Predicting barrier island dune habitats from simplified groundwater lens geometry and abiotic gradients

Authors:

Abstract
Coastal dune habitats are critical for biodiversity and coastal protection. Despite their importance, coastal dune habitats face pressures from infrastructural development, land-use management, and climate change. Accurately delineating their spatial distribution is essential for understanding ecosystem functioning under climate change and informing integrated coastal management. However, current modelling attempts either emphasize morphodynamic processes without hydrology or rely on computationally expensive numerical groundwater models. Here, we developed and tested a transferable workflow for predicting Natura 2000 dune habitats on sandy barrier islands, using the Dutch Wadden Sea as a case-study system. In addition, we tested how habitat suitability may respond to groundwater rise under sea-level-rise scenarios. We first evaluated the performance of a simplified freshwater-lens model based on the Ghyben–Herzberg principle in sandy dune systems, against the more complex iMOD-WQ groundwater model. The simplified approach reproduced island-scale lens geometry sufficiently well to serve as an indicator for habitat delineation. We then fitted logistic generalized additive models (GAMs) to predict eight habitats from five abiotic predictors including groundwater depth. Across islands, distance to the coast and groundwater depth consistently dominated habitat discrimination. A comparison between global and island-specific models demonstrated a trade-off between generality and local accuracy, with most errors concentrated in late-successional habitats. Under projected groundwater rise from sea-level rise, models projected a redistribution of habitat suitability, most robustly a decline for dune slacks. The resulting framework is computationally efficient, interpretable, and directly supports spatial planning and adaptive management of Natura 2000 dune habitats under changing groundwater and sea-level conditions.

This publication has 54 references indexed in Scilit: