Isotopic composition and emission characteristics of CO2 and CH4 in glacial lakes of the Tibetan Plateau

Carbon dioxide (CO2) and methane (CH4) emissions from freshwater ecosystems are predicted to increase under climate warming. However, freshwater ecosystems in glacierized regions differ critically from those in non-glacierized regions. The potential emissions of CO2 and CH4 from glacierized environments in the Tibetan Plateau (TP) were only recently recognized. Here, the first direct measurement of CO2 and CH4 emission fluxes and isotopic composition during the spring of 2022 in 13 glacial lakes of the TP revealed that glacial lakes were the previously overlooked CO2 sinks due to chemical weathering in glacierized regions. The daily average CO2 flux was −5.1 ± 4.4 mmol m−2 d−1, and the CO2 consumption could reach 38.9 Gg C-CO2 yr−1 by all glacial lakes in the TP. This consumption might be larger during summer when glaciers experience intensive melting, highlighting the importance of CO2 uptake by glacial lakes on the global carbon cycle. However, the studied glacial lakes were CH4 sources with total emission flux ranging from 4.4 ± 3.3 to 4082.5 ± 795.6 μmol m−2 d−1. The large CH4 range was attributed to ebullition found in three of the glacial lakes. Low dissolved organic carbon concentrations and CH4 oxidation might be responsible for the low CH4 diffusive fluxes of glacial lakes without ebullition. In addition, groundwater input could alter CO2 and CH4 emissions from glacial lakes. CH4 in glacial lakes probably had a thermogenic source; whereas CO2 was influenced mainly by atmospheric input, as well as organic matter remineralization and CH4 oxidation. Overall, glacial lakes in the TP play an important role in the global carbon cycle and budget, and more detailed isotopic and microbial studies are needed to constrain the contributions of different pathways to CO2 and CH4 production, consumption and emissions.


Introduction
Inland waters, including rivers, lakes, wetlands and reservoirs, have been considered important greenhouse gas (GHG) sources to the atmosphere (Cole et al 2007, Bridgham et al 2013, Raymond et al 2013, Wang et al 2022, Lin et al 2023. The emission of carbon dioxide (CO 2 ) and methane (CH 4 ) from inland waters changes local atmospheric GHG levels and influences the heat exchange among different ecosystems (Walter et al 2008, Tranvik et al 2009, Michaud et al 2017. Currently, GHG emissions from melting cryospheres have drawn increasing attention on a global scale; however, the glacierized regions are poorly investigated. So far, several studies of GHGs from the cryosphere indicated that ice sheets were suggested as CH 4 storage and release media. For instance, the gas hydrate stability zone beneath the Ice Sheet could serve as a CH 4 sink, whereas, CH 4 in deep water is penetrating the window between the subsea and subglacial gas hydrate stability zone (Weitemeyer andBuffett 2006, Portnov et al 2016). The continuous export of CH 4 supersaturated water was found in the Greenland Ice Sheet, with high CH 4 concentration, which coincided with the subglacier flushing events (Lamarche-Gagnon et al 2019) and at the onset of the glacier melt (Christiansen et al 2021). Significant CH 4 production and release also occurred in a temperate glacier bed due to suitable sub-oxic conditions (Burns et al 2018). With glacier recession, subglacial ecosystems gradually turn into proglacial environments and net CH 4 production are observed (Bárcena et al 2010, Zhu et al 2020, Zhang et al 2021b, Du et al 2022. Exceptions exist in the Antarctic, where subglacial marine sediments and lakes were acting as potential CH 4 reservoirs or sink via aerobic, bacterial oxidation (Wadham et al 2012, Michaud et al 2017. Various CH 4 biological production and consumption pathways are mainly responsible for CH 4 source and/or sink during ice sheet and glacier melting. Whereas, CO 2 in glacierized regions has different characteristics. Besides biological production, abiotic processes also contribute to CO 2 . Meltwater in the Greenland Ice Sheets was found to be undersaturated, especially in the summer with large discharges (Meire et al 2015). The same phenomenon exists in glacierfed rivers and lakes in the Arctic (Pierre et al 2019), and this CO 2 undersaturation was caused by carbonate and silicate weathering reactions. As a result, the proglacial rivers and lakes were CO 2 sinks (Pierre et al 2019). However, different sources of subglacial CO 2 could make meltwater a CO 2 source. For instance, the enriched organic matter and its remineralization made the subglacial regions of Greenland Ice Sheet supersaturated with CO 2 (Pain et al 2021). Different pathways to CO 2 production and consumption determine CO 2 source or sink in glacier meltwater, which is worth a comprehensive investigation of carbon cycles in glacierized regions.
