v>EPA
United States
Environmental Protection
Agency
Targeted National Sewage Sludge Survey
Statistical Analysis Report
January 2009
-------
U.S. Environmental Protection Agency
Off ice of Water (4301T)
1200 Pennsylvania Avenue, NW
Washington, DC 20460
EPA-822-R-08-018
-------
Acknowledgments and Disclaimer
This report has been reviewed and approved for publication by the Office of Science and
Technology. This report was prepared with the support of Battelle and Computer Sciences
Corporation, under the direction and review of the Office of Science and Technology.
Neither the United States government nor any of its employees, contractors, subcontractors, or
other employees makes any warranty, expressed or implied, or assumes any legal liability or
responsibility for any third party's use of, or the results of such use of, any information,
apparatus, product, or process discussed in this report, or represents that its use by such a third
party would not infringe on privately owned rights.
iii January 2009
-------
TABLE OF CONTENTS
FIGURES v
TABLES v
APPENDICES vi
EXECUTIVE SUMMARY vii
1.0: INTRODUCTION 1
2.0: DATA COLLECTION 2
2.1 Selection of Analytes for In-Depth Statistical Analysis 2
2.2 Target Population and Sample Frame 5
2.3 Selection of Facilities 6
2.3.1 Number of Facilities (Sample Size) 6
2.3.2 Final Sample Size 7
2.3.3 Partial Treatment Facilities 8
2.3.4 Facilities with Wastewater Treatment Ponds 9
2.4 Biosolids Collection 9
2.4.1 Multiple Biosolids Treatment Systems 10
2.4.2 Field Duplicates 10
2.4.3 Aggregating Data Across Multiple Samples 11
3.0: OVERVIEW OF STATISTICAL METHODOLOGY 13
3.1 Survey Weights 13
3.2 Statistical Analysis Approaches 14
3.2.1 Lognormal Approach 14
3.2.2 Nonparametric (Distribution Free) Approach 14
3.3 Quality Assurance 15
4.0: FINDINGS FROM IN-DEPTH STATISTICAL ANALYSES 16
4.1 National Estimates of Detection Percentages 16
4.2 Statistical Graphics: Bar Charts and Box Plots 19
4.2.1 Metals 20
4.2.2 Organics 23
4.2.3 Classicals 25
4.2.4 PBDEs 26
4.2.5 Pharmaceuticals, Steroids, and Hormones 28
4.3 Data Review 36
4.3.1 Statistical Outliers 36
4.3.2 Statistical Tests for Lognormality 37
4.3.3 Findings from the Data Review 39
4.3.3.1 Metals 39
4.3.3.2 Organics 39
4.3.3.3 Classicals 40
4.3.3.4 PBDEs 40
4.3.3.5 Pharmaceuticals, Steroids, and Hormones 40
4.4 National Estimates 41
4.5 Comparison of Metals to Current Standards 44
4.6 Comparison of Metals, Organics, and Classicals to NSSS Results 45
5.0: CONCLUSIONS 47
iv January 2009
-------
6.0: REFERENCES 48
FIGURES
Figure 4-la. Bar Charts for Metals 21
Figure 4-lb. Box Plots for Metals and Classicals 23
Figure 4-2a. Bar Charts for Organics 24
Figure 4-2b. Box Plots for Organics 25
Figure 4-3. Bar Chart for Classicals (Nitrate/Nitrite) 25
Figure 4-4a. Bar Charts for PBDEs 27
Figure 4-4b. Box Plots for PBDEs 28
Figure 4-5a. Bar Charts for Pharmaceuticals 29
Figure 4-5b. Bar Charts for Steroids and Hormones 33
Figure 4-5c. Box Plots for Pharmaceuticals, Steroids, and Hormones 35
TABLES
Table ES-1. Nationally-Representative Estimates for 34 Analytes viii
Table 2-1. Analytes With Reported Data for Biosolids Samples in the TNSSS 2
Table 2-2. The Eight Target Analytes Within the TNSSS 4
Table 2-3. Pharmaceuticals, Steroids, and Hormones Selected for In-Depth Statistical
Evaluation 4
Table 2-4. Original and Final Sample Sizes for the TNSSS 7
Table 2-5. Facilities in the Original Sample that Employed Partial Treatment, and Their
Replacement Facilities 8
Table 2-6. Facilities from the Original Sample that Employed Wastewater Treatment Ponds,
and How EPA Handled These Facilities 9
Table 2-7. Summary of Situations Where Multiple Samples Were Selected at POTWs 10
Table 2-8. Original Set of Eight Facilities Selected for Field Duplicate Sampling 11
Table 2-9. Determining the Classification of Aggregated Measurements as Detected or Not
Detected 11
Table 3-1. Final Set of Survey Weights 14
Table 4-1. 34 Analytes Considered for In-Depth Statistical Analysis 16
Table 4-2. Nationally-Representative Estimates of Detection Percentages in Biosolids for
Analytes Included in the In-Depth Statistical Analysis 17
Table 4-3. Nationally-Representative Estimates of Detection Percentages in Biosolids for
Analytes Not Included in the In-Depth Statistical Analysis 18
Table 4-4. Listing of Detected Measurements Labeled as Statistical Outliers for Analytes
Subject to In-Depth Statistical Analysis 37
Table 4-5. Results of Shapiro-Wilk Tests for Normality of Log-Transformed Biosolids Data
for Analytes Subject to In-Depth Statistical Analysis 38
Table 4-6. Nationally Representative Estimates of the Mean, Standard Deviation, and Selected
Upper Percentiles of the Distribution of Concentrations for 34 Analytes in the
TNSSS 42
Table 4-7. Land Application Ceiling Standards for Nine Metals, and Maximum Concentrations
As Observed in Samples Collected in the TNSSS 44
Table 4-8. Comparison of Distributional Parameter Estimates Between the NSSS and the
TNSSS, Obtained Using Nonparametric (Distribution-Free) Approaches 46
January 2009
-------
APPENDICES
Appendix A: List of Facilities in the TNSSS and Sample Measurements by Analyte
Appendix B: Data Review and Preparation for Statistical Analysis
Appendix C: Statistical Methodology
Appendix D: Estimates of Mean, Standard Deviation, and Selected Percentiles as Calculated
Under Both Statistical Approaches
Appendix E: TNSSS Sample Design Report
Appendix F: Estimates from the 1988 National Sewage Sludge Survey (NSSS)
vi January 2009
-------
EXECUTIVE SUMMARY
This document provides technical background, statistical methods, and resulting estimates of pollutant
concentrations in treated sewage sludge ("biosolids") that represent Publicly Owned Treatment Works
(POTWs) in the contiguous United States with flow rates of at least 1 million gallons per day (MOD).
Estimates were produced using data from a national probability sample of 74 POTWs that statistically
represent 3,337 POTWs that met the study criteria. This sampling effort is known as the Targeted
National Sewage Sludge Survey (TNSSS). Estimates presented in this document are generally
exploratory, and they provide important input to EPA's efforts to evaluate biosolids generated by the
nation's POTWs. The results also may support the development of pollutant limitations, regulatory
impact analysis (RIA), and aggregate risk analysis related to biosolids under Part 503 of 40 CFR.
This report presents the concentrations for 145 analytes, including metals, classicals, organics,
polybrominated diphenyl ethers (PBDEs), pharmaceuticals, steroids, and hormones. For 34 of the
analytes measured in this survey, including eight "target" analytes, this report discusses an in-depth
statistical analysis that yielded nationally-representative estimates. For all other analytes, Appendix B.3
provides preliminary summaries and national estimates derived from the concentration data. Because
EPA has not performed an in-depth statistical analysis on the 111 analytes listed in Appendix B.3, the
reader should exercise caution when interpreting the preliminary summaries. If information becomes
available at some later time that warrants further evaluation of these analytes, or if other analytes become
the basis for any decision-making activities, EPA will perform in-depth statistical analyses and possibly
revise the preliminary results.
For each of the 34 analytes, Table ES-1 presents nationally-representative estimates of the 50th percentile
(i.e., median) of the underlying distribution of measurements across POTWs, as well as the 90th, 95th, 98th,
and 99th percentiles. Table ES-2 provides selected nationally-representative estimates of the mean and
standard deviation, along with the minimum and maximum measurements that were encountered among
the samples collected in this survey. For 33 of the 34 analytes (i.e., all but nitrate/nitrite), EPA's
statistical approach assumed an underlying lognormal distribution for the measurements. Because
lognormality was a poor fit to the observed distribution of nitrate/nitrite data, EPA used a distribution-free
nonparametric approach to generate its estimates.
vii January 2009
-------
Table ES-1. Nationally-Representative Estimates for 34 Analytes ~
Estimates Statistically Adjusted to Represent 3,337 POTWs (>1 MGD)
Analyte
Observed Values
Minimum
Maximum
Estimates
Percentiles
99th
98th
95th
90th
50th
Summary Statistics
Mean
Standard
Deviation
Percent
POTWs
with
Detected
Cone
Metals (mg/kg)
Barium
Beryllium
Manganese
Molybdenum
Silver*
77
0.04
35
2.51
2
2,117
2.34
14,900
86.4
195
2,230
1.81
9,700
68.7
105
1,848
1.45
6,904
55.6
82
1,396
1.04
4,156
40.5
57
1,088
0.77
2,648
30.6
42
452
0.27
540
11.4
13
572
0.38
1,165
15.3
20
443
0.37
2,231
13.8
22
100
98.5
100
100
100
Organics (ug/kg)
4-Chloroaniline
Fluoranthene
Pyrene
51
45
44
5,900
12,000
14,000
12,013
13,173
15,918
8,288
9,112
10,894
4,762
5,256
6,184
2,912
3,226
3,742
513
575
634
1,284
1,421
1,654
2,946
3,211
3,981
74.4
89.5
84.9
Classicals (mg/kg)
Nitrate/Nitrite
2
6,120
6,120
2,750
960
463
14
219
828
100
PBDEs (ng/kg)
BDE-47 (2,2',4,4'-
tetrabromodiphenyl)
BDE-99 (2,2',4,4',5-
pentabromodiphenyl)
BDE-153(2,2',4,4',5,5'-
hexabromodiphenyl)
BDE-209
(decabromodiphenyl)
73,000
64,000
9,100
150,000
5,000,000
4,000,000
410,000
17,000,000
2,650,430
2,696,928
265,395
15,836,435
2,212,077
2,248,181
220,098
11,645,502
1,688,881
1,713,370
166,454
7,360,103
1,329,167
1,346,295
129,902
4,898,034
570,448
574,559
54,117
1,162,523
709,174
716,362
68,334
2,181,237
523,791
533,447
52,685
3,462,942
100
100
100
98.5
Pharmaceuticals (ug/kg)
4-Epitetracycline (ETC)
Azithromycin
Carbamazepine
Cimetidine*
41
8
9
4
4,380
5,205
6,030
8,330
8,026
8,717
1,234
19,128
5,937
5,811
856
10,975
3,787
3,172
497
4,789
2,540
1,853
306
2,294
620
278
55
171
1,135
831
135
1,332
1,741
2,342
298
10,314
96.0
96.0
96.0
89.9
Vlll
January 2009
-------
Table ES-1 (cont.)
Analyte
Observed Values
Minimum
Maximum
Estimates
Percentiles
99th
98th
95th
90th
50th
Summary Statistics
Mean
Standard
Deviation
Percent
POTWs
with
Detected
Cone
Pharmaceuticals (ug/kg) (cont.)
Ciprofloxacin
Diphenhydramine
Doxycycline
Erythromycin-Total
Fluoxetine
Miconazole
Ofloxacin
Tetracycline (TC)
Triclocarban
Triclosan
75
37
34
2
10
7
25
38
187
334
40,800
5,730
5,090
180
3,130
9,210
58,100
5,270
441,000
133,000
79,636
5,255
7,021
264
1,555
16,931
85,562
10,042
276,708
197,288
57,975
4,021
5,046
194
1,178
10,083
57,929
7,250
205,043
124,176
36,095
2,696
3,082
123
778
4,652
32,363
4,458
131,079
62,217
23,703
1,891
1,989
82
539
2,341
19,304
2,895
88,120
33,693
5,367
541
424
19
147
207
3,113
630
21,677
3,862
10,501
871
877
36
245
1,239
8,573
1,278
39,433
16,097
17,658
1,101
1,588
58
329
7,311
21,998
2,255
59,924
65,135
100
100
92.8
92.9
96.1
95.8
98.5
97.5
100
92.4
Steroids and Hormones (ug/kg)
Beta Stigmastanol
Campesterol
Cholestanol
Cholesterol
Coprostanol
Epicoprostanol
Stigmasterol
3,440
2,840
3,860
2,340
7,720
868
455
1,330,000
524,000
4,590,000
5,390,000
43,700,000
6,030,000
568,500
1,651,188
842,112
7,874,368
13,376,891
57,794,254
25,579,800
4,606,900
1,123,256
598,919
5,071,045
8,538,884
35,060,035
13,441,281
2,646,615
632,009
360,119
2,629,149
4,369,111
16,626,022
5,143,938
1,157,099
379,365
229,283
1,467,636
2,410,541
8,574,467
2,193,143
555,217
62,547
46,547
187,244
295,092
827,108
108,028
41,513
168,079
100,879
680,046
1,129,268
4,366,714
1,702,708
321,199
419,232
193,964
2,374,369
4,171,366
22,636,715
26,783,520
2,464,383
98.5
100
100
96.9
100
98.5
90.1
* The calculation of these estimates excludes one sample whose concentration was considered a statistical outlier (silver) or whose concentration was missing
(cimetidine).
