Scientific Investigations Report 2007–5179
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2007–5179
The behavior of contaminants within the environment is influenced by many factors such as the chemical and physical properties of the contaminant and the physical, geochemical, and biological nature of the surrounding environment in which the contaminant is exposed. An important environmental parameter that can have great influence on the toxicity and mobility of contaminants is redox condition. The presence of oxygen generally is indicative of an aerobic or oxidizing environment, and the absence of oxygen is indicative of an anaerobic or reducing environment. Oxygen-reduction (Redox) condition varied among the NAWQA study units, possibly influencing the occurrence of nutrient, pesticides, and VOCs. For example, nitrate can be removed by denitrification from anoxic ground water even though fertilizers might be contributing substantial amounts of nitrate to the ground-water system. As a result, distinguishing the effects of redox condition from land use may be complex. Dissolved-oxygen (DO) concentration was the most important explanatory variable for the detection frequency of 6 of the 14 most frequently detected VOCs in shallow ground water in recently urbanized areas throughout the United States (Squillace and others, 2004), and in addition to VOCs and nitrate, some pesticides also are sensitive to redox condition (Langmuir, 1997; Barbash, 2003).
Dissolved-oxygen concentrations were not significantly different in water from wells in agricultural and urban land-use settings (fig. 3, p = 0.554). This is because of the large variability in DO concentration among and within each of the study units. In four of the study units, DO concentrations ranged from less than 0.2 to greater than 7.0 mg/L (fig. 4). Within individual study units, DO concentrations between water from wells from agricultural and urban areas can be strikingly different. In the NVBR and SACR study units, DO concentrations are significantly lower in the samples collected from agricultural wells than from the urban wells (p < 0.05), and the converse is true in the RIOG study unit (p = 0.001).
The ratio, expressed as a percentage, of nitrite to nitrite plus nitrate concentration was used to determine whether nitrite or nitrate is the dominant species in sampled ground water and whether there was a difference between agricultural and urban areas. Nitrite concentration is measured and reported independently, and as a component of the nitrite plus nitrate analysis. The ratio of the two measurements was calculated for each sample in the database, where concentrations of nitrite plus nitrate were measurable. The ratios show that nitrate is the dominant species in both agricultural and urban samples (fig. 5A, B). Because nitrate is the dominant species identified in all samples, measured nitrite plus nitrate concentrations will be referred to as nitrate concentrations in the remainder of the report.
The ratio of the ammonia concentration to the Kjeldahl nitrogen (ammonia plus organic nitrogen) concentration (fig. 5C, D) indicates that the reduced nitrogen is predominantly organic; ammonia was the dominant species in only 32 percent of the samples. The data indicate that reduced nitrogen species is the same between agricultural and urban areas. However, reduced-nitrogen speciation varies among study units. For example, a comparison of the percent ammonia from agricultural sites in the NVBR and RIOG study units shows that the median is 57 percent in the NVBR study unit (n = 39), and only 8 percent in the RIOG unit (n = 37).
The ratio of orthophosphorus to total-phosphorus concentrations in filtered ground water (fig. 5E) indicates that orthophosphorus is the dominant phosphorus species. The percentage of orthophosphorus commonly exceeds 100 percent of the total phosphorus concentrations, particularly at low concentrations. This is the result of rounding analytical results, accuracy effects that occur at low-concentration determinations, and possible matrix interferences. The data indicate that agricultural samples contained higher amounts of orthophosphorus than samples from urban areas (fig. 5F, p = 0.016).
Reduced forms of nitrogen were the least-frequently detected nutrients (fig. 6A); only 15 to 30 percent of the samples had detectable concentrations of ammonia, Kjeldahl nitrogen, and nitrite. In 80 to 90 percent of the samples there were detectable concentrations of nitrate, total phosphorus, and orthophosphorus. To determine whether the data are biased by repeated collection of samples from some sites, repeat analyses from individual sites were deleted, leaving only the most recent sample collected. The greatest change was for ammonia, where removal of repeat analyses increased the detection frequency from 15 to 21 (fig. 6B).
The data in figure 6 have been recensored, and as such do not necessarily represent how frequently contaminants would be detected in a new sampling effort undertaken today in similar areas. Comparison of detection frequencies for nitrite, nitrate, total phosphorus, and orthophosphorus shows that values for recensored and non-recensored data are within a few percent of each other. For ammonia and Kjeldahl nitrogen, however, the differences are much greater. Prior to recensoring, ammonia and Kjeldahl nitrogen were detected in 42 and 52 percent of the samples, respectively. After recensoring, these nitrogen species were detected in only 15 and 29 percent of the samples, respectively.
Breakdown of the data by land-use classification shows that nitrite, nitrate, total phosphorus, and orthophosphorus are detected slightly more frequently in agricultural areas than in urban areas (fig. 7). Conversely, ammonia and Kjeldahl nitrogen are more frequently detected in urban settings than agricultural ones.
