Woodland recovery following drought-induced tree mortality across an environmental stress gradient.

Recent droughts and increasing temperatures have resulted in extensive tree mortality across the globe. Understanding the environmental controls on tree regeneration following these drought events will allow for better predictions of how these ecosystems may shift under a warmer, drier climate. Within the widely distributed piñon-juniper woodlands of the southwestern USA, a multiyear drought in 2002-2004 resulted in extensive adult piñon mortality and shifted adult woodland composition to a juniper-dominated, more savannah-type ecosystem. Here, we used pre- (1998-2001) and 10-year post- (2014) drought stand structure data of individually mapped trees at 42 sites to assess the effects of this drought on tree regeneration across a gradient of environmental stress. We found declines in piñon juvenile densities since the multiyear drought due to limited new recruitment and high (>50%) juvenile mortality. This is in contrast to juniper juvenile densities, which increased over this time period. Across the landscape, piñon recruitment was positively associated with live adult piñon densities and soil available water capacity, likely due to their respective effects on seed and water availability. Juvenile piñon survival was strongly facilitated by certain types of nurse trees and shrubs. These nurse plants also moderated the effects of environmental stress on piñon survival: Survival of interspace piñon juveniles was positively associated with soil available water capacity, whereas survival of nursed piñon juveniles was negatively associated with perennial grass cover. Thus, nurse plants had a greater facilitative effect on survival at sites with higher soil available water capacity and perennial grass cover. Notably, mean annual climatic water deficit and elevation were not associated with piñon recruitment or survival across the landscape. Our findings reveal a clear shift in successional trajectories toward a more juniper-dominated woodland and highlight the importance of incorporating biotic interactions and soil properties into species distribution modeling approaches.