Hach EZ7800 TOPHO

The EZ7800 TOPHO is different from all the other analyzers because it is the only analyzer designed to target TP. TP requires an additional digestion step to further break down phosphorus-containing compounds such as organics. It should be noted that the EZ7800 TOPHO does not measure TP directly because it has a 100-μm filter, and TP laboratory methods are unfiltered (Patton and Kryskalla, 2003).

The EZ7800 TOPHO uses the blue method and can measure a range of 0–5 milligrams per liter (mg/L) of phosphorus nearly every 23 minutes. For this study, the EZ7800 TOPHO sampled every 30 minutes. It is a cabinet-based analyzer measuring 68.8 centimeters (cm) tall, 46 cm wide, and 34 cm deep and weighing about 25 kilograms (kg), depending on analyzer options. For surface waters, the analyzer requires the addition of the EZ9020 fast loop filtration system. This filtration system is a continuous flow cell with a stainless steel 100-μm filter mounted inside of it and requires the use of compressed air, which is used to backflush the filter. The sample is drawn from a small, continuously circulating side vessel and sent to the EZ7800 TOPHO’s digestion chamber, where the persulfate digestion takes place. The digestion uses heat and pressure to boil the sample and reagent mixture. To cool the sample back down for absorbance measurements, the analyzer required the use of continuously circulating cold water. Typically, this would be done with continuously flowing tap water at a municipal plant; however, tap water was not available at the testing site. To compensate for this, a water chiller was installed to continuously supply cooling water to the EZ7800 TOPHO. Because of the large amounts of reagent required, the reagent was supplied by the operator and mixed according to manufacturer specifications (Hach, 2019). The analyzer has manual cleaning and calibration settings with a variety of communication options. For this study, Modbus TCP was used in conjunction with a Campbell Scientific CR6 datalogger.

Hach Phosphax sc

The Phosphax is a cabinet style orthophosphate analyzer originally designed for use at municipal water treatment facilities (Hach, 2021). The cabinet is 72 cm tall, 54 cm wide, and 37 cm deep; it weighs 29 kg and can be adapted to a wide variety of environmental conditions. For surface waters, the Phosphax requires the addition of the Filtrax, which is a filtration unit that feeds sample water to the Phosphax. The Filtrax is a separate cabinet measuring 53 cm tall, 43 cm wide, and 22 cm deep and weighing 22 kg. The Filtrax continuously switches between pulling samples from two different 0.45-μm filters equipped with water circulation. The maximum head difference between the filters and Filtrax is about 3 meters (m) with clean filters. This analyzer uses the yellow method and can sample every 5 minutes; however, for this study, it was set to a 15-minute interval. The analyzer has automatic cleaning that was run daily as recommended by the manufacturer. The Phosphax was controlled with a Hach SC4200 and equipped with Claros–Mobile Sensor Management, which allowed for cellular communication. For datalogging purposes, the Phosphax was additionally connected to a Campbell Scientific CR6 datalogger via Modbus TCP.

Sea-Bird Scientific HydroCycle-PO₄

The HydroCycle is different from cabinet-based analyzers. It is the only self-contained analyzer that does not have a separate filtration unit and is specifically designed to be deployed in situ. It is a wet-chemical analyzer that uses the blue method to supply continuous orthophosphate measurements. The instrument range is 0–0.3 mg/L of phosphorus, but the upper range can be increased to 1.2 mg/L if validated with user samples. The analyzer is cylindrical, measuring 56 cm tall with an 18-cm diameter weighing about 7.6 kg with the protective casing. The sample is drawn through the bottom of the analyzer through a coarse copper mesh screen and then through a 5–10-μm filter. This analyzer is the only one without a heated reaction chamber, and therefore, the only one with a variable reaction time with a mean of 15 minutes per sample. Additionally, the analyzer has a calibration standard check that it performs at an operator-specified interval. This standard check is for operator sample verification and does not change the output values of the analyzer. The calibration standard check was run at the manufacturer-recommended interval of every six samples. For this study, the Hydrocycle was run every hour to increase the lifespan of the filter and length between required maintenance. The analyzer was controlled via SDI–12 communications by a Campbell Scientific CR6 for this study.