The distribution of concentrations of nutrient species differs substantially among study units (fig. 8; table 5). Median nitrate concentrations within each study unit were below the USEPA MCL of 10 mg/L and varied substantially (p < 0.05). Median nitrite concentrations were below the NWQL LRL of 0.01 mg/L for all study units. Ammonia and Kjeldahl nitrogen were rarely detected in the CAZB, GRSL, SACR, and SANJ study units, for example, but were commonly detected in the NVBR, RIOG, and SOCA study units. In only one study unit, SOCA, were the median ammonia and Kjeldahl nitrogen concentrations greater than the NWQL LRL (0.2 mg/L).
In all study units, median orthophosphorus and total-phosphorus concentrations were less than 0.1 mg/L (fig. 8; table 5) and were determined to be significantly different among study units (p < 0.05). The greatest phosphorus concentrations, greater than 2 mg/L, were measured in the NVBR and SANJ study units (fig. 8). Orthophosphorus concentration in nearly 10 percent of the samples from the NVBR study unit exceeded 1 mg/L. Phosphorus concentrations were low in ground water because phosphorus typically binds to soil particles. Most of the NVBR agricultural samples are from the Fallon area, where ground water is commonly sulfate reducing and orthophosphorus is being released as iron and manganese oxyhydroxides dissolve (Seiler and others, 2005). Because of the relatively oxic nature of water from wells sampled from the SANJ, different processes are likely governing the occurrence of orthophosphorus in these systems.
With the exception of nitrite concentrations in urban land-use areas (p = 0.620), there were significant differences among study units with respect to nutrient concentrations in agricultural and urban areas (fig. 9, p < 0.05). Differences between agricultural and urban areas were examined in the three study units where both agricultural and urban samples were collected (NVBR, RIOG, and SACR). Comparison of land use in these study units indicates that there are differences in nutrient concentrations between agricultural and urban lands (fig. 9); however, the differences are inconsistent. For nitrate, concentrations from agricultural sites were significantly greater than those from urban sites for the RIOG unit (fig. 9, p < 0.05); however, the reverse was true for the NVBR (p < 0.05) and SACR units (p = 0.048). For ammonia and Kjeldahl nitrogen, concentrations from agricultural sites were significantly greater than those from urban sites for the NVBR study unit (fig. 9, p < 0.05). Ammonia concentrations in the RIOG study unit were higher in urban samples (p < 0.05) and Kjeldahl nitrogen was higher in agricultural areas (p = 0.016). Ammonia and Kjeldahl nitrogen concentrations were similar (p > 0.05) in agricultural and urban areas in the SACR study unit. For orthophosphorus and total phosphorus, concentrations from agricultural sites were significantly greater than those from urban sites for the NVBR study unit (p < 0.05). Although there appears to be slightly higher concentrations of these nutrients in agricultural areas than in urban areas within the RIOG study unit, the differences were not determined to be significant (p > 0.05). There were no statistical differences in median total phosphorus or orthophosphorus concentrations between the agricultural and urban areas in the SACR study unit (p > 0.05).
The distribution of nutrient concentrations among the study units probably results from differences in local area geochemistry, in particular ground-water redox conditions, rather than from differences in fertilizer application rate. Phosphorus adsorption and many of the reactions involving nitrogen are redox sensitive. Consider two areas in which initial nitrate concentrations are identical: the area with anaerobic ground water likely will have lower nitrate concentrations because of nitrogen loss through denitrification, which occurs only in anaerobic settings (table 6). Release of ammonia following oxidation of organic matter can increase ammonia concentrations in anaerobic settings because ammonia is not oxidized to nitrate.
The median nitrate concentration in samples collected from agricultural areas (2.7 mg/L) was higher (p = 0.02) than that in samples collected from urban areas (2.2 mg/L) (fig. 8). The interquartile ranges differ, with more than 25 percent of the samples in agricultural areas exceeding the MCL of 10 mg/L and less than 10 percent of the samples in urban areas exceeding the MCL. All study units had samples that contained nitrate concentrations exceeding the USEPA MCL of 10 mg/L. The median nitrate concentration exceeded the USEPA MCL for nitrate only in the CAZB study unit, but was near the MCL in the GRSL and SANJ study units (fig. 8; table 5). In three study units, the maximum nitrate concentration exceeded 50 mg/L.
Comparisons were made of the median and upper-quartile summary statistic between agricultural and urban areas. The data indicate that higher nitrate concentrations are more frequently observed in agricultural areas. The median of the upper-quartiles from agricultural areas (12.5 mg/L) is more than twice that of the urban areas (5.8 mg/L). Possible reasons that nitrate concentrations exceed the nitrate MCL more frequently in agricultural areas include greater application of nitrogen fertilizer (fig. 10A) and a greater number of houses using septic systems in agricultural areas (fig. 10B).
