Amplified Production and Export of Dissolved Inorganic Carbon During Hot and Wet Subtropical Monsoon

Understanding the origins and processes of riverine dissolved inorganic carbon (DIC) is crucial for predicting the global carbon cycle with projected, more frequent climate extremes yet our knowledge has remained fragmented. Here we ask: How and how much do DIC production and export vary across space (shallow vs. deep, uphill vs. depression) and time (daily, seasonal, and annual)? How do the relative contributions of biogenic (soil respiration) and geogenic (carbonate weathering) sources differ under different temperature and hydrological conditions? We answer these questions using a catchment‐scale reactive transport model constrained by stream flow, stable water isotopes, stream DIC, and carbon isotope data from a headwater karstic catchment in southwest China in a subtropical monsoon climate. Results show climate seasonality regulates the timing of DIC production and export. In hot‐wet seasons, high temperature accelerates soil respiration and carbonate weathering (up to a factor of three) via elevating soil CO2 and carbonate solubility, whereas high discharge enhances export by two orders of magnitude compared to cold‐dry seasons. Carbonate weathering is driven more by soil CO2 than water flow. At the annual scale, 92.9% and 7.1% of DIC was produced in shallow and deep zone, respectively, whereas 64.5% and 35.5% of DIC was exported from shallow and deep zone, respectively. These results highlight the uniqueness of subtropical karst areas as synchronous reactors and transporters of DIC during the hot‐wet monsoon, contrasting the asynchronous production and export in other climate regions. A future hotter and wetter climate with more intensive storms in the region may further intensify DIC production and export, accentuating the potential of subtropical karst regions as global hot spots for carbon cycling.

Riverine DIC originates from two major sources: biogenic carbon via soil respiration and geogenic carbon via chemical weathering (Campeau et al., 2017;Gaillardet et al., 1999;Mayorga et al., 2005).Soil respiration transforms organic matter into soil CO 2 , which further dissolves in water to become DIC and reduces pH.Carbonate weathering consumes soil CO 2 , increases DIC concentrations, often elevates pH, and further modifies the speciation of DIC (bicarbonate  HCO − 3 , carbonate  CO 2− 3 , and dissolved CO 2 (aq)) (Keller, 2019;Z. Liu et al., 2018).The potential sources of DIC can be deciphered from the distinct signatures of stable carbon isotopes in δ 13 C, with a typical range of −26‰ to −18‰ for organic matter from C3 plants and 0‰ for carbonate minerals of marine sedimentary origin (Land et al., 1980).The relative contributions of different carbon sources can therefore lead to variations in stream δ 13 C from −26 to 0‰.For example, in South China with large karst area (∼550,000 km 2 ), ∼35%-60% of DIC concentrations was estimated to derive from carbonate weathering, much higher than the percentage in non-karst systems (Qin et al., 2019;Zhong et al., 2017).A generic ratio of 50% (i.e., one half of DIC) has been commonly used to designate the contribution from carbonate weathering in karstic catchments and at regional scales (Z.Liu et al., 2018;S. Zeng et al., 2019).
Rates of soil respiration and carbonate weathering are regulated by a multitude of interacting drivers.Soil respiration rates generally increase with temperature but often peak at intermediate soil water content of 50%-70%, depending on organic matter characteristics, soil types, and biological activity (Davidson & Janssens, 2006; Z. F. Yan et al., 2018).Low soil water content facilitates CO 2 emission back to the atmosphere, and therefore reduces soil CO 2 and DIC concentrations (Ilstedt et al., 2000).For carbonate weathering, in addition to the general influence of subsurface spatial heterogeneities (Wen andLi, 2017, 2018), laboratory studies in well-controlled, small-scale reactors have identified the direct control of temperature on reaction affinity, kinetics, and thermodynamics (Kirstein et al., 2016;Plummer et al., 1978).Recent field studies highlighted the indirect control of climate conditions on carbonate weathering via soil CO 2 that in turn depends on water availability, microbial activity, and root respiration (Calmels et al., 2014;Romero-Mujalli et al., 2019a;Wen, Sullivan, et al., 2021;White & Blum, 1995).Global carbonate weathering rates are often estimated assuming carbonate weathering at equilibrium with dependence on soil CO 2 (Gaillardet et al., 2019;Romero-Mujalli et al., 2019b).These multiple, interacting drivers challenge the identification of most influential factors and the assessment of the impact of individual factors.
Understanding the sources, timing, and magnitude of riverine DIC export presents another challenge (Duvert et al., 2020;Qin et al., 2019).During wet periods, especially during storms, biogenic dissolved carbon is primarily transported through shallow soils, promoting lateral export to streams (Qin, Li, et al., 2020;Wen et al., 2020;Zhong et al., 2017).Additionally, soil water rich in CO 2 can recharge rapidly into the deeper subsurface, increasing water-rock interactions and geogenic carbon fluxes into the stream (Clow & Mast, 2010;Wen et al., 2022).During dry times, deeper groundwater becomes the major venue for DIC export to streams (e.g., Stewart et al., 2022).The export of DIC is also influenced by factors such as mineral abundance, weathering agents (e.g., carbonic acid), solute transport, and water residence time (Raymond & Hamilton, 2018).Given the limited accessibility of the subsurface, it remains challenging to pinpoint when, where, and how much stream DIC comes from different subsurface zones and how the flow paths and sources change with flow regimes, among other conditions.
Here we ask the questions: How and how much do DIC production and export vary across space (shallow vs. deep, uphill vs. depression) and time (daily, seasonal, and annual)?How do the relative contributions of biogenic (soil respiration) and geogenic (carbonate weathering) sources differ under changing temperature and hydrological conditions?Previous work has shown that in a forested catchment under temperate climate, the production and export of dissolved organic carbon (DOC) are asynchronous, with production mostly occurring in dry summers whereas its export primarily in wet springs and winters with high discharge (Wen et al., 2020).Here we draw upon the recent development of a watershed-scale reactive transport model (RTM) BioRT-Flux-PIHM (L.Li, 2019;Zhi et al., 2022).The model integrates hydro-biogeochemical processes and can distinguish the role of individual processes and elucidate mechanisms at the catchment scale (L.Li et al., 2017;Steefel et al., 2015).The recently developed isotope-enabled feature in RTMs complements the traditional mixing approach that often neglects the complex interactions of flow and biogeochemical reactions (Druhan, Guillon, et al., 2021;Druhan, Lawrence, et al., 2021).The study catchment is Chenqi, a catchment within the karst Critical Zone observatories in southwestern China.Field data was used to calibrate BioRT-Flux-PIHM, from which we quantified flow paths and reaction rates over time and space and provided insights on hydro-biogeochemical processes.The catchment offers a test bed to understand inorganic carbon cycling and biogeochemical processing in limestone terrain and under subtropical monsoonal climate.

Study Site: A Karstic Catchment With a Subtropical Monsoonal Climate
Chenqi is located in central Guizhou, China (Figure 1), experiencing a subtropical monsoonal climate with the mean annual precipitation of 1,308 ± 244 mm and mean annual temperature of 15.4 ± 0.5°C.The catchment has an underground stream in a peak-depression karstic landscape.Encircled by star-distributed conical hills, it can be divided into two units (Chen et al., 2018;Zhang et al., 2019): the flat depression areas at low elevation (1,320-1,340 m) and the steeper uphills at high elevation (1,340-1,520 m).The uphill (0.88 km 2 , ∼83% of the catchment area) is covered by thin soils (<0.5 m), deciduous forest, and shrub-grassland (Figure 1).The depression area (0.37 km 2 ) has much thicker soils (0.5-2.0 m) and is dominated by farmland (corn and rice paddy).Carbonate bedrock, predominantly limestone, situates above a formation of impermeable marlite and spreads extensively with a thickness ranging from 150 to 200 m (Zhang et al., 2011).Quaternary soils are irregularly developed on carbonate bedrock, with outcrops of carbonate rocks covering 10%-30% of the catchment.

