Fish farm sensors: checking for measurement drift

A reassuring oxygen reading is only useful while the instrument remains reliable. Equally, a sudden alarm may reflect a genuine water-quality event, sensor fouling or a fault elsewhere in the measurement chain. For fish farms, the practical challenge is to check the instrument while protecting the stock. This guide explains how to separate those possibilities and retain the evidence needed to improve maintenance and emergency response.

Start with the question the sensor is meant to answer

A dissolved oxygen sensor reports conditions at a particular location. It cannot automatically represent every part of a production tank. Water near an inlet may differ from water that has already passed through a substantial concentration of fish. Before interpreting a reassuring number, establish which exposure the monitoring point is intended to describe and whether its location serves that purpose.

The US Geological Survey guidance by Wagner and colleagues links monitoring-site selection to knowledge of local water movement and variability. Its original setting is environmental monitoring, rather than aquaculture. Applying that reasoning to a farm is a practical adaptation. A useful site record identifies the monitoring position, depth, nearby aeration, circulation pattern and access arrangements. Moving a sensor should become a recorded change in the monitoring system.

This also makes comparisons between units more meaningful. Two identical instruments installed in different hydraulic positions are not necessarily measuring equivalent conditions. An apparent difference between tanks may require a positioning review before it is attributed to fish performance or a treatment-system problem. The farm should preserve enough context to examine both explanations.

Keep fouling separate from calibration drift

Fouling occurs when material accumulates on the sensing surface. Deposits and biological growth can change the response of the instrument. Calibration drift is a change in measurement response that remains after cleaning and becomes apparent when the sensor is checked against an appropriate reference. Both processes can occur during the same deployment, but they call for different evidence.

USGS procedures distinguish readings taken before cleaning, after cleaning and during calibration checks. Cleaning and immediately recalibrating an instrument without recording those intermediate observations loses the opportunity to separate the causes. On a farm, an intervention record can capture the instrument identifier, time, visible deposits, successive readings and action taken. Cleaning products and calibration procedures must remain appropriate to the particular instrument and its manufacturer’s instructions.

Water conditions may change while the operator is servicing the equipment. A difference between the first and last reading therefore cannot automatically be assigned to fouling. Keeping a checked reference meter in the water helps establish whether the environment changed during the intervention. That reference also needs its own quality controls: disagreement between two instruments does not reveal which one is correct.

Optical sensing still needs local validation

Johnston and Williams compared two Clark-cell oxygen sensors with three optical sensors in Oregon’s Tualatin River. During the three-week comparison, the optical instruments showed less fouling-related and calibration drift. Measurements were also checked against a chemical reference method. The findings support the usefulness of optical measurement under the conditions of that deployment.

They do not establish a maintenance interval for every fish farm. Only one optical instrument from each manufacturer was tested, and the study was conducted in a river rather than a production system. The authors explicitly caution against assuming that every unit would behave identically. Different suspended material, salinity and installation conditions can alter performance. The appropriate lesson is to validate maintenance needs locally, rather than assume that a technology eliminates them.

Respond to uncertain alarms on two tracks

A rapid oxygen decline, a flat signal that no longer responds or an abrupt unexplained step should trigger a predefined response. We suggest checking an independent measurement at the same location, inspecting circulation and oxygen-supply equipment, and observing the fish. Changes in ventilation or distribution provide biological context. However, apparently normal behaviour does not establish that oxygen conditions are acceptable.

Where oxygen shortage is credible, protective measures in the farm’s emergency plan should proceed alongside the instrument check. Waiting for proof of sensor failure can waste time during a real event. Conversely, discovering that one instrument is faulty should not end the investigation if fish remain affected. The record should distinguish a confirmed equipment fault, a confirmed water-quality event and an unresolved discrepancy.

Action thresholds require a separate assessment for the species, life stage and production setting. Instrument acceptance criteria from environmental-monitoring guidance are not veterinary safety limits. Concentration in milligrams per litre and percentage saturation should also remain clearly identified. They are related measurements with different meanings; the associated compensation settings and water parameters belong in the retained record.

Use the record to improve the next deployment

Raw observations, corrected values and maintenance notes should remain distinguishable. Quietly replacing a questionable value prevents later reconstruction of an incident. Leaving a known erroneous reading labelled as valid can be equally misleading. A useful audit trail records what changed, why it changed and what evidence supports the decision.

For recirculating farms, we recommend reviewing these records alongside feeding, biomass changes, pump interruptions and work on the water-treatment system. Repeated discrepancies following a particular operation can help direct an investigation, although timing alone does not prove causation. The review can inform sensor placement, reference checks and servicing arrangements without inventing a universal schedule.

Someone must also own the follow-up. An alarm acknowledged at night but never reviewed can recur indefinitely. Assigning responsibility for checking unresolved events helps distinguish an isolated interruption from a recurring weakness. This is a practical management recommendation, not an additional finding claimed from the USGS comparison.

The same need for traceability applies to pH monitoring, with reference materials and checks suited to the particular probe model.

A reliable number includes its limitations

Good monitoring combines representative placement, documented maintenance and independent confirmation. Cleaning checks and calibration checks answer different questions. Optical instruments can perform well while still requiring oversight, and an uncertain alarm calls for simultaneous attention to fish welfare and measurement reliability.

How Vetofish can help

Vetofish can work with technical teams to relate monitoring points to stock-health risks, investigate unexplained events and develop a practical response to water-quality alarms. The aim is to make measurements useful for prevention and veterinary decisions while keeping the limits of each instrument explicit.