Turning Optical Backscatter into Reliable Sediment Time Series
Optical backscatter sensors provide a fast way to observe changes in suspended sediment concentration. Instead of collecting a water sample every few hours, an instrument can record fluctuations caused by storms, dredging, vessel traffic, tidal currents, or changing river discharge. The raw output, however, is not automatically a concentration record. It is an optical response that must be cleaned, interpreted, calibrated, and checked against conditions in the field.
A defensible sediment time series therefore depends on more than a conversion equation. Sensor configuration, deployment depth, particle characteristics, fouling, bubbles, instrument range, and sampling frequency can all affect the relationship between backscatter and suspended solids. Processing should preserve genuine short-term events while removing readings that reflect the instrument or deployment rather than the water column.
The workflow below applies to optical backscatter measurements from marine, freshwater, and dredging environments. It is also relevant to turbidity monitoring systems and OEM instruments that deliver voltage, digital counts, or internally calculated optical values.
Start With the Measurement Rather Than the Equation
Raw optical backscatter data usually represent the intensity of light scattered toward a detector after an instrument emits light into the surrounding water. Depending on the sensor, the recorded value may be an analog voltage, digital count, frequency, or processed signal such as turbidity in nephelometric units. These outputs are related to suspended particles, but they are not interchangeable with suspended sediment concentration.
The first processing step is to document exactly what the channel means. Record the instrument model, optical wavelength, beam geometry, gain setting, calibration range, firmware, units, and any internal filtering. A value called “backscatter” may already include an offset correction or logarithmic transformation. Applying a second correction without understanding the signal path can produce a biased result.
Deployment information is equally important. Store sensor depth, orientation, location, water level, flow direction, instrument serial number, and installation dates with the data. If the instrument was moved between sites or depths, treat each deployment as a distinct calibration population unless testing demonstrates that the same response applies.
Build a Traceable Raw Data Record
Preserve the original data file before applying edits. A useful archive contains the unmodified instrument export, a decoded working file, processing scripts or spreadsheet formulas, calibration records, and a quality-control log. Every removed or modified observation should have a reason code rather than disappearing silently.
Convert timestamps to a consistent standard, preferably Coordinated Universal Time, while retaining the original logger time in a separate field. Check for clock drift, daylight-saving changes, duplicated records, gaps, impossible dates, and irregular sampling intervals. Align optical measurements with pressure, flow, stage, conductivity, temperature, and wave or velocity data when those variables are available.
Sampling frequency should reflect the process being monitored. A dredging plume or rapidly rising storm hydrograph may change within minutes, while a seasonal lake record can tolerate longer intervals. If the logger samples rapidly but stores averages, retain the averaging period and determine whether maximum, minimum, or variance information is needed to identify short-lived events.
A practical dataset includes at least these fields:
- Timestamp and time zone
- Raw optical signal and converted engineering units
- Sensor depth and deployment identifier
- Battery, diagnostic, or instrument-status fields
- Water temperature and other supporting measurements
- Quality flags and processing notes
- Laboratory suspended-solids results used for calibration
Remove Instrument and Deployment Artifacts
Optical sensors can respond to conditions unrelated to the average sediment concentration. Air bubbles create sharp spikes, biofouling causes gradual drift, and a tilted or buried sensor can change the illuminated particle field. Sunlight, electrical interference, wiper movement, cable motion, and reflections from a mounting frame can also create abnormal observations.
Begin with simple physical tests. Flag values outside the instrument’s valid range, negative readings after offset correction, missing values, repeated constants, and abrupt steps that coincide with maintenance or power events. Do not automatically remove every spike. A sharp peak during a known dredging pass or rainfall-driven flow increase may be the signal of interest. A spike with no corresponding change in water level, velocity, or neighboring sensors deserves closer review.