YSI Inc. Alyza IQ PO4

The Alyza is an online, cabinet-style analyzer originally designed for the continuous monitoring of orthophosphate at municipal water treatment facilities (YSI, 2021). The cabinet is about 77.5 cm tall, 67.5 cm wide, and 44.5 cm deep. It weighs about 37 kg (depending on the equipment) and can be equipped for a wide variety of environmental conditions. The sample is drawn into the analyzer through an onboard pump via a large rectangular 0.45-μm filter that measures about 36.6 cm tall and 25.5 cm wide. With a clean filter, a total head difference of about 2 m is possible between the filter and main cabinet. The onboard pump runs continuously, pumping water through the filter into a small interior reservoir from which the sample is collected. The sample is analyzed using the yellow method by detecting the change in absorbance at a wavelength of 400 nanometers (nm). The analyzer can sample every 5 minutes but was set to a 15-minute interval for this study. It was also set to automatically clean and complete a two-point calibration daily. The analyzer is controlled by the IQ Sensor Net, which offers a wide variety of communication options. For this study, the analyzer was set up for Modbus RTU communications with a Campbell Scientific CR6.

Laboratory Trials

Laboratory trials were designed to verify each analyzer’s accuracy and precision across a large selection of known phosphorus concentrations. Additionally, standard tests with dissolved organics were performed because they are a known source of interference. Other laboratory tests, such as temperature or turbidity, were not performed. Although temperature affects the rate of the chemical reaction in the blue and yellow methods (Nagul and others, 2015; Baird and others, 20171), the three analyzers used in the laboratory trials all compensate for this by heating the reaction chamber; therefore, manufacturers can more reliably predict the rate of a sample reaction. Previous laboratory testing for the Sea-Bird Hydrocycle showed no clear pattern of the temperature’s effect on accuracy (Johengen and others, 2017). Turbidity was also not considered in laboratory testing because a wide range of turbidity values was expected in field trials.

Field Verifications

Field verifications took place at the Vermilion River near Danville, Illinois (USGS station 03339000) from November 20, 2020, to August 12, 2021. During this time, the analyzers were verified for accuracy, matrix considerations, and maintenance needs. Analyzers were compared to NWQL samples under a wide variety of site conditions to assess accuracy. The river matrix was verified based on turbidity effects on the analyzers. Maintenance needs were performed according to manufacturer recommendations, and data gaps were tracked.

Accuracy and Precision

Laboratory testing results of various phosphorus standard concentrations analyzed using the EZ7800 TOPHO, Phosphax, and Alyza are provided in table 5 and figure 2. Residuals were calculated by subtracting the mean concentration determined by NWQL from the concentration measured by the analyzer for the same standard. Accuracy and precision decreased as the phosphorus concentration increased for all analyzers. It should be noted that NWQL results for standards 2.5 and 5.0 mg/L of phosphorus are from TP tests because NWQL OP results were about 9 and 7 percent higher than TP for the 2.5 and 5 mg/L of phosphorus standards, respectively. Having OP results higher than TP is impossible by definition because OP is a part of TP, meaning that there was an unknown source of error in the NWQL OP analysis of the 2.5 and 5 mg/L of phosphorus standards.

As the EZ7800 TOPHO does not have a stated accuracy by the manufacturer, discrete samples are recommended to determine data quality. The EZ7800 TOPHO had substantial issues performing calibrations and deviated substantially from standard values. The range of absolute differences between the mean analyzer concentration and the National Water Quality Laboratory concentration were 0.003–1.187 mg/L of phosphorus. The problem was determined to be sample contamination from the previous run, meaning the analyzer was not draining or rinsing effectively. After working with the manufacturer, it was determined to be a design flaw. The sample contamination was most problematic with large, consecutive changes in concentration, such as the recommended 0 and 5 mg/L concentrations for calibration. The EZ7800 TOPHO uses a two-point calibration where the mean of three measurements of each standard is used to determine the linear regression coefficients. It is possible to manually change the EZ7800 TOPHO’s linear regression coefficients, and several attempts were made using the standard results to develop a linear regression as shown in figure 3. To minimize the effects of sample contamination, the analyzer was run for more than 12 hours, and the first sample results were removed as the analyzer stabilized. Linear regressions using the 0 and 5 mg/L standards as in the EZ7800 TOPHO’s calibration procedure or incorporating all standards produced regressions with various levels of bias depending on standard concentration. The only regression that caused all results to vary about zero was a quadratic regression, which is not how the analyzer was designed. All calibration regressions showed poor data quality and each regression showed bias, which is evident in figure 3. The factory calibration was kept as the final regression to avoid manipulating the analyzer in a way the manufacturer did not intend.

