SEARCH

SEARCH BY CITATION

Keywords:

  • stream gaging;
  • monitoring;
  • streamflow;
  • nonpoint source pollution

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Summary and Conclusions
  7. Acknowledgments
  8. Literature Cited

Abstract:  Nonpoint source pollution (NPS) studies, such as total maximum daily loads development, often require quantification of flow in small first-order and second-order streams. Frequently, stream-gaging techniques are implemented in flows that are below the manufacturer’s recommended minimum velocity. A comparative analysis of the accuracy of current technologies used in NPS pollution stream-gaging applications and their applicability in low-flow conditions was conducted. Nine stream-gaging methods were evaluated for their field and laboratory performance and control structures were used as the statistical control. Analysis of the field investigation data indicated that Marsh McBirney current meter and the One-orange method were the most accurate in the field while the results of the laboratory experiments found that the Starflow acoustic Doppler and Valeport Braystoke current meter performed best among the 10 methods. Overall, the Marsh McBirney and Valeport Braystoke current meters exhibited the best performance for both field and laboratory situations.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Summary and Conclusions
  7. Acknowledgments
  8. Literature Cited

Open channel flow measurement is necessary for a wide variety of water management applications including water supply management, pollution control, irrigation, flood control, energy generation, and industrial uses (Herschy, 2002). A current meter with rotating propeller or cups has been the basic method of measuring flow velocities in open channels for over 100 years (Costa et al., 2000). The velocity of flow at a point is proportional to the rate of rotation of the rotor during a measured period of time (Rantz, 1982). Depth and velocity measurements are taken at multiple points across a channel, and these measurements are used to calculate the stream discharge. A relationship between the discharge and depth is used to develop a rating curve (Costa et al., 2000). Other devices more recently used to measure stream velocity include electromagnetic current meters and acoustic Doppler devices. Comparisons of some methods have been performed by various researchers (Fulford et al., 1994; Thibodeaux, 1994; Fulford, 2001); however, none cover the broad range of technologies currently employed to measure low flows.

Fulford (2001) compared four types of current meters, the Price Type-AA, Price Pygmy, Marsh McBirney 2000, and Swoffer 2100, and six of each type for variation of meter performance within each type and percent velocity error. Steady flow velocities, ranging from 0.25 to 8.0 ft/s, were tested. The study found that the Price types performed more consistently and within accuracy limits over a range of test velocities compared with other current meters. Price meters also measured flow velocities with significantly better accuracy at velocities <0.5 ft/s while the March McBirney most frequently exhibited measurement errors exceeding accuracy specifications at velocities <0.5 ft/s.

Thibodeaux (1992) found that the majority of the research on current meters has been performed on mechanical current meters. Very little work has been reported on the flow measuring characteristics of electromagnetic current meters (Thibodeaux, 1992) or the performance characteristics of acoustic point velocity meters. Thibodeaux (1992) noted poor current meter alignment in the Price Type-AA which created internal friction and caused the meter to underestimate velocity. Ackers et al. (1978) found that even though electromagnetic and ultrasonic techniques had been developed for use in small rivers and tributaries, these applications are less suited to the technology than when the section is fixed and constant.

The demand for accurate streamflow measurement has increased along with rising concerns about nonpoint source pollution (NPS). Total maximum daily loads are being developed to assess water quality problems, identify pollution sources, and determine pollution reductions needed to restore and protect rivers, streams, and lakes (USEPA, 2007). Stream discharge measurements are used to calibrate models that estimate the necessary reductions in both point and nonpoint source pollutants. Because of the lack of observed flow data, there is considerable uncertainty in the pollutant allocation process (NRC, 2002), particularly in the first-order and second-order stream reaches dominated by NPS pollution. In small streams located in upland watersheds, frequently flows are below minimum recommended velocities for common stream-gaging methods, thus it is critical to evaluate the applicability of various gaging methods to these conditions.

Open channel flow measurement techniques employed today range from the standard current meter to complex ultrasonic methods. The goal of this study was to conduct a comparative analysis of the accuracy of current technologies used in stream-gaging applications and their applicability during low-flow conditions. The objectives were to compare the accuracy of various methods for estimating stream discharge in the small first-order and second-order streams and in a controlled laboratory environment.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Summary and Conclusions
  7. Acknowledgments
  8. Literature Cited

Both field and laboratory investigations were conducted to compare the accuracy of various methods for estimating stream discharge. Field investigations were conducted to evaluate the performance of various techniques for estimating the stream discharge in the first-order and second-order streams, and laboratory tests were conducted to examine the techniques for estimating stream discharge in a controlled environment. This study compared four current meters: Price Type-AA, Price Pygmy, Valeport Braystoke Model 001 (Valeport Developments, Dartmouth, Massachusetts), Marsh McBirney Flo-mate model 2000 (Marsh McBirney Inc., Frederick, Maryland); two acoustic Doppler devices: Isco 4150 Flow Logger with standard velocity-area sensor (Isco, Inc., Lincoln, Nebraska) (Isco, 1993) and Starflow model 6526B Ultrasonic Doppler Instrument with Micrologger (Unidata, Australia); the One-orange float method (Christensen, 1994); a Global Water flow probe FP101 (Global Water, Inc., Gold River, Colorado); and a dilution technique (Rhodamine WT dye, slug injection). Stream gages were not replicated; however, prior to the collection of discharge measurements data, equipment was field tested to ensure proper operation (Mitchem, 1999). A pressure transducer and a hydraulic structure equipped with an FW-1 stage recorder and potentiometer were used as the field investigation statistical controls. The laboratory trials were conducted in a hydraulic flume and a thin plate v-notch weir was installed to measure discharge and served as a control.