The heterogeneous characteristics of CH 4 and CO 2 in glacier meltwater are controlled by multiple mechanisms mentioned above, including those that either emit or consume CO 2 and CH 4 (Pierre et al 2019, Pain et al 2021. It is well known that bacterial CH 4 is primarily produced via two metabolic pathways, acetoclastic and hydrogenotropic methanogenesis. The former generates CH 4 from acetate, while the latter reduces CO 2 to CH 4 (Whiticar et al 1986). Meanwhile, CH 4 in freshwater could be generated through thermogenic processes as well (Wu et al 2014). These pathways could be evaluated via GHG carbon isotopic signatures. Sources of CO 2 are much more complex compared to CH 4 , and its isotopic signature should be discussed together with that of CH 4 , dissolved organic carbon (DOC) and dissolved inorganic carbon (DIC) to determine possible sources of CH 4 oxidation, DOC remineralization and potential removal by chemical weathering (Whiticar 1999, Pierre et al 2019, Christiansen et al 2021. The Tibetan Plateau (TP) has the largest extent of glaciers in the low and middle latitudes with the average elevation above 4000 m a.s.l. Rapid warming and changes in precipitation patterns are reshaping the hydrological process in the TP, facilitating the transport of the carbon from other terrestrial to water ecosystems (Yang et al 2019, Song et al 2020, Yao et al 2022. Due to these changes, glaciers in the TP are experiencing severe melting and rapid retreat in many mountain regions (Yao et al 2012, Kang et al 2020, Bhattacharya et al 2021, Miles et al 2021. Glacier melting favors the formation and development of a large number of glacial lakes mostly in the low-lying land. The number of glacial lakes in the TP increased at a rate of 306 lakes per year from 2008 to 2017 (15 348 lakes in 2017) (Chen et al 2021). The continuous number increase and area expansion of glacial lakes are expected, particularly in the high elevation glacier regions with enhanced glacier shrinkage (Chen et al 2021). The glacial lake related investigations can not only provide insights into both the current status of mountain glaciers but also the projection of further deglaciation (Carrivick and Tweed 2013). Several studies conducted in glacierized regions of the TP suggested that glacier meltwater could release CH 4 and mostly absorb CO 2 during glacier extensive melting periods , Zhang et al 2021b, Du et al 2022, which has important implications for the global carbon budget. However, it is still unclear how the formation of glacial lakes might impact the carbon balance among organic/inorganic carbon and CO 2 /CH 4 , as well as the production and consumption pathways of CH 4 and CO 2 .
Here, we conducted the first direct measurements of CO 2 and CH 4 emission fluxes and their isotopic composition (δ 13 C) over 13 glacial lakes in the TP during the glacier onset melting period (figure 1 and table S1). The aim is to understand the emission characteristics of CO 2 and CH 4 , and explore their potential production and consumption pathways in glacial lakes of the TP.

Study areas
GHG and water samples were collected from 13 typical glacial lakes, representing different climate regions and geological conditions (figure 1 and table S1). These lakes were chosen because they are relatively easy to access, since most glacial lakes are located Figure 1. The location of the studied glacial lakes in each glacierized region. Note: STBL and KLKL glacial lakes are around 10 km distant from each other. Two glacial lakes at SP glacial terminal were investigated named SP inner lake (SP-in) and SP outer lake (SP-out). Two glacial lakes located in up and down slope of KQGR glacier terminal named KQGR-up and KQGR-down glacial lakes were investigated. The database of glacial lake distribution in the TP was obtained from Chen et al (2021), which can be downloaded via https://zenodo.org/record/4275164. The glacier distribution database was obtained from Liu et al (2012), which is available via http://poles.tpdc.ac.cn/zh-hans/data/f92a4346-a33f-497d-9470-2b357ccb4246/.
in remote and high-altitude regions with complex terrain and beyond in situ observation. The elevation of collected glacial lakes range from 3500 to 5500 m a.s.l.

