Active acoustic sensors such as acoustic Doppler profilers and acoustic Doppler current profilers of varying frequencies have been used by the USGS over the past two decades to measure SSC in rivers and streams. These profilers are designed to measure water column velocity, and backscatter intensity is a parameter associated with the internal calculations. The ABS sensor differs from these profilers because acoustic backscatter intensity is the primary product output, and a transducer is used to convert the strength of the backscatter signal to a standardized measure of SSC. The transducer in the ABS sensor is factory calibrated for SSCs ranging from 0 to 10,000 mg/L using a proprietary sediment suspension tank (Sequoia, 2016). Factory calibration uses Ballontini glass impact beads (glass beads) sieved to a range of 75 to 90 µm in calibration suspensions; thus, the standardized concentration is reported in units of milligrams per liter of 75- to 90-µm glass beads.

Initial testing by the USGS indicates the ABS sensor has utility as a novel, cost-effective, off-the-shelf tool for monitoring SSC in surface waters (Snazelle, 2017; Manaster and others, 2020), and its use within the agency has increased since its introduction around 2016. However, initial testing did not account for the potential of transducer calibration drift over longer deployments. Comparison of factory calibration checks in ABS sensors recently deployed by the USGS Wyoming-Montana Water Science Center has indicated average calibration drifts ranging from 5 to 61 percent per year; and maximum drifts from individual concentration bins have ranged from 10 to 92 percent (table 1). Of the four ABS sensors tested, only one stayed within 10 percent per year across all concentration bins (serial no. 6083).

These magnitudes of calibration drift suggest a need to develop methods for checking calibration drift in ABS sensors deployed in the field. Currently (2023), the manufacturer recommends their lab perform calibrations, which requires ending the deployment and returning the sensor to the manufacturer (Sequoia, 2016). Returning the sensor thus requires the user to obtain a second sensor or accept data gaps of weeks to months, which is the approximate common range of time reported by USGS users for the manufacturer to calibrate and return the instrument. These issues could be avoided if the user could apply basic calibration checks in the field or remove the ABS sensor for short periods (less than 1 day) for checks in a laboratory.

As part of its mission to unify and standardize research and development activities of Federal agencies involved in fluvial sediment studies, the Federal Interagency Sedimentation Project partnered with USGS Wyoming-Montana and New Mexico Water Science Centers to examine if standard, low-tech laboratory equipment could be used to perform calibration checks on ABS sensors under long-term deployment.

The purpose of this report is to provide interim guidance for using the ABS sensor. This report summarizes the results of laboratory experiments that tested materials and methods to check calibration drift of the ABS sensor. Based on the criteria of Giesen (2015), the experiments tested the robustness, repeatability, and reliability of the materials and methods. The experiments did not test reproducibility of the materials and methods because the experiments were performed by the same operator, which is a common scenario for USGS continuous water-quality instrument deployments. An example application of one of the methods at a current deployment site is also presented. All data collected for this report are available in a USGS data release (O’Connell and others, 2023).

Four ABS sensors were used for the experiments described in this report. The manufacturer uses a unique four-digit serial number to track its ABS sensors. Three sensors, serial numbers 6122, 6178, and 6211, were used for the laboratory experiments; sensor 6223 was used for the field calibration check. The sensors used in the laboratory experiments were all calibrated by the manufacturer in February 2022 and were not field deployed before the experiments. The sensor used for the field calibration check was calibrated by the manufacturer in September 2020 and then remained in the office until deployment in April 2022.

Each ABS sensor was suspended vertically with a benchtop lab stand into a 2-L beaker with the sensor head submerged about 1 cm into the water column (fig. 2). The inside diameter of the beaker was 12.7 cm. Water agitation to suspend standardized particles in the beaker was achieved using a magnetic stir bar propelled by a Fisher Scientific 11–502–49SH magnetic stir plate. The sample volume of the ABS sensor is 5.5 cm from the transducer head, but the total sample distance can extend to 15 cm (Sequoia, 2016). To avoid acoustic interference with the stir bar or sidewalls of the beaker, a minimum distance of 15 cm was maintained from the transducer face to the top of the stir bar.

