Habitat mapping of coastal dunes with deep learning

About one-third of the world's shoreline is defined by sandy coasts with developed dune ecosystems. These ecosystems have drastically degraded due to anthropogenic pressures. To develop strategic management that counteracts this degradation, it is essential to closely monitor ongoing habitat changes. Traditionally, coastal dune monitoring is based on field observations, which are labour intensive and costly. While automated analyses of aerial imagery could reduce monitoring efforts and enhance spatial coverage, to date, its application has remained limited to a single small-scale trial (<2 km 2 ). Here, we trained a Convolutional Neural Network to map the Dutch coastal dunes (562 km 2 ) at 25 cm resolution using six habitat classes: bare sand, shrubs, fresh water, grass, broadleaf trees, and needleleaf trees. Training the network on only RGB imagery resulted in predictions with 92 % accuracy, 80 % average recall and 70 % precision. Model performance increased when the network was trained on all available data – RGB imagery, near-infrared, distance to sea, digital surface model, and canopy height - resulting in 95 % accuracy, 88 % averaged recall and 80 % precision. Finally, we compared the predictions with 499 in-field observations across the Dutch coastal dunes and found 88 % accuracy, 74 % averaged recall and 62 % precision. We used this model to create a map of the entire Dutch coastal dunes, which enables rapid and precise assessments of habitat diversity and extent. As habitat and species diversity are intrinsically linked, our results showcase how automated image analysis can enable biodiversity monitoring on a national scale. • We trained a deep learning model with multispectral imagery and elevation data. • It predicts dune habitats with 95 % test set accuracy and 88 % in-field accuracy. • Using the model, we mapped the Dutch coastal dunes (562 km 2 ) with 25 cm resolution. • Our method enables labor-efficient monitoring on a national scale. • We showcase how ecosystem monitoring can be advanced and upscaled.