High Riverine CO 2 Outgassing affected by Land Cover Types in the Yellow River Source Region

Mingyang Tian, Xiankun Yang, Lishan Ran, Yuanrong Su, Lingyu Li, Ruihong Yu, Haizhu Hu, Xi Xi Lu Inner Mongolia key laboratory of river and lake ecology, School of ecology and environment, Inner 5 Mongolia University, Hohhot, 010021, China School of Geographical Sciences, Guangzhou University, Guangzhou, 510006, China Department of Geography, The University of Hong Kong, Hong Kong, China Department of Geography, National University of Singapore, 117570, Singapore 10


Introduction
Rivers connect land and oceans, acting as pipes and containers transporting carbon and other substances from terrestrial ecosystems to the oceans.Existing studies on riverine CO2 evasion focus mainly on the spatial and temporal dynamics of partial pressure of CO2 (pCO2) and CO2 efflux (FCO2) (Cole et al.,2001;Aufdenkampe et al., 2011;Raymond et al., 2013;Abril et al., 2014).Many researchers have argued that river water CO2 is primarily derived from respiration of terrestrial ecosystems and decomposition of organic matter in river (Raymond et al., 2013;Hotchkiss et al., 2015;Schelker et al., 2016;Ran et al., 2017).For example, Abril et al. (2014) pointed that wetlands are the primary source of riverine CO2 emissions in the Amazon river.However, the sources and underlying mechanisms of riverine CO2 dynamic for many rivers remain largely unknown.Therefore, to more accurately estimate riverine CO2 outgassing and understand its driving factors, more studies focusing on rivers in particular climates (i.e., alpine climate) and regions (e.g., headwater region or intermitted rivers) are strongly needed to gain deeper insights into global carbon balance processes.
With respect to global-scale CO2 outgassing, available estimates are characterized by great uncertainty.
For example, recent global CO2 outgassing fluxes from rivers and streams range from 0.65 to 3.2 P g C yr −1 (Raymond et al., 2013;Lauerwald et al., 2015;Swakuchi et al., 2017;Drake et al., 2017), which are considerably higher than the earlier estimate by Cole et al. (2007) (i.e., 0.23 P g C yr -1 ).A major reason for the huge range is likely the absence of a global CO2 outgassing database which includes direct CO2 emission measurements over different rivers and under different climate and land cover types (Raymond et al., 2013;Cole et al., 2007;Aufdenkampe et al., 2011;Drake et al., 2017).More direct field measurements are therefore strongly needed to better refine global CO2 efflux estimates.
Yet, there have been few studies on CO2 effluxes of rivers in extreme geographical and climatic conditions, such as alpine rivers (Wu et al., 2008;Zhang et al., 2013).Crawford et al. (2013) investigated the riverine CO2 outgassing in the Alaska region and explored its temporal and spatial changes under different land use types.Crawford et al. (2015) further studied carbon emissions from the rivers and lakes in alpine areas around the Estes Park in the United States and found that the average pCO2 was only 417 μatm.They concluded that high altitude and low vegetation coverage are the primary factors limiting CO2 outgassing.Weyhenmeyer et al. (2015) concluded that production of CO2 in lakes was usually half of the CO2 emissions and most of the degassed CO2 was derived from dissolved inorganic carbon (DIC).Humborg et al. (2010) surveyed rivers in central and northern Sweden and determined that the average pCO2 and FCO2 was 1445 μatm and 3033 g C m -2 yr -1 , respectively.Overall, compared with temperate and tropical rivers, riverine CO2 outgassing under alpine climate is at a relatively low level.This is largely due to the cold climate with low temperature and high altitude that hamper riverine CO2 emissions (Peter et al., 2014).