The differences in nutrient concentrations may be influenced by the number of samples collected from the individual study units. The effect on the summary statistics of unequal numbers of samples from the different study units is shown by comparing nitrate concentrations from the completely agricultural SANJ study unit with agricultural samples from all the other study units. The median nitrate concentration for all 273 agricultural samples in the data set is 2.7 mg/L. If, however, the 110 agricultural samples from the SANJ study unit are removed, then the median concentration falls to only 0.7 mg/L (fig. 11). The sizeable effect of samples from the SANJ study unit on the combined summary statistics results from it having a statistical distribution greatly different from that of the remainder of the samples (median 8.3 mg/L, fig. 11), and the large number of samples from the SANJ study unit. The difference in nitrate concentration in the SANJ study unit relative to other agricultural areas included in this investigation may partly be a result of greater agricultural intensity and aquifer characteristics. Agricultural fertilizer use within the SANJ study unit was similar to that in RIOG (p = 0.727), less than that in CAZB (p < 0.05), and higher than that in NVBR or SACR (p < 0.05). Samples from the SANJ study unit make up nearly 40 percent of all the agricultural samples. Similar skewing occurs with the other nutrient species but is not as pronounced because the statistical distributions of the sample groups are not as distinctly different.
Ground water at agricultural sites in the NVBR and SACR study units tends to be anoxic, with the median dissolved-oxygen concentration being less than 0.3 mg/L (fig. 4), whereas urban sites tend to be oxic, with the median dissolved-oxygen concentration exceeding 2 mg/L (fig. 4). The converse is true in the RIOG study unit, where agricultural sites tend to be more oxic than the urban sites (fig. 4). The effects of dissolved oxygen and fertilizer use on nitrate concentrations are shown in figure 12. Regardless of land use, the highest median nitrate concentrations occur in areas with the highest median dissolved-oxygen concentrations. The effect of fertilizer use on nitrate concentration is not as clearly defined as that for dissolved oxygen.
Previous studies have shown that nitrate contamination is caused by complex interactions among nitrogen sources and aquifer characteristics (Nolan 2001, p. 298-299). Correlations between nutrient concentration and various factors such as well depth and aquifer and land-use characteristics (table 4) were computed using non-parametric univariate and multivariate methods. Nutrient concentrations were significantly correlated (p < 0.05) with some factors; however, the absolute value of the Spearman rho (ρ) was generally less than 0.40, indicating minimal correlation. Scatter plots showing univariate correlations between rankings of nitrate and selected well, aquifer, and land-use data for agricultural and urban wells are shown in figures 13 and 14, respectively. These plots are typical of the data and show that there is little predictability, even though statistically significant correlations may exist between the nutrient concentrations and the variables.
During univariate statistical analyses, where the p-value is less than 0.05 and the correlation coefficient is relatively high, care must be taken in interpreting the results. For example, one would expect a significant positive relation in agricultural areas between nitrate concentrations and the amount of fertilizer applied. This is observed in the data set when all of the study units are included, with the rank of nitrate concentration significantly increasing as the rank of fertilizer application increases (fig. 13). However, regression lines for individual study units on the plot show slightly negative relations for three study units, and slightly positive relations for the other two (fig. 15). The SACR and NVBR agricultural areas appear to dominate the lower end and the CAZB and SANJ agricultural data dominate the high end. Concentrations of all the nutrient species are affected by redox conditions (table 6), and an analysis of effects of well, aquifer, and land-use characteristics cannot be made without including differences in redox among the areas. Additionally, nitrate concentrations could be influenced by more than one environmental factor.
Multivariate-logistic regression was used to develop two models for predicting the probability that nitrate concentrations would exceed the USEPA nitrate MCL of 10 mg/L in agricultural areas. The two models were based on agricultural land use, with and without the SANJ nitrate data. Similar models for predicting exceedances in urban areas could not be developed because no significant explanatory factors could be elucidated.
These logistic regression models are intended to show the relation among relevant environmental and anthropogenic factors influencing the occurrence of nitrate above the USEPA MCL in ground water underlying agricultural areas within the arid and semiarid Western United States. Model coefficients will change in relation to irrigation practices, fertilizer use, aquifer redox conditions, and, in some cases, soil characteristics.
