Performance of aerobic treatment units: monitoring results from the Florida Keys.

Introduction

The Florida Keys are an ecologically sensitive environment and a destination for growing numbers of permanent and transient visitors. To address the increasing negative ecological effects resulting from increased population, statutory onsite wastewater treatment requirements have increased over the last 15 years. Consequently, the use of aerobic treatment units (ATUs) has become widespread. From 1985 through April 2001, approximately 1,200 such systems were installed in the Florida Keys. In Florida, a biennial operating permit and a maintenance agreement between a maintenance entity and the system owner are required for ATUs. The county health department permits and inspects onsite systems, and until 2001 sampled ATUs yearly. The objective of the present study was to assess the performance of installed ATUs and to draw from these experiences to develop programs for monitoring and enforcing the performance of more advanced systems that will be required to address nutrient concerns.

Performance expectations for ATUs are based on the National Sanitation Foundation's (NSF's) Standard 40 for residential wastewater treatment systems. Florida requires testing to this standard for approval of aerobic treatment units (Standards for Onsite Sewage Treatment and Disposal Systems, 2004). Several studies have indicated that effluent from ATUs frequently exceeds these expected limits (Hutzler, Waldorf, & Fancy, 1978; Kellam, Boardman, Hagedorn, & Reneau, 1993: Maxfield, Daniell, Treser, & VanDerslice, 2003; Otis, Boyle, & Sauer, 1974; Sexstone et al., 2000).

In 2000, the second author analyzed the results of 584 grab samples that had been taken in 1999 from 580 ATUs by the Monroe County Health Department (MCHD). These samples were analyzed for five-day carbonaceous biochemical oxygen demand (CBO[D.sub.5]), total suspended solids (TSS), total nitrogen (TN), and total phosphorus (TP). For CBO[D.sub.5], one-sixth of the samples, and for TSS, close to half the samples, gave values that exceeded limits set by Florida secondary treatment standards (20 mg/L each). The effectiveness of ATUs in removing nitrogen was expectedly low, with nitrogen exceeding 10 mg/L in 72 percent of samples. The analysis suggested that the two most important factors influencing effluent sampling results were the method of discharge and the location of the sampling port. The type and manufacturer of ATU did not appear to influence the sampling results.

In the study reported here, the authors review results from the 2000 and 2001 sampling campaigns. Answers to the following questions were sought: How meaningful are grab samples? Are ATUs consistently meeting expectations for NSF-certified, secondary wastewater treatment units? Does it matter where and when a sample is taken? How variable are concentrations with respect to time?

Methods

MCHD designed the study to sample each system older than six months once during each year in accordance with statutory requirements. The systems were distributed between three field offices of the county health department. To maximize the effectiveness of staff time, the department used a convenience sampling scheme, sampling systems in close vicinity to each other on the same day The objective in sampling-point selection was to obtain a sample "as close to the end of the treatment stream as possible" (Florida Department of Health, 2000). For systems discharging into injection wells, a special feature of the Florida Keys, the required treatment includes a sand filter and chlorinator after the ATU. Sampling ports (mostly "Tee" or "Cross" connectors tied into the effluent line) have been required in systems installed after January 1995, but were frequently found dry. Therefore, sampling points included the treatment unit, sampling ports, pump chambers, and injection wells. Sampling ended April 2001, when legislative changes reduced permit fees and thus funding for sampling.

MCHD staff visited each system and took a grab sample from the sampling point. In some cases, access problems to the property and to the unit precluded sampling. Sampling-point location, time, and date were noted in the chain-of-custody information for each sample. The samples were then cooled on ice and sent overnight by express freight to analytical laboratories for analysis within 24 hours. The samples were analyzed for CBO[D.sub.5] (Standard Method [SM] 5210B), TSS (U.S. EPA Method 160.2), TN (U.S. EPA Method 300.0 and U.S. EPA Method 351.2), and TP (U.S. EPA Method 365.2). MCHD staff entered the results reported by laboratories onto the chain-of-custody document or into an Excel spreadsheet. The sampling time was not usually entered into the spreadsheet. In the fall of 2004, Florida Department of Health (DOH) staff completed data entry from paper records and aggregated all data into one spreadsheet.