IX
January 2009
-------
-------
1.0: INTRODUCTION
Biosolids are the nutrient-rich solid, semisolid, or liquid organic materials that result from the treatment
of domestic wastewater by municipal wastewater treatment plants, also known as Publicly Owned
Treatment Works (POTWs). Local municipalities typically decide how best to manage the treated
sewage sludge ("biosolids") that their POTWs generate, such as to recycle them as a fertilizer, incinerate
them, or bury them in a landfill. The U.S. Environmental Protection Agency (EPA) is responsible for
providing the public with educational information, based on the best science, on the safe recycling and
disposal of these biosolids. Furthermore, Section 405(d) of the Clean Water Act requires EPA to identify
and regulate toxic pollutants that may be present in biosolids at levels that may negatively impact public
health and the environment.
In 1988, EPA conducted the National Sewage Sludge Survey (NSSS) to obtain information on pollutant
levels in treated biosolids (USEPA, 1992). EPA used information collected in this survey when
promulgating Round 1 of regulations in 1993, which established standards for the final use and disposal
of biosolids. EPA completed Round 2 of regulations, which focused on land-applied biosolids containing
dioxin and dioxin-like compounds, in October 2003.
Following these first two rounds of regulations, EPA performed a screening assessment of chemical
pollutants in biosolids. From this assessment, EPA identified a subset of pollutants for possible
regulation. However, additional data were needed for these pollutants. In addition, EPA and other
organizations, such as the National Research Council (NRC, 2002), recognized the need for collecting
data on other non-regulated analytes that had not been previously assessed. Examples of such analytes
included polybrominated diphenyl ethers (PBDEs), pharmaceuticals, steroids, and hormones. To obtain
these data, EPA initiated a new survey called the Targeted National Sewage Sludge Survey (TNSSS). In
this survey, EPA collected physical samples of biosolids from a statistically representative subset of the
nation's POTWs and analyzed these samples for a series of environmental pollutants and contaminants.
This report presents statistical methodology and evaluations related to the data collected in the TNSSS.
For selected analytes, it provides estimates of concentrations in biosolids that are representative of the
nation's largest 3,337 POTWs. A companion report describes the sampling and chemical analyses
(USEPA, 2008: Sampling and Analysis Report for the Targeted National Sewage Sludge Survey, EPA-
822-R-08-016).
This report about the statistical methodology has six chapters. This first chapter provides background and
organization of the report. Chapter 2 provides a summary of the selected analytes for the in-depth
statistical evaluation, the target population, and selection of facilities for the survey. Chapter 3 provides
an overview of the statistical methodology used to derive the survey weights and model the concentration
data. (Appendix C provides the statistical equations and derivations.) Chapter 4 presents the results of the
statistical analyses. Finally, Chapter 5 presents a summary of the results and conclusions with Chapter 6
providing references.
January 2009
-------
2.0: DATA COLLECTION
This chapter lists the set of analytes (i.e., pollutants and contaminants) for which EPA collected
concentration measurements within the TNSSS. Then, it provides an overview of the survey design.
Section 2.2 defines the target population of POTWs; and Section 2.3 describes the plan for selecting a
statistically representative sample of facilities from this target population and deviations during the study.
Finally, Section 2.4 describes the biosolids collection and the numbers of biosolids samples collected at
each sampled facility.
2.1
Selection of Analytes for In-Depth Statistical Analysis
This report evaluates the concentrations for 145 analytes, including metals, classicals, organics,
polybrominated diphenyl ethers (PBDEs), pharmaceuticals, steroids, and hormones. Table 2-1 identifies
the 145 analytes measured in TNSSS, and asterisks denote the 34 analytes with in-depth statistical
evaluation presented in Chapter 4. This section describes EPA criteria for selecting specific analytes for
the in-depth evaluation of the statistical results. As a result of performing the in-depth evaluations, EPA
verified or modified distributional assumptions and data selections as described in this document.
Although EPA presents preliminary summaries of the remaining 111 analytes in Appendix B.3, it has not
thoroughly reviewed these summaries to determine if distributional assumptions are appropriate or
statistical outliers are present. As a consequence, the reader should exercise caution when interpreting
these preliminary summaries. If information becomes available at some later time that warrants further
evaluation of certain analytes, or if other analytes become the basis for any decision-making activities,
EPA will perform in-depth statistical analyses and possibly revise the preliminary results at that time.
Table 2-1. Analytes With Reported Data for Biosolids Samples in the TNSSS
Metals
Organics
Classicals
(inorganic ions)
PBDEs
Steroids and
Horniones
Aluminum
Antimony
Arsenic
Barium*
Beryllium*
Boron
Cadmium
Calcium
Chromium
Cobalt
2-Methylnaphthalene
4-Chloroaniline*
Fluoride
Nitrate/Nitrite*
BDE-28
BDE-47*
BDE-66
BDE-85
17 Alpha-Dihydroequilin
17 Alpha-Estradiol
17 Alpha-Ethinyl-Estradiol
17 Beta-Estradiol
Androstenedione
Androsterone
Beta Stigmastanol*
Copper
Iron
Lead
Magnesium
Manganese*
Mercury
Molybdenum*
Nickel
Phosphorus
Benzo(a)pyrene
Bis(2-ethylhexyl) phthalate
Water-Extractable
Phosphorus
BDE-99*
BDE-100
BDE-138
BDE-153*
Campesterol*
Cholestanol*
Cholesterol*
Coprostanol*
Desmosterol
Epicoprostanol*
Equilenin
Selenium
Silver*
Sodium
Thallium
Tin
Titanium
Vanadium
Yttrium
Zinc
Fluoranthene*
Pyrene*
BDE-154
BDE-183
BDE-209*
Ergosterol
Estriol
Estrone
Norethindrone
Norgestrel
Progesterone
Stigmasterol*
January 2009
-------
Table 2-1. (continued)
Beta-Estradiol 3-Benzoate
Beta-Sitosterol
Equilin
Testosterone
Pharmaceuticals
1,7-Dimethylxanthine
4-Epianhydrochlortetracycline
(EACTC)
4-Epianhydrotetracycline (EATC)
4-Epichlortetracycline (ECTC)
4-Epioxytetracycline (EOTC)
4-Epitetracycline (ETC)*
Acetaminophen
Albuterol
Anhydrochlortetracycline (ACTC)
Anhydrotetracycline (ATC)
Azithromycin*
Caffeine
Carbadox
Carbamazepine*
Cefotaxime
Chlortetracycline (CTC)
Cimetidine*
Ciprofloxacin*
Clarithromycin
Clinafloxacin
Cloxacillin
Codeine
Cotinine
Dehydronifedipine
Demeclocycline
Digoxigenin
Digoxin
Diltiazem
Diphenhydramine*
Doxycycline*
Enrofloxacin
Erythromycin-Total*
Flumequine
Fluoxetine*
Gemfibrozil
Ibuprofen
Isochlortetracycline (ICTC)
Lincomycin
Lomefloxacin
Metformin
Miconazole*
Minocycline
Naproxen
Norfloxacin
Norgestimate
Ofloxacin*
Ormetoprim
Oxacillin
Oxolinic Acid
Oxytetracycline
(OTC)
Penicillin G
Penicillin V
Ranitidine
Roxithromycin
Sarafloxacin
Sulfachloropyridazine
Sulfadiazine
Sulfadimethoxine
Sulfamerazine
Sulfamethazine
Sulfamethizole
Sulfamethoxazole
Sulfanilamide
Sulfathiazole
Tetracycline (TC)*
Thiabendazole
Triclocarban*
Triclosan*
Trimethoprim
Tylosin
Virginiamycin
Warfarin
Analytes for which EPA performed in-depth statistical analyses of the survey data.
Eight "target" analytes were identified as an outgrowth of the December 2003 review of biosolids
regulations (68 FR 75531) where EPA identified 15 toxic pollutants as warranting additional evaluation
of potential risks using more up-to-date sludge concentration and occurrence data. For these pollutants,
EPA conducted an exposure and hazard assessment using available sewage sludge data (USEPA, 2004).
EPA concluded that a new survey, the TNSSS, would be needed to collect more data for eight of the
analytes. Table 2-2 identifies the eight target analytes: four metals, three organics, and one classical.1'2
Because of the importance of these analytes to the study, EPA determined that in-depth statistical
analyses were appropriate.
1 In this document, nitrate/nitrite is counted as one analyte because the chemical analysis generated one value for the combined analytes. In other
documents, EPA often refers to them as two separate analytes. Nitrate and nitrite can be analyzed separately in wastewater, but only on a very
short holding time (24-48 hours). This was essentially impossible for the survey without raising the shipping and analytical costs dramatically.
The two species can undergo transformations back and forth in environmental samples, with nitrate reduced to nitrite under certain conditions,
and nitrite oxidized to nitrate under others. It is difficult to look for the two separately in sludge since the process of leaching the sludge with
water to make measurements is likely to lead to some conversion of nitrite to nitrate.
2 EPA has used the term "classicals" to refer to nitrate/nitrite, fluoride, and water-extractable phosphorous. In other documentation for this study,
EPA has referred to these analytes as "inorganic ions."
January 2009
-------
Table 2-2. The Eight Target Analytes Within the TNSSS
Metals
Organics
Classicals
Barium
Beryllium
4-Chloroaniline
Fluoranthene
Manganese
Silver
Pyrene
Nitrate/Nitrite
EPA selected an additional metal, molybdenum, for the in-depth statistical evaluation. EPA is currently
re-evaluating this metal using updated information to determine the need for a revised numeric standard
in land applied biosolids.
Four PBDEs were identified for in-depth statistical evaluation because they are most prevalent in various
environmental media and acceptable human health benchmarks exist that may be useful for any future
risk assessment purposes. The four PBDEs are BDE-47 (2,2',4,4'- tetrabromodiphenyl), BDE-99
2,2',4,4',5- pentabromodiphenyl), BDE-153 (2,2',4,4',5,5'-hexabromodiphenyl), and BDE-209
(decabromodophenyl).
The pharmaceuticals, steroids, and hormones (including some that are naturally occurring) were measured
using new chemical analytical methods that were recently developed to monitor POTWs. The data from
this method should be considered to be tentative results, pending further study of the chemical analytical
method. Consequently, the statistical analyses of these analytes presented in this report should be
considered to be exploratory in nature. For the group of 97 pharmaceuticals, steroids, and hormones, EPA
used the survey data to estimate the percentage of POTWs, nationally, with detectable levels of the
analytes.3 EPA then conducted an in-depth statistical review of the analytes estimated to be detected at 90
percent4 or more of the POTWs in the target population.
Table 2-3. Pharmaceuticals, Steroids, and Hormones Selected for In-Depth Statistical Evaluation
Analyte
Azithromycin
Beta Stigmastanol
Campesterol
Carbamazepine
Cholestanol
Cholesterol
Cimetidine
Ciprofloxacin
Coprostanol
Diphenhydramine
Doxycycline
Percent of POTWs
Nationally Estimated
to have Detected
Concentrations
96.0
98.5
100
96.0
100
96.9
89.9
100
100
100
92.8
3 The detection limit is generally considered to be the smallest quantity of the analyte that can be reliably measured with that particular method.
Thus, detection is related to the sensitivity of the chemical analytical method, rather than a determination of the presence or absence of a
particular analyte.
4 When rounded to 90 percent.
January 2009
-------
Table 2-3. (continued)
Analyte
4-Epitetracycline (ETC)
Epicoprostanol
Erythromycin-Total
Fluoxetine
Miconazole
Ofloxacin
Stigmasterol
Tetracycline (TC)
Triclocarban
Triclosan
Percent of POTWs
Nationally Estimated
to have Detected
Concentrations
96.0
98.5
92.9
96.1
95.8
98.5
90.1
97.5
100
92.4
2.2
Target Population and Sample Frame
The target population for the statistical analysis of biosolids data from the TNSSS consisted of all
POTWs that:
• were in full operation in 2002 and/or 2004,
• had flow rates greater than 1 million gallons per day (MOD),
• employed a minimum of secondary treatment5, and
• were located in the contiguous United States.
The target population excluded privately-owned, non-publicly owned, and Tribal facilities. The sample
design in Appendix E describes EPA's rationale for focusing the survey on POTWs that met these
criteria. For example, EPA excluded POTWs with less than 1 MGD because they collectively contribute
only about six percent of the total flow among all POTWs in the nation, suggesting that their potential
impact to the environment is minor.
A principal task in the development of a sample survey design is establishing a sample frame that
identifies the entities within the target population. EPA's sample frame consisted of 3,337 facilities
which it identified from one of two sources: the 2004 Clean Watersheds Needs Survey (CWNS)6 and the
2002 version of the Permits Compliance System (PCS).7 Within this sample frame, EPA uniquely
identified all members of the target population, and each member had a known chance of being included
in the sample (Kish, 1965). EPA then used statistical survey sampling techniques to select a sample of
POTWs from the sample frame that would be representative of the entire target population. By applying
At a POTW, all wastewater first must go through the primary treatment process, which involves screening and settling out large
particles. The wastewater then moves on to the secondary treatment process, during which organic matter is removed by
allowing bacteria to break down the pollutants.
6 CWNS is a joint EPA-State survey that collected information on water quality programs and projects that may be eligible for
funding under the Clean Water State Revolving Fund (CWSRF).
PCS is EPA's computerized information management system that tracks permit issuance, permit limits, monitoring data, and
other data pertaining to facilities regulated under the National Pollutant Discharge Elimination System (NPDES).