The general logistic model is described by Hosmer and Lemeshow (1989) and by Rupert (2003). The general form of the logistic-regression equation is, P = [(ex)/(1+ex)], where “ex” denotes the exponential function, x, that contains the significant explanatory variables (Vn). P is the probability of exceeding the USEPA MCL for nitrate. Using nitrate data from all agricultural areas, excluding SANJ, the probability of exceeding the USEPA standard can be estimated by P = [(ex)/(1+ex)],
where
x = [-3.604 + 0.250(V1) + 0.849(V2) – 0.259(V3)], (1)
and
| V1 | is the amount of applied fertilizer, in tons/km2, |
| V2 | is the area irrigated by sprinkler irrigation system(s), km2*10, and |
| V3 | is the iron concentration, µg/L. |
The probability of exceeding the MCL can be estimated by using all available data from agricultural areas within the regional-study area, including SANJ, using, P = [(ex)/(1+ex)],
where
x = [-4.040 + 0.227(V1) + 0.387(V2) – 0.077(V3) + 0.023(V4)], (2)
and
| V1, V2, and V3, | are as described above, and |
| V4 | are the percentage of sand in the soil within the 500-meter buffer zone around the well. |
Generally, the probability of exceeding the USEPA MCL for nitrate is influenced by the interaction of three key parameters: irrigated area, amount of fertilizer applied, and redox condition of the aquifer. Flood- and sprinkler-irrigation practices were examined during the development of the nitrate MCL exceedance models in this study. Although flood irrigation had a negative effect on the occurrence of nitrate above the MCL, it was not a significant explanatory variable within either agricultural model (equation 1, p = 0.09; equation 2, p = 0.19). Therefore, flood irrigation was not included in either model. Alternatively, in equation 1 (the model excluding SANJ data) the addition of sprinkler irrigation was highly significant (p = 0.001) and significantly improved the strength of the model (p < 0.05). However, within the second model (equation 2), sprinkler irrigation was marginally significant at the 95-percent confidence level (p = 0.062) and only slightly improved the overall strength of the model (p = 0.06). Although sprinkler irrigation was marginally significant in the second model, because of the significance of this parameter in equation 1 and because it had the greatest positive influence on nitrate concentrations above the MCL in both models, sprinkler irrigation was retained in both models.
Fertilizer-application data were available on a county-scale basis and extrapolated into the 500-meter buffer area around each of the wells. Within each model, fertilizer application showed a similar positive relation to the occurrence of nitrate exceeding the MCL (p < 0.05). Although the addition of percent sand did not significantly influence the exceedance of the nitrate MCL in the first model (equation 1, p = 0.092), sand did affect MCL exceedance in the second model (p = 0.031).
Although dissolved oxygen was not determined to be a significant explanatory variable in the logistic-regression models, dissolved-iron concentrations are influenced by the redox condition within the well (fig. 16). Under anaerobic and reducing conditions, iron oxides undergo reductive dissolution whereby iron is released into surrounding ground water. Other studies have used the relation between redox-sensitive elements, such as iron, to develop redox models (Suzanne Paschke, U.S. Geological Survey, written commun., 2006). Redox condition, as evidenced by iron concentration, showed a negative influence on the exceedance of the nitrate MCL in both models (p < 0.05).
Although naturally occurring nitrate may play a role in the elevated nitrate concentrations observed in the shallow ground-water samples discussed in this report, at least some of the differences in nutrient concentrations among study units and between agricultural and urban land use likely results from differences in and interactions among ground-water redox condition, fertilizer application rates, and irrigation practices (table 7).
All pesticide concentrations were below the USEPA MCLs for atrazine (3 µg/L), simazine (4 µg/L), and dinoseb (7 µg/L). MCLs have not been promulgated by USEPA for other pesticides discussed in this section, but all concentrations were less than 8 µg/L, and were commonly less than 1 µg/L.
The most commonly detected pesticides in ground water belonged to the triazine, urea, amide, and carbamate classes. Predominant in ground water beneath both agricultural and urban areas were the triazine herbicides (fig. 17). Of the water samples collected from the 272 agricultural and 176 urban wells examined, 38 and 43 percent, respectively, contained detectable amounts of at least one triazine herbicide. Although water from urban wells showed a slightly higher detection frequency than water from agricultural wells for triazine herbicides, the difference was not significant (fig. 17, p = 0.215). The next most commonly detected pesticides were the urea herbicides, which were detected in water from 12 percent of the agricultural wells and 7 percent of the urban wells. Detection frequency for none of the other pesticides exceeded 10 percent. Because of relatively low detection frequencies of different pesticides and varying pesticide-use patterns, correlations between pesticide detection frequencies and the ancillary data could be made only for the triazine, urea, and metolachlor pesticides. The relation between detection frequency of these particular pesticides and the ancillary data is described below.
Simazine is used on a wide variety of crop types including corn, citrus, and vineyards (Gilliom and others, 2006) and, in addition to agricultural use, simazine also is used as a nonselective herbicide in industrial areas (Extoxnet, 1993a). Simazine was detected in water from 28 percent of the wells in agricultural areas and 17 percent of the urban wells (fig. 18). The difference in detection frequency between the two land uses was significant (p = 0.010). Simazine was detected in at least one ground-water sample collected from each of the agricultural areas. Concentrations in water from agricultural wells ranged from less than detection to 0.230 µg/L, and concentrations in water from urban areas ranged from less than detection to 0.051 µg/L (table 8). Univariate statistical analysis used to examine the relation between the ancillary data (table 4) and simazine detection frequency did not show strong correlations. However, by using nonparametric multivariate logistic regression, simazine showed stronger correlation with agricultural than urban land use, and within the agricultural model dissolved oxygen, temperature, depth to screened interval, and percent row crop were important explanatory variables (table 8). A strong model could not be made by using the urban dataset; however, general soil permeability characteristics within the buffer area around the well were determined to be important within the urban land-use setting (table 8).