The riverine CO2 emissions from the Yellow River Basin have been preliminarily studied.Su et al. (2005) reported that the mainstream pCO2 was between 1100 and 1700 μatm, which were in intermediate-low level of world rivers.The main controlling factor was its carbonate system.Zhang et al. (2008) measured the pCO2 of 1570 μatm at Lijin Hydrological Station on the lower Yellow River during sediment regulation period (June-July), which was higher than in other periods.Zhang et al. (2009) measured the FCO2 of the Yellow River and concluded that the Yellow River waters were a source of atmospheric CO2 during autumn and the flux was about 0.0174 Tg C, which was similar to that of the Ottawa River but far less than that of the Amazon in autumn.Ran et al. (2015b) estimated that the annual CO2 emissions of the whole Yellow River system at 7.9 Tg C, which is close to the basin-wide carbon deposition of 8.7 Tg C while larger than the marine import (i.e., 6 Tg C).Ran et al. (2017) further studied the Wuding River, a tributary of the middle Yellow River, and concluded that lateral carbon derived from soil respiration and chemical weathering played a central role in controlling the riverine pCO2.In addition, radiocarbon analyses of the degassed CO2 suggest the release of old carbon previously stored in soil horizons (Ran et al., 2018).
These studies on CO2 emissions from the Yellow River were mainly confined to its middle and lower reaches.In contrast, to date little has been done on the upper reaches, especially the source region on the Tibetan Plateau.The Yellow River source region is located in the alpine zone with the Yellow River mainstream flowing through a variety of land cover types, including grassland, wetland, glacier, and permafrost.Affected by increasing temperature as a result of global warming, the alpine rivers in this region have become hot spots of riverine carbon cycle studies and warrant a thorough understanding of their implications for global climate change (Ulseth et al., 2018;Peter et al., 2014;Hood et al., 2015).
Although Ran et al. (2015b) have estimated its pCO2 and FCO2 by using water chemistry data, there are no field-based direct measurements of CO2 emissions from these alpine rivers.
To accurately determine the magnitude of riverine CO2 outgassing and understand its underlying control mechanisms in this alpine climate region, we conducted in situ measurements of riverine CO2 emissions under different land cover types, including grassland, peatland, glacier, and permafrost, in the Yellow River source region.The objectives of this study were to examine (1) the spatiotemporal patterns of CO2 emissions under different land cover types; (2) the magnitudes of stream CO2 emissions; and (3) the sources of riverine CO2 in this alpine river system.Clearly, the obtained findings will lead to a greater understanding of riverine carbon export and CO2 emissions, especially for alpine rivers, which will help refine the global estimates of riverine FCO2.

Site description
The Yellow River originates from the Bayanhar Mountains in Tibetan Plateau, flows through the Loess Plateau and North China Plain, and eventually empties into Bohai Sea.Generally, the drainage basin above the Tangnaihai hydrological station is called the Yellow River source region (Figure1).The study area is situated from 32°3'N 95°5'E to 36°1'N 103°3'E (Figure 1).In this region, most of the rivers flow through the Tibetan Plateau at an altitude of 3000-4000 m with meandering river channels.The study area is about 1.32×10 5 km 2 , accounting for about 17.6 % of the Yellow River basin.The Yellow River source region is located in an alpine zone with a typical plateau continental climate affected by plateau monsoon (Yang et al.,1991).Its lithology is homogeneous and predominantly composed of shale and granite rocks (Chen et al., 2005).The climate is characterized by a pronounced seasonal variation with the wet season starting from June to September and the dry season from October to next May.Major land cover types of the source region include glacier, permafrost, wetland, and grassland.
Precipitation is the dominant source of runoff in the Yellow River source region.Its annual mean precipitation is 486 mm, accounting for approximately 96% of the total runoff (Liu et al., 2005).The annual evaporation varies from 800 to 1200 mm.Although the area of the source region represents only 17.6% of the whole Yellow River basin, it supplies over 33% of the basin's total water discharge (Sun et al., 2009).In recent decades, precipitation in the source area has slightly increased owing to accelerating glacier melting (Chang et al., 2007), which has increased its relative importance of water flux for the whole Yellow River basin (Zhang et al., 2012).

