Multivariate Control Charts for Ecological and Environmental Monitoring

Abstract

Ecological and environmental monitoring has become increasingly important, with increasing threats from human disturbances. Monitoring usually involves sampling from several sites of a similar habitat at regular (or irregular) intervals through time. The purpose of monitoring is to determine where and when an impact may have occurred or, once detected, may still be occurring. Sequential statistical methods, including control charts, as developed for industrial applications, offer some promise in this regard. These provide a way of identifying when a system (e.g., in a factory) is going “out of control,” so as to trigger an alarm to stop the system and employ appropriate remedial measures. Such techniques clearly would be useful in the context of environmental monitoring. Traditional control charts, however, cannot be used for many ecological applications because they do not handle multivariate data, and individual counts of species abundances do not generally fulfill the necessary statistical assumptions. A distance‐based multivariate control chart method is described here, with some examples of its use in monitoring coral reef fish assemblages of the Great Barrier Reef, Australia. The method is flexible, as it can be based on any dissimilarity measure of choice, and useful, as it does not require any specific assumptions regarding distributions of variables. Bootstrapping techniques are used to provide control‐chart limits for an appropriate multivariate distance‐based criterion through time. The method is designed to identify impacts at individual sites as quickly as possible, thus triggering an “alarm bell” in the context of ecological monitoring. It can also be applied at several spatial scales in hierarchical designs.

40 Years of benthic community change on the Caribbean reefs of Curaçao and Bonaire: the rise of slimy cyanobacterial mats

“…A two-way RELATE test was used to assess correlation between patterns at different depths and sites by comparing submatrices of the individual quadrats (M = 9999 simulations). In addition to the seriation tests, multivariate control chart analysis (Anderson and Thompson 2004) was applied to the Bray-Curtis matrices to determine when a community had become significantly different from the start of the time series (ESM Fig. S2).…”

“…Bootstrapping (1000 bootstraps) was used to provide percentile limits (50, 75, 90 and 95%). Points positioned above the 90% limit were considered to have a significantly different community compared to the start of each time series (Anderson and Thompson 2004). R code for the multivariate control chart analysis was developed by EM.…”

Long-Term Deer Exclusion Has Complex Effects On A Suburban Forest Understory

“…Cao et al (1997) used log 10 , but the current function uses natural logarithms so that the values are approximately 2.30 times higher than with 10-based logarithms. Anderson & Thompson (2004) give an alternative formulation of Cao index to highlight its relationship with Binomial index (above).…”

From fronds to fish: the use of indicators for ecological monitoring in marine benthic ecosystems, with case studies from temperate Western Australia

“…MJ Anderson). The data from the monitoring program is being analysed using multivariate control chart techniques (Anderson and Thompson 2004) to track sequential estimates of the centroid of the community data as an indicator of change in ecosystem state with time.…”