Salinity causes widespread restriction of methane emissions from small inland waters

Methane (CH₄) is a potent greenhouse gas responsible for a substantial fraction of current atmospheric radiative forcing, and inland waters are the largest natural source. Global aquatic CH₄ emission estimates are largely based on measurements from solute-poor (freshwater) systems, even though salt-rich inland waters comprise a large share of global inland water volume and area. There is clear evidence that salinity—particularly sulfate (SO₄²⁻)—inhibits methanogenesis and can also promote anaerobic methane oxidation by coupling with alternative electron acceptors (e.g., sulfate, nitrate, iron), thereby potentially reducing net CH₄ emissions. Salinity also interacts with organic matter (OM) and nutrient availability through processes such as sorption and redox competition, integrating multiple pathways of CH₄ suppression. Despite mechanistic indications from experiments and coastal/salt-rich wetlands, the regional-to-continental impact of salinity on inland-water CH₄ emissions remains poorly quantified. To address this gap, the study surveys diverse aquatic ecosystems across the Canadian Prairies, augments observations with peripheral pond sampling and eddy covariance measurements, and evaluates the role of salinity—alone and interacting with OM—in regulating CH₄ dynamics and emissions, with implications for regional and global CH₄ budgets.

Prior work demonstrates that salinity shapes microbial community composition, including methanogens and methanotrophs, and that methanogenesis is the least energetically favorable pathway in anaerobic carbon mineralization. Elevated ionic strength favors alternative redox processes (e.g., sulfate and iron reduction), which can outcompete methanogens for labile carbon and enhance anaerobic methane oxidation coupled to sulfate, nitrate, or iron reduction. Experimental studies show suppression of net sediment CH₄ production with increased salinity (NaCl) and strong SO₄²⁻-linked inhibition of methanogenesis in salt-rich and coastal wetlands. Salinity often covaries with DOC and nutrients due to catchment sources and evapoconcentration, complicating causal inference. Nevertheless, numerous studies across wetlands and lakes (including on the Tibetan Plateau and coastal systems) report reduced CH₄ production/emissions with increasing salinity or sulfate, and altered nutrient availability under salinization. These findings suggest that multiple ions—not solely sulfate—can influence CH₄ dynamics across inland waters, but the extent to which these mechanisms scale to regional/global emissions remains uncertain.

Study region and design: The study was conducted across the Canadian Prairie ecozone (Alberta, Saskatchewan, Manitoba; 467,029 km²), characterized by semi-arid to continental climate, extensive agriculture, and numerous hardwater lentic systems. Primary survey: 193 sites sampled once in summer (2011–2021): 23 rivers, 17 lakes, 45 wetlands, and 108 agricultural ponds, spanning wide gradients in morphometry, hydrology, chemistry, and trophic status. Peripheral survey: 48 wetlands (16 per province) repeatedly sampled (≥3–5 times) during the May–September 2021 open-water season; two Manitoba wetlands instrumented with eddy covariance towers for continuous CH₄ fluxes over a full year. Environmental measurements: Near-surface water temperature, dissolved oxygen, pH, and specific conductivity measured with YSI or Hydrolab DS5 multiparameter probes (routinely calibrated). Salinity was measured directly or derived from specific conductivity and temperature (R oce::swSCTp). Water samples (<0.5 m depth) were collected and analyzed for DOC (filtered 0.45 μm; acidified; TOC analyzers), TP and TN (persulfate digestion; various analyzers). For a subset, dissolved nutrients (0.45 μm) were used and assumed representative given high dissolved fractions (mean 77% P, 89% N). Chlorophyll a measured by spectrophotometry after filtration and alcohol extraction. Sulfate (SO₄²⁻) measured at 118 sites by discrete analyzer or ion chromatography. CH₄ concentration (pCH₄): Determined via headspace equilibration (air-to-water ratio 0.05–0.25; ambient air or N₂ headspace; ≥2 min shaking), followed by GC analysis. In-situ water CH₄ concentrations were back-calculated using temperature- and salinity-dependent solubility, local pressure, and headspace ratio, and expressed as partial pressure pCH₄. Emissions calculations: Diffusive flux F estimated as F = k(C_eq − C_w), using gas transfer velocity k = 1.64 ± 1.14 m d⁻¹ from floating chamber measurements. Diffusive flux datasets compiled for 139 ponds and wetlands from published sources. Ebullition: At 10 sites (5 wetlands, 5 ponds), two funnel traps per site collected bubbles June–August; seasonally averaged ebullition from deep-zone traps was computed from bubble volume and CH₄ content (GC) to areal daily rates. Eddy covariance: Two flux towers (open-path LI-7700 CH₄ IRGA; LI-7500A CO₂/H₂O IRGA; RM Young 81000 sonic anemometer) operated May 20, 2021–May 19, 2022. High-frequency (20 Hz) data processed to 30-min averages (EddyPro v7.0.6), filtered by friction velocity thresholds (REddyProc), stationarity/turbulence flags, spike removal, instrument quality (signal strength >20%), and wind-direction filtering. Meteorological gaps filled with nearby stations; CH₄ gaps filled via random forest. Final datasets are filtered and gap-filled fluxes. Statistical analyses: All analyses in R. Variables log₁₀-transformed where needed. Multiple linear regressions (lm) related pCH₄ or flux (diffusive, ebullitive, total) to predictors including salinity, DOC, TP, temperature, and area (for small lentic). Residual diagnostics (Shapiro–Wilk) and standardized-coefficient comparisons performed. Marginal effects of salinity visualized with sjPlot. Model testing: Observed pCH₄ and fluxes compared to predictions from literature models developed largely for freshwater systems: pCH₄ model using area, temperature, TN; diffusive flux models using area and chlorophyll a or area and temperature; ebullition models using chlorophyll a (global) and sediment temperature (regional; sediment temperature approximated as surface temperature −1°C). Deviations quantified and regressed against salinity. Upscaling: For the Canadian Prairies, total CH₄ flux (diffusion + ebullition) was simulated using the empirical salinity–flux regression (from 10 small lentic sites) at two salinity levels (0.5 ppt vs 0.1 ppt), applied over a conservative 3-month summer window (91 days) and multiplied by mapped regional area of small open-water lentic systems <0.1 km² (2869 km²) derived from CWI and ABMI/DUC inventories. For global first-order scaling, estimated small salt-rich lentic area (166,120 km²; 12.8% of small lentic area) was combined with the two flux scenarios (1.57 vs 8.42 mmol m⁻² d⁻¹) over three months to illustrate potential overestimation if salinity is ignored. Temporal trends: Long-term SO₄²⁻ trends (1990–2020) in Saskatchewan lentic systems assessed with Sen slope analysis (R zyp) under inclusion criteria (≥10 observations, ≥5-year span, ≥1 post-2010 sample); significance determined by 95% CI of slopes.