Field Studies

Field investigations were performed at the outlets of two small agricultural watersheds, selected because of the existence of functioning control structures at the sites. To reliably measure the stage in each stream, an FW-1 chart stage recorder and potentiometer were installed in existing stilling wells. Field sites were selected based on the presence of the artificial control structures, however, at these two locations, the stream sections were not always straight, velocities were usually below 0.5 ft/s (0.015 m/s), and stream bottoms were not uniform. Significant accumulation of sediment was present at both sites. Each method was tested in random order at each location.

The first gaging site is located at the outlet of the Crab Creek watershed in Montgomery County, Virginia (Figure 1). The drainage area of the watershed is 318 ha (786 acres) with an average slope of 12% (Mitchem, 1999). The U.S. Department of Agriculture (USDA) instrumented and monitored this watershed from August of 1957 until 1979. The installed flow control structure is a double 6 ft by 6 ft rectangular culvert modified with a twin Virginia V-notch weir. A stilling well with a small instrument shelter and a staff gage are still present at the gaging station. The sections at Crab Creek were fairly uniform, with a few large stones and the stream bottom consisted mostly of smaller cobbles. Multiple measurement sections were used to facilitate the simultaneous use of multiple measurement techniques. Current meter data were collected at the sections, above and below the riffle (Figure 1). Acoustic Doppler devices and the One-orange method were tested above the riffle, while the Global flow probe was tested below the riffle.

image

Figure 1.  Plan View of the Crab Creek Gaging Station (not to scale).

Download figure to PowerPoint

The second gaging site is in Thorne Springs Branch watershed, located in Pulaski County, Virginia (Figure 2). The drainage area of this watershed is 1,236 ha (3,054 acres) and the average slope in the watershed is 10% (Mitchem, 1999). Monitoring of this watershed began by the USDA in June of 1957 and in 1969 station monitoring responsibilities were passed to the U.S. Geological Survey (USGS). The existing hydraulic control at this watershed is a large concrete dam with a broad-crested v-notch weir. In addition to the control structure, the site is equipped with a stilling well and a staff gage. The stream above the dam was more favorable for measurement, but deep sediment accumulation prevented wading. Stream measurement sections were instead taken approximately 21.3 m downstream of the dam. Below the dam, the stream dropped about 1.52 m into a large pool and eddy currents developed as water flowed out of the pool and into a culvert. A curved transition region was constructed from rocks in and around the stream to straighten the flow lines. Flow was allowed to stabilize for two days prior to measuring discharge at the site and the site was inspected for flow into and out of the stream between the control structure and the measurement sections located below the culvert. The best measurement section for each device was selected and is shown in Figure 2. The Global flow probe was used to estimate discharge upstream of the dam in a section of the stream with higher flows. Acoustic Doppler flow measurement devices were tested at the downstream end of the transition structure and the One-orange method was tested just downstream of the culvert. Current meters were tested at the mouth of the culvert because this section had the most uniform stream bed and the best distribution of discharge across the section. The measurement section contained flow that was fairly uniform and free of eddies, slack water, and excessive turbulence, but large rocks were present in and around the section.

image

Figure 2.  Plan View of the Thorne Springs Branch Gaging Station (not to scale).

Download figure to PowerPoint

Flow Measurement Techniques

Current Meters.  The procedures described by the USGS (Rantz, 1982) were followed in collecting data by all current meters used in this study. Because of the small size of the streams, some deviation from the USGS procedures was necessary; most notably a reduction in the number of measurement verticals. Rantz (1982) stated that at least 25 verticals are desirable for a discharge measurement, with each vertical representing <10% of the total flow. A semi-permanent tag line was placed at each measurement section and observation verticals were established every 0.076 m (0.25 ft). The number of verticals and velocity measurements varied by site and meter but a minimum of 30 discharge estimations were performed with each device at each of the two field sites to provide adequate data for statistical analyses (G. Steeno, 1999, personal communication). At some verticals a discharge measurement could not be collected (usually due to low flows) but if flow was encountered between verticals, additional discharge measurements were obtained. Velocity was measured using the 0.6-depth method for the all measurements except for the Pygmy current meter at Thorne Springs Branch gaging station. In this case, the flow velocities were very low (0.15 ft/s and less) at the 0.6-depth; thus the 0.2-depth method was employed. Area was defined using the midpoint method, where the vertical is considered the center of a partial area of discharge (Pierce, 1941; Rantz, 1982). The relationship between the mean velocity and the velocity at 0.2-depth can vary with depth and discharge and this method is not as reliable as the 0.6-depth or two-point methods if flow conditions are equally favorable for all methods (Rantz, 1982). At each vertical, the current meter was given a minimum of three seconds to adjust to the water velocity. The device was then operated for at least 40 s to obtain a velocity at the point of measurement and filter out potential surges in current velocity.

Acoustic Doppler Devices.  Mounting plates were fabricated to support the acoustic Doppler devices (Mitchem, 1999) and installed in the stream prior to installation of the sensors. The measurement section was surveyed to develop stage-area tables required for operation of each of the Doppler devices. The acoustic Doppler devices were allowed to operate and collect data for at least 1.5 hours. During this observation time the device was monitored continuously and a sample of 30 random discharge readings were recorded. The gage heights corresponding to the recorded stream discharges were paired for analysis.

Float Method.  The One-orange method was used to measure discharge following the procedure presented by Christensen (1994). The orange was held to the bottom of the stream at the deepest vertical in the section. The orange was released and the time and horizontal distance of travel after emergence of the orange were measured and recorded. The orange was then allowed to continue traveling until it reached a total of 16 feet downstream and the total elapsed time of travel to the downstream section was recorded. The maximum velocity, mean velocity, and discharge were then derived as described by Christensen (1994). The method was repeated 30 times to yield 30 discharge estimates for analysis.

