Outgassing of carbon dioxide (CO2) from inland waters has been estimated to offset approximately 20% of net uptake of carbon into the terrestrial biosphere. However, this calculation is based on estimates of source strength at the air-water interface that is highly uncertain [Ciais et al., 2013]. One of the largest unknowns is the accuracy and precision of freshwater partial pressure of CO2 (pCO2) estimates. To better understand observational uncertainty in CO2 flux from inland waters, here we conduct an analysis of historical observations and take new field observations at the North Temperate Lake Long Term Ecological Research (NTL LTER) site to quantify random errors in freshwater carbon system. We further estimate errors contribution to the precision of pCO2 derived from multiple carbonate equilibria and identify other sources of uncertainty in pCO2 calculations.
The net air-water CO2 exchange is calculated as a product of the CO2 gas transfer velocity (k), the CO2 solubility constant (K0) and the gradient between pCO2 in the atmosphere and water (ΔpCO2). The aquatic component of ΔpCO2 in current estimates of carbon evasion from inland waters relies on calculating pCO2 using thermodynamic equilibria in the carbonate system due to scarcity of direct pCO2 measurements at regional and global scales [Butman and Raymond, 2011; McDonald et al., 2013; Raymond et al., 2013]. Carbonate equilibria use temperature and the combination of two CO2-related parameters (i.e., pH, alkalinity (ALK), dissolved inorganic carbon (DIC)) to calculate pCO2 in surface waters [Parkhurst and Appelo, 1999; van Heuven et al., 2011]. Knowledge on the magnitude and sources of errors is necessary to quantify the effect of those uncertainties on estimates of aquatic pCO2 concentrations. Oceanographers have made significant efforts to standardize and reduce errors, resulting in thermodynamically consistent measurements of CO2-related parameters, highly precise and accurate estimates of the seawater carbonate system [Lueker et al., 2000; Millero, 2007], and thus of the ocean sink for anthropogenic carbon [Sabine et al., 2004; Ciais et al., 2013; Khatiwala et al., 2013]. Similar efforts however have not been undertaken for the carbon dioxide system in freshwaters. Moreover, there would be an additional challenge of addressing the more diverse chemical composition of inland waters [Dickson and Riley, 1978]. Therefore, an error analysis for freshwaters is critical to identify key uncertainties in measurements of CO2-related parameters and pCO2 calculations. Given the high accuracy and precision of atmospheric pCO2 measurements [Andrews et al., 2014], uncertainties attributed to measurement errors in aquatic carbon system parameters are likely the largest source of uncertainty in ΔpCO2 calculated …show more content…
We perform random error analysis using paired field observations taken under identical conditions from 16 to 26 year records in the North Temperate Lakes Long-Term Ecological Research (NTL LTER) dataset to quantify and characterize the random error in carbon system parameters. We propagate parameter errors onto three carbonate equilibria using a Monte Carlo approach in order to investigate the effect of those uncertainties on calculated pCO2 in four lake groups across a broad gradient of alkalinity and dissolved organic carbon concentrations. We also determine which carbonate equilibrium and alkalinity group is least sensitive to variation in input parameters and gives the most precise estimates of pCO2. We also compare calculated pCO2 with direct pCO2 observations to investigate if random error magnitudes among the two are comparable. Finally, we determine if random errors are the major source of uncertainty in pCO2
The net air-water CO2 exchange is calculated as a product of the CO2 gas transfer velocity (k), the CO2 solubility constant (K0) and the gradient between pCO2 in the atmosphere and water (ΔpCO2). The aquatic component of ΔpCO2 in current estimates of carbon evasion from inland waters relies on calculating pCO2 using thermodynamic equilibria in the carbonate system due to scarcity of direct pCO2 measurements at regional and global scales [Butman and Raymond, 2011; McDonald et al., 2013; Raymond et al., 2013]. Carbonate equilibria use temperature and the combination of two CO2-related parameters (i.e., pH, alkalinity (ALK), dissolved inorganic carbon (DIC)) to calculate pCO2 in surface waters [Parkhurst and Appelo, 1999; van Heuven et al., 2011]. Knowledge on the magnitude and sources of errors is necessary to quantify the effect of those uncertainties on estimates of aquatic pCO2 concentrations. Oceanographers have made significant efforts to standardize and reduce errors, resulting in thermodynamically consistent measurements of CO2-related parameters, highly precise and accurate estimates of the seawater carbonate system [Lueker et al., 2000; Millero, 2007], and thus of the ocean sink for anthropogenic carbon [Sabine et al., 2004; Ciais et al., 2013; Khatiwala et al., 2013]. Similar efforts however have not been undertaken for the carbon dioxide system in freshwaters. Moreover, there would be an additional challenge of addressing the more diverse chemical composition of inland waters [Dickson and Riley, 1978]. Therefore, an error analysis for freshwaters is critical to identify key uncertainties in measurements of CO2-related parameters and pCO2 calculations. Given the high accuracy and precision of atmospheric pCO2 measurements [Andrews et al., 2014], uncertainties attributed to measurement errors in aquatic carbon system parameters are likely the largest source of uncertainty in ΔpCO2 calculated …show more content…
We perform random error analysis using paired field observations taken under identical conditions from 16 to 26 year records in the North Temperate Lakes Long-Term Ecological Research (NTL LTER) dataset to quantify and characterize the random error in carbon system parameters. We propagate parameter errors onto three carbonate equilibria using a Monte Carlo approach in order to investigate the effect of those uncertainties on calculated pCO2 in four lake groups across a broad gradient of alkalinity and dissolved organic carbon concentrations. We also determine which carbonate equilibrium and alkalinity group is least sensitive to variation in input parameters and gives the most precise estimates of pCO2. We also compare calculated pCO2 with direct pCO2 observations to investigate if random error magnitudes among the two are comparable. Finally, we determine if random errors are the major source of uncertainty in pCO2