Converting Relative Turbidity to Suspended Sediment Concentration
Turbidity sensors installed on dredgers, river gauging stations, and offshore platforms produce readings in Relative Turbidity Units, Nephelometric Turbidity Units, or Formazin Nephelometric Units, yet almost every compliance report, environmental impact statement, and hydrological study in Australia asks for a number in milligrams per litre. The disconnect sits at the heart of sediment monitoring: optical instruments respond to how light scatters off particles, while suspended sediment concentration (SSC) reflects the dry mass of those particles in a known volume of water. A single conversion factor cannot serve every catchment, port, or mine site, because the particles look different in each place.
Building a robust site-specific relationship between relative turbidity and SSC starts with recognising that the two measurements describe overlapping but distinct properties. Sand stirred up during a flood in the Fitzroy River in Queensland, organic-rich mud lifted from the floor of Sydney Harbour, and the bright red tailings plume leaving a Pilbara processing plant will all register on an optical sensor, but each will deliver a different ratio of scattered light to dry mass. A calibration developed in one environment will systematically over- or under-estimate SSC in another, which is why Australian practitioners rely on paired field sampling and regression modelling rather than factory-default curves.
Why Optical Readings Cannot Stand In for Mass Concentration
Relative turbidity is a comparative signal. The sensor shines light into the water, captures what bounces back, and expresses the result against an internal reference, often a calibrated suspension of formazin or a stable polymer bead standard. The output is useful for tracking change over time at the same location because the optics, geometry, and electronics stay constant. It becomes unreliable the moment the question shifts from "is the water dirtier than yesterday" to "how much sediment is in the water right now in mg/L".
Several physical properties of the suspended material shift the optical reading without changing the dry mass. Fine clays in the upper reaches of the Murray-Darling system scatter light many times more efficiently per gram than coarse quartz sand washing out of a Hawkesbury-Nepean tributary after a bushfire-debris flow. Dark-coloured particles from a coal seam near Newcastle absorb rather than reflect light, depressing the turbidity signal even when SSC is high. Organic matter from algal blooms in the Gippsland Lakes adds a fluffy, low-density component that inflates the optical response. Because the same sensor sees all of these, a generic conversion equation will always misrepresent at least some of them.
For this reason, Australian laboratories and water authorities insist on gravimetric SSC values derived from filtered and dried water samples, and they require turbidity to be calibrated against those values at the actual deployment site. Sensor manufacturers publish guidance on this point, with application notes and technical background available through D & A Instruments for practitioners building defensible site-specific relationships rather than relying on transfer curves carried over from unrelated deployments.
Designing a Paired Sampling Campaign
The conversion from relative turbidity to SSC is only as good as the dataset behind it, and that dataset comes from physically collecting water at the same depth and time the sensor is logging. A typical Australian campaign begins with a site walkover to identify the dominant sediment sources: is the material coming from a freshwater inflow, an estuary, an engineered channel, or a working dredge? In Brisbane, for example, the Brisbane River receives pulses of urban runoff carrying fine silt, while the Port of Brisbane experiences coarser material resuspended by ship traffic, and a single regression built for both settings would be misleading.
Sampling needs to span the full range of conditions the site experiences. Low-flow baseline samples taken during the dry season in Darwin Harbour establish the lower end of the relationship. Storm events captured by automatic samplers in catchments like the Yarra or the Swan River push the upper end, sometimes above several thousand mg/L during a single flush. Rising-stage sampling is critical because hysteresis is common: the relationship during the rising limb of a flood differs from the falling limb, as heavier particles settle first and fine material travels further. A campaign that misses high events produces a curve that extrapolates poorly when conditions finally peak.
Practical logistics matter. Sample bottles must be cleaned, pre-labelled, and stored in a dark cooler in the field, then filtered through pre-weighed glass-fibre filters in a NATA-accredited laboratory within 24 to 48 hours. The filter is dried, reweighed, and the mass difference divided by the filtered volume to give SSC in mg/L. Concurrent sensor readings are pulled from the data logger with a matching timestamp, and any gap larger than a few minutes between sample and logger entry is flagged for later review. Planners should expect to collect between 30 and 100 paired points before the data is robust enough to fit a curve, and field teams can engage the support portal for advice on log configuration, sample timing, and the practicalities of synchronising logger clocks with field sampling rounds.
