Monitoring coastal vegetation dynamics from space


The aim of this project is to improve our ability to monitor the changing condition of natural coastal erosion and flood defences by developing automated approaches for the identification, and characterisation of coastal biogeomorphic systems from satellite remote sensing data using machine learning based approaches.

Low-lying coastal areas are some of the most vulnerable to the effects of sea level rise. Communities, business, infrastructure and ecosystems at the coast are threatened by hazards such as erosion and flooding during storms worldwide. These hazards are expected to become more frequent and more severe in the coming decades with anthropogenic climate change1, resulting in escalating multi-sector damage costs. In many locations, the natural shoreline configuration and topography provides protection against such damage, particularly when there are dune or saltmarsh systems at the coast. They provide valuable ecosystem services as natural barriers or buffers against wave energy and high water levels, trap sediment, encourage carbon storage and support biodiversity2. For example, in Scotland, it is estimated that £14.5 Billion of assets are protected from the risks of coastal erosion and flooding by natural coastal defences3. Sea level rise is expected to almost double the number of residential properties at risk from coastal flooding by 2080 in Scotland (28,000 to 55,000. SEPA National Flood Risk Assessment 2018), these are underestimates as they do not consider erosion-enhanced flooding. There is a clear management need to better monitor the function of these natural coastal protection systems in the light of joint nature and climate emergencies, particularly given future anticipated coastal change, and the potential for detecting thresholds in the system so that we can identify potential tipping points for management intervention and adaptation.

The health and function of these biogeomorphic systems is vital to coastal protection in our warming world, and understanding their dynamics is key for implementing and assessing Nature-based Solutions (NBS) approaches to maintaining resilient coasts, mitigating and adapting to climate change4. In particular, how these systems respond to the effects of anthropogenic climate change at interacting temporal and spatial scales will be key to understanding their future protective capacity at the coast. Remote sensing approaches provide the opportunity to monitor our coastal zones at local to global scale at regular intervals in time5,6.

This project aims to develop automated approaches to the identification, classification and monitoring of coastal vegetation dynamics and coastal geomorphology at national scale in order to quantify the current and future extent and health of these natural assets. The specific objectives will be to:
• Automate the identification and mapping of coastal vegetation dynamics from satellite imagery, verified by manual and ground-based observations.
• Investigate variation in the health and extent of natural coastal defences across single storm event to seasonal to multiannual timescales to identify natural assets most at risk.
• Develop and apply approaches to modelling Nature-based interventions to provide change intelligence to managers to maximise future coastal resilience and blue carbon interventions.


This project will focus on the use of remote sensing to identify and monitor vegetated coastal environments across a range of spatial and temporal scales. Identification and monitoring of coastal vegetation at coarse resolution over decadal timescales (1985-present)7 will take place using freely available Landsat and Sentinel platforms. PlanetScope satellite data will also be used as it allows for more detailed mapping and a higher temporal and spatial resolution, ideal for event-based analysis8. Consideration will also be given to the substantial high-resolution aerial image back-catalogue over the last twenty years licensed by NatureScot.

In the first phase of the project, the student will develop routines for automatic identification and mapping of coastal vegetation zones. This will begin with manual mapping of coastal vegetation zones using field-based mapping including the collection of aerial imagery and topography using an unmanned aerial vehicle/LiDAR system. Machine Learning (ML) and Deep Learning (DL) approaches routinely applied to land use land cover mapping9, such as Random Forests and Convolutional Neural Networks, supported by training datasets developed through manual mapping will then allow automated classification of coastal vegetation. Edge detection routines will allow important boundaries such as the dune/beach transition to be monitored accurately8.

The second phase of the project will focus on the automation and roll out of monitoring coastal vegetation over a variety of time and space scales. Building on the first phase and leveraging the Dynamic Coast evidence base, software systems will be developed to provide coastal vegetation monitoring at national scale for Scotland for both saltmarsh and dune environments using freely available satellite data (Sentinel) and the student will investigate long-term changes for the duration of the satellite record (1985-present). Shorter-term, high-resolution investigations will be targeted at sites thought to be most vulnerable, particularly areas subjected to coastal erosion and/or flooding events associated with storms, preferably that take place during the project (so as to allow pre- and post- storm field surveys to be conducted).

Finally, by combining information from coastal vegetation change, topography and climate projections, the project will use numerical modelling identify locations where NBS that promote the health of dunes and saltmarshes are likely to be most effective in promoting flood and erosion protection services, as climate change increasingly threaten these natural coastal defences.