Atrazine, a restricted-use herbicide, was detected in water from 16 and 24 percent of agricultural and urban wells, respectively (fig. 18). The difference in detection frequency of atrazine within the agricultural and urban land-use settings was significant (p = 0.031). Nationally, atrazine is ranked number one in agricultural use and is associated with the cultivation of corn (Gilliom and others, 2006). In addition to agricultural uses, atrazine is also used as a nonselective herbicide in industrial areas (Extoxnet, 1993b). Atrazine was detected in at least one well in four of the five agricultural areas (CAZB, NVBR, SACR, SANJ) and in at least one well in all the urban areas. Concentrations ranged from less than detection to 1.2 µg/L in agricultural areas and from less than detection to 1.58 µg/L in urban areas (table 8). Atrazine detection frequency was evaluated for correlation with the ancillary data by using multivariate-logistic regression. Generally, detection of atrazine in urban areas was correlated with dissolved oxygen, temperature, permeability of aquifer materials, and percentage of households using public-sewer systems (p < 0.05; table 8). A strong agricultural model could not be developed but temperature and permeability of aquifer materials were important factors in agricultural areas (p < 0.05; table 8).
The degradation product of atrazine, deethylatrazine (CIAT), is more stable under oxic conditions than is atrazine (Barbash, 2003). Detectable concentrations of CIAT were measured in water from both agricultural (less than detection to 0.088 µg/L) and urban (less than detection to 0.320 µg/L) wells. Evaluation of the relation between atrazine, CIAT, and dissolved oxygen showed that the differences in detection frequency between these two compounds were not significantly different (p ≥ 0.10) at either the low (< 0.5 mg/L) or high (> 4.0 mg/L) range of dissolved oxygen (fig. 19A, B). The number of wells represented within the low and high dissolved-oxygen categories for each of the two land-use settings was similar (fig. 19C). The difference between the detection of atrazine in agricultural samples within the low-oxygen category and high-oxygen category was not significantly different (p = 0.236). However, the detection frequency of CIAT was significantly higher in samples in the high dissolved-oxygen range than in the low dissolved-oxygen range in both agricultural (p = 0.003) and urban (p < 0.05) wells. Atrazine detection frequency in ground-water samples from urban areas also was determined to be significantly higher in the higher dissolved-oxygen category when compared to the lower dissolved-oxygen category (p < 0.05). Although the data indicate that atrazine is being degraded into CIAT, there is insufficient evidence to show whether this degradation is significantly influencing the detection frequency of atrazine within these land-use settings.
Prometon was detected more frequently in urban ground-water samples than in agricultural samples. This herbicide was detected in water from 25 percent of the urban (range less than detection to 0.518 µg/L) and 7 percent of the agricultural (range less than detection to 0.230 µg/L, table 8) wells, (fig. 18). Prometon was detected in four of the five urban areas (GRSL, NVBR, RIOG, SOCA) and was detected in at least one agricultural well from four of the five agricultural areas (CAZB, NVBR, RIOG, SANJ). Although there are no known prometon on crops, prometon is used to control weeds near fences, buildings, and rights-of-way in agricultural areas (Gilliom and others, 2006). Although separate multivariate logistic-regression models could not be developed for the occurrence of prometon, a general model considering all agricultural and urban land-use data was produced. The detection frequency of prometon was associated predominantly with percent urban land use, depth-to-screened interval, and average percent clay and sand in the soil within the buffer area (p < 0.05; table 8). Unlike the other triazine herbicides, dissolved oxygen was not determined to be a significant explanatory variable in its detection.
Pesticide-use data are available for all study units in the regional area on a county-scale basis, but also are available on a township-and-range scale in areas in California. With the assumption that the township-and-range-scale data portray a more accurate representation of pesticide use, comparison of pesticide-use data in California on the two scales indicates that county-scale data generally underestimate simazine use in areas where the agricultural wells were located (figs. 21A-C). There is greater variability in the application rates of simazine in the data set at the township-and-range scale (8.1 to 23,196 lbs/acre) than at the county scale (5.4 to 59.4 lbs/acre) (fig. 21C, p < 0.05). Although, simazine is used in this report as an example, other pesticides for which data were available at both scales showed similar results. On a national scale, the diversity of agricultural crop types, climate, and pesticide use made it impossible to elucidate differences among the different agricultural settings (Gilliom and others, 2006).
