Publications: Research reports and publications

Review of river ecosystem threshold models

1 November, 2012
Cawthron Report 2259. Prepared for the 'management of cumulative effects of stressors on aquatic ecosystems' research programme (C01X1005).


Scientists have increasingly become interested in the detection and quantification of ecological thresholds in river ecosystems with a view to defining scientifically defensible criteria for resource management. Ecological thresholds are defined as points where there is an abrupt change in an ecosystem property or where small changes in an environmental driver produce a large response in an ecosystem. This report outlines the range of situations where the ecological threshold concept has been used in the scientific literature in respect to river ecosystems. In most cases detection of thresholds is approached through the spatial study of driver-state relationships.

Thresholds are associated with nonlinear behaviour and can take the form of a step, an abrupt but continuous change, or a subsidy-stress response. Multiple analytical approaches to threshold determination are available. They can be coarsely divided into exploratory and inferential analytical approaches, but they also differ in the number of drivers they can take into account and whether they base threshold definition on univariate or multivariate response variables or on multiple taxa responses. Exploratory approaches are available that describe nonlinear threshold response relationships between drivers and ecological response variables, or that define thresholds based on multiple taxa responses.

Inferential threshold models which underlie the most commonly used approaches, nonparametric changepoint analysis and breakpoint regression analysis and their variants; test a formal hypothesis of the existence of a threshold using simple linear threshold models. Application of these models should be guided by the expected threshold response and associated with an ecologically sound hypothesis. Failure to use a representative threshold model and to meet statistical assumptions underlying the modelling approach may lead to spurious threshold detection and location.

In this report we review exploratory and inferential analytical approaches used to identify ecological thresholds as reported in the scientific literature, in view of gaining a better understanding of the data assumptions, the strengths and the limitations of specific models. We also conduct a pilot analysis on an existing driver-response data set and provide guidance on the most appropriate approach to apply to future data sets.

The four recommendations made as a result of this review are:

  • investigating and testing for the existence of statistical thresholds in support ofecological thresholds in a suite of stressor-response relationships
  • using analyses that account for multiple drivers
  • application of exploratory methods that use information from multiple taxa responsesto define thresholds where community structure is likely to change
  • providing multiple lines of evidence.