Measurements
All measurements were undertaken at the outlet of the underground stream (Figure 1).Discharge was measured daily with v-notch weirs from 1 September 2007 to 31 December 2008.Stream chemistry samples were collected monthly between September 2007 and August 2009.Rainwater chemistry was measured for individual precipitation events.Water samples were filtered through 0.45 μm using Nylon syringe filters, transferred to acid-washed plastic bottles, and stored immediately in darkness at 4°C prior to lab analysis.
The pH was measured at the time of sampling by a hand-held water quality meter (WTW Multiline P3 pH/LF-SET).Alkalinity was measured by in situ titration with the Aquamerck Alkalinity Test and Hardness Test, with analytical precision of ±0.1 mM C. Major cations (e.g., Ca) concentrations were measured using inductively coupled plasma optical emission spectrometer.Stable isotopic ratios of carbon in DIC (δ 13 C) was measured using GV Isoprime continuous flow mass spectrometer (CF-IRMS).All δ 13 C values were reported as per mil (‰) deviations from the standard Vienna Pee Dee Belemnite, with a standard deviation (1σ) of <0.2‰.Stable isotopic ratios of oxygen in water (δ 18 O) was analyzed using CF-IRMS and calibrated to the Vienna Standard Mean Ocean Water, with a standard deviation less than 0.2‰.Further details about sampling and measurements are available in previous publications (Hao et al., 2019;Zhang et al., 2011;Zhao et al., 2010Zhao et al., , 2015Zhao et al., , 2018)).

Watershed Reactive Transport Model
We used the recently developed model BioRT-Flux-PIHM (Zhi et al., 2022).The code has been applied to understand solute transport, chemical weathering, and microbe-mediated redox reactions at the catchment scale (Bao et al., 2017;Saberi et al., 2021;Wen et al., 2020;Zhi & Li, 2020;Zhi et al., 2019).The code includes three modules: the surface hydrological module PIHM, the land surface module Flux, and the multicomponent reactive transport module BioRT.Flux-PIHM calculates water variables (e.g., water storage, soil saturation, and water table depth) in shallow soil water and deep groundwater zones (Figure 2).The soil water zone (SZ) is most conductive to water flow (e.g., interflow) and responsive to hydroclimatic forcing; the deep groundwater zone (deeper zone [DZ]) is the deeper, less weathered, and lower permeability layer that harbors the old groundwater.The Flux-PIHM module uses daily meteorological data (precipitation, temperature, radiation, etc.) as input, simulates hydrological processes, and outputs daily evapotranspiration (ET), surface runoff Q surf , shallow soil water interflow Q SZ , and the deeper groundwater flow Q DZ .The BioRT module uses water and temperature output from Flux-PIHM at each time step and integrates transport and biogeochemical reactions through the governing mass conservation equations.
The model uses watershed characteristics including topography (e.g., soil depth, surface elevation), subsurface properties in SZ and DZ (e.g., hydraulic conductivity, porosity, macropore fraction, van Genuchten parameters), and vegetation properties to set up the domain.Geochemical characteristics include the geochemistry of rainwater, soil water, deep groundwater, soils, and bedrocks as initial conditions, as well as their thermodynamics and kinetics of geochemical reactions.The model can produce outputs at the timescales of hours to years, including water state variables (e.g., soil saturation, water table depth), water fluxes (e.g., Q), geochemical reaction rates, and solute concentrations in the SZ, DZ, and stream.

Reaction Networks
Here DIC, the sum of CO 2 (aq),  HCO − 3 , and  CO 2− 3 , is assumed to derive primarily from soil respiration and carbonate weathering (Figure 2).Soil respiration includes the heterotrophic respiration, or the decomposition of soil organic carbon (OC), and the autotrophic root respiration (Ekblad & Högberg, 2001;Jones et al., 2004).Soil CO 2 can diffuse vertically into the atmosphere and/or dissolve in water and export to streams as DIC.We do not have data to track CO 2 emission back to the atmosphere.Therefore, Reaction 1 in Table 1 represents the net production of CO 2 from soil respiration that eventually dissolves in water and contributes to DIC, which likely underestimates the total rates of soil respiration as we ignored the vertical soil CO 2 fluxes.With abundant OC and O 2 in soils serving as electron donors and acceptors, a simple rate law is applied to describe the net CO 2 production rate from soil respiration (  p_bio ): where k bio is the kinetic rate constant of the net biogenic CO 2 production (mol C/m 2 /s); and A is a lumped "surface area" (m 2 /m 3 ) that quantifies OC content, biomass, and root abundance.The function f(T) describes the rate , where Q 10 quantifies the rate increases with T and is set at 4.0, which is within the typical range for karst areas (Davidson & Janssens, 2006;Wang et al., 2020).The function f(S w ) describes the dependence on soil saturation (S w ) with a threshold form (Z. F. Yan et al., 2018).
Soil CO 2 acidifies water and accelerates carbonate weathering (Reaction 2-6).Carbonate weathering follows a transition state theory rate law (Lasaga, 1984): where the local carbonate rate r cal (i.e., local DIC production rate from geogenic source  p_geo ) is determined by the kinetic rate constant (k cal , mol/m 2 /s), the activity of hydrogen ion (    + ), the carbonate bulk surface area (A, m 2 /m 3 ), and the extent of disequilibrium represented as the ratio of the ion activity product (IAP) and the equilibrium constant (K eq ).The reaction parameters were calibrated using water chemistry data (Sections 2.6 and 3.2).
To differentiate the contribution from soil respiration and carbonate weathering, carbon isotopes ( 12 C and 13 C isotopes) were explicitly simulated as independent, individual "species" that are distributed in appropriate ratios based on δ 13 C values and radiocarbon compositions measured at Chenqi (Hao et al., 2019;Qin et al., 2022).The stoichiometry in soil respiration and carbonate weathering (Reaction 1 and 6 in Table 1) therefore reflects the distribution of carbon isotopes, resulting in a reported δ 13 C value of −22.0 and 0‰, respectively.Note that the value of −22.0‰ refers to the observations in soil solution as the eventual outcome of soil CO 2 dissolution and evasion (Campeau et al., 2017;Qin et al., 2019Qin et al., , 2022)).Values of δ 13 C can vary via complex interactions between plants, microbes, and environmental conditions.The two end-member δ 13 C values simplify isotope variations in natural systems but capture the key characteristics between biogenic and geogenic sources without overburdening the model with too many parameters.The equilibrium expressions of DIC (Reaction 2-5) are repeated twice to keep track of both 12 C and 13 C isotopes.This multi-isotope simulation approach has been benchmarked and applied in 1D soil and karst cave systems (Druhan, Guillon, et al., 2021;Druhan, Lawrence, et al., 2021).

