A Review of Control Charts and Exploring Their Utility for Regional Environmental Monitoring Programs

Abstract

Industrial control charts are used in manufacturing to quickly and robustly indicate the status of production and to prompt any necessary corrective actions. The library of tools available for these tasks has grown over time and many have been used in other disciplines with similar objectives, including environmental monitoring. While the utility of control charts in environmental monitoring has been recognized, and the tools have already been used in many individual studies, they may be underutilized in some types of programs. For example, control charts may be especially useful for reporting and evaluating data from regional surveillance monitoring programs, but they are not yet routinely used. The purpose of this study was to promote the use of control charts in regional environmental monitoring by surveying the literature for control charting techniques suitable for the various types of data available from large programs measuring multiple indicators at multiple locations across various physical environments. Example datasets were obtained for Canada’s Oil Sands Region, including water quality, air quality, facility production and performance, and bird communities, and were analyzed using univariate (e.g., x-bar) and multivariate (e.g., Hotelling’s T2) control charts. The control charts indicated multiple instances of unexpected observations and highlighted subtle patterns in all of the example data. While control charts are not uniquely able to identify potentially relevant patterns in data and can be challenging to apply in some monitoring analyses, this work emphasizes the broad utility of the tools for straightforwardly presenting the results from standardized and routine surveillance monitoring.

Assessing the Interpretability–Performance Trade-Off of Artificial Neural Networks Using Sentinel Fish Health Data

“…In one approach, an expected or normal range is derived from any existing data and used to iteratively define the range of expected future observations [5,6]. Similar to industrial control charts, these environmental normal ranges can, however, also use covariates to account for background variation and enhance the sensitivity of the tool to the stressor of interest, to identifying the emergence of novel sources of variation, or to changes in the relationships between indicators and natural factors [7][8][9].…”