National Water-Quality Assessment (NAWQA) Project
U.S. Geological Survey Circular 1291
Appendix 7C. Statistical summaries of organochlorine pesticide compounds in bed sediment, 1992-2001
Summary statistics for organochlorine pesticide compounds and selected pesticide groups in bed sediment are presented below as a function of land-use class (agricultural, urban, undeveloped, and mixed land-use sites). The summary statistics include detection frequencies at selected thresholds (detections greater than or equal to 5 µg/kg dry weight, detections greater than or equal to 2 µg/kg, detections greater than or equal to 1 µg/kg, and detections at any concentration), and selected concentration percentiles (50th, 75th, 90th, 95th, and maximum) for each land-use class.
Statistical summaries by land-use class:
Table 7C-1. Agricultural land use.
Table 7C-2. Urban land use.
Table 7C-3. Undeveloped land use.
Table 7C-4. Mixed land use.
How the summary statistics were calculated:
All statistics are based on one sample per stream site. In general, low-level detections of pesticides were not censored at the reporting level by the laboratory. However, because one Study Unit (the Potomac River Basin)—had reporting levels in sediment that were substantially below the normal reporting level, the Potomac sediment data were censored to make them comparable to sediment data from the other Study units. Specifically, the Potomac sediment data were censored at the lowest estimated concentration in samples from the other (non-Potomac) Study Units—in other words, any concentration in a Potomac sample that was below the lowest concentration detected in any other Study Unit sample was treated as a nondetection. If a compound was detected only in Potomac sediment samples, then this compound was censored in Potomac samples at the reporting level for that compound in non-Potomac samples. A detailed description of the approach for analyzing organochlorine pesticides in bed sediment and computing statistical summaries is provided in Appendix 8C.
For each of the four land use classes, and for each detection threshold, the frequency of detection for each pesticide compound in bed sediment was calculated as follows:
(n / N) * 100
Where N is the number of samples analyzed for the pesticide compound and n is the number of samples in which the specific pesticide compound was detected at or above the threshold. Samples with missing data were dropped from the analysis. In computing detection frequency for “all” detections, samples with raised reporting levels (that is, reported as not detected above a specified concentration that is higher than the normal reporting level, such as <10 µg/kg for a compound with a reporting level of 5 µg/kg) were treated as missing data, and were dropped from the analysis. Samples with "raised" reporting levels as high as <2 µg/kg were included (counted as nondetections) in calculating detection frequencies at or above 2 µg/kg, and samples with "raised" reporting levels as high as <5 µg/kg were included (counted as nondetections) in calculating detection frequencies at or above 5 µg/kg.
Percentiles of concentration were computed using the Statistical Analysis System (SAS) Proc Univariate procedure. Samples with missing data were dropped from the analysis. Samples were ranked from low to high concentrations within the land-use class, and all nondetections were ranked lower than any detection within the class. Any calculated percentile less than the lowest detected concentration in the land-use class was considered to be a nondetection at the reporting level and is expected to be less than the lowest detection in the land-use class.
Summary statistics are presented for individual pesticide compounds analyzed in bed sediment, as well as for two groups of related pesticide compounds, total DDT and total chlordane. (The component compounds making up each of these two pesticide groups are provided in footnotes to Appendix 1B.) In each sample, concentrations of total DDT and total chlordane were determined by summing concentrations of the individual components of each group, with missing data and nondetections treated as zero concentrations. Therefore, summed concentrations are lower-bound estimates of the concentrations for total DDT and total chlordane.
Notes for using tables:
Reporting level: This is the concentration consistently reported by the NWQL for a given pesticide compound in a given medium, which reflects the sensitivity of the method for that compound. It is determined by the NWQL on the basis of quality-control data. For samples in which a given compound was not detected, the value is reported as the reporting level, along with a "less than" remark code. For example, a routine non-detection of dieldrin in bed sediment would be reported as <1 µg/kg (dry weight).
