Shallow water UAV based habitat monitoring of seagrass meadows in the Baltic Sea

Seagrass meadows are crucial habitats since they serve as fish nurseries and food sources for many marine species. They prevent nearshore erosion and are an important CO₂ sink. As the plants are bound to the photic zone, seagrass meadows normally populate the shallow coastal zones. Unmanned aerial vehicles (UAV) are gaining popularity within the earth sciences community. Most surveys are of terrestrial nature and carried out by using the camera of the UAV to obtain orthophotos and three-dimensional surface models of a survey area. In comparison to space-borne systems, UAVs are capable of higher resolution image quality and time independent measurements, which enables an event-based surveying approach. We here present a submarine habitat mapping study, obtained by using an UAV flying 75 m above the water surface. Within the frame of the BONUS ECOMAP project, we aim to conduct repeated UAV surveys over the seasonal cycle to observe changes within coastal seagrass bed habitats. The key study area is located in the Baltic Sea offshore Heidkate (near Kiel, Germany). For data acquisition, we are using a commercial DJI Inspire 2 UAV with a gimbal mounted 20.8 megapixel Zenmuse X5S camera with a 15 mm/ 1.7 ASPH lens. For less reflection and distortion at the air-water interface, we are using a B&W circular polarized filter. Ground control points are measured and leveled with a Leica RTK system, which has a lateral resolution of ~2 cm. We process the data with the commercial software Pix4D™ and Agisoft PhotoScan™ to compute orthomosaic images and digital elevation/surface models. Since February 2018, we were able to conduct repeated surveys offshore Heidkate and Wendtorf (Germany). The average resolution of the orthomosaic data is better than 5 cm/px. First results show that we can obtain high-resolution images of habitats within water depths less than ~4 m in the Baltic Sea. Penetration is limited to factors like wave action, suspended sediment load and angle of the solar radiation. We perform supervised classification and pattern detection for habitat identification and discrimination. The data show the presence of seagrass, algae but also rocks, which are exposed at the seafloor. All scenes show a seasonal variability of the extent of seagrass meadows which are affected by migrating sand bars and major storm events. These data are the basis for a long-term monitoring framework, we are currently establishing in the working area.