Predicting the consequences for stability of connecting ecosystems: What do spatiotemporal patterns of variables tell us?

Background/Question/Methods

Connection, or reconnection, of isolated ecosystem fragments happens under many natural scenarios (e.g., afforestation, altered migratory routes) and human-induced ones (e.g., wildlife corridors).  Spatial connection can either boost or impair the persistence and stability of species populations and has been a focus of conservation theory.  However, managing ecosystems requires an understanding of how connection affects the stability of a full suite of system variables, not just of focal species.  We therefore ask, is stabilization or destabilization of a variable predictable from its pre-connection behaviour?  We tested the stability responses of ten ecosystem variables (e.g., NPP, dissolved O₂, invertebrate populations) in aquatic microcosms connected into arrays with plastic tubing.  We note that any change in temporal variation at the landscape scale due to connection (variance_connected / variance_unconnected) arises from either: (1) changes in mean magnitudes, (2) changes in variances of individual patches or (3) changes in how much patches covary.  Predicting each type of change would be important for ecosystem management. We thus asked if aspects of a variable’s spatial and temporal variation could predict how overall temporal variation, or its components, would change with landscape connection.

Results/Conclusions

Variables differed in their response to spatial connection.  Whether measured as temporal variance or dimensionless estimators of stability, some variables varied significantly less when connected and others more.  But these contrasting responses were predictable. They were reconciled and predicted by (1) the strength of spatial gradients established by a variable, and (2) the degree of synchrony a variable showed among patches.  Strong spatial gradients indicate large differences in magnitude and/or dynamics among patches that, when allowed to mix, lead to jumps both in means (p = 0.02) and in synchrony (p < 0.01).  Meanwhile, more environmentally driven variables (e.g., temperature) had highly synchronized dynamics when unconnected, and often similar magnitudes among patches, and so were relatively unaffected by connection.  For these synchronous variables, changes in patch variances (p < 0.01) were more likely than changes to synchrony or means.  Results point to the notion that we might predict consequences of connection on stability from two related spatiotemporal parameters, namely; (1) the strength of spatial gradients as a proxy for the potential for change upon mixing and (2) the amount of patch synchrony as a barometer of whether the variable’s dynamics are driven from within patches (sensitive to connection) or externally (insensitive to connection).