Green algal mat satellite detection and comparison of satellite estimates with field data for validation

In this update,  Philippe and Maël are validating the algal mat spatial detection toolkit. From their base in ARGANS France, they are gathering data from key partners to ground truth and  validate the images from the algal mats satellite detection.

The interest in this piece is growing from national regulatory groups as well as from academia, research, industry, citizen scientist, or the naturally curious.

Mapping algal mat coverage can tell us the extend of intertidal habitats impacted by eutrophication and the management measures that should be considered for achieving good ecological status (GES) of these transitional and coastal areas.

Healthy coastal and transitional areas support a good balance of ecosystems that includes certain macro algae, seagrass, saltmarshes, and shellfish that functionally removes nutrients (nitrates, phosphates, etc.) and capture carbon to mitigate adverse climate changes.

Knowing or predicting the locations of excessive algal mat coverage, and what extend or conditions it is in, will help management groups to identify priority areas for restoration or additional strategic protection.

Across southern England and northern France, colleagues are working to detect and map algal mat coverage as well as to model the hydrodynamic processes that are supporting these ecosystems. Ground truthing by using field surveys, drone data and drop cameras has been promising. Together these will speed up the process of creating accurate maps.

This work is creating answers that will support our government’s actions to halt and reverse biodiversity loss by 2030 and to protect 30% of the world’s land and ocean by the same date. In the longer term, nutrient remediation as well as protecting and restoring the ecosystems of transitional and coastal areas so that they can deliver benefits for people, nature, and the climate.

The maps will provide an important evidence baseline for years to come.

Validation through ground truthing is being done using our latest artificial intelligence techniques and comparative models (fuzzy classification).

On the validation data for July and August 2021 there is 790 974 m2 covered by algae, this includes 721 638 m2 with a 0.4 threshold, and 803 952 m2 with a 0.3 threshold within the fuzzy classification.

This fuzzy classification programmes have difficulties to detect shiny algae. But the results stay coherent with the validation data as shown in the following images for Ledano estuary and Langstone harbour.

RGB PlanetScope Image of July 22, 2021

Fuzzy classification Ledano (0.4 threshold), PlanetScope Image of July 22, 2021

Validation data August 2021 (source CEVA)

Additional fuzzy classification and validation are taking place. During August 2021, the validation data included and area of 790 974 m2 covered by algae (493 623 m2 with a 0.4 threshold, and 721 674 m2 with a 0.1 threshold). However, from observation, it seems that there are less algae in the areas under consideration, as shown where the classification doesn’t detect some pixels covered by a thin algae layer. Some of these areas are confirmed as water channels from the ground truthing data.

RGB PlanetScope Image of August 8, 2021

Fuzzy classification Ledano (0.4 threshold), PlanetScope Image of August 8, 2021

Validation data August 2021 (source CEVA)

The validation journey also includes using ground truth field data that were collected on the 14th , 15th , 20th and 21st of September 2022 by  partners at the Institute of Marine Sciences at the University of Portsmouth (UoP).

Dominant algae location, Fuzzy classification Langstone Harbour 17/09/2022, PlanetScope

Ground truth algae position taken (source: Institute of Marine Sciences from the University of Portsmouth)

Dominant algae location, Fuzzy classification Langstone Harbour 18/09/2022, Sentinel-2

Ground truth algae position taken (source: Institute of Marine Sciences from the University of Portsmouth)

Dominant algae location, Fuzzy classification Langstone Harbour 18/09/2022, Sentinel-2

Ground truth algae position taken (source: Institute of Marine Sciences from the University of Portsmouth)

For the three different classifications, most locations containing algae in the ground truth data, were identified, however, some locations containing macro algae according to the ground truth data are not being classified as macro algae, and this will be the focus of refining the next stages of this work to ensure accuracy of the AI tool and classification model for algal mat coverage.

Our aspirations are to have these accurate maps accessible from the Natural England Open Data Geoportal (naturalengland-defra.opendata.arcgis.com/) and IGN France (data.gouv.fr portal) for all public access and use. IGN maps are France’s equivalent of the UK’s Ordnance Survey.

Overall, this work is helping to reduce evidence gaps for ecosystem management of our coastal and transitional waters.

Our ambitions are that this evidence will feed into global and regional decision support tools which are being used by stakeholders and decision makers.

This work continues in 2023 to better identify saltmarshes, as well as seagrass, and intertidal substrates (sand or mud).

Find out more about RaNTrans by signing up to our Newsletter (rantransproject.com/#signup) visiting the website (rantransproject.com) and following us on twitter (@ProjectRaNTrans)