Improving Intertidal Reef Mapping Using UAV Surface, Red Edge, and Near-Infrared Data

Type Article
Date 2019-09
Language English
Author(s) Collin Antoine1, Dubois StanislasORCID2, James Dorothée1, Houet Thomas3
Affiliation(s) 1 : Ecole Pratique des Hautes Etudes, PSL Université Paris, CNRS LETG, 35800 Dinard, France
2 : IFREMER, laboratoire d’écologie benthique côtière (LEBCO), 29280 Plouzané, France
3 : CNRS, Université Rennes 2, Unité Mixte de Recherche 6554 LETG, 35000 Rennes, France
Source Drones (2504-446X) (MDPI AG), 2019-09 , Vol. 3 , N. 3 , P. 67 (12p.)
DOI 10.3390/drones3030067
WOS© Times Cited 20
Keyword(s) reefs, red edge, near-infrared, digital surface model, classification, Sabellaria alveolata

Coastal living reefs provide considerable services from tropical to temperate systems. Threatened by global ocean-climate and local anthropogenic changes, reefs require spatially explicit management at the submeter scale, where socioecological processes occur. Drone surveys have adequately addressed these requirements with red-green-blue (RGB) orthomosaics and digital surface models (DSMs). The use of ancillary spectral bands has the potential to increase the mapping of all reefscapes that emerge during low tide. This research investigates the contribution of the drone-based red edge (RE), near-infrared (NIR), and DSM into the classification accuracy of five main habitats of the largest intertidal biogenic reefs in Europe, built by the honeycomb worm Sabellaria alveolata. Based on photoquadrats and the maximum likelihood algorithm, overall, producer’s and user’s accuracies were distinctly augmented. When isolated, the DSM provided the highest gain percentage (3.42%), followed by the NIR (2.58%), and RE (2.02%). When joined, the combination of the DSM with both RE and NIR was the best contributor (4.98%), followed by the DSM with RE (4.80%), DSM with NIR (3.74%), and RE with NIR (3.22%). At the class scale, all datasets increasingly advantaged sand, gravel, reef, mud and water. The rather low effect of the DSM with NIR (3.74%) was assumed to be linked with a statistical noise originated from redundant information in the intertidal area

Full Text
File Pages Size Access
Publisher's official version 12 5 MB Open access
Top of the page