tions when the recovery rate decreases as a regime shift is approached (Scheffer et al., 2012; van Nes and Scheffer, 2007). Changes in other system properties, such as variance, spatial correlation, autocorrelation, and skewness, may also prove useful in detecting approach to critical thresholds (Scheffer et al., 2012). But this field is quite new and does not yet have a body of well-established results to guide management or policy.
Other research has explored the concept of functional diversity (Allen et al., 2005), response diversity (Petchey and Gaston, 2009), and population densities in stochastic systems (Ives, 1995). These approaches require species-level ecological information about function and response. For some groups of organisms, such as birds, this information may be inferred from size and morphology (Cumming and Child, 2009).
When a system is well understood, it may be more straightforward to define appropriate metrics for ecosystem resilience. With respect to tidal wetlands, for example, it seems prudent to adopt practical metrics that are inclusive of numerous ecosystem services and overall ecosystem function. Practical metrics with these characteristics would be (1) change in total wetland area by plant community type and (2) relative elevation (relative to mean sea level). Similarly, critical components of the ecosystem service under study that can be used to measure resilience should be identified. The ability to do this links directly to the understanding of the ecological production functions associated with the service and links directly back to the understanding of ecosystem dynamics.
Finding 3.9. The measurement of resilience poses a number of conceptual and practical challenges. Measures of speed of recovery to pre-disturbance conditions (engineering resilience) depend on having good baseline data and may vary depending on what ecosystem service or system component is measured and what type and severity of disturbance is considered. Measuring how likely a system is to cross a critical threshold and undergo a regime shift (ecological resilience) is a topic that is at the frontier of science. Consensus on practical measures that can be used to predict the location of critical thresholds or the probabilities of regime shifts does not yet exist. However, for specific systems, it may be possible to define a set of metrics that measure key conditions or processes linked to system dynamics that can predict the resilience of the system and the return of provision of ecosystem services.
Resilient ecosystems can lead to the resilience of specific ecosystem services. However, ecosystems typically generate multiple ecosystem services. Changes in ecosystem structure and function will generally not affect all ecosystem services in the same manner. Therefore, different services could have different outcomes after a disturbance. As noted by Carpenter et al. (2001), evaluation of a system’s resilience requires identification of the initial state of the system (regime) and the disturbances that can impact that system. If the system is viewed through the lens of ecosystem services, then there is a need to identify the most important services and how they may be affected by potential alternate states of the system. For example, coastal marsh habitat of the GoM provides several significant services, two of which are storm surge protection and fishing opportunities, both recreational and commercial. The quality of these services is directly linked to the structural condition of the marsh. Fragmented marsh,