In this study, we intended to make seed-based wetland restoration predictive and inform management by (1) seeking integrated experimental evidence generalizing stage-specific causative factors for demographic loss/mortality and (2) developing predictors oriented toward site-specific bottlenecks. Specifically, this study is focused on seed retention and seedling emergence, since they represent the most vulnerable life stages that follow sowing. Firstly, by means of large-scale field experiments, we tested how seed retention and seedling emergence were affected by varied management options (i.e., seed-planting depth and species selection) and a wide range of physical settings (i.e., elevation, hydrodynamic intensity, bed-level dynamics and sediment properties). Variable screening was then implemented to identify stage-specific governing factors. Secondly, the resulting insights and dataset were used to develop stage-specific predictive models using machine learning. Model experiments under various scenarios were then conducted to assess site-specific feasibility of potential seed-based restoration practices.
These files include the data used to create each figure in the manuscript, organized as follows:
1. Field experiments
a) Physical conditions at all study locations
b) Manipulated experiment concerning seed retention
c) Manipulated experiment concerning seedling emergence
2. Machine learning
a) Developing machine learning predictors on seed retention
b) Developing machine learning predictors on seedling emergence
For a complete description, see 'Data description.docx'
%I 4TU.ResearchData