Diffuse Groundwater Discharge Dominates Terrestrial Dissolved Inorganic Carbon Export and CO2 Evasion From a Semiarid Headwater Stream

Groundwater discharge to headwater streams and concomitant terrestrial dissolved inorganic carbon (DIC) export play a significant role in headwater stream CO2 evasion. However, previous studies rarely examined diffuse groundwater discharge and its impact on headwater stream CO2 evasion, thereby lacking the understanding of the role of diffuse groundwater discharge in terrestrial DIC export and stream CO2 evasion. This study quantified diffuse groundwater discharge along a 43 km semiarid headwater stream by combining hydraulic, isotopic (radon-222) and chemical (electrical conductivity) approaches, and estimated the reach-level CO2 budgets of the stream. Reach-scale water and mass balance modeling yielded highly variable diffuse groundwater discharge rates (n = 16, range: 1.08-7.80 m2/d, mean ± 1 sd: 4.57 ± 1.81 m2/d). Groundwater was supersaturated with CO2 at all sites, with strongly variable CO2 partial pressure (pCO2) and DIC concentrations at 1,223-27,349 μatm and 30-119 mg/L, respectively. Diffuse groundwater discharge dominated terrestrial DIC export to the stream (12-111 g C m-2 d-1, normalized to water surface area). A portion of groundwater dissolved CO2 transported to the stream was emitted to the atmosphere with evasion rates varying at 0.62-3.18 g C m-2 d-1. However, most dissolved CO2 was transformed into HCO3through carbonate buffering because of the regulation of carbonate equilibrium. Overall, the stream CO2 evasion was driven by carbon transfer but limited by carbon supply. This study provides a bottom-up perspective to understand terrestrial DIC export and stream CO2 evasion in arid and semiarid areas.


45
Reach-scale groundwater discharge to the Hailiutu River was estimated by modeling the 161 stream water balance and the mass balances of 222 Rn and EC simultaneously (Cook, 2013;Cook 162 et al., 2006). The stream water balance is given by 163 (1) 164 9 where Q is the stream flow rate (m 3 /d), x is the distance in the direction of flow (m), I is the 165 groundwater discharge per unit length (m 2 /d), Tri is the tributary inflow rate per unit length 166 (m 2 /d), E is the evaporation rate (m/d) , and w is the stream width (m). Tri is equal to the 167 tributary flux (positive, i.e. Bulang River flux) or irrigation diversion flux (negative, we found 168 three irrigation diversion points) divided by the length between two adjacent stream 169 measurement points. 170 The environmental tracer 222 Rn has been used frequently to quantify groundwater discharge 171 to surface water (Cook, 2013;Cook et al., 2003;Cook et al., 2006;Hofmann et al., 2011;Xie et 172 al., 2016). 222 Rn is a radioactive noble gas with a half-life of 3.8 days. It is a decay product of 173 uranium series isotopes. Given the extensive existence of uranium in aquifer sediment, 222 Rn is 174 produced continuously in groundwater. Once groundwater discharges to the stream, 222 Rn 175 activity is affected by several factors including gas exchange with the atmosphere, radioactive 176 decay and dispersive mixing. The mass balance of 222 Rn is given by the following equation 177 (Cook,

Hydrogeochemical modeling 275
We modeled carbonate buffering process after groundwater discharged to the stream 276 through the PHREEQC simulation program (Parkhurst and Appelo, 2013). The mean water 277 temperature, pH, Ca 2+ , Mg 2+ and alkalinity of our groundwater samples were assigned as the 278 initial model parameters, then we modeled the re-equilibrium processes between the groundwater 279 and the air for different pCO 2 values and calculated the corresponding calcite saturation (Ion 280 Activity Product / Solubility Product Constant of calcite, IAP/K calcite, similarly hereinafter). 281 Through the change of calcite saturation, we can explore the shift in carbonate equilibriums.

288
Furthermore, we used one-way analysis of variance (ANOVA test) to compare the carbon 289 concentration differences between the stream and the groundwater, and different carbon budget 290 components at the significance level of p < 0.05.

