Predicting the duration of stream tracer experiments

Code written by Clarissa Glaser (February 2025) in R. The code was developed as a Shiny app to provide an interactive interface.

This application uses four random forest models to predict the duration of experiments using two different tracers (chloride and uranine). Specifically, the models predict the duration required to achieve transport metrics (coefficient of variance (CV) and holdback) from tracer-truncated breakthrough curves (BTCs) that deviate by less than a specified percentage from the actual transport metrics of non-truncated BTCs. The models were trained on data from a virtual experiment using tracer-specific mass-to-streamflow ratios (see publication below). Each tab in this application represents a different random forest model. The models differ in the transport metrics (CV and holdback) and tracers. The required input parameters include prevailing stream characteristics (D, ATS, Q, and α), which can be approximated prior to the experiment, as well as the desired percentage deviation between transport metrics of tracer-truncated and non-truncated BTCs.

Select the 'CV chloride' tab to predict the duration of a slug tracer experiment where the CV from a chloride-truncated BTC deviates by less than a specific percentage (defined by you) from the actual value of a non-truncated BTC. Define your stream by changing the values of D, ATS, Q, and α according to your stream characteristics.

The folder "Predicting_the_duration_of_stream_tracer_experiments_(version 1.0)" contains the following files:

R file "app"

random forest model "RF_chloride_CV_TSI.rds"

random forest model "RF_chloride_Holdback_TSI.rds"

random forest model "RF_uranine_CV_TSI.rds"

random forest model "RF_uranine_Holdback_TSI.rds"

Open the R file "app" and press the "run app" button to open the interactive interface.