An Exploratory Data Analysis (EDA) Tool for Tracer Selection in Sediment and Pollution Fingerprinting Studies

Abstract

Tracer selection is a critical step for any mixing model that uses the information of multiple variables to find the most likely solution for sediment or pollution source apportionment. The purpose of this work was to design an Exploratory Data Analysis (EDA) tool for qualitative and quantitative interpretation of tracer selection based on boxplots. This tool establishes a set of conditions presented as a decision tree to support the user in tracer selection decisions where current systems are either too statistically driven, which limits user interaction with the data, or rely too heavily on judgement and prior knowledge of tracer behaviour. This method extends upon the classical “range test” by incorporating a set of conditions and thresholds, creating a semi-quantitative method with a visual component to improve the tracer selection process and enable targeted consideration of geochemical behaviours for conservativism purposes. We present the results from experimental mixtures using the mixing model MixSIAR to demonstrate the method and illustrate the value of this approach for improved tracer selection. The tool, which is written in R, is freely available for use across a wide range of source apportionment applications.