Nonlinear self-organizing processes, like this session of talks, are interesting because they are hard to predict. Though it seems risky to recap an event that has not yet happened, it is likely that the session can be summarized by considering three questions. (Q1) Do indicators or early warnings of ecosystem regime shifts show common patterns across systems or are they idiosyncratic? (Q2) What is the theoretical and empirical evidence about indicators, and how consistent are theory and observation? (Q3) Do early warnings have implications for policy that differ from the uses of other kinds of ecological information?

Results/Conclusions

Important changes can result from crossing critical thresholds, which are in principle measurable. In practice, however, thresholds appear, change locations, or disappear over time. Statistical estimates of thresholds can be highly imprecise.

Certain statistics of ecological time series or spatial patterns change predictably as a critical threshold is approached. These early warning signals are observable before the ecosystem changes. Early warning statistics are similar in diverse kinds of ecosystems and regime shifts (answering Q1). Cautions apply. If an ecosystem is forced rapidly across the threshold, the signals is not detected in time. Not all big changes may involve critical thresholds. Key variables to monitor differ among ecosystems and regime shifts.

Theory of early warnings is ahead of empirical field tests (answering Q2). Some empirical studies support theory of early warnings, but field tests are few. Mechanisms that could produce false positives or false negatives exist, and these must be considered in field studies.

Moving unknown thresholds pose unique challenges for management. The precautionary principle fails, because one does not know what to be cautious about. Instead resilience thinking is required to maintain good conditions when possible, adapt to unavoidable breakdowns, and switch modes of operation when the old mode has failed.

In resilience thinking, early warnings are used to avoid dangerous thresholds, or exploit opportunities to shift a system to a preferred state (answering Q3). Use of early warnings requires anticipation of what thresholds might exist; attention to indicators through monitoring, analysis, and interpretation; and appropriate response. Thus use of early warnings in management requires knowledge of what to expect (for example findings presented by today’s speakers); capability to monitor, analyze, and interpret early warnings; and resources (such as people, equipment, authority, and time) to respond.