Sample collection and measurements
Field expeditions were conducted in May 2022 to investigate the characteristics of GHGs during the early glacier melting. The related in situ measurements and sample collections were generally carried out every 2-3 h for a whole day except for XXJLC glacial lake due to a terrible sampling condition of snow avalanche. The headspace equilibration method (Kling et al 1991, Raymond et al 1997 was adopted for CO 2 and CH 4 concentrations and carbon isotope measurements using Picarro G2201-I isotopic analyzer (O'Dwyer et al 2020). Standards for CO 2 and CH 4 concentrations and isotopes were analyzed prior to and during the sample measurements to ensure the stability of the instrument and make corrections for the measured data.
The carbon isotopic fraction (a c ) between CO 2 and CH 4 was calculated with the following equation (Whiticar et al 1986): where δ 13 C-CO 2 and δ 13 C-CH 4 are stable carbon isotope of CO 2 and CH 4 . CO 2 source apportionment was determined with a Bayesian tracer mixing model (MixSIAR) using δ 13 C-CO 2 in the atmosphere, δ 13 C-DOC and δ 13 C-CH 4 by assuming CO 2 in glacial lakes was mainly from atmospheric input, DOC remineralization and CH 4 oxidation. The δ 13 C values for three endmembers were measured based on our in situ collected samples.
Water samples were collected at an approximate depth of 10 cm below the water surface. DOC and cation samples were collected in 60 ml acid and pure-water washed polycarbonate bottles after being filtered through pre-combusted glass fiber filters and stored at −20 • C until analysis (Raymond et al 2004). Samples for DIC concentration and isotope measurements were collected in 125 ml pre-cleaned and precombusted (450 • C, 4 h) brown, gastight glass bottles free from air bubbles with the addition of 0.2‰ saturated HgCl 2 solution (Raymond et al 2004). DOC and DIC concentrations were analyzed with a TOC-500A analyzer (Shimadzu Corp, Kyoto, Japan). The DIC carbon isotopes were analyzed with a MAT-253 mass spectrometer coupled with a Gas Bench II automated device (Li et al 2021a). The DOC isotope composition was measured by ISO TOC CUBE-ISOPRIME100 (Elementar). The major cations were measured using a Dionex-6000 Ion Chromatograph (Yan et al 2018).
Water pH, conductivity, DO, TDS and temperature were measured in situ using a portable Multi meter (Multi 3630 IDS, Germany). 222 Rn was measured using RAD7 H 2 O (Durridge, Co.) in situ. Air temperature and wind velocity were recorded with a portable anemometer (LZ836).

CO 2 and CH 4 flux estimations
Total CO 2 and CH 4 emission fluxes from glacial lakes were evaluated using the updated floating chambers in Wang et al (2021) (figure S1). Six gas samples of approximately 100 ml were automatically collected into 200 ml pre-evacuated airtight gas sampling bags from the air headspace at the 10 min interval over a total sampling duration of 60 min in each gas sampling. The concentrations of CO 2 and CH 4 were measured using the same method as the equilibrated dissolved gas samples. The fluxes of CO 2 and CH 4 were calculated as below (Miller and Oremlan 1988): where dP/dt is the slope of gas accumulation in the chamber over the sampling period, V and A are the chamber volume (m 3 ) and chamber enclosed water surface area (m 2 ), R is the universal gas constant (8.314 m 3 atm −1 k −1 mol), and T is the temperature in the chamber (K). For glacial lakes with a distinct nonlinear concentration increase in chambers caused by ebullition (Yang et al 2020, Wang et al 2021, total fluxes were calculated using the initial and final headspace gas concentrations (Miller andOremlan 1988, Zhu et al 2016).

