Correcting optical sensor drift with periodic in-situ reference samples

Optical turbidity monitors and suspended-solids sensors are valued because they provide continuous measurements where laboratory sampling alone cannot capture short-lived sediment events. A sensor positioned near a dredging project, river intake, estuary or seabed research site can record plume movement at intervals of seconds or minutes. That resolution is especially useful in Australian waters, where rainfall, tidal exchange and vessel activity can change water quality rapidly.

The measurement is not completely fixed once an instrument has been calibrated. Optical windows accumulate biofilm, fine sediment and air bubbles; light sources age; detectors respond differently as temperature changes; and the relationship between turbidity or backscatter and suspended solids varies with particle size, colour and mineral composition. These influences create measurement drift, meaning that the reported value gradually moves away from the actual condition of the water.

Periodic in-situ reference samples provide a practical way to detect and correct that movement without removing a deployed instrument every time its performance needs checking. A reference can be a stable turbidity standard measured under controlled field conditions, a carefully collected water sample with a laboratory result, or a repeatable local suspension whose properties are documented. The best choice depends on whether the objective is regulatory turbidity, suspended-solids mass concentration, optical backscatter or trend detection.

For Australian operators, the procedure must also account for long travel distances, remote telemetry, strong seasonal changes and demanding marine environments. A monitor operating through the Darwin wet season, a dredging campaign in the Port of Brisbane, or a catchment study in the Murray–Darling Basin may have very different fouling and particle conditions. A drift-correction program therefore combines reference measurements, site knowledge, maintenance records and clear limits for accepting or rejecting data.

Why optical measurements drift in the field

An optical sensor estimates water quality by sending light into the sample and measuring scattered or transmitted light. Turbidity instruments commonly report the response in nephelometric turbidity units, while suspended-solids instruments may be calibrated against gravimetric concentrations in milligrams per litre. Neither output is a direct universal measurement of “dirt” in water. The signal depends on the number, size, shape, colour and refractive properties of particles in the optical path.

Fouling is often the most visible source of drift. Algae, bacterial films, mineral deposits and fine organic material can coat the window or lens. A thin film can reduce emitted light, scatter light before it reaches the water, or create an unstable background signal. In seawater, salt deposits may form during drying periods, while freshwater sites can develop rapidly changing biofilms. Wipers and copper components can reduce fouling, but they do not remove the need for verification.

Electronic and optical components also change over time. LED intensity can decline, detector sensitivity can shift, and small changes in internal alignment can alter the response. Temperature affects both electronics and the physical properties of water. Pressure, vibration and cable damage can add intermittent errors that look like environmental events. At a shallow site near Sydney Harbour, for example, boat wake may produce genuine spikes, whereas repeated spikes at a fixed interval may indicate mechanical interference or a telemetry problem.

The calibration model introduces another form of drift. A sensor calibrated with one sediment type may lose accuracy when the site changes from clay-rich runoff to coarse sand or organic particles. This is particularly important in Australian rivers and coastal zones, where wet-season catchment material can differ substantially from dry-season sediment. A reference check reveals whether the instrument remains stable, but it cannot by itself make an unsuitable sediment calibration physically correct.

Designing a useful reference sample

A reference sample should be stable enough to produce a repeatable optical response and representative enough to test the measurement that matters. Commercial turbidity standards based on approved reference materials are suitable for many turbidity checks, provided their storage, expiry and temperature requirements are followed. They should be protected from contamination, mixed gently where required and allowed to reach a suitable temperature before use.

For suspended-solids monitoring, a site-specific water sample can be more informative than a generic standard. The operator collects a well-mixed sample near the sensor, records the instrument reading, and sends subsamples for laboratory filtration and weighing. Several paired observations across low, medium and high concentrations are more useful than a single sample because they reveal whether the calibration slope has changed. Laboratory handling must preserve the sample and document holding times, container type and analytical method.

Some field programs prepare a local reference suspension from retained site sediment. This can provide a consistent challenge sample for repeated checks, but the material must be homogenised and documented. Settling, flocculation and changes in particle size distribution can make the suspension unstable. A local suspension should therefore be treated as a control material rather than assumed to be an internationally traceable turbidity standard.

The container and measurement geometry matter as much as the material. The vessel should be large enough to prevent the sensor from viewing the wall or base, and the probe should be positioned at the same depth and orientation for each check. Remove bubbles from the optical path, avoid direct sunlight where it can affect the reading, and record mixing time. A reference result is meaningful only when the sample, instrument position and handling method are repeatable.

Before field deployment, operators should define the acceptable difference from the reference value. The limit may be expressed as a percentage, an absolute turbidity value, or a combination of both. Low-range readings often need an absolute limit because a small numerical change can otherwise become a large percentage. The sensor’s usable operating range also needs to be understood; guidance on the linear range limits helps prevent a high-concentration check from being mistaken for evidence of good performance throughout the range.