Use diagnostic variables where possible. A sensor with a fouling indicator, reference detector, optical clarity channel, or wiper status can reveal gradual contamination. Compare the optical record with photographs, field notes, cleaning dates, and grab samples. A slow increase followed by an immediate decline after cleaning is more likely fouling than a sustained increase in sediment.
Filtering should be conservative and transparent. A short median filter can suppress isolated electrical spikes while retaining event structure. Moving averages are useful for matching laboratory sample intervals, but they blur peaks and shift event timing if applied carelessly. Keep both the unfiltered and filtered series, and record the window length, end treatment, and missing-data rule.
Correct and Convert the Optical Signal
After quality control, convert the raw channel into a physical optical variable. The exact calculation depends on the instrument. Some systems use a linear voltage relationship; others report counts that must be transformed using factory coefficients. A common conceptual form is:
[ B = k(S-S_0) ]
where (S) is the measured signal, (S_0) is the dark or clean-water offset, and (k) is an instrument scaling factor. In some instruments, the signal is expressed on a logarithmic scale, so the manufacturer’s equation must be followed instead of assuming linearity.
The offset should come from a documented zero or clean-water measurement. Do not assume that zero voltage means zero scattering. Water itself, the sensor window, optical noise, and dissolved material can contribute to the baseline. If the offset changes during a deployment, investigate whether the cause is fouling, electronic drift, temperature sensitivity, or a change in the optical environment.
The relationship between optical backscatter and suspended sediment concentration is usually site-specific. A general empirical model may be written as:
[ SSC = aB + c ]
or, when particle behavior and dynamic range require it:
[ SSC = aB^b ]
Here, (SSC) is commonly expressed in milligrams per liter, (B) is the corrected optical signal, and the coefficients are estimated from paired field measurements. A model should be selected from observed data, not imposed because it is convenient.
Particle size, mineral composition, color, shape, concentration, and settling behavior influence scattering. Fine clay may produce a very different response from coarse sand at the same mass concentration. Salinity and dissolved organic matter can also affect optical properties. For that reason, a calibration developed in a clear freshwater lake should not automatically be applied to a turbid estuary or dredging channel.
Calibrate Against Suspended Solids Samples
Collect water samples across the full expected range of conditions, including low background levels, rising limbs, falling limbs, and high-concentration events. Pair each sample with the sensor reading at the same depth and time. If the water column is stratified or vertically variable, sample depth must match the optical measurement or be documented as a separate limitation.
Laboratory analysis typically involves filtering a known water volume, drying the retained material, and calculating suspended solids by mass per unit volume. Follow a consistent method for filter type, drying temperature, balance precision, sample volume, and handling. Replicate samples help distinguish laboratory variation from sensor response.
A calibration dataset should be broad enough to support the intended model. If samples cluster at low concentrations, the resulting equation may perform poorly during storms or dredging. If sediment composition changes by season or flow regime, test separate regressions or use a conditional calibration. The lake nutrient context illustrates why a turbidity signal can be useful while still requiring careful interpretation of what the particles represent.
Plot measured concentration against corrected backscatter before fitting a model. Look for curvature, distinct groups, hysteresis, and changing variance. A single regression may hide two populations, such as fine suspended material during baseflow and coarse resuspended sediment during high flow. In that situation, a flow-dependent or particle-class calibration may be more realistic than forcing every observation into one line.
Validate the calibration with samples that were not used to estimate its coefficients. Report bias, root mean square error, mean absolute error, coefficient of determination, and the number and range of validation samples. Also inspect residuals over time and concentration. A strong overall correlation can conceal systematic underestimation of peaks or poor performance at the low end.
