Applications | ATH9010 Enables Water Quality and Membrane Fouling Monitoring for Desalination During Algal Blooms

Harmful algal blooms (HABs) bring far more than discoloured seawater and ecological risks. For seawater desalination plants, algal cells and their released organic matter and sticky polymers can accelerate reverse osmosis membrane fouling, increase pretreatment burdens, and raise operational costs. A recent study published in Water Research demonstrates that UAV‑based hyperspectral imaging combined with deep learning can transform traditionally lab‑dependent membrane fouling‑related water quality indicators into spatially continuous distribution maps – offering a new technological approach for desalination intake monitoring.

01  Why Do Algal Blooms Threaten Seawater Desalination Systems?

With climate change and increased nutrient inputs to coastal waters, the frequency and duration of harmful algal blooms (HABs) are on the rise.

During bloom events, large numbers of algal cells, extracellular organic matter (EOM), and transparent exopolymer particles (TEP) enter seawater desalination intake areas, easily causing filter unit clogging, reverse osmosis membrane flux decline, increased cleaning frequency, and even compromised water production stability.

In seawater reverse osmosis (SWRO) systems, the following three indicators are closely linked to membrane fouling risk:

• SDI (Silt Density Index): reflects the potential degree of membrane clogging by particulate matter in the water;
• TOC (Total Organic Carbon): characterises the overall level of organic matter in the water – an important reference for organic fouling;
• TEP (Transparent Exopolymer Particles): produced by algae and microorganisms, highly adhesive, and serving as important precursors to biofilm and organic membrane fouling.

Traditional methods rely mainly on manual sampling, water collection, and laboratory analysis. While accurate, the limited number of monitoring points makes it difficult to capture rapid water quality changes and localised pollution hotspots around intakes.
 

02  What New Method Did the Study Propose?

A research team from a Korean university published a paper in Water Research titled "UAV‑based hyperspectral imaging and deep learning for mapping fouling‑related water quality indicators in seawater desalination during HABs" , establishing a monitoring framework integrating "UAV hyperspectral + synchronous sampling + deep learning" .

The researchers conducted UAV aerial surveys before, during, and after algal bloom events, acquiring hyperspectral images in the 400–1000 nm range. Meanwhile, water samples were collected at corresponding locations for laboratory analysis to obtain measured data for SDI, TOC, and TEP.

For model construction, the team first used Partial Least Squares Regression (PLSR) to screen important wavelengths correlated with the target indicators, then weighted the spectral features using VIP (Variable Importance in Projection) scores, and finally fed the weighted spectral data into a one‑dimensional Convolutional Neural Network (1D‑CNN) to extract nonlinear relationships between spectral information and membrane fouling indicators.

The overall technical workflow can be summarised as:

UAV hyperspectral aerial survey → Synchronous water sampling → Spectral preprocessing → PLSR/VIP feature selection → 1D‑CNN modelling → Pixel‑level inversion → Fouling risk spatial mapping.

03  From "Point‑Based Measurement" to "Area‑Wide Water Monitoring"

The study results showed that the VIP‑weighted 1D‑CNN model achieved varying degrees of predictive performance for the three indicators:
 

Among them, SDI showed the most outstanding inversion performance, while TOC also demonstrated good predictive capability. TEP, due to its complex formation mechanisms and pronounced spatial variability, proved more challenging to predict, but the model still showed spatial identification potential.

More importantly, the hyperspectral inversion results revealed water quality gradients and local hotspots that were not detectable with conventional RGB imagery. Even after water treatment, some areas may still carry residual risks from organic matter and biopolymers.

This means that the value of UAV hyperspectral imaging extends beyond "detecting algal blooms" – it can further help answer:

• Which areas have higher membrane fouling risk?
• Are pollutants spreading toward the water intake?
• Do high‑risk residues remain after pretreatment?
• Should intake location, depth, or treatment processes be adjusted?

Hyperspectral water quality monitoring is thus expanding from ecological environment surveys to operational management and risk decision‑making for desalination facilities.

04  Building an Integrated Air‑Ground‑Model Monitoring Solution

Optosky

For seawater desalination intakes, harmful algal blooms, and coastal water quality monitoring, Optosky offers a comprehensive technology solution centred on UAV‑based hyperspectral imaging.

Rapid UAV Hyperspectral Aerial Surveys

Equipped with Optosky's UAV hyperspectral imaging system, users can conduct mobile surveys around intakes, coastal waters, aquaculture zones, and algae‑prone areas, continuously acquiring water spectral and spatial location data.

Compared with conventional visible‑light or multispectral cameras, hyperspectral imaging records hundreds to thousands of contiguous bands, providing a more complete data foundation for identifying spectral differences related to chlorophyll, suspended solids, organic matter, and algae.

For projects emphasising rapid on‑site assessment, the ATH9030 real‑time UAV hyperspectral system can also be deployed, allowing operators to view imagery and spectral information from the remote control terminal, assisting field personnel in promptly identifying anomalous areas and sampling locations.

Ground Sampling and Spectral Synchronous Validation

Using field spectroradiometers, water quality testing equipment, and standardised sampling procedures, the system establishes correlations between image pixels and indicators such as SDI, TOC, TEP, chlorophyll‑a, suspended solids, and turbidity – providing reliable samples for model training and regional calibration.

Hyperspectral Data Analysis and Intelligent Modelling

Employing spectral preprocessing, feature band selection, PLSR, Random Forest, Support Vector Machines, and deep learning algorithms, the system builds inversion models tailored to specific water bodies and monitoring objectives, ultimately generating:

• Algal bloom distribution maps
• SDI membrane fouling risk maps
• TOC and organic matter anomaly distribution maps
• TEP high‑risk zone maps
• Water intake safety level maps
• Multi‑temporal water quality change maps

From Research Data to Operational Decision‑Making

Through periodic or emergency UAV surveys, hyperspectral results can be integrated with intake scheduling, pretreatment processes, and membrane system operational data – providing desalination plants with spatially targeted decision support.

When a high‑risk water mass is detected approaching the intake area, plant managers can proactively increase monitoring frequency, optimise coagulation and filtration parameters, or adjust intake strategies – transforming passive treatment into active early warning.

This study highlights an emerging direction worth watching: UAV‑based hyperspectral imaging is moving beyond algal bloom identification and routine water quality inversion toward membrane fouling risk monitoring and industrial process management for seawater desalination.

Looking ahead, as UAV platforms, real‑time hyperspectral imaging, and artificial intelligence algorithms further converge, water quality monitoring will no longer be confined to a limited number of sampling points – it will evolve into continuous risk maps covering entire intake areas.

Optosky will continue to advance the application of hyperspectral technology in marine environments, coastal water quality, harmful algal blooms, seawater desalination, and smart water management – providing research institutions, environmental protection agencies, and water treatment enterprises with integrated solutions spanning equipment acquisition, ground validation, and intelligent analysis.

For more information, please contact:

Email: optoskyphotonics@gmail.com

Web: www.optosky.net