Contributed Session.

Lead Organizer: Serghei Bocaniov , University of Waterloo (sbocaniov@uwaterloo.ca)

Presentations

06:00 PM

ASSESSMENT OF PREDICTED RELATIVE STREAMFLOW HYDROGRAPH BASED ON TIME-LAPSE IMAGERY (9708)

Primary Presenter: Alicia Bateman, Yale School of the Environment (abateman222@gmail.com)

Recently, there have been developments in implementing machine learning approaches to estimate relative stream discharge in the US using fixed, ground-based cameras through the USGS Flow Photo Explorer (FPE) website. This approach can improve our understanding of streamflow dynamics in small headwater streams where continuous discharge data is otherwise sparse. Previous research has focused on method development and application of this method in new areas. Here, using various statistical approaches, we assess the model’s effectiveness in capturing relative discharges and inter-stream dynamics based on a number of physical environmental conditions. New sites in Vermont and Virginia will be assessed alongside established public sites in the FPE. Local precipitation data will be used from some sites to consider factors such as active precipitation rate, estimated antecedent wetness, and recurrence of large rain events on the model’s performance. Additional parameters to be assessed are seasonality, temperature, specific conductance, and watershed characteristics. This assessment will help identify key information required when implementing this method and how the model and/or model training could be improved to better accommodate temporal variations in channel morphology. We believe these insights are valuable in better understanding how this new approach captures dynamic stream systems, informs future best practices when developing new sites, and opens the door to further applications of this method.

View Additional Details