Longitudinal patterns of stream flow and groundwater discharge 293
Field measured values for reach-scale water and mass balance modeling are listed in SI 294 Table S1, and the spatial variations in Q, 222 Rn activities and EC are depicted in Figure 2a

316
Model parameters are defined in SI Table S2. Parameters E, θ and λ were assumed to be 317 constant. For each stream reach, the upstream and downstream sampling sites were used to 318 calculate w, d, and C gw . There are three small irrigation canals along the Hailiutu River where we 319 did not measure the 222 Rn activities and EC. We assumed that the 222 Rn activities and EC in the 320 irrigation canals were the same as those of the nearest stream sampling site ( = Tri CC ). This 321 assumption is reasonable as these values were only used to account for water and mass losses from the study stream. Q 0 , C 0 at the first sampling site (Hailiutu-01 in SI Table S1)  modeling results indicate that the groundwater discharge occurred along the entire stream other 336 than concentrating on some local areas, and I varied between 1.08 and 7.80 m 2 /d with the mean ± 337 1 standard deviation at 4.57 ± 1.81 m 2 /d. The highest and lowest I occurred at Reach 2 and 338 Reach 10, respectively. The uncertainty of I is approximately 2 m 2 /d (shaded area in Figure 2d). 339

Longitudinal patterns of carbon concentrations in stream and groundwater 340
Stream DIC concentrations show a slightly decreasing trend from 62 mg/L at the upstream 341 end to 43 mg/L at the downstream end (Figure 3a). In comparison, groundwater DIC 342 concentrations (66 ± 24 mg/L) fluctuated more strongly than those of stream water (48 ± 5 mg/L) ( Figure 3). Notably, DIC was the main carbon species in both the stream and the adjacent 344 groundwater, because the DIC concentrations were approximately nine times higher than DOC 345 concentrations in both the stream and groundwater (ANOVA, n = 34, F = 1012 and 97, 346 respectively, both p values < 0.0001). Stream and groundwater DOC concentrations were 347 relatively constant along the stream with the values at 5 ± 1 and 7 ± 3 mg/L, respectively. Both 348 the stream and the adjacent groundwater were supersaturated with CO 2 with pCO 2 at 719 ± 168 349 μatm and 9,343 ± 7,050 μatm, respectively, when compared with the average atmospheric pCO 2 350 of 390 μatm. Furthermore, groundwater pCO 2 correlates well with groundwater DIC (Figure 3b, 351 R 2 = 0.91, p < 0.0001). 352 Overall, the DIC, DOC and CO 2 concentrations in the groundwater were significantly 353 higher than those in the stream (Figure 4, ANOVA, n = 34, F = 9.01, 6.70 and 25.42, 354 respectively, all p values < 0.05). Particularly, pCO 2 in the groundwater was an order of 355 magnitude higher than that in the stream with the mean values at 9,343 and 719 μatm, 356 respectively. We also found that DIC concentrations and pCO 2 in the riparian groundwater were 357 higher than those in the groundwater from the wells (SI Table S3). groundwater re-equilibrates with the air of -log 10 (pCO 2 ) (atm) at 2.03 (i.e., 9,343 μatm, the average pCO 2 of 378 the 17 groundwater samples in our study), 2.5, 3.0, 3.41 (i.e., 390 μatm, the atmospheric pCO 2 ), respectively.

379
The model parameters used in the PHREEQC simulation are defined in SI Table S4. 380

Terrestrial carbon export and stream CO 2 evasion 381
The measured data for quantifying the reach-scale carbon budget are listed in SI Table S3,  382 and these results are depicted in Figure 6-8. The comparison between external and internal CO 2 383 contributions indicates that the external CO 2 input was higher than the net internal CO 2 384 production ( 2 CO gw F : 3.73 ± 2.52 g C m -2 d -1 , 2 CO m F : 1.08 ± 4.66 g C m -2 d -1 , Figure 6). It should be 385 noted that the net internal CO 2 production at all the stream sections except Reaches 5, 8, 9 and 13 386 made positive contribution to the stream CO 2 balance (Figure 6a). Since DIC is the main carbon 387 species in both the groundwater and the stream as discussed above, the terrestrial carbon export 388 is primarily in the form of DIC. The reach-scale terrestrial DIC export ( DIC gw F , the product of the 389 groundwater DIC concentrations and I) was 48.78 ± 28.78 g C m -2 d -1 , and varied between 12.20 390 and 111.13 g C m -2 d -1 (Figure 7a). I at Reach 2 was the highest (Figure 2d), and DIC gw F was also 391 the highest (Figure 7a). Conversely, where I was limited (e.g., Reaches 10 and 11), DIC gw F was 392 also constrained ( Figure 7a). 393 Notably, both the stream and the groundwater were high in pH (8.50 ± 0.10 and 7.60 ± 0.25, 394 SI Table S3) and alkalinity (3.98 ± 0.46 and 5.04 ± 1.71 meq/L, SI Table S3)