The Four Grid-Cell Domain Setup
Following Figure 1b, we conceptually simplified the watershed into four prismatic grid cells representing the depression and uphill at each side of the stream (Figure 2).The areas of the depression and uphill grid cells are All parameters refer to the condition at 25°C.b Values of K eq were interpolated using the EQ3/6 database (Wolery et al., 1990), except Reaction 3, 5, and 6.The K eq values related to 13 C in Reaction 3, 5, and 6 were from Druhan, Guillon, et al. (2021) and Druhan, Lawrence, et al. (2021).c The kinetic rate parameters in Reaction 1 and 6 for soil respiration and carbonate weathering (i.e., biogenic and geogenic sources of DIC) were from Palandri and Kharaka (2004) and Plummer et al. (1978), respectively.d The specific surface areas were calibrated by fitting stream chemistry data.The calibrated surface areas are much lower than field measurements because they only reflect the effective surface area that are truly contacting and reacting with water (Wen & Li, 2017).e The stoichiometry in soil respiration and carbonate weathering reflects the distribution of carbon isotopes and is calculated based on field measurements at the study site (Hao et al., 2019;Qin et al., 2022).
0.18 and 0.44 km 2 , respectively.The corresponding slopes are 7° and 24°, averaged based on the digital surface elevation map.The hydrometeorology data are from the China meteorological forcing data set at hourly frequency (He et al., 2020); the leaf area index is from the Moderate Resolution Imaging Spectroradiometer; soil properties are from the soil survey database from the Puding Karst Ecohydrological Observation (Chen et al., 2018).
Each prismatic land element within the modeling domain was discretized into two layers to represent the shallow soil and the deeper underlying bedrock (the SZ and DZ in Figure 2), based on characterization and measurement data (Bai & Zhou, 2020;Chen et al., 2018;Qin, Li, et al., 2020;Qin et al., 2022;Zhang et al., 2019).Soil depths (i.e., the SZ) were set at 0.20 and 1.75 m for the uphill and depression, respectively, while corresponding depths of the deep groundwater zone (i.e., the uphill and depression DZ) were set at 125.0 and 28.0 m.The OC and root content in depression soils were set at 8% (v/v solid phase) compared with 4% of uphill soils and 0.8% of the DZ.Considering the widely observed outcrops of carbonate rocks across the catchment, carbonate abundance was set at 50% (v/v solid phase) in SZ and 75% in DZ.The subsurface matrix properties were parameterized based on the reported median values at Chenqi (Zhang et al., 2011).

Model Calibration
We first calibrated the hydrological module to reproduce stream discharge, followed by the calibration of biogeochemical module using stream chemistry, stable water isotope (i.e., nonreactive tracer 18 O), and carbon isotope data.Stream discharge (daily) and chemistry (monthly) data from September 2007 to August 2009 were used for model calibration.The year 1977-2006 was used as spin-up until a "steady state" for both discharge and water chemistry was reached.The "steady state" here refers to a state where the inter-annual difference between stored mass within the catchment is less than 5% of the total mass.To reduce the potential bias from monthly chemistry data, such as missing dynamics in response to intense precipitation, we further validated the model using high-frequency (hourly) stream discharge and chemistry data collected during rainfall events in June 2018 from Qin, Ding, et al. (2020) and Qin, Li, et al. (2020).The model performance was evaluated using the Nash-Sutcliffe efficiency (NSE), RMSE-observations standard deviation ratio (RSR), and percent bias (PBIAS) (Gupta et al., 1999;Nash & Sutcliffe, 1970;Singh et al., 2005).We used the general satisfactory range of NSE > 0.5, RSR ≤ 0.7, |PBIAS| < 25% (Moriasi et al., 2007), similar to those used in Wen et al. (2020).
Two types of parameters were adjusted (Table S1 in Supporting Information S1), including Zilitinkevich coefficient related to the land-atmosphere interaction, and subsurface properties including van Genuchten parameters α and n, porosity, hydraulic conductivity, and macropore fraction.To reproduce stream chemistry data (pH, Ca, alkalinity, and δ 13 C), we adjusted the specific surface area for the DIC production rates from soil respiration and carbonate weathering (Reaction 1 and 6).Because not all water is fully in contact with OC, roots and carbonate minerals, the calibrated surface area here (Table 1) represents the effective solid-water contact area at the catchment scale, and can be orders of magnitude lower than those measured in labs and fields (Pennell et al., 1995;Wen & Li, 2017, 2018).The stream alkalinity and δ 13 C data together constrained the contribution from biogenic and geogenic sources.

Water Transit Time, Stream Concentrations, and Catchment-Scale Rates
Water transit time.We estimated the transit time of water from the rainfall to the catchment outlet via a reactive tracer-based approach developed by Wen et al. (2022).The tracer concentration in the rainfall (C 0 at 0.9 mol/L) is assumed to decay following zero-order kinetics    = −decay , where k decay is the decay rate (=0.1 mol/m 3 /yr) and t is the time (year).Consequently, the mean transit time τ i can be determined through the corresponding tracer concentration C i , calculated as = 0 − decay .Following the above equation, the mean transit time in the shallow soil (τ SZ ), deep groundwater (τ DZ ) and stream (τ stream ) were determined using the tracer concentrations coming out of the SZ (C SZ ), DZ (C DZ ), and stream (C stream ), respectively.
Stream DIC from distinct sources.Stream DIC concentrations (C) reflect the contributions of different source waters from distinct flow paths including surface runoff, soil water flow from SZ and deep groundwater flow from DZ (Stewart et al., 2022).The source water chemistry is influenced by both soil respiration and carbonate weathering in these subsurface zones.To differentiate their contribution to stream DIC, we carried out two additional simulations based on the calibrated case: (a) the transport-only model (i.e., no reactants) without reactions (C r ); and (b) soil respiration only without carbonate weathering (C bio+r , k cal = 0 in Reaction 6).Comparison of these two cases with the calibrated base case with both soil respiration and carbonate weathering can quantify the effects of different reactions.Thus, the DIC concentrations from soil respiration and carbonate weathering can be estimated using C bio = C bio+r − C r and C geo = C − C bio+r , respectively.
Production and export rates.Reaction rates R at the catchment scale were calculated as the sum of local rates in individual grid cells (r i ) multiplied by their corresponding volume (V i ).The DIC production rate from biogenic source is R p_bio =  ∑ (p_bio, ×  ).The DIC production rate from geogenic source is the carbonate weathering rate: R p_geo = R cal = ∑(r cal,i × V i ).The overall DIC production rate (R p ) is the sum of R p_bio and R p_geo .The DIC input from the rainfall R r (mol/d) is the precipitation rate (m/d) times the rainfall DIC concentration (5.0 × 10 −2 mol C/m 3 ) and the catchment drainage area (m 2 ).
The overall DIC export rate R e is the product of discharge and the DIC concentration at the stream outlet, including contributions from rainfall, soil respiration, and carbonate weathering.The export rate from these sources was calculated using their corresponding export concentrations (C r , C bio , and C geo ).

Water Balance and Hydrological Processes From Daily to Annual Time Scales
The annual precipitation was 1,361 mm from 1 September 2007 to 31 August 2008 and 1,462 mm from 1 September 2008 to 31 August 2009.The average discharge approximated 10.0 mm/d in the hot and wet seasons (summer and fall), compared to about 0.2 mm/d in the cold and dry seasons (spring and winter) (Figure 3a).Stream discharge was highly responsive to intense precipitation events in the hot and wet seasons.The model captured the dynamics of daily discharge, with NSE, RSR, |PBIAS| of 0.57, 0.65, and 0.01, respectively.
The model estimated that the average annual ET is 442 mm, about 31.3% of precipitation, with the rest contributing to stream discharge of about 921 mm.The discharge originated from three flow paths: surface runoff (Q surf ), shallow soil water (Q SZ ), and deep groundwater (Q DZ ), contributing about 0.4%, 75.6%, and 24.0%, respectively, consistent with field estimation (C.Zeng et al., 2014).During dry seasons, discharge mostly originated from the old Q DZ at a water age of around 30 years; during wet seasons, the relatively young Q SZ (<1 year old) responded faster than Q DZ and contributed to up to ∼95% of streamflow during intense rainfall events (Figures 3b and 3c), leading to young streamflow (∼2 years).The simulated daily soil saturation and T followed the temporal trends of discharge and ET, respectively (Figure 3d).