Magnetic stir bars are commonly used in laboratories to generate shear and turbulent forces to mix a liquid with dissolved or solid constituents. For dissolved constituents, stir bars are effective at maintaining consistent vertical and horizontal mixtures. However, for suspensions of solids, differences in settling velocity between grains can create vertical differences in sediment concentration, and localized vortices can allow for coarser grains to settle in patches. The standardized particles used by the manufacturer to calibrate the ABS sensor range from 75 to 90 µm and have settling velocities of about 0.1 to 1 centimeter per second (Dietrich, 1982). These settling velocities are fast enough to settle out of the mixture and cause a time-dependent trend in measured standard particle concentrations, thus requiring some investigation of the efficacy of different stir bar geometries.

Three stir bar styles were chosen to compare relative effectiveness of suspending stable concentrations of standardized particles: polygon, cross, and wedge (fig. 3). These stir bars are available from various vendors, and the stir bars used in our experiments ranged from about 6 to 8 cm in length (table 2). Vendors have different descriptions of the stirring characteristics and uses of each shape of stir bar (table 2). The stir bar descriptions shown in table 2 are summaries of various vendor descriptions and were not verified by the USGS. Accuracy of SSCs produced by each stir bar was most important across the range of typical higher calibration concentrations (100, 1,000, and 10,000 mg/L) because these are the conditions when settling of grains was hypothesized to have the greatest effect on measured concentrations.

The sensitivity of the ABS sensor signal is nearly constant between particle sizes of about 60 to 100 μm in diameter (coarse silt to very fine sand; Agrawal and others, 2019). Because of the lack of sensitivity of the ABS sensor in this grain-size range, the manufacturer uses 40- to 90-μm glass beads, sieved to 75 to 90 μm, to calibrate the sensor across a range of concentrations. Typical calibration concentrations are 0, 1.1, 3.2, 10.4, 100, 1,000, and 10,000 mg/L. The term “standardized particles” is used in this report in reference to materials for which the distribution of nominal diameters of the b-axis are known with a reasonable level of precision. For the purposes of this report, standardized particles were used to create suspensions of known concentration (called “standard concentrations” in this report).

Three standardized particle sets were used to create standard concentrations with which to check ABS sensor measurements (table 3). The first set of standardized particles was the 75- to 90-μm glass beads sieved and furnished by the manufacturer (called “manufacturer standards” in this report). This set of particles was used to test the potential of the apparatus to accurately reproduce the manufacturer’s calibration concentrations. The second set of standardized particles was raw (direct from glass bead manufacturer), 40- to 90-μm glass beads, which were used to test the accuracy of off-the-shelf standards (called “raw standards” in this report). The third set of standardized particles were the raw standards sieved by the USGS New Mexico Water Science Center to 75- to 90-μm (called “USGS standards” in this report). The USGS standards were used to provide some estimate of error between particles sieved by two operators.

Sediment standard concentration test intervals for particle standards were chosen to replicate the ABS manufacturer’s calibration intervals: 0, 1.1, 3.2, 10.4, 100, 1,000, and 10,000 mg/L. Standard concentrations were achieved by weighing particle mass using a Mettler AE160 digital analytical balance. Water volume was quantified by taring the balance with a dry beaker and incrementally adding room-temperature deionized water. After mass was determined, standards were added to the water column and allowed to mix for 2 minutes before submerging the sensor face. Live readings were made using the ABS sensor manufacturer’s software and were monitored for stability before collecting 20 sequential measurements (Sequoia, 2016). If a trend or instability was observed in the data, the cause was often bubble nucleation on the sensor face, or a beaker adjustment was needed to ensure the stir bar was rotating in the center of the beaker. If the stir bar was not precisely in the center of the beaker, sediment pockets often developed on the bottom of the beaker, and adjustments were needed to ensure suspension of all particles. Sediment pockets were most likely to develop at concentrations greater than or equal to 1,000 mg/L. Measurements at lower concentrations, even if inaccurate, did not seem to destabilize by minor misalignment of the stir bar.

One of the calibration checks developed using laboratory experiments was applied to an ABS sensor currently deployed in the field at the Animas River near Cedar Hill, New Mexico, streamgage (USGS streamgage 09363500 [not shown]; USGS, 2023). The deployment includes the ABS sensor alongside turbidity sensors, which are used with SSC measurements to build statistical regression models for predicting SSC for a separate project. Predicted values of SSC are then compared with samples analyzed for suspended metals to understand metal transport in the Animas River. The ABS sensor and turbidity data are foundational for estimates of SSC and metal loads for that project. It was thus ideal to use a method for verifying calibration of the ABS over the term of the deployment.

