Accounting for temporal changes in carbon dioxide (CO2) emissions from freshwaters remains a challenge for global and regional carbon budgets. Here, we synthesize 171 site-months of eddy covariance flux measurements of CO2 from 13 lakes and reservoirs in the Northern Hemisphere (NH) and quantify the magnitude and dynamics at multiple temporal scales. We found pronounced diel and sub-monthly oscillatory variations in CO2 flux at all sites. Diel variation converted sites to daily net sinks of CO2 in only 11% of site-months. Upscaled annual emissions had an average of 25% (range 3-58%) interannual variation. Given temporal variation remains under-represented in inventories of CO2 emissions from lakes and reservoirs, revisions in CO2 flux are needed using a better representation of sub-daily to interannual variability. Constraining short- and long-term variability is necessary to improve detection of temporal changes of CO2 fluxes in response to natural and anthropogenic drivers.

Accounting for temporal changes in carbon dioxide (CO2) emissions from freshwaters remains a challenge for global and regional carbon budgets. Here, we synthesize 171 site-months of eddy covariance flux measurements of CO2 from 13 lakes and reservoirs in the Northern Hemisphere (NH) and quantify dynamics at multiple temporal scales. We found pronounced sub-annual variability in CO2 flux at all sites. Accounting for diel variation, only 11% of site-months were net daily sinks of CO2. Annual CO2 emissions had an average of 25% (range 3-58%) interannual variation. Nighttime emissions regularly exceeded daytime emissions. Sources of CO2 flux variability were delineated through mutual information analysis. Sample analysis of CO2 fluxes indicate importance of continuous sampling. Constraining short- and long-term variability is necessary to improve detection of temporal changes of CO2 fluxes in response to natural and anthropogenic drivers.

The isotopic composition of surface water vapor flux (δ) is a quantity frequently used to investigate the local and regional water cycle. In this study, the δ determined with the Keeling method was evaluated against the flux-gradient method and the Craig-Gordon model prediction. Previous studies have shown that the choice of regression fitting methods can bias the δ intercept results and precision of the Keeling method. Here, the Keeling method was applied to high-frequency (0.2 to 1 Hz) data measured at a cropland and a lake site to test different regression methods. Results show that the Keeling method with the York’s solution (YS) and the ordinary least squares (OLS) regression produced robust estimates of δ when compared with the flux-gradient method. Increasing concentration range reduced the standard error of estimate but did not bring obvious improvement to the bias error for the YS and OLS regression. The Keeling result was better using data from two sampling heights than only one. There was evidence that the Keeling method with the OLS regression slightly outperformed the flux-gradient method during periods with small vertical vapor gradient. Results also show that the Keeling method with the geometric mean regression gave highly biased estimate of δ for the types of isotope ratio infrared spectroscopy analyzer deployed in this study. These results can inform δ calculations and future experimental designs.