Using remote sensing tools to study ocean health and whale strandings (SENSE CDT)

Scientific background and motivation: Causes of cetacean strandings (animals live-stranded or washed ashore dead) are poorly understood, but information gained from these events represents an important resource for understanding marine mammal health, diet, environmental contaminants, regional oceanography, social structuring and climate change [1]. Many coastal nations have marine mammal stranding networks. This is a crucial means by which strandings of local marine mammals can be monitored, providing first notice of potential marine contamination or pathogen spread. However, rapid detection is a major challenge in many areas, with resources often concentrated on activities associated with a posteriori carcass sampling and analysis, rather than on initial detection. The scale of strandings can also be hard to identify, particularly in remote places. Very High Resolution (VHR) remote sensing technology can offer a lower cost, large-scale means of monitoring whale stranding events. Potentially, this enables more rapid response times in order to identify and collect data on mass mortality events, and improves understanding of their temporal and spatial extent, particularly for inaccessible regions. This approach has been trialled, using archived satellite images to retrospectively analyse the number of whales stranded at a mass stranding event in Golfo de Penas, Chile [2]. Here, VHR images provided accurate, large-scale information on the numbers of whales stranded, and helped to resolve the timeframe over which the stranding occurred [3]. Satellites are an innovative, lower-cost way to monitor strandings in remote areas, but at present there are time-consuming steps (in particular, manual review of images) which prohibit this being used as a rapid response tool. Strandings detection is time-sensitive, because whale carcasses degrade rapidly and there is a short window of opportunity to collect biological samples and identify causes. Automation of the detection of whales on satellite imagery is therefore required to make this approach useful for large-scale monitoring.

The growing body of high-resolution earth observation data provides an opportunity to study the oceanic conditions associated with whale strandings. Stranded whales tend to be the most visible indicator of contamination events [4], so high-resolution observations of the local marine environment [5] associated with stranding events (e.g. sea-surface salinity and temperature, ocean colour and chlorophyll [6]) could point to local variations in ocean health and provide new tools for future monitoring of strandings and their causes.

Aim and objectives: The aim of this project is to automate the identification of stranded whales on satellite imagery, and to identify ocean health-related parameters associated with whale stranding events. The objectives are:

Develop an automated identification technique for stranded whales from satellite imagery;
Find relationships between local marine environmental variables and whale strandings;
Assess impact of local ecosystem health on stranding patterns and identify predictors of future strandings.
Methodology: In this project, we will develop analytical tools for identifying and classifying stranded whales on satellite imagery, using machine learning approaches (e.g. convolutional neural networks, among others), applied to existing archival images of stranded whales (including down-sampled images from aerial and drone surveys). We will then collate high-resolution remotely-sensed earth observation data from the vicinity of stranding hotspots (such as WorldView or Pleiades), to assess which environmental parameters are associated with stranding periods. The project will focus in particular on sites in Chile [2] and New Zealand [7], where strandings regularly occur. During the PhD, the student will benefit from an excellent supervisory team with expertise in whale population biology (Dr Jen Jackson); remote sensing of coastal systems (Dr Encarni Medina-Lopez); use of satellite imagery to study animals (Dr Peter Fretwell); whale strandings and New Zealand marine ecology (Dr Karen Stockin); Chilean whale populations and marine ecology (Dr Carlos Olavarría); earth observation with satellites (Dr Gwawr Jones, JNCC CASE Partner).


This PhD is part of the NERC and UK Space Agency funded Centre for Doctoral Training "SENSE": the Centre for Satellite Data in Environmental Science. SENSE will train 50 PhD students to tackle cross-disciplinary environmental problems by applying the latest data science techniques to satellite data. All our students will receive extensive training on satellite data and AI/Machine Learning, as well as attending a field course on drones, and residential courses hosted by the Satellite Applications Catapult (Harwell), and ESA (Rome). All students will experience extensive training on professional skills, including spending 3 months on an industry placement. See http://www.eo-cdt.org