The effect of random parameter errors on predictability of long-term change in freshwater pCO2 calculated from carbonate equilibria

Most estimates of CO₂ evasion from inland waters rely on calculating pCO₂ using carbonate equilibria models. Therefore, the quality of input parameters directly influences uncertainty in pCO₂ estimates and detection level of pCO₂ temporal trends. We used North Temperate Lakes Long Term Ecological Research datasets to quantify random errors in the measurements of pH, alkalinity and dissolved inorganic carbon. Monte Carlo simulations were used to propagate uncertainties into long-term records in parameters and pCO₂ calculated from three thermodynamic equilibria models to determine the resultant precision of pCO₂ estimates. Random parameter errors were generally below 2% and varied by lake type. Temporal trends in pCO₂ differed across lakes and thermodynamic equilibrium model type. Each carbonate equilibrium model showed different sensitivities to random uncertainties and many trends were insignificant. Our results highlight the possible challenges in predicting long-term change in aquatic carbon efflux with existing long-term data.