Potential drivers for terrestrial carbon export and release 426
Both the terrestrial DIC export and the stream CO 2 evasion can be controlled by either 427 carbon transfer (i.e., groundwater discharge rate and CO 2 gas transfer velocity, Zone A in Figure  428 9) or carbon supply (i.e., groundwater DIC concentration and stream pCO 2 , Zone C in Figure 9). 429 The reaches located at Zone B in Figure 9 are hotspots for carbon fluxes and driven by both the 430 transfer and the supply, whereas the reaches located in Zone D are limited by both the transfer 431 and the supply and so are not important for carbon fluxes. In our study, most reaches are hotspots 432 (located at Zone B in Figure 9a) for terrestrial DIC export except Reaches 10 and 11 (limited by 433 groundwater discharge, Figure 2d). Stream CO 2 evasion rates are mainly located close to the threshold for dividing Zone A and Zone B in Figure 9b, indicating that the carbon fluxes are 435 driven by the transfer but limited by the supply. 436 Terrestrial DIC export to headwater streams is mainly controlled by groundwater discharge 437 . In our study, we found that the terrestrial DIC export is 439 positively correlated with the groundwater discharge, but no significant correlation between the 440 terrestrial DIC export and the groundwater DIC concentration (Figure 10a and 10c). In 441 comparison, the stream CO 2 evasion is positively correlated with both the CO 2 gas transfer 442 velocity and stream pCO 2 , with the former more significant than the latter (Figure 10b and 10d).   DIC export in not only our headwater stream but also many other headwater streams. 478

Terrestrial DIC export as the major carbon source for streams 479
As demonstrated by previous studies, supersaturated CO 2 * in streams and rivers is mainly 480 derived from external input (i.e., terrestrial DIC export) or internal metabolism (Hotchkiss et al., vegetation growth than areas that are relatively far from the streams. Therefore, soil respiration is 490 more active in the riparian zone than in the rest of the catchment, causing the higher CO 2 491 concentrations (Hope et al., 2004;Leith et al., 2015). 492 This finding was also supported by groundwater δ 13 C DIC values (-11.90 ± 1.98 ‰, see SI 493 Table S5), which fall in the potential δ 13 C DIC range for C4 plants (corn in our case) grown in the 494 riparian zone (Clark and Fritz, 1997). Furthermore, our δ 13 C DIC data also suggest that terrestrial 495 DIC export is the main carbon source of stream DIC pool. After terrestrial DIC was exported to 496 the stream, the CO 2 gas exchange between the stream and atmosphere and the internal 497 metabolism resulted in more positive δ 13 C DIC values in stream than in groundwater (ANOVA, n 498 = 34, F = 8.64, p < 0.01) (Deirmendjian and Abril, 2018). 499