Fieldwork and laboratory analyses
In this study, four field work campaigns in the Yellow River source region were conducted in April, June, August, and October 2016.The riverine pCO2 and related environmental factors, including water temperature, pH, dissolved oxygen (DO), were monitored in the field under different land cover types.
In total, there are 36 sampling points (Figure 1) and they can be categorized on the basis of complexity of river network structure and land cover types (i.e., glacier, permafrost, wetland, and grassland) (Table 1).In addition, three groundwater samples in grassland covered sub-catchments were collected to determine the pCO2 in groundwater.The temperature, pH, and DO were measured by using a Multi 3420 analyzer (WTW GmbH, Germany) with the accuracies of ±0.2 °C, ±0.004, and ±1.5%, respectively.Before measurement, the pH probe was calibrated with three pH buffers (i.e., pH4.01, pH7.00, and pH10.01,respectively).
Prior studies suggested that, when pH ranges from 7 to 10, HCO3 -represents 96% of alkalinity and alkalinity can be used to calculate DIC (Hunt et al., 2011).Alkalinity was determined by on-site titration in this study.The collected water samples were subjected to low-pressure suction filtration through a prefired glass fiber filter (Whatman GF/F, GE Healthcare Life Sciences, USA) with a pore diameter of 0.7 μm.For each water sample, the alkalinity was titrated with 0.1 mol L -1 HCl within 12 hours after sampling.Triplicate titrations with Methyl orange as the indicator suggest that the analytical error below 3%.Beside alkalinity analysis, the remaining filtered water was transferred into 100 ml amber glass vials, poisoned with nitric acid, and preserved in refrigerator at 4 ᵒC condition for dissolved organic carbon (DOC) measurement in laboratory.DOC was analyzed using a total organic carbon (TOC) analyzer (Elementar Analysensysteme GmbH, Germany), which has a precision better than 3%.

Determination of CO 2 emission
The CO2 emission flux FCO2 was measured using the floating chamber method (Ran et al., 2017) with a Li-7000 CO2/H2O gas analyzer (Li-Cor, Inc, USA), which has a precision better than 1%.The Li-7000 gas analyzer was calibrated with standard CO2 gases of 500 ppm and 2000 ppm before each measurement.
The rectangular floating chamber has a volume of 17.8 L and a water surface area of 0.09 m 2 .The chamber walls were lowered 3 cm into water and mounted with plastic foams that had streamlined ends to limit artificial disruptions to near-surface turbulence.The chamber is covered with tin foil to reduce the influence of sun light's heating.Temperature inside chamber was measured with a waterproof thermometer.Prior to each deployment, the chamber was placed in air and the air inside the chamber was continuously circulated in a closed loop that was connected to the infrared Li-7000 gas analyzer through rubber-polymer tubes.The instrument automatically records the air CO2 concentration and ambient atmospheric pressure.When the chamber was placed on water surface, the accumulating CO2 concentration inside the chamber was recorded every 2 seconds, and each deployment lasted for 6-10 mins.In large rivers with relatively favorable flow conditions, the chamber was tied to a small rubber boat and freely drifted with flow to measure FCO2.In contrast, we used the static chamber method to measure FCO2 in small rivers or streams which may have caused an overestimation of CO2 evasion (Lorke et al., 2015).While the chamber was freely drifting at 32 sampling sites, we used the static deployment method only at 4 sampling sites, accounting for about 10% of the all sites.
The CO2 efflux from water was calculated using following equation (Frankignoulle et al., 1988): where, dpCO2/dt is the slope of CO2 change within the chamber (Pa d -1 ; converted from μatm min -1 ), V is the chamber volume (17.8 L), R is the gas constant, T is the chamber temperature (K), and S is the area of the chamber covering the water surface (0.09 m 2 in this study).
