Background and objectives:
The Indian Ocean (IO) is unique among the tropical oceans, as it is one of the least well-understood and characterised in terms of marine ecosystems processes sustaining fisheries and productivity. The western IO is also the fastest warming part of the ocean due to climate change. The underlying mechanisms of Indian Ocean productivity and fisheries are complex, variable and inherently dependent on physical processes. Changes in the IO occurring at the basin scale, such as variations in temperature or in the strength of the reversing monsoonal winds, are likely to alter nutrient availability and impact species differently at sub-regional scales. Biogeochemical (BG) provinces have the potential of improving understanding of present and future changes in marine ecosystems. Monitoring such changes could be very useful to policy makers in developing management strategies in the IO countries where coastal populations depend on fishery resources for livelihoods and food security. However, the existing BG provinces, which have been determined globally, subdivide the IO into very large areas, which do not necessarily reflect the complexity of the environmental conditions in different parts of the basin. The aim of this research is to establish a comprehensive trans-basin characterization of the IO BG spatiotemporal distributions based on the objective identification of biophysical similarities related to pelagic fisheries abundance. There is considerable flexibility in the direction that the research may ultimately take. The PhD will explore the following key research questions:
1) How are the changing IO environmental conditions influencing the spatial distributions of the BG provinces and how will these provinces will evolve in the future?
2) How do the boundaries of these provinces shift between years and seasons? This will include examining the effects of the El-Niño Southern Oscillation, the Indian Ocean Dipole and the monsoonal winds.
3) Given that much of the IO biological productivity occurs in upwelling regions, can changes in upwelling be detected and characterised?
This PhD will investigate the applicability of unsupervised machine learning techniques to a set of historical and new EO data, numerical model outputs and fisheries records in order to capture the past, current and future of IO ecosystem productivity across the BG provinces. The nonlinear interactions of the BG provinces make unsupervised Machine Learning (ML) methods (deterministic and / or stochastic) well suited to objectively determine BG provinces. The main ML methods that will be explored are Self Organizing Maps, K-means clustering, E/M mixture of Gaussians, and variational autoencoders.
Both historical and recent satellite datasets will be exploited. These will cover ocean colour derived data, Sea Surface Temperature [SST], Sea Surface Salinity [SSS], winds, altimetry derived Sea Surface Height [SSH] and currents. The understanding derived from satellite observations will be compared to output from a high resolution (1/12°) ocean model (NEMO) that includes biogeochemical processes (MEDUSA). Using environmental factors inferred from numerical model future projections, the evolution of the BG areas will be assessed and the potential impact on fisheries.
The fisheries data that will be used include (but are not limited to) the AIS [Automatic Identification System] open source vessel tracking data which provide data on fishing effort at fine spatial and temporal resolution (https://globalfishingwatch.org/map-and-data/).
This project will enable the PhD student to develop skills in several techniques including statistics, computer programming and artificial neural network, with the support of a team with a diverse range of expertise.
The student will be based at the National Oceanography Centre (NOC) in Southampton and work closely with supervisors based at NOC and Leeds University, where the student will be registered.
Jebri F., Jacobs Z., Raitsos D.E., Srokosz M., Painter S.C., Kelly S., Roberts M., Scott L., Taylor S., Palmer M., Kizenga H., Shaghude Y., Wihsgott J. & Popova E. 2020 Interannual monsoon wind variability as a key driver of East African small pelagic fisheries, Nature Sci. Rep., 10, doi: 10.1038/s41598-020-70275-9
Kingma, D.P . and W elling, M., 2013. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114.
Razavi, A., van den Oord, A. and Vinyals, O., 2019. Generating diverse high-fidelity images with VQ-VAE-2. In Advances in Neural Information Processing Systems (pp. 14866-14876). Vancouver Reygondeau G., Longhurst A., Martinez E., Beaugrand G., Antoine D. and Maury O (2013). Dynamic biogeochemical provinces in the global ocean. Global Biogeochem. Cycles 27, 1046–1058.
Sonnewald, M., Dutkiewicz, S., Hill, C., and Forget, G. (2020). Elucidating ecological complexity: Unsupervised learning determines global marine eco-provinces. Science Advances 6, 22, eaay4740.