Urea and carbamate pesticides were detected in water from both agricultural and urban wells (fig. 17). The urea pesticide diuron was detected in 13 percent of the samples from agricultural wells but was detected in only 3 percent of the samples from urban wells. The difference in detection frequency between the samples from agricultural and urban wells was significant (p = 0.003). Diuron is a general-use herbicide used on forage crops, field crops, fruits, vegetables, nuts, and ornamental crops. Nonagricultural uses of diuron include use on rights-of-way and on irrigation and drainage ditches (Extoxnet, 1993c). Diuron concentrations ranged from less than detection to 0.920 µg/L in agricultural ground-water samples and from less than detection to 0.325 µg/L in urban samples (table 8). Important factors associated with the detection of diuron were percent land use (agricultural and urban), dissolved oxygen, temperature, general soil-permeability characteristics, and population density (table 8). Like simazine, the detection frequency of diuron was more strongly correlated with percent agricultural land use than urban land use (p < 0.05). Tebuthiuron is a broadspectrum urea herbicide used to control woody and herbaceous plants in noncropland areas, rangelands, rights-of-way, and industrial areas. Tebuthiuron was detected significantly more frequently (p = 0.008) in water from urban wells (4.5 percent) than agricultural wells (0.7 percent). Tebuthiuron was detected in water from wells located within urban areas of GRSL, NVBR, SACR, and SOCA. Tebuthiuron is used in some areas to control alfalfa (Extoxnet, 1993d), an important crop in most agricultural areas within the regional study area (CAZB, NVBR, RIOG, and SANJ). Tebuthiuron was detected in water from only two agricultural wells (one each in SACR and SANJ). Because of the relatively low detection frequency, a sufficiently strong logistic model could not be developed for this herbicide. However, the association of tebuthiuron occurrence and “recreational grasses” showed a positive correlation in urban areas (p < 0.05; table 8).
Of the carbamate pesticides, carbofuran and thiobencarb were detected in water from 1.5 and 0.7 percent of the agricultural wells, respectively (fig. 20). The agricultural detections of thiobencarb were in samples collected from two wells within the SACR study unit. Less than 1 percent of the samples from urban wells had a detection of either carbofuran (0.6 percent) or thiobencarb (0.6 percent). The differences between agricultural and urban carbofuran and thiobencarb were not determined to be statistically significant (p > 0.05). Carbofuran is a broadspectrum insecticide used to control pests such as insects, mites, and nematodes. Use of granular carbofuran was banned in 1994; however, restricted use of liquid formulations is still permitted for use on field, fruit, vegetable, and forest crops (Extoxnet, 1993e). The herbicide thiobencarb is used primarily for the cultivation of rice (Pesticide Action Network North America, 2006). The most commonly detected carbamate herbicide was molinate, which was detected in water from 2 percent of the agricultural wells (fig. 20). However, molinate was only detected in agricultural area of the SACR study unit and is likely associated with rice cultivation. The thiocarbamate EPTC was detected in water from 1 percent of the agricultural wells (fig. 20) but was only detected in the SANJ study unit. EPTC is a selective herbicide used to control annual grasses, perennial weeds, and some broadleaf weeds in beans, forage legumes, corn, and sweet potatoes (Extoxnet, 1993f).
Bentazon, metolachlor, dinoseb, and norflurazon were detected in water from 6, 6, 3, and 2 percent of the agricultural wells, respectively (fig. 17). Bentazon was detected primarily in water from wells within the SACR agricultural area. Bentazon is used primarily on soybeans, rice, and corn (Extoxnet, 1993g). The amide, metolachlor, was detected in water from 6 percent of the agricultural wells, all from the RIOG and SANJ study units, and none from the urban wells. Nationally, metolachlor was ranked number two in agricultural use and primarily is used in the cultivation of corn (Gilliom and others, 2006), but is also used on soybeans, peanuts, potatoes, cotton, and woody ornamentals (Extoxnet, 1993h). Nonagricultural uses of metolachlor include turf, nursery stock, fence lines, and landscaping (Gilliom and others, 2006). By using logistic regression, metolachlor occurrence in shallow ground water underneath agricultural areas was positively correlated to percent row crop and negatively correlated to sprinkler irrigation methods (p < 0.05, table 8). Generally, dinoseb has been used as a nonselective contact herbicide in orchards and vineyards, a selective contact herbicide in alfalfa, clover, onions, garlic, and small grains, and as a pre-emergent herbicide (Extoxnet, 1993i). Although dinoseb was banned from use in the United States in 1986, it was detected in water from eight wells in the SANJ study unit. Dinoseb does not strongly adsorb to soil and is susceptible to microbial degradation but has been determined to be persistent in some areas (Extoxnet, 1993i). The source of dinoseb to the ground water in this area is unclear. Norflurazon was detected in ground water from the SOCA and SANJ study units. Norflurazon is a pre-emergent herbicide used to control germinating annual grasses and broadleaf weeds in crops such as cotton, soybeans, citrus, nectarines, and almonds. Nonagricultural uses of norflurazon are permitted on airports and rights-of-way (Pesticide Management Education Program, 1984a).
Important factors in pesticide occurrence at a national scale include pesticide use, climate, recharge source, soil characteristics, irrigation practices, pesticide residence time within an aquifer, environmental redox conditions, pesticide physical and chemical properties, and attenuation processes (Barbash, 2003; Gilliom and others, 2006). Overall pesticide use was determined to be greatest in corn and soybean crops and lowest in wheat and alfalfa crops (Gilliom and others, 2006). Most of the crops in areas included in this study were wheat and alfalfa, with four of the five agricultural areas in this study (CAZB, NVBR, RIOG, SACR, and SANJ) reporting alfalfa as a primary crop type. Therefore, the relatively low detection frequencies of pesticides in samples collected from the arid to semiarid Western United States are not unexpected.