Global Flow Probe.  The manufacturer’s guidelines (Global Water, 1997) for use of the Global Water flow probe FP101 were followed to estimate stream discharge. This involved measuring the cross-sectional area of the measurement section and then operating the flow probe throughout the section for at least 40 s until a steady value was recorded. The resulting average velocity was multiplied by the section area to obtain an estimate of the discharge. The flow probe was used to estimate the discharge at least 30 times at each site. The flow probe did not operate in small flow velocities, and appeared to only calculate an average velocity when the propeller was spinning; resulting in an overestimation of the average flow velocity. At Thorne Springs Branch, the flow probe was used upstream of the dam, in a section of the stream with faster flow.

Dilution Technique.  The streamflow discharge was determined by dilution of Rhodamine WT 20% that was injected into the stream by slug injection. The measurement reach was selected to avoid dead areas that would prevent elongation of the tracer cloud. Ideal mixing reaches have sections that are narrow and deep. The optimum reach lengths were calculated (Kilpatrick and Cobb, 1985) to be 65 feet and 175 feet at Crab Creek and Thorne Springs Branch, respectively. The recommended peak concentration of rhodamine WT at the sampling section was 10-20 μg/l and the injection solution was prepared by mixing stream water with the rhodamine WT 20% (Kilpatrick and Cobb, 1985). The injection solution was then poured into the centroid of flow in one slug at the upstream end of the designated reach. Sampling was performed at three points across the measurement section, representing approximately one-third of the stream discharge. Samples were collected every 5 s prior to detection of the dye cloud and until the peak passed. Sampling frequency was reduced but continued for 2-3 times the amount of time it took the peak to arrive at the sampling section. This ensured total recovery of the dye by fully accounting for the low concentrations on the concentration-time curve.

A Turner model 111 fluorometer (Testwave LLC®, Sparks, Nevada) was equipped with a U-type ultraviolet lamp. The excitation (primary) filter selected was a 546 nm filter and the emission (secondary) filter combination consisted of a 325-700 nm band pass filter in conjunction with a > 570 nm sharp cut filter. This filter combination was chosen specifically for use with Rhodamine WT, while the lamp is appropriate for a broad range of fluorometric analyses. The fluorescence readings were converted to dye concentrations and plotted against elapsed time, resulting in time-concentration curves (Mitchem, 1999). The mean area under the three time-concentration curves was then determined to calculate stream discharge.

Pressure Transducers and Potentiometer.  Two pressure transducers were installed at the Crab creek site. One transducer was placed directly in the stream adjacent to the stilling well and the other inside the stilling well to evaluate the effects of the stilling well in damping the change in water level. A stainless steel shroud was fabricated to protect each of the pressure transducers while in service. Only one pressure transducer was installed in the stilling well at the Thorne Springs Branch gaging station, because sediment accumulation prevented the installation of a transducer in the stream above the weir.

After the transducers were mounted in the stream and stilling wells, the devices were connected to a Campbell Scientific 21X datalogger (Campbell Scientific, Logan, Utah). The potentiometer attached to the FW-1 stage recorder was also connected to the datalogger. Following installation, each device was field calibrated to match the stage reading of the staff gage and stage data were recorded by the datalogger for at least a week. Data were downloaded using PC208W software designed to interface with the Campbell dataloggers (Campbell Scientific, 1996). A sample of 30 data points were randomly extracted from the collected data for comparison. Stage values obtained with the pressure transducers were compared to those obtained with the potentiometer attached to the FW-1 since the potentiometer was considered to be the standard.

Recommended Operating Velocities

The various methods investigated in this study have different recommended operating requirements. These guidelines are typically conservative estimates, so caution is suggested when using the devices in streams with velocities below these limits. Table 1 lists the suggested velocity operating ranges for each method.

Table 1.   Suggested Minimum and Maximum Operating Flow Velocities for Each Method.
MethodMinimum Velocity (ft/s)Maximum Velocity (ft/s)
  1. *A velocity value could not be found.

Marsh McBirney−0.5119.991
Price Type-AA0.22*
Pygmy0.22*
Valeport Braystoke0.5316.43
Global flow probe0.34*
One-orange method**
Isco acoustic Doppler**
Starflow acoustic Doppler0.066516.45

Laboratory Studies

While field investigations provide a more realistic setting for evaluating various methods of streamflow measurement, they also expose these methods to rapidly changing velocities and flow conditions. Laboratory experiments are necessary to assess the performance of all the methods under nearly identical flow conditions. All methods, except for the tracer dilution method, evaluated in the field investigations were also studied in laboratory experiments.

The laboratory trials were conducted in a 0.305 meter (1-foot) hydraulic flume (Engineering Laboratory Design Inc., Lake City, Minnesota). A thin plate v-notch weir was installed at the far end of the flume to measure discharge and served as a control. A point gage was used to measure the water depth in the flume during the experiment. The flume was set at a given discharge and allowed to stabilize for approximately two minutes before any measurements were collected. After the flow had stabilized, each method was randomly employed to determine the discharge. Three discharge readings were recorded for each device, and the mean of these three discharge values was used in the analysis. The flow in the flume was increased at random increments, fourteen times, until the maximum discharge of the flume was reached. After the maximum discharge of the flume was achieved, the discharge rate in the flume was decreased 15 times in random increments until there was no flow in the flume. The control discharge values ranged from 0 to 0.591 ft3/s (0-0.0167 m3/s), while the average velocity in the flume ranged from 0 to 0.498 ft/s (0-0.152 m/s).

