Automatic Segregation of Pelagic Habitats

It remains difficult to segregate pelagic habitats since structuring processes are dynamic on a wide range of scales and clear boundaries in the open ocean non-existent. However, improve our knowledge about existing ecological niches shaping enormous diversity marine plankton, we need better understanding driving forces behind plankton patchiness. Here describe new machine-learning method detect quantify based hydrographic measurements. An Autoencoder learns two-dimensional, meaningful representations higher-dimensional micro-habitats, which characterized by variety biotic abiotic measurements from high-speed ROTV. Subsequently, apply density-based clustering algorithm group similar micro-habitats into associated macro-habitats German Bight North Sea. Three distinct macro-habitats, “surface mixed layer,” “bottom an exceptionally “productive layer” consistently identified, each with its community. We provide evidence that model detects relevant features like doming thermocline within Offshore Wind Farm or presence tidal mixing front.