Saltmarsh Restoration

Under the ESA LaunchPad Initiative, ARGANS has developed a methodology of identifying current and historic extents of UK saltmarsh ecosystems using Earth Observation data.

Click the title to keep on reading...

Background

Saltmarsh ecosystems provide a valuable asset to UK coastlines as a natural flood defence, a combat to coastal erosion and benefits to the global carbon issues. However UK saltmarsh ecosystems are under threat from sea level rise and land reclamation. There is an estimated 32,462 ha of saltmarsh within England alone which needs to be monitored and protected. This study evaluates how Remote Sensing and Earth Observation can play it's role in surveying these important ecosystems.

Methodology

To help assess the current extent of UK saltmarsh, satellite imagery from ESA's Copernicus Sentinel-2, USGS Landsat Missions and AIRBUS Pleiades were acquired giving a range of different spatial and spectral resolutions with a historic catalogue to help assess change of saltmarsh extent through time. 

Our first task was to assess the best methodology for identifying saltmarsh extents using earth observation data. Land cover classification provides the obvious and best solution for this task and by incorporating both unsupervised and supervised classification methods we were able to map, with high accuracy, the current and historical extent of UK saltmarshes and apply this methodology to historical imagery . 

By moving forward with machine learning algorithms, ARGANS was able to produce classification maps for Sentinel-2, Landsat and Pleiades dating back as far as each sensor would allow in a very short time achieving high accuracy. ARGANS is now aiming to improve classification accuracy with more detailed ground truth data while making the methodology more robust to be applied to the entire UK coastline to help monitor and assess current and historic changes of the saltmarsh ecosystem. Machine learning algorithms rely on accurate input data to correctly interpret the results therefore there is a high need for quality data sets.

Land Cover Classification of Sentinel-2 imagery

This project nicely shows how earth observation data can highlight and map saltmarsh extent but why is earth observation data important to this?

Earth observation data provides a methodology of assessing and monitoring hard to reach and also dangerous locations from the safety of a desk. We can observe, measure and manipulate the data to show how historic and current trends in saltmarsh ecosystems may develop into the future, providing models of future growth/decline, highlighting areas of concern to ensure that these valuable environments thrive in today's climatic conditions.