Main content area

Simple regression models can act as calibration-substitute to approximate transient storage parameters in streams

Femeena, P.V., Chaubey, I., Aubeneau, A., McMillan, S., Wagner, P.D., Fohrer, N.
Advances in water resources 2019 v.123 pp. 201-209
dispersibility, equations, meta-analysis, models, monitoring, prediction, regression analysis, solutes, streams, tracer techniques, watersheds, Germany, United States
Transient storage models in combination with tracer tests are widely used to study solute transport dynamics in streams. Storage parameters included in such models are typically calibrated for one or more study reaches by monitoring solute concentrations and fitting breakthrough curve data. Since stream characteristics vary spatially and temporally, it is challenging to generalize these calibrated parameters for another stream reach. This study investigates the ability of simple regression models to predict transient storage parameters such as dispersion coefficient (D), transient storage area (As) and storage exchange coefficient (α). A meta-analysis of 834 tracer studies from 67 published papers was used to develop parsimonious non-linear regression models that relate storage parameters to easily available stream parameters such as discharge, velocity, flow width and flow depth. Correlation analysis showed moderate correlation of D with velocity, depth and width; and high correlation of As with the ratio of discharge to depth. Exchange coefficient (α) did not show significant correlation with available stream parameters. The models were tested using a subset of meta-analysis data and experimental tracer data from Hubbard Brook Experimental Forest located in the US and Kielstau Catchment located in Germany. We predicted storage and breakthrough curves with reasonable accuracy (R2 > 0.5) by using new regression equations and incorporating it into an advection-dispersion-storage model. These equations provide a viable alternative to approximate transient storage parameters in circumstances where time and cost-intensive reach-specific calibration is impossible. Therefore, such regression-based estimates of storage parameters can also form an integral part of larger watershed scale models by predicting solute transport and storage in stream reaches.