Publication

Two decades of research has shown that the global river network emits significant amounts of greenhouse gas. Despite much progress, there is still large uncertainty in the temporal dynamics of gas exchange and thus carbon emissions to the atmosphere. Much of this uncertainty stems from a lack of existing tools for studying the spatiotemporal dynamics of gas exchange velocity k600 ${k}_{600}$ (the rate of this diffusive transport). We propose that the NASA/CNES/UKSA/CSA Surface Water and Ocean Topography (SWOT) satellite can provide new insights to fluvial gas exchange modeling upon launch and subsequent data collection in 2022. Here, we exploit the distinct geomorphology of SWOT-observable rivers (>50 m wide) to develop a physical model of gas exchange that is remotely sensible and explains 50% of log-transformed variation across 166 field measurements of k600 ${k}_{600}$. We then couple this model with established inversion techniques to develop BIKER, the "Bayesian Inference of the k600 ${k}_{600}$ Exchange Rate" algorithm. We validate BIKER on 47 SWOT-simulated rivers without an in-situ calibration, yielding an algorithm that predicts the k600 ${k}_{600}$ timeseries solely from SWOT observations with a by-river median Kling-Gupta Efficiency of 0.21. BIKER is better at inferring the temporal variation of gas exchange (median correlation coefficient of 0.91), than reproducing the absolute rates of exchange (median normalized RMSE of 51%). Finally, BIKER is robust to measurement errors implicit in the SWOT data. We suggest that BIKER will be useful in mapping global-scale fluvial gas exchange and improving CO2 $\left[\mathrm{C}{\mathrm{O}}_{2}\right]$ emissions estimates when coupled with river CO2 $\left[\mathrm{C}{\mathrm{O}}_{2}\right]$ models.