Biogeochemical Dynamics: Stream Water Chemistry Varies Most in the Hot and Wet Monsoon
Stream chemistry varied substantially between cold-dry and hot-wet seasons.Due to large variations in the rainfall isotopic signature, stream δ 18 O values were more variable during the hot-wet season, alternating between enriched and depleted from −9.5 to −6.6‰ (Figure 4a).In contrast, stream δ 18 O during the cold-dry season remained relatively constant around −8.0‰.Stream pH, Ca, alkalinity and δ 13 C values varied in similar ways.They were relatively high during the cold-dry season (7.4 ± 0.2, 3.0 ± 0.7 mM, 4.2 ± 0.2 mM C, and −11.1 ± 1.0‰), and dropped during the hot-wet season (7.2 ± 0.3, 2.7 ± 0.4 mM, 3.7 ± 0.3 mM C, and −13.9 ± 1.7‰; Figures 4b-4d  and 4f).The model captured the general trends of stream isotope (δ 18 O and δ 13 C), pH, Ca, and alkalinity data (Figure 4; See Figure S1 in Supporting Information S1 for scatterplots comparing model outputs to measurements), with NSE, RSR, |PBIAS| values of 0.51, 0.69, and 0.20, respectively.The simulated stream chemistry exhibited more temporal dynamics than the monthly measurements in 2007-2009 because measurements occurred on non-rain days and do not provide the full picture of stream chemistry during intense rainfalls.The consistency between model output and hourly data from the 2018 summer (Figures 4d and 4f) further validated the model performance in wet seasons.The stream water chemistry reflected the mixing of SZ and DZ, as demonstrated by model outputs (lines in Figure 4).Solute concentrations were stable in SZ (dark dash lines) and fluctuated significantly in SZ (light dash lines), with high values in the cold-dry spring and winter and low values in rainfall events and in the hot-wet summer and fall.The variation of non-reactive tracer 18 O was pronounced in rainwater and was much damped in groundwater.Relatively low stream concentrations of solutes and relatively negative δ 13 C during hot-wet seasons were attributed to greater contributions from SZ.

Spatial Patterns of Reaction Rates and Water Chemistry: Uphill Versus Depression and Shallow Versus Deeper Zones (SZ and DZ)
Spatial patterns of modeled variables at uphill versus depression in SZ generally followed similar cold-dry and hot-wet seasonal dynamics.In the cold-dry season (from December to May), soil saturation at depression was constantly around 0.62, about 30% higher than that at uphill (Figure 5a), whereas the difference in soil T between the two zones was much smaller (<2°C).Both uphill and depression soils were hydrologically disconnected from the stream, with the lateral flow Q SZ less than 0.1 mm/d (Figure 5b).Spatial distributions of soil respiration rates r p_bio in SZ resembled that of OC and soil saturation, and were higher in depression due to higher water content.With minimal water flow and export to the stream, high r p_bio in depression led to the accumulation and high concentrations of soil CO 2 (aq) and H + , reaching a maximum of 5 times higher than that of the uphill at the end of dry periods (Figures 5e-5g).Spatial patterns of Ca and DIC followed that of soil CO 2 , indicating its predominant control on carbonate dissolution.High soil CO 2 and low pH effectively reduced bicarbonate concentration and increased carbonate solubility, leading to higher weathering rates r cal in depression (Figures 5d and 5h)., (e) pH, (f) dissolved inorganic carbon (DIC), (g) aqueous and gaseous soil CO 2 , and (h) ion activity product/K eq .Gaseous soil CO 2 (g) in panel (g) was estimated using Henry's law.Soil CO 2 and r p_bio were high in depression soils that had relatively high organic carbon (OC) and water content, leading to correspondingly high H + (lower pH), DIC and carbonate weathering rates compared to the uphill.These spatial contrasts were more significant during the cold-dry season when the uphill and depression was mostly disconnected and most products accumulated in soils.The DZ with more abundant carbonate had much less temporal and spatial variations because of less abundant OC content, slower groundwater flow, and longer water transit time.Compared to uphill-depression differences, the shallow SZ versus deeper DZ contrasts were much more pronounced.For example, rates in SZ and DZ can reach several orders of magnitude differences (c, d).
In the hot-wet season (June to November), soil T increased to ∼18°C in both depression and uphill.Soil saturation at depression increased by 20%-40% responding to intense rainfall events while variations in uphill were less than ∼10%.R p_bio increased by over four times across the entire catchment but dropped sharply at depression under very wet conditions when soil saturation exceeded 0.7.Soil CO 2 and DIC concentrations dropped by a factor of more than four times and was relatively homogeneous from uphill to depression.This is because the fast soil water flow (Q SZ, uphill and Q SZ, depression in Figure 5b) exported soil CO 2 and DIC from uphill into the stream rapidly, leading to less accumulation and lower concentrations.Correspondingly, carbonate weathering rates were high as the water was at disequilibrium with low IAP/K eq .
The DZ showed much less temporal and spatial variations in water chemistry (dashed line) than the SZ (solid line).With low OC in DZ, soil respiration rates r p_bio were about two orders of magnitude lower than that in SZ.The deeper flows (Q DZ, uphill and Q DZ, depression in Figure 5b) were relatively constant compared to shallow flow and had higher concentrations of CO 2 , H + , and DIC (Figures 5e-5g).The carbonate in DZ was in contact with water that was close to equilibrium (IAP/K eq = ∼0.95, Figure 5h) and had low r cal at ∼10 −5 mol C/m 3 /d (Figure 5d).With faster groundwater flow into stream (Q DZ, depression , dark dashed line in Figure 5b), soil CO 2 (aq), H + , Ca, and DIC concentrations at depression were slightly lower than those at uphill.The soil respiration r cal at the depression DZ fluctuated more and were higher in wet compared to dry seasons, because of the elevated, fluctuating recharge that brought soil CO 2 -enriched water from SZ in wet seasons.Compared to these uphill-depression differences, the SZ versus DZ contrasts were much more pronounced.For example, rates in SZ and DZ can reach several orders of magnitude differences (Figures 5c and 5d). ) and (B) corresponding rates of soil respiration (R p_bio , R e_bio ) and carbonate weathering (R p_geo , R e_geo ) in SZ and DZ.The DIC input from rainfall (R in_r ) was negligible, about 1.0% of R p .Temporal patterns of R p , R p_bio and R p_geo varied within an order of magnitude and generally followed that of soil T, whereas R e , R e_bio and R e_geo mostly followed that of discharge, increasing by up to two orders of magnitude with Q.The relative magnitude of production and export rates in different seasons determines the temporal patterns of the overall DIC storage, which peaks in cold-dry seasons with minimal flow and reached minima in the hot moments of export during hot-wet periods.Rates and mass in SZ generally fluctuated more and were one-to-two orders of magnitude higher than those in DZ.