Project Timeline

Year 1

Knowledge development in coastal biogemorphology and research philosophy, alongside training in specific techniques and methods required for the project. The student will develop and refine the initial scientific problem informed by reviewing existing literature on coastal monitoring using remote sensing methods, as training in the preparation of a research paper. The student will learn how to download, manipulate, and analyse satellite imagery to extract information about coastal vegetation. Assisted by Dynamic Coast, NatureScot and partners the student will conduct fieldwork at targeted sites with dunes and/or saltmarsh coasts. The student will develop skills in machine learning for image classification. Attendance at the BSG Windsor workshop and international remote sensing training course.

Year 2

Development and application of procedures for monitoring coastal and vegetation change at a local and a national scale. Learn how to perform validation and calibration of image classification approaches. Consider the effects of different products in terms of timespan and image resolution. Consider approaches to numerical modelling of biogeomorphic coastal systems for predicting future trajectory. Assisted by Dynamic Coast, NatureScot and partners the student will conduct fieldwork at targeted sites with dunes and/or saltmarsh coasts. Attendance and presentation of preliminary results at a UK conference (e.g. British Society for Geomorphology).

Year 3

Develop framework for continuous monitoring of coastal vegetation. Finalise results and prepare papers for publication. Attendance and presentation at an international conference (e.g. EGU or AGU), writing up results as drafts for academic publications.

Year 3.5

Year 3.5: Finalise results, prepare papers for publication, write and submit thesis.

& Skills

The student will be trained by leading experts in coastal erosion monitoring and automated image analysis. The student will receive training in customising and automating GIS and appropriate computer programming languages (e.g. Python, C++, Matlab) required to develop and deploy algorithms to analyse satellite data and perform automated image classification with machine learning/deep learning. They will learn how to handle and analyse large environmental datasets. The student will have the option to be trained in running coastal evolution models. The student will also receive training in project management academic writing, writing funding proposals. The student will also gain training and experience in sharing their results with government agencies and local authorities. The student will benefit from gaining exposure to the extensive non-academic networks the supervisory team has through the Dynamic Coast project and liaison with NatureScot regarding opportunities for nature conservation site management, and, to assist with sharing the implications of their results for coastal managers.

The student will emerge from the PhD process with skills making them highly suited to a career in the Environmental Sciences, including the ability to manipulate and interpret large datasets, and conduct numerical modelling. There are obvious career paths in natural hazards and land management, for example, as well as research.

References & further reading

1. IPCC. Summary for Policymakers. Clim. Chang. 2021 Phys. Sci. Basis. Contrib. Work. Gr. I to Sixth Assess. Rep. Intergov. Panel Clim. Chang. (2021).
2. Kirwan, M. L., Temmerman, S., Skeehan, E. E., Guntenspergen, G. R. & Fagherazzi, S. Overestimation of marsh vulnerability to sea level rise. Nat. Clim. Chang. 6, 253–260 (2016).
3. Rennie, A. F. et al. Dynamic Coast Research Summary. (2021).
4. Slinger, J., Stive, M. & Luijendijk, A. Nature-based solutions for coastal engineering andmanagement. Water (Switzerland) 13, 3–7 (2021).
5. Fitton, J. M., Rennie, A. F., Hansom, J. D. & Muir, F. M. E. Remotely sensed mapping of the intertidal zone: A Sentinel-2 and Google Earth Engine methodology. Remote Sens. Appl. Soc. Environ. 22, 100499 (2021).
6. Vos, K., Splinter, K. D., Harley, M. D., Simmons, J. A. & Turner, I. L. CoastSat: A Google Earth Engine-enabled Python toolkit to extract shorelines from publicly available satellite imagery. Environ. Model. Softw. 122, 104528 (2019).
7. Jia, M. et al. Rapid, robust, and automated mapping of tidal flats in China using time series Sentinel-2 images and Google Earth Engine. Remote Sens. Environ. 255, 112285 (2021).
8. Rogers, M. S. J., Bithell, M., Brooks, S. M. & Spencer, T. VEdge_Detector: automated coastal vegetation edge detection using a convolutional neural network. Int. J. Remote Sens. 42, 4809–4839 (2021).
9. Maxwell, A. E., Warner, T. A. & Fang, F. Implementation of machine-learning classification in remote sensing: An applied review. Int. J. Remote Sens. 39, 2784–2817 (2018).

Further Information

For more details contact Martin Hurst, School of Geographical and Earth Sciences, University of Glasgow, University Avenue, Glasgow, G12 8QQ.
Tel: +44 (0)141 330 2326;

Apply Now