Trihalomethanes (THMs), solvents (tetrachloroethene, trichloroethene), and the fuel oxygenate methyl tert-butyl ether (MTBE) were detected more frequently in ground water from urban areas than agricultural areas (p < 0.05). BTEX (benzene, toluene, ethylbenzene, xylene), chlorofluorocarbons (CFCs), and fumigants (dibromochloropropane; 1,2-dichloropropane; 1,2,3-trichloropropane) were detected more commonly in agricultural areas (fig. 22). The differences between BTEX and CFC detection frequencies within agricultural and urban areas were not determined to be significant (p > 0.05).
THMs are commonly produced during the disinfection of drinking water with chlorine (Zogorski and others, 2006) when the chlorine reacts with naturally occurring organic matter present in the water (Bellar and others, 1974). The current USEPA criterion for total THM concentration is 80 µg/L (U.S. Environmental Protection Agency, 2006a); that value was not exceeded in any samples included in this investigation. The concentration of the most commonly detected THM, trichloromethane (chloroform), in urban areas ranged from less than 0.2 to 40.4 µg/L and from less than 0.2 0.34 µg/L in agricultural areas (table 9). The difference between urban and agricultural chloroform detection frequency was significant (p < 0.05). Where THMs were detected, chloroform predominated (fig. 23), which is similar to the findings of the national assessment of VOC occurrence and distribution (Zogorski and others, 2006). Brominated THM concentrations ranged from less than 0.2 to 3.39 µg/L in urban ground-water samples and were not detected in agricultural areas. Mono- and di‑chlorinated THM compounds were detected only in ground water in urban areas. As bromifications within the THM molecule increased, detection frequency decreased (fig. 23), which also was observed during the national VOC assessment (Zogorski and others, 2006). Although the detections of dichlorobromomethane (CHBrCl2) and dibromochloromethane (CHClBr2) could result from spills during their historical use as industrial solvents, reagents, and (or) flame retardants (Agency for Toxic Substances and Disease Registry, 1989; GreenFacts, 2006), Zogorski and others (2006) attributed their detection to the use of chlorine in the disinfection of water process. The THM compounds detected within the study area likely are disinfection by-products resulting from the treatment of water with chlorine and subsequent use of the treated water on landscaping (Squillace and others, 1999; Edmonds and Gellenbeck, 2002; Thiros, 2003a; Ivahnenko and Barbash, 2004).
With a logistic model, the detection frequency of chloroform in urban areas was associated with DO, pH, percentage of households on public sewer, and percentage of hydric soil within the 500-meter buffer areas (p < 0.05; table 9). Nationally, chloroform was determined to be associated with percent urban land use, aerobic conditions within the aquifer, areas with relatively high precipitation rates (high aquifer recharge areas), public supply-wells, low amounts of anaerobic soils, percentage of households on septic systems, and proximity to Resource Conservation and Recovery Act (RCRA) facilities (Zogorski and others, 2006).
The minimum DO concentration in water from wells with detectable amounts of chloroform was 0.5 mg/L (fig. 24). Under aerobic conditions, chloroform is more stable than its degradation products (Zogorski and others, 2006). In both agricultural and urban areas, chloroform was more commonly detected in ground water with DO concentrations above 1.5 mg/L than in ground water with DO concentrations below 1.0 mg/L (fig. 24).
Similar to the national results (Zogorski and others, 2006), solvents were the second most frequently detected class of VOCs (fig. 22). Tetrachloroethene (PCE) and trichloroethene (TCE) were the most commonly detected solvents in ground-water samples (fig. 25). Concentrations of PCE ranged from less than 0.2 to 89.0 µg/L and TCE concentrations ranged from less than 0.2 to 19.19 µg/L (table 9). Water from five wells in urban areas within NVBR and one well in GRSL had PCE concentrations exceeding the USEPA drinking-water criterion (5 µg/L). Water from 2 of the 5 wells in the urbanized land-use areas within the NVBR that exceeded the PCE criterion also exceeded the 5 µg/L TCE criterion. PCE is much less soluble in water (150 mg/L) than TCE (1,100 mg/L) (German Federal Ministry for Economic Cooperation and Development, 2007; Toxic Use Reduction Institute, 2007); however, both compounds can readily migrate through the unsaturated zone and move with ground water (Russell and others, 1992, p. 1-6).
PCE and TCE were detected in ground water in urban areas (GRSL, NVBR, and SOCA) and in agricultural areas (CAZB and SANJ). There are approximately 19 Superfund (CERCLA) site listings within the regional study area; however, it is unclear to what extent these CERCLA sites are influencing the ground-water quality at sampling locations included in this investigation. In the CAZB study unit, PCE and TCE detections were suspected to be from a solvent plume upgradient from the wells (Gellenbeck and Anning, 2002). The source(s) of PCE detected in samples collected from the GRSL study unit have not been determined (Thiros, 2003a) and is (are) currently under investigation. Within the urbanized areas considered in this report, PCE detection frequency was determined to be associated with DO, percentage of industrial land use, and pH (p < 0.05; table 9). Although a sufficiently strong logistic assessment could not be determined for the detection of TCE in our investigation, percentage of industrial land use was a significant explanatory variable (p = 0.001; table 9).
