ARTICLE https://doi.org/10.1038/s41467-022-31219-1 OPEN Groundwater discharge as a driver of methane emissions from Arctic lakes Carolina Olid 1,2,3 ✉, Valentí Rodellas 4, Gerard Rocher-Ros 1,2, Jordi Garcia-Orellana 4,5, Marc Diego-Feliu4,5,6,7, Aaron Alorda-Kleinglass 4, David Bastviken 8 & Jan Karlsson 1 1234567890():,; Lateral CH4 inputs to Arctic lakes through groundwater discharge could be substantial and constitute an important pathway that links CH4 production in thawing permafrost to atmo- spheric emissions via lakes. Yet, groundwater CH4 inputs and associated drivers are hitherto poorly constrained because their dynamics and spatial variability are largely unknown. Here, we unravel the important role and drivers of groundwater discharge for CH4 emissions from Arctic lakes. Spatial patterns across lakes suggest groundwater inflows are primarily related to lake depth and wetland cover. Groundwater CH4 inputs to lakes are higher in summer than in autumn and are influenced by hydrological (groundwater recharge) and biological drivers (CH4 production). This information on the spatial and temporal patterns on groundwater discharge at high northern latitudes is critical for predicting lake CH4 emissions in the warming Arctic, as rising temperatures, increasing precipitation, and permafrost thawing may further exacerbate groundwater CH4 inputs to lakes. 1 Climate Impacts Research Centre, Department of Ecology and Environmental Science, Umeå University, 90187 Umeå, Sweden. 2 Department of Forest Ecology and Management, Swedish University of Agricultural Sciences, 90183 Umeå, Sweden. 3 UB-Geomodels Research Institute, Departament de Dinàmica de la Terra i l’Oceà, Facultat de Ciències de la Terra, Universitat de Barcelona, 08028 Barcelona, Spain. 4 Institut de Ciència i Tecnologia Ambientals, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain. 5 Departament de Física, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain. 6 Department of Civil and Environmental Engineering, Universitat Politècnica de Catalunya, 08034 Barcelona, Spain. 7 Associated Unit: Hydrogeology Group, UPC-CSIC, 08034 Barcelona, Spain. 8 Department of Thematic Studies—Environmental Change, Linköping University, 58183 Linköping, Sweden. ✉email: carolina.olid@ub.edu NATURE COMMUNICATIONS | (2022)13:3667 | https://doi.org/10.1038/s41467-022-31219-1 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31219-1 A major challenge to forecast future climate is constraining (ANOVA, df = 1, F = 0.50, p = 0.48). Similar to CH4, 222Rn and regulating fluxes of greenhouse gases (GHG) such as concentrations in groundwater (median 3500 Bq m−3, IQR methane (CH4)1. CH4 is a potent GHG responsible for 2100–8800 Bq m−3) were at least an order of magnitude higher one-quarter of the radiative forcing by all long-lived GHGs2. than in lake waters (110 Bq m−3, IQR 78–160 Bq m−3) and inlet Arctic lakes represent a large and climate-sensitive natural source streams (520 Bq m−3, IQR 260–2100 Bq m−3) (Supplementary of CH4 to the atmosphere, with emissions comparable to those Fig. 2), suggesting that 222Rn can be used as tracer for quantifying from northern high-latitude wetlands3. In the context of climate groundwater inflows into the lakes, as previously done in two warming, CH4 emissions from Arctic lakes are expected to studies in Alaska12,13. Similar ranges in 222Rn concentration were increase 2–3 fold by the end of the twenty-first century4,5, observed in summer (3700 Bq m−3, IQR 2600–8300 Bq m−3) and potentially constituting a strong positive climate feedback. Yet, autumn (3200 Bq m−3, IQR 2000–8200 Bq m−3) (ANOVA, df = the sensitivity of CH4 emissions from Arctic lakes to climate 1, F = 0.087, p = 0.77). change is highly uncertain because of a poor understanding of the The 222Rn mass balance (Supplementary Fig. 3) revealed that underlying mechanisms controlling lake CH4 cycling. groundwater discharge was an important water source for the CH4 emissions from lakes are generally regarded to be con- lakes (Fig. 2), except for the two shallowest lakes (BD09 and trolled by the imbalance between in-lake processes, including BD12). Groundwater inflows varied between lakes, with median CH4 production6,7 and CH4 oxidation8–10. However, high CH4 rates ranging from 0.18 to 6.4 cm d−1. Groundwater inflows concentrations in lake waters can also result from the large supply were within the range of the water discharge through the inlet of terrestrial CH4 through groundwater discharge11–13. CH4 streams (0.69 cm d−1, IQR 0.20–3.0 cm d−1, normalizing the inputs to lakes through groundwater and resulting emissions to point-source stream discharge by lake area) and comparable to the atmosphere can be important in the Arctic where wetlands those found in two other lakes in Alaska using a similar (hotspots of CH4 production) are abundant, and water flow paths approach (0.6–2.1 cm d−1)12,13. Groundwater inflows were are constrained within the shallow active layer (i.e., soil layer that higher in summer (range of 1.6–6.4 cm d−1) compared to thaws and refreezes annually), resulting in the supply of high autumn (range of 0.18–3.4 cm d−1). loads of inorganic and organic carbon (C) to surface waters11–13. CH4 inputs supplied by groundwater into the study lakes ranged While external inputs of terrestrial C through groundwater dis- from 28 to 120 mg CH4 m−2 d−1 in summer and 2.0 to 59 mg charge have been identified to have a critical influence on lake C CH4 m−2 d−1 in autumn, exceeding up to one order of magnitude cycling14,15, the significance of groundwater inflows for CH4 the CH4 inputs through the inlet streams (range of <0.01–1.3 mg emissions from lakes has rarely been addressed. CH4 m−2 d−1) (Fig. 3a). Similar summer groundwater CH4 inputs Two recent studies on single lakes demonstrated that were found in Landing Lake in Alaska (32–128 mg CH4 m−2 d−1)12. groundwater discharge is an important source of CH4 for lakes in Lower CH4 inputs through groundwater were found in Toolik Lake Alaska, suggesting that groundwater CH4 inputs could entirely in summer (1.6–11 mg CH4 m−2 d−1)13, likely due to the lower CH4 sustain CH4 evasion rates in summer12,13. However, since both concentrations in groundwater from the active layer (8–35 μM) mire CH4 production and export to receiving waters are highly compared to lakes in this study (50–210 μM). dependent on environmental factors such as temperature, water To better understand the significance of groundwater discharge table depth, active layer thickness, and topographic features16,17, for lake CH4 cycling, we compared groundwater CH4 inputs with snapshot observations in single lakes or seasons may not fully total CH4 emissions from the lakes. Diffusive CH4 fluxes to the represent the relevance of large-scale or year-round CH4 inputs atmosphere in summer ranged from 0.70 to 7.6 mg CH4 m−2 d−1. via groundwater. Extended studies covering multiple lakes and Lower diffusive CH4 fluxes were observed in autumn, ranging from seasonal variability are thus needed to better assess the role of <0.01 to 2.3 mg CH4 m−2 d−1. As