Discussion

Conservation management plans that are based on current patterns of species distribution will become less effective under climate change (Carroll et al. 2017), so new approaches are needed for managing natural resources (Lawler 2009; Choe et al. 2018). We applied a climate network analysis to examine the climate links among 48 wetlands and found five climate-connection categories that can inform natural resource management strategies.
Wetlands lend themselves well to a network analysis because they are distinct landscape features occurring across multiple climates that are composed of different species at most nodes. While wetlands themselves are ecological islands that may have enhanced capacity to persist under climate change (Cartwright 2019), for conservation management of the vertebrate species using wetlands, climate dynamics will likely become an important challenge. Shorebirds in particular are largely dependent on the NWRS-managed network due to substantial wetland loss and their unique habitat requirements (Schaffer-Smith et al. 2018). Therefore, the five climate categories we found in the network analysis provide useful context for developing climate-adaptive management strategies for vertebrate species, particularly birds within the refuge network system, considering its unique ecosystem and management approach.
Most small-size refuges were climatically disappeared in the future time periods, or had only one other unit with a future analogous climate. For refuges with only one future unit, it may be beneficial to conduct further research into the resilience and species climate vulnerability at both the current and future locations. For climatically isolated units, some may have suitable climates in units beyond the boundaries of our study area (Choe and Thorne 2019), such as in northern refuges whose climates may shift into adjoining states (Lenoir and Svenning 2015). Alternatively, new refuges may need to be identified to account for species inhabiting these climatically isolated refuges, particularly if their climates do not appear anywhere in the larger network of protected areas.
The refuge network contains four climate hubs, where the current climate conditions of many refuges converge in the future (Bitter Creek NWR, North Central Valley WMA, Hopper Mountain NWR, and Desert NWR). These nodes could be targets for managed relocation (Schwartz et al. 2012), and target species lists can be developed from the species lists of those refuges whose climate conditions become unsuitable, but shift to these arriving node units in the future. Conversely, the climate in some nodes (Desert NWR, North Central Valley WMA, San Diego NWR) appears in many other refuges (dispersing climates), suggesting greater flexibility in species relocation for the species found in this class.
Not surprisingly, the three refuges with enduring climates include the two largest, whose range of climate conditions is broad enough to contain their own current climate conditions in the future. Although the North Central Valley WMA is only designated as a refuge area of interest and therefore the lands are mostly not under conservation management, its role as an important node in the climate networks shows the value that conserving this large area could provide. The third unit to retain its own climate, the San Diego NWR, is moderately sized (eighth largest). It is located in coastal southern California, and maritime influence may be a factor in the lower warming in this area than elsewhere in our study domain (Fig. S2). Management for this climate class of refuges, should focus on external threats such as invasive species, habitat fragmentation, and maintaining ecosystem functions.
This study could be considered a coarse filter approach (Groves et al. 2002; Khoury et al. 2011) to support wetland climate-conservation. Further research could consider wetland hydrological processes and additional climate variables to better define the climate conditions of each refuge. However, when we conducted a sensitivity analysis using up to nine climate variables, we found more variables narrowed the areas identified as climate analogs and rendered most refuge units climatically isolated. We decided to use the two most fundamental climate variables (temperature and climate). In addition, our study does not include projected shifts in species’ ranges (Choe and Thorne 2017; Choe et al. 2017), or their sensitivity and adaptive capacity (Thorne et al. 2016) which are commonly used frameworks for understanding the vulnerability of individual species to climate change (Glick et al. 2011). These approaches have their own strengths and limitations, including assumptions about dispersal success and biotic interactions (Perez-Garcia et al. 2017). Instead this study focuses on the climate conditions that are part of the exposure metrics for species, but that are useful for describing the sites species occupy.
Our discovery of five classes of climate risk to existing wetland conservation features emphasizes the importance of including climate information when developing management strategies for protected area networks. It may be possible to quickly identify these classes for other conservation areas based on size, topographic complexity, and proximity to maritime influences. For example, the >2300 Ramsar wetlands which cover over 2.5 million km2 globally (https://rsis.ramsar.org/), could be good candidates for climate network analyses. Finally, although some of the refuges we studied are very small, they intrinsically have a high adaptive capacity because the majority are managed wetlands into which additional waters could be pumped, which may offset climate impacts as projected here, at least to some degree. Although this study was limited to the refuge network, our next steps would be to explore the locations and time schedules for additional refuges.