Search

Extensive efforts have been undertaken in the northern Gulf of Mexico to restore coastal wetlands that have been lost rapidly. The evaluation of these restorations mostly focused on individual-project scales. A modeling framework that can coherently synthesize multi-scale monitoring data and account for various uncertainties would improve quantitative evaluations at broader spatial scales needed for regional decision-making. We aim to develop such a framework to investigate the impact of different restoration methods (hydrological alteration, breakwater infrastructure, vegetative planting, or marsh creation using dredged materials) on wetland loss on the outermost mainland coastlines in Louisiana. We did this by implementing multi-level Bayesian models to predict areal wetland loss (1996-2005 before Hurricane Katrina) as a function of local geophysical variables (relative sea-level rise, wave height, tidal range) and a dummy variable indicating presence/absence of restoration. We assumed the effects of these variables varied by broader watershed scales. The restoration's effect also depended on temporal scales of implementation. The results indicate the sites with hydrological alteration, when implemented for longer than 7 years, had significantly smaller areal wetland loss, compared to the reference sites controlled for the local geophysical variables, in the Chenier Plain watershed, but not in the lower Mississippi River watershed. The effects of the other restoration methods on wetland loss were not significant based on limited numbers of sites. The Bayesian modeling framework we developed can integrate monitoring data/key drivers across projects with uncertainties accounted for, it is adaptable, and presents a useful tool in restoration evaluations spatially and temporally.