Isotopic composition of different dissolved components in glacial lakes
The δ 13 C-CO 2 of glacial lakes were relatively stable with the values ranging from −11.8 ± 0.7 to −16.7 ± 0.1‰ (table S1). The source carbon isotope of CO 2 from the Miller-Tans plot was −20.69‰ (figure 3). The δ 13 C-CH 4 had large range from −17.2 ± 1.9 to −52.0 ± 3.8‰ with depleted values found in KQGR-down, KLKL and STBL, where ebulltions were captured, and relatively enriched values in the rest of glacial lakes (table S1). The carbon source signature of CH 4 from the Miller-Tans plot was −46.12‰. The carbon isotopic fraction between CO 2 and CH 4 was less than or equal to 1.04. The δ 13 C-DIC was enriched with the average value of −5.1 ± 1.8‰ ranging from −2.6 ± 0.1‰ in QY to −7.6 ± 0.8‰ in XLH glacial lake. Whereas, the δ 13 C-DOC was depleted between −23.1 ± 0.6‰ in KLKL and −31.9 ± 0.7‰ QY glacial lake. As a tracer of groundwater, 222 Rn was consistently low with an exception in SP-out glacial lake (963.1 ± 379.8 Bq m −3 ).

Glacial lakes of the TP as CH 4 sources and CO 2 sinks
The positive total CH 4 fluxes suggest that the studied glacial lakes are acting as CH 4 sources. Given the relatively constant wind speed during chamber deployments, the larger CH 4 emissions in KQGR-down, STBL and KLKL glacial lakes were attributed to ebullitions with clear steplike increases in headspace CH 4 concentration (figures 2 and S2). Once the ebullitions were captured, total emission fluxes of these three lakes could be 100-700-fold larger than those measured in the rest of the glacial lakes. Although small bubbles also contribute to CH 4 increase in headspace of the chamber without changing concentration increase, total CH 4 fluxes were comparable in those glacial lakes without evident ebullitions (figure 2). The CH 4 ebullitions are stochastic; thus, not every floting chamber could capture the bubbles. Previous studies often adopt the gas transfer model to estimate diffusive flux used in evaluating the relative contributions of ebullition and diffusive fluxes , Wang et al 2021. In this case, the average fluxes from chambers without evident ebullition were defined as diffusive flux similar to that adopted by Zhu et al (2016) to avoid the method-caused divergences (Miller and Oremlan 1988). As a result, the ebullition fluxes in KQGR-down, STBL and KLKL glacial lakes could contribute as much as 91.4%, 94.8% and 81.8%, respectively, to total CH 4 fluxes, which were comparable to those in thermakarst lakes of the TP (Zhu et al 2016, Yang et al 2023. This result indicates that although total CH 4 emissions are relatively low (figure 2), when the ebullition occurs, it becomes the predominant CH 4 emission pathway and contributes largely to the total CH 4 emission during the onset melt of glacial lakes.
Total CH 4 emissions from glacial lakes were lower compared to other lakes worldwide. For instance, the investigated values were lower than global small ponds (800 ± 180 µmol m −2 d −1 ) and thermakarst lakes in western Siberia  types and study periods, thus, divergence existed in CH 4 flux comparisons between glacial and other typical lakes (Miller and Oremlan 1988, Kling et al 1992, Yan et al 2018. Despite this, our study shows that glacial lakes in the TP are acting as sources of CH 4 with much enhanced fluxes when ebullition occurs. The studied glacial lakes were acting as CO 2 sinks except SP-out glaciers located downstream of SP glaciers (figure 2). The depleted δ 13 C-DIC compared to the bulk carbonate indicated the CO 2 consumption in glacial lakes was attributed to the chemical weathering in glacierized regions (Skidmore et al 2004, Pierre et al 2019. This phenomenon was also found in glacier-fed rivers and lakes in the Arctic (Pierre et al 2019), and the carbonate tributary of a large northern Ottawa River (Telmer and Veizer 1999), where chemical weathering mainly caused CO 2 consumption. These results were consistent with a finding that chemical weathering was common in Asian glacierized regions (Li et al 2022). Besides chemical weathering, other processes, such as organic carbon oxidation, may also contribute to CO 2 flux. However, organic carbon respiration results in oversaturated CO 2 , thus, chemical weathering within the glacial lakes was the predominant process accounting for CO 2 undersaturation. To illustrate the chemical weathering process, four different contributions including atmospheric input, silicate weathering, carbonate dissolution and sulfide oxidation to major cations in glacial lakes were estimated based on the cation concentrations. The contributions of silicate weathering, carbonate dissolution and sulfide oxidation to water solutes have the range of 3%-31%, 31%-83% and 8%-57%, respectively, across the studied glacial lakes (figure S4), indicating CO 2 consumption during in situ measurements was probably induced by carbonate dissolution.