For the purposes of this analysis the authors assumed that all ATUs are similar to each other without regard to manufacturer. The data analysis consisted of assessments of variability as indicated by relative standard deviations, summary statistics, checks on the normality of the distribution with the Lilliefors correction to the Kolmogorov-Smirnov test, and cross-tabulations. Because of the non-normal distribution of measured parameters, the authors performed non-parametric statistics, Kendall's tau, the Kruskal-Wallis one-way analysis of variance (ANOVA) (National Resource Conservation Service, 2002), and a multiple-comparison procedure of the Kruskal-Wallis test (Cabilio & Masaro, 2001) with SPSS 12 and Excel 2003. The significance level was usually .05, unless stated otherwise in the text below.

Results

Precision of Grab Samples

The coefficient of variation or relative standard deviation (standard deviation/average) indicates variability around a mean. When two samples are involved, the coefficient of variation translates into a factor between the lower and the higher value. In seven instances, duplicate samples had been taken at the same time and location. This duplication allowed an assessment of the precision of analytical results for the conditions in the treatment system at that time. The results in Table 1 indicate high precision (variability of <5 percent) for nitrogen and phosphorus measurements, and higher variability, mostly within a factor of 2, for TSS and CBO[D.sub.5] measurements.

Repeat Sampling of the Same Unit

Because sampling results from two years are included in the data set, data were available from two sampling events (n = 104) and in one case from three sampling events from the same system. This repeat sampling allowed an assessment of the consistency of sampling results from any given system. As indicated in Table 1, the variability was larger than for duplicates of individual samples. CBO[D.sub.5] and TSS values showed the highest variability, typically within a factor of 4; TN varied typically within a factor of 3, and TP within a factor of 2.

Summary Statistics

Figure 1 shows the distribution of analytical results from 901 sampling events on a log-concentration scale. It illustrates that observed concentrations varied over several orders of magnitude. CBO[D.sub.5] and TSS distributions appear to be similar to each other, as do TN and TP distributions. CBO[D.sub.5] concentrations were lower relative to TSS than other studies have found. Summary statistics are given in Table 2. Coefficients of variation had values between 3 and 8 and show that the variability overall is much larger than the variability between repeat samples from the same unit shown in Table 1. This result suggests that differences between units were more important than sampling technique and variations in the performance of one system.

[FIGURE 1 OMITTED]

The means were far larger than the medians for all parameters, indicating distributions skewed strongly to the right. The distribution of analytical results for all four parameters was not normal. Normal distribution for the log(x + 1) distributions for CBO[D.sub.5], TSS, and TP was also rejected. For this reason, the subsequent analysis relied on non-parametric statistics, in particular a one-way ANOVA using the Kruskal-Wallis test, which compares average ranks rather than numerical values.

Comparison with Performance Expectations

Table 3 compares the observed levels of CBO[D.sub.5] and TSS with limits set by several related standards. The standardized test of NSF-40 relies on 24-hour composite samples and assesses the performance of a newly installed unit under regular loading from the sidestream of a wastewater treatment plant. NSF standards require multiple composite samples from one system, while in the study reported here, grab samples were taken from multiple units (NSF International, 2000). The standard has changed over time (Brown & McClelland, 1977). Since the 1996 edition of the standard, there has been no individual sample limit (NSF International, 1996). While the NSF-40 protocol is used in the regulatory approval of ATU designs, its sampling scheme is too costly for use in a yearly inspection program of installed units.

Florida grab sample standards for secondary treatment and advanced secondary treatment assume the same sampling scheme used in this study. A substantial fraction of samples exceeded either standard for CBO[D.sub.5] and TSS.

