Scrupulous proxies: Defining and applying a rigorous framework for the selection and evaluation of a suite of ecological indicators

Abstract Ecosystem indicators are key to understanding, assessing, and managing ecosystems and to implementing an Ecosystem Approach to Management. We present a guidance framework for the selection and evaluation of indicators for assessment of ecosystem status, trends monitoring and reporting and apply it to the Scotian Shelf Bioregion, Nova Scotia, Canada. In particular, we focus on the selection of a parsimonious suite of complementary indicators using clear criteria and hierarchical cluster analysis to reduce redundancy in the indicator suite. In applying the framework to the Scotian Shelf Bioregion we separate the region into a range of spatial scales for indicator estimation to determine the impact of spatial aggregation on indicator redundancy. Overall results suggest patterns of redundancy among indicators were generally consistent at all spatial scales. Differences in indicator trends for eastern and western Scotian Shelf were observed, reflecting their different environmental properties and fishing histories. In a few cases, spatial differences in indicators trends were large enough to remove these indicators from a cluster representing all spatial scales. We recommend that the Redundancy Analysis should be repeated every 5–10 years to ensure that the clusters are still strong and that the indicator selected to represent the cluster is still optimal. Overall, the assessment of the Scotian Shelf Bioregion highlighted large coherent trends across indicators and attributes at all spatial scales indicating that this system has undergone massive, negative change since 1970. Despite reductions in fishing pressure in many fisheries through resource management actions, most attributes have not improved in recent years. This Framework represents a general approach that can be extended to additional system goals, attributes and indicators and is widely applicable across data rich and data poor systems.