Applying a correction without hiding a fault

A periodic reference check normally begins with an observed value and a known or independently measured reference value. If a clean instrument reads 102 NTU against a 100 NTU reference, the result may be within the agreed tolerance. If the same instrument later reads 125 NTU, the difference indicates a possible bias. A simple multiplicative correction can be calculated as reference value divided by observed value, but this is appropriate only when the error is proportional and the sensor remains within its valid range.

An additive correction may be more suitable when the error is close to a fixed offset. For example, a sensor that consistently reports 3 NTU in clean reference water may require an offset correction at the low end. A slope-and-intercept model is useful when two or more reference levels are available. These calculations should be applied transparently, with the raw measurement retained and the corrected result stored as a separate data field.

Correction factors should not be used to disguise fouling, a damaged window or a failing lamp. If the reference response changes abruptly, varies between replicate checks, or returns to normal after the sensor is cleaned, the appropriate action is maintenance and investigation. Applying a factor to all historical data can create a false impression of precision, especially when the underlying problem is intermittent.

A reference check also needs a decision rule for the field team. Within tolerance means the data can generally continue with the check recorded. A small but consistent bias may justify a documented correction and closer follow-up. A large, unstable or physically implausible result should trigger cleaning, inspection, repeat testing and possible data qualification. The same logic applies to optical backscatter sensors used as proxies for suspended solids, although the relationship to concentration may require a separate sediment calibration.

For higher-confidence programs, combine reference checks with independent samples and instrument diagnostics. The secondary standard checks can help distinguish a sensor response problem from errors in sample preparation or field handling. Store the date, operator, material identification, lot or preparation details, temperature, raw output, corrected output and any maintenance performed.

Building the checks into an Australian deployment

The interval between checks should reflect risk rather than a universal calendar. A heavily fouling estuary may need verification every few days, while a protected freshwater installation may be checked fortnightly or monthly. During a dredging campaign, reference measurements may be scheduled before work starts, after maintenance, following major storm runoff and whenever the time series shows an unexpected step change. Remote stations can use a shorter interval when an operator is already travelling to the site.

Local conditions strongly affect the schedule. Warm tropical water around Darwin can accelerate biological growth, while the wet season can deliver abrupt sediment pulses that challenge a calibration developed during the dry season. In the Great Barrier Reef region, plume monitoring may require careful separation of natural river discharge, resuspension and dredging effects. In the Murray–Darling Basin, long deployments and variable flow conditions make sample traceability and consistent field technique particularly important.

Marine construction near the Port of Brisbane, Newcastle or Fremantle may involve high vessel traffic, tides, saline water and strict reporting requirements. A monitor should be checked at a stage of the tide and operational cycle that can be reproduced, especially where sediment concentration changes quickly. Australian field crews also need practical plans for access, battery capacity, telemetry outages and the time required to transport samples from remote locations to an accredited laboratory.

A useful record links every corrected result to the exact instrument, deployment location and calibration version. Campbell Scientific now provides product-management and contact information for the D & A Instruments product line, which supports continuity when operators need historical specifications or assistance with established systems. Data platforms should preserve the original signal, the reference observation, the correction method and the quality flag rather than replacing the raw value.

Interpreting corrected data with confidence

Correction improves comparability, but it does not remove uncertainty. A turbidity value may be stable against a reference standard while the conversion to suspended solids remains inaccurate because the field particles have changed. Conversely, a sensor may show a modest reference bias yet still identify plume timing reliably. The required level of accuracy should therefore be stated before the program begins: compliance reporting, process control, research, alarm generation and environmental trend analysis do not all have the same tolerance.

Review the time series around every reference event. A gradual change suggests ageing or progressive fouling; a sudden jump may indicate cleaning, relocation, a cable disturbance or a change in water matrix. Compare the optical signal with depth, conductivity, temperature, rainfall, tide and operational records where available. A correction that improves agreement with reference samples but creates implausible jumps in the continuous record deserves further investigation.

For an effective quality-control file, retain the deployment calibration, reference material details, replicate readings, laboratory results, maintenance notes and acceptance decision. Record whether the sample was taken at the sensor depth and whether the instrument was checked in its normal orientation. If the probe was removed for testing, note that the result may not represent the optical path under in-situ pressure, flow and fouling conditions.

The strongest programs use periodic in-situ reference samples as one part of a control system. Routine cleaning limits physical drift, secondary standards test repeatability, grab samples check the relationship with real suspended material, and diagnostics identify electronic or communication faults. Together, these practices support defensible turbidity and suspended-solids records across Australian freshwater, estuarine and marine deployments while keeping corrections traceable and technically modest.