A 30/30 benchmark, similar to NSF Class I standards, has been used in several previous studies (Kellam et al., 1993; Maxfield et al., 2003; Sexstone et al., 2000). Effluent from ATUs in the Florida Keys appears to have somewhat lower CBO[D.sub.5] and about the same TSS concentrations.

In 2000--not affecting the systems sampled for the study reported here--the Florida Legislature set performance standards as annual averages for new onsite systems in the Florida Keys to 10mg/L CBO[D.sub.5], 10 mg/L TSS, 10 mg/L TN, and 1 mg/L TP.

These comparisons serve as yardsticks of expected performance. While many sampling results from ATUs in the Keys met NSF expectations, a significant fraction exceeded even the more lenient secondary wastewater treatment standards for grab samples.

Sampling Ports

Table 4 gives a summary of results distinguished by discharge method and sample point. To assess how representative a sample from the sampling port is, compared with sampling from the unit, the authors looked at the differences between samples from tee ports and cross ports and samples from the ATU itself. The analysis considered only systems discharging to a drainfield to eliminate the confounding factor of additional treatment steps in systems with an injection well. Samples from tee ports and cross ports had significantly higher CBO[D.sub.5], TSS, and TN (p < .1) concentrations than samples from the treatment unit itself. The low relative standard deviations for sampling-port results suggest that the precision of these samples is fairly high. The high TSS results in sampling-port samples suggest that solids were included in the water sample. This result confirmed the analyses performed by the second author of the earlier data set and is consistent with observations during sampling that found tee ports and cross ports dry or filled with settled solids. As a result of such experiences during the 1999 sampling campaign, a new U-trap type design ("Holmes port") for sample ports was recommended in 2000 to enhance the accuracy of sampling. Future studies should test this sampling-port design.

Additional Treatment After the ATU

A comparison of results of samples taken from treatment units, pump chambers, and injection wells allows an assessment of whether additional treatment is occurring in the tank volume of a pump chamber, or the sand filter and chlorination unit for an injection well (Table 4). For systems discharging to drainfields, levels for all four parameters were significantly lower in pump chamber samples than in treatment unit samples (p < .1 for TSS) or sampling-port samples (p < .15 for tee ports and cross ports). For systems discharging to injection wells, levels for all four parameters were significantly lower in samples from both pump chambers and injection wells than in samples from the treatment unit. Only TSS levels differed significantly between pump chamber and injection well samples. These results indicate that the pump chamber and sand filter perform additional treatment.

Overall, samples showing additional treatment effects more than compensated for the higher concentrations in samples from T- and X-ports in the overall results. For CBO[D.sub.5] and TSS, respectively, more than a quarter and more than a half of samples from the treatment unit itself exceeded the Florida secondary-treatment grab sample standards.

Age of System

While systems from all manufacturers have undergone NSF testing, one could expect that older systems perform worse than newer ones as a result of wear and tear. Information on the age of systems was not directly available. Therefore, the permit number was used as a substitute for relative age--that is, the higher the permit number, the more recent the system. No significant correlation was found between permit number and any parameter in injection well samples. In samples from treatment units, significant positive correlations were found between the permit number and CBO[D.sub.5], TSS, and TN--that is, samples from more recently installed ATUs tended to have higher concentrations. This result suggests that newer systems do not perform better than older ones.

Effects of Time of Sampling

To investigate the effect of sampling time of day on effluent concentrations, the authors considered the relatively few samples overall (n = 113) and the samples from treatment units (n = 29) that had been coded for the time of sampling in two-hour increments (8 a.m.-10 a.m., 10 a.m.-12 noon, 12 noon-2 p.m., etc.). Only four samples had been taken between 8 a.m. and 10 a.m., and all samples from the treatment unit itself had been sampled between 10 a.m. and 2 p.m. Differences in concentrations among the three two-hour intervals between 8 a.m. and 2 p.m. were not significant at the .05 level, although the four samples from the period 8 a.m. to 10 a.m. appeared to have slightly higher concentrations. Therefore, sampling time did not seem to be an important factor in the observed values. This result contrasts with results found by Maxfield and co-authors (2003), who found higher concentrations of TSS in morning samples than afternoon samples.