Choose the Right Processing Approach
Different optical outputs require different handling. The comparison below separates the measured signal from the derived water-quality quantity and identifies where uncertainty enters the workflow.
| Data product | What it represents | Main processing need | Typical limitation |
|---|---|---|---|
| Raw voltage or digital counts | Instrument detector response | Decode units, apply offset and range checks | Not comparable across instruments without metadata |
| Corrected optical backscatter | Scattering response after baseline correction | Check drift, fouling, bubbles, and geometry | Still depends strongly on particle properties |
| Turbidity estimate | Optical response expressed in a standardized reporting unit | Verify factory coefficients and sensor condition | May not track mass concentration consistently |
| Suspended sediment concentration | Estimated particle mass per water volume | Apply site-specific calibration and validate samples | Conversion changes with sediment type and flow |
| Event-averaged concentration | Mean concentration over a defined interval or storm | Define interval, handle gaps, and preserve peaks | Averaging can hide short high-concentration pulses |
Turbidity and suspended sediment concentration are related but distinct. Turbidity is an optical property, while suspended solids are a mass-based laboratory result. A turbidity sensor can be useful for continuous monitoring, yet its reading should not be presented as a mass concentration unless a suitable relationship has been established. Sensor selection issues in a treatment setting are discussed in this comparison of optical and conventional sensors, where operating conditions and measurement purpose affect the preferred approach.
For high-frequency records, calculate summary products that match the monitoring objective. Useful outputs include hourly or daily means, event maxima, percentile concentrations, duration above a threshold, and sediment load. Retain the original sampling interval as a separate product so that averaging does not erase short plume passages or peak transport.
Integrate Quality Flags and Environmental Drivers
A concentration time series becomes more useful when every value carries a quality status. A simple flag structure might distinguish valid observations, suspect observations, corrected observations, missing data, sensor maintenance, and values outside the calibration range. Keep suspect values available for review, but exclude them from formal statistics unless the project’s data policy allows otherwise.
Cross-check the processed series against environmental drivers. In rivers, concentration often changes with discharge, stage, velocity, and hysteresis between rising and falling flow. In coastal water, tide, wave height, current direction, and vessel activity can explain rapid changes. In dredging applications, equipment position and operating status help separate the intended plume from unrelated background variability.
When several sensors are deployed, compare their timing and relative response rather than expecting identical concentrations. A sensor nearer the bed may record resuspension earlier than a midwater instrument. Differences can reveal vertical structure, but they can also indicate unequal fouling, different optical ranges, or local flow disturbance around the mount.
Sediment load can be estimated by combining concentration with water discharge:
[ Load = SSC \times Q \times C ]
where (Q) is discharge and (C) converts the units to a desired mass rate. Use synchronized data and state the assumptions clearly. Missing flow values, concentration values beyond the calibration range, and changing cross-sectional concentration profiles can create more uncertainty in load estimates than the optical measurement itself.
Recommendations for a Defensible Workflow
- Preserve the untouched raw file and maintain a processing log with versioned formulas or code.
- Document sensor settings, deployment geometry, maintenance, timestamps, and every calibration sample.
- Use conservative quality-control tests that distinguish genuine sediment events from bubbles, fouling, and electronic artifacts.
- Develop calibration equations from paired samples spanning the complete concentration and particle range.
- Validate independently and publish uncertainty, calibration limits, quality flags, and averaging intervals with the final series.
A robust workflow should be repeatable by another analyst. Store calibration coefficients with their valid site, season, sensor, and sediment conditions rather than placing one universal equation in a project template. Recalibrate after major changes in particle source, sensor configuration, optical window condition, or deployment location.
For long deployments, schedule periodic cleaning and verification samples. Compare the instrument with a reference measurement before and after maintenance when possible. These checks can identify drift before it becomes embedded in a multi-month sediment record.
Reliable processing turns an optical detector into evidence that can support environmental research, dredging compliance, hydrology, and operational decisions. D & A Instruments’ optical sensing heritage, together with Campbell Scientific’s current product and support information, provides a useful foundation for building monitoring systems around documented measurements rather than unexplained converted values.
Begin with the raw channel, preserve its context, and build the concentration series through transparent corrections and site-specific validation. When the resulting dataset shows its assumptions, limitations, and quality history, it can support confident interpretation of sediment transport and help operators respond to changing water conditions in real time.