ebullition fluxes were not directly groundwater discharge on lake CH4 emissions under different measured here, we used the results from 9 years of flux environmental conditions, not least in the Arctic, which experi- measurements in three lakes within the Stordalen mire, close to ences strong seasonality in terrestrial CH4 dynamics, precipita- the study lakes (Fig. 1), to estimate the potential contribution of tion, and runoff. To fill this gap, we combined measurements of ebullition in our lakes19. For lakes in Stordalen, diffusive CH4 CH4 and radon (222Rn), a natural tracer of groundwater, in 10 emission accounted for 17–52% of the ice-free CH4 flux, with the lakes and adjacent groundwater in the Arctic region of Sweden remainder being emitted via ebullition. The Stordalen lakes are (Fig. 1) to provide regional estimates of rates, patterns, and dri- situated in a unique palsa mire complex and are not fully vers of groundwater CH4 inputs during the ice-free season. representative of this landscape, likely leading to ebullition contribution more to the total lake emissions than other lakes in the region. Hence, to estimate maximum ebullition and maximum Results total lake CH4 emission—thereby making our comparisons between Groundwater CH4 inputs into lakes. We found a consistent groundwater CH4 inputs and total emissions conservative—we enrichment of CH4 and 222Rn in groundwater relative to surface assumed that diffusive flux and ebullition in all lakes accounted for waters. Concentrations of CH4 in groundwater (median 150 μM, 17 and 83% of the total atmospheric CH4 emission, respectively. interquartile range (IQR, 25th and 75th percentiles) 49–210 μM) Thus, potential ebullition rates were estimated to range from < 0.1 were more than two orders of magnitude higher than in lake to 37 mg CH4 m−2 d−1. This range is similar to ebullition fluxes waters (median 0.19 μM, IQR 0.02–0.48 μM) and inlet streams reported for small and midsize glacial and post-glacial lakes (IQR (median 0.02 μM, IQR 0.02–0.37 μM) (Supplementary Fig. 1). 0–15 and 3–27 mg CH4 m−2 d−1) across the boreal and Arctic Concentrations of CH4 in groundwater are similar to those in region20. These potential ebullition rates yield maximum total CH4 nearby sedge and Sphagnum permafrost mires (13–160 μM)18, and emissions that ranged from 4.1 to 44 mg CH4 m−2 d−1 in summer in the active layer from the continuous permafrost zone in Alaska and from <0.1 to 13 mg CH4 m−2 d−1 in autumn. Overall, these (Toolik Lake: 0.63–150 μM13; Landing Lake: 7.8–610 μM12). The potential total CH4 emissions were of the same order of magnitude high CH4 concentrations in groundwater suggests that even rela- as the CH4 inputs supplied by groundwater (Fig. 3a), suggesting tively low groundwater inflows into lakes may disproportionately that groundwater CH4 discharge can match total lake CH4 affect lake CH4 budgets. There was no evidence that groundwater emissions at a regional scale. CH4 concentration differed between seasons (120 μM and IQR Our results show that groundwater discharge to Arctic lakes is 18–240 μM in summer, 150 μM and IQR 88–210 μM in autumn) a pervasive external source of CH4, with the potential to sustain 2 NATURE COMMUNICATIONS | (2022)13:3667 | https://doi.org/10.1038/s41467-022-31219-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31219-1 ARTICLE Fig. 1 Map of the sampled lakes in Northern Sweden. Blue color indicates the study lakes and red lines show the corresponding catchments. The yellow star indicates the location of the Stordalen mire (Image source: © Lantmäteriet, 2021). Year 2018 2019 BD03 BD04 BD05 BD06 BD09 12 12 12 12 12 9 9 9 9 9 6 6 6 6 6 Groundwater discharge (cm d−1) 3 3 3 3 3 0 0 0 0 0 July September July September July September July September July September BD10 BD11 BD12 BD13 BD15 12 12 12 12 12 9 9 9 9 9 6 6 6 6 6 3 3 3 3 3 0 0 0 0 0 July September July September July September July September July September Fig. 2 222Rn-derived groundwater inflows into the study lakes. Values and error bars are obtained from the median and the 25th and 75th percentiles, respectively, of the 1000 Monte Carlo simulations for each lake and season. total lake CH4 emissions. However, other lake processes that can controlling lake CH4 budgets. These results help to understand substantially control lake CH4 emissions need to be considered. the disproportionate role of Arctic lakes as a landscape source of For instance, large amounts of CH4 are produced in sediments CH4 and highlight the need to consider groundwater CH4 inputs and transferred to lake waters, and a large share of the lake CH4 is to understand lake CH4 emissions at the catchment level. The consumed via oxidation in the water column8–10. Nevertheless, importance of groundwater inflows on controlling lake CH4 CH4 production21–24 and oxidation rates9,21–23,25 in lakes across emissions is further supported by the positive correlation the Arctic are comparable in magnitude to groundwater CH4 (R2 = 0.61, p = 0.005) between groundwater discharge rates and inputs found in this study (Fig. 4), which emphasizes the total CH4 emissions during summer (Fig. 3b). Unlike previous relevance of groundwater discharge as an important mechanism investigations based on single lakes12,13, this study reveals that NATURE COMMUNICATIONS | (2022)13:3667 | https://doi.org/10.1038/s41467-022-31219-1 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31219-1 a b July September total CH4 emission (mg CH4 m−2 d−1) 125 40 CH4 flux (mg CH4 m−2 d−1) 100 b 30 75 20 50 a 10 25 c c 0 0 Atm flux Groundwater Inlet Outlet 0 2 4 6 (Diff+Ebull) inflow streams streams Groundwater discharge (cm d−1) Fig. 3 CH4 fluxes from different sources and relationship between total CH4 emissions and groundwater inflow rates. a Inputs of CH4 to the study lakes are groundwater inflow and inlet streams, while outputs are emissions to the atmosphere (Atm flux) (by diffusion (Diff) and ebullition (Ebull)) and outlet streams. Box plots for groundwater CH4 inputs were generated by considering median values reported at each lake for each season. The boundaries of each box plot indicate the 25th and 75th percentiles of these fluxes, and the solid line in each box marks the median. Different lower-case letters indicate differences between water sources. The data used to generate the box plots is represented with gray circles. b In summer, as groundwater inflow rates increase, higher atmospheric CH4 emissions are found. The solid line represents the linear regression between total CH4 emissions and groundwater inflow rates (y = (6.5 ± 1.7) x + (0.1 ± 6.2), df = 8, F = 15, p < 0.005). The shaded area represents 95% confidence intervals. groundwater discharge is a key process controlling lake CH4 groundwater CH4 inputs into lakes based on spatial variables, which emissions at high latitudes and represents an important source of opens the door to future inclusions in regional assessments and/or CH4 at a regional scale. Therefore, including groundwater Earth system models. discharge to lakes in the global CH4 cycling may