The effect of sampling day of week (Monday through Thursday) was significant for all parameters. Table 5 summarizes the day-of-week data for all samples. Wednesday concentrations were significantly higher than those of all other days. Monday's CBO[D.sub.5] values were significantly lower than Tuesday's and Thursday's. Treatment unit samples by themselves showed that Wednesday's concentrations of CBO[D.sub.5] and TSS were significantly higher than Tuesday's. For injection well samples, only CBO[D.sub.5] concentrations showed significant daily variations, and Thursday's samples were significantly higher than Tuesday's. A possible cause is changes in influent concentrations over the course of the week--for example because of laundry days on the weekend--and flow through the treatment system. Data from the water supply authority for the Florida Keys for this time period show that overall water supply flows are less than 5 percent higher on weekends than during weekdays. Maxfield and co-authors (2003) found slightly higher concentrations of CBO[D.sub.5], TSS, and fecal coliform in the first half of the week than in the second half of the week (Thursday and Friday). Further studies are necessary to reproduce these results and determine the causes.

Differences between months were assessed for February through November and were significant; no samples were taken in January, and only three samples were taken in December. Water supply data from the water supply authority for the Florida Keys for this time period showed no significant correlation between median concentrations and average monthly water supply flow. Separate consideration of treatment unit samples and injection well samples also showed no consistent monthly concentration patterns. Therefore, further studies should be undertaken to confirm the significance of monthly variations in effluent concentrations.

Conclusions

MCHD performed monitoring sampling at ATUs for CBO[D.sub.5], TSS, TN, and TP during 16 months in 2000 and 2001. Sampling was very precise for nutrients and varied mostly within a factor of 2 for CBO[D.sub.5] and TSS. Repeat samples taken from the same unit in different months varied typically by a factor of 4 for CBO[D.sub.5] and TSS, 3 for TN, and 2 for TP, and indicate variability of the performance of individual systems.

Samples taken from tee ports and cross ports had higher concentrations of CBO[D.sub.5], TSS, and TN than did samples taken from the treatment unit itself. This result suggests that samples taken from such ports are not representative of treatment unit effectiveness and that alternative sampling-port designs should be evaluated. The effect of additional treatment (tank, sand filter, chlorination) between the treatment unit and pump chambers and injection wells was significant for all parameters. Such effects should be considered in defining the performance of systems.

Between 8 a.m. and 2 p.m., time of day had no significant effect on sampling results. Concentrations for all parameters and all samples were higher on Wednesdays. Concentrations appeared also to vary significantly by month. Future studies of individual systems should clarify the relationship between effluent concentrations and loading variations.

Effluent from a substantial fraction of systems exceeded limits set by treatment expectations based on NSF-40, secondary treatment, or new local standards. The assumption that all these systems work well would appear to be an oversimplification. There was no evidence that newer systems perform better than older systems.

Opportunities to improve performance through user education and better maintenance and enforcement should be explored. Monitoring is crucial for determining and enforcing performance in an onsite sewage program. Currently, three elements are missing for an effective monitoring program: funding for collection and evaluation of samples, definition of performance boundary and consistency in sampling (treatment unit, pump chamber or injection well), and definition of an enforceable performance standard, which has been traditionally based on far more frequent sampling than is practicable for OSTDS.

Acknowledgements: The authors thank Gerald Briggs, Sonia Cruz, David Hammonds, Kevin Sherman and an anonymous reviewer for their comments on earlier versions of this paper. Thanks also go to the Florida Keys Aqueduct Authority for providing water supply data.

