Project Background
Managing UK seas to maintain clean, healthy, safe, productive and biologically diverse, requires comprehensive information about the state of our seas. Accurate prediction/forecast of spatiotemporal phenomena such as spring blooms, harmful algal blooms from environmental parameters such as temperature, salinity, chlorophyll, upwelling, and rainfall indices allows to better understand ecosystem variability. This requires combination of numerical models and observational techniques. There exist multiple methods including fixed observation points (for e.g. L4 station), ship based oceanography and remote sensing using satellites to aid. Along with all these still being used, the use of Autonomous Underwater Vehicles (AUVs) for oceanographic sampling of such environmental parameters has multiple benefits: (i) accessibility to a wider 3 dimensional region at a greater scale (at required depth), (ii) gathering of high quality in-situ dynamic information/data and (iii) greater endurance. A key aspect in the use of AUVs depends on the locations at which it needs to take measurements. Ensuring the traverse of the AUVs for gathering data along the most probable regions and how the gathered information assimilated to the existing short term forecast models to reduce uncertainty in prediction are key to a successful programme. The proposed project aims to address some of the challenges in this line of research. Beyond the prediction/monitoring of presence of chlorophyll indicating the algal blooms, the concept is expandable to other environmental problems such as deoxygenation, acidification and eutrophication. Plymouth Marine Laboratory and University of Exeter has been collaborating in this area of research since 2018.
Project Aims and Methods
The envisaged main components of the research are: a short term forecast model, an iterative algorithm to plan the paths for the AUVs, data based validation plans. The evolution of the features over the 3 dimensional space follows a complex spatiotemporal dynamics. This dynamics and internal principles associated with the phenomenon needs to be learned with in the short term forecast model. Machine learning based methods, which use available historical data, have often found to be helpful for making predictions with associated level of confidence bound. Gaussian process based method is an interesting candidate. The research candidate will carry out an initial literature review on different useful available methods for spatiotemporal sequence forecasting, assess the advantages and disadvantages of the methods from the perspective of current problem listed above. Based on the trade off on computational complexity and accuracy, the research student will develop advanced methods and codes for generating appropriate short term spatiotemporal sequence forecast model of the spring blooms over a region (regular and irregular grids cab be considered). The research student will receive support on machine learning and numerical model development from University of Exeter. The research student will have access to all the relevant historical data set, and continuous support based on the domain expertise on the physical-biogeochemical spatiotemporal features from the PML team. The student will have the opportunity to test the tools/codes for generating the spatiotemporal sequence forecast model with other sets of environmental parameters such as oxygen, pH, etc. Based on the outcome of the mathematical model, an iterative algorithm will be developed by the research student to find a trajectory for the AUVs that maximises an information quality metric for e.g. this could be variance reduction, information gain or the mutual information. The research student will have the opportunity to work at Unmanned Systems Control & Autonomy Lab at CFCM. The research student can benefit from the opportunity the lab facility provides to test some of the algorithms in an indoor emulated environments using existing available unmanned robotic systems, and the realistic testing depends entirely on the preference of the research student (as it is not a primary goal of the project as stated above). Further different constraints such as fuel, energy, time, no-go zones, and bathymetry constraints will also be considered in a step by step manner. A balance of exploration and exploitation of the search space will also be considered. The data gathered while traversing through the proposed trajectory will be made use to assimilate with the forecast model to enhance the confidence in predictions (or reduce the levels of uncertainty). A data based validation of the proposed probabilistic short term forecast model with the iterative algorithms to plan the paths for AUVs will be carried out and demonstrated to wider community.
Candidate requirements
Indicate if specific skills or disciplines are required. Mechanical/Electrical/Mathematics/Computer Science Coding of MATLAB/Simulink necessary (Python expertise also useful) Interest in AUVs, Oceanography, Machine Learning
Project partners
Highlight the exciting research collaborations, resources and student opportunities provided by the GW4+ Research Organisations, GW4 Universities, CASE partners and Collaborative Partners on the project
Training
Describe any specialist training, fieldwork and overseas opportunities.
Useful links
For information relating to the research project please contact the lead Supervisor via [Email Address Removed] https://emps.exeter.ac.uk/engineering/staff/pmp204
How to apply
In order to formally apply for the PhD Project you will need to go to the following web page.
https://www.exeter.ac.uk/study/funding/award/?id=4242
The closing date for applications is 1600 hours GMT on Friday 10th January 2022.
Interviews will be held between 28th February and 4th March 2022.
If you have any general enquiries about the application process please email [Email Address Removed] or phone: 0300 555 60 60 (UK callers) or +44 (0) 1392 723044 (EU/International callers). Project-specific queries should be directed to the main supervisor
Continue with Facebook
