SAIC Plankton Predict

The Scottish salmon sector is being negatively impacted by an apparent increase in blooms of
harmful plankton.
Early warning helps farmers respond with management practices to mitigate the impacts of
these events on gill health. Typically, this is achieved by collection of seawater samples, with
microscope plankton analysis. However, such monitoring acts only as a “nowcast”, and often
does not provide sufficient time to coordinate a mitigation response.
To address this, within previous SAIC-funded projects collectively called “SAIC-HABs” we have:
1) Developed approaches for standardisation and sharing of farm-collected microscope counts
of harmful algal (HAB) taxa and their rapid sharing between industry partners on the
www.HABreports.org website.
2) Deployed an automated plankton counter (the Imaging FlowCytoBot, IFCB) at a Shetland
salmon farm. This allows cell imaging and AI-based identification, with near real-time
(every 20 minutes) uploaded to www.HABreports.org for public access.
3) Developed an unstructured grid oceanographic model (WeStCOMS) to produce short-term
(~ 5 days) prediction of harmful bloom development when initiated with cell counts from
industry or IFCB.
4) Developed the NORSCOM model that extends WeStCOMS from its original west coast
domain to include the important aquaculture region of the Shetland and Orkney Isles.
Our new Plankton-Predict project will build on the SAIC-HABs developments above to provide
micro-jellyfish monitoring, true HAB early warning and water column profiling at farm sites.
Specifically, Plankton-Predict will:
1) Develop a bespoke holographic imaging system to detect harmful micro-jellyfish in-situ,
providing better identification and early warning of these harmful organisms in real-time.
2) Evaluate the added benefit of depth profile monitoring the water column for harmful plankton
taxa and environmental conditions at aquaculture sites.
3) Enhance our model-based bloom predictions using statistical-based ensemble modelling that can
provide a probability of bloom appearance in advance of occurrence.

Rationale:
Micro-jellyfish are an important harmful component of the plankton that are not easily
enumerated by microscopy or IFCB. Parallel to a UKRI-funded project, we shall develop a highly
customised holographic micro-jellyfish detection system that will be deployed at an aquaculture
site or sites.
Currently, plankton monitoring is undertaken at either a fixed depth (bottle or IFCB) or integrated
across the photic zone (net or tube). However, in the typical summer conditions of Scottish sea
lochs and voes stratification results in “thin layers” of potentially harmful plankton and
detrimental environmental conditions (e.g. depleted dissolved oxygen) at specific depths.
Deploying an IFCB and environmental sensors (temperature, salinity, pressure, oxygen,
chlorophyll) on a computer-controlled winch system will allow controlled water column profiling
to detect these thin layers, providing better management information to the aquaculture industry.
Profiling will provide enhanced understanding of environmental control of blooms, strengthening
expert risk assessment.
Current harmful plankton alerts rely on the identification of a bloom at an aquaculture site to
initiate a WeStCOMS/NORSCOM model forecast of subsequent dispersion of the bloom. Bayesian
and machine learning ensemble-modelling approaches (recently developed using Scottish shellfish
farm monitoring data) offer the potential for probability estimates of HAB development at salmon
farms, providing warning of events before they occur

SAIC
Industry Contribution