improve climate The impact of groundwater discharge on lake CH4 emissions predictions26. varied seasonally, as shown by the strong positive correlation between groundwater inflow rates and CH4 evasion in summer Spatial and temporal patterns of groundwater CH4 inputs to that was not observed in autumn (Fig. 3b). This seasonal lakes. There was a large variation in the magnitude of groundwater divergence indicates that the mechanisms driving groundwater inflows among lakes and seasons. The partial least squares regres- CH4 inputs to lakes may be sensitive to climatic conditions and sion (PLS) showed that selected catchment (percentage of wet zones likely reflect seasonal changes in hydrological and biological and open mires, catchment area, and slope) and lake characteristics drivers. For instance, snow melt during spring typically increases (mean depth) together with precipitation explained 72% of the groundwater recharge compared to the frozen period31–33, variability in groundwater inflows among lakes (Supplementary consistent with seasonal water table level variations in surficial Fig. 4). The best multiple linear regression model showed that the aquifer piezometers in the Arctic34,35. The lower magnitude of best predictors of groundwater inflows were mean lake depth groundwater inflows in autumn compared to summer observed in (depth), wet area coverage (wetzone) in the catchment, and this study agrees with the general hydrological cycle in the catchment slope (slope), with the resulting model explaining 45% of region36, with higher water flows in early summer compared to the variance (Qgw = 7.3 + 3.5·depth − 5.7 ·log10(wetzone) − 0.27· the autumn (Supplementary Fig. 5). log10(slope); df = 20, F = 7.2, p < 0.002). Groundwater inflows were Furthermore, changes in biological processes may also affect the positively related to lake depth, likely reflecting the higher water CH4 pool available for export via groundwater discharge. Produc- interception and groundwater connectivity of large and deep lakes tion rates of CH4 in mires are expected to be higher in summer compared to shallow systems. Contrary to our expectations, the compared to autumn due to warmer soil temperatures37,38. strongest explanatory variables, wet area, and mire catchment, However, CH4 concentration in mire groundwater measured in showed a negative relationship with groundwater inflows. We this study did not differ between seasons, possibly because of the initially expected that mire cover would positively affect ground- lower groundwater discharge (i.e., longer water residence time in water inflows to lakes as the cover of wet areas would represent the the soil) in combination with lower CH4 production rates in hydrological connectivity of lakes with the catchment. A possible autumn, which could result in similar levels of accumulated CH4. explanation for this result is that increasing mire cover is related to Overall, the strong correlation between groundwater inflows and flatter catchments, reducing the hydrological gradient and, conse- atmospheric CH4 evasion likely results from a higher discharge in quently, reducing the lateral water transport and increasing the early summer, although the role of biological processes in CH4 groundwater residence time in peatlands27. Our findings highlight cycling needs further attention8,39. The seasonal variability in the need to consider how groundwater inflows overlay landscape groundwater inflows into lakes observed here emphasizes the need patterns of CH4 production to better assess lake CH4 emissions, as for a better characterization of temporal variations of this external they may contrast and lead to unexpected responses on terrestrial driver for its inclusion in broad-scale estimates of lake CH4 CH4 export through groundwater. The complexity of hydrological emissions. pathways in a catchment and the number of variables involved in groundwater transport (e.g., permafrost coverage, preferential flow Implications for Arctic lakes under climate change. This study paths, hydraulic conductivity, and geological heterogeneities)27–30 unraveled the important role and drivers of groundwater dis- prevents any further assessment of the spatial variance of ground- charge for CH4 emissions from Arctic lakes. The comparison of water inflows. Regardless, we show that it is possible to predict the magnitudes of groundwater CH4 inputs to lakes with other 4 NATURE COMMUNICATIONS | (2022)13:3667 | https://doi.org/10.1038/s41467-022-31219-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31219-1 ARTICLE Sediment Diffusive flux Ebullitive flux Methane oxidation Groundwater inputs production 10 000.00 Permafrost category Continuous Sporadic Discontinuous Non permafrost 1 000.00 CH4 flux (mg CH4 m−2 d−1) 100.00 10.00 1.00 0.10 Dabrowsky Paytan This study et al. et al. Fig. 4 Comparison of groundwater CH4 inputs with other CH4 fluxes in Arctic lakes. Atmospheric fluxes (diffusion and ebullition) were extracted from the Boreal-Arctic Wetland and Lake Methane Dataset (BAWLD-CH4)20. Sediment production21–23 and oxidation9,21,23 rates were obtained from incubation experiments found in the literature (see Supplementary Data 1). potential CH4 sources and sinks (Fig. 4) suggests that ground- About 15% of the total catchment area is composed of lakes, concentrated in water discharge is a major mechanism controlling CH4 cycling in low elevation areas where this study was performed53. We selected lakes (n = 10) across the precipitation gradient (Fig. 1) to investigate seasonal patterns on CH4 Arctic lakes. Spatial variability in rates of groundwater inflow is inputs supplied by groundwater to Arctic lakes. All selected lakes were small (area mainly derived from the physical-hydrological characteristics of between 1.8 and 11.6 ha), with an average mean depth of 2.9 m and water volumes the catchment-lake continuum (lake morphology and vegetation ranging from 24,000 to 760,000 m3. cover). Seasonally, groundwater CH4 inputs to lakes are influ- enced by hydrological (groundwater recharge) and biological Estimating groundwater CH4 inputs into lakes using 222Rn. We quantified drivers (CH4 production rates). This information on the spatial groundwater inflow rates into the study lakes using the noble gas radon (222Rn) as tracer of groundwater inputs. 