Abstract:
This thesis shows the global marine ecosystems mapping at ocean surface through an unsupervised clustering analysis method with global marine environmental variables and their seasonal variations, and then applying the mapped ecosystems with other data layers to prioritize areas for planning MPAs at a global scale. It provides a quantitative approach with numerous variables to identify marine ecosystems globally, and uses the ecosystems in prioritizing where marine protected areas should be best located. In contrast to previous world ocean classifications which were mostly based on expert opinion and ad hoc management areas, this study used data analysis to map the ecosystems. First, data were normalized, and PCA removed the correlations between 20 annually averaged environmental variables and reduced their number to eight components. The components were analysed by k-means clustering, and cosine similarity as the distance function. I found that a seven clusters (mapped ecosystems) solution supported silhouette values. Some areas of the same ecosystem were divided by continents, symmetrically distributed on both sides of the equator at similar latitudes. The mapped marine areas were named as Ecosystems and were ready for further applications such as the ecosystem assessment, MPAs planning, and climate change, etc. Mapping of the seasonal distribution of global marine ecosystems can follow a similar method. To map the distribution of the ecosystems in June-August and December-February, 18 environmental variables within the two periods were collected from multiple datasets. There were 11 Ecosystems identified in each season. In addition to the features discovered in the mapping by annually averaged variables, some fragmentary parts of Ecosystems were found around oceanic islands. The little changes in tropic, the latitudinal changes at high latitudes, and the more complicated changes in other places were discovered from the seasonal variations. The seasonal mapping could enrich the understanding of global marine ecosystems from a seasonal scope, and provide a most appropriateness when seasonally epipelagic context is needed. The prioritization for MPAs planning then utilized the seven mapped Ecosystems, the Realms for species endemicity (Costello et al., 2017), the four Biomes (seagrass, kelp, mangrove, and shallow water coral reefs), and species richness with the software Zonation to map the top 30% most prioritized areas. The highly prioritized areas were generally distributed at the coastal areas, marginal seas, and oceanic islands. They were covered by the EEZs but few were covered by current protected areas. The responsible countries in the EEZs could thus expand the areas for protection following coverage of the prioritized areas. Work could be advanced through the addition of more variables with ecological criteria such as threatened species and anthropogenic disturbances included in the prioritization analysis.