A field calibration check station for the ABS sensor was set up inside the gage house of USGS streamgage 09363500, which is an about 8- by 10-foot steel shed with shelving and a countertop (fig. 4). The check station was set up on the countertop and included the following: battery-powered Fisher Scientific stir plate, 1-L Pyrex beaker, polygon stir bars, deionized water for rinsing and mixing calibration solutions, and a laboratory stand with flask clamps to hold the ABS sensor in the 1-L beaker in proper orientation to avoid interference with the sides and bottom of the beaker. Glass scintillation vials (20 milliliters) were used to hold pre-weighed glass beads to make standard concentrations of 10.4, 100, 1,000, and 10,000 mg/L. The standardized particles were weighed in the USGS New Mexico Water Science Center Sediment Laboratory and were added to a known volume of water in the field to arrive at the exact concentration. The 1.1 and 3.2 mg/L concentrations were omitted in field calibration checks because it was a range of instrument operation that was not critical to the project objectives, and preliminary results from laboratory experiments indicated substantial potential for error at those standard concentrations.

During each monthly site visit, field personnel removed the ABS sensor from the housing at the river’s edge, cleaned the instrument to avoid introducing sediment to the calibration check process, and brought the instrument to the gage house. The materials for calibration checks included the 7.5-cm polygon stir bar and glass beads sieved by the USGS to nominal diameters of 75 to 90 µm; the primary measure of error was recorded as MPD between the expected concentration and 20 to 60 measured concentrations.

The wedge stir bar had difficulty maintaining speed in the mixing chamber, creating erratic readings, and experimentation with that stir bar was abandoned. The polygon and cross-shaped bars combined with the raw standardized particles provided the most consistent results; MAPDs for higher concentrations were generally less than 10 percent. The cross-shape bar had slightly smaller errors across the range of standard calibration concentrations, but these errors were not judged to be significant. Speed of the stir bar played a critical role in suspending the standardized particles. Low speed resulted in particles settling out; too high of a speed resulted in the formation of a vortex that would interfere with the ABS sensor backscatter signal. A moderately high speed of 8 on the stir plate produced consistent mixing results. Optimal mixing speeds may vary by stir plate mode and manufacturer and should be adjusted as necessary.

MPD relative to the standard concentration using ABS manufacturer standardized particles varied by instrument and standard concentration, but a general pattern of the magnitude of MPD being inversely proportional to standard concentration was consistent across instruments (fig. 5). The MPDs of measured concentrations were all positive, indicating the instrument was consistently reading higher than the standard concentration. At the lowest standard concentrations of 1.1 and 3.2 mg/L, MPDs varied from 58 to 219 percent (table 5) but decreased to between 6.4 and 29 percent at the highest standard concentrations of 1,000 and 10,000 mg/L. Behavior of RMSE was not consistent across instruments; instruments 6178 and 6211 had smaller relative RMSE values at lower concentrations relative to instrument 6122, and RMSE values generally aligned at higher standard concentrations.

Experiments using the ABS manufacturer standardized particles and the cross-shaped stir bar showed accuracy and behaviors similar to those with the polygon-shaped stir bar in that the MPDs were consistently higher at lower concentrations and improved with increasing standard concentration (fig. 5). Magnitudes of MPD were like those produced using the polygon-shaped stir bar except for measurements at the 100-mg/L standard concentration, which were consistently lower than the expected concentration (table 6). Root-mean square errors were consistently close between instruments for concentrations as much as 1,000 mg/L but differed by as much as a factor of four between instruments at the 10,000-mg/L standard concentration (table 6).

MPD relative to the standard using raw standardized particles exhibited similar patterns as those for sensor manufacturer standardized particles whereby the largest percent differences were observed at low concentrations and the smallest at higher concentrations (fig. 6), but maximum percent differences were mostly negative at low concentrations. The magnitude of percent differences relative to the standard also declined substantially relative to those observed for Sequoia standardized particles; maximum MAPD was 20 percent across the range of concentrations and 10 percent across concentrations greater than 10.4 mg/L (table 4). MPD did not change substantively among stir bar types over the range of standard concentrations and for high concentrations (greater than 10.4 mg/L).