January 2009
-------
appropriate statistical techniques, EPA was able to generate statistical estimates from data collected from
this sample which could be extrapolated to cover the entire target population.
In the original sample design for the TNSSS (see Appendix E), EPA considered a target population that
differed slightly from the final definition. EPA had originally excluded 46 facilities from the target
population that utilized either partial treatment (17) or wastewater treatment ponds (29) as the final stage
of treatment. This was done because such facilities would have either not produced final biosolids, or the
biosolids would have been too difficult to sample. Because EPA expected that the sample frame had
some imperfections, as most do, EPA incorporated a slight upward adjustment to the sample size (of
approximately four percent) to account for the possibility that some facilities in the sample would actually
be ineligible for the target population. In fact, EPA encountered more ineligible facilities in its sample
than it had initially anticipated. Thirteen facilities (16 percent) were initially ineligible because they
utilized either partial treatment (Section 2.3.3) or wastewater treatment ponds (Section 2.3.4). For
reasons discussed in Sections 2.3.3 and 2.3.4, EPA later redefined the target population to include
facilities that utilized partial treatment or ponds as the final stage of treatment.
2.3 Selection of Facilities
From the sample frame, EPA used statistical sampling techniques to select 80 facilities from which to
collect biosolids samples within the TNSSS. To ensure that the sampled facilities covered the entire
range of flow rates, the sampling design divided the sample frame into three flow groups (or strata):
• Facilities exceeding 100 MOD (>100 MOD);
• Facilities exceeding 10 MOD but no higher than 100 MOD (10 to 100 MOD);
• Facilities exceeding 1 MOD but no higher than 10 MOD (1 to 10 MOD).
Most POTWs are located in the eastern part of the country. To ensure that the sample contained POTWs
from all parts of the nation, EPA selected the sample according to the following two-step process:
1. The facilities were sorted within each stratum by EPA Region (e.g., Region 1, Region 2, etc.),
then by state name within each Region.
2. A systematic sample of facilities was selected within each stratum. If TV denotes the size of
the stratum and n denotes the stratum's target sample size, systematic sampling involves
dividing the stratum into n equal-sized subgroups, generating a random number k between 1
and N/n, and selecting the kth facility within each of the n subgroups
The following sections describe the original sample size selected for the study, the actual number of
facilities selected, and deviations from the original target population definition.
2.3.1 Number of Facilities (Sample Size). To determine an appropriate number of facilities to
sample, EPA employed a sample design that was based upon the binomial distribution. The binomial
distribution applies to situations in which only two outcomes are possible (e.g., yes or no), and it is of
interest to estimate the percentage of the target population achieving the outcome of interest. In
determining a sample size, EPA assumed that the true value of this percentage was 50 percent (e.g.,
pyrene was detected in the biosolids samples at 50 percent of the facilities). This assumption yields the
maximum sample size necessary to achieve the following two precision criteria:
• Overall Criteria: If the true value of the percentage is 50 percent, then a 90% confidence
interval on the percentage is no more than +/- 10% (i.e., the estimated value will be within the
January 2009
-------
range of 40% to 60%). In other words, the sample size must ensure that the unknown
percentage for the target population is estimated to within 20% of its true value with 90%
confidence.
• Within Stratum Criteria: If the true value of the percentage was 50 percent, then a 90%
confidence interval on the percentage is no more than +/- 30% (i.e., the estimated value will
be within the range of 20% to 80%). In other words, the sample size must ensure that an
unknown stratum-specific percentage is estimated to within 60% of its true value with 90%
confidence. (EPA recognizes that this level of precision is not sufficient to produce stratum-
level estimates, but it helps ensure that certain facilities within each stratum are represented
within the sample.)
To achieve both precision criteria, EPA determined that it needed to sample a minimum of 74 facilities.
EPA increased this by four percent (to 80 facilities total) in anticipation of possible ineligible facilities
within the sample. (The size of this upward adjustment was determined by the number of ineligible
facilities encountered in the NSSS sample.) Table 2-4 summarizes the original and final sample sizes by
strata. Appendix A. 1 lists the plant IDs assigned to the selected facilities, the strata in which they
belonged, and the geographic region in which they were located.
As is relatively common in sampling, EPA's sample contained some facilities that did not fall within the
survey's initial definition of the target population. As a result, EPA replaced some facilities with others;
these situations and the replacement facilities are noted within Table 2-4 and Appendix A.I. As first
noted in Section 2.2, selected facilities that were outside of the target population were one of two types:
facilities that conducted only partial treatment, and facilities that utilized wastewater treatment ponds as
final treatment. The following subsections describe how EPA handled these two types of "ineligible"
facilities within its sample.
Table 2-4. Original and Final Sample Sizes for the TNSSS
Stratum
>100 MOD
10 to 100 MOD
1 to 10 MOD
TOTAL
Stratum
Size
51
543
2,743
3,337
Original
Sample
Size
8
12
60
80
No. of
Ineligibles
3
0
8
11
No. of
Replace-
ments
3
0
2
5
Final
Sample
Size
8
12
54
74
% Change
from the
Original
Sample Size
0%
0%
-10.0%
-7.5%
2.3.2 Final Sample Size. Field sampling involved visiting each of the selected POTWs and
collecting a single sample of treated biosolids. Separate documents exist on the procedures that were
used in the TNSSS to contact the selected POTWs, to arrange for a field visit, and to collect biosolids
samples from these POTWs during the field visit. Although EPA had adjusted the sample size upwards
for ineligible facilities selected from the sample frame (e.g., partial treatment, ponds as final treatment),
early contacts with the facilities indicated that the 4% adjustment was an underestimate. To maintain the
target sample size of 74, EPA made the following changes to the facility selection criteria:
• Because the objective was to obtain pollutant concentrations in final biosolids, EPA reevaluated
its decision to consider partial treatment as ineligible. Attempts were made to "follow the
January 2009
-------
sludge," or to replace the facility with the facility receiving the sludge for final treatment as
described in Section 2.3.3.
• Because the sample design incorporated locality (through the sorting of the sample frame), EPA
reevaluated its decision to eliminate a location because the selected facility used a wastewater
treatment pond or lagoon. If another facility within the same system and/or locality produced
biosolids on a regular basis, EPA collected biosolids samples at that location as described in
Section 2.3.4.
2.3.3 Partial Treatment Facilities. Within the original sample of 80 facilities, five utilized partial
treatment. For four of these five facilities, EPA collected biosolids samples from "replacement" facilities.
This section describes EPA decisions about replacements for these facilities.
Upon encountering the first facility found to use partial treatment (ID 84), EPA replaced it with another
facility in the same municipality (ID 53). Compared to the original facility, the replacement facility
shared the same geographic location and management, had similar hydraulic capacity, and was
approximately the same size (in terms of flow). Although EPA did not select this replacement using a
probabilistic sampling approach, it considers the replacement reasonable because of the similarities
between the two facilities.
As it encountered a greater number of partial treatment facilities in its original sample, EPA re-evaluated
its earlier decision to exclude such facilities from its target population. As a result of biosolids
regulations and other factors, it is possible that partial treatment is more common than it was during the
1988 NSSS. Because the objective was to characterize final biosolids, EPA was concerned that it might
be excluding a growing treatment practice, and thus, determined that it would be appropriate to "follow"
the partially treated biosolids to the facility that applied full treatment, and then collect a biosolids sample
from that facility. In reaching this conclusion, EPA also considered whether the co-mingling of wastes
from other facilities would provide misleading results. However, co-mingling is part of the treatment
process for the partially treated biosolids. Thus, because the study objective was to measure the pollutant
concentrations present in the final treated biosolids at each facility, EPA concluded that concentrations at
the facility applying full treatment would appropriately represent treated biosolids for each partial
treatment facility. Therefore, EPA replaced three facilities employing partial treatment (IDs 81, 83, and
99) with the facilities that provided final treatment of the biosolids.
EPA did not replace one facility that performed partial treatment (ID 77). This was due to the inability to
schedule sample collection at a replacement facility without incurring additional study costs.
Table 2-5 summarizes the partial treatment facilities, the replacements, and the selection criteria.
Table 2-5. Facilities in the Original Sample that Employed Partial Treatment,
and Their Replacement Facilities
Original Facility
ID
84
81
83
99
77
Stratum
>100 MOD
1 to 10 MOD
>100 MOD
>100 MOD
1 to 10 MOD
Replacement Facility
ID
53
31
73
61
Stratum
>100 MOD
1 to 10 MOD
10 to 100 MOD
MGD>100
~
Type of Replacement Facility
Similar facility within the same system
Facility that performed final treatment of the biosolids
Facility that performed final treatment of the biosolids
Not replaced
January 2009
-------
2.3.4 Facilities with Wastewater Treatment Ponds. Eight of the original sample of 80 facilities
utilized wastewater treatment ponds (or lagoons) in its treatment process. The bottom layer of a pond
includes deposits of biosolids and supports anaerobic organisms. Facilities remove solids from their
ponds only when they consider the solids to be fully treated. Because the removal process is extensive,
facilities tend to perform it infrequently. For example, ID 5 removes its treated biosolids once every five
years.
EPA and the wastewater treatment industry have long recognized ponds as an effective method for
treating biosolids. However, in the design stage of this survey, EPA felt that coordinating sample
collection with the facility's scheduled removal of biosolids from the ponds would be too difficult. For
this reason, EPA excluded facilities utilizing ponds from the target population. However, EPA re-
evaluated this decision after encountering a greater number of facilities with ponds than expected in the
sample. Because ponds provide an effective final treatment of biosolids, EPA decided that it should
attempt to collect samples of biosolids from these facilities whenever possible.
As noted in Table 2-6, EPA collected biosolids from two facilities that utilized ponds (IDs 3 and 5). This
was possible because EPA was able to schedule physical sampling activities at these facilities as they
were recovering the treated biosolids from the ponds. EPA replaced one facility (ID 82) with its "sister"
facility (ID 21) in the same system. Compared to the original facility, the replacement facility shared the
same geographic location and management, and it was of the same approximate size (in terms of flow).
Although EPA did not select this replacement using a probabilistic sampling approach, it considers the
replacement to be reasonable because of the similarities between the two facilities. The remaining five
facilities with ponds were neither sampled nor replaced, and generally were dropped early in the study
before EPA had reevaluated the eligibility requirements. It also was not feasible to return to the general
area to sample a replacement. To reduce sampling costs, the contractor had grouped its site visits by
region. Consequently, when EPA reevaluated the eligibility requirements, it would have increased the
study costs substantially to incorporate several new sampling trips to the affected regions.
Table 2-6. Facilities from the Original Sample that Employed Wastewater Treatment Ponds,
and How EPA Handled These Facilities
ID
3
5
75
76
78
79
80
82
Stratum
1 to 10 MOD
10 to 100 MOD
1 to 10 MOD
1 to 10 MOD
Final Outcome
Biosolids sampled as planned
Excluded from the study and not replaced
Replaced by ID 21, its sister facility
2.4
Biosolids Collection
EPA collected grab samples of biosolids from the 74 facilities between August 2006 and March 2007.
Because EPA collected samples during one day in a relatively short period of time, the concentration data
associated with these samples allow EPA to evaluate levels in biosolids at a fixed point in time, rather
than to examine trends overtime.
January 2009
-------
EPA collected a single biosolids grab sample from most facilities. However, at ten facilities, EPA
collected two grab samples. This was done either when the facility had more than one treatment system
(implying different types of biosolids generated), or for quality assurance purposes. The following two
sections describe each situation in more detail; and the third section describes how the multiple
measurements were used in the statistical analyses. Table 2-7 summarizes all situations where two
samples were collected at a given POTW.
Table 2-7. Summary of Situations Where Multiple Samples Were Selected at POTWs
Stratum
MOD > 100
10 < MOD < 100
1 < MOD < 10
TOTALS
Final
Sample
Size
8
12
54
74
# POTWs
Having Field
Duplicates
Sampled
o
1
(ID 49)
5
(ID 2, 11, 19*,
28, and 32)
6
# POTWs Having
Solid and Liquid
Products
Sampled from
Different
Treatment
Systems
1
(ID 53)
0
1
(ID 74)
2
# POTW Having
Different
Locations
Sampled from
Different
Treatment
Systems
1
(ID 18)
0
1
(ID 48)
2
Total No.
POTWs
With
Multiple
Samples
Collected
2
1
7
10
*Data for the duplicate sample for ID 19 were excluded from statistical analyses for classicals, anions, and metals (see Section
4.3.3).
2.4.1 Multiple Biosolids Treatment Systems. Most sampled facilities had only one treatment
system that produced biosolids. However, four facilities in the sample utilized two treatment systems.
These facilities are represented in columns 4 and 5 of Table 2-7. Because analyte concentrations have the
potential to differ among biosolids generated by different systems, EPA obtained a grab sample from both
systems within each of these four facilities.
Two facilities (IDs 53, 74) produced biosolids in both liquid and solid forms, both of which were
sampled. The other two facilities produced biosolids in solid form from both of their systems.
2.4.2 Field Duplicates. Within its sample, EPA randomly selected eight facilities (10 percent) for
the collection of duplicate grab samples. Field duplicates allow EPA to assess sampling procedures as
part of its field quality assurance evaluations. EPA does not consider any decisions about field duplicates
to affect the conclusions from the study.
While EPA had planned to sample a field duplicate from eight facilities, field personnel were successful
in collecting field duplicate samples from six facilities (as noted in the third column of Table 2-7). Table
2-8 lists each of the eight facilities and when field duplicate samples were successfully obtained at each.