The manufacturer-stated accuracy for phosphorus concentrations measured by the Phosphax is 2 percent ± 0.05 mg/L. For all standards tested, the Phosphax measurements were within the manufacturer-stated accuracy. The range of absolute differences between the mean analyzer concentration and the National Water Quality Laboratory concentration were 0.003–0.008 mg/L of phosphorus. The analyzer had a downward pattern in median values as the concentration of the standard increased (fig. 2, table 5), which is likely due to a bias in the calibration values. This analyzer does not have automatic calibrations; only automatic cleanings that were performed daily. Additionally, it is evident that precision decreases as the concentration of the standard increases, whereas the standard deviation increased from 0.004 mg/L for the 0.06-mg/L standard to 0.047 mg/L for the 5.0-mg/L standard (table 5).

The mean concentrations (table 5) for standards 1.0 and 2.5 mg/L measured by the Alyza were higher than the manufacturer-stated accuracy of 2 percent or ±0.02 mg/L (whichever is higher) at 2.4 and 3.6 percent, respectively. However, for the 0.06-, 0.25-, and 5-mg/L standards, mean phosphorus concentrations measured by the Alyza fell within these bounds. It is likely that this extra source of error resulted from the daily cleaning and calibration cycles being completed between standard runs. Evidence for this conclusion is provided in figure 2, where no patterns exist, and none of the boxplots for any of the standards is associated with the zero line. This highlights the importance of the automated cleaning and calibration cycles and their effect on the data quality from the Alyza. The range of absolute differences between the mean analyzer concentration and the National Water Quality Laboratory concentration were 0.006–0.094 mg/L of phosphorus. No patterns in precision were observed with a standard deviation of 0.005 mg/L (table 5).

Dissolved Organic Matter Response

The effects of three concentrations of IHSS Suwannee River natural organic matter on analyzer-measured phosphorus concentrations are shown in figure 4 and table 6. The EZ7800 TOPHO and Phosphax indicated a decrease in precision as DOM concentrations increased, and their standard deviations were about three times higher on the 250-mg/L DOM standard concentration when compared to the 50-mg/L DOM standard concentration (table 6). It is unclear why the EZ7800 TOPHO appears to have performed better at the 50-mg/L DOM concentration, but it may be related to the number of samples processed; it had 12 runs completed for the 10-mg/L DOM standard compared to 32 runs for the 50-mg/L DOM standard. The standard deviation of the first 16 runs was more than double the latter 16 runs, indicating a long stabilization period. The Alyza indicated a considerable bias in accuracy at the 250-mg/L DOM concentration, where the mean value of the 250-mg/L standard is more than double the 10-and 50-mg/L DOM standards. This analyzer does have an additional background correction setting for interferences caused by elevated levels of organic matter, which was active during testing.

Waste Production

Manufacturer-specified total volume and reagent volume is listed in table 7. A wide variety of total and reagent volumes exists among analyzers per sample. The EZ7800 TOPHO uses the most total and reagent volumes, whereas the Alyza uses the least. It is important to note that analyzer design typically determines waste volumes. For example, the EZ7800 TOPHO was designed to target total phosphorus and, as a result, has a large digestion vessel that would require special design considerations, including larger volumes. The Alyza targets orthophosphate and uses microfluidics and separate waste tubing to reduce total waste and reagent volumes.

Analyzer and Discrete Data Quality

Linear regressions between OP discrete samples and each analyzer are shown in figure 6. Strong correlations (coefficient of determination equal to or greater than 0.9) were observed from the Alyza, Phosphax, and Hydrocycle, along with a moderate correlation (coefficient of determination of 0.47) for the EZ7800 TOPHO. Regressions between OP concentrations in discrete samples and the difference (residual) in OP concentrations between the analyzer and discrete samples are shown in figure 7. Weak but statistically significant linear relations (p-values less than 0.001 based on 95-percent confidence interval) exist for the Phosphax and Alyza but not for the EZ7800 TOPHO or Sea-Bird Scientific HydroCycle-PO₄.

All analyzers in the study used some combination of automated cleaning or calibration. Because of the complexity and importance of these processes, discrete samples become paramount to verifying accuracy and proper operation of the analyzer. This can be problematic because most operators send verification samples to laboratories for analysis, and it may take weeks to receive results. Additionally, multiple check samples are usually required to conclude whether or not there is a problematic pattern. Combined, these issues indicate that long periods of poor data quality are likely before a problem is diagnosed.