The negative CO 2 flux in glacial lakes of the TP is in stark contrast to that of global rivers and lakes which are generally considered as the ubiquitous carbon sources in the atmosphere (Bastviken et al 2004, Raymond et al 2013, Johnson et al 2022, Yang et al 2023, but consistent with some glacial rivers with undersaturated CO 2 in the TP (Zhang et al 2021b, Du et al 2022. The CO 2 consumption rate in the studied glacial lakes was almost comparable to that of glacial rivers and lakes in the Arctic despite different sampling periods (Pierre et al 2019), around 40% of that in Amazon rainforest (Espirito-Santo et al 2014, Brienen et al 2015, but much lower than that in the saline lake of the TP (Li et al 2021b). When extending this flux to the glacial lakes of the whole TP using glacial lake area from Chen et al (2021), CO 2 consumption could reach 38.9 Gg C-CO 2 yr −1 . This carbon consumption is expected to increase, as the TP glaciers' melting intensifies, if glacial lakes remain CO 2 sinks, because the chemical weathering rate would be much enhanced with increasing temperature (Yu et al 2021). This CO 2 consumption may offset the carbon emission from glacial lakes of the TP, playing an important role in the cryospheric carbon cycle.
As a natural tracer of groundwater and a driver of CO 2 and CH 4 emission from coast and lake waters (Santos et al 2012, Olid et al 2022, 222 Radon in SPout glacial lake which acted as CO 2 source was much larger than that of other glacial lakes (figure 2 suggesting the potential contribution from groundwater (with the 222 Radon of 2360 ± 555 Bq m −3 ) nearby SP glaciers. This phenomenon indicated groundwater input could alter GHG emissions from lake waters, which would be important in lake carbon budget as temperature rise is exacerbating groundwater inputs to lakes of the TP (Lei et al 2022). Other environment factors, such as conductivity and water temperature, had negative and positive influences on CO 2 and CH 4 fluxes, respectively (table S2).

Sources of CO 2 and CH 4 in glacial lakes of the TP
There are several potential sources of CO 2 in waters including the influx of atmospheric CO 2 (Pierre et al 2019), respiration in lake sediment and water column, organic matter remineralization, and oxidation of CH 4 in deep water (Tranvik et al 2009, Michaud et al 2017, Pain et al 2021. The source isotope of CO 2 from the Miller-Tans plot was between that of DOC and atmospheric CO 2 (with the average δ 13 C-CO 2 of −13.4 ± 0.53‰) (figure 3 and table S1). Therefore, although DOC concentrations were relatively low in some glacial lakes (figure 2), both processes exist in lake water. However, the consistent CO 2 undersaturation suggested CO 2 replenishment from the atmosphere was the main contributor to CO 2 in glacial lakes. The significant regression in Miller-Tans plot also indicated the importance of the atmospheric CO 2 input (figure 3), because concentration and isotope values of atmospheric CO 2 could be relatively stable; whereas, the two values of CO 2 from organic matter remineralization vary largely depending on organic matter quantity, composition and the remineralization rates, leading to the scatter rather than the linear regression in Miller-Tans plot, like those in the glacier meltwater of the Greenland Ice Sheet (Pain et al 2021).
The major biological CH 4 production is anaerobic fermentation based on simple organic compounds by acetoclastic or methylotrophic methanogenesis (Lamarche-Gagnon et al 2019). An alternative biological pathway is the reduction of CO 2 by H 2 with hydrogenotrophic methanogenesis in a lowsulfate anoxic environment (Michaud et al 2017). Additionally, the oxidation of CH 4 to CO 2 is a major CH 4 reduction pathway reducing net CH 4 emission. All these pathways can be largely evaluated using δ 13 C-CH 4 , a c value and ε c values (defined as (δ 13 C-CO 2 )-(δ 13 C-CH 4 )) due to the different isotopic signatures of CO 2 and CH 4 (Whiticar et al 1986, Pain et al 2021. However, the source signature from the Miller-Tans plot for the studied glacial lakes indicates the thermogenic origins of CH 4 (the process with δ 13 C-CH 4 heavier than −50‰) (Whiticar et al 1986, Wu et al 2014. This was quite different from that in glacial meltwaters of the Antarctic and Greenland Ice sheets (Michaud et    isotopic signature may indicate thermal development beneath the lake sediment after glacial lake formation. Meanwhile, the low a c values indicated that CH 4 production was not attributed to either acetate fermentation or CO 2 reduction (two processes with a c range of 1.04-1.09) (Whiticar et al 1986), further suggesting the potential thermogenic origins of CH 4 in glacial lakes. This is preliminary analysis of CH 4 sources, and further constraints with δD-CH 4 or CH 4 clumped isotopes are urgent because divergences for CH 4 sources exist in different studies of thermokarst lakes in the TP using mere δ 13 C-CH 4 . For instance, Wu et al (2014) first reported CH 4 in thermokarst lakes of the TP was of thermogenic origin, recently Mu et al (2023) and Yang et al (2023) suggested CH 4 from these lakes was derived from acetate fermentation and hydrogenotrophic pathways, respectively.
