loading page

How Much Does Stream-Groundwater Exchange Influence Whole-Stream Metabolism in a Small Mountain Stream?
  • Erin Jenkins,
  • Michael Gooseff
Erin Jenkins
Institute of Arctic and Alpine Research

Corresponding Author:[email protected]

Author Profile
Michael Gooseff
University of Colorado at Boulder
Author Profile


To estimate whole-stream metabolism, the open-channel oxygen method has traditionally provided underlying assumptions for modeled estimates of gross primary productivity (GPP) and ecosystem respiration (ER). The open-channel oxygen method employs the diel dissolved oxygen (DO) curve, which attributes stream metabolism to four processes: photosynthesis by primary producers, oxidative respiration, reaeration, and groundwater flux. Of these processes, groundwater flux is often assumed to be negligible when modeling whole-stream metabolism, which may introduce bias in estimates of GPP and ER. For example, if net groundwater flux is into the main channel, we may expect an overestimation of modeled ER due to dissolved oxygen dilution effects from influent groundwater. Although this error is recognized, there is a lack of continuous and spatial data that quantifies the extent of bias that is introduced by not including groundwater flux in model parameters. To investigate this bias, we measured whole-stream metabolism and groundwater flux in Como Creek, a headwater catchment 26 km west of Boulder, CO. DO sensors were deployed in the stream and groundwater wells in June 2018 at 3 sites along 500 m of the reach. BASE (Bayesian Single-station Estimation), a package available through R, was used for modeling whole-stream metabolism between peak streamflow and baseflow. BASE also optimizes the reaeration coefficient, which was estimated both including and neglecting groundwater discharge and DO concentration. Preliminary results indicate that Como Creek has a net groundwater flux out of the stream, resulting in higher rates of GPP in the groundwater-corrected model output, and indicating the potential for bias in uncorrected models.