The phosphorus concentrations measured by the EZ7800 TOPHO had a moderate relation with OP discrete samples (fig. 6), and no bias with discrete OP concentration (fig. 7). This analyzer was designed to target TP with its large 100-μm pore size filter and persulfate digestion. However, the EZ7800 TOPHO had no relation with TP with a coefficient of determination of 0.01. It is unclear why the phosphorus concentration measured by the EZ7800 TOPHO did not indicate any relation with TP although it is likely procedural and related to the sample contamination problem described in the “Laboratory Verification Results” section.

The OP concentrations measured by the Phosphax had a consistent high bias compared to OP concentrations determined in discrete samples and a significant negative linear correlation with increasing concentration (figs. 6 and 7). The Hach Phosphax does not have automated calibrations; it only has cleaning cycles, so it is likely that the initial calibration was slightly off even though it was within the stated analyzer accuracy. However, this revelation led to noticeable biases in the dataset, which demonstrated how critical the initial calibration correction is for this analyzer.

The OP concentrations measured using the Sea-Bird Scientific Hydrocycle-PO4 had a strong regression with OP concentrations determined in discrete samples (fig. 6), and residuals indicate no correlation with concentration (fig. 7). This analyzer does have a slight low bias (fig. 6), and this has been a consistent finding through multiple studies (Johengen and others, 2017; Snazelle, 2018). All discrete sample concentrations with one exception were within the manufacturer’s stated range for the Hydrocycle; however, all concentrations were within the theoretical operating limits (table 1).

The OP concentrations measured by the Alyza have a strong linear regression with orthophosphate concentrations in discrete samples and has a consistent low bias that increases with concentration (figs. 6 and 7). Reasons for the strong regression and low bias are unclear, but it is hypothesized that the problem is in the daily calibration correction procedure, which would be expected to have nullified any bias. This may be related to the laboratory findings indicating the daily calibration correction-induced error in the measurements.

Turbidity Response

A large portion of phosphorus is associated with particulates in the Vermilion River. Discrete sample data indicate that the mean OP concentration is about 63 percent of TP concentration at turbidities less than 100 formazin nephelometric units (FNUs) but only about 17 percent at turbidities greater than 100 FNUs. Additionally, TP concentrations have a moderately strong relation with turbidity (fig. 8). Regression results between analyzers and turbidity did not indicate any relation (coefficient of determinations less than 0.1), and neither did the regressions between the difference in analyzer discrete OP concentrations and turbidity (coefficients of determination less than 0.05). This is an unexpected finding because the EZ7800 TOPHO and the Sea-Bird Scientific HydroCycle-PO₄ had filter pore sizes larger than the conventional 0.45 μm, and therefore, were expected to have OP concentrations greater than those in the discrete samples because they were analyzing more particulate phosphorus.

The EZ7800 TOPHO has a filter with a mean pore size of 100 μm and an added persulfate digestion. Although the EZ7800 TOPHO does not measure a true TP concentration, this analyzer is designed to analyze more particulates and break down phosphorus-containing compounds to make its analysis more comparable to TP. The EZ7800 TOPHO had no relation with any other properties except OP. This includes turbidity, which did have a moderately strong coefficient of determination with TP (fig. 8). Reasons why the EZ7800 TOPHO did not relate with turbidity or TP likely result from an issue with turbidity because it did not meet manufacturer specifications when turbidity was greater than 100 FNUs. For the 26 discrete samples that had turbidity values greater than 100 FNUs, the EZ7800 TOPHO was functional for 16 of them and measured 0 mg/L of phosphorus for 9 out of these 16 samples. Further investigation indicates that TP concentrations for discrete samples have a moderate correlation (coefficient of determination=0.45) with the TP concentrations measured by the EZ7800 TOPHO for turbidity values less than 100 FNUs. These conditions also exist when OP makes up a larger part of TP. This evidence indicates that turbidity or another characteristic associated with higher turbidity levels is interfering with the operation of this analyzer.