Surface water pCO2 was calculated using the headspace equilibrium method (Ran et al., 2017).By using an 1100 mL conical flask, 800 mL of water were collected 10 cm below water surface and the remaining volume of 300 mL was filled with ambient air.The flask was immediately closed with a lid and vigorously shaken for 1 min to equilibrate the gas in water and air.The equilibrated gas was then injected into the calibrated Li-7000 gas analyzer.Triplicate measurements were performed at each site and the average was calculated (analytical error below ±3%).Surface water pCO2 was calculated based on the equations from Dickson et al. (2007): where, the superscripts i and f represent the initial and final pCO2 (μatm), Vh and Vw are the headspace volume and water volume, respectively, K0 is the solubility of CO2 in water calculated on the basis of solubility constants for CO2 from Weiss (1974), K1 and K2 are the thermodynamic reaction constants (Lueker et al., 2000), [H + ] represents the total concentration of hydrogen ions in final solution.R is the universal gas constant (8.314J mol -1 K -1 ), and T is the water temperature (K).Temperature in the flask after equilibration was measured to correct for temperature changes relative to that of in situ river water.
The initial pCO2 was taken as the CO2 concentration in ambient air before the headspace equilibration measurement.
Conventionally, FCO2 can also be estimated from the following equation.
where, k is the gas transfer velocity (m d -1 ), KH is the Henry's constant for CO2 at a given temperature, FCO2 is the measured riverine CO2 efflux, and the △pCO2 is the difference between the surface water and the atmosphere.Using the field-measured pCO2 in surface water and air, k can be computed by rearranging Equation (3).To compare our calculated k value with other studies, it was standardized to a Schmidt number of 600 (k600) by assigning the Schmidt number exponent to be 0.5 (Jähne et al., 1987).
We also predicted the k600 (m d -1 ) through the Model 5 developed by Raymond et al. (2012).
where, V is the stream velocity (m s -1 ), S is the slope of rivers (unitless).
Previous studies indicate that k600 is affected by a number of environmental factors, such as wind speed, slope, flow velocity, depth, and discharge (Wanninkhof et al., 1992;Zappa et al., 2007;Raymond et al., 2012).Using only flow velocity and slope of river channels would have caused overestimation for mountainous rivers due to their relatively high channel slope and thus higher flow velocity.Therefore, the extremely high k600 values calculated from Equation (3) were excluded from the comparison between our calculated k600 and the modeled k600.
Spatial variability of the air temperature was consistent with that of the water temperature at almost all the sites, although it could be as high as 33 °C.The annual average air temperature in 2016 was 16.7±6.3°C.
Water pH ranged from 6.97 to 9.02 with an average of 7.89±0.64(Table 1 1).Alkalinity ranged from 600 to 7600 μmol L -1 with an average of 2871±1381 μmol L -1 (Table 1).Alkalinity was higher in the cold months (3378 μmol L -1 in April and 2941 μmol L -1 in October) than in the warm months (2644 μmol L -1 in June and 2326 μmol L -1 in August).
The pCO2 value showed different temporal variation characteristics for the four land cover types (Figures 2a, 3a, and 2c).In grassland, the average river pCO2 value in April, June, August, and October was 836±258 μatm, 609±297 μatm, 1086±551 μatm, and 734±253 μatm, respectively.In comparison, the average peatland river pCO2 in April, June, August, and October was 875±436 μatm, 792±436μatm, 1156±630 μatm, and 926±285 μatm, respectively.The pCO2 in these two land cover types showed the same temporal pattern with the highest pCO2 occurring in August and the lowest in June.