222Rn (T1/2 = 3.82 d) is a radioactive isotope pro- and temporal patterns of groundwater discharge in permafrost duced in the uranium (238U) decay series. Owing to its high enrichment in environments is fundamental to understanding the dynamics of groundwater and its conservative behavior in waters54, 222Rn is an ideal geo- landscape-level lake CH4 emissions and associated responses to chemical tracer to detect and quantify groundwater inflow rates into surface climate warming. waters55,56. We estimated groundwater inflow rates to the study lakes using a 222Rn Our results indicate that the multiple facets of climate change in mass-balance approach57,58. Major model assumptions include steady-state con- ditions over a relatively short period (1–3 days, comparable to 222Rn residence time the Arctic may exacerbate the magnitude of groundwater CH4 in the system59) and a well-mixed water column. inputs to lakes and subsequent emissions to the atmosphere. For Our survey showed that the 222Rn signal was relatively uniformly distributed instance, rising temperatures result in extensive degradation of horizontally and vertically in all the lakes, indicating well-mixed water columns. permafrost40,41, deepening active layer thickness, and higher The mass-balance approach is based on accurately constraining all the 222Rn sources (groundwater inflow (Fgw), diffusion of 222Rn from bottom sediments amounts of organic C exposed to decomposition, which can fuel (Fdiff), discharge from the inlet streams (Finlet), and in situ 222Rn production from methanogenesis in anoxic environments42. At the same time, CH4 decaying 226Ra dissolved in the water column (FRa)), and sinks (evasion to the production is highly dependent on temperature43 and thus may atmosphere (Fatm), losses through the outlet streams (Foutlet), and radioactive decay also increase with current warming, increasing CH4 concentration (Fdecay)). The change in 222Rn content over time [Bq d−1] can thus be described as: in groundwater from the active layer. Furthermore, precipitation is ∂ðC Rn;lake VÞ ¼ Fgw þ F diff þ F inlet þ F Ra  F atm  F outlet  F decay ð1Þ expected to increase up to 40% in the Arctic by the end of this ∂t century44 which, together with enhanced groundwater recharge This equation can also be expressed as: due to permafrost thaw and tundra greening45, may reshape the ∂ðC Rn;lake VÞ hydrology of the Arctic and increase groundwater inflows into ¼ Qgw C Rn;gw þ f diff A þ Qinlet C Rn;inlet þ λ VC Ra;lake ∂t ð2Þ lakes. Taken together, all these long-term changes can potentially  f atm A  Qoutlet CRn;outlet  λ VC Rn;lake increase the magnitude of lake CH4 emissions derived from where Qgw [m3 d−1] is the advective groundwater inflow; CRn,gw, CRn,lake, CRn,inlet, groundwater inputs. Thus, groundwater CH4 inputs constitute an and CRn,outlet [Bq m−3] are the 222Rn concentrations in groundwater, lake water, often-overlooked feedback in the ongoing climate change that inlet and outlet streams, respectively; Qinlet and Qoutlet are the mean water flow rates needs to be recognized and addressed appropriately to better [m3 d−1] for the inlet and outlet streams, respectively; CRa,lake [Bq m−3] is the 226Ra forecast future CH4 emissions from Arctic lakes. concentration [Bq m−3] in lake water; fdiff is the molecular diffusion flux of 222Rn from underlying sediments [Bq m−2 d−1]; fatm is the atmospheric flux of 222Rn to the atmosphere [Bq m−2 d−1], λ is the 222Rn decay constant [d−1]; and A [m2] and Methods V [m3] are the lake surface and volume, respectively. Note that this approach Study site. The Torneträsk catchment in Arctic Sweden (68.40°N 18.90°E; Fig. 1) lies assumes that all lakes are not losing water via groundwater, resulting in estimates of within the discontinuous permafrost zone along the 0 °C isotherm46. Mountain the minimum amount of groundwater discharging into the lakes. 222 permafrost is found approximately above 880 m a.s.l., whereas at lower elevations  We assumethat the Rn concentration in lake water is nearly in steady-state ∂ðCRn;lake VÞ permafrost is present on north-facing slopes and wind-exposed areas because of the ∂t ¼ 0 during the residence time of 222Rn in the water column. The 222Rn lack of an insulating snow cover during winter47 (Supplementary Fig. 6). The area is residence time (τ, [d]) in the lakes was estimated following the equation59: characterized by the transition between tundra and nordic mountain birch forest48–50 (Supplementary Fig. 7). The historical mean annual temperature for the study lakes 1 for the 1970–2000 period ranged between −0.1 to −1.4 °C51. During the past decade, τ¼ ð3Þ λ þ Qoutlet V þh k however, the temperature has risen above 0 °C52. The area exhibits a strong climatic gradient, with a general decrease in regional precipitation and annual temperature where h [m] and k [m d−1] are the mean depth of the lake and the gas transfer amplitudes eastwards. The highest precipitation (~1000 mm yr−1) occurs near the velocity for 222Rn (see below, section “Atmospheric fluxes”), respectively. The rest Norwegian border, while the lowest (~300 mm yr−1) is around Abisko. of the parameters are described in Eq. 2. Considering an average k and h of NATURE COMMUNICATIONS | (2022)13:3667 | https://doi.org/10.1038/s41467-022-31219-1 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31219-1 0.67 m d−1 and 2.8 m, respectively, and using the average Qoutlet/V measured in the Discharge estimates from inlet and outlet streams were measured using an study lakes (0.06 d−1), the average residence time of 222Rn for the study lakes is electrode magnetic flow meter (model 801 EC Meter; Valeport, Devon, U.K.) (in ∼2–3 days. Steady-state conditions over three days are thus a reasonable 2018) and based on salt slug injections64 (in 2019). Wind-speed, rainfall, air assumption considering the relatively stable environmental conditions in the days temperature, and air pressure data were acquired from weather stations before the sampling (e.g., no precipitation events, minor changes in wind regimes permanently installed at the shore close to each lake. and temperatures). The mass balance in Eq. 2 was used to estimate the flux of 222Rn supplied by Sediments. In July 2019, lake sediment cores were collected from all the lakes groundwater (Fgw = QgwCRn,gw) into each lake under steady-state conditions. (except lake BD05) using a standard sediment corer made from PVC pipes. Three Uncertainties associated with Fgw were deterministically estimated by propagating sediment cores were sliced into 1 cm thick sections, weighted, and dried to calculate the uncertainties of the individual terms in Eq. 2 (see Supplementary Methods). porosity and dry bulk density. The remaining sediment cores were reserved for The estimated 222Rn flux (Fgw) and its uncertainty were then used for laboratory sediment incubation experiments65,66. The incubation experiments were quantifying groundwater inflows and associated CH4 inputs to each lake based on a used to constrain the diffusive 222Rn inputs from underlying sediments (fdiff) for Monte Carlo analysis. The analysis consisted of generating 1000 values of Fgw for the 222Rn mass balance and to obtain an independent estimate of the 222Rn con- each lake (following a normal distribution based on calculated Fgw and its centration in the groundwater end-member (CRn,gw) (used only for comparison uncertainty). Each generated Fgw was then divided by a 222Rn concentration in with direct measurements of groundwater samples; Supplementary Fig. 10). groundwater (CRn,gw) to derive a groundwater flow, and by a 222Rn to CH4 concentration ratio in groundwater (i.e., CRn,gw/CCH4,gw), for calculating CH4 Lake and catchment characteristics. Echo sounding was done at transects 10–20 m inputs. Both CRn,gw and CRn,gw/CCH4,gw were randomly selected from all apart for bathymetric analysis. Lake average depth and surface area were calculated groundwater samples collected (n = 41), producing a 1000-length list of using ReefMaster v2.0 and the add-on volumes and areas67. Catchment delinea- groundwater flows and groundwater CH4 fluxes for each lake. Final fluxes are tions were made from a 2-m digital evaluation model68 using Whitebox GAT69, reported as the range of the median values for all the lakes. Groundwater inflows allowing to burn channels through road culverts70. Catchment slope was calculated and associated CH4 inputs are reported in the manuscript as the median and the using the “slope” function in the “Spatial Analyst” toolset in Arcmap 10.8 (ESRI). 25th and 75th percentiles of the 1000 simulations for each lake and season. Catchment forest and mire cover were calculated by overlying vegetation maps68 to the catchment areas. Catchments for each lake were delimited using a flow Sampling and analyses. The first survey in 2018 included three lakes (BD04, BD13, direction and flow accumulation model for the landscape, derived from the BD15) sampled in June (21–25) and September (8–12). In July (27–28) 2018, a short national digital elevation model (DEM) with a horizontal resolution of 2 m sampling campaign was conducted to collect groundwater samples from the active (Lantmäteriet; https://www.lantmateriet.se/). This analysis was performed using the layer. For the second survey in 2019, nine lakes (BD03, BD04, BD05, BD06, BD09, hydrological toolbox from ArcMap 10.8 (ESRI 2019 Redlands, CA: Environmental BD11, BD12, BD13, BD15) were sampled in July (9–13) and September (13–20). Systems Research Institute). Once catchments were delineated, several catchment Dates were selected to capture summer high flow and autumn base flow conditions. properties were extracted from the DEM (elevation range, average catchment slope, average aspect). Another landscape property used was modeled soil moisture (Soil Surface water. Surface lake water samples were collected from the shore and open- moisture map, Dept. of Forest Ecology and Management, Swedish University of water areas of the lakes using a submersible pump (Supplementary Fig. 8). A deep Agricultural Sciences). This machine learning product represents soil wetness in a (4 m depth) water sample was collected from the deepest lake point. Water samples scale from 0 to 10071, and here we used it as a proxy of surface hydrological were also collected from the main inlet and outlet streams. connectivity, calculating the average value for each catchment. We characterized For 222Rn analyses, water samples were filled into 1.5 L polyethylene the fraction of the catchment above 70% wetness to quantify the catchment terephthalate (PET) bottles, minimizing water-air contact to prevent 222Rn hydrological connectivity, which captures water-saturated zone areas (Supple- degassing. Shortly after collection, 222Rn activities were determined using a mentary Fig. 11). Maps were produced using QGIS (QGIS.org, 2021. QGIS Geo- Durridge RAD7 electronic radon-in-air monitor coupled to the RAD7 Soda bottle graphic Information System. QGIS Association. http://www.qgis.org), using as aerator kit accessory. 222Rn measurements were decay-corrected and converted to layers the global permafrost map72, Swedish land cover, and national orthophotos water concentrations using the air-water partitioning of 222Rn corrected for water (Lantmateriet; https://www.lantmateriet.se/). salinity and temperature60. The concentration of 222Rn supported by 226Ra decay in the lake water column Atmospheric fluxes. The flux of 222Rn and CH4 to the atmosphere were calculated was determined by measurements of the 226Ra concentration in lake water from as: five of the study lakes. Large volumes (30–50 L) of lake water were collected using a   submergible pump and filtered slowly (<1.0 L min−1) through a column loosely f atm ¼ kgas Cgas;lake  Cgas;air ð4Þ filled with MnO2-impregnated acrylic fiber (ca. 25 g dry) to quantitatively extract Ra isotopes61,62. Fibers were rinsed with Milly-Q water, incinerated (820 °C, 16 h), where kgas [cm d−1] is the gas transfer velocity for the corresponding gas at the ground, and transferred to hermetically sealed counting vials. Samples were measured temperature, Cgas,lake and Cgas,air [Bq m−3] are the gas concentration analyzed using a well-type Ge detector (Canberra model GSW120) after storing the measured in the lake and the concentration expected when the lake is in equili- samples for a minimum of three weeks to ensure the radioactive equilibrium brium with the atmosphere, respectively. between 226Ra and its daughters. For k, we used a wind-based model developed by Klaus and Vachon73 based on Dissolved CH4 concentrations were determined by analyzing the headspace of empirical k estimates from 46 globally distributed lakes data (see Supplementary gas-tight vials (22 mL vials, PerkinElmer Inc., U.S.) after addition of 20 μL of 4% Methods). This model fitted the study lakes as the range of conditions in terms of HCl to 4 mL sampled water, using a gas chromatograph (Clarus 500, PerkinElmer wind-speed (from 0 to 16 m s−1) and lake surface area (from 0.018 to 0.11 km2) Inc., USA). A gas mixture with known concentrations of CH4 (10 and 500 ppm) cover a substantial range of the calibration dataset (wind-speed from 0 to 13 m s−1; was prepared, stored, and analyzed as standards together with each batch of lake surface area from 0.018 to 1342 km2). To weigh the different influences on samples. Triplicate analyses of the standards were within 2% coefficient of 222Rn budgets of degassing events depending on their proximity to the sampling variation. In a few samples (n = 4), CH4 concentrations were below atmospheric time, we used a weighting factor to the hourly wind-speed data59. To evaluate the saturation and outside the detection limits of the instrument. Those values were uncertainties associated to k parametrization, two other empirical equations for k assumed to be in equilibrium with the atmosphere. estimates commonly used in lakes were used74,75. Groundwater. Groundwater (20–40 cm deep) samples (n = 41) were collected from Compilation of existing data of CH4 fluxes from Arctic lakes. We compiled data on mire areas right at the lake shoreline (Supplementary Fig. 9) using a direct-push CH4 fluxes (groundwater inputs, diffusion, ebullition, sediment production, and well-point piezometer coupled to a gas-tight syringe and tubing, minimizing the oxidation) from several studies across the Arctic to evaluate the importance of water-air contact57. For 222Rn analysis, 10 mL of filtered (0.45 μm) groundwater groundwater discharge in lake CH4 cycling. This compilation includes warm- were collected and directly transferred to 20 mL polyethylene vials prefilled with a season (May–October depending on the location) diffusive and ebullition fluxes 10 mL high-efficiency liquid scintillator cocktail63. Concentrations of 222Rn were extracted from the Boreal-Arctic Wetland and Lake Methane Dataset (BAWLD- analyzed using an ultra-low-level