Data Analyses

The field investigation data were collected over a three-month period, including one month at Crab Creek and two months at Thorne Springs Branch. During that time period, there was very little significant rainfall and little variation in discharge was observed. Therefore, the majority of the data points fall within a narrow discharge range. The percent relative error was calculated to compare the performance of the various methods to the control discharge at each site. The percent relative error was calculated using Equation 1.

  • image(1)

where Qobs is observed discharge, in cfs; Qcontrol is control discharge, in cfs.

Descriptive statistics were calculated on the percent relative error values, including the mean, median, standard deviation, and range of the percent relative error values, and a confidence interval for each stream-gaging method at each field investigation site. A one-way analysis of variance combined with Duncan’s multiple range test (Ott and Longnecker, 2001) was used to test in pairwise fashion whether the field trial mean residuals of the different methods were significantly different from each other. A threshold of < 0.05 was used to assess statistical significance.

The laboratory portion of the study was analyzed by regression analysis and analysis of covariance (Walpole and Myers, 1993). The analysis of covariance was performed to improve evaluation of the regressions between the measured and control discharge values. The experiment was a completely randomized design with multiple treatments and a single covariate, the control discharge. Also included in the analysis was the test of fixed effects to estimate the slope of the regressions between the treatment and control data. The slope of the regressions between the treatment and control data, the Y-intercept and its p-value, and the R2-values were compared to identify the closest relationship between the measured and control discharges.

Results and Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Summary and Conclusions
  7. Acknowledgments
  8. Literature Cited

Field Studies

The mean control discharge measured at the Crab Creek gaging station was 0.402 cfs (0.0114 m3/s), with values ranging from 0.182 to 1.196 cfs (0.00515-0.0339 m3/s). The average measured velocity for the various methods at Crab Creek ranged from 0.104 ft/s (0.032 m/s) to 0.403 ft/s (0.123 m/s). At Thorne Springs Branch, the mean control discharge was 0.377 cfs (0.0107 m3/s), with values ranging from 0.348 to 1.198 cfs (0.00985 to 0.0339 m3/s). The average measured velocity for the various methods at Thorne Springs Branch ranged from 0.075 ft/s (0.023 m/s) to 0.563 ft/s (0.172 m/s). Similar discharge rates obtained for the two field sites would seem to indicate that the streams were very similar. However, the stream cross-section at Thorne Springs Branch was much larger than that of Crab Creek. Because of the limited range of control discharge values, the mean residual and percent relative error were calculated to compare performance of the various methods to the control discharge at each site. Table 2 presents the mean residual values and descriptive statistics associated with the percent relative error. The mean and median values give an idea of the central tendency of the data while the standard deviation and range indicate the degree of scatter the data exhibit.

Table 2.   Descriptive Statistics of Percent Relative Error Data and Mean Residual in Field Trials Listed by Site and Method.
 Percent Relative ErrorMean Residual cfs (m3/s)
Mean1 (%)Median (%)Standard Deviation (%)Range
Min (%)Max (%)
  1. 1Positive values indicate overestimation of discharge, and negative (−) values indicate underestimation of discharge.

  2. 2Significantly different according to Duncan’s multiple range test at < 0.05.

  3. 3Pressure transducers were analyzed for statistical differences between the two transducers installed at Crab Creek but were not analyzed for statistical differences between other measurement techniques.

Crab Creek
Price Type-AA60.160.930.71.9116.90.131 (0.0037)e2
Pygmy21.025.916.9−29.345.00.055 (0.0016)f
Valeport Braystoke46.946.313.520.377.40.135 (0.0038)e
Marsh McBirney23.724.020.5−17.774.20.075 (0.0021)ef
Global flow probe137.0143.028.477.1182.30.525 (0.0149)b
One-orange36.535.113.212.273.00.071 (0.0020)ef
Isco109.3116.743.3−76.4146.80.934 (0.0264)a
Starflow59.957.719.932.2107.10.389 (0.0110)c
Dilution52.639.846.5−0.2126.50.278 (0.0079)d
Pres. transducer 13.50.05.3−2.611.80.008 (2.3 × 10−4)z3
Pres. transducer 2−3.9−4.21.1−5.7−1.60.007 (2.0 × 10−4)z
Thorne Springs Branch
Price Type-AA−36.4−43.040.1−100.051.40.171 (0.0048)e2
Pygmy−70.7−84.424.2−96.2−21.90.270 (0.0077)cd
Valeport Braystoke−27.9−26.228.3−87.817.40.110 (0.0031)f
Marsh McBirney−19.8−24.030.3−61.565.00.124 (0.0035)ef
Global flow probe138.3141.417.199.0175.30.506 (0.0143)a
One-orange35.632.017.79.669.20.132 (0.0037)ef
Isco117.2122.226.789.1163.30.412 (0.0117)b
Starflow64.264.519.314.399.50.237 (0.0067)d
Dilution65.077.643.4−1.8101.30.317 (0.0090)c
Pres. transducer10.13.315.5−3.140.00.003 (8.5 × 10−5)3

The Global flow probe consistently resulted in overestimation of discharge values, as almost all the data collected by this device indicate a relative error >100%. While using this device in the field, it was noted that the probe only registered velocity in very fast flow. Because only the registered velocities were used to estimate the average velocity, the entire cross-section was overestimated. The acoustic Doppler devices also consistently overestimated discharge values and did not perform well at either field study site. The relative errors for the Isco acoustic Doppler at the Crab Creek site ranged from −76% to 146%; however, by removing three observations, the minimum relative error is 109%. The relative errors at the Thorne Springs Branch site ranged from 89 to 163%. The method prescribed by the manufacturer directs the user to position the sensor in the middle of the section, where velocity is the highest. The velocity recorded by the sensor is then multiplied by the total area based on the measured stage. In natural streams, the edges often contain “dead” spots where there is very low flow velocity. The prescribed method resulted in a consistent overestimation of discharge by the device.

