Dissolved Oxygen Signals in Spatially Heterogeneous Rivers

The genesis and propagation of dissolved oxygen signals in river networks is of particular interest to stream ecologists seeking to understand ecosystem energetics, and the attendant biological and chemical functions. In open-channel applications, gross primary production (GPP) and ecosystem respiration (ER) are computed from sub-daily variation in dissolved oxygen, subject to adjustments for physical gas exchange (in turn controlled by the gas exchange rate, K600). Uncertainties in process rates can be vexing, particularly given equifinal fits to observed time series for parameters that describe K600 and ER. Compounding the challenge of inferring three unknowns (GPP, ER, K600) from one time series is the embedded assumption of the method that the upstream reach is homogeneous over the integration length of the oxygen signal. While this assumption is tenable in some settings, particularly those with high K600 and thus short integration lengths, the presence of biological, hydraulic, and source water heterogeneity (e.g., stream confluences) alters the timing and shape of the dissolved oxygen time series. This in turn is expected to strongly influence the inference of stream metabolism. To explore these effects, we coupled a 1-dimensional model of stream advection and dispersion with various scenarios of spatial variation in the rates of primary production, respiration and gas exchange. In each scenario, where the metabolic inputs are known, we also estimated stream metabolism from the resulting dissolved oxygen time series. Our results suggest that errors are small when spatial heterogeneity occurs over short length scales compared with the integration length. However, as the length scales of heterogeneity and integration converge, errors can be substantial. We propose best practices to minimize these errors at reach and network scales.