liquid scintillation counter (Quantulus 1220) with CH4)20 that is available at the Arctic Data Center (https://doi.org/10.18739/ alpha-beta discrimination counting (background of 0.02–0.07 cpm; efficiency of A2C824F9X). Sediment production21–23 and oxidation9,21,23 rates were obtained 1.5–3.0, depending on the quenching factor of the sample). Samples for dissolved from incubation experiments found in the literature and include results from 46 CH4 were collected simultaneously following the sampling procedure lakes across continuous, discontinuous, sporadic, and non-permafrost regions. The described above. compiled values and the original sources and further details are found in Sup- plementary Data 1 and Data 2. Physicochemical parameters. Water temperature, dissolved oxygen (DO), and specific conductivity were measured in situ in lake and stream waters using a Statistics and error estimates. Lake 222Rn and CH4 concentrations were reported as calibrated handheld water monitor (Yellow Springs Instrument ProSolo). A depth the mean of collected samples (±1 standard deviation). Differences in 222Rn and profile of temperature and DO was measured every 0.5 m from the surface at the CH4 concentrations between water sources (groundwater, streams, and lake waters) deepest point of each lake. Water pH and conductivity were measured using pH and between seasons (summer and autumn) were tested by analysis of variance and conductivity electrodes in the laboratory. (ANOVA), followed by Tukey-Kramer HSD post hoc test to identify differences 6 NATURE COMMUNICATIONS | (2022)13:3667 | https://doi.org/10.1038/s41467-022-31219-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31219-1 ARTICLE between groups. All statistical tests were considered statistically significant at 15. Striegl, R. & Michmerhuizen, C. Hydrologic influence on methane and carbon p < 0.05. We followed a two-step procedure to analyze spatial patterns in dioxide dynamics at two north-central Minnesota lakes. Limnol. Oceanogr. 43, groundwater inflows. First, partial least squares regression (PLS) was conducted to 1519–1529 (1998). detect correlation structures in the dataset and to rank the relative importance of 16. O’Connor, M. T., Bayani Cardenas, M., Neilson, B. T., Nicholaides, K. D. & catchment and lake characteristics on groundwater discharge rates. PLS is espe- Kling, G. W. Active layer groundwater flow: The interrelated effects of cially suitable for correlated covarying predictor variables and when there are more stratigraphy, thaw, and topography. Water Resour. Res. 55, 6555–6576 (2019). predictor variables than observations76,77. Catchment (area, slope, percentage of 17. Saarnio, S., Alm, J. & Silvola, J. Seasonal variation in CH4 emissions and wetzone and peatland cover) and lake (area, depth) characteristics, and pre- production and oxidation potentials at microsites on an oligrotrophic pine cipitation were used as predictor variables. Cross-validation was used for selecting fen. Oecologia 110, 414–422 (1997). the optimal number of components that minimize the prediction errors (RMSE). 18. Lupascu, M., Wadham, J. L., Hornibrook, E. R. C. & Pancost, R. D. The variable importance in projection (VIP) coefficients was calculated to classify Temperature sensitivity of methane production in the permafrost active layer predictors according to their explanatory power of the dependent variable. Then, at Stordalen, Sweden: a comparison with non-permafrost Northern wetlands. multiple stepwise linear regression analysis was used to identify relationships Arct., Antarct. Alp. Res. 44, 469–482 (2012). among the most relevant predictor variables and groundwater inflows. Models were 19. Jansen, J. et al. Climate-sensitive controls on large spring emissions of CH4 selected by considering all subsets on adjusted R2 and goodness-of-fit using the and CO2 from northern lakes. J. Geophys. Res. G Biogeosci.124, 2379–2399 Akaike Information Criteria (AIC), such that the most parsimonious model yielded (2019). the lowest AIC value. AIC measures both the model fit and complexity and is used 20. Kuhn, M. et al. BAWLD-CH4: a comprehensive dataset of methane fluxes in model selection to reduce over-fitting. Predictor variables were evaluated for from boreal and arctic ecosystems. Earth Syst. Sci. Data 13, 5151–5189 (2021). multicollinearity using Spearman correlation. Correlations between predictor 21. Hershey, A. E., Northington, R. M., Hart-Smith, J., Bostick, M. & Whalen, variables with a p-value less than 0.05 were considered multicollinear and removed S. C. Methane efflux and oxidation, and use of methane-derived carbon by from the models. Response and predictor variables were visually inspected for normality and log-transformed to improve normality. All data processing, statis- larval Chironomini, in arctic lake sediments. Limnol. Oceanogr. 60, 276–285 tics, and figures were done in R (version 4.1.0)78, using the PLS79, and ggplot280 (2015). packages. 22. Bretz, K. A. & Whalen, S. C. Methane cycling dynamics in sediments of Alaskan Arctic Foothill lakes. Inl. Waters 4, 65–78 (2014). 23. Lofton, D. D., Whalen, S. C. & Hershey, A. E. Effect of temperature on methane dynamics and evaluation of methane oxidation kinetics in shallow Data availability Arctic Alaskan lakes. Hydrobiologia 721, 209–222 (2014). Source data can be found in the supplementary materials of this paper. 24. Gentzel, T., Hershey, A. E., Rublee, P. A. & Whalen, S. C. Net sediment production of methane, distribution of methanogens and methane-oxidizing bacteria, and utilization of methane-derived carbon in an arctic lake. Inl. Received: 16 July 2021; Accepted: 6 June 2022; Waters 2, 77–88 (2012). 25. Cunada, C. L., Lesack, F. W. & Tank, S. E. Methane emission dynamics among CO2-absorbing and thermokarst lakes of a great Arctic delta. Biogeochemistry 156, 375–399 (2021). 26. Lecher, A. L. Groundwater discharge in the Arctic: a review of studies and implications for biogeochemistry. https://doi.org/10.3390/hydrology4030041 References (2017). 1. Dlugokencky, E. J., Nisbet, E. G., Fisher, R. & Lowry, D. Global atmospheric 27. Woo, M.-K. Permafrost Hydrology (Springer-Verlag, 2012). methane: budget, changes and dangers. Philos. Trans. R. Soc. A Math. Phys. 28. Drexler, J. Z., Bedford, B. L., Scognamiglio, R. & Siegel, D. I. Fine-scale Eng. Sci. 369, 2058–2072 (2011). characteristics of groundwater flow in a peatland. Hydrol. Process. 13, 2. IPCC. Climate change 2013: the physical science basis. Contribution of 1341–1359 (1999). Working Group I to the Fifth Assessment Report of the Intergovernmental 29. Cheng, G. & Jin, H. Permafrost and groundwater on the Qinghai-Tibet Panel on Climate Change. Intergov. Panel Clim. Chang. Work. Gr. I Contrib. to Plateau and in northeast China. Hydrogeol. J. 21, 5–23 (2013). IPCC Fifth Assess. Rep. (AR5) (Cambridge Univ Press, New York, 2013). 30. Cochand, M., Molson, J. & Lemieux, J. M. Groundwater hydrogeochemistry in 3. Wik, M., Varner, R. K., Anthony, K. W., MacIntyre, S. & Bastviken, D. permafrost regions. Permafr. Periglac. Process. 