Amplified DIC Production and Export in the Hot and Wet Monsoon
The catchment-scale rates varied significantly over time.The daily production rate R p (= R p_bio + R p_geo , 8.6 × 10 3 mol C/d) was much higher than the daily rainfall input R in _ r (9.3 × 10 1 mol C/d) such that rainfall input was negligible.R p_bio and R p_geo were 5.1 × 10 3 and 3.5 × 10 3 mol C/d, respectively, accounting for 59.1% and 40.9%, respectively, of the overall DIC production rates R p .These reactions mostly occurred in SZ (Figures 6B1 and 6B2), typically one to two orders higher than those in DZ.They followed the general trend of soil T, except that R p_geo also increased during big storms as excessive water flow drove carbonate dissolution far from equilibrium.The export rate R e (= R e_r + R e_bio + R e_geo ), R e_bio , and R e_geo primarily followed the seasonality of discharge.During the dry-to-wet transition, DIC was flushed out mostly through SZ as the catchment became wetter (Figures 6B3 and 6B4), substantially reducing the overall DIC storage to ∼1.35 × 10 7 mol.During the cold dry periods, although R p was relatively low, the DIC pool increased, reaching as high as ∼1.55 × 10 7 mol due to the minimal flow and export at the time.During the longer dry periods (Q << 1.0 mm/d) from November 2007 to May 2008, DIC accumulated more compared to the same period in 2008-2009.The biogenic rates varied significantly in SZ with seasonal variations but were almost constant in DZ (Figure 6B1).The geogenic rates had similar variations in SZ but varied also in DZ (Figure 6B2), especially during the wet season, following the temporal patterns in SZ.The annual DIC storage changed by about 2.6 × 10 5 mol between the two years, about 0.4% of the overall DIC production, indicating a general mass balance at the catchment scale.(d-f) R e and daily soil T, soil saturation S w , mean water transit time τ stream , and discharge.All dots are daily model outputs; the symbol size in panels (a, b) represents the magnitude of S w and T. R p_bio , R p_geo , and R p all generally increased with soil T, whereas R e_bio , R e_geo , and R e all depended more on τ stream and S w , except the cold-dry periods with low soil T and S w (<15°C and 0.53).In the cold-dry periods, R p_geo decreased with decreasing S w even when T increased; export rates were relatively constant because stream water was mostly contributed by the old deeper zone.

Production Rates Depend More on Temperature Than Water Content
The daily production rates R p_bio , R p_geo , and R p generally depended more on soil T (R 2 > 0.80) than on water content, as illustrated by water variables (S w , discharge and τ stream with R 2 < 0.20) (Figures 6 and 7).All production rates correlated positively with soil T, and the R 2 values decreased from 0.98 to 0.94 to 0.81 for R p_bio , R p , and R p_geo , respectively.At low T and S w conditions (<15°C and 0.53), R p_geo depended more on S w than T and decreased with decreasing S w even when T increased (circled region in Figures 7a and 7b).Export rates correlated more with S w , Q, and τ stream than soil T, rising by over two orders of magnitude as Q increased (or τ stream decreased) by about two orders of magnitude.And under low S w (<0.53) when most stream DIC was contributed by the old DZ (∼30 years), DIC export rates were almost constant at ∼10 3 mol C/d.

Stream Water Sources Switching From Primarily Old, Deep Water in the Cold Dry Season to Young, Shallow Water Under Wet Conditions
Stream DIC exhibited a slight dilution behavior with decreasing concentrations at high discharge, with a power law CQ slope b = −0.12(C = aQ b , Figure 8).The CQ relationship indicated that the stream water sources switched from primarily old water in DZ almost at equilibrium with carbonate (IAP/K eq close to 1, large dots in Figure 8A1) under dry conditions, to primarily young water in SZ far from equilibrium (IAP/K eq << 1.0, small Figure 8. (A) Relationship between daily stream discharge and stream dissolved inorganic carbon (DIC) concentrations (A1), and δ 13 C (A2); (B) Relative contribution of daily carbonate weathering rates/fluxes to the corresponding daily export rates R e (B1) and production rates R p (B2) as a function of soil temperature and water-related measures.All dots are daily model outputs.In panel (A), the dot size represents the saturation state of carbonate weathering.Concentration-discharge relationships were quantified using a power law equation C = aQ b , with b representing the slope of CQ patterns.From low Q (cold dry) to high Q (hot wet), the stream water transitioned from the old groundwater approaching carbonate equilibrium (large dots and close-to-1 ion activity product [IAP]/K eq values) to young, abundant-soil CO 2 water far from the equilibrium (small dots and IAP/K eq ).With increasing discharge, the proportion of CO 2 -abundant soil water (i.e., biogenic source) increased in stream DIC, resulting in less dilution of dissolved carbon from soil respiration compared to carbonate weathering.Correspondingly, R e_geo /R e was almost constant at 0.43 under low discharge (20 years).Beyond that, R e_geo /R e decreased significantly with discharge.R p_geo /R p showed more variations, decreased with soil T, and decreased further under extremely dry conditions (S w < 0.53).dots) under wet conditions.The pattern of δ 13 C versus Q exhibited similar decreasing trend (Figure 8A2), indicating stream DIC primarily from DZ with high δ 13 C of geogenic origin (−10.8 ± 0.1‰) under low discharges and from SZ with more negative δ 13 C of biogenic origin (−13.0 ± 0.7‰) under wet conditions.Correspondingly, the relative contribution of geogenic DIC export to the overall daily DIC export (R e_geo /R e ) was almost constant at around 0.43 under dry conditions (Q < 1 mm/d) when R e predominately originated from DZ (Figure 8B1).Beyond that, as R e increased with more contributions from abundant-soil CO 2 young SZ, R e_geo /R e gradually decreased to ∼0.37.
The geogenic carbon (calcite weathering) contribution to the production, R p_geo /R p , showed more variations with soil T and hydrological regimes compared to the export ratio R e_geo /R e (Figure 8B).The ratio R p_geo /R p was generally lowest under warmest conditions when soil respiration dominated DIC production (Figure 8B2).Minimal geogenic contribution (low ratio) can occur and reach as low as 0.25 under cold conditions (T < 15°C), when the cold conditions coincided with the very dry conditions (S w < 0.53) such that carbonate weathering was mostly at equilibrium with slow flow.

Discussion
Understanding sources (biogenic vs. geogenic) and processes (production and export) of riverine DIC is essential for comprehending global carbon cycling and carbon fluxes.It is however challenging to understand the reactions and transport processes in the subsurface that control when, where, and how much DIC production and export occur (Campeau et al., 2017;Zamanian et al., 2016).These processes vary with climate forcing, landscape, and subsurface characteristics.DIC from karst catchments is often thought to be about 50% geogenic, because carbonate weathering consumes 1 mol of soil CO 2 (aq) and produces 2 mol of DIC.Here we use hydrometeorological, stream flow and chemistry data to constrain a catchment-scale RTM.This integrated model-data approach illuminates the coupled reactive transport processes in the subsurface and distinguishes the effects of individual factors, including the monsoon climate, on the sources, timing, and magnitude of DIC production and export (Figure 9).Results here additionally highlight the distinct contributions of depression-versus-uphill and shallow-versus-deep carbon sources, portraying catchments as hydro-biogeochemical reactors that have been conceptually proposed but not explicitly quantified (L.Li, 2019;L. Li et al., 2021).