Maximum MPDs observed using the raw standardized particles and the polygon-shaped stir bar were largest for low concentrations and diminished for concentrations greater than or equal to 10 mg/L (fig. 6). MPDs for each standard concentration ranged from −80 to 16 percent at low standard concentrations of 1.1 and 3.2 mg/L but decreased to a range of −2.5 to 14 percent for concentrations of 100 mg/L or greater (table 7). Instrument 6211 read consistently lower than the other instruments when the polygon-shaped stir bar was used, but no consistent bias was observed over the range of concentrations in the other instruments (fig. 6). Root-mean square errors were generally consistent across instruments in that values fell within a factor of two across instruments (table 7).

MPDs observed using the combination of raw standardized particles and the cross-shaped stir bar exhibited the same behavior as those for polygon-shaped stir bar, whereby the largest errors were observed at low concentrations and diminished with increasing concentration (fig. 6). MPDs for each standard concentration ranged from −66 to 28 percent at low standard concentrations of 1.1 and 3.2 mg/L but decreased to a range of −5.9 to 23 percent for concentrations of 100 mg/L or greater (table 8). Instrument 6211 read consistently lower than the other instruments for concentrations of 10.4 mg/L or less, but no consistent bias was observed among other instruments or standard concentrations (fig. 6). Magnitude of RMSE was generally consistent across instruments and concentrations up to 1,000 mg/L but differed by as much as a factor of five for standard concentrations of 10,000 mg/L (table 8).

Experiments using USGS standardized particles were only done with the cross-shaped stir bar. The magnitude and pattern of MPDs relative to the standard concentrations were like those observed for the experiments with the raw standardized particles because errors where largest and negative for low concentrations and lowest (positive or negative) at higher concentrations (fig. 7). Maximum MAPD values for USGS standardized particles were 21 percent across the range concentrations and 8 percent across concentrations greater than 10.4 mg/L (table 4). MPDs for each standard concentration ranged from −70 to 17 percent for standard concentrations less than 10.4 mg/L and −10.1 to 17 percent for standard concentrations greater than 100 mg/L (table 9). Instrument 6211 had the largest errors on average, but MAPDs for that instrument were less than 10 percent for concentrations greater than 10.4 mg/L.

To test the precision of the methods described above, two sequential experiments were run within the same day. Raw standardized particles and a cross-shaped stir bar were chosen to create the conditions for quantifying repeatability of experiments. This combination was chosen because ideal standardized particles would be those that have the least cost and best accuracy. Raw standardized particles meet the criteria of low cost because they are readily available from commercial vendors and do not require additional processing. Likewise, the results of the experiments described above indicate that the raw standardized materials and cross-shaped stir bar produced results equal to or better than those observed using other combinations of stir bar and standardized particles.

For the purposes of testing repeatability of creating standardized conditions, standardized conditions were reproduced for concentrations of 1.1, 10.4, 100, 1,000, and 10,000 mg/L. Only 2 of the 15 MAPD values between first and second experiments exceeded 10 percent, and both were for concentrations of 10.4 mg/L or less, where previous experiments demonstrated large errors (fig. 8). For standardized concentrations of 100 mg/L or greater, MAPD values were all less than 10 percent, and eight out of nine measurements were less than 7.5 percent (table 10), indicating conditions were reproducible within the maximum limits generally accepted by the USGS for requiring re-calibration.

Despite the relatively small errors between the repeated experiments, unpaired, nonparametric statistical hypothesis tests indicate that the differences between experiments were significant at the 95-percent confidence level for 13 of 15 experiments (table 10). These data suggest that the experimental conditions produced by the experiments described here were not stable enough, or within-instrument precision was not high enough, to produce conditions that were statistically indistinguishable within the 10-percent error bounds accepted by the USGS.

Monthly field calibration checks were completed at USGS streamgage 09363500 (USGS, 2023) from July through November of 2022 (table 11). All concentrations measured during the calibration checks were higher than the standard concentration; however, all MAPDs except one were within 10 percent, and 12 of the 20 checks were 5 percent or less. As expected, errors were highest for the smallest concentration of 10.1 mg/L, but only one of the checks exceeded the 10 percent threshold at that concentration. Across the 5 months, none of the calibration check values indicated a consistent increasing or decreasing trend, suggesting that the sensor calibration was not drifting.