The mean percent error values for each method used at the Crab Creek gaging station are shown in Figure 3. The error bars represent the 95% confidence interval for the population mean. The 95% confidence interval for the dilution method is very broad and can be attributed to the small sample size during the field trials. The Isco and Global devices have the largest sample mean percent relative errors among all devices. The pressure transducers installed to assess the difference in accuracy between a transducer installed in a stilling well and one installed in the stream exhibited a very small relative error. The pressure transducer installed in the stilling well slightly overestimated the stage compared to the float and weight recorder equipped with a potentiometer. The transducer installed in the stream underestimated the stage, with a small percent relative error. The pressure transducer in the stilling well had a mean percent relative error of 3.5 (Table 2), while the transducer in the stream had a mean percent relative error of −3.9.

image

Figure 3.  Mean Percent Relative Error With 95% Confidence Intervals for the Crab Creek Site.

Download figure to PowerPoint

At Crab Creek, the mean residuals of the Pygmy current meter, Marsh McBirney current meter, and One-orange method were not statistically different from one another (Table 2). These methods were also the most accurate with mean residual differences <0.08 cfs (0.0022 m3/s). The Price Type-AA and Valeport BFM001 current meters were not significantly different from the One-orange method and the Marsh McBirney current meter, but had slightly higher values of 0.131 cfs (0.0037 m3/s) and 0.135 cfs (0.0038 m3/s), respectively. The mean residuals for the remaining methods were substantially higher, and were significantly different from the other methods. The mechanical current meters performed well at the Crab Creek site because the stream velocities encountered were more suitable for their use. The Marsh McBirney current meter seemed to be more sensitive in the lower velocities at the edges of the stream sections. The One-orange method performed very well, given the low level of technological sophistication associated with the method. The dilution method had the next highest mean residual value at 0.278 cfs (0.0079 m3/s). The reach length may have been too short for complete mixing, although the reach length used was calculated as prescribed by Kilpatrick and Cobb (1985).

At the Thorne Springs Branch the mean percent relative error for each of the current meters was negative, indicating that the current meters generally underestimated the discharge at this site (Figure 4). This could likely be attributed to low flow velocities at Thorne Springs Branch. At the Crab Creek gaging station, which generally exhibited higher flow velocities, the current meters overestimated the discharge. All other methods overestimated the discharge for both watersheds and had similar sample means and population mean confidence intervals, implying that the performance of the methods was consistent across the two sites. The transducer error at Thorne Springs Branch was considerably higher than at Crab Creek, registering a mean relative error of 10.1% and maximum relative error of 40% (Table 2).

image

Figure 4.  Mean Percent Relative Error With 95% Confidence Intervals for the Thorne Springs Site.

Download figure to PowerPoint

At Thorne Springs Branch, the Valeport Braystroke current meter, the Marsh McBirney current meter, and the One-orange method performed best with mean residuals ranging from 0.11 cfs (0.0031 m3/s) to 0.13 cfs (0.0037 m3/s, Table 2). The Price Type-AA current meter was not significantly different from the One-orange method and Marsh McBirney current meter, but had a slightly higher mean residual value of 0.17 cfs (0.0048 m3 s−1). These four were significantly different from the remaining methods. The lower flow velocity at Thorne Springs Branch had a significant effect on the performance of the Pygmy current meter, drastically increasing the mean residual compared to Crab Creek. The measurement section at Thorne Springs Branch was less desirable than Crab Creek, which could have contributed to the higher mean residual values for some methods. The dilution method had a higher mean residual value at Thorne Springs Branch than at Crab Creek, probably due to the irregular streambank shape caused by cattle entering and exiting the stream, which might have created small eddies that impeded the travel of the dye.

Considering the data collected at both sites, the Marsh McBirney and One-orange methods performed the most accurately under the range of conditions encountered. If a stream has a slightly higher flow velocity, the Pygmy current meter would be preferred over these two methods. The Valeport Braystroke current meter would be the preferred mechanical current meter under conditions of extremely slow flow, since it performed best in the slow velocities at Thorne Springs Branch. The Marsh McBirney current meter exhibited the best accuracy overall at both field study sites. The One-orange method also performed very well at both sites. Neither of the acoustic Doppler devices seemed to perform well in the field study, likely due to the procedure prescribed by the manufacturer.

Laboratory Studies

Laboratory experiments provide an assessment of the accuracy of the methods evaluated in the field under controlled conditions. Table 3 presents the descriptive statistics of the percent relative error values. Each method was tested at very low velocities to identify the lowest measurable flow velocity. At low velocities, some of the devices were unable to register the flow in the flume. To ensure that the performance of each method was evaluated within its operating range, nonregistering data points were omitted from the analysis.

Table 3.   Descriptive Statistics of Percent Relative Error Data in Laboratory Trials Listed by Stream-Gaging Technique.*
 Mean (%)Median (%)Standard Deviation (%)Data Range
Min (%)Max (%)
  1. *Data with nonregistering data points omitted.