Amplified DIC Production and Export During Subtropical Monsoon Season
The subtropical monsoonal climate is characterized by alternating cold-dry winter and spring and hot-wet summer and fall (Green et al., 2019;Zhang et al., 2011).Transitioning from the cold-dry to hot-wet season, soil saturation (S w ) increased by almost a factor of two (0.45-0.75), and soil T rose from 4 to 22°C (Figures 3 and 7).In addition, daily DIC production rates R p escalated by an order of magnitude, and seasonal average increased by almost two times (3.8 × 10 −3 to 5.9 × 10 −3 mol C/m 2 /day).High soil saturation (S w > 0.7) only reduced the rates by less than 10%.Under extremely dry conditions, carbonate weathering is limited by low effective rock-water contact and approaches to equilibrium, regardless of temperature conditions and biological activities.Under hot-wet conditions, elevated soil CO 2 promotes carbonate weathering by increasing water acidity and therefore carbonate solubility.Additionally, the rapid water flow drives carbonate weathering away from equilibrium.This indicates that carbonate weathering is largely driven by biological activity instead of water flow under wet conditions but by water flow under dry conditions.This supports the hypotheses that weathering rates are limited by the availability of reactive gas such as CO 2 in humid climates and by accessible water in arid environments (Calmels et al., 2014;L. Li et al., 2017).DIC production rates R p , whether biogenic or geogenic, depend more on T (R 2 > 0.8) than hydrological conditions (R 2 < 0.2, as shown in Figures 6 and 7), similar to those observed in catchments in temperate climate (Wen et al., 2020).The export rates R e , however, largely depend on discharge into the stream.In cold-dry spring and winter, DIC export is mostly from the slow flow at the deep subsurface.In the hot-wet season, uphill and depression soils are hydrologically connected to the stream, flushing out accumulated DIC produced in spring and winter and enhancing export by orders-of-magnitude.
The temporal patterns of DIC production and export rates therefore are predominantly controlled by the climate seasonality (Figures 6 and 9).Both production and export rates peak in the hot-wet summer and fall, and plummet in the cold-dry winter and spring.This is similar to catchments in boreal climates characterized by moderately warm-moist summers and extremely cold-dry winters.They have been shown to behave similarly due to their synchronous temperature and precipitation patterns (Bowering et al., 2023).This contrasts with catchments in temperate climates, where warmer seasons often coincide with drier conditions, resulting in high production but low export, as shown in the asynchronous pattern of production and export for DOC in Shale Hills in Pennsylvania, USA (Wen et al., 2020).In Shale Hills, the production rates peak in summer but export peak in spring during snow melt events.The catchment acts as a reactor (producer) in hot-dry summer and as a transporter (exporter) in cold-wet winter (L.Li et al., 2022;Wen et al., 2020).In Chenqi, the catchment is a reactor and transporter simultaneously in the hot-wet summer and fall.In tropical ecosystems, where temperature variation is negligible but precipitation varies significantly across seasons, the production of dissolved carbon was observed to be predominantly influenced by water content (Ilstedt et al., 2000;W. J. Zhou et al., 2016), with elevated concentrations and fluxes of dissolved carbon (DIC, DOC) during rainy seasons (Neu et al., 2016).

More Significant Vertical (Shallow-Deep) Contrasts Than Landscape Positions (Uphill-Depression)
Uphill versus depression.Under hot-wet conditions, high soil water saturation and fluxes (>0.1 mm/d) enhance hydrological connectivity and minimize differences between uphill and depression.Under cold-dry conditions when depression and uphill soils are hydrologically disconnected, the differences are amplified.Although with similar temperature, depression soils with high OC abundance and soil water content have higher production rates (up to a factor of two) than the uphill (Figure 5), especially in the cold-dry season.Similarly, solute concentrations in depression and uphill differ more during the cold-dry season.The depression zone generally is more acidic Figure 9. Conceptual representation of the catchment under cold-dry and hot-wet conditions.The subtropical monsoonal climate determines synchronous daily production and export rates of dissolved inorganic carbon (DIC) (R p and R e ) that peak in the hot-wet season and plummet in the cold-dry season.Daily DIC production rates R p are driven predominantly by temperature and vary by about one order of magnitude across the year.Conversely, daily R e is mostly driven by stream discharge, peaking in the hot and wet season with rates higher than those in the cold dry season by more than two orders of magnitude.Soil respiration dominates the production (R p_bio /R p > 0.5) and contributes more under hot-wet conditions; higher soil CO 2 productivity at high T also elevates carbonate solubility and accelerates carbonate weathering.The contribution of biogenic versus geogenic DIC in the stream largely depends more on flow paths.In the hot-wet season, stream water has low DIC concentrations with 13 C depletion, indicating biogenic from biogeochemically active soils.In the cold-dry season, stream DIC mostly originates from the deep subsurface with old groundwater (∼30 years), where geogenic carbon contributes constantly and about half of stream DIC.
(lower pH), and has higher DIC and CO 2 concentrations by a factor from 1.2 to 6.3 compared to the uphill, due to its generally higher soil moisture.In DZ, however, the landscape position difference is dampened and concentrations are similar in these two zones.
Shallow versus deep zones.The shallow-deep contrasts are more pronounced than the uphill-depression differences (Figures 5 and 6).On a daily to seasonal scale, compared to SZ, DZ is more stable, generally with higher pH and DIC concentrations and close to equilibrium.Similarly, respiration rates in DZ are stable and orders of magnitude lower than SZ.Surprisingly, rates of carbonate weathering in DZ are low but vary substantially.This is especially the case during the hot-wet season when carbonate weathering is predominantly driven by large water events and recharge of high soil CO 2 abundant-water.In contrast, soil respiration rates in SZ vary substantially with seasons (Figures 5c, 5d, and 6B1).The annual DIC production rate (R p ) in SZ is about 2.9 × 10 6 mol C/yr, over one order of magnitude higher than that in DZ (2.2 × 10 5 mol C/yr).The DIC export rate through DZ is also lower, at about 1.1 × 10 6 mol C/yr, about half of that through SZ (2.0 × 10 6 mol C/yr) due to generally lower water flow (∼24.0% of Q).
The carbonate weathering rates in the DZ are typically one to two orders of magnitude lower than those in the SZ (Figures 5 and 6), contrasting the common perception of high carbonate weathering rates at the deep weathering front in karst systems (G.Zhou et al., 2015).This can be attributed not only to the abundant carbonate rocks in shallow soils (i.e., outcrops, or shallow reaction front) in Chenqi, but also high flow that enables the relatively persistent disequilibrium conditions such that carbonate can dissolve and uptake CO 2 in soil zones.Such combination of shallow carbonate weathering front and rapid flow in shallow soil accelerate carbonate weathering in SZ, contrasting carbonate weathering in deep zones that is typically limited by slow flow and at-equilibrium water (Figure 5h).
Carbonate weathering is at disequilibrium with soil CO 2 most of the time in the SZ (Figure 5h), especially in the hot-wet season.This contrasts the general perception from small-scale lab systems that carbonate dissolution reaches equilibrium rapidly, often within hours (Plummer et al., 1978).This may largely arise from the combination of high carbonate content in karst formations, high temperature that promotes the generation of soil CO 2 , and high discharge that drives the water to disequilibrium.Together they enhance carbonate weathering and consume more soil CO 2 than expected.This indicates that the common equilibrium assumption for the estimation of riverine concentrations in regional, continental or global models may not hold and can potentially lead to estimation biases (Maher & Chamberlain, 2014;Romero-Mujalli et al., 2019a;S. Zeng et al., 2019).
Although the results highlight the significant differences in solute concentrations and rates between shallow and deep zones, the magnitude of this contrast may vary with a variety of factors that influence DIC production, including soil properties and land cover (Calmels et al., 2014;Pokrovsky et al., 2015;Shan et al., 2021).Low DIC (∼3.0 mM) in thin soils compared to higher concentrations (∼5.0 mM) in thick OC-rich soils have been attributed to limited soil respiration in karst terrains in southeast China (Green et al., 2019;J. H. Yan et al., 2011).DIC levels in shallow zones can also vary tremendously with land cover, with lower concentration ∼0.5 mM in bare soils and high concentrations ∼6.0 mM in grasslands (Macpherson et al., 2008;Q. Zeng et al., 2017).

