Restored saltmarsh trajectories; Restored SMART; machine learning evaluation of in situ saltmarsh restoration

* Please note that this PhD will be hosted at University of Reading*

Saltmarsh habitat provides key ecosystem services such as carbon storage and flood defence through attenuation of wave and tidal energy. However, almost 50% of saltmarsh globally has been lost and degraded due to human activity such as land claim and erosion caused by sea level rise [1]. The restoration of saltmarsh is a major nature-based solution to the challenges associated with climate change, and is being driven by Biodiversity Net Gain policies and the growing recognition of the benefits provided.

To date, much of the focus on saltmarsh restoration has been on managed realignment. This restoration approach involves breaching of flood defences to allow tidal inundation of the terrestrial, typically agricultural, land behind. These studies have indicated that the saltmarsh in managed realignment sites is less diverse and has a lower abundance of key species. This has been associated with poor drainage because of compaction caused by the earlier agricultural land use, leading to anoxia.

In contrast, other methods of saltmarsh restoration may not have the limitations associated with managed realignment as many of them take a more localised approach and attempt to restore saltmarsh in situ. These include the use of structures such as brushwood fencing and coir rolls to reduce wave and current velocities and trap sediment, sediment recharge projects, and transplanting vegetation [2].

Despite the potential of these methods to deliver successful restoration and a biologically functioning marsh, the ecological trajectory and recovery of saltmarsh restored using these methods is not properly understood.

The proposed PhD project, Restored SMART, will provide new insights into these methods of saltmarsh restoration through the development of machine learning approaches to spatially interrogate satellite, uncrewed aerial system and other remote sensing datasets. The use of machine learning and AI techniques to evaluate coastal wetlands has emerged as a rapidly developing research area as it supports automated, near real-time assessments of change over a large spatial scale.

For example, machine learning has previously been used to classify saltmarsh community composition and structure in pre-existing marshes and has been used to detect the change from unvegetated to vegetated marsh in managed realignment sites [3].

This project will develop machine learning models applied to localised interventions to provide a new understanding of the trajectory of restoration efforts that do not rely on managed realignment. Specifically, the plant community will be classified in terms of its composition, diversity, heterogeneity and spatial structure of restored marshes, and the delivery of other services such as carbon storage. To achieve this, pre-existing datasets, including biodiversity measurements collected by the supervision team, will be utilised and a range of restoration attempts will be considered.

The temporal variability before and after restoration, and the cause and indicators of any variations, will be evaluated to provide a new understanding of the restoration trajectory. The models developed during the study will then be used to evaluate the restoration potential of other sites, informing decisions around prioritising investment and increasing the success of restoration attempts.

The research will provide an exciting and novel assessment of the in-situ saltmarsh restoration methods. Through the collaborations between the project partners and a possible placement opportunity with Natural Resources Wales, the project findings will provide the tools required to make near real-time assessments of the impacts of saltmarsh restoration interventions. The newly developed models will also allow coastal managers to make data-informed management decisions in response to variations in the restoration trajectory and recovery of the restored saltmarsh, supporting the prioritisation of interventions to maximise the ecological impact.

Training Opportunities:

A comprehensive training programme will be provided, comprising training both in applied AI and biodiversity, and transferable professional and research skills. The project includes a placement with an AI-INTERVENE project partner of between 3-18 months in duration. The student will present at national and international conferences, placing the student at the forefront of the discipline, leading to excellent future…