Marsh McBirney−23.2−27.514.7−37.326.6
Price Type-AA−32.3−3014.5−540
Pygmy−43.1−45.215.8−67.90
Valeport Braystoke−25.2−27.38.7−37.10
Global−58.1−62.125.1−90.80
One-orange−56.1−58.122−84.93.3
Isco−27.5−29.217.4−67.447
Starflow−4.4−11.228.5−29104.1
Pressure transducer−0.6−0.60.7−2.10.5

The mean and median percent relative error values were negative for all methods, indicating that they consistently underestimated the discharge in the flume. Relative error values were lower than those obtained in the field trials, except for the One-orange method. There is no clear reason for the higher percent error associated with the use of the One-orange method in the laboratory; however, this method does involve characterization of the stream cross-section, and the flume has a simple rectangular profile. The pressure transducer, as in the field trials, exhibited a very low percent relative error.

Figure 5 shows the mean percent relative error for each device in the laboratory trial, along with the 95% confidence interval. The Starflow acoustic Doppler device is the only velocity-measuring method whose 95% population mean confidence interval contains the zero value. This implies that this device is the only method tested that is capable of estimating the discharge with minimal error under the given flow conditions.

image

Figure 5.  Mean Percent Relative Error With 95% Confidence Interval for the Laboratory Studies.

Download figure to PowerPoint

The data collected in the laboratory experiment contained a discharge value for each method corresponding to a common flow discharge. The correlation between the control and measured discharge values was evaluated. The plots of the laboratory data include the ideal trend line to provide a visual comparison of the data with ideal conditions. Figure 6 shows the measured discharge values plotted against the control discharge for all methods evaluated in the laboratory portion of this study. All current meters underestimated the discharge in the flume when the control discharge was between 0.3 cfs (0.009 m3/s) and 0.55 cfs (0.0156 m3/s). One can also see that none of the current meters, except the Marsh McBirney, were able to measure discharge below 0.2 cfs (0.0057 m3/s). This lack of ability to measure very low velocity and low discharge rates was also observed in the field, which confirms the assertion that slow flow at the Thorne Springs Branch site was responsible for underestimation of discharge by the current meters.

image

Figure 6.  Comparison of Laboratory-Obtained Measurements to Control Discharge Data Represented by an Ideal Trend Line.

Download figure to PowerPoint

As the discharge rates increased, the variability in the discharge data collected by the One-orange method increased, as did the gap between the data and ideal trend line. Most of the data points for the Global flow probe lie on the y-origin, indicating the device did not measure flow velocity for most of the control discharge range. The minimum recommended operating velocity of this device is 0.3 ft/s (0.09 m/s) (Table 1). The majority of the flow velocity values in this experiment were below 0.3 ft/s (0.09 m/s), which explains the Global flow probe’s failure to register velocity for much of this experiment. When the probe did register velocity, the variability was still quite high. The data for the Starflow device seem to follow the 1:1 trend line better than any other method used in the study. The Starflow data points fell on both sides of the line for the majority of the control discharge range. The data for the Isco device diverged from the ideal trend line as the discharge increased. The Isco data exhibited a linear trend with little variability, but the slope of the trend line was <1.

Regression analyses were conducted between the measured discharge data by each device and the corresponding control discharge data. Measurements where no flow velocity registered were omitted from the data prior to the regression analysis to avoid skewing regression results and to create a regression representative of the operating range of each method. The ideal result for each analysis was a linear equation with an intercept of zero and slope of one. The p-values presented in Table 4 statistically reject the null hypothesis that the Y-intercept is equal to zero when < 0.10. A higher p-value, on the other hand, indicates a greater chance that the Y-intercept is equal to 0, and a failure to reject the null hypothesis. The standard error in Table 4 gives an indication of the reliability of the regression and is a measure of the scatter of the data about the regression line. Greater error implies more variability of the data about the regression.

Table 4.   Data Describing Regressions Between Measured and Control Discharge by Technique.
 Y-Interceptp-ValueSlopeStandard Error
Marsh McBirney0.0050.5590.7050.137
Price Type-AA0.1290.0071.0440.024
Pygmy0.1560.0001.1140.132
Valeport0.0480.0351.1860.067
Global0.2390.0591.4020.588
One-orange0.0820.2182.1440.489
Isco−0.0130.2071.4980.039
Starflow−0.0040.8961.1300.078
Pressure transducer−0.0040.8021.0100.013

The pressure transducer data resulted in an almost ideal regression with a Y-intercept of −0.004, a slope of 1.010, and a standard error of 0.013. Among the velocity-area methods of discharge measurement, the Starflow acoustic Doppler device exhibited a trend closest to the ideal conditions, which confirms the visual observations in Figure 6. The following methods also have Y-intercepts that are not statistically different than zero: Marsh McBirney current meter, the One-orange method, and the Isco acoustic Doppler device. The high p-values (> 0.5) of the Marsh McBirney current meter, Starflow acoustic Doppler device, and pressure transducer provide strong evidence that the Y-intercept is not statistically different than zero. All the current meters evaluated exhibited a regression slope close to 1, ranging from 0.705 to 1.186 (Table 4). Aside from the current meters and the Starflow acoustic Doppler device, the remaining velocity-area methods resulted in a slope >1.4. The One-orange method exhibited the highest slope value (2.144). Among the velocity-area discharge measurement methods, the lowest standard error values resulted from the Price Type-AA current meter (0.024), and the Isco acoustic Doppler device (0.039). The Global flow probe and the One-orange method had the highest values of standard errors (0.588 and 0.489, respectively), indicating high variability in the data about the regression line.

The regression parameters of the various treatments were compared to the ideal conditions (slope of 1, intercept of 0) and the 95% confidence intervals for the differences between the treatment regression parameters and the ideal regression parameters were calculated. The methods for which the confidence intervals included a zero value for the difference between the treatment and control slope parameter include the Price Type-AA, Pygmy, and Valeport Braystoke current meters, and the Starflow acoustic Doppler device. Similarly, the methods whose 95% confidence intervals for difference of intercept include the value of zero include Marsh McBirney current meter, Valeport Braystoke current meter, Starflow acoustic Doppler device, and Isco acoustic Doppler device. This implied that these methods could have the ideal regression parameters stated.