Spatial and temporal variations of FCO 2
CO2 emissions exhibited spatial and seasonal variations among the 36 stream sites (Table 1, Figures 2b,   3b, and 3d).The CO2 effluxes ranged from -221 to 1469 g C m -2 yr -1 in April, -144 to 6892 g C m -2 yr -1 in August, and -34 to 2321 g C m -2 yr -1 in October.While the highest FCO2 was measured at the wetland sites (Site Pt 3 in August, 6892 g C m -2 yr -1 ), the lowest FCO2 was observed at permafrost sites (Site Pm 3 in April, -221 g C m -2 yr -1 ) (Table 1).The averaged FCO2 of all sites was 479±436, 261±205, 873±1220, and 714±633 g C m -2 yr -1 in April, June, August, and October, respectively.Clearly, rivers in the Yellow River source region were net carbon sources for the atmosphere, despite the great spatial and seasonal FCO2 variations.When grouped by land cover types, the mean CO2 efflux shows a clear decreasing trend from wetland (767±1644 g C m -2 yr -1 ) through grassland (679±610 g C m -2 yr -1 ) and glacier (508±588 g C m -2 yr -1 ) to permafrost (302±±349 g C m -2 yr -1 ).Because the intensity of CO2 emissions depends on river pCO2, the FCO2 showed a similar spatial and temporal pattern to the pCO2, although the highest and lowest pCO2 and FCO2 value were not found at the same sampling sites.

Impact of land cover types on riverine pCO 2 and CO 2 outgassing
This study shows that the lowest FCO2 appeared in the permafrost covered region among all land cover types, with the annual average at FCO2 of 302±349 g C m -2 yr -1 .It is well known that a large quantity of riverine CO2 is derived from land (Dinsmore and Billett., 2013;Hope et al., 2004).Particularly, rivers flowing through permafrost are characterized by higher organic carbon input from soils (Zeng et al., 2004), which can support higher riverine DOC export and lead to stronger CO2 outgassing.The correlation analysis between hydro-chemical parameters and pCO2 in the permafrost region showed that, while alkalinity, DO and DOC were not significantly correlated with pCO2, pH exhibited a statistically significant relationship with pCO2 (Figure 4).The negative relationship between pCO2 and pH is likely because dissolved CO2 itself acts as an acid in water (Stumm and Morgan., 1996).In poorly buffered systems like the study area, CO2 can be a strong control on river water pH (Neal et al., 1998;Waldron et al., 2007).The DOC concentrations in the permafrost rivers (mean: 5.0±2.4mg L -1 ) were relatively higher than that in the glacier rivers (mean: 3.6±1.1 mg L -1 ) and the grassland rivers (4.6±2.3 mg L -1 ) but were comparable to the peatland rivers of 5.1±3.7 mg L -1 in peatlands.Additionally, the average alkalinity concentration in the permafrost region is the highest among the four land cover types.However, the pCO2 and FCO2 values in this region were always the lowest during the four campaigns.One potential explanation is that its low temperature (i.e., annual average water temperature: 9.9 ℃) because of high elevation may have constrained soil respiration and riverine organic matter degradation (Battin et al., 2008).Furthermore, although there is sufficient dissolved CO2 in the river water, it may be difficult for CO2 to degas from rivers in view of the low temperature (thus strong solubility) and low flow velocity (average: 0.8±0.5 m s -1 ) (Alin et al., 2014).The lower temperature is likely the major reason for the high riverine DOC concentrations while low CO2 outgassing rates in the permafrost region.