Corresponding Author: Eberhard Roeder, Professional Engineer III, Florida Department of Health, Bureau of Onsite Sewage Programs, 4052 Bald Cypress Way, Bin# A-08, Tallahassee, FL 32399-1713. E-mail: [e-mail]._

REFERENCES

Brown, R.M., & McClelland, N.I. (1977). NSF: Its role in defining and coordinating objectives. In N.I. McClelland (Ed.), Proceedings of the First National Conference, 1974: Individual Onsite Wastewater Systems (pp. 37-42). Ann Arbor, MI: Ann Arbor Science Publishers.

Cabilio, P., & Masaro, J. (2001). Basic statistical procedures and tables (10th ed.). Retrieved October 4, 2004, from http://ace.acadiau.ca/math/cabilio/StatLabs/BSPT(02).pdf.

Florida Department of Health. (2000, March 20). Sampling of aerobic treatment units (ATU's) (HSEWOS 00-004) [Interoffice Memorandum]. Retrieved June 12, 2006, from http://www.doh.state.fl.us/environment/OSTDS/pdfiles/memos/2000/00-004.pdf.

Hutzler, N.J., Waldorf, L.E., & Fancy, J. (1978). Performance of aerobic treatment units. In Proceedings of the Second National Home Sewage Treatment Symposium (ASAE Publication 5-77) (pp. 149-163). St. Joseph, MI: American Society of Agricultural Engineers.

Kellam, J.L., Boardman, G.D., Hagedorn, C., & Reneau, R.B. (1993). Evaluation of the performance of five aerated package treatment systems (Research Bulletin No. 178). Blacksburg, VA: Virginia Water Resources Research Center, Virginia Polytechnic Institute and State University.

Maxfield, M., Daniell, W.E., Treser, C.D., & VanDerslice, J. (2003) Aerobic residential onsite sewage systems: An evaluation of treat-ed-effluent quality. Journal of Environmental Health, 66(3), 14-19.

NSF International. (1996). Residential wastewater treatment systems (NSF Publication No. ANSI 40-1996). Ann Arbor, MI: Author.

NSF International. (2000). Residential wastewater treatment systems (NSF Publication No. ANSI 40-2000). Ann Arbor, MI: Author.

National Resource Conservation Service. (2002, February). Part 615: Analysis of water quality monitoring data, National water quality handbook (Publication No. 450-VI-NWQH). Washington, DC: U.S. Department of Agriculture.

Otis, R.J., Boyle, W.C., & Sauer, D.K. (1974). The performance of household wastewater treatment units under field conditions. In Proceedings of the National Home Sewage Treatment Symposium (ASAE Proc-175) (pp. 191-201). St. Joseph, MI: American Society of Agricultural Engineers.

Sexstone, A., Aiton, M., Bissonnette, G., Fleming, K., Kinneer, K., Hench, K., Bozicevich, T., Cooley, B., & Winant, E. (2000). A survey of home aerobic treatment systems operating in six West Virginia counties. Small Flows Quarterly, 1(4), 38-46.

Standards for Onsite Sewage Treatment and Disposal Systems, Florida Administrative Code [section] 64E-6 (2004). Retrieved June 12, 2006, from http://www.doh.state.fl.us/environment/ostds/pdfiles/forms/64e6.pdf.

Eberhard Roeder, Ph.D., P.E.

William Brookman, M.P.H.

TABLE 1 Relative Standard Deviation (Coefficient of Variation) as
Indicator of Precision of Sampling and Consistency of Treatment Unit
Performance

Replicate
Sample      Relative Standard   CBO[D.sub.5]  TSS     TN      TP
Population  Deviation           (mg/L)        (mg/L)  (mg/L)  (mg/L)

Duplicate at same sampling event (n = 7)
            Average             0.35          0.39    0.024   0.030
            Median              0.39          0.31    0.022   0.025
            Standard deviation  0.26          0.28    0.016   0.016
Repeat visit at same unit (n = 105)
            Average             0.70          0.79    0.55    0.47
            Median              0.68          0.79    0.52    0.35
            Standard deviation  0.44          0.43    0.41    0.39

