provided in the introduction, rainwater-corrected river data from the Gaoping River are plotted in Fig. 2a showing the mixing of the four end members, which adhere to theoretical expectations in the case of quaternary mixing of solute derived from silicate and carbonate weathering via sulphuric and carbonic acid in the absence of evaporite contributions (Fig. 2a). The relatively high concentrations of chloride ions are similar to those of New Zealand rivers within catchments that receive high amounts of precipitation, which is also attributed to cyclic salt input 27 . Based on the measured DI 13 C values, soil organic matter and kerogen degradation are considered to contribute subordinately to the DIC pool. Similarly low contributions are also inferred for other Taiwanese rivers where most DI 13 C values suggest a predominantly carbonate source for riverine DIC 12 , supporting the assumption in our model that DIC is exclusively carbonate mineral-derived. Figure 2b shows the effect of ternary mixing between reactions (2), (4), and (5). Excess sulphate is introduced into the system via reaction (3), which has no effect on the DI 14 C composition, yet causes the data points to lie outside the boundaries for ternary mixing. By applying the equations outlined in the methods and using the stoichiometric constraints summarised in Table 1 on the data reported in Table S1, the relative contributions stemming from mineral weathering pathways are calculated. These calculations deconvolve the excess sulphate contributions in order to allow quantification of the weathering contributions controlling DI 14 C under the constraint of ionic composition adhering to the quaternary mixing line. The weathering apportionment did not reveal any clear patterns as a function of sampling season or location. Averaged over the catchment and the seasons, and based on molar mineral unit-normalised weathering, α Silicate,H2CO3 and α Carbonate,H2CO3 averaged 34 ± 5% and 0.5% (−0.5/+2%) respectively, while α Silicate,H2SO4 and α Carbonate,H2SO4 averaged 14 ± 9% and 53 ± 8%, respectively (see also Supplemental Fig. S1-S13). Total weathered silicates (α Silicate,H2CO3 + α Silicate,H2SO4 ) contribute 46 ± 8% and carbonates (α Carbonate,H2CO3 + α Carbonate,H2SO4 ) contribute 54 ± 8% to the total dissolved ion load, which is broadly consistent with previous observations of major cation chemistry from the Gaoping River 2,19,28 (see Appendix Fig. S16). Integrated over the Gaoping River catchment, the weathering effect of carbonic acid (α Silicate,H2CO3 + α Carbonate,H2CO3 = 33 ± 6%) and sulphuric acid The Gaoping River data collected over multiple years in relation to the quaternary end-member mixing line. Literature data points where ionic compositions reflect incursion of seawater plot away from the quaternary mixing line. One data point from this study (datapoint with highest mineral unit normalised sulphate) deviated from the mixing line due to high ammonium amounts that are attributed to anthropogenic inputs and therefore wasn't considered further in this study. Removing these anomalies, the data from this study correlates with R 2 = 0.99, and for all studies with R 2 = 0.99. (b) Mineral unit-normalised sulphate versus pMC of DIC. The three mineral weathering reactions producing bicarbonate span a ternary mixing triangle. All of the data lie outside of this mixing ternary due to the addition of sulphate from sulphuric acid weathering of silicates. The latter process, which leaves DI 14 C on the x-axis unaffected, adds sulphate until a theoretical maximum of one mole per weathered mineral unit, representing 100% silicate weathering by sulphuric acid. www.nature.com/scientificreports www.nature.com/scientificreports/ (α Silicate,H2SO4 + α Carbonate,H2SO4 = 67 ± 6%) account for one-third and two-thirds of the total weathering, respectively. The results reveal that carbonates are almost exclusively weathered by sulphuric acid. This is most likely due to the limited abundance of rock carbonate (as siliciclastic metasedimentary units dominate in the catchment of the Gaoping River), a high supply of sulphuric acid, and the rapid reaction kinetics of sulphuric acid with carbonates. The excess sulphuric acid continues to weather silicates. The weathering patterns observed for the Gaoping catchment operate comparably for other Taiwanese catchments including the Liwu 8 , Taimaili 12 , and Chenyoulan 12 rivers. This is evidenced by their adherence to the quaternary mixing trend, encouraging the extrapolation of the results generated here to the entire Taiwan orogenic belt (see appendix).