The filter for the Sea-Bird Scientific Hydrocycle-PO₄ has a mean pore size of 7.5 μm, which is more than 10 times larger than the 0.45-μm filter used to determine OP. The larger filter size was hypothesized to have caused this analyzer to have a high bias when compared to discrete OP samples because it allowed more particulate-associated phosphorus into its analysis. That relation had been observed previously with Cycle-PO₄, the predecessor to this analyzer (Terrio and others, 2015). This finding was not observed with the Sea-Bird Scientific Hydrocycle-PO₄, and OP data showed a low bias (fig. 6) and no correlation with turbidity.

The Sea-Bird Scientific Hydrocycle-PO₄ has extensive quality control software that allows the user to analyze sample runs. One of those tools has the ability to analyze light absorption during the chemical reaction of each sample. Using this tool, it is hypothesized that the full color development may not always be present during the correct part of its analysis. An example of the problem with an onboard calibration standard check is shown in figure 9. This run was chosen because the analyzer had just been serviced by the manufacturer before the deployment, and a discrete OP sample was collected within the hour of the analyzer sample for comparison. The spike just before minute eight in figure 9 indicates the part of the reaction in question. According to the user’s manual, in this section, the analyzer should be rinsing and preparing for the next sample. However, additional light absorption is present, which may indicate further color development. A second calibration check that does not have this issue was included for comparison (fig. 9). By not allowing full color development, the analyzer will underestimate light absorption and in turn underestimate OP concentration. From direct communications with the manufacturer, it was agreed that the spike is an unknown solution that is outside the optical chamber during the reaction phase, but then was swept through during the final flushing phase resulting in the spike. The manufacturer has not performed an analysis on what that spike may be, but it is accepted as normal behavior of the analyzer at the time of writing. It is unclear if the Hydrocycle would have indicated any correlation with turbidity had this potential issue not existed.

Data Coverage and Maintenance

Data coverage is listed in table 9. Missing data because of laboratory trials was not included in the total possible data points. Differences in the total possible data points are the result of the removed standards testing periods and analyzer sample intervals. It should be noted that the EZ7800 TOPHO and Phosphax were serviced quarterly by Hach personnel onsite. YSI Inc. and Sea-Bird Scientific did not have onsite service available at the time of the study. All analyzers experienced user error and various malfunctions during the study period, which resulted in data loss.

The EZ7800 TOPHO had 76-percent data coverage. Despite quarterly servicing, the tubing was subject to clogging, meaning the daily cleanings were not effective. The operator error was much higher largely because of the amount of reagent consumption. From table 7, the amount of reagent consumption per run is much higher than the other analyzers requiring more frequent replacement. Replacement rates depend on several factors including container size, operator- or manufacturer-supplied reagents, and transportation and site storage safety considerations.

The Phosphax had 95-percent data coverage. Most of the missing data were from a period where internal leaking took place and needed repair. Additionally, the communications hardware malfunctioned, but the analyzer was able to continue sampling during this period, which allowed the data to be recoverable. If the data had not been recoverable, the Phosphax would have had about 86-percent data coverage during the study. Many internal components were replaced quarterly, which likely resulted in the high percentage of data coverage.

The Sea-Bird Scientific Hydrocycle-PO4 had 33-percent data coverage. The Hydrocycle initially operated above the recommended range of 0.3 mg/L of phosphorus, but within the hypothetical manufacturer extended range of 1.2 mg/L of phosphorus (table 1, fig. 5). It became clear that the operator could not adequately clean the analyzer at 2-week intervals, with several trial-and-error periods from March through May 2021 trying to diagnose and fix the problem. During this time, the analyzer produced very little usable data. The analyzer was sent back to the manufacturer for analysis and repair in June 2021. After repair, the analyzer was fully functional and put back into normal operation in July 2021. It is unclear what river matrix characteristics caused the analyzer’s function to degrade because the degradation slowly became apparent during the months of usage. In addition, the analyzer could have benefitted from a more frequent cleaning interval. This is the only analyzer in the study that relied on manual cleaning by the operator rather than an automated cleaning procedure.

The Alyza had 78-percent data coverage. Hardware malfunctions were mainly related to clogged parts and halted further analysis until the issue was rectified. It is possible the daily cleaning should have been run more frequently to help with hardware issues. Annual maintenance, which includes hardware replacement, was not done because the study period was less than a year. Many of the hardware solutions were part of the annual maintenance routine, indicating the Vermilion River may be harsher than the environment intended for the analyzer. The analyzer also had several software malfunctions related to communication errors between different parts of the analyzer. These were often fixed by resetting the analyzer or firmware upgrades.