Remote sensing and DEEP learning for early warning of WATER hazards (DeepWater)

Current satellite sensors allow global monitoring of water resources at an unprecedented optical and spatial resolution, paving the way for operational monitoring of potential hazards from space. Of particular interest is the potential value of satellite observations during and following episodic events such as heavy wind and rain, which may lead to harmful algal blooms, sewage overflow and poor visibility.

Aims and Methods
The overarching objectives of this project are to develop image segmentation and object-based satellite image processing techniques for cases where water quality issues are evident. A successful monitoring solution would provide hazard warning information to the affected companies and end users.

The first such case concerns mapping river plumes and their ‘sphere of influence’. Tracing river plumes from their source to the furthest extent, visible as an optical and/or radar signature, directly informs commercial (e.g. aquaculture) and recreational (e.g. diving, fishing, surfing) use of potential risks in the event of strong storm runoff.

The second case focuses on mapping of dynamic features such as potentially harmful algal and cyanobacterial blooms, and relatively stable features such as shallow areas (bottom visibility), and floating vegetation, in coastal and inland water systems. Object oriented mapping would classify these optically dominant structures and create a novel approach to spatial binning on satellite imagery.

The project will seek to tackle the challenge by exploiting the new generation of high resolution satellite imagery (Sentinel-1 (radar), Sentinel-2 (optical), and Landsat optical missions) and numerous advanced computing techniques which have not yet been applied in remote sensing of water bodies [1], e.g. deep learning, sub-pixel level endmember extraction methods, local parallelisation and distributed processing techniques etc.

Candidate
The project would suit a student with a first degree in Computer Science and a desire to develop a range of skills such as remote sensing, machine learning and computer vision techniques. A candidate with the background of remote sensing and satellite image processing is also encouraged to apply.

Case Award
This is a CASE award. The student will spend a minimum of 3 months at the CASE Partner, Pixalytics (Satellite imagery expert). Dr Lavender is the Managing Director of Pixalytics Ltd, a Trustee of SAHFOS, Chairman of the British Association of Remote Sensing Companies and Honorary Reader of Geomatics at Plymouth University. She has 20+ years research experience in the fields of this project.

Training
The funded student will benefit from collaboration and training opportunities through the EU H2020 EOMORES project and by working closely with the CASE partner. The student will attend a training course on Computer Vision and Machine Learning in the first year, usually hosted by the University of Exeter and IEEE Computer Vision Society. Dr Simis will guide them through his training course on Aquatic Optics, previously taught at the University of Helsinki, and relevant chapters of the book ‘Light and Photosynthesis in Aquatic Ecosystems’ by JTO Kirk (1994). The student will be notified of summer school opportunities organized by the International Ocean Colour Coordinating Group (IOCCG), and be encouraged to join the Remote Sensing and Photogrammetric Society (RSPSoc) Wavelength group that supports remote sensing students and early career professionals.

More information on the application process can be found here: http://nercgw4plus.ac.uk/research-themes/prospective-students/