Landscape metrics as indicators of coastal morphology: A multi-scale approach

• I use a multi-scale approach changing extent, resolution and moving-window size.
• Four landscape metrics are good indicators for coastal features.
• The smallest extent and resolutions achieve better validation results.
• Moving-window size affects sensitivity and specificity of the models.
• Raster metrics from vectorial coastlines are effective as coastal shape predictors.

In this study, the aim was to assess how commonly used landscape metrics perform as predictors of coastal shape. I examined nine metrics computed in FRAGSTATS to model the distribution of three coastal features of the Iberian Peninsula: beaches, capes and gulfs. A multi-scale approach was used combining three extents, three resolutions and five moving-window sizes to implement generalized linear models (GLMs). This study has found that three landscape metrics (edge density, mean perimeter-area ratio and percentage of landscape) were good indicators for the three coastal features, while mean shape index was only for beaches and gulfs. Differences in performance were found among the coastal features and scales studied. GLMs revealed that the smallest extent (Levante coast) and resolutions (250 m2 and 1 km2) achieved better validation results, suggesting a higher suitability of these scales for detecting changes in vectorial shorelines. Differences in sensitivity and specificity were also found among models estimated from different moving-window sizes. The present study confirms previous findings on the high multicollinearity of landscape metrics, and the convenience of testing correlations in advance. Raster-based metrics computed from vectorial coastlines were effectively incorporated in spatial modeling. This research provides new insight into the use of coastal shape to predict species distributions and other coastal processes, serving as a base for future studies.