- Salinity–pCH₄ relationships vary by ecosystem type: Multiple linear regressions showed salinity had no significant effect on pCH₄ in rivers and larger lakes (>0.1 km²), but was a key predictor in small lentic systems (wetlands and agricultural ponds, <0.1 km²). - Small lentic systems: In ponds, log₁₀ pCH₄ = 0.47 + 0.57 log₁₀ DOC − 0.60 log₁₀ Salinity + 0.28 log₁₀ TP + 0.01 Temperature (p<0.001; R²adj=0.27). In wetlands, log₁₀ pCH₄ = 3.5 − 0.50 log₁₀ DOC − 0.82 log₁₀ Salinity + 0.04 log₁₀ TP + 0.02 Temperature (p=0.001; R²adj=0.31). Combined small lentic: log₁₀ pCH₄ = 1.7 − 0.16 log₁₀ DOC − 0.56 log₁₀ Salinity + 0.33 log₁₀ TP + 0.03 Temperature − 0.06 Area (p<0.001; R²adj=0.31). - Interaction with organic matter: pCH₄ scaled positively with the DOC-to-salinity ratio in small lentic systems: log₁₀(pCH₄) = 2.03 (±0.22) + 0.63 (±0.12) log₁₀(DOC/Salinity), p<0.001, R²adj=0.15; validated by repeated peripheral sampling across provinces. - Nonlinear sensitivity: Modeled pCH₄ response to salinity is most sensitive at low salinity. Example at DOC=10 mg L⁻¹: increasing salinity from 0.5 to 1.0 ppt reduced pCH₄ by ~35% (710→458 ppm), whereas 3.5→4.0 ppt reduced pCH₄ by ~8% (207→190 ppm). - Emissions pathways: In small lentic systems, both diffusive and ebullitive CH₄ fluxes declined with salinity (log₁₀–log₁₀ regressions; Fig. 2): diffusion (n=139) p=0.001, R²adj=0.25, slope=-0.7, intercept=0.24; ebullition (n=10) p=0.02, R²adj=0.47, slope=-3.0, intercept=-1.9; total flux (n=10) p=0.002, R²adj=0.69, slope=-1.0, intercept=-0.12. A 10-fold salinity increase (0.1→1 ppt) led to ~4.5-fold (diffusion) vs ~1000-fold (ebullition) declines. - Eddy covariance confirmation: Two wetlands with salinities 0.5 vs 2.3 ppt showed ~2 orders of magnitude difference in integrated CH₄ fluxes despite both being hypereutrophic and DOC-rich. Mean (median) CH₄ emissions: 20.7 (8.7) mg C m⁻² d⁻¹ at low salinity vs 0.96 (0.07) mg C m⁻² d⁻¹ at higher salinity. - Cross-region support: Meta-analysis shows inverse salinity–CH₄ relationships in other hardwater landscapes (Tibetan lakes, Indian ponds/lakes, Mexican ponds), consistent with Prairie patterns. - Model evaluation: Freshwater-derived empirical models explained limited variance for Prairie small lentic systems (pCH₄ R²=0.32; diffusion R²=0.15; ebullition R²=0.03). Deviations from model predictions correlated with salinity: slopes of -0.80 (pCH₄), -0.60 (diffusion), and -3.0 (ebullition) vs log₁₀ salinity with p-values 6.6×10⁻⁸, 5.4×10⁻⁵, and 0.025, explaining 20%, 24%, and 42% of deviations, respectively. - Regional scaling: Applying observed salinity effects to small open-water lentic systems in the Canadian Prairies (<0.1 km²; 2869 km²) over 3 summer months yielded 6555 vs 35,169 Mg CH₄ for salt-rich (0.5 ppt) vs freshwater (0.1 ppt) scenarios—a >5-fold difference (~0.97 Tg CO₂-eq), comparable to major national emissions sources. - Global implication: Using a conservative 3-month window and assuming 12.8% of small lentic area is salt-rich (166,120 km²), freshwater-based scaling could overestimate CH₄ emissions by ~1.7 Tg CH₄ yr⁻¹ (~4% of recent global lentic estimates).

