Mapping Arctic benthic community distribution patterns in relation to environmental drivers, using the model-based multi-species ’Regions of Common Profile’ (RCP) approach

• Main Author: Pantiukhin, Dmitrii
• Format: Thesis
• Language: unknown
• Published: M.Sc. Program for Polar and Marine Science POMOR 2019
• Subjects: Arctic; Arctic Ocean; Barents Sea; Kara Sea; Laptev Sea; Climate change; laptev
• Online Access: https://epic.awi.de/id/eprint/50558/; https://hdl.handle.net/10013/epic.459fdb72-e7b3-47c3-9f38-44efe7bfda76

The last report of the Intergovernmental Panel on Climate Change (IPCC) with high confidence postulated the changes of the marine ecoregions around the world. These shifts are particularly pronounced in polar regions, as they are expected to accumulate the impacts of climate change. The organisms that are most likely to be affected are benthic communities due to the predicted weakening of pelago-benthic coupling. In order to understand and predict these changes, a continuous classification of marine ecoregions is essential. However, the majority of modern classifications are based on qualitative assumptions of scientists or scientific groups, and, unfortunately, these are not suitable for dynamically developing systems. Moreover, delineating the boundaries of ecoregions remains a statistically challenging task as well. However, new machine learning methods, which are currently being actively applied in many fields of science, could potentially solve the problem of ecoregional classification. Such methods are particularly effective in studying the patterns of complex interactions of biotic and abiotic systems. In this study, multi-species data of macrobenthic species are grouped in relation to environmental variables via the novel unsupervised machine learning method of ‘Regions of common profile modelling with mixtures of experts’ (RCP). A RCP is a probabilistic area in which species share the same probability of occurrence, given certain environmental conditions. This method has many advantages, such as the possibility of predicting regions in a probabilistic form, handling of different datasets, performing forward model selection and prediction of regions under environmental changes. The RCP analysis was performed with an existing extensive dataset of biological data on macrobenthic species encompassing the Eurasian Arctic. All environmental data potentially affecting the distribution of macrobenthic communities were collected and processed in the framework of this study. The results provided fine-scale map of the distribution of the macrobenthic communities over the large spatial area of the Eurasian Arctic shelf. Four RCPs were identified over the study area (Barents Sea, Abyssal Arctic, Coastal Kara Sea and Laptev Sea, Transitional High Arctic). The results consider to have high congruence with proposed theoretical approaches and should bring deeper insight on the structures of the macrobenthic communities, as well as establishing a continuous modelling survey of macrobenthic bioregions in the Arctic Ocean. The developed model is proposed to be extrapolated over the entire Arctic Ocean, established on a continuous basis and used for systematic conservation planning.