TR-535 Comparison of Daily Streamflow Estimation Methods in the Thompsons Creek Watershed

Authors: Michael Schramm, Anna Gitter, Jackie Rambo, Stephanie DeVilleneuve, Ed Rhodes, Lucas Gregory

To facilitate development of a total maximum daily load in the Thompsons Creek watershed, we assessed methods for estimating mean daily streamflow in absence of a suitable long-term streamflow gage within the watershed. Continuous water levels were recorded at three sites in the watershed and combined with periodic 15-minute streamflow measurements to develop streamflow rating curves. The rating curves were used to calculate mean daily streamflow from measured depth data from May 2020 through March 2021. Three methods — drainage area ratio, linear regression, and generalized additive models — were used to estimate daily streamflows, and performance of each method was assessed against the rating curve calculated streamflows. Due to the lack of suitable streamflow gages, the drainage area ratio was determined to perform poorly in the Thompsons Creek watershed. Linear regression and generalized additive models provide acceptable performance for predicting daily streamflows and flow exceedance values. One station with a high proportion of no-flow days required the use of a two-step hurdle model that predicts the likelihood of flow and the volume of flow on a given day. Based on the results, ease of use, and model interpretability, we suggest using linear regression models at two sites and a generalized additive model at the third site to develop estimated daily streamflows for future total maximum daily loads.