Although both PCE and TCE are used as industrial solvents, through reductive dehalogenation PCE can be transformed into TCE and eventually vinyl chloride in anaerobic settings (Russell and others, 1992; Vogel and McCarty, 1985). By using univariate correlation statistical analysis, PCE and TCE concentrations were positively correlated to each other (ρ = 0.391 at α = 0.05), and multivariate logistic regression indicated that PCE and TCE detections were associated with each other (p < 0.05). Dissolved-oxygen concentration in water from wells where both PCE and TCE were detected ranged from 0.2 to 8 mg/L (fig. 26). It is unclear whether the detections of TCE result from spills of TCE itself, or from the degradation of spilled PCE.
Because of air pollution caused by automobile exhaust, 1990 Clean Air Act Amendments required that oxygenates be added to gasoline (Squillace and others, 1996; Cozzarelli and Baehr, 2003). Fuel oxygenates that have been used include methyl tert-butyl ether (MTBE), ethanol, ethyl tert-butyl ether (ETBE), tert-amyl methyl ether (TAME), and di-isopropyl ether (DIPE) (Cozzarelli and Baehr, 2003). The most commonly detected oxygenate in the environment is MTBE (Cozzarelli and Baehr, 2003). In this study, MTBE was more commonly detected in urban ground water (8 percent) than agricultural ground water (less than 1 percent; p < 0.05). Most MTBE detections occurred in water from wells with low dissolved-oxygen concentrations (fig. 27). When using the urban dataset and multivariate analysis, MTBE was positively correlated with population density and percent industrial land use and negatively correlated to DO concentration (p < 0.05; table 9). MTBE concentrations exceeded the USEPA drinking-water advisory of 20–40 µg/L in samples collected from two wells in the NVBR study unit (140 µg/L and 220 µg/L) and ranged from 0.2 to 12.7 µg/L in samples from all other wells.
BTEX (benzene, toluene, ethylbenzene and xylene) compounds are aromatic-organic chemicals associated with crude oil (Cozzarelli and Baehr, 2003) and frequently are associated with fossil fuels and car exhaust. Benzene was detected in two samples from urban ground water in the NVBR study unit. Toluene was detected in water from 3 of the 238 agricultural wells (all within the SANJ study unit) and in 2 urban ground-water samples from NVBR. Ethylbenzene was not detected at concentrations greater than 0.2 µg/L in any well sampled as part of this assessment. Xylene was the most commonly detected of the BTEX compounds—50 percent of the BTEX detections—but was only detected in ground water collected from the RIOG agricultural area. The low detection frequencies and distribution pattern of these BTEX compounds make it difficult to assess factors resulting in their detection patterns.
The CFCs, trichlorofluoromethane (CFC-11) and 1,1,2-trichloro-1,2,2-trifluoroethane (CFC-113, also known as Freon), were detected in ground water as part of this study. Although CFC-113 was detected slightly more frequently in water from urban wells (1 percent) than agricultural wells (0.4 percent), this difference was not statistically significant (p = 0.445). CFC-11 was detected significantly more often in agricultural than urban areas (p = 0.021). CFC-11 has been used as a refrigerant and as an aerosol propellant during the application of insecticides (Spectrum Laboratories, Inc., 2006) and was detected in water from 12 wells in the SANJ agricultural area and 3 times in the NVBR urban area.
The most commonly detected fumigant was dibromochloropropane (DBCP) which, with the exception of one well in the CAZB study unit, was detected only in ground water from the SANJ study unit. Nationally, this fumigant also has been detected in Hawaii (Zogorski and others, 2006) and was used as a soil fumigant for the control of nematodes in vegetable and ornamental crops (U.S. Environmental Protection Agency, 2006b) until it was banned in the late 1970s, except for use on pineapples (Pesticide Management Education Program, 1984b). The other fumigants, 1,2-dichloropropane and 1,2,3-trichloropropane, were detected in water from 2 and 3 percent of the agricultural wells, respectively, with all detections being in the SANJ study unit. 1,2-dichlorpropane has been used as a soil fumigant, industrial solvent, and chemical intermediate for the synthesis of PCE (Agency for Toxic Substances and Disease Registry, 1999). The fumigant, 1,2,3-trichloropropane, was used as an industrial solvent and cleaning/degreasing agent (Agency for Toxic Substances and Disease Registry, 1995). One sample collected from the SANJ study unit showed co-occurrence of PCE, TCE, 1,2-dichloropropane, and 1,2,3-trichloropropane.
U.S. Department of the Interior | U.S. Geological Survey
URL: https://pubs.usgs.gov/sir/2007/5179
Page Contact Information: Publications Team
Page Last Modified: Thursday, 01-Dec-2016 19:50:00 EST