Number of sites: This is the number of sites (which also equals the number of samples) used in calculating percentiles and detection frequencies both (1) for “all” detections and (2) at the normal reporting levels of the compounds analyzed in sediment. For detection frequencies at thresholds higher than the normal reporting level for a given compound, the number of samples may be slightly higher (on average 5 percent higher). This occurs because some samples have raised reporting levels, and such samples may be treated as missing data in detection frequency calculations at some detection thresholds, but not at others. For example, the normal reporting level for dieldrin in sediment is 1 µg/kg, so a sample reported as <3 µg/kg has a raised reporting level. This sample would be considered missing data when computing detection frequencies for “all” detections, and at thresholds of 1 µg/kg and 2 µg/kg, but it would be considered a nondetection at a threshold of 5 µg/kg. Thus, the site with this sample would not be counted in the number of sites listed for dieldrin in Table 7C (which applies to detection frequencies both for “all” detections and at the 1 µg/kg threshold), although it would be included when computing the detection frequency at the 5 µg/kg threshold.
Detection frequencies: The detection frequency indicates the percentage of sites ( samples) at which a given compound was detected, which is a function of the sensitivity of the analytical method used for that compound. Detection frequencies in bed sediment were calculated for four detection thresholds: (1) “All” detections (detections at any concentration—some as low as 0.2 µg/kg dry weight); (2) detections at or greater than 1 µg/kg dry weight; (3) detections at or greater than 2 µg/kg; and (4) detections at or greater than 5 µg/kg. The values for "All" detections should not be directly compared among compounds because reporting levels varied among compounds. Because reporting levels varied, detection frequencies were calculated using three common detection thresholds (1, 2, and 5 µg/kg). The use of detection thresholds facilitates comparisons among compounds by censoring detections to a common reference concentration. Adjustments of this type are essential in order to answer questions like "is compound x detected more often than compound y?" (Also see discussion of "censoring" data in Appendix 8C).
Only one pesticide has a higher reporting level than 5 µg/kg in sediment—toxaphene, which actually is a complex mixture and has a reporting level of 200 µg/kg. Because its total detection frequency is biased low relative to pesticide compounds with much lower reporting levels, detection frequencies for toxaphene cannot be directly compared to those of the other pesticide compounds analyzed in sediment.
Frequency of detection, in percent, was rounded to the hundredths place.
Percentiles of concentration: Concentrations measured for each pesticide are summarized using percentiles and the maximum concentration. The 50th, 75th, 90th, and 95th concentration percentiles for each land-use class are provided. The tables also include a column for "Maximum," which is the highest measured concentration at any site within this land-use class.
Percentiles provide information about the concentrations at selected points in the cumulative frequency distribution of the ranked concentrations. For example, concentrations of dieldrin in bed sediment were less than or equal to 27 µg/kg at 95 percent of urban sites (or samples), less than or equal to 12 µg/kg at 90 percent of urban sites, less than or equal to 2.1 µg/kg at 75 percent of urban sites, and nondetections at 50 percent of urban sites. If percentiles are detected concentrations (as opposed to nondetections), then the percentiles also may be interpreted as the percentage of samples in which concentrations were greater than or equal to a given concentration. For example, dieldrin concentrations in sediment were greater than or equal to 27 µg/kg at 5 percent of urban sites, greater than or equal to 12 µg/kg at 10 percent of urban sites, and greater than or equal to 2.1 µg/kg at 25 percent of urban sites (table 7C-2).
Nondetections are denoted as “ND.”
Pesticide names are presented in Appendix 1B.
Information on sampling sites and their characteristics is presented in Appendix 5C.
Downloadable concentration data are presented in Appendix 6C.
Additional information on approach and methods is presented in Appendix 8C.
For more information, contact:
Lisa H. Nowell
U.S. Geological Survey
NAWQA Pesticide Synthesis Project
6000 J Street
Sacramento , CA 95819-6129
voice: (916) 278-3096
fax: (916) 278-3070