Because the glacier region exhibits similar temperatures and elevations to the permafrost, its pCO2 and FCO2 values were also relatively low, with the average only at 657±240 g C m -2 yr -1 .This is probably because all the sampling sites are located on the 1-2 order streams characterized by strong hydrologic connection with the terrestrial landscape (Sorribas et al., 2017;Smits et al., 2017) , and the surrounding catchment is lack of exogenous terrestrial carbon input.The river water alkalinity of the glacier rivers showed constantly the lowest level throughout the study year (Table 1), due largely to the low coverage of carbonate rocks.For the glacier rivers, only the DOC was significantly related to pCO2 (Figure 5d, r 2 =0.56, p＜0.001).The glacier sampling sites are mainly located around the Aemye Ma-chhen Range (Figure 1).Wang (1998) discovered that these rivers are predominantly supplied by glacier melting that is characterized by significant seasonal variability.The sampled glacier rivers showed the lowest annual average DOC concentration among the four land cover types (3.6±1.1 mg L -1 ).This is probably because the sub-catchments around the Aemye Ma-chhen Range do not have sufficient vegetation coverage as a result of high elevation and low temperature, limiting the terrestrial source of DOC.Poor soil, short water retention time, and low precipitation are the main reasons for the low vegetation coverage in this region (Lu et al., 2001).The rivers flowing down the snow mountain cut deep into the B horizon of soils because of strong glacial erosion and retreat.Almost all the glacial sampling sites are characterized by gravel channel, limiting the supply of terrestrial organic carbon into river carbon pools.As a result, the measured DOC concentrations in most of the sampled glacier rivers were very low.For glacial rivers, if there is no external supply of DOC, a complete decomposition of the river water DOC can only produce 0.34 μmol L -1 CO2.This suggests that the CO2 produced by DOC degradation in the glacial river cannot maintain such a high CO2 outgassing rate.The modern snow and ice which are important water sources in the Aemye Ma-chhen Range do not have enough DOC, DIC, or CO2 contents (Wu et al., 2008).Instead, chemical weathering may have played a crucial role in supporting glacier riverine CO2 (Wu et al., 2005;Wu et al., 2008).Previous studies have shown that glaciers contain large amounts of CO2 (Meese et al., 1997) and DOC (Hood et al., 2009;Singer et al., 2012), which are important sources of CO2 for glacial rivers.Our observations found that, with increasing distance from the glaciers, the riverine pCO2 exhibited a decreasing trend, which is likely caused by the dilution of glacier-related pCO2.
The FCO2 was highest in the peatland rivers among the studied 4 land cover types.Only the pH showed a negative linear relationship with the pCO2, while the alkalinity had a weak linear relationship with the pCO2 (Figure 6).For peatland rivers, terrestrially-derived organic carbon has been widely recognized an important source of riverine CO2 (Abril et al., 2014;Müller et al., 2015;Billett et al., 2015, Hu et al., 2015).There are a variety of sources for DOC in the peatland.First, the soil in the wetland ecosystem is rich in peat soil.The amount of peat stock in the Zoige Peatland is estimated to be 1.9 billion tons, accounting for about 40% of China's marsh wetland carbon storage (Wang et al., 2012).These carbon supplies to river carbon pools are an important driver for the high FCO2 in the wetland rivers.In addition, soil pore water enriched with high concentrations of dissolved CO2 continues to enter river waters, which can provide enough riverine CO2 (Butman et al., 2011).Furthermore, vegetation in the peatland region can import large amounts of CO2 into the river water through two mechanisms.On one hand, vegetation litter and root exudates release degradable organic matter into rivers.Decomposition of these organic matter serves as a carbon source for heterotrophic microorganisms.During this process, heterotrophic organisms release CO2 into water (Abril et al., 2014).On the other hand, respiration of plant roots and soil microorganisms that are submerged in wetland soils could also release CO2 directly into river water (Abril et al., 2014).The combined effects of these factors have resulted in rivers with high DOC and FCO2values in wetlands.
The average FCO2 in the grassland rivers of 818±394 g C m -2 yr -1 is at a moderate level, lower than the wetland FCO2 but considerably higher than that in the glacier and permafrost rivers.Correlation analyses between water chemistry parameters and riverine pCO2 for the grassland rivers showed that both pH and DOC had weak correlations with pCO2 (Figure 7).This also suggests that pCO2 is partially affected by the water pH.Compared to the other three land cover types, grassland has been substantially affected by human activity (i.e., grazing).Consequently, besides the DOC derived from physical erosion, the pollutants produced by grazing are also important sources of riverine DOC.The average pCO2 in peatland is 15% higher, but the average DOC concentration in wetland is 11% higher than that in grassland, and the alkalinity in grassland rivers is 46% higher than that in the wetland rivers.In addition, DIC is an important source of riverine CO2 for grassland rivers.While stream DIC source are highly variable across space and time (Smits et al., 2017), most of the HCO3 -in the Yellow River source region is derived from carbonate and silicate weathering (Wu et al., 2005;Wu et al., 2008;Wu et al., 2008), which largely reflects the contribution of groundwater inflow (Marx et al., 2017).Our groundwater samples from grassland region show an average pCO2 of 1976 μatm, which is 2.5 times the average pCO2 of the whole Yellow River source region.Therefore, the CO2 excess in the grassland rivers is more likely maintained by both the terrestrial organic carbon input and the inorganic carbon from groundwater.