The ε c value can not only reflect methanogenesis but also the extent of CH 4 oxidation (Pain et al 2021). The ε c values were less than 30‰ except TGL, KQGR-down, STBL and KLKL glacial lakes suggesting that CH 4 ebullitions could escape from oxidation during release (Miller and Oremlan 1988) (figure 5). The lower ε c in rest glacial lakes (generally <30‰) was attributed to the occurrence of CH 4 Figure 5. Isotopic composition difference between CO2 and CH4 (εc = (δ 13 C-CO2)-(δ 13 C-CH4)) (a); significant correlation between δ 13 C-CH4 and water DO in studied glacial lakes of the TP. oxidation (Whiticar 1999). The significantly negative correlations between DO and δ 13 C-CH 4 further indicated CH 4 oxidation by archaeal in anaerobic water conditions (figure 5). As such, the weak CH 4 emission without ebullitions could be partly attributed to CH 4 oxidation.
Based on the discussion above, the potential contribution to glacial lake CO 2 includes atmospheric input, DOC remineralization and CH 4 oxidation.
With δ 13 C values of three contributors as endmember input data, MixSIAR quantified the major contribution of atmospheric input to glacial lake CO 2 (figure 6). Although there are some uncertainties in this estimate due to the limited data for each lake, the significant correlation between contribution ratios of atmospheric input and δ 13 C-DIC (p < 0.01, figure  S5) except SP-out glacial lake indicated that strong chemical weathering caused large atmospheric input of CO 2 . In conclusion, more than 60% CO 2 in glacial lakes of the TP during the study period came from chemical weathering induced atmospheric input, which is significant in the cryospheric carbon cycle.

Conclusions
This study investigated the emission fluxes and isotope composition of CH 4 and CO 2 in glacial lakes of the TP. The results suggest the distinct roles of CO 2 sinks and CH 4 sources in the studied glacial lakes. The CO 2 consumption rate was comparable to that of glacial rivers and lakes in the Arctic (Pierre et al 2019), which indicated CO 2 consumption may be a common phenomenon in glacierized regions. CO 2 consumption was attributed to chemical weathering. Chemical weathering rates beneath glaciers are expected to increase with enhanced glacier melting under a warming climate (Li et al 2022), therefore, if glacial lakes are consistent CO 2 sinks, the carbon sequestration would be larger than that estimated in this study. Meanwhile, air temperature increase in the TP may affect the relative abundances of some bacteria in glacial lakes which further influences GHG emission or consumption (Liu et al 2017, Yang et al 2023. Although ebullitions were captured in three of the studied lakes, the rest of glacial lakes in the TP are generally acting as slight CH 4 sources, and this carbon emission could probably be counteracted by CO 2 consumption, having a negative effect on global warming. The potential CH 4 anaerobic oxidation and low DOC content could partly explain this low CH 4 emission. The CH 4 thermogenic origin is still waiting for further constraint using δD-CH 4 or clumped isotopes. All these findings fill a critical gap in understanding GHGs in fresh water related carbon cycling in the TP. As the first in situ investigation for glacial lake CO 2 and CH 4 emissions and isotopic composition, the lakes in this study represented a minor glacier catchment of the TP, and long-term investigations should be carried out for large glacierized regions in the future to comprehend this unknown cryosphere carbon interaction and feedback in glacierized regions.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).