Techniques for Reducing Flow-Induced Noise in In-Situ Turbidity Measurements
In-situ turbidity sensors provide continuous information about suspended particles in rivers, channels, reservoirs, estuaries, and coastal waters. Their value depends on detecting changes caused by sediment transport rather than changes created by the deployment itself. Fast currents, turbulence, bubbles, vibration, and unstable mounting can all produce fluctuating optical signals that appear to be genuine changes in turbidity.
Flow-induced noise is especially important during storm events, dredging, flood monitoring, and hydrological research. These applications often involve rapidly changing water levels and high suspended-solids concentrations, so the sensor must remain representative while resisting hydraulic disturbance. A carefully selected instrument is only the starting point; installation, sampling configuration, maintenance, and data processing have equal influence on measurement quality.
The most effective approach combines hydraulic awareness with optical discipline. The sensor should be placed where water is representative but not directly exposed to wakes, recirculation zones, or debris impacts. Its output should then be checked against site observations and sediment samples so that a stable signal is also a meaningful one.
Why moving water disturbs optical measurements
A turbidity monitor generally estimates particle concentration by transmitting light into the water and measuring scattered or attenuated light. In clear, steady water, the optical path is relatively consistent. In a turbulent flow, the path can be interrupted by air bubbles, rapidly changing particle orientations, eddies, and material passing unevenly through the sensing volume.
Bubbles are among the most disruptive sources of interference. They can scatter light strongly and create short, sharp spikes that are unrelated to sediment concentration. Surface turbulence, hydraulic jumps, weirs, aeration structures, and pump discharges are common sources. Even when bubbles are not visible at the surface, entrained air can pass the submerged optics during high discharge.
Mechanical movement adds another form of noise. A loose bracket, flexible cable, or vibrating pole can cause the sensing head to move relative to the water. That movement changes the particle population in the optical field and may repeatedly expose the instrument to its own wake. In fast rivers, pressure fluctuations can also make a poorly supported mounting assembly oscillate.
The signal can be affected by the flow around the instrument as well. A sensor body, guard, wiper, or cable creates a local disturbance that may draw sediment into the measurement zone or leave it in a low-velocity shadow. These effects are difficult to remove after deployment, which is why hydraulic positioning is usually more valuable than aggressive data filtering.
Choose a stable and representative deployment position
A good location balances two requirements: water must be moving sufficiently to represent the cross-section, but the sensor must avoid severe turbulence and local hydraulic anomalies. The ideal position is often below the surface and away from the bed, bank, and structures. The exact depth depends on channel geometry, stratification, sediment size, and the intended monitoring objective.
Avoid placing the optical window directly downstream of a support, cable, piling, bend, intake, or other obstruction. These features can produce alternating vortices and recirculation. A short relocation can make a substantial difference, especially when the sensor is mounted immediately behind a frame or facing into the wake of a vertical member.
Sensor orientation should be considered with the current direction. In many installations, presenting a streamlined profile to the flow reduces separation and vibration. The optical path should remain clear of the boundary layer near the mounting surface, while the sensor should not face directly into a region where bubbles or floating material accumulate. A small change in angle can reduce glare, particle shadowing, and direct exposure to entrained air.
Fixed-platform installations require particular care because a rigid structure can still transmit vibration. Use a robust bracket, minimize unsupported cable length, and secure the cable so that it cannot sweep through the sensing volume. Practical guidance on high-flow deployments can help translate these principles into a river installation plan.
Control bubbles, fouling, and optical interference
Bubble avoidance begins with site selection, but it can also involve depth and timing. If the water column is strongly aerated near the surface, lowering the sensor may reduce false spikes, provided the deeper location remains representative. In tidal or regulated systems, compare measurements across water levels because the depth of the bubble-rich zone may change with discharge and structure operation.
The optical face should be protected from sediment deposition and biological growth without creating a hydraulic pocket. A wiper or cleaning system may be useful where fouling is persistent, but it should be assessed for vibration and power requirements. A moving wiper can briefly disturb the signal, so the data logger should record maintenance events or apply a defined settling interval after cleaning.
A protective guard can shield the sensing head from impact and large debris, yet an overly dense guard may trap particles or generate a recirculation zone. Open, hydraulically clean protection is preferable to a cage that blocks the flow. Inspect the guard for accumulated leaves, algae, and sediment because these materials can produce long-term bias rather than obvious short-term spikes.
Optical interference also includes ambient light, reflective surfaces, and changing water color. Install the sensor according to its optical geometry and manufacturer’s recommended clearance. Keep the measurement path away from the bed when bright mineral surfaces or dense aquatic vegetation could enter the field of view. A stable dark background is not always required, but a consistent and unobstructed optical environment improves repeatability.
Use sampling and signal-processing methods carefully
A single instantaneous reading can exaggerate turbulence-related variation. Configuring the logger to collect a burst of readings and store a median, mean, or other robust statistic can make the reported value more representative. The appropriate interval depends on the turbulence scale and the monitoring objective. Short bursts may capture fast sediment pulses, while longer averaging can suppress noise at the cost of temporal detail.
Median filtering is useful when the record contains isolated bubble spikes. A moving median is less influenced by extreme values than a moving average, although it can flatten genuine short-duration events. Standard deviation, range, or spike counts should be retained as diagnostic metrics where possible. A smooth record without quality indicators can conceal a failing mount or a fouled optical window.