In the case of Taiwan and major orogenic phases (e.g. Himalayan orogeny), meta-sediments are uplifted bearing the mineral ingredients to both sequester CO 2 and also release it to the atmosphere 4,10,29 . Based on Taiwanese long-term discharge measurements (4.98 × 10 13 l/yr 1 ), the average mineral-unit equivalent normalised dissolved ion load (2.5 ± 0.9 mmol/l n = 63) observed for the Gaoping River catchment (this study 2,19,28 ), and the calculated mineral weathering apportionment with propagated uncertainty, 0.50 ± 0.19 MtC/yr are removed from the atmosphere as a result of carbonic acid weathering of silicates following carbonate precipitation in the oceans. In contrast, 0.40 ± 0.16 MtC/yr are released long term as a result of sulphuric acid weathering of carbonates. While sulphuric acid weathering of silicates does not directly involve carbon, for each silicate mineral unit weathered, ≈0.45 mineral units of carbonate can be precipitated by combining the dissolved cations generated with bicarbonate, which then releases CO 2 to the atmosphere. Based on the ideal mineral unit formula, 0.30 moles of calcium are released upon weathering of one silicate unit and 0.66 moles of calcium are needed for one carbonate unit. Over geological timescales, sulphuric acid weathering of silicates on Taiwan may thus result in the release of 0.09 ± 0.07 MtC/yr (see second panel in Fig. 2). The net release of CO 2 to the atmosphere from chemical weathering of minerals as interpreted in this study is consistent with previous work incorporating considerations of alkalinity and DIC delivered to the ocean 30 . CO 2 release is expected to occur on timescales longer than that of www.nature.com/scientificreports www.nature.com/scientificreports/ carbonate burial (10 5 -10 6 years), yet shorter than pyrite burial (10 7 years) 30 (see Fig. S14 in appendix). In addition to chemical weathering of minerals, the weathering of kerogen (CO 2 source), quantified by river dissolved rhenium, results in the release of 0.27-0.47 MtC/yr 29 . To counterbalance this, the burial of terrestrial biospheric carbon in adjacent ocean sediments 31,32 , quantified previously using radiocarbon as a tracer for modern sedimentary organic matter, removes 0.5-0.6 MtC/yr 32,33 . In terms of overall carbon balance, the long-term weathering effect of silicates, carbonates, and kerogen on land, together with offshore burial of terrestrial biospheric organic carbon, results in the removal of 0.24 ± 0.13 MtC/yr from the atmosphere. However, the total weathering effect on land acts as a net source of CO 2 to the atmosphere, releasing 0.31 ± 0.12 MtC/yr (see Fig. 3). These quantities illustrate the fine balance between weathering mechanisms, and of the need to disentangle underlying processes for accurate assessment of the net effects of weathering budgets on atmospheric chemistry. The Taiwan orogeny shows that the weathering of metasedimentary catchments rich in pyrite and kerogen significantly influence the carbon budget of an orogenic entity.
The carbon budget presented here should be viewed conservatively, as two additional sources of CO 2 to the atmosphere, authigenic silicate formation 34 and metamorphic degassing 35 , are not considered. Considerable uncertainty lies in the former, as this process is limited by reactive silica availability in the oceans 34 . In the case of the latter process, the emission of CO 2 and volatile hydrocarbons (e.g. CH 4 ) further offset net carbon sequestration as a result of earth surface geochemical pathways and may tip the integrated surface and deep Earth geochemical cycles of an orogeny towards acting as a carbon source to the atmosphere 35,36 . Changes through time can be expected with erosional events 2 as a result of major typhoons 37 and mineral weathering pathways modified by fertiliser usage 23 . From the available data, the key variable compensating the release of CO 2 to the atmosphere due to chemical weathering is the export and preservation efficiency of terrestrial biospheric organic carbon in offshore sediments -a particularly dynamic component that is responsive to tectonic 38 , climatic 31,39 and anthropogenic 40 influences.
The relative importance of sulphuric acid-mediated weathering depends on the presence and weathering of pyrite in exposed lithologies. The combined contributions of the Gaoping, Mackenzie (northern Canada), and Jialing (southern China) rivers deliver around 12% of sulphide-derived sulphate to the oceans, while their catchments encompass less than 2% of the Earth's land surface 7 , highlighting the uneven distribution and localized occurrence of sulphuric acid-mediated weathering. In metasedimentary catchments such as those of the Gaoping River on Taiwan, weathering by sulphuric acid may contribute significantly to mineral dissolution. Here, it accounts for approximately two-thirds of total mineral dissolution, with carbonates almost entirely dissolved by sulphuric acid. The chemical weathering carbon budget of Taiwan supports the hypothesis that sulphuric acid mediated weathering in tandem with orogenic activity may well have affected atmospheric chemistry over geologic time 10 . Since the rise in atmospheric oxygen in the Precambrian, pyrite oxidation and sulphuric acid evolved therefrom has affected ocean chemistry, biologic activity, and atmospheric oxygen concentrations [41][42][43] . Mineral weathering by sulphuric acid has important implications for atmospheric CO 2 inventories over geologic time 10,30 , with CO 2 uptake and release governed by the fine balance between organic carbon burial in sediments and silicate, carbonate, and kerogen weathering controlled by the activity of oxygen, carbonic acid, and sulphuric acid. The total effect of orogenic activity on atmospheric chemistry remains a matter for discussion 36 , yet it is clear that the presence of sulphuric acid and its involvement in carbonate and silicate mineral weathering acts as a strong buffer on atmospheric CO 2 uptake.