Findings show salinity is a dominant regulator of CH₄ dynamics in small lentic inland waters, with minimal direct effect in rivers and larger lakes where covarying DOC/nutrients and water-column processes may override salinity’s influence. In shallow, small systems, close coupling of surface pCH₄ to sediment methanogenesis, combined with salinity-driven competition for electron donors (favoring sulfate/iron reduction) and potential enhancement of anaerobic methane oxidation, substantially lowers CH₄ concentrations and emissions. The DOC/salinity ratio highlights how abundant organic substrates can partly counteract salinity inhibition, whereas low DOC relative to salinity strengthens suppression of methanogenesis and net CH₄ production. Ebullition is particularly sensitive to salinity, likely due to strong controls on sediment CH₄ production and oxidation prior to bubble release, while diffusion integrates additional water-column processes that dampen salinity’s effect. Eddy covariance data corroborate that salinity-associated reductions persist at ecosystem scales and across seasons, even in highly productive wetlands. The inverse salinity–emissions patterns observed in other hardwater regions suggest broader applicability, implying that traditional freshwater-based models overestimate CH₄ emissions where salinity is elevated. Incorporating salinity as a predictor significantly improves empirical model performance and can refine national and global CH₄ budgets. Given ongoing and projected salinization from agriculture, mining, urbanization, and climate-related hydrological change, salinity trends may partially offset expected CH₄ increases due to warming and eutrophication, underscoring the need to jointly consider multiple drivers in future projections and inventories.

This study demonstrates that salinity imposes widespread and strong constraints on methane emissions from small inland waters, especially via ebullition, across a major global hardwater region. Salinity’s effect interacts with organic matter availability, with the DOC/salinity ratio emerging as a robust indicator of pCH₄. Freshwater-derived empirical models substantially overpredict emissions from salt-rich ponds and wetlands; adding salinity markedly improves predictions. Regionally, ignoring salinity inflates Canadian Prairie small-lentic CH₄ budgets by amounts comparable to major national sources; globally, similar biases likely contribute to overestimation of lentic CH₄ emissions. Future research should expand high-resolution CH₄ and ion-speciation datasets across diverse hardwater landscapes and seasons; deploy ecosystem-scale flux measurements (eddy covariance) across salinity and productivity gradients; disentangle mechanistic pathways (e.g., sulfate vs multi-ion effects, anaerobic oxidation) with targeted experiments; integrate salinity and OM interactions into process-based and empirical models; and improve spatial inventories of small waterbodies for robust upscaling.

- Primary survey sites were sampled once per summer, potentially underrepresenting intra-seasonal variability, though peripheral surveys and flux towers provide temporal context. - Emissions upscaling was restricted to small open-water lentic areas (<0.1 km²) and a conservative 3-month summer window, excluding emergent vegetation zones; thus annual and whole-ecosystem fluxes may differ. - Ebullition estimates were based on 10 sites and deep-zone traps; shallow-zone data were often lost due to trap instability, introducing uncertainty in within-site spatial variability. - CH₄ diffusive fluxes relied on a single k value (1.64 ± 1.14 m d⁻¹) from floating chamber measurements; site-specific k variability may affect absolute flux estimates. - Headspace methods and instrumentation varied among teams; although differences are minor relative to large gradients, methodological heterogeneity introduces measurement uncertainty. - Global extrapolations are first-order due to limited salt-rich inland water data; ion composition and co-varying factors (DOC, nutrients) differ among regions, constraining generalizability. - Model testing used freshwater-based empirical models not designed for salt-rich systems; comparisons illustrate bias but do not constitute optimized predictive frameworks for hardwater ecosystems.