Effects of cooling and internal wave motions on gas transfer coefficients in a boreal lake

Abstract

A B S T R A C TLakes and other inland waters contribute significantly to regional and global carbon budgets. Emissions from lakes are often computed as the product of a gas transfer coefficient, k 600 , and the difference in concentration across the diffusive boundary layer at the airÁwater interface. Eddy covariance (EC) techniques are increasingly being used in lacustrine gas flux studies and tend to report higher values for derived k 600 than other approaches. Using results from an EC study of a small, boreal lake, we modelled k 600 using a boundarylayer approach that included wind shear and cooling. During stratification, fluxes estimated by EC occasionally were higher than those obtained by our models. The high fluxes co-occurred with winds strong enough to induce deflections of the thermocline. We attribute the higher measured fluxes to upwelling-induced spatial variability in surface concentrations of CO 2 within the EC footprint. We modelled the increased gas concentrations due to the upwelling and corrected our k 600 values using these higher CO 2 concentrations. This approach led to greater congruence between measured and modelled k values during the stratified period. k 600 has a well-resolved and Âcubic relationship with wind speed when the water column is unstratified and the dissolved gases well mixed. During stratification and using the corrected k 600 , the same pattern is evident at higher winds, but k 600 has a median value of Â7 cm h(1 when winds are less than 6 m s (1 , similar to observations in recent oceanographic studies. Our models for k 600 provide estimates of gas evasion at least 200% higher than earlier wind-based models. Our improved k 600 estimates emphasize the need for integrating within lake physics into models of greenhouse gas evasion.

Consequences of gas flux model choice on the interpretation of metabolic balance across 15 lakes

“…The eddy covariance studies cited here have argued that CC98 significantly underestimates gas flux (MacIntyre et al 2010a, Heiskanen et al 2014, and that k models that incorporate a buoyancy flux term better represent realistic gas flux (MacIntyre et al 2010b, Tedford et al 2014. Uncertainties in NEP have consequences for the interpretation of carbon budgets in lakes, and choice of k model may impact those interpretations.…”

“…For instance, in low-wind environments, such as the bog lakes, gas exchange can be driven largely by diel heating and cooling (Read et al 2012, Heiskanen et al 2014, Podgrajsek et al 2015. R12S, T14, and H14, which incorporate a convective forcing component, are potentially able to reproduce the diel variability in k caused by buoyant mixing (Fig.…”

“…For 5 of the published models (abbreviated as CC98, CW03, VP13, T14, and H14), we adhered to the published methods and parameters in our analysis (Cole and Caraco 1998, Crusius and Wanninkhof 2003, Vachon and Prairie 2013, Heiskanen et al 2014, Tedford et al 2014. Because CW03 does not describe any single model, rather 4 different models based on the same dataset, we chose the exponential form because it been used elsewhere in the literature , Trolle et al 2012.…”

“…Each model has a different set of assumptions and varies in the complexity of terms and data requirements, and many models based on empirical data were developed for specific limnological and meteorological conditions. Only recently have studies begun to incorporate multiple k models (Staehr et al 2010, Heiskanen et al 2014, Bartosiewicz et al 2015.…”

“…The exchange of soluble gases, including oxygen and carbon dioxide (CO 2 ), across the air-water interface is influenced by physical processes such as wind stress, convection, and currents and is described by the gas transfer coefficient (k; Zappa et al 2007) of magnitude (Wanninkhof 1992), primarily driven by diel and seasonal variations in mass and energy fluxes at the lake surface. Methods to determine k include gas tracers (Cole and Caraco 1998), floating chambers , Vachon et al 2010, and eddy covariance techniques (Jonsson et al 2008, MacIntyre et al 2010a, Heiskanen et al 2014. If observational data are unobtainable, k must be estimated using models.…”