Biogenic Versus Geogenic DIC Production
This work reveals that on average, about 60% of the riverine DIC is derived from soil respiration.The annual production rate is 2.5 mol C/m 2 /yr, with 1.5 mol C/m 2 /yr being biogenic.Their daily contributions vary significantly with weather conditions (Figures 7-9).Under cold-dry conditions where most stream water is from the old, deep groundwater, the relative contribution of geogenic carbon peaks and can reach almost half of the total DIC production.In the hot-wet season, geogenic carbon can contribute as low as ∼34%.In other words, as the catchment becomes wet and hydrologically connected, stream DIC is mostly from the shallow soil that is more biogenic.Similar dry-wet biogenic versus geogenic shifts were observed in the Congo basin and Howard River catchment of Australia (Bouillon et al., 2014;Duvert et al., 2020), suggesting such changes may be characteristic of dry-wet climates in (sub)tropical regions.The overall mean contributions of geogenic carbon are around 40%, lower than the generally reported 50% for karst formations (Hartmann, 2009;Z. Liu et al., 2010;S. Zeng et al., 2019), indicating generally overestimated geogenic contribution, especially under hot-wet conditions.The δ 13 C signatures in stream DIC ranged from −15.2‰ to −10.8‰.Similar ranges of δ 13 C signatures in stream DIC were observed in catchments primarily composed by carbonate rocks (Jin et al., 2014;J. Liu & Han, 2020), potentially indicating the typical dominance of biogenic carbon in karst systems.The main sources of DIC in headwater streams have been observed to change with a variety of factors, including lithology and climate, resulting in a broader range of δ 13 C signatures (Calmels et al., 2014;Pokrovsky et al., 2015;Shan et al., 2021).Sulfuric acid from oxidation of pyrite and acid rain and other acids (such as nitric acid and organic acids) can enhance carbonate weathering and produce 13 C-enriched DIC (Qin et al., 2019;Song et al., 2020).Strong enrichment from −13.5‰ to −4.8‰ have been observed in southwest China due to the oxidation of pyrite and acid rain (S.-L.Li et al., 2008;Xu & Liu, 2007).Silicate weathering accelerated by soil CO 2 can result in depleted δ 13 C signatures between −22.0‰ and −16.0‰ (Duvert et al., 2020;W. Wu et al., 2008).Co-existing carbonate and silicate rocks often result in carbonate precipitation in dry seasons and in arid and semi-arid regions but dissolution under wet conditions (Monger et al., 2015;Wen et al., 2022;Zamanian et al., 2016), which can further complicate DIC sources and δ 13 C signatures.
Future climate projection indicates southwest China will be warmer and wetter, with an increase of ∼190 mm in precipitation and 3.5°C in temperature during 2051-2100, respectively (Duan et al., 2021).The wetter and hotter conditions may further elevate soil respiration, carbonate weathering, and overall DIC production.If the relationship between temperature and rates in Figure 7a holds, the increase in DIC production is expected to be around 10 5.2 mol C/yr.Meanwhile, extreme precipitation and temperatures are predicted to increase at much higher rates than their mean values (S.-Y.Wu et al., 2019), further amplifying DIC production and export during the hot wet monsoon season.More and intensified precipitation extremes can also reduce riverine DIC concentrations but enhance DIC lateral export at higher discharge.

Global Hot Spots for Soil CO 2 Sink But Riverine CO 2 Sources?
The consumption of soil CO 2 by carbonate weathering, equivalent to the carbonate weathering rates here, has been considered as a sink for atmospheric CO 2 at short timescales (years -millennia) (Gaillardet et al., 1999;Hartmann, 2009;Z. Liu et al., 2018;J. H. Yan et al., 2011;S. Zeng et al., 2019).The carbonate weathering rate is estimated to be 1.0 mol C/m 2 /yr in Chenqi, higher than most of the estimation in the literature (Figure 10).Carbonate weathering rates are 0.48 mol C/m 2 /yr at the global scale (Amiotte-Suchet et al., 2003), 0.43 mol C/m 2 / yr in Chinese karst systems (0.67 mol/m 2 /yr for tropical and subtropical karst, 0.34 mol/m 2 /yr for plateau karst, 0.06 mol C/m 2 /yr for semiarid karst and 0.02 mol C/m 2 /yr for temperate humid karst) (Jiang & Yuan, 1999;Z. Liu & Zhao, 2000), 0.50 mol C/m 2 /yr for the Alpine river basins (Donnini et al., 2016), 0.66 mol/m 2 /yr in Konza Tallgrass Prairie (Macpherson & Sullivan, 2019), and 0.58-1.53mol/m 2 /yr for 5 karst hydrosystems in France (Binet et al., 2020).This indicates that karst formations in subtropical or tropical climate with abundant precipitation may be global hot spots for soil CO 2 sink (Figure 10).On the other hand, high DIC export rates at high discharge  2016), Jiang and Yuan (1999), Z. Liu and Zhao (2000), and Macpherson and Sullivan (2019).The dashed line represents the linear regression trend.Karst formations in subtropical or tropical climate (red cycles) with abundant precipitation tend to have high carbonate weathering rates and therefore may act as global hot spots for soil CO 2 sink.can also lead to high CO 2 evasion in rivers and streams (Duvert et al., 2018;Hotchkiss et al., 2015), which again imposes large uncertainties on whether karst regions are CO 2 sources or sinks.

Model Limitations and Implications for Future Research
The soil respiration rates estimated here correspond to the rates of producing DIC.It does not include processes such as soil CO 2 effluxes back to the atmosphere on land, and in-stream processes such as CO 2 evasion and stream respiration.As such, the soil respiration rates here likely underestimate the total soil respiration rates significantly.For example, the soil CO 2 efflux measured at the ground surface is usually considered as approximating soil respiration rates (Jian et al., 2021).These rates are usually within the range of 3.2-208.3mol C/m 2 /yr at subtropical regions, much higher compared to the soil respiration estimated here.In addition, riverine CO 2 evasion and stream respiration are known to affect DIC concentrations and fractionation of δ 13 C (Duvert et al., 2018;Hotchkiss et al., 2015;Marx et al., 2017;Solano et al., 2023).About 11% of dissolved carbon was estimated to evade the stream in karst formations of southwest China (S.-L.Li et al., 2010).CO 2 evasion and stream respiration can also lead to a more enriched δ 13 C signature.CO 2 evasion and stream respiration however usually occur at slower rates in small catchments with underground streams compared to aboveground rivers with direct interactions with the atmosphere (Butman & Raymond, 2011;Finlay, 2003).In the future hotter and wetter climate (Duan et al., 2021), storms, high discharge, and more contribution of biogenic sources (e.g., CO 2 (aq)) in the riverine DIC can potentially promote CO 2 evasion.Detailed, co-located measurements of soil CO 2 efflux to the atmosphere, riverine CO 2 evasion, and dissolved carbon in streams are often lacking but are essential to gain a comprehensive understanding and quantification of carbon cycling and fluxes in terrestrial systems and terrestrial-aquatic-atmospheric interfaces (Wen et al., 2022).
The four-grid cell model here aims to capture the key dynamics of the systems without overparameterizing the model without sufficient data (Wen, Brantley, et al., 2021).The model simplified the representation of the watershed by having only two cells of depressions and hillslopes in the landscape direction and two layers of soil and bedrock in the vertical direaction.This similarity aligns more with a hillslope-based model than with spatially distributed models that usually consist of hundreds or more cells.Although with some uncertainties (Figure 4), the model reproduced water chemistry data satisfactorily in both dry and wet seasons, indicating the essential features that lead to key stream dynamics are captured.This supports the hypothesis that representing essential landscape functioning units (e.g., hillslope) using simple or lumped models may be essential to capture key dynamics (Wen, Brantley, et al., 2021).In particular, model results show that the shallow and deep contrasts are much more pronounced than the uphill-depression contrasts.This suggests that representing the shallow and deep features accurately might be more crucial than depicting variations in the landscape direction.It shows the promise of using relatively simple hydro-biogeochemical models to represent salient features of watersheds in order to capture the key dynamics.Similar ideas have been implemented to simulate complex hydrological dynamics in karstic catchments through multiple conceptual boxes as the fast, medium, and slow flow paths (Husic et al., 2019;Zhang et al., 2019Zhang et al., , 2020Zhang et al., , 2021)).