30, 90–103 (2018). Climate-sensitive northern lakes and ponds are critical components of 31. Yamazaki, Y., Kubota, J., Ohata, T., Vunglinsky, V. & Mizuyama, T. Seasonal methane release. Nat. Geosci. 9, 99–105 (2016). changes in runoff characteristics on a permafrost watershed in the southern 4. Tan, Z. & Zhuang, Q. Arctic lakes are continuous methane sources to the mountanous region of eastern Siberia. Hydrol. Process. 20, 453–467 (2006). atmosphere under warming conditions. Environ. Res. Lett. 10, 054016 (2015). 32. Peterson, B. J. et al. Increasing river discharge to the Arctic. Ocean. Sci. (80-.). 5. Tan, Z. & Zhuang, Q. Methane emissions from pan-Arctic lakes during the 298, 2171–2173 (2002). 21st century: an analysis with process-based models of lake evolution and 33. Déry, S. J., Stieglitz, M., McKenna, E. C. & Wood, E. F. Characteristics and biogeochemistry. J. Geophys. Res. Biogeoscience 120, 2641–2653 (2015). trends of river discharge into Hudson, James, and Ingava Bays. J. Clim. 18, 6. Bastviken, D., Cole, J., Pace, M. & Tranvik, L. Methane emissions from lakes: 2540–2557 (2005). dependence of lake characteristics, two regional assessments, and a global 34. Lemieux, J. M. et al. Groundwater dynamics within a watershed in the estimate. Glob. Biogeochem. Cycles 18, GB4009 (2004). discontinuous permafrost zone near Umiujaq (Nunavik, Canada). Hydrogeol. 7. DelSontro, T., del Giorgio, P. A. & Prairie, Y. T. No longer a paradox: the J. 28, 833–851 (2020). interaction between physical transport and biological processes explains the 35. Cochand, M. et al. Rapido groundwater recharge dynamics determined from spatial distribution of surface water methane within and across lakes. hydrogeochemical and isotope data in a small permafrost watershed near Ecosystems 21, 1073–1087 (2018). Umiujaq (Nunavik, Canada). Hydrogeol. J. 28, 853–868 (2020). 8. Bastviken, D. in Encyclopedia of Inland Waters. (ed. Likens, G. E.) 783–805 36. Giesler, R. et al. Catchment-scale dissolved carbon concentrations and export (Academic Press, 2009). estimates across six subarctic streams in northern Sweden. Biogeoscience 11, 9. Martinez-Cruz, K., Sepulveda-Jauregui, A., Water Anthony, K. & Thalasso, F. 525–537 (2014). Geographic and seasonal variation of dissolved methane and aerobic methane 37. Crill, P. M. et al. Methane flux from Minnesota peatlands. Global Biogeochem. oxidation in Alaskan lakes. Biogeoscience 12, 4595–4606 (2015). Cycles 2, 371–384 (1988). 10. Thottathil, S. D., Reis, P. C. J., del Giorgio, P. A. & Prairie, Y. T. The extent 38. Dunfield, P., Knowles, R., Dumont, R. & Moore, T. R. Methane production and regulation of summer methane oxidation in northern lakes. J. Geophys. and comsumption in temperate and subarctic peat soils: Response to Res. Biogeosciences 123, 3216–3230 (2018). temperature and pH. Soil Biochem. 25, 321–326 (1993). 11. Connolly, C. T., Bayani Cardenas, M., Burkart, G. A., Spencer, R. G. M. & 39. Bastviken, D., Cole, J. J., Pace, M. L. & Van de Bogert, M. C. Fates of methane McClelland, J. W. Groundwater as a major source of dissolved organic matter from different lake habitats: Connecting whole-lake budgets and CH 4 to Arctic coastal waters. Nat. Commun. 11, 1479 (2020). emissions. J. Geophys. Res. Biogeosciences 113, https://doi.org/10.1029/ 12. Dabrowski, J. S. et al. Using radon to quantify groundwater discharge and 2007JG000608 (2008). methane fluxes to a shallow, tundra lake on the Yukon-Kuskokwim Delta, 40. Romanovsky, V. E., Smith, S. L. & Christiansen, H. H. Permafrost thermal Alaska. Biogeochemistry 148, 69–89 (2020). state in the polar northern hemisphere during the international polar year 13. Paytan, A. et al. Methane transport from the active layer to lakes in the Arctic 2007-2009: a synthesis. Permafr. Periglac. Process. 21, 106–116 (2010). using Toolik Lake, Alaska, as a case study. Proc. Natl Acad. Sci. USA 112, 41. Grosse, G. et al. Vulnerability and feedbacks of permafrost to climate change. 201417392 (2015). Eos Trans. Am. Geophys. Union 92, 73–74 (2011). 14. Kling, G. W., Kipphut, G. W. & Miller, M. C. The flux of CO2 and CH4 from 42. Christensen, T. R. Thawing sub-arctic permafrost: effects on vegetation and lakes and rivers in arctic Alaska. Hydrobiologia 240, 23–36 (1992). methane emissions. Geophys. Res. Lett. 31, L04501 (2004). NATURE COMMUNICATIONS | (2022)13:3667 | https://doi.org/10.1038/s41467-022-31219-1 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31219-1 43. Duc, N. T., Crill, P. & Bastviken, D. Implications of temperature and sediment resolution national-scale soil moisture mapping of the Swedish forest characteristics on methane formation and oxidation in lake sediments. landscape. Geoderma 404, 115280 (2021). Biogeochemistry 100, 185–196 (2010). 72. Obu, J. et al. Northern Hemisphere permafrost map based on TTOP 44. Bintanja, R. et al. Strong future increases in Arctic precipitation variability modelling for 2000-2016 at 1 km2 scale. Earth-Sci. Rev. 193, 299–316 (2019). linked to poleward moisture transport. Sci. Adv. 6, eaax6869 (2020). 73. Klaus, M. & Vachon, D. Challenges of predicting transfer velocity from wind 45. Young, N. L., Lemieux, J.-M., Delottier, H., Fortier, R. & Fortier, P. A measurements over global lakes. Aquat. Sci. 82, 53 (2020). conceptual model for anticipating the impact of landscape evolution on 74. Cole, J. J. & Caraco, N. F. Atmospheric exchange of carbon dioxide in a low- groundwater recharge in degrading permafrost environments. Geophys. Res. wind oligotrophic lake measured by the addition of SF6. Limnol. Oceanogr. 43, Lett. 47, e2020GL087695 (2020). 647–656 (1998). 46. Brown, J., Ferrians, Jr., O. J., Heginbottom, J. A. & Melnikov, E. S. Circum- 75. Vachon, D., Prairie, Y. T. & Smith, R. The ecosystem size and shape arctic of permafrost and ground ice conditions. https://doi.org/10.3133/cp45 dependence of gas transfer velocity versus wind speed relationships in lakes. (1997). Can. J. Fish. Aquat. Sci. 70, 1757–1764 (2013). 47. Johansson, M., Christensen, T. R., Akerman, H. J. & Callaghan, T. V. What 76. Carrascal, L. M., Galvan, I. & Gordo, O. Partial least squares regression as an determines the current presence or absense of permafrost in the Torneträsk alternative to current regression methods used in ecology. Oikos 118, 681–690 region, a sub-arctic landscape in Northern Sweden? Ambio 35, 1–9 (2006). (2009). 48. Berggren, M. et al. Systematic microbial production of optically active 77. Sobek, S., Algesten, G., Bergström, A. K., Jansson, M. & Tranvik, L. J. The dissolved organic matter in subarctic lake water. Limnol. Oceanogr. 65, catchment and climate regulation of pCO2 in boreal lakes. Glob. Chang. Biol. 951–961 (2020). 9, 630–641 (2003). 49. Virtanen, R. et al. Where do the treeless tundra areas of northern highlands fit 78. R Core Team. R: A Language and Environment for Statistical Computing (R in the global biome system: toward an ecologically natural subdivision of the Core Team, 2019). tundra biome. Ecol. Evol. 6, 143–158 (2016). 