To improve evaluation of the regressions between the measured and control discharge values an analysis of covariance was also performed (Walpole and Myers, 1993). The experiment was a completely randomized design with multiple treatments and a single covariate, the control discharge. Also included in the analysis was the test of fixed effects to estimate the slope of the regressions between the treatment and control data. The Price Type-AA current meter, Pygmy current meter, Valeport Braystoke current meter, and the Starflow acoustic Doppler device exhibited slopes nearest to one (Table 4). Thus, these methods most closely reflect the performance of the thin-plate weir used as the control in this experiment. The test of fixed effects also implied that the slopes of these methods most closely match the slope of the control.

To further compare the regressions for these four methods (Price Type-AA, Pygmy, Valeport Braystoke, and Starflow), the regression parameters were compared to the control regression parameters and ideal regression conditions by calculating the difference between the slopes of the method and control regressions (Table 5). Next, the Y-intercept and its p-value were compared and the results indicated that the Valeport Braystoke current meter and the Starflow acoustic Doppler devices performed better than the Price Type-AA and Pygmy current meters. Finally, the R2-values were evaluated to determine the relationship between the two variables in the regression (measured and control discharge). The R2-values for the Valeport Braystoke and Starflow devices were 0.917 and 0.874, respectively. These values are very similar and when considered with the previous two criteria they do not distinguish the two devices from one another.

Table 5.   Laboratory Data Analysis of Covariance Summary.
 Difference in Slope (method-control)Y-Interceptp-Value1R2
  1. 1p-Value <0.05 indicates rejection of the null hypothesis that the Y-intercept is not statistically different from zero.

  2. 2Data were not listed for methods eliminated from consideration as “most accurate” using the previous criterion (reading the table left to right).

Marsh McBirney−0.2562
Price Type-AA0.074−0.2090.003
Pygmy−0.051−0.1760.000
Valeport Braystoke−0.187−0.0300.2230.917
Global−0.507
One-orange−0.929
Isco−0.369
Starflow−0.056−0.0330.3450.874

Based on the slope, Y-intercept, and R2-value, among the velocity-area discharge measurement devices tested, the Starflow acoustic Doppler device is the best-performing method for measuring discharge. The Y-intercepts and R2-values are similar for the Starflow (0.874) and Valeport Braystoke (0.917) devices, but a higher slope (0.94) was obtained using the Starflow, compared with the 0.81 slope for the Valeport device.

From the results of this laboratory experiment, a few observations can be made. The Marsh McBirney performs very reliably, even in extremely slow-moving flows. However, for the flow velocities in the range of those seen in this experiment (<0.50 ft/s or <0.150 m/s), the Valeport Braystoke current meter and the Starflow acoustic Doppler device are the most accurate among the devices tested. Some devices, such as the Starflow acoustic Doppler device, can safely be used below their recommended minimum flow velocities; however, the Global flow probe should not be used below the recommended minimum velocity.

Summary and Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Summary and Conclusions
  7. Acknowledgments
  8. Literature Cited

A comparative analysis of the accuracy of technologies currently used in stream-gaging applications during low-flow conditions was conducted. Nine stream-gaging methods were evaluated for their field performance and laboratory performance and control structures were used as the statistical control. Field studies were conducted to provide a realistic setting for evaluating the various methods of flow measurement. Studies were conducted at two separate locations, but dry weather conditions resulted in similar flows during the three months of field testing. While field investigations evaluated the performance of various methods of streamflow measurement when exposed to rapidly changing velocities and flow conditions, laboratory experiments were necessary to assess the performance of all the methods under nearly controlled and identical flow conditions. All methods, except for the tracer dilution method, were also studied in laboratory experiments.

The Marsh McBirney current meter and One-orange method exhibited the highest accuracy in the field study, while the Starflow acoustic Doppler and Valeport BFM001 current meter performed best in the laboratory experiment. The Marsh McBirney, while not the most accurate, did perform well in the laboratory experiment as well. Similarly, the Valeport Braystoke current meter performed well in the field study, though it was not the most accurate method. The Pygmy and Price Type-AA current meters also performed well in both the field and laboratory. The One-orange method did not perform well in the laboratory experiment, while the Starflow acoustic Doppler device did not estimate discharge well in the field. The Global flow probe provided the least accuracy under the conditions encountered in both the field study and laboratory experiments. The Marsh McBirney and Valeport Braystoke current meters exhibited the best combined field and laboratory accuracy. However, water depth must also be considered when using the Valeport Braystoke current meter, given the large diameter of the sensor. Minimum water depth to ensure submersion of the propellor is 15.2 cm (6 inches), which may not always be possible in some small streams. Therefore, based on the results of this comparative study, the Marsh McBirney electromagnet current meter is recommended for flow measurements in small order streams.

While this study compared nine velocity-area discharge measurement techniques and one dilution technique, evaluation of additional methods in the field and laboratory experiments is recommended for future studies. Many additional techniques are available for measuring streamflow than those tested in this study, including the SONTEK Flowtracker. While some techniques are not commonly used in NPS pollution studies, evaluation of these might result in the discovery of new applications for a given method. Incorporating additional field sites would have allowed for the comparison of a broader range of flow conditions and collecting field data over a longer time period would have been more representative of annual climatic variation. However, results from this study should prove useful in identifying optimal stream-gaging techniques to employ during low-flow conditions.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Summary and Conclusions
  7. Acknowledgments
  8. Literature Cited

The authors would like to thank the Virginia Department of Conservation and Recreation, Division of Soil and Water Conservation for partial funding of this research. Eugene Powell and his staff at the Virginia Department of Environmental Quality and Harold Henderlite of the U.S. Geological Survey contributed greatly to this study. Thanks to Jan Carr, Julie Jordan, and Philip McClellan, and Julie Petruska from Virginia Tech for their assistance.

