Habitat classification: approximation or reality?
Intensive coastal and marine resource management, climate change investigations and nature conservation needs, boost the demand for adequate seafloor mapping. Habitat mapping encompasses a variety of ocean floor parameters such as sediment texture, bedforms and benthic biota allows complementing and supporting scientific studies as well as stakeholder decisions. Politics or directives are often reliant on data and experiences from science. In this course, diverging scientific and stakeholder interests increase debates about the three focal questions in habitat mapping: which map resolution to choose, which parameter to consider and which classification to apply? This conflicts arise because stakeholders have to work with simplified classifications to fulfill e.g. European guidelines (e.g. EUNIS) while scientists work with a higher degree of complexity. In the southeastern North Sea, considerable effort went into highly resolved area-wide acoustic investigation of sediment-dynamic processes. While a huge amount of grab samples have been taken through the past 50 years (~ 500,000 station in the German Bight), the systematic collection of precisely referenced backscatter data has not commenced before the turn of the millennia. Results of both approaches offer useful comparison and are matter for future discussions. Grab data support statistical model approaches, in which grain-size fraction are modeled as continuous variables with the resolution and accuracy highly depending on the method used. Sidescan sonar data provide very precise high-resolution maps. However, the backscatter data require categorization into distinct sediment classes which is a difficult, and time-consuming task since automated standard-classification routines usually fail and (sometimes subjective) expert knowledge must be implemented. This study presents advantages and disadvantages of the different approaches (grab-sample data analysis, modelling, acoustic-data analysis) and how these complement each other in a case study undertaken in the Sylter Outer Reef, a protected (FFH) area in the German Bight. The southwest of the study area is characterized by homogeneous distributed muddy fine sands while the northeastern part is shaped by multiple Pleistocene glacial advances and retreats. As a result, the seafloor is characterized by heterogeneous glacial deposits (medium sand to gravel size with stones) partly covered with Holocene marine sediments. We discuss the different options, including automatic (unsupervised classification) and manual classification of back scatter data (following a national mapping program in Germany; cf. Propp et al.) or statistical models (based on principle component analysis). Modelling is able to provide the distribution of ‘unclassified fractions', but may not reach the accuracy and resolution of hydroacoustic methods (particularly in areas with heterogeneous sediment distribution). Finally we discuss the value that can be added when merging the three methods.