Fitting the Site-Specific Regression
Once the paired dataset is in hand, the next step is choosing a functional form. The simplest model is a straight line through the origin, SSC = k × RTU, where k is an empirical slope with units of mg/L per RTU. This works reasonably well for narrow ranges and uniform particle populations, such as a sheltered intake pond fed by a single mine dewatering line. For broader ranges, a log-transformed model, log(SSC) = a + b × log(RTU), often performs better because it allows the scatter to grow with concentration, which is the typical pattern when both very clear and very turbid waters are included.
More complex polynomials or piecewise models are occasionally justified, particularly in estuaries where freshwater and marine end-members produce two distinct clusters. The decision should be driven by residual analysis rather than by chasing a higher R-squared value. Plots of predicted versus observed SSC, residuals versus predicted value, and residuals versus time of year reveal whether the model is missing a pattern. A high R-squared with a fan-shaped residual plot indicates that the uncertainty grows with concentration, which is information the user needs in order to report confidence intervals honestly to regulators and stakeholders.
Spreadsheets, plotting templates, and worked examples available from the downloads section can speed up this work considerably, especially for teams managing multiple sites in different jurisdictions. Site-specific equations should be stored with metadata that records the date range of the calibration, the number of paired samples, the mean particle size or settling velocity if available, and any major rainfall or dredging events that occurred during the campaign. This metadata is essential when the relationship is reviewed during an audit by the New South Wales Environment Protection Authority or the Victorian Environment Protection Authority, both of which expect documented evidence that the conversion is appropriate for the receiving environment.
Accounting for Local Variability Beyond the Calibration Window
A calibration completed in autumn does not necessarily describe winter or summer conditions. Tropical Australia experiences a pronounced wet season between November and April, when monsoonal rains deliver fresh sediment loads to rivers such as the Mitchell and the Daly, and the particle size distribution can shift dramatically compared with the dry-season baseline. Southern Australia, by contrast, sees winter rainfall and spring snowmelt in the alpine catchments that feed the Murray and its tributaries, with different sediment characteristics again. A static regression will gradually drift out of calibration if these seasonal cycles are ignored.
Operators should plan for periodic re-calibration, typically on an annual cycle or after any major catchment disturbance such as a bushfire, flood, or land-use change. The 2019-2020 bushfires across eastern Australia mobilised huge volumes of ash and charred soil into waterways from the Hawkesbury to the Snowy, and several councils found that existing turbidity-to-SSC relationships no longer held once the post-fire debris moved through their catchments. Updating the dataset with a fresh batch of paired samples, then refitting the curve, restored defensible numbers within a season.
Site-specific factors also include biofouling, sensor alignment, and the influence of bubbles during heavy rainfall or dredging. Optical windows in tropical waters around Cairns and Townsville foul quickly with algae and biofilm, suppressing the signal and biasing the SSC estimate downward if the operator does not follow a cleaning schedule. Routine checks, wiper maintenance, and reference-standard verification keep the optical baseline stable, and a documented cleaning log forms part of the evidence package submitted alongside every calibration report.
Validating, Documenting, and Maintaining the Conversion
The final step is independent verification. After the regression has been fit, a small set of samples withheld from the calibration dataset, typically 10 to 20 percent of the total, should be run through the model and compared with the laboratory SSC values. Acceptance criteria are usually expressed as a percentage error or a root-mean-square error in mg/L, and these thresholds are often set by the project regulator rather than by the operator. Gladstone Ports Corporation, for example, publishes specific performance criteria for turbidity monitoring during capital dredging campaigns, and contractors must demonstrate compliance before mobilisation and after each major works phase.
Documentation is what turns a working equation into a defensible monitoring tool. Every calibration event should produce a report that records the instrument serial number, firmware version, optical configuration, the raw paired dataset, the fitted model with coefficients and confidence intervals, the residual diagnostics, and the names of the analysts and field operators. Reports are then archived alongside the monitoring plan so that any future inquiry, whether from a state environment agency, a port authority, or an internal audit, can trace the conversion back to its evidence base rather than treating it as an inherited black box.
Long-term reliability depends on treating the conversion as a living object rather than a one-off task. A drift check on every site visit, an annual re-fit using the most recent paired samples, a log of all sensor maintenance, and a clear escalation path to technical advisors when results look anomalous will keep the relationship between relative turbidity and suspended sediment concentration trustworthy across years of changing conditions. In Australian monitoring practice, where rivers, estuaries, and working harbours all behave differently and where regulators expect robust evidence, that ongoing discipline is what makes an optical sensor a credible source of SSC data.