Conclusion
This work asks the questions that have not been well addressed in subtropical karst regions: How and how much do sources, production and export of DIC vary across space (shallow vs. deep, uphill vs. depression) and time (daily, seasonal, and annual)?What is the relative contribution of biogenic (via soil respiration) and geogenic (via carbonate weathering) sources?Multiple observations (weather, stream discharge and chemistry, and carbon isotope) and a catchment-scale RTM enable the illumination of subsurface hydrological and biogeochemical processes that determine the timing and magnitude of DIC production and export in a small karst catchment.
Results show that annually, soil respiration was the primary contributor of DIC, accounting for about 60% of the total DIC production (the rest 40% from carbonate weathering), which exceeds the typical estimated 50% in karst formations.The hot-wet monsoon season simultaneously enhances soil respiration and carbonate weathering via disequilibrium conditions amplified by high flow and soil CO 2 levels.Compared to the deeper subsurface zone, the shallow soil typically has significantly higher production and export rates, often up to two orders of magnitude.The disparity between the depression near the stream and the uphill area is much less pronounced, typically within a factor of two.Carbon isotope data and model analysis indicate that the sources of riverine DIC shift from more biogenic (∼58%-66% of the total) from the shallow young soil water (<1 year old) in hot-wet seasons to about half biogenic (∼50%) from the deeper old groundwater (∼30 years old) in cold-dry seasons.These findings highlight the significance of time (hot-wet vs. cold-dry) and distinct contribution of the shallow and deep subsurface to carbon production and lateral export.They also suggest that, in the future, hotter and wetter climates and intensive storms may further elevate biogenic DIC production and lateral export.

Figure 2 .
Figure 2. A schematic representation of major processes in the watershed reactive transport model BioRT-Flux-PIHM.The Chenqi watershed is discretized into four prismatic grid cells, representing the steep uphill and flat depression (green and yellow) on each side of the stream.Each prismatic land element has two layers, representing the shallow soil (shallow zone [SZ]) and the deeper underlying bedrock (deeper zone [DZ]).Stream discharge Q is the sum of surface runoff Q surf , soil water interflow Q SZ , and groundwater flow Q DZ .Dissolved inorganic carbon (DIC) in the stream comes from rainfall, soil respiration (biogenic source), and carbonate weathering (geogenic) with distinct isotope signatures.Soil CO 2 from soil respiration (root respiration and organic carbon mineralization) can dissolve in water and become DIC or further react with carbonate minerals.DIC can export laterally via SZ into stream (Q SZ ), or vertically recharge into DZ and eventually enter stream via a longer flow path (Q DZ ).

Figure 4 .
Figure 4. Temporal dynamics of stream chemistry in 2007-2009, including (a) δ 18 O, (b) pH, (c) Ca, (d) alkalinity, (e) dissolved inorganic carbon, and (f) δ 13 C, as well as the model output and the hourly data during the 2018 summer from Qin, Ding, et al. (2020) and Qin, Li, et al. (2020) (the right side of panels (d, f)).In cold-dry spring and winter, stream discharge was mostly derived from deeper zone with almost constant stream chemistry.Under hot-wet summer and fall, stream chemistry resembled soil water chemistry from soil CO 2 -abundant shallow zone with more fluctuations and negative δ 13 C.

Figure 3 .
Figure 3. Temporal series of hydrological variables at the catchment scale in 2007-2009, including (a) daily precipitation, discharge, and evapotranspiration (ET), (b) normalized flux at the outlet, (c) mean water transit time, and (d) soil water saturation and soil temperature (soil T).Dots are data; lines are model outputs.Stream discharge was primarily from the young soil water Q SZ (<1 year old) in summer and fall with intense rainfall and high ET and from the old deeper water Q DZ (∼30 years) in spring and winter.

Figure 5 .
Figure5.Modeled variables in the shallow zone (SZ) and deeper zone (DZ) at uphill and depression, including (a) soil saturation and T, (b) flow rates, (c) soil respiration rates r p_bio , (d) carbonate weathering rates r p_geo (= r cal ), (e) pH, (f) dissolved inorganic carbon (DIC), (g) aqueous and gaseous soil CO 2 , and (h) ion activity product/K eq .Gaseous soil CO 2 (g) in panel (g) was estimated using Henry's law.Soil CO 2 and r p_bio were high in depression soils that had relatively high organic carbon (OC) and water content, leading to correspondingly high H + (lower pH), DIC and carbonate weathering rates compared to the uphill.These spatial contrasts were more significant during the cold-dry season when the uphill and depression was mostly disconnected and most products accumulated in soils.The DZ with more abundant carbonate had much less temporal and spatial variations because of less abundant OC content, slower groundwater flow, and longer water transit time.Compared to uphill-depression differences, the shallow SZ versus deeper DZ contrasts were much more pronounced.For example, rates in SZ and DZ can reach several orders of magnitude differences (c, d).

Figure 6 .
Figure6.Model output of temporal dynamics for (A) temperature, discharge, dissolved inorganic carbon (DIC) storage, production and export rates at the catchment scale (including both shallow zone [SZ] and deeper zone[DZ]) and (B) corresponding rates of soil respiration (R p_bio , R e_bio ) and carbonate weathering (R p_geo , R e_geo ) in SZ and DZ.The DIC input from rainfall (R in_r ) was negligible, about 1.0% of R p .Temporal patterns of R p , R p_bio and R p_geo varied within an order of magnitude and generally followed that of soil T, whereas R e , R e_bio and R e_geo mostly followed that of discharge, increasing by up to two orders of magnitude with Q.The relative magnitude of production and export rates in different seasons determines the temporal patterns of the overall DIC storage, which peaks in cold-dry seasons with minimal flow and reached minima in the hot moments of export during hot-wet periods.Rates and mass in SZ generally fluctuated more and were one-to-two orders of magnitude higher than those in DZ.

Figure 7 .
Figure 7. Relationships between daily (a-c) R p and(d-f) R e and daily soil T, soil saturation S w , mean water transit time τ stream , and discharge.All dots are daily model outputs; the symbol size in panels (a, b) represents the magnitude of S w and T. R p_bio , R p_geo , and R p all generally increased with soil T, whereas R e_bio , R e_geo , and R e all depended more on τ stream and S w , except the cold-dry periods with low soil T and S w (<15°C and 0.53).In the cold-dry periods, R p_geo decreased with decreasing S w even when T increased; export rates were relatively constant because stream water was mostly contributed by the old deeper zone.

Figure 10 .
Figure 10.Carbonate weathering rates (R cal ) in karst formations as a function of annual precipitation.Data in different regions are from Amiotte-Suchet et al. (2003), Binet et al. (2020), Donnini et al. (2016),Jiang and Yuan (1999), Z.Liu and Zhao (2000), andMacpherson and Sullivan (2019).The dashed line represents the linear regression trend.Karst formations in subtropical or tropical climate (red cycles) with abundant precipitation tend to have high carbonate weathering rates and therefore may act as global hot spots for soil CO 2 sink.