With respect to the k600, the computed k600 showed statistically significant but weak correlation with the modeled results (Figure 8a) when the high k600 values (>70 m d -1 ) were removed from analysis.Given the chamber's dampening effect of wind (Matthews et al., 2003), there was no any statistically significant relationship between wind and k600 for streams.Instead, flow velocity is a relatively good predictor of k600 and can approximately explain 15% of its variability (Figure 8b).Although we deployed the floating chamber very carefully, the statistical analysis could not reflect the complex interactions of various environment factors except the four land cover types through our 36 sampling sites.Additionally, it is worth noting that the Model 5 of Raymond et al. (2012) has overestimated the k600, especially for mountainous rivers.This is probably because of low water temperature that has constrained CO2 degassing although the steeper channel slope has caused stronger flow turbulence (Battin et al., 2008).A low temperature will limit the rate of Brownian motion and reduce the CO2 exchange with the atmosphere.
Meanwhile, a low temperature will increase the solubility of dissolved CO2, thus reducing the outgassing of CO2.

Significance and implications for riverine carbon budgets
This study demonstrates that the annual average pCO2 is 771±380 μatm and FCO2 is 590±766 g C m -2 yr -1 in the Yellow River source region.In comparison, Ran et al. (2015a;2015b) estimated a considerably lower pCO2 value of 241±79 μatm and an areal CO2 efflux of is -221±112 g C m 2 yr -1 for the Yellow River source region, indicative of a strong carbon uptake from the atmosphere.Combining the seasonal difference of water surface area between the wet season (122 days and a water surface area of 770 km 2 ) and the dry season (243 days and a water surface area of 560 km 2 ), we estimated a total CO2 efflux from the Yellow River source region at 0.37±0.49T g C yr -1 .This suggests a net carbon source for the atmosphere.Our CO2 effluxes contrast with the earlier estimate by Ran et al. (2015b) which reported a carbon sink of -0.17±0.08T g C yr -1 .
Unlike our systematic sampling within the Yellow River source region, Ran et al. (2015 b) estimated its riverine CO2 outgassing by using only results at five sampling sites.There may have caused the huge CO2 efflux difference.Firstly, the sampling by Ran et al. (2015b) was confined to the mainstem and major tributaries, which may have underestimated CO2 emissions from lower-order headwater streams that usually present strong CO2 degassing (Butman and Raymond, 2011).For example, our sampling in the Zoige peatland rivers demonstrated that the lower-order rivers exhibit substantially higher FCO2 (767±1144 g C m -2 yr -1 ) than the Yellow River mainstem (351±306 g C m -2 yr -1 ).This reveals the impact of strong flow turbulence and land-river connectivity of low-order streams on sustaining the high CO2 effluxes (Crawford et al., 2013).In addition, the importance of groundwater inflow may decline with increasing stream orders, leading to a decreasing pCO2 and thus lower FCO2 (Marx et al., 2017).Another potential reason is that the number of sampling sites has limited the accuracy of CO2 emissions.This is highly possible for the Yellow River source region with the pCO2 in groundwater (1976 μatm) 2.5 times higher than that in the river (771±380 μatm).The CO2 originating from groundwater can be quickly released to the atmosphere within a short distance (Hotchkiss et al., 2015).Obviously, it is considerably challenging to detect the impact of groundwater inflow without high-resolution sampling.
