Researchers from the University of Stirling have developed a new remote sensor for analysing the health of water bodies using artificial intelligence (AI).
Called the 'meta-learning method', this new AI algorithm uses data from satellites to analyse more accurately and efficiently monitor the condition of freshwater sources.
The new sensor is able to track the growth of harmful algae bloom (HABs) and the toxicity to local marine life.
Bypassing the challenge
Using satellite data to monitor freshwater reservoirs is not new. For example, the Copernicus Sentinel-3 Ocean and Land Color Instrument (OLCI) can currently do this by measuring phytoplankton concentration via the optical pigment, chlorophyll-a.
However, using chlorophyll-a to retrieve this data presents several challenges, namely the diverse nature of global water.
"We have developed a method that bypasses the chlorophyll-a retrieval and enables us to estimate water health status directly from the signal measures at the remote sensor."
The team used the OLCI using a waveband configuration at 49 global locations, both at inland and nearshore waters, resulting in 567 observations.
"We have developed a method that bypasses the chlorophyll-a retrieval and enables us to estimate water health status directly from the signal measures at the remote sensor," said lead author Mortimer Werther, a PhD researcher in biological sciences at Stirling’s Faculty of Natural Sciences.
Currently, many environmental protection agencies use the 'trophic state' of water - the total weight of biomass in a water body - to determine its health.
If bodies of water go unchecked, it can lead to eutrophication, mass growth of cyanobacteria and HABs, which can be deadly to the local environment and inhabitants.
The research was funded by the European Union Horizon 2020 programme and was carried out by five different research bodies and institutions. It is the first of its kind to show that that tropic status can be learnt by machine learning algorithms from the OLCI.
Werther added: "Our method outperforms a comparable state-of-the-art approach by five-12 per cent on average…it estimates trophic status with over a 90 per cent accuracy for highly affected eutrophic and hypereutrophic waters."
While these initial results are encouraging, the full paper adds that their algorithm had lower accuracy for measuring mesotrophic waters.
The Mesotrophic waters (lakes) are priority habitats of areas of standing water that have moderate alkalinity and nutrient levels and feature a high diversity of aquatic plant and macroinvertebrate species.
AI revolutionising water monitoring
For surface water monitoring, AI holds a great deal of potential, and despite the algorithm still needing fine-tuning, these initial results will certainly bode well for the future of the technology.
Algae and HABs are becoming a growing concern, with increasing critical events happening more regularly. Earlier this year, Aquatech Online reported how Singapore’s upmarket Sentosa Cove neighbourhood experienced an algal bloom, with pink water resulting in the deaths of thousands of fish and some unpleasant odours in the district.
In 2020, a 'Red Tide' algae bloom event happened in Florida, killing over 100 manatees, 127 dolphins and 589 sea turtles, as well as thousands of fish.
Further back in 2014 in Ohio, Lake Erie was taken over by blue-green algae called Microcystis Aeruginosa, a form of cyanobacteria. The state spent $132 million on improving its water treatment plant to deal with the HAB's.
AI and machine learning are offering a solution to help better monitor our freshwater bodies, and it's learning all the time.
- Understanding algal blooms
- Surface water: our essential guide to surface water and the impact of algae
- Big data tracks harmful algal blooms
- ETRI develops drones for water algae analysis