Pinpointing stage-specific causes of recruitment bottlenecks to optimize seed-based wetland restoration.
Abstract
Attaining the goals of 'The UN-Decade on Ecosystem Restoration' requires efficient methods for large-scale restoration of degraded ecosystems. Seed-based approaches may offer opportunities for massive recovery of native vegetation but are prone to failure when applied to highly valued coastal wetlands such as salt marshes. Pinpointing the impact of early life stage transitions on recruitment variation across species and contexts is a critical first step toward amplifying seed-based restoration efficiency. Large-scale field experiments were conducted in 100 microhabitats across eight salt marshes to investigate root causes of variation in seed retention and seedling emergence, using four globally occurring salt marsh species as models. The resulting insights and dataset were then translated into predictors using machine learning, for targeted application in disentangling recruitment bottlenecks. Seed retention, regardless of species, was identified as the principal bottleneck in recruitment with hydrodynamic intensity, bed-level dynamics, and burial depth as critical governing factors. Seedling emergence was discerned as the critical bottleneck driving cross-species recruitment variability and was pivotally influenced by soil salinity and burial depth. Predictions using machine learning under different restoration scenarios indicated that simple management, such as seed burial or species selection, can create opportunities to bypass potential recruitment bottlenecks. Synthesis and applications. Our results suggest that the failure of seed-based coastal wetland restoration should be attributed to multiple recruitment bottlenecks that arise from different life stage transitions and are context/species dependent. In planning future seed-based restoration practices, managers should assess the variability of life stage-specific dominant factors at target sites to identify site-specific recruitment bottleneck(s). Our work underscores the need for strategic management that buffers against recruitment bottlenecks to improve restoration efficiency and advances the application of data-driven techniques to make seed-based restoration predictive.