A new approach to selecting sustainable ecosystem indicators: A neural network

A management plan designed to maintain ecosystem health requires measurable indicators, as they indicate when certain components of the ecosystem begin to deviate from the norm. Current methodology for selecting ecosystem components to monitor and then creating the indicators has been haphazard, with only the final indicator set presented in most studies. This dissertation contributes to scientific knowledge in a few key areas, including the establishment of a strict methodology for the selection of key ecosystem components of concern, establishment of acceptable values per indicator, and refinement of the potential indicator set to a minimum, realistic number. Presentation is via the case study of Canadian Forces Base Shilo, a military base requiring a minimal set of indicators to ensure the continuation of ecologically sustainable training. The science of indicators and ecosystem management is further advanced through the creation and use of an artificial neural network. This modelling approach determined the required measured indicators by a non-statistical method while minimizing the potential for researcher bias. The neural network model reduced the indicator set from 62 to 12, while still ensuring the persistence of a healthy ecosystem.