Stream CO 2 evasion was driven by carbon transfer but limited by carbon supply 500
The terrestrial CO 2 export via the diffuse groundwater discharge directly sustained the 501 stream CO 2 evasion. However, considering the high pH and high alkalinity setting in our study 502 area, most of the terrestrial DIC exported to the stream were in the form of HCO 3 -. Thus, the 503 transformation between CO 2 * and HCO 3 -(carbonate buffering) can also indirectly enhance or 504 limit the stream CO 2 evasion by regulating stream CO 2 pool (conceptual model in Figure 11 balance results indicate that the carbonate buffering caused most CO 2 * to be transformed into 507 HCO 3 after the CO 2 -rich groundwater discharged to the stream, thereby increasing the calcite 508 saturation of the stream water ( Figure 5) (Jacobson and Usdowski, 1975;Lorah and Herman, 509 1988;Lu et al., 2000). Although most reaches are the hotspots for the terrestrial DIC export 510 (Figure 9a), most CO 2 * loss occurred through the carbonate buffering, causing the limited carbon 511 supply for the stream CO 2 evasion. The limited CO2 evasion was supported by very close mean 512 δ 13 C DIC values of stream water and groundwater (-10.46 ‰ and -11.90 ‰, respectively, SI Table  513 S5). This CO 2 loss mechanism is attributed to the high alkalinity and pH setting in groundwater 514 and stream. This diffuse groundwater discharge pattern is different from previous studies where 515 most CO2 was emitted to the atmosphere due to focused groundwater discharge (Duvert et al., 516 2018; Johnson et al., 2008). Thus, the stream CO 2 evasion in our study catchment was driven by 517 the carbon transfer but limited by the carbon supply (most reaches have high CO 2 gas transfer 518 velocity but relatively low stream pCO 2 ). 519 manuscript submitted to Water Resources Research 31 520 Figure 11. The conceptual model demonstrates that diffuse groundwater discharge dominates terrestrial DIC 521 export, and carbonate buffering process regulates stream CO 2 pool through transformation between CO 2 * and 522 HCO 3 -. This carbonate buffering process can either enhance (i.e., HCO 3 transformed into CO 2 ) or limit (i.e., 523 CO 2 transformed into HCO 3 -) stream CO 2 evasion. In our study stream, the carbonate buffering largely limited 524 the stream CO 2 evasion. Wallin et al., 2013) due to high internal production in these environments. However, headwater 530 streams in arid and semiarid regions are likely to be a significant "transfer station" for terrestrial 531 carbon export and release to the atmosphere because of their close connection with terrestrial 532 ecosystem through diffuse groundwater discharge. Our reach-scale carbon budget results indicate 533 that stream CO 2 evasion rates (0.62-3.18 g C m -2 d -1 in our study) could be comparable to the 534 average CO 2 efflux of conterminous US streams (2.42-10.98 g C m -2 d -1 ) (Butman and Raymond, 535 2011). Comparison in headwater stream CO 2 evasion rates between our study, peatland and 536 forested headwater streams suggests that headwater stream CO 2 evasion from arid and semiarid 537 regions may be as important as that from humid regions (SI Table S6). 538 Former studies pointed out semiarid headwater streams may also be hotspots for CO 2 539 evasion (Gómez-Gener et al., 2015; Schiller et al., 2014). Our CO 2 evasion rates are higher than 540 those reported in these studies (Mediterranean rivers, 0.20-2.63 and 0.49-1.15 g C m -2 d -1 , 541 respectively). We attributed the higher stream CO 2 evasion rates in our study to the greater 542 diffuse groundwater discharge rates and higher CO 2 gas transfer velocities (transfer driven). As 543 our survey was conducted during the dry season, our results may represent the lower bound of 544 the Hailiutu River CO 2 evasion rates. Larger CO 2 evasion rates are expected to occur when 545 groundwater discharge is higher during the wet season. 546

547
In this study, we discovered that headwater streams in arid and semiarid areas are 548 significant sources of CO 2 to the atmosphere. These understudied streams received a 549 considerable amount of dissolved CO 2 from terrestrial ecosystems via diffuse groundwater 550 discharge. Interestingly, a large portion of dissolved CO 2 was not directly and quickly emitted to 551 the atmosphere, but transformed into HCO 3 through carbonate buffering. The stream CO 2 552 evasion was driven by fast carbon transfer processes between terrestrial ecosystems, stream and 553 atmosphere, but limited by relatively small carbon supply in stream due to the inhibition of 554 carbonate buffering. To the best of our knowledge, previous studies seldom integrated the vital 555 contribution of terrestrial carbon export via diffuse groundwater discharge to headwater stream 556 carbon budget, which may underestimate headwater stream CO 2 evasion rates (Duvert et al.,