79. Mevik, B. J. & Wehrens, R. The pls package: principal component and partial 50. Wielgolaski, F. E (ed.). in Plant Ecology, Herbivory, and Human Impact in least squares regression. R. J. Stat. Softw. https://doi.org/10.18637/jss.v018.i02 Nordic Mountain Birch Forests . 3–18 (Plant Ecology, Herbivory, and Human 1–24 (2007). Impact in Nordic Mountain Birch Forests, 2005). 80. Wickham, H., Chang, W. & RStudio. ggplot2: Create Elegant Data 51. Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G. & Jarvis, A. Very high Visualisations Using the Grammar of Graphics (Version 2.2.1.) (2016). resolution interpolated climate surfaces for global land areas. Int. J. Climatol. 25, 1965–1978 (2005). 52. Kuhn, M., Lundin, E. J., Giesler, R., Johansson, M. & Karlsson, J. Emissions Acknowledgements from thaw ponds largely offset the carbon sink of northern permafrost We thank Karl Heuchel, Joan Manuel Bruach, Julia Papenhausen, Maike Strack, Alberto wetlands. https://doi.org/10.1038/s41598-018-27770-x (2018). Zannella, and Mathilde Schnuriger for invaluable assistance in the field. Sven Norman and 53. Rocher-Ros, G. et al. Large lakes dominate CO2 evasion from lakes in an Marcus Klaus for bathymetry maps and geographical information. Danny CP Lau and Isolde Arctic catchment. Geophys. Res. Lett. 44, https://doi.org/10.1002/ Puts for their help with the PLS analyses, and the staff at the Abisko Scientific Research 2017GL076146 (2017). Station (ANS) for infrastructure, logistic and technical support. This study has been made 54. Burnett, W. C. Quantifying submarine groundwater discharge in the coastal possible by data provided by Abisko Scientific Research Station (ANS) and the Swedish zone via multiple methods. Sci. Total Environ. 367, 498–543 (2006). Infrastructure for Ecosystem Science (SITES). This study was financially supported by 55. Dimova, N. T., Burnett, W. C., Chanton, J. P. & Corbett, J. E. Application of FORMAS (Grant no. 2018-01217) with a grant awarded to C.O. and by the Swedish Research radon-222 to investigate groundwater discharge into small shallow lakes. J. Council (2016-05275) and Knut and Alice Wallenberg Foundation (2016.0083) with grants Hydrol. 486, 112–122 (2013). awarded to J.K. D.B. was supported by the European Research Council (ERC; grant 725546), 56. Dimova, N. & Burnett, W. Evaluation of groundwater discharge into small the Swedish Research Council (grant 2016-04829), and FORMAS (grant 2018-01794). J.G.O. lakes based on the temporal distribution of radon-222. Limnol. Oceanogr. 56, acknowledges the financial support of the Spanish Ministry of Science, Innovation and 486–494 (2011). Universities, through the “Maria de Maeztu” program for Units of Excellence (CEX2019- 57. Rodellas, V. et al. Groundwater-driven nutrient inputs to coastal lagoons: The 000940-M) and the Generalitat de Catalunya (MERS; 2017 SGR–1588). V.R. acknowledges relevance of lagoon water recirculation as a conveyor of dissolved nutrients. financial support from the Beatriu de Pinós postdoctoral program of the Catalan Government Sci. Total Environ. 642, 764–780 (2018). (Generalitat de Catalunya) (2019‐BP‐00241). M.D.F. acknowledges the economic support 58. Sadat-Noori, M. et al. Intermittently closed and open lakes and/or lagoons from the FI‐ 2017 fellowships of the Generalitat de Catalunya autonomous government (ICOLLs) as groundwater-dominated coastal systems: evidence from seasonal (2017FI_B_00365). A.A.K. acknowledges financial support from ICTA “Unit of Excellence” radon observations. J. Hydrol. 535, 612–624 (2016). (MinECo, MDM2015‐0552‐17‐1) and PhD fellowship, BES‐2017‐080740. 59. Rodellas, V. et al. Conceptual uncertainties in groundwater and pore-water fluxes estimated by radon and radium mass balances. Limnol. Oceanogr. 1–19 Author contributions https://doi.org/10.1002/lno.11678 (2021). C.O. and J.K. designed the study. C.O., V.R., J.G.-O., M.D.-F., A.A.-K. performed 60. Schubert, M., Paschke, A., Lieberman, E. & Burnett, W. C. Air-water fieldwork and lab analysis with infrastructure support from C.O., J.K., and J.G.O. C.O. partitioning of 222Rn and its dependence on water temperature and salinity. performed data analyses, and G.R.-R. helped generate figures. C.O., V.R., and M.D.-F. Environ. Sci. Technol. 46, 3905–3911 (2012). performed uncertainty analysis. J.K. and D.B. provided input on CH4 fluxes inter- 61. Moore, W. S. Sampling 226Ra in the deep ocean. Deep. Res. 23, 647–651 pretation. C.O. interpreted results and drafted the manuscript with main inputs from (1976). G.R.-R., V.R., and J.G.-O. C.O. wrote the final version of the manuscript with con- 62. Moore, W. S. & Reid, D. F. Extraction of radium from natural-waters using tribution from all co-authors. manganese-impregnated acrylic fibers. J. Geophys. Res. 36, 8880–8886 (1973). 63. Cable, J. E. & Martin, J. B. In situ evaluation of nearshore marine and fresh pore water transport into Flamengo Bay, Brazil. Estuar. Coast. Shelf Sci. 76, Funding 473–483 (2008). Open access funding provided by Umea University. 64. Moore, R. D. Slug injection using salt in solution. Watershed Magage Bull. 8, 1–6 (2005). 65. Corbett, D. R., Burnett, W. C., Cable, P. H. & Clark, S. B. A multiple approach Competing interests The authors declare no competing interests. to the determination of radon fluxes from sediments. J. Radioanal. Nucl. Chem. 236, 247–253 (1998). 66. Chanyotha, S., Kranrod, C. & Burnett, W. C. Assessing diffusive fluxes and Additional information pore water radon activities via a single automated experiment. J. Radioanal. Supplementary information The online version contains supplementary material Nucl. Chem. 301, 581–588 (2014). available at https://doi.org/10.1038/s41467-022-31219-1. 67. Reffmaster Software Ltd. West Sussex, PO20 7ES, United Kingdom. http:// reffmaster.com.au/ (Reffmaster Software Ltd, 2017). Correspondence and requests for materials should be addressed to Carolina Olid. 68. Lantmäteriet. GSD-Vegetationsdata (GSD vegetationdata, 2012). 69. Lindsay, J. B. Whitebox GAT: a case study in geomorphometric analysis. Peer review information Nature Communications thanks the anonymous reviewer(s) for Comput. Geosci. 95, 75–84 (2016). their contribution to the peer review of this work. Peer reviewer reports are available. 70. Lidberg, W., Nilsson, M., Lundmark, T. & Ågren, A. M. Evaluating preprocessing methods of digital elevation models for hydrological modelling. Reprints and permission information is available at http://www.nature.com/reprints Hydrol. Process. 31, 4660–4668 (2017). 71. Ågren, A. M., Larson, J., Paul, S. S., Laudon, H. & Lidberg, W. Use of multiple Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in LIDAR-derived digital terrain indices and machine learning for high- published maps and institutional affiliations. 8 NATURE COMMUNICATIONS | (2022)13:3667 | https://doi.org/10.1038/s41467-022-31219-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-31219-1 ARTICLE Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/. © The Author(s) 2022 NATURE COMMUNICATIONS | (2022)13:3667 | https://doi.org/10.1038/s41467-022-31219-1 | www.nature.com/naturecommunications 9