EPA had excluded one of the eight facilities (ID 75) from the study because it utilized a pond (Table 2-6).
At another facility (ID 18), EPA collected samples of two types of biosolids (Section 2.4.1) rather than a
field duplicate sample, without identifying another facility from which to sample a field duplicate. While
the number of facilities with field duplicate samples collected was less than planned, EPA determined that
this number was sufficient to meet its quality assurance objectives.
10
January 2009
-------
Table 2-8. Original Set of Eight Facilities Selected for Field Duplicate Sampling
ID
2
11
18
19
28
32
46
75
Stratum
1 to 10 MOD
1 to 10 MOD
>100 MOD
1 to 10 MOD
10 to 100 MOD
1 to 10 MOD
Final Outcome of Field Duplicate Sampling
Duplicate was collected as planned.
Samples were collected from each of two treatment systems at the
facility, rather than a field duplicate sample.
Duplicate was collected as planned.
Duplicate was collected at another facility instead (ID 49).
Excluded from the study and not replaced (Table 2-6).
EPA often grouped facilities in nearby locations into a single sampling trip for convenience. On one trip,
the field team discovered that the first facility visited (ID 48) produced two types of biosolids. As a
result, they collected two samples from this facility, one of each biosolids type. To allow for the
additional sampling, the team used equipment that had been designated for collecting a field duplicate
sample at another facility (ID 46) to be visited later in the trip. Because the field team did not expect to
receive the replacement equipment until after visiting ID 46, EPA collected the field duplicate sample
from another facility that the team visited after visiting ID 46. This facility (ID 49) was in the same flow
group and geographic area as the originally selected facility.
2.4.3 Aggregating Data Across Multiple Samples. When a facility had two biosolids samples
collected, either for quality control purposes or because the facility generated two types of biosolids
products. EPA investigated whether the two data values for a given analyte could be aggregated into a
single value prior to performing the data review and analysis. This was done to achieve the objective of
characterizing a facility's average analyte concentration within its final treated biosolids at any single
point in time.
Aggregation of field duplicate measurements within a facility: For each analyte in each chemical
classification, EPA aggregated the data values within a facility when a field duplicate was collected with
the regular sample (Section 2.4.2). The aggregation involved calculating a simple arithmetic average of
the two data values for each analyte. If one or both samples contained non-detected levels of the given
analyte, then the sample-specific detection limit entered into the calculation of this average. EPA
classified the aggregated (average) result as "detected" or "not detected" as specified in Table 2-9.
Table 2-9. Determining the Classification of Aggregated Measurements as
Detected or Not Detected
If the two sample data
values are ...
Both detected
Both not-detected
A mixture of detected and not-
detected samples
The aggregated value
is calculated as the ...
Arithmetic average of the measured values
Arithmetic average of the sample-specific
detection limits
Arithmetic average of the measured value (for
the detected sample) and sample-specific
detection limit (for the not-detected sample)
This result is
labeled ...
Detected
Not detected
Detected
11
January 2009
-------
Averaging the duplicates might artificially dampen the variability seen in the concentrations. The effects
of averaging would be most pronounced if the duplicates were taken under different wastewater treatment
conditions. Because EPA's objective was to characterize a facility's average concentration at a single
point in time, the duplicates were collected on the same day from biosolids that were treated under the
same process. The effects of averaging the duplicates, therefore, are minimal. EPA determined that the
average concentration best represented the concentration of the POTW at that point in time.
Aggregation of measurements for multiple treatment systems within a facility: When multiple samples
were collected at a facility having multiple treatment systems (Section 2.4.1), the data from these samples
were aggregated for some analytes, but not for others. For analytes within the classicals, metals, and
organics classifications, EPA aggregated the two measurements in the same way as field duplicates.
However, for the remaining analytes (i.e., PBDEs, pharmaceuticals, steroids, and hormones),
measurements often differed considerably between the two biosolids samples generated by different
systems. This difference was especially apparent between solid and liquid samples. (These differences
can be seen in the data listings and reviews presented in the appendices and Chapter 4.) Therefore, for the
statistical analyses, EPA did not aggregate the measurements for PBDEs, pharmaceuticals, steroids, and
hormones. That is, the individual sample measurements were included in the analysis as reported, rather
than their average. EPA assigned one-half of the facility's assigned survey weight to each sample
measurement within the statistical analysis.
12 January 2009
-------
3.0: OVERVIEW OF STATISTICAL METHODOLOGY
This chapter describes the statistical techniques which EPA applied to the collected sample measurements
for selected analytes. Section 3.1 describes the derivation of the survey weights assigned to the selected
facilities. Section 3.2 describes the distributional assumptions used to estimate the means and summary
statistics. Section 3.3 describes the quality assurance aspects associated with this report.
3.1 Survey Weights
Each POTW in the sampling frame had a nonzero probability of being selected for the sample. However,
as a result of the stratified sampling design, some POTWs had a different probability of being selected
than others. Therefore, EPA assigned a survey weight to each POTW contributing a biosolids sample to
the survey. The survey weight corresponds to the total number of POTWs in the sampling frame that the
selected POTW represents. The sum of all survey weights equals the total number of POTWs in the
sampling frame. By incorporating survey weights in the statistical analysis, EPA obtained estimates that
represented the entire target population.
As a first step in assigning survey weights, EPA assigned an initial "base weight" to each stratum.
Because each POTW within a stratum had an equal probability of being selected for the sample, each
selected POTW in a stratum received the same base weight. Because stratum and sample sizes differed
among the strata, different strata had different base weights.
Once all field sampling was completed, EPA calculated a final set of survey weights. This involved
adjusting the base survey weights to account for deviation between EPA's original and final sample of
POTWs (Table 2-4). The final weights should be used, rather than the base weights, when analyzing data
from this survey.
As noted in Table 2-4, the final sample size for one stratum (1 to 10 MGD) differed from its original
targeted sample size. The sample size was reduced by six facilities. This required an adjustment to the
base survey weight for this stratum. Because EPA considered the six excluded facilities to fall within the
survey's target population, this adjustment corresponded to dividing the stratum size by the actual sample
size.
The replacement of five POTWs with other facilities had no effect on the final survey weights for the
three strata. In each incidence that a replacement occurred, EPA determined that the biosolids sampled by
the replacement POTW were representative of the biosolids generated by the POTW that it replaced.
Thus, the replacement had no net impact on the sample size. To each replacement POTW, EPA assigned
the survey weight associated with the stratum for the POTW that it replaced.
Table 3-1 provides the final set of survey weights for each stratum. Within a given stratum, EPA
assigned the final survey weight to each POTW that contributed one or more biosolids samples to the
survey.
As detailed in Section 4.4.3.1, the final statistical analysis for silver excluded one plant from the
"10
-------
Table 3-1. Final Set of Survey Weights
Stratum
>100 MOD
10 to 100 MOD
1 to 10 MOD
Stratum
Size
51
543
2,743
Original
Sample Size
8
12
60
Base Weight
51/8 = 6.375
543/12 = 45.25
2,743/60 = 45.7167
Final
Sample
Size
8a
12
54
Final Weightb
51/8 = 6.375
543/12 = 45.25
2,743/54 = 50.80
a One of the eight POTWs performed final treatment of the partially-treated biosolids of a facility originally selected from
the "> 100 MOD" stratum. Thus, this replacement facility was assigned the final weight for the ">100 MOD" stratum.
b Assigned to each POTW within the final sample. The final weight, rather than the base weight, is utilized in all statistical
analyses. For silver, the final weight for the "10 to 100 MOD" stratum was 543/11 = 49.36.
3.2 Statistical Analysis Approaches
As noted in Section 2.1, EPA applied an in-depth statistical analysis to concentration data for 34 analytes,
including the survey's eight target analytes. The primary objective of the statistical analysis was to
generate national estimates of the mean, standard deviation, and selected percentiles of analyte
concentrations (i.e., 50th, 90th, 95th, 98th, 99th percentiles). EPA used one of two statistical approaches for
obtaining these national estimates: a lognormal-based approach and a nonparametric approach that did not
assume any underlying distributional form in the data. To decide which statistical approach was more
appropriate, EPA performed a preliminary investigation of the data as described in Section 4.2. In
general, EPA selected the lognormal approach unless it was clear for a particular analyte that its data were
not consistent with lognormality. An overview of each approach is given in the following subsections,
with details provided in Appendix C.
3.2.1 Lognormal Approach. This was EPA's primary statistical approach. The lognormal
approach assumed that, for a given analyte, average concentrations of biosolids among the nation's
POTWs follow a lognormal distribution. This is equivalent to assuming that the log-transformed
concentrations follow a normal distribution. Experience has shown that for a variety of environmental
media and POTW-generated discharges, including biosolids, concentrations at a given point in time
generally follow a lognormal distribution. The NSSS found that a lognormal distribution was a
reasonable assumption for concentrations of target pollutants in biosolids.
The lognormal approach takes into account the stratified sample design and the survey weights assigned
to each facility. The approach uses established equations associated with the lognormal distribution to
obtain stratum-specific estimates of the mean, standard deviation, and percentiles. These equations are
provided in Section C.I of Appendix C.
The lognormal approach treats non-detects as observations that are censored at the sample-specific
detection limit. Appendix C notes how the approach is modified in the presence of non-detects.
3.2.2 Nonparametric (Distribution Free) Approach. As an alternative to the lognormal
approach, the nonparametric approach does not assume that the data follow any particular function. The
estimates of the mean, standard deviation, and percentiles are determined solely from the observed data,
while taking into account the stratified sample design and the survey weights. The mathematical formulas
for estimating these statistics are provided in Section C.2 of Appendix C.
14
January 2009
-------
3.3 Quality Assurance
While performing the statistical analyses presented in this report, we adhered to all procedures specified
in a formal EPA-approved Quality Assurance Project Plan. The statistical analysis approach followed an
analysis plan that EPA approved prior to implementation.
For the statistical summaries and analyses presented in this report, we downloaded and utilized the final
version of the TNSSS data in SAS dataset format from EPA's mainframe computer without modification.
EPA had previously performed a comprehensive review of the laboratory data packages for data
completeness and compliance with project and method specifications. The overall objective of the data
review process was to identify any limitations apparent in the results that might affect their end use. This
information was encoded in the database through a series of qualifiers. In a few instances, EPA
determined that the laboratory results were so seriously flawed that no reasonable use could be made of
the concentration values. In these instances, EPA excluded the concentration values, but the dataset
includes the qualifiers that led to its exclusion. In all other cases, the database included the concentration
values and any qualifiers. Appendix B. 1.1 identifies and defines the qualifier codes used in the database.
Prior to statistical analysis, we further assessed the quality and integrity of the survey data relative to their
acceptability for use for the analysis. This assessment included:
• Performing exploratory analyses in which definitions of the data variables are reviewed, their
appropriate units of measure were noted, and any known relationships were assessed.
• Utilizing statistical and graphical techniques to characterize the data distribution, noting
presence of missing data, identifying outliers, and influential data points, assessing the type
and degree of censoring in the data and any censoring patterns, and determining deviation
relative to assumed underlying distributions (i.e., lognormal).
Chapter 4 documents the outcome of these assessments.
All data analysis programs were written and tested using good programming practices. Programs were
constructed to be modular, to include sufficient comments, and to include code which performs interim
validation steps on the summaries and analyses.
15 January 2009
-------
4.0: FINDINGS FROM IN-DEPTH STATISTICAL ANALYSES
As noted in Section 2.1, EPA identified eight target analytes, along with molybdenum, 4 PBDEs, 14
Pharmaceuticals, and 7 steroids and hormones, for in-depth statistical evaluation. Table 4-1 lists these 34
analytes. This chapter presents the results of the in-depth statistical analyses, including an evaluation of
detection frequency (Section 4.1); statistical graphics (Section 4.2); data review of outliers and
distributional assumptions (Section 4.3); national estimates of means and percentiles (Section 4.4); and
comparisons to current standards for land application under 40 CFR 503 and to previous survey data
(Sections 4.5 and 4.6).
Table 4-1. 34 Analytes Considered for In-Depth Statistical Analysis
Metals*
Organics*
Classicals*
PBDEs
Pharmaceuticals
Steroids and
Hormones
Barium
Beryllium
Manganese
4-Chloroaniline
Fluoranthene
Molybdenum
Silver
Pyrene
Nitrate/Nitrite
BDE-47 (2,2',4,4'- tetrabromodiphenyl)
BDE-99 (2,2',4,4',5- pentabromodiphenyl)
4-Epitetracycline (ETC)
Azithromycin
Carbamazepine
Cimetidine
Ciprofloxacin
Diphenhydramine
Doxycycline
Beta Stigmastanol
Campesterol
Cholestanol
Cholesterol
BDE-153(2,2',4,4',5,5'-hexabromodiphenyl)
BDE-209 (decabromodiphenyl)
Erythromycin-Total
Fluoxetine
Miconazole
Ofloxacin
Tetracycline (TC)
Triclocarban
Triclosan
Coprostanol
Epicoprostanol
Stigmasterol
With the exception of molybdenum, the analytes listed for these chemical classes represent the survey's target analytes.
Concentration measurements for biosolids samples collected in the TNSSS are found in six SAS datasets,
one for each analyte classification (shaded column of Table 4-1). For a given analyte, the SAS dataset
contained one record for each biosolids sample. If a sample result met EPA's quality requirements, its
data record specified either a measured value ("detected") or a sample-specific detection limit ("non-
detected"). The datasets also include any qualifier flags that EPA assigned to the sample measurements
in its quality assurance review of the laboratory data packages (see Section 3.3). Appendices A.2 through
A.6 list the sample measurements for each analyte along with detection indicators and qualifier flags.
Appendix B. 1.1 defines the qualifier flags that EPA assigned during the data quality review. Because the
SAS datasets reported the concentration values on a dry-weight basis, the data are directly comparable
across facilities without the need to consider the percentage of solids present in each sample.
4.1
National Estimates of Detection Percentages
The detection limit is generally considered to be the smallest quantity of the analyte that can be reliably
measured with that particular method. Thus, detection is related to the sensitivity of the chemical
16
January 2009
-------
analytical method, rather than a determination of the presence or absence of a particular analyte. EPA is
sometimes interested in this aspect of the data. Consequently, for each analyte, EPA used the survey data
to estimate the percentage of POTWs nationally that had detectable concentrations. Because these
estimates take into account the final survey weights, they are representative of detection percentages for
biosolids generated by POTWs within EPA's target population. Table 4-2 provides estimates of these
percentages for each of the 34 analytes listed in Table 4-1. Table 4-3 provides these estimates for the
remaining analytes.
Table 4-2 shows that all eight target analytes had detection percentages of at least 74 percent, with four
achieving 100 percent. Nine of the 11 PBDEs had detection rates of 100 percent; BDE-209 had a
detection rate of 98.5 percent (Table 4-2), while BDE-138 had a detection rate of 65.5 percent (Table 4-
3). All of the 21 pharmaceuticals, steroids, and hormones subject to in-depth statistical analysis had
estimated detection percentages of at least 90 percent (when rounded).
Table 4-2. Nationally-Representative Estimates of Detection Percentages in
Biosolids for Analytes Included in the In-Depth Statistical Analysis
Metals
Organics
Classicals
PBDEs
Pharmaceuticals
Steroids and
Hormones
Analytes
Barium
Beryllium
Manganese
Molybdenum
Silver
4-Chloroaniline
Fluoranthene
Pyrene
Nitrate/Nitrite
BDE-47
BDE-99
BDE-153
BDE-209
4-Epitetracycline (ETC)
Azithromycin
Carbamazepine
Cimetidine
Ciprofloxacin
Diphenhydramine
Doxycycline
Erythromycin-Total
Fluoxetine
Miconazole
Ofloxacin
Tetracycline (TC)
Triclocarban
Triclosan
Beta Stigmastanol
Campesterol
Cholestanol
Cholesterol
Coprostanol
Epicoprostanol
Stigmasterol
Detection Percentage
100%
98.5%
100%
100%
100%
74.4%
89.5%
84.9%
100%
100%
100%
100%
98.5%
96.0%
96.0%
96.0%
89.9%
100%
100%
92.8%
92.9%
96.1%
95.8%
98.5%
97.5%
100%
92.4%
98.5%
100%
100%
96.9%
100%
98.5%
90.1%
17
January 2009
-------
Table 4-3. Nationally-Representative Estimates of Detection Percentages in
Biosolids for Analytes Not Included in the In-Depth Statistical Analysis
Metals
Organics
Classical*
PBDEs
Pharmaceuticals
Analytes
Aluminum
Antimony
Arsenic
Boron
Cadmium
Calcium
Chromium
Cobalt
Copper
Iron
Lead
Magnesium
2-Methylnaphthalene
Benzo(a)pyrene
Fluoride
BDE-28
BDE-66
BDE-85
BDE-100
1 ,7-Dimethylxanthine
4-EACTC
4-EATC
4-ECTC
4-EOTC
Acetaminophen
Albuterol
ACTC
Anhydrotetracycline (ATC)
Caffeine
Carbadox
Cefotaxime
Chlortetracycline (CTC)
Clarithromycin
Clinafloxacin
Cloxacillin
Codeine
Cotinine
Dehydronifedipine
Demeclocycline
Digoxigenin
Digoxin
Diltiazem
Enrofloxacin
Flumequine
Gemfibrozil
Ibuprofen
Isochlortetracycline (ICTC)
Lincomycin
Detection
Percentage
100%
87.8%
100%
97.1%
100%
100%
100%
100%
100%
100%
100%
100%
40.9%
77.1%
100%
100%
100%
100%
100%
4.7%
0%
38.8%
1.4%
11.3%
3.0%
1.5%
1.5%
64.9%
47.4%
0%
0%
1.4%
54.8%
0%
0%
23.3%
47.4%
23.0%
4.6%
0%
0%
83.1%
15.8%
0%
87.8%
64.4%
1.4%
4.6%
Analytes
Mercury
Nickel
Phosphorus
Selenium
Sodium
Thallium
Tin
Titanium
Vanadium
Yttrium
Zinc
Bis(2-ethylhexyl)
phthalate
Water-Extractable
Phosphorus
BDE-138
BDE-154
BDE-183
Lomefloxacin
Metformin
Minocycline
Naproxen
Norfloxacin
Norgestimate
Ormetoprim
Oxacillin
Oxolinic Acid
Oxytetracycline (OTC)
Penicillin G
Penicillin V
Ranitidine
Roxithromycin
Sarafloxacin
Sulfachloropyridazine
Sulfadiazine
Sulfadimethoxine
Sulfamerazine
Sulfamethazine
Sulfamethizole
Sulfamethoxazole
Sulfanilamide
Sulfathiazole
Thiabendazole
Trimethoprim
Tylosin
Virginiamycin
Warfarin
Detection
Percentage
100%
100%
100%
100%
100%
94.1%
94.1%
98.5%
100%
100%
100%
100%
100%
65.5%
100%
100%
2.9%
6.5%
48.2%
50.5%
36.2%
0%
1.5%
0%
0.2%
38.2%
0%
0%
60.6%
3.0%
2.9%
3.1%
4.5%
7.0%
0.1%
2.8%
0%
40.8%
12.0%
0.1%
71.7%
27.3%
0%
18.9%
0%
18
January 2009
-------
Table 4-3. Nationally-Representative Estimates of Detection Percentages in
Biosolids for Analytes Not Included in the In-Depth Statistical Analysis (Continued)
Steroids mid
Hormones
Analytes
17 Alpha-Dihydroequilin
17 Alpha-Estradiol
17 Alpha-Ethinyl-Estradiol
17Beta-Estradiol
Androstenedione
Androsterone
Beta-Estradiol 3-Benzoate
Beta-Sitosterol
Desmosterol
Detection
Percentage
1.6%
7.2%
0%
10.6%
41.5%
65.4%
24.6%
85.5%
65.9%
Analytes
Equilenin
Equilin
Ergosterol
Estriol
Estrone
Norethindrone
Norgestrel
Progesterone
Testosterone
Detection
Percentage
1.6%
15.5%
61.3%
23.1%
75.0%
6.3%
4.7%
20.2%
21.6%
4.2
Statistical Graphics: Bar Charts and Box Plots
To provide an initial view of how the survey measurements for the 34 analytes were distributed, this
section presents two types of statistical graphical data displays: bar charts and boxplots. Both display
data without considering survey weights. As a result, these displays portray only the distribution of
analyte concentrations among the survey samples. EPA did not apply survey weights when preparing
these graphics because their purpose was to explore the distributional properties of the actual data
collected. As such, they provide insight into which statistical methods should be used. Once the
appropriate methods are decided upon, the survey weights were appropriately applied within each
approach to obtain national estimates.
In these evaluations, EPA has assumed that non-detects have the same value as the sample-specific
detection limit (or more correctly, sample-specific reporting limit). The detection limit is generally
considered to be the smallest quantity of the analyte that can be reliably measured with that particular
method. If the value could be measured with more specificity, it would have a value between zero and
the detection limit. However, for convenience, EPA has assumed the upper bound for every non-detected
value. Thus, for datasets with many non-detected values, the results from the graphical displays should
be viewed with caution.
Bar charts partition the observed range of values into groups and use vertical bars to express the number
(and percentage) of values within each group. The bar charts consider aggregated data within a POTW,
as described in Section 2.4. To distinguish between them, the bars have gray diagonals for the measured
(detected) portion and black for non-detected portion at each value range.
Within each bar chart, the horizontal axis represents the range of observed concentrations. The axis is
logarithmic, with powers of 10 equally spaced along the axis. Thus, the observed shape of the bar chart is
actually associated with the log-transformed data. If the bar chart resembles a symmetric, bell-shaped
curve, this suggests a lognormal distribution assumption is appropriate. Because the purpose of the
graphical analysis is to view the general shape of the concentrations, the vertical axis does not indicate the
number of values associated with each bar.
Boxplots provide a visual summary of the key parameters of the data distribution. The boxplots display
the sample-specific measurements as originally reported (i.e., without aggregation), with non-detects
represented by their detection limits. One boxplot represents each analyte and is interpreted as follows:
• The length of the box represents the interquartile range of the observed log-transformed
data (i.e., the distance between the 25th and the 75th percentiles).
19
January 2009
-------
• The asterisk represents the mean of the observed log-transformed data.
• The horizontal line within the box represents the median of the observed log-transformed
data.
• The vertical lines (or "whiskers") extending from both ends of the box extend to the most
extreme data value in that direction that falls within 1.5 interquartile ranges from the end
of the box.
• The open circles denote data values that exceed 1.5 interquartile ranges from the end of
the box. (While they suggest possible extreme values, they are not necessarily statistical
outliers unless they fall quite far from the end of the vertical line.)
Each boxplot is plotted along a logarithmic vertical axis. Thus, like the bar charts, the distribution
represented in each boxplot represents log-transformed data. If data originate from a lognormal
distribution, their boxplot would have the following properties:
• The asterisk (mean) and horizontal line within the box (median) would be plotted on top
of each other, midway through the vertical length of the box.
• The "whiskers" would be of equal length on each side of the box.
• The number of any open circles would be very limited and distributed equally on both
sides of the box.
4.2.1 Metals. Figure 4-la contains bar charts for each of the four metals among EPA's target
analytes, along with molybdenum. The bar chart for silver contains an isolated bar at the far right end of
the chart, suggesting a possible statistical outlier. Otherwise, each bar chart is relatively symmetric and
unimodal. This suggests that a lognormal assumption is plausible for these metals.
Figure 4-lb includes boxplots for the five metals. The boxplot for barium suggests the data are tightly
clustered around the mean and median. Thus, while its distribution may resemble a lognormal
distribution in shape, its tails (i.e., and lower and upper ends of the curve) may be "skinnier" than what is
typical for a lognormal distribution. This suggests that lognormal-based estimates for upper percentiles
may be slightly lower than what would be estimated from the observed data alone. The boxplot for silver
indicates that a few large values could influence the calculation of upper percentiles.
20 January 2009
-------
Barium
Beryllium
I
1
i
I
p
*
to
o
o
Concentration (mg/kg)
Manganese
I
I
^
2?
??
2?
%
%
m
w
%
V
%
v/
m
%
#/
%
%
m
w
%
V
.
1
1
W/
%
W
V
%
v/
%
m
m
%
m
m
%
m
V
%
v/
%
m
V
\
\
\
\
\
i
I
1
1
1
\
I
I
I
1
\
\
I
Concentration (mg/kg)
Silver
Concentration (mg/kg)
Note: Sample-specific detection limits are noted in solid black for samples associated with not detected outcomes.
Figure 4-la. Bar Charts for Metals
Concentration (mg/kg)
I
I
I
1
1
I
I
\
I
-------
Molybdenum
I
I
I
to
to
Concentration (mg/kg)
Figure 4-la. Bar Charts for Metals (Continued)
to
o
o
-------
Beryllium Manganese Nitrate
/Nitrite
Analyte
Silver Molybdenum
Figure 4-lb. Box Plots for Metals and Classicals
4.2.2 Organics. Figure 4-2a includes bar charts for the three organics that are among the target
analytes, while Figure 4-2b contains the boxplots for these analytes. They suggest that there are no
apparent outliers. However, because of the presence of a fair number of non-detected values, it is difficult
to definitively draw conclusions about distributional assumptions, especially for 4-chloroaniline with 25.6
percent of the samples being classified as non-detected. Because the shape of the detected values tend to
support lognormality, EPA concluded that this distributional assumption was likely to be appropriate.
23
January 2009
-------
4-chloroaniline
Fluoranthene
I,
Concentration (ug/kg)
Pyrene
I
Concentration (ug/kg)
Concentration (ug/kg)
to
O
O
Note: Sample-specific detection limits are noted in solid black for samples associated with not detected outcomes.
Figure 4-2a. Bar Charts for Organics
-------
4-Chloroaniline
Fluoranthene
Analyte
Pyrene
Figure 4-2b. Box Plots for Organics
4.2.3 Classicals. The target analytes include one classical compound: nitrate/nitrite. Its bar chart
appears in Figure 4-3. Its boxplot was included with the metals in Figure 4-lb. These plots demonstrate
how measurements for nitrate/nitrite are spread out along the entire range of measurements. Furthermore,
the distribution tends to be bimodal (i.e., has two distinct peaks). A few large data values appear
separated from the other values at the rightmost end of the bar chart. The boxplot suggests that the
distribution is skewed, with a few large values extending beyond the other values. Thus, the nitrate/nitrite
data do not appear to resemble a lognormal distribution.
Nitrate/Nitrite
I
I
1
1
I
I
1
1
I
1
Concentration (mg/kg)
Figure 4-3. Bar Chart for Classicals (Nitrate/Nitrite)
25
January 2009
-------
4.2.4 PBDEs. Figure 4-4a contains bar charts for the four PBDEs on which EPA performed in-
depth statistical analysis, while Figure 4-4b contains boxplots. Each bar chart has a unimodal, symmetric
shape with no obvious outliers. This suggests that lognormality is a reasonable assumption for these
PBDEs.
The boxplots in Figure 4-4b also provide strong evidence for lognormality. For each PBDE, the mean
and median log-concentrations hold similar values, while the "whiskers" on each end of the box appear
similar in length. Furthermore, measurements represented by open circles are not extreme.
26 January 2009
-------
BDE47
BDE 153
I
I
I
I
I
I
I
to
Concentration (ug/kg)
BDE 99
Concentration (ug/kg)
1
1
1
Concentration (ug/kg)
BDE 209
i
i
i
i
i
1
1
!
I
l
I
\
I
%\
%\
J
1
\
1
1
1
1
1
I
I
I
1
I
i
I
Concentration (ug/kg)
Note: The bar charts summarize n=78 measurements. Each of the six POTWs having field duplicate samples collected is represented by the average of the two sample
measurements. Sample-specific detection limits are noted in solid black for POTWs associated with not-detected outcomes
Figure 4-4a. Bar Charts for PBDEs
to
o
o
-------
Figure 4-4b. Box Plots for PBDEs
4.2.5 Pharmaceuticals, Steroids, and Hormones. Figures 4-5a and 4-5b contain bar charts for
the 14 pharmaceuticals and the seven steroids and hormones, respectively, that were included among the
analytes on which in-depth statistical analyses were performed. Figure 4-5c presents boxplots for these
21 pharmaceuticals, steroids, and hormones.
Triclocarban is the only analyte having extreme measured (detected) values on the low end of its
distribution range. Other analytes occasionally have one or two measurements that are high compared to
the others, but they do not appear to be overly extreme. Overall, considering the shapes for the detected
values in the bar charts, the lognormal assumption seems plausible for this set of 21 analytes.
In reviewing the shapes of the bar charts, EPA noted a different pattern for this set of 21 analytes than it
generally had noted for the other analytes. For the other analytes, the sample-specific detection limits and
detected values were consistent. In contrast, for some analytes in this set, the histograms show the
sample-specific detection limits to be clustered separately from the detected values.
28
January 2009
-------
4-EPITETRACYCLINE (ETC)
AZITHROMYCIN
]
I
1
I
\
1
1
1
I
1
I
I
I
I
Concentration (ug/kg)
CARBAMAZEPINE
1
I
\
\
!
m
*
\
I
'//
1
\
ft
•'/.
1
\
\
%
1
\
\
I
I
1
I
I
I
\
I
1
1
1
I
1
1
s
1
1
I
I
I
1
1
I
I
II
I
\
Concentration ug/kg)
CIMETIDINE
1
111
I
1
1
1
\
I
$
\
I
1
\
\
I
*
Concentration (ug/kg)
Concentration (ug/kg)
to
O
O
Figure 4-5a. Bar Charts for Pharmaceuticals
-------
CIPROFLOXACIN
DIPHENHYDRAMINE
I
I
I
I
1
I
1
I
I
I
I
OJ
o
DOXYCYCLINE
ERYTHROMYCIN-TOTAL
1
i.
!
I
1
I
I
\
\
\
i
1
1
1
1
i
\
\
\
1
I
i
I
I
I
Concentration (ug/kg)
Concentration (ug/kg)
Figure 4-5a. Bar Charts for Pharmaceuticals (continued)
to
O
o
-------
FLUOXETINE
MICONAZOLE
li lilll
I
I
\
\
\
I
\
ll
Concentration (ug/kg)
OFLOXACIN
Concentration (ug/kg)
1 1 1
I
I
I
I
I
p
j
s
I
I
\
1
I
\
1
p
I
I
I
Concentration (mg/kg)
TETRACYCLINE (TC)
Concentration (ug/kg)
Figure 4-5a. Bar Charts for Pharmaceuticals (continued)
ll,
I
I
I
1
I
I
^
^
^
^
%
^
%
^
%
I
I
1
1
I
I
I
I
1
to
o
o
-------
TRIG LOG ARE AN
TRIG LOS AN
I
1
\
\
\
\
\
\
\
\
\
1
J
x3 %
SA "S
M X
SA "S
I
1
\
I
I
I
\
1
I
I
\
1
1
I
\
I
s
1
I
I
$
I
1000 0.1
Concentration (mg/kg)
Concentration (mg/kg)
to
o
o
Figure 4-5a. Bar Charts for Pharmaceuticals (continued)
-------
BETA STIGMASTANOL
CAMPESTEROL
OJ
OJ
P
to
o
o
Concentration (mg/kg)
CHOLESTANOL
I
I
s
I
I
I
1
I
I
I
I
I
I
1
1
1
I
Concentration (mg/kg)
I
\
I
1
1
I
1
I
I
I
1
1
1
1
I
\
\
Concentration (ug/kg)
CHOLESTEROL
Concentration (mg/kg)
Figure 4-5b. Bar Charts for Steroids and Hormones
-------
COPROSTANOL
EPICOPROSTANOL
1
Concentration (mg/kg)
Concentration (mg/kg)
STIGMASTEROL
II
I
I
1
1
1
1
I
to
o
o
Concentration (mg/kg)
Figure 4-5b. Bar Charts for Steroids and Hormones (continued)
-------
1E+6 •
1E+5 '
1E+4-
1
| 1E+3-
1E+2-
1E+1 •
1E+0 '
]
O
!
€
c
i
!
>
ETC AZITHR- CARBAM- CIMET1- CIPROF-
OMYCIN AZEPINE DINE LOXACIN
Analyte
1E+6
1E+5
1E+4
•a
1
*= 1E+3
8
1E+2
1E+1
1E+0
I
I
I
(
O
DIPHENH-
YDRAMINE
1E+8
1E+7
1E+6
_ 1E+5
1
|
1
1E+3
1E+2
1E+1
1E+0
BETA CAMPES-
ST1GMAS- TEROL
TANOL
0
O
8
O
I
I
0
1
0
0
o
>
T I 1
11 n T
1 I 1
8 T 1
1 o
o o
e
CHOLES- CHOLES- COPROS- EPICOPR- STlGMA-
TANOL TEROL TANOL OSTANOL STEROL
Analyte
TRICLOCA- TFJICLOSAN
to
o
o
Figure 4-5c. Box Plots for Pharmaceuticals, Steroids, and Hormones
-------
4.3 Data Review
In addition to reviewing the measurement data visually through bar charts and boxplots, we applied
statistical techniques to review data for the 34 analytes in Table 4-1. The primary objectives of this
review were 1) to identify statistical outliers and evaluate whether to include them in the statistical
analysis, and 2) to decide whether to take a lognormal or nonparametric approach to estimate means and
percentiles for a particular analyte (Section 2.6 and Appendix C).
4.3.1 Statistical Outliers. We applied two statistical techniques to identify the presence of
statistical outliers among detected measurements: the "generalized extreme-Studentized deviate (BSD)
many-outlier" procedure (Rosner, 1983), and analysis of variance (ANOVA) modeling approach. Both
assumed a lognormal distribution to the data. The Rosner test was capable of identify multiple outliers
simultaneously among the observed measurements. The ANOVA model expressed average log-
transformed measurements as a linear function of the stratum (flow group) and geographic region in
which the POTW was classified. This yielded a "studentized residual" for each measurement. The
studentized residual equaled the difference between the observed log-transformed measurement and what
the model predicts for this value, divided by the estimated standard error of this difference. A
measurement could be considered extreme if its studentized residual exceeded three in absolute value.
The ANOVA model approach was applied to both unaggregated and aggregated measurements, as well as
to aggregated measurements that were weighted by the survey weights.
Not all measurements flagged by one of these statistical approaches appeared to be extreme. We assessed
these findings with information from the bar charts and boxplots (Section 4.2) and data lists (Appendices
A.2 through A.6). In each case, no analytical concerns existed that would suggest excluding these values
from the statistical analysis. However, if the validity of an extreme sample measurement was brought
into question, such as when a given sample had extremely high measurements for multiple analytes, this
could lead to excluding the measurement(s). Table 4-4 lists extreme data values that had the greatest
potential of being highly influential to the outcome of the statistical analyses, with more details beginning
in Section 4.3.3.
When considering only samples collected from plants with flow rates between 1 and 10 MGD, the liquid
sample from ID 74 contained the largest concentrations of the PBDEs subject to in-depth analysis.
Within this sample, the values for three of these PBDEs (BDE-47, BDE-99, and BDE-153) were highest
among all samples and were extreme enough to be detected as statistical outliers in Table 4-4. The fourth
PBDE (BDE-209) had a concentration of 15,000,000 ng/kg, which was relatively closer in value to the
next largest value (11,000,000 ng/kg) among plants with flow rates between 1 and 10 MGD, and it was
not extreme enough to be detected as an outlier (e.g., one sample had a higher concentration).
36 January 2009
-------
Table 4-4. Listing of Detected Measurements Labeled as Statistical Outliers for
Analytes Subject to In-Depth Statistical Analysis
Analyte
Barium
Silver
BDE47
BDE99
BDE 153
Azithromycin
Carbamazepine
Ciprofloxacin
Fluoxetine
Tetracycline (TC)
Triclocarban
Cholesterol
Plant
ID
74
27
74
74
74
69
74
21
23
21
70
21
20
48
61
74
Flow Group
1100
1
-------
Table 4-5. Results of Shapiro-Wilk Tests for Normality of Log-Transformed
Biosolids Data for Analytes Subject to In-Depth Statistical Analysis
Analyte
P-value of Shapiro-Wilk Test
Performed on log-
transformed
unaggregated
detected data
Performed on all log-
transformed
aggregated data
Metals
Barium
Beryllium
Manganese
Molybdenum
Silver
0.0152*
0.7796
0.1489
0.6203
0.0010*
0.0291*
0.9844
0.1207
0.7402
0.0007*
Organics
4-Chloroaniline
Fluoranthene
Pyrene
0.0068*
0.3604
0.6515
0.0433*
0.4850
0.7069
Classicals
Nitrate/Nitrite O.0001* 0.0001*
PBDEs
BDE47
BDE99
BDE 153
BDE 209
0.2600
0.2130
0.6190
0.2650
0.2504
0.2678
0.7709
0.4056
Pharmaceuticals
4-Epitetracycline (ETC)
Azithromycin
Carbamazepine
Cimetidine
Ciprofloxacin
Diphenhydramine
Doxycycline
Erythromycin-total
Fluoxetine
Miconazole
Ofloxacin
Tetracycline (TC)
Triclocarban
Triclosan
0.4439
0.1810
0.0006*
0.3294
0.0001*
0.2147
0.1092
0.5804
0.0738
0.1279
0.2360
0.1900
0.0002*
0.0002*
0.2019
0.0747
0.0004*
0.1466
0.0001*
0.2827
0.0024*
0.0636
0.0210*
0.0445*
0.0176*
0.0584
0.0004*
0.0004*
Steroids and Hormones
Beta Stigmastanol
Campesterol
Cholestanol
Cholesterol
Coprostanol
Epicoprostanol
Stigmasterol
0.1737
0.1834
0.0091*
0.0008*
0.0033*
0.0023*
0.3952
0.2594
0.3016
0.0244*
0.0001*
0.0127*
0.0058*
0.0001*
' P-value is below 0.05, indicating that the hypothesis of normality in log-transformed data can be rejected at the 0.05 level.
38
January 2009
-------
4.3.3 Findings from the Data Review. The following sections provide general findings and
conclusions made from data reviews for the 34 analytes, along with decisions made on the statistical
analysis approach.
4.3.3.1 Metals. For silver, we identified one extreme data value (856 mg/kg; ID 27) that was
over four times larger than the next largest value of 195 mg/kg. All tests identified this as a statistical
outlier, and it is clearly extreme within the bar chart and boxplot for silver. Upon excluding this value,
the p-value reported from the Shapiro-Wilk test was 0.20, suggesting that lognormality appeared
reasonable. Thus, we initially applied the lognormal approach to the silver data both with and without
this extreme value. EPA performed an additional evaluation of the laboratory's analytical data and
documentation, including results from a similar chemical method (Method 200.7) that supported a value
close to 856 mg/kg in the database from Method 200.8. EPA then contacted the POTW to ask if they had
ever had high silver results in their sludge before, and was told no. Their only major industrial
contributors are a dog food plant and a tire manufacturer. (The latter has a pretreatment system.) EPA
would not expect either industry to contribute much silver to the plant. Because photo processors are
known to contain silver in effluents, EPA also asked if any large photo processors discharged on the
system, and the POTW was not aware of any. Because the extreme value of 856 mg/kg appears to be an
anomaly that may not reflect normal operations at the POTW, we excluded the value from the final
statistical analyses.
In addition to silver, we decided to take the lognormal approach for each of the other three metals:
• Although the Shapiro-Wilk test formally rejected the hypothesis of lognormality for
barium at the 0.05 level, any observed deviation from lognormality appeared to be minor.
This deviation was not significant enough to warrant taking a nonparametric approach.
• For beryllium and manganese, the assumption of normality in the log-transformed data
was reasonable.
A field duplicate collected at one plant (#68357, collected at ID 19) had high concentration values for
several metals, organics, and classical compounds, especially when compared to the other sample from
the plant with which it was paired. In particular, for all metals, the concentrations associated with this
sample were two to four times higher than its paired sample and were frequently high compared to
samples from other POTWs. EPA performed an additional evaluation of the laboratory's analytical data
and documentation, and they appeared to be acceptable for both samples. The documentation indicated
that the samples were liquid sludge products collected from a large storage tank. In addition, the data for
total solids suggest that the liquid product was not particularly homogeneous, and the sampling
procedures used for this facility did not result in true duplicate samples. Because of sizeable differences
from the results of its paired sample and the data from other POTWs, the representativeness of sample
#68357 was put into question. As a result, EPA decided to exclude data from this field duplicate sample
from the statistical analyses for all metals (as well as for anions and classicals, as noted below). Thus, ID
19 was represented in the statistical analysis by one sample result rather than two.
While we identified one statistical outlier for barium (3,460 mg/kg; ID 74), it was one of two samples
collected at this POTW. Upon averaging this measurement with the other sample result for this POTW, it
lowered its influence on the distribution. In other words, the average value did not appear to be a
statistical outlier, and thus, was retained for the statistical analyses.
4.3.3.2 Organics. As was done with the metals, EPA determined that all data associated with
the field duplicate sample from ID 19 (#68357) would be excluded from the statistical analyses for the
three organics.
39 January 2009
-------
No statistical outliers were identified among the other samples for the three organics.
We selected the lognormal approach for each of the organics. Although the p-value of the Shapiro-Wilk
test for 4-chloroaniline was slightly below 0.05, its observed distribution for detected values was similar
to that for the other organics, for which lognormality appeared to be sufficient.
4.3.3.3 Classicals. The distribution of nitrate/nitrite concentrations appeared to deviate
considerably from the symmetric bell-shaped curve that signifies a lognormal distribution. The bar chart
exhibited two peaks in the data and a long right tail. The outcome of the Shapiro-Wilk test verified the
lack of lognormality. Therefore, we selected a nonparametric approach for nitrate/nitrite data analysis.
This compound may have a different distribution from others, because it is the combination of two
analytes (nitrate and nitrite).
As was done for the metals and organics, EPA excluded data associated with the field duplicate sample
(#68357) for one plant (ID 19) from the statistical analysis applied to the nitrate/nitrite data. While we
identified no other statistical outliers among the nitrate/nitrite data, this could be the result of high
variability in the data.
4.3.3.4 PBDEs. Within the four facilities having samples collected from different treatment
systems (Section 2.4.1), we observed that variability in measurement values within a facility appeared to
be a significant component of total variability for PBDEs. For several PBDEs, one sample's
measurement was two to six times higher than the other sample for that facility. If these paired
measurements were averaged, the statistical analysis would have ignored this potentially significant
source of variability. Therefore, for each of these four facilities, we included the measurements for both
samples in the statistical analyses without averaging them together. We assigned a weight to each sample
result equal to the facility's survey weight divided by two. (Deviations of this magnitude were
considerably less prevalent among the six facilities having regular and field duplicate samples collected.
Thus, measurements for the paired samples within these facilities were averaged.)
• For one of these facilities (ID 74), its liquid sample consistently had the highest
concentrations among all samples in the survey for each PBDE. For three of the four
PBDEs included in the in-depth statistical analysis, these measurements were flagged as
statistical outliers (Table 4-4). However, none of these measurements had data qualifiers
assigned to them that suggested validity concerns. Thus, none were excluded from the
O OO J '
statistical analysis.
Data for each of the four PBDEs included in the in-depth statistical analysis well-resembled a lognormal
distribution. Thus, we took the lognormal-based approach for each of these PBDEs.
4.3.3.5 Pharmaceuticals, Steroids, and Hormones. Among the set of 21 pharmaceuticals,
steroids, and hormones included in the in-depth statistical analyses, one analyte (cimetidine) had no
reported measurement for ID 34 because the laboratory result for this facility did not meet EPA's quality
assurance criteria. As noted in the data listings within Appendix A.6.2, EPA excluded at least one
measurement for 28 other pharmaceuticals, steroids, and/or hormones that were not selected for the in-
depth statistical analyses, also for quality assurance reasons.
For the six facilities with field duplicate samples collected, the measurement for one sample was
frequently no more than two times the other. Some exceptions occurred, such as for total erythromycin
and miconazole at ID 32, and for several steroids and hormones at ID 19. Even with the exceptions, EPA
40 January 2009
-------
considered the results to be reasonable, and continued to average field duplicate measurements within a
facility for pharmaceuticals, steroids, and hormones.
For some of the pharmaceuticals, steroids, and hormones, within a facility with two treatment systems,
the extent of differences in measurements between the two samples was similar to what we observed with
the PBDEs. For six analytes (azithromycin, cholestanol, cholesterol, ciprofloxacin, diphenhydramine,
ofloxacin), each of these four facilities had one sample measurement that exceeded twice the value of the
other sample, while all facilities with field duplicate samples had smaller deviations. As a result, we did
not average measurements within these facilities. Instead, we used the measurements individually in the
statistical analysis. Like the PBDEs, we assigned a weight to each sample result equal to the facility's
survey weight divided by two.
We chose the lognormal approach for each of the 21 pharmaceuticals, steroids, and hormones for the in-
depth statistical analysis. Although the p-values from the Shapiro-Wilk test were occasionally below 0.05
for some of these analytes, the bar charts and boxplots suggested that any deviation from lognormality
tended to be minor.
Although our statistical tests identified some potential outliers among seven of the 21 analytes (Table 4-4;
Appendix B.1.2), there was not sufficient evidence to warrant exclusion of any of these values from the
statistical analysis. Unlike the other analyte classifications, the outliers associated with this set of 21
analytes were on both the low and high side. No single facility was the primary source of these outliers.
4.4 National Estimates
By applying the statistical approach specified in Section 4.3, we obtained estimates of the mean, standard
deviation, and selected percentiles (99th, 98th, 95th, 90th, and 50th percentiles) for each of the 34 analytes
specified in Table 4-1. Appendix C provides details on how these estimates were calculated within each
approach (lognormal and nonparametric). Each method incorporated the final survey weights assigned to
the POTWs. Therefore, these estimates are representative of the distribution of concentrations in
biosolids for the entire target population (i.e., they represent "national" estimates).
Table 4-6 summarizes the statistical estimates for each of the 34 analytes. We list the number of data
points used in the analysis within the column labeled 'n'. The 'n' column lists three different values: 74
when multiple measurements at some POTWs were averaged, 73 under the same conditions with one
value excluded during the chemical quality assurance review, and 78 when some multiple measurements
were used separately with half of the survey weight. (See Section 2.4.3.) We also provide the estimated
number of POTWs in the target population which they represent, within the column labeled 'Est. N.'
These values equal the sum of the survey weights. When we applied the nonparametric approach (for
nitrate/nitrite only), we represented non-detects by one-half of the sample-specific detection limit.
For each of the 34 analytes, we produced two sets of estimates by applying both statistical approaches.
This was done to investigate how the estimates may differ if a different approach was taken. Both sets of
estimates are presented in Appendix D. However, EPA considers the set of estimates presented in Table
4-6 as the final set for each analyte.
For nitrate/nitrite, whose underlying data distribution did not appear to be lognormal, the estimates
presented in Table 4-6 for the mean and percentiles are higher than the estimates generated under the
lognormal approach. The standard deviation estimates for nitrate/nitrite, however, were similar between
the two methods. Because our sample size was less than 100, the nonparametric approach sets the 99th
percentile for a given analyte to the largest reported measurement.
41 January 2009
-------
Table 4-6. Nationally Representative Estimates of the Mean, Standard Deviation, and Selected Upper Percentiles of the Distribution of
Concentrations for 34 Analytes in the TNSSS
Analyte
Observed Values
Minimum
Maximum
Estimates
Percentiles
99th
98th
95th
90th
50th
Summary Statistics
Mean
Standard
Deviation
Percent
POTWs
with
Detected
Cone
Metals (mg/kg)
Barium
Beryllium
Manganese
Molybdenum
Silver*
77
0.04
35
2.51
2
2,117
2.34
14,900
86.4
195
2,230
1.81
9,700
68.7
105
1,848
1.45
6,904
55.6
82
1,396
1.04
4,156
40.5
57
1,088
0.77
2,648
30.6
42
452
0.27
540
11.4
13
572
0.38
1,165
15.3
20
443
0.37
2,231
13.8
22
100
98.5
100
100
100
Organics (ug/kg)
4-Chloroaniline
Fluoranthene
Pyrene
51
45
44
5,900
12,000
14,000
12,013
13,173
15,918
8,288
9,112
10,894
4,762
5,256
6,184
2,912
3,226
3,742
513
575
634
1,284
1,421
1,654
2,946
3,211
3,981
74.4
89.5
84.9
Classicals (mg/kg)
Nitrate/Nitrite
2
6,120
6,120
2,750
960
463
14
219
828
100
PBDEs (ng/kg)
BDE-47 (2,2',4,4'-
tetrabromodiphenyl)
BDE-99 (2,2',4,4',5-
pentabromodiphenyl)
BDE-153(2,2',4,4',5,5'-
hexabromodiphenyl)
BDE-209
(decabromodiphenyl)
73,000
64,000
9,100
150,000
5,000,000
4,000,000
410,000
17,000,000
2,650,430
2,696,928
265,395
15,836,435
2,212,077
2,248,181
220,098
11,645,502
1,688,881
1,713,370
166,454
7,360,103
1,329,167
1,346,295
129,902
4,898,034
570,448
574,559
54,117
1,162,523
709,174
716,362
68,334
2,181,237
523,791
533,447
52,685
3,462,942
100
100
100
98.5
Pharmaceuticals (ug/kg)
4-Epitetracycline (ETC)
Azithromycin
Carbamazepine
Cimetidine*
41
8
9
4
4,380
5,205
6,030
8,330
8,026
8,717
1,234
19,128
5,937
5,811
856
10,975
3,787
3,172
497
4,789
2,540
1,853
306
2,294
620
278
55
171
1,135
831
135
1,332
1,741
2,342
298
10,314
96.0
96.0
96.0
89.9
to
o
o
-------
Table 4-6. Nationally Representative Estimates of the Mean, Standard Deviation, and Selected Upper Percentiles of the Distribution of
Concentrations for 34 Analytes in the TNSSS (Continued)
Analyte
Observed Values
Minimum
Maximum
Estimates
Percentiles
99th
98th
95th
90th
50th
Summary Statistics
Mean
Standard
Deviation
Percent
POTWs
with
Detected
Cone
Pharmaceuticals (ug/kg) (cont.)
Ciprofloxacin
Diphenhydramine
Doxycycline
Erythromycin-Total
Fluoxetine
Miconazole
Ofloxacin
Tetracycline (TC)
Triclocarban
Triclosan
75
37
34
2
10
7
25
38
187
334
40,800
5,730
5,090
180
3,130
9,210
58,100
5,270
441,000
133,000
79,636
5,255
7,021
264
1,555
16,931
85,562
10,042
276,708
197,288
57,975
4,021
5,046
194
1,178
10,083
57,929
7,250
205,043
124,176
36,095
2,696
3,082
123
778
4,652
32,363
4,458
131,079
62,217
23,703
1,891
1,989
82
539
2,341
19,304
2,895
88,120
33,693
5,367
541
424
19
147
207
3,113
630
21,677
3,862
10,501
871
877
36
245
1,239
8,573
1,278
39,433
16,097
17,658
1,101
1,588
58
329
7,311
21,998
2,255
59,924
65,135
100
100
92.8
92.9
96.1
95.8
98.5
97.5
100
92.4
Steroids and Hormones (ug/kg)
Beta Stigmastanol
Campesterol
Cholestanol
Cholesterol
Coprostanol
Epicoprostanol
Stigmasterol
3,440
2,840
3,860
2,340
7,720
868
455
1,330,000
524,000
4,590,000
5,390,000
43,700,000
6,030,000
568,500
1,651,188
842,112
7,874,368
13,376,891
57,794,254
25,579,800
4,606,900
1,123,256
598,919
5,071,045
8,538,884
35,060,035
13,441,281
2,646,615
632,009
360,119
2,629,149
4,369,111
16,626,022
5,143,938
1,157,099
379,365
229,283
1,467,636
2,410,541
8,574,467
2,193,143
555,217
62,547
46,547
187,244
295,092
827,108
108,028
41,513
168,079
100,879
680,046
1,129,268
4,366,714
1,702,708
321,199
419,232
193,964
2,374,369
4,171,366
22,636,715
26,783,520
2,464,383
98.5
100
100
96.9
100
98.5
90.1
* Nitrate/nitrite estimates were estimated using the non-parametric model with not-detected values replaced with one-half of the sample specific detection limit. All other estimates
were calculated using the lognormal model.
** Outlier removed for ID 27.
-------
4.5
Comparison of Metals to Current Standards
EPA established the current standards for land application (40 CFR 503) as a ceiling (i.e., upper limit) on
the dry-weight concentrations for nine distinct metals. Table 4-7 documents these nine metals and their
land application ceiling standards, along with the maximum observed concentrations among samples
collected in the TNSSS for these nine metals. The maximum concentrations are calculated by considering
both the individual sample results (presented under "unweighted statistics") and after averaging results
within each POTW when data values for multiple samples were reported (presented under "weighted
statistics"). The number of POTWs in the sample with data values exceeding the ceiling, as well as an
estimate of the total number of POTWs in the target population that exceed the ceiling, are reported.
As noted in the table, only three metals have maximum observed concentrations exceeding their
respective land application ceiling concentrations: molybdenum, nickel, and zinc. The maximum
observed concentration for all other metals in this table are well below their respective land application
regulatory limits.
Table 4-7 also shows the number of POTWs in the survey with concentrations exceeding the specified
land application ceiling. After excluding sample #68357 for ID 19 as explained in Section 4.3.3.1, only
four samples in this study had concentrations that were greater than the land application ceiling
concentrations. Two of the samples were from a single POTW (ID 71), which exceeded the limits for
both molybdenum and nickel, while the other two samples were from other POTWs (ID 2 exceeded the
nickel standard, and ID 57 exceeded the zinc standard). When we apply the survey weights to these
POTWs to obtain national estimates, we determine that less than three percent of POTWs in the survey's
target population might be expected to exceed the land application standards for any of these three metals.
EPA notes that three percent is likely to be an overestimate, because the regulations apply only to land
application, and many facilities use other methods of disposal.
Of the POTWs observed exceeding these standards in the survey, one incinerated its treated biosolids on
site, while the others sent their biosolids to landfills. Thus, results from this survey indicate that POTWs
were generally complying with the existing land application standards for metals.
Table 4-7. Land Application Ceiling Standards for Nine Metals, and Maximum
Concentrations As Observed in Samples Collected in the TNSSS
Analyte
Arsenic
Cadmium
Copper
Lead
Mercury
Molybdenum
Nickel
Selenium
Zinc
CAS No.
7440382
7440439
7440508
7439921
7439976
7439987
7440020
7782492
7440666
Land
Application
Ceiling
(mg/kg)
75
85
4300
540
57
75
420
100
7500
Max..
Cone, of
Individual
Samples
(mg/kg)a
49.2
11.8
2580.0
450.0
8.3
86.4*
526.0*
24.7
8550.0*
Max. Cone.,
After
Averaging
(mg/kg)
49.2
11.8
1720.0
350.0
7.5
86.4*
526.0*
24.2
8550.0*
# Sampled
POTWs
Over the
Ceiling
0
0
0
0
0
1
2
0
1
Estimated #
POTWs
Nationally
Over the
Ceiling
0
0
0
0
0
45
96
0
51
Estimated
% POTWs
Nationally
Over
Ceiling
0.0
0.0
0.0
0.0
0.0
1.4
2.9
0.0
1.5
Exceeds the land application ceiling.
44
January 2009
-------
4.6 Comparison of Metals, Organics, and Classicals to NSSS Results
EPA conducted the 1988 National Sewage Sludge Survey (NSSS) to obtain national estimates of over 400
pollutants in biosolids that POTWs have treated and prepared for disposal or some other use (e.g., land
application). EPA used the data collected in the NSSS to support its development of pollutant limitations,
regulatory impact analysis, and aggregate risk analysis in the Final Standards for the Use or Disposal of
Sewage Sludge (40 CFR 503).
While it is computationally possible to compare distributional estimates between the NSSS and the
TNSSS, it is not entirely appropriate to do so. The two studies were designed with different statistical
objectives and target populations. More importantly, the TNSSS was not designed in a manner that
would allow for statistical inferences about the differences between the two studies. As a result, any
observed differences between the two sets of estimates do not necessary imply that levels have changed
from the time that the NSSS occurred. The following differences between how EPA designed the NSSS
and how the Agency designed the TNSSS result in limitations on how estimates obtained from the two
surveys can be compared:
• The target population sizes were considerably different. In 1988, EPA identified 11,307 POTWs
from the 1986 NEEDS survey, contrasted with the 3,337 POTWs for the TNSSS.
o The NSSS included an additional stratum that represented POTWs with flow rates below
1.0 MOD. One-quarter of the sample for this survey (46 out of a total of 185 POTWs
sampled) came from this stratum. The TNSSS did not represent such POTWs at all
within its sample, and therefore, in its results. Because neither survey was designed to
report stratum-specific results, distributional estimates from the NSSS cannot be obtained
from existing documents for only those POTWs with flow rates above 1.0 MGD.
o The NSSS excluded POTWs using lagoons and partial treatment from its sample, but the
TNSSS considered such POTWs as eligible.
• The TNSSS required laboratories to measure percent solids first, then adjust the aliquot of wet
sludge used to get both a consistent and a manageable amount of dry solids to extract. In the
NSSS, laboratories analyzed a standard aliquot volume, and a dry weight concentration was
obtained by mathematically adjusting the analytical measurement using the sample's percent
solids.
• The TNSSS achieved greater sensitivity in measuring organic concentrations. In addition,
TNSSS also required the laboratories to run a specific cleanup technique, called gel permeation
chromatography (GPC), to remove much of the lipid content from the raw sample extracts before
they were concentrated. It was very effective at removing the lipids as well as other
interferences. GPC was used on some samples in the NSSS, but at the discretion of the
laboratory. For the TNSSS, it was required for all samples.
• The two studies primarily focused on different sets of analytes. NSSS evaluated 412 analytes,
including many organics. TNSSS evaluated 145 analytes which largely consisted of PBDEs and
Pharmaceuticals, steroids, and hormones that were not measured in the NSSS. It also included a
few analytes that had previously been evaluated in NSSS.
Despite its concerns about the validity of comparisons between the two studies, EPA has presented a
comparison of the two surveys in Table 4-8. One set of estimates is based on data from the NSSS (SAIC,
45 January 2009
-------
2003). The second set of estimates originates from data collected in the TNSSS. Of the eight target
analytes (i.e., barium, beryllium, manganese, silver, fluoranthene, pyrene, 4-chloroaniline, and
nitrate/nitrite), only beryllium had lognormal model-based estimates reported in USEPA (1992) for its
distributional parameters, which were calculated from 1988 NSSS data. As a result, Table 4-8 presents
estimates derived from nonparametric (distribution-free) approaches. These numbers were taken from
Appendix D for TNSSS and Appendix F for NSSS, with the sample-specific detection limits substituted
for non-detected outcomes.
Table 4-8. Comparison of Distributional Parameter Estimates Between the NSSS and the TNSSS,
Obtained Using Nonparametric (Distribution-Free) Approaches
Analyte
Survey
11
N
%
Det.
Est.
Mean
Est.
S.D.
Estimated Percentiles
99th
98th
95th
90th
50th
Metals (mg/kg)
Barium
Beryllium
Manganese
Silver
NSSS
TNSSS
NSSS
TNSSS
NSSS
TNSSS
NSSS
TNSSS
176
74
176
74
176
74
176
74
7,750
3,337
7,750
3,337
7,750
3,337
7,750
3,337
100%
100%
22%
99%
100%
100%
84%
100%
673
575
1.84
0.386
538
1247
48.2
32.3
840
454
2.43
0.374
1040
2228
112
101
3000
2117
8.56
2.34
4060
14900
546
856
2370
2060
8.33
1.23
3720
7690
218
195
1730
1700
6.00
1.17
1620
3430
128
71.6
1230
1240
5.00
0.89
929
3020
75.8
42.3
499
426
0.56
0.27
276
449
25.5
13.5
Organics (ug/kg)
4-
Chloroaniline
Fluoranthene
Pyrene
NSSS
TNSSS
NSSS
TNSSS
NSSS
TNSSS
176
74
176
74
176
74
7,750
3,337
7,750
3,337
7,750
3,337
5%
76%
5%
91%
5%
85%
8640
1099
8950
1420
8850
1647
13800
1051
13400
2247
13400
2593
46700
5900
46700
12000
46700
14000
43000
3700
43000
9700
43000
10000
33300
3200
32800
6700
33000
8700
28800
2500
27700
3500
28000
3900
4760
865
4760
550
4760
620
Classicals (mg/kg)
Nitrate
Nitrite
Nitrate/Nitrite
NSSS
NSSS
TNSSS
176
176
74
7,750
7,750
3,337
95%
83%
100%
1420
201
219
5040
1210
828
26500
2920
6120
15500
2910
2750
5020
462
960
1890
215
463
96.5
12.9
13.8
Source of estimates from NSSS: SAIC (2003). Some numbers are approximate.
46
January 2009
-------
5.0: CONCLUSIONS
For a variety of targeted chemicals, the primary goal of the TNSSS was to characterize mean
concentration levels and selected percentiles of analytes in biosolids generated by the nation's POTWs
(having flow rates of at least 1.0 MOD). EPA successfully collected 84 biosolids samples from its
targeted sample size of 74 POTWs. EPA collected these samples from 69 of the 80 POTWs in its original
sample, and from five POTWs that served as replacement facilities. While EPA had anticipated that it
would find some ineligible facilities in its sample, it encountered more ineligible facilities than initially
anticipated. As a result, EPA redefined its target population to include wastewater treatment ponds as a
final treatment stage. Also, if a facility utilized partial treatment, it "followed" the biosolids to the facility
that applied a final treatment stage and sampled from that facility instead. EPA determined that the
biosolids generated by the five replacement POTWs in this survey were characteristic of final treated
biosolids for the POTWs that they replaced. Although EPA had excluded facilities known to use ponds
or to conduct only partial treatment from the sample frame, the sample design report (Appendix E) notes
that only 46 facilities with flow rates exceeding 1.0 MGD were excluded as a result (29 facilities with
ponds, and 17 facilities conducting partial treatment). This is less than 1.5 percent of the 3,337 facilities
that EPA included within the sample frame. Even if these 46 facilities had been included in the sample
frame and one or more had been selected for the sample, it is expected that the results presented in this
report would have been impacted in only a very minor way, if at all. Therefore, EPA has concluded that
the estimates generated from the survey data retain the original statistical properties of the original sample
design.
The TNSSS database, available from EPA, contains concentration measurements for 145 different
analytes. These analytes included three classicals, 28 metals, six organics (PAHs and semivolatiles), 11
PBDEs, 72 pharmaceuticals and 25 steroids and hormones. As necessary, the database qualifies each
reported measurement by noting any quality-control issues associated with the laboratory analysis. None
of these qualifications were severe enough to warrant excluding any measurement from EPA's statistical
analyses that was reported in the database.
At 64 of the 74 facilities, EPA collected a single biosolids grab sample. At the other ten facilities, EPA
collected two grab samples on a single day. The second sample represented either a field duplicate (at six
facilities) or a sample collected from a second treatment system or product generated by the facility (at
four facilities). In general, EPA found that a field duplicate sample's measurements were similar to those
of the regular sample collected at the same facility, with one measurement seldom more than twice the
other. (One exception occurred, however, at ID 19, with the field duplicate measurement being
considerably higher than its paired sample measurement.) However, measurements could differ
considerably between the two samples collected from different treatment systems or products within a
facility. This was especially true for PBDEs, pharmaceuticals, steroids, and hormones.
EPA performed an in-depth statistical analysis on data for 34 analytes. The statistical analysis produced
national estimates by incorporating survey weights based upon the statistical sample design. Thus, the
estimates represent analyte concentrations from the target population of POTWs. Except for
nitrate/nitrite, EPA assumed that lognormality procedures were appropriate for percentile estimates, based
upon its review of statistical graphics and goodness of fit tests. Nitrate/nitrite had an observed data
distribution that deviated considerably from what would be expected under lognormality. For this
analyte, EPA applied a nonparametric (distribution-free) approach to deriving estimates. The percentiles
are presented in Table ES-2 and Table 4-6 for the 34 analytes selected for in-depth review. Although
EPA was less interested in the remaining analytes and did not perform in-depth evaluation of the
statistical results for them, it has provided preliminary data summaries in Appendix B.3. The reader
should exercise caution in interpreting these preliminary summaries.
47 January 2009
-------
6.0: REFERENCES
Kish, L. (1965). Survey Sampling. New York: John Wiley and Sons, Inc.
NRC (2002). Biosolids Applied to Land: Advancing Standards and Practices. Report by the National
Research Council. Washington, DC: National Academy Press. July 2002.
Rosner, B. (1983). "Percentage Points for a Generalized BSD Many-outlier Procedure." Technometrics
25(2): 165-172.
SAIC (2003). National Pollutant Concentration Percentile Estimates from the National Sewage Sludge
Survey: Candidate Pollutants for Round Two Regulations. Facsimile to Henry Kahn from T. Stralka.
January 30.
Shapiro, SS, and Wilk, MB. (1965). "An Analysis of Variance Test for Normality (Complete Samples)."
Biometrika. 52(3,4):591-607.
United States Federal Register. 68 FR 75531. Wednesday December 31, 2003, pp. 75531-75552.
USEPA (2008). Sampling and Analysis Report for the Targeted National Sewage Sludge Survey. Final
Report, Office of Science and Technology, U.S. Environmental Protection Agency. EPA-822-R-08-
016.
USEPA (2004). Technical Background Document: Refined Biosolids Exposure and Hazard Assessment
Interim Final. Contract No. 68-C-04-006. Office of Water, Washington, DC. October 8, 2004.
USEPA (2002a). Statistical Support Document for the Development of Round 2 Biosolids Use or
Disposal Regulations. Final Report, Office of Science and Technology, U.S. Environmental
Protection Agency. EPA-822-R-02-034. April 2002.
USEPA (2002b). Wastewater Technology Fact Sheet: Facultative Lagoons. U.S. Environmental
Protection Agency. Retrieved from http://www.epa.gov/OW-OWM.html/mtb/faclagon.pdf
December 2008
USEPA (1992). Statistical Support Document for the 40 CFR, Part 503: Final Standards for the Use or
Disposal of Sewage Sludge. Volumes I and II. Final Report, Engineering and Analysis Division,
Office of Science and Technology, U.S. Environmental Protection Agency. 11 November 1992.
48 January 2009
------- |