Literature Cited

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Summary and Conclusions
  7. Acknowledgments
  8. Literature Cited
  • Ackers, P., W.R. White, J.A. Perkins, and A.J.M. Harrison, 1978. Weirs and Flumes for Flow Measurement. John Wiley & Sons, Chichester, New York.
  • Campbell Scientific, 1996. CR10 Measurement and Control Module Operator’s Manual. Campbell Scientific, Inc., Logan, Utah.
  • Christensen, B.A., 1994. Velocity Measurements by the “One-Orange Method.” In : Fundamentals and Advancements in Hydraulic Measurements and Experimentation Proceedings of the Symposium on Fundamentals and Advancements in Hydraulic Measurements and Experimentation, (1994: Buffalo, New York), Clifford.A.Pugh (Editor). American Society of Civil Engineers, New York, New York, pp. 76-85.
  • Costa, J.E., K.R. Spicer, R.T. Cheng, P.F. Haeni, N.B. Melcher, E.M. Thurman, W.J. Plant, and W.C. Keller, 2000. Measuring Stream Discharge by Non-Contact Methods: A Proof-of-Concept Experiment. Geophysical Research Letters 27(4):553-556.
  • Fulford, J.M., 2001. Accuracy and Consistency of Water-Current Meters. Journal of the American Water Resources Association (JAWRA) 37(5):1215-1224.
  • Fulford, J.M., K.G. Thibodeaux, and W.R. Kaehrle, 1994. Comparison of Current Meters Used for Stream-Gaging. In : Fundamentals and Advancements in Hydraulic Measurements and Experimentation Proceedings of the Symposium on Fundamentals and Advancements in Hydraulic Measurements and Experimentation, (1994: Buffalo, New York), Clifford.A.Pugh (Editor). American Society of Civil Engineers, New York, New York, pp. 376-385.
  • Global Water, 1997. Global Flow Probe FP101-FP201 Instruction Manual. Global Water, Inc., Gold River, Colorado.
  • Herschy, R.W., 2002. The Uncertainty in a Current Meter Measurement. Flow Measurement and Instrumentation 13:281-284.
  • Isco, 1993. 4150 Flow Logger Instruction Manual. Isco, Inc., Lincoln, Nebraska.
  • Kilpatrick, F.A. and E.D. Cobb, 1985. Measurement of Discharge Using Tracers. In : Techniques of Water-Resources Investigations of the United States Geological Survey. Book 3, Chapter A16, United States Government Printing Office, Washington, D.C., 52 pp. Available at: http://pubs.usgs.gov/twri/, accessed September, 2008.
  • Mitchem, C.J., 1999. A Comparative Study of Stream-Gaging Methods Employed in Nonpoint Source Pollution Studies in Small Streams. M.S. Thesis. Virginia Polytechnic Institute and State University, Blacksburg, Virginia.
  • NRC, 2002. The Twenty Needs Report: How Research Can Improve the TMDL Program. NAS, Washington, D.C.
  • Ott, R.L. and M. Longnecker, 2001. Statistical Methods and Data Analysis. 5th. Wadsworth Group, Pacific Grove, California.
  • Pierce, C.H., 1941. Investigations of Methods and Equipment Used in Stream-Gaging: Part 1 – Performance of Current Meters in Water of Shallow Depth. USGS-Water Supply Paper 868-A. U.S. Geological Survey, Washington, D.C. Available at: http://pubs.er.usgs.gov/usgspubs/wsp/wsp868A, accessed September 2008.
  • Rantz, S.E., 1982. Measurement and Computation of Streamflow: Volume I. Measurement of Stage and Discharge. Geological Survey Water-Supply Paper 2175. U.S. Government Printing Office, Washington, D.C.
  • Thibodeaux, K.G., 1992. A Brief Literature Review of Open-Channel Current Meter Testing. In : Proceedings of the Hydraulic Engineering Conference 1992. American Society of Civil Engineers, New York, New York, pp. 458-463.
  • Thibodeaux, K.G., 1994. Review of Literature on the Testing of Point-Velocity Current Meters. U.S. Geological Survey Open-File Report 94-123. U.S. Geological Survey, Denver, Colorado, 44 pp. Available at: https://igsrglib03.er.usgs.gov/ipac20/ipac.jsp?session=1222B54V44878.59537&profile=r&source=~!horizon&view=subscriptionsummary&uri=full=3100001~!155337~!1&ri=2&aspect=subtab13&menu=search&ipp=20&spp=20&staffonly=&term=Thibodeaux%2C+Kirk+G.&index=&uindex=AL&aspect=subtab13&menu=search&ri=2#focus.
  • Unidata, 1995. Ultrasonic Doppler Instrument With Micrologger-User’s Manual. Lynn MacLaren Publishing, Australia.
  • USEPA, (U.S. Environmental Protection Agency), 2007. TMDL Definition -- What is a Total Maximum Daily Load (TMDL)? United States Environmental Protection Agency, Washington, D.C., Available at: http://www.epa.gov/owow/tmdl/intro.html, accessed September, 2008.
  • Valeport, 1995. The Basic Design and Operation Instructions. Valeport Developments, Dartmouth, Massachusetts.
  • Walpole, R.E. and R.H. Myers, 1993. Probability and Statistics for Engineers and Scientists. Macmillan Publishing Company, New York, New York.