TABLE 2 Summary Statistics for ATU-Sampling Results in the Florida Keys

                             CBO[D.sub.5]  TSS      TN      TP
Statistic                    (mg/L)        (mg/L)   (mg/L)  (mg/L)

N                            901             900    900     900
10% percentile (mg/L)          1               2.4    5.0     1.8
Median (mg/L)                  5              32     26       7.8
90% percentile (mg/L)        117           1,087    106      23
Mean (mg/L)                   48             567     69      15
Relative standard deviation    4.1             5.9    7.4     3.4
Skewness                      12              20     28      13

TABLE 3 Frequency and Extent to Which Limits Set by Related Standards
Were Exceeded

Standard                               CBO[D.sub.5] (mg/L)  TSS (mg/L)

NSF-40: ATU Class II 10% exceedance    60 mg/L 17%          100 mg/L 31%
(until 1996 Class I individual-sample
limit)
Florida secondary-treatment standard:  60 mg/L 17%           60 mg/L 38%
grab sample (62-600 and 64E-6 Florida
Administrative Code)
NSF-40: ATU class I 30-day average     25 mg/L 25%           30 mg/L 50%
Florida advanced-secondary-treatment   20 mg/L 29%           20 mg/L 58%
standard: grab sample (64E-6 Florida
Administrative Code)
Florida Keys standard                  10 mg/L 37%           10 mg/L 70%

TABLE 4 Differences in Sampling Results Between Sampling Points

Discharge   Sample             CBO[D.sub.5]  TSS      TN      TP
Method      Point   Statistic  (mg/L)        (mg/L)   (mg/L)  (mg/L)

Drainfield
            Treatment unit (n = 154)
                    Median     14.2             76.5   40.2   11
                    Mean       75.4            868     64.3   21
                    RSD         3.3              2.5    1.1    2.1
            Tee port (n = 42)
                    Median     32.2            141     70.2   13.7
                    Mean       67.1            747     88.5   35.3
                    RSD         1.3              1.9    1.2    3.3
            Cross port (n = 85)
                    Median     23.3            232     51.8    9.9
                    Mean       66.8          1,016    115     19.6
                    RSD         1.6              2.0    1.4    1.8
            Pump chamber (n = 42)
                    Median      4.1             27.0   23.0    7.2
                    Mean       29.7            887    386     28.9
                    RSD         3.3              5.0    6.0    4.6
Injection well
            Treatment unit (n = 35)
                    Median     11.4             54     38.5    8.2
                    Mean       40              336    102     41.1
                    RSD         1.5              1.6    2.2    3.3
            Pump chamber (n = 107)
                    Median      2.0             12.0   18.0    4.7
                    Mean        7.7            163     24.0    7.3
                    RSD         2.5              5.5    1.5    2.0
            Injection well (n = 171)
                    Median      2.0              4.5   13.8    4.6
                    Mean        5.1             17.5   24.4    5.4
                    RSD         2.5              3.8    1.2    0.81

RSD = relative standard deviation.

TABLE 5 Differences Between Concentrations by Day for All Samples

                    CBO[D.sub.5]
Weekday  Statistic  (mg/L)        TSS (mg/L)  TN (mg/L)  TP (mg/L)

Monday (n = 41)
         Median      1.1            9.0        20.0       5.9
         Mean        3.8           26.2        28.0       6.7
         RSD         2.7            1.7         0.95      0.58
Tuesday (n = 81)
         Median      2.9           22.0        23.0       6.1
         Mean       11.1           88.7        32.0       7.5
         RSD         2.7            2.9         0.92      0.79
Wednesday (n = 467)
         Median     12.6           74.0        35.5       9.0
         Mean       75.2          920         101        21.4
         RSD         3.5            4.9         7.0       3.3
Thursday (n = 310)
         Median      3.0           15.0        19.9       6.9
         Mean       23.4          235          34.9       9.7
         RSD         2.9            3.7         1.5       1.8

RSD = relative standard deviation.