Do not use filtering as a substitute for correcting the installation. If the sensor is repeatedly exposed to bubbles, a filter may remove valid high-concentration events along with artifacts. Flagging data based on impossible rates of change, diagnostic status, cleaning cycles, or simultaneous pressure and flow observations is generally safer than deleting all rapid variation.
Synchronizing turbidity data with discharge, stage, velocity, rainfall, and maintenance records helps identify the source of instability. Flow-related noise often follows a predictable pattern: it may begin at a particular stage, occur only during rising water, or appear when a pump or gate changes state. These relationships can guide a physical modification much more effectively than examining turbidity alone.
| Source of disturbance | Typical symptom | Preferred physical response | Useful data treatment |
|---|---|---|---|
| Entrained air | Sharp positive spikes and erratic bursts | Move below the aerated layer or away from hydraulic structures | Median or spike flagging |
| Mount vibration | Repeating oscillation or unstable baseline | Stiffen the bracket and restrain the cable | Review frequency and range |
| Sensor wake | Persistent bias or unusual difference between orientations | Reorient or relocate the sensing head | Compare with reference samples |
| Bed or bank proximity | Strong dependence on stage or local scour | Increase clearance and reassess representativeness | Stratify data by water level |
| Fouling or sediment coating | Gradual drift and slow recovery after cleaning | Improve cleaning access and inspection frequency | Track diagnostics and cleanout times |
| Debris impact | Sudden offset, dropouts, or physical damage | Add suitable protection or relocate | Flag the affected interval |
Verify performance with site-specific calibration
Reducing variance does not automatically produce accurate suspended-solids data. Turbidity is an optical response, while suspended-solids concentration is a mass concentration. The relationship between the two depends on particle size, mineralogy, shape, color, organic content, and aggregation. Flow changes can alter that particle population, so a calibration developed during low flow may not apply during a flood.
Collect water samples across the expected range of conditions, including rising and falling limbs of events where practical. Pair each sample with the sensor reading, stage, discharge, and time since the last cleaning. Laboratory analysis of total suspended solids or suspended sediment concentration can then be used to create a site-specific relationship and identify whether separate ratings are needed for different flow regimes.
A reliable calibration also requires attention to sampling location. If the sensor is near the bank but the sample is collected in the thalweg, differences may reflect spatial variability rather than instrument error. Replicate samples and repeated measurements near the sensor help distinguish optical noise from real heterogeneity. The guidance on calibrating suspended-solids sensors provides a useful framework for connecting optical output with local sediment samples.
Quality checks should include zero or clean-water checks when feasible, inspection of raw readings, and review of instrument diagnostics. Compare the sensor with a second instrument only after considering differences in optical design, measurement geometry, and calibration. Agreement between two sensors can be reassuring, but it does not replace validation against physical samples.
Diagnose noisy records before changing thresholds
A noisy turbidity record should be investigated in chronological order. First examine whether the variation coincides with changes in stage, current speed, wind, rainfall, gate operation, pumping, or maintenance. Next review installation photographs and field notes for cable movement, sediment accumulation, or changes in the bracket. Finally inspect the sensor’s optical surface and diagnostic records.
The FAQ resources can help with common questions about sensor behavior, installation, and measurement interpretation. Manufacturer documentation is particularly valuable when a signal problem may involve optical saturation, communication faults, temperature effects, or an operating condition outside the intended range.
Separate random noise from systematic bias. Random noise produces scatter around a stable mean, while bias shifts the entire signal upward or downward. A partially blocked optical window may create a gradual drift; a bubble cloud may create isolated spikes; a vibrating mount may create periodic variation. Each pattern calls for a different response.
When possible, perform a controlled field test. Record the sensor while the flow is undisturbed, then change its orientation or temporarily add mechanical support while observing the raw output. A second nearby sensor, a handheld measurement, or timed water samples can provide additional evidence. Document every change so that improvements can be attributed to a specific intervention rather than to natural changes in the river.
Practical actions for a quieter measurement record
The following measures are usually effective when adapted to the site rather than applied as a fixed checklist:
- Mount the sensor on a rigid, low-profile support with the cable secured outside the optical path.
- Keep the sensing window clear of wakes, recirculation zones, bed contact, vegetation, and heavy bubble entrainment.
- Use burst sampling with a robust statistic, while retaining raw or diagnostic values for quality control.
- Inspect and clean the optical face and protective guard on a schedule linked to fouling rate and flow conditions.
- Validate readings with site-specific suspended-solids samples collected across changing discharge and sediment regimes.
These actions work best as a feedback cycle. Start with the hydraulic setting, collect a short diagnostic record, make one physical change at a time, and compare the resulting signal with independent observations. This approach prevents data processing from concealing a mounting problem and makes the final monitoring method easier to defend.
For long-term networks, define acceptance criteria before deployment. Examples include a maximum allowable spike rate, a minimum proportion of valid readings, a stable response after cleaning, and an acceptable difference from paired samples. Clear criteria help operators decide when to service, relocate, recalibrate, or temporarily suspend a sensor.
A well-designed installation preserves the natural sediment signal while reducing disturbances created by the instrument and its support system. Contact Campbell Scientific for current product-management and support information for D & A Instruments technologies, and use the resulting guidance to plan a deployment that delivers dependable turbidity and suspended-solids data in real flow conditions.