While the Yellow River source region occupies 17.6% of the whole Yellow River basin, it accounts for only around 4% of the basin's total CO2 efflux (Ran et al., 2015a;2015b).The CO2 efflux of the Yellow River source region is also small compared with the effluxes from boreal river catchments (Teodoru et al., 2009;Butman and Raymond., 2011;Crawford et al., 2013;2015;Kokic et al., 2015;Looman et al., 2016) or even smaller relative to the global CO2 efflux (Aufdenkampe et al., 2011).Nevertheless, there is a huge carbon emission potential in the coming decades.Since the permafrost and wetland in the Yellow River source region are abundant in huge quantities of carbon storage.Continuously increasing temperature due to global warming will accelerate not only the mobilization of organic carbon in permafrost, but also the degradation of organic carbon by soil microorganisms.As a consequence, increasing riverine CO2 effluxes are highly anticipated and warrant further studies to comprehensively understand their implications for global carbon cycle and climate change.
We have comprehensively evaluated the riverine carbon dynamics within the Yellow River source region by means of in situ measurement of CO2 emissions under four different land cover types.However, it must be noted that there are still great uncertainties to be properly addressed in future studies.Despite the significant increase in the number of sampling sites compared with previous studies, less research on single watersheds that are spatially representative has been performed.Moreover, temporally continuous sampling involving the diel dynamics of riverine carbon export remains lacking.For example, prior studies suggest CO2 efflux during the daytime would be completely different from that at night and floods may have a huge shift on CO2 emissions (Geeraert et al., 2017;Smits et al., 2017).

Conclusions
Based on four rounds of field direct measurements of CO2 outgassing within the Yellow River source region, the average pCO2 in the study area was estimated at 771±380 μatm and the average FCO2 was 590±766 g C m -2 yr -1 .The FCO2 and pCO2 are lower than other rivers in the world and at a relatively low level compared to the middle and lower reaches of the Yellow River.The results showed that the rivers in the Yellow River source region were net sources of atmospheric CO2.Both the pCO2 and FCO2 showed strong spatial and temporal variations.The largest riverine CO2 efflux was found in August, followed by October and April, and the lowest was observed in June.When grouped into different land cover types, the FCO2 in the permafrost river was the lowest among the four types of land cover.The highest FCO2 was found in peatland rivers, followed by rivers in the grassland and glacier regions.
For the Yellow River source region with an alpine climate, the low temperature conditions have played a crucial role in limiting its biological activity and reducing CO2 emissions.As a consequence, these procedures control both the riverine CO2 sources and gas transfer velocity across the water-air interface.
The DOC concentration acts as an important control on riverine CO2 dynamics under all the four land cover types.In the permafrost region, the large amounts of terrestrially-derived DOC supported its high pCO2 levels.While in the glacier region, the glacial DOC and CO2 may have played an essential role in determining CO2 outgassing.In the peatland and grassland regions, decomposition of plant-derived organic matter is an important source of riverine CO2.Moreover, groundwater inflow and chemical weathering played an important role in supporting riverine CO2 for the whole Yellow River source region.
By integrating the seasonal changes of water surface area, the riverine CO2 efflux of the Yellow River source region was estimated at 0.37±0.49T g C yr -1 , which is significantly higher than earlier estimates (e.g., -0.168±0.084T g C yr -1 by Ran et al. (2015a;2015b).To date, very few studies have focused on the dynamics of riverine carbon cycling on the Tibetan Plateau river systems.This study provides insight into the riverine CO2 outgassing in the Yellow River source region, which will improve our current understanding of CO2 emissions from alpine rivers in the world, in particular these located on the Tibetan Plateau.
Table 1.Land cover types, altitude, stream types, pH, alkalinity, DOC, pCO2, and FCO2 of the 36 stream sites within the Yellow River source region, expressed in the order of April, June, August, and October in 2016.

Figure 2 .Figure 3 .
Figure 2. Spatial and temporal variations of annual average pCO2 (a) and FCO2 (b) within the Yellow River source region in 2016.