Salinity causes widespread restriction of methane emissions from small inland waters

Abstract

Inland waters are one of the largest natural sources of methane (CH₄), a potent greenhouse gas, but emissions models and estimates were developed for solute-poor ecosystems and may not apply to salt-rich inland waters. Here we combine field surveys and eddy covariance measurements to show that salinity constrains microbial CH₄ cycling through complex mechanisms, restricting aquatic emissions from one of the largest global hardwater regions (the Canadian Prairies). Existing models overestimated CH₄ emissions from ponds and wetlands by up to several orders of magnitude, with discrepancies linked to salinity. While not significant for rivers and larger lakes, salinity interacted with organic matter availability to shape CH₄ patterns in small lentic habitats. We estimate that excluding salinity leads to overestimation of emissions from small Canadian Prairie waterbodies by at least 81% ( ~ 1 Tg yr^-1 CO₂ equivalent), a quantity comparable to other major national emissions sources. Our findings are consistent with patterns in other hardwater landscapes, likely leading to an overestimation of global lentic CH₄ emissions. Widespread salinization of inland waters may impact CH₄ cycling and should be considered in future projections of aquatic emissions.

Fig. 1. The relationship between pCH₄, salinity, and DOC across ecosystem types.

Marginal effect of salinity on pCH₄ derived from multiple linear regressions (Table 1, Fig. S2), when holding other factors constant at their average value in different system types (a). Regressions were non-significant for lakes and rivers (shown as dashed lines). Shaded areas represent the 95% confidence interval around the slope. b For small lentic systems (ponds and wetlands), pCH₄ (ppm) scaled positively with the DOC (mg L⁻¹) to salinity (ppt) ratio (p-value « 0.001, R²_adj = 0.15; linear regression equation with standard errors: log₁₀ (pCH₄) = 2.03 (±0.22) + 0.63 (±0.12) log₁₀ (DOC/Salinity)). c For small lentic ecosystems, modeled pCH₄ is shown as a function of salinity for varying levels of DOC, based on the empirical equation model from panel b.

Fig. 2. Negative relationship between salinity and aquatic CH₄ emissions in different global salt-rich landscapes.

Significant linear regressions (on log₁₀ scales) exist for diffusion, ebullition, and total flux (ebullitive plus diffusive) from small lentic Canadian Prairie ecosystems (wetlands and ponds; n = 139, 10, and 10; p = «0.001, 0.02, and 0.002; R²_adj = 0.25, 0.47, and 0.69; slope = −0.7, −3.0, and −1.0; intercept = 0.24, −1.9, and −0.12, respectively), for diffusive fluxes from lakes of the Tibetan Plateau (n = 18; p = 0.04; R²_adj = 0.19; slope = −0.8; intercept = −0.01), and for diffusive flux from southern Indian ponds and lakes (n = 14; p = 0.06, R²_adj = 0.2; slope = −0.58; intercept = 0.48). No statistics are presented for diffusive fluxes from the Mexican ponds of the Cuarto Cienegas Basin (n = 5) since one of the values was zero and could not be log-transformed (dashed line). Median emissions values from two Canadian Prairie wetland sites with eddy co-variance towers are also shown.

Fig. 3. Testing existing empirical models in hardwater Prairie systems.

Observed versus predicted (based on published models^,,) values of CH₄ partial pressure (a), diffusion (b), and mean summer ebullition rates (c) in small lentic sampling sites. The corresponding deviations from predictions (measured - modeled) are shown as a function of salinity (d–f). The gray dashed line represents a perfect (1:1) correspondence between predicted and observed values. Linear regression lines (solid black) have p-values = 6.6 × 10⁻⁹, 5.4 × 10⁻¹⁰, and 0.025, R²_adj = 0.20, 0.24, and 0.42, and slopes = −0.80, −0.60, and −3.0 for panels d–f, respectively.