Published in J. Environ. Qual. 34:420-428 (2005).
© ASA, CSSA, SSSA
677 S. Segoe Rd., Madison, WI 53711 USA
TECHNICAL REPORTS
Atmospheric Pollutants and Trace Gases
Empirical Relationship between Use, Area, and Ambient Air Concentration of Methyl Bromide
LinYing Li*,
Bruce Johnson and
Randy Segawa
California Environmental Protection Agency, Department of Pesticide Regulation, Environmental Monitoring Branch, Post Office Box 4015, Sacramento, CA 95812-4015
* Corresponding author (lli{at}cdpr.ca.gov)
Received for publication June 3, 2004.
 |
ABSTRACT
|
|---|
Methyl bromide (MeBr) is one of the most widely used soil fumigants. Human exposure to MeBr above threshold values can cause serious health problems. The exposure assessment of MeBr depends on estimation or measurement of its air concentrations. This study proposed a methodology for systematically exploring the empirical relationship between MeBr use intensity and ambient air concentrations. Monitored air concentrations were regressed to MeBr use over various spatiotemporal scales that step-wise increased around the monitoring site and monitoring period. The results showed that the goodness-of-fit varied with the spatiotemporal scale of MeBr use. The best fit was Y = 0.46 + 0.00120X (R2 = 0.95, n = 11), where Y was the 8-wk average ambient air concentration (µg/m3), and X was the weekly average use (kg/wk) over an area of 11.3 x 11.3 km (7 x 7 mi). The model was calibrated with air-monitoring data and use data of 2000, and verified with the same type data of 2001. The model estimated subchronic air concentration with reasonable accuracy.
Abbreviations: MeBr, methyl bromide
|
INTRODUCTION
|
|---|
METHYL BROMIDE is a colorless and odorless gas at normal pressure and temperature. As a broad-spectrum soil fumigant, MeBr provides excellent and reliable pest and disease control before planting (USDA, 1993). However, MeBr is an ozone-depleting compound, and is very toxic to nontarget organisms as well, including humans. Acute, subchronic, and chronic exposures to the chemical occur through inhalation. During the 1990s, California used between 6800 and 8000 Mg/yr (California Department of Pesticide Regulation, 2000). Substantial amounts of MeBr escaped into the atmosphere after soil fumigation (Yagi et al., 1995; Williams et al., 1999), and MeBr was listed as a toxic air contaminant (TAC) in California. Although the production and importation of MeBr have declined significantly in recent years as required by the Montreal Protocol of 1993 and by the U.S. Clean Air Act of 1990, the use of MeBr is likely to continue in production of important economic crops under critical use exemptions.
The exposure assessment of MeBr depends on estimation or measurement of its air concentrations. One way to estimate the air concentration is through the use of numerical modeling methods. The USEPA developed a simulation model, the Industrial Source Complex—Short Term (ISCST) model (USEPA, 1995), to estimate the distribution of pollutants based on source characteristics, weather conditions, and terrain types. For MeBr, the ISCST model has been used in field studies (Ross et al., 1996), where the source is clearly identified, its geometry and emission rates are well characterized, and other model inputs can be obtained. The modeling approach outputs spatial and temporal variability of air concentrations, which are then compared with monitored air concentrations at various receptor locations and times in an effort to refine initial flux loss estimates.
Ambient air monitoring for subchronic or chronic exposure assessment usually lasts weeks or months. The monitoring sites are generally located in the vicinity of public activities, such as residential areas and schools. Methyl bromide fumigation occurred as multiple intermittent events scattered over a large area, unlike the field-specific studies of Ross et al. (1996) that focused on a single field. The source of pollution is no longer an isolated point or area, and measured ambient air concentrations can reflect contributions from many fields at various emission rates. The source locations, strengths, and timings responsible for the observed air concentration distributions in large airsheds are rarely known with certainty. Although the ISCST model has the capability to handle multiple emission sources and schedules, it is difficult to track all fumigation activities in a degree of detail that satisfies the input requirement of the model (Honaganahalli and Seiber, 2000). As a result, mechanistic modeling using ISCST3 was not used, but rather an empirical approach correlating air concentration to MeBr usage was explored in this study.
The seasonal air concentration of MeBr is likely related to the use pattern in the nearby area, such as application amount, frequency, and density. In this study, empirical methods were explored to develop a correlation between the subchronic air concentrations of MeBr and the amount of MeBr use. Off-gassing of MeBr from soil may reach kilometers away from the application site and it is important to choose an appropriate MeBr-use area and time interval to correlate measured air concentrations with use. This statistical analysis relates the measured air concentrations to the local MeBr use, where independent parameters are the airshed size (with the monitoring station at the centroid) and time interval. By tracking model parameters while changing the MeBr-use area and period in a stepwise manner, the MeBr-use area and period responsible for the observed air concentrations can be determined. The objectives of this study are to (i) estimate the size of area (and time interval) surrounding a monitoring site where MeBr applications significantly affect air concentrations and (ii) provide a mechanism to estimate subchronic air concentrations as a function of use.
|
MATERIALS AND METHODS
|
|---|
Air Monitoring
Under the Toxic Air Contaminant Program, the California Air Resources Board conducted ambient air monitoring of MeBr in Monterey, Santa Cruz, and Kern Counties in 2000 and 2001. The California Department of Pesticide Regulation recommended sites and periods for the air monitoring. These recommendations specified historically heavy-use areas and times of peak use in these regions (California Air Resources Board, 2001a, 2001b, 2002a, 2002b).
The California Air Resources Board conducted two-year sampling and subsequent lab analysis for six sites in Monterey and Santa Cruz counties and six sites in Kern County each year (Table 1). Most monitoring sites remained the same for the two monitoring years. Only one site was replaced with a new site in the second year in each region.
The air monitoring was conducted during a time interval of intense field fumigation, which differed between geographic regions (Table 2). One monitoring period lasted 7 to 9 wk. During a monitoring period, air sampling was usually taken daily for four consecutive days each week. In the 2000 monitoring, sampling days were Monday through Thursday, while sampling days included some weekend days in the 2001 monitoring.
The daily average air concentration data used in this analysis were provided by the California Air Resources Board in its summary reports for this monitoring project (California Air Resources Board, 2001a, 2001b, 2002a, 2002b). Air concentrations of four consecutive days were averaged, and the 4-d average was assumed to be the average air concentration for that week. Monthly average air concentrations were calculated from weekly average air concentrations of 3 to 5 wk (mostly 4 wk). Two-month average air concentrations were calculated in the same way, except they included all weekly concentrations in the air-monitoring period. Average air concentrations over various time periods were calculated separately for each monitoring site.
Methyl Bromide Use
Monitoring sites of 2000 in Kern, Monterey, and Santa Cruz are illustrated in Fig. 1 and 2
, respectively, along with MeBr use in surrounding township sections during the monitoring periods. In California, the location of agricultural pesticide use is reported by section. A section is a basic land unit in the Public Land Survey System (PLSS), which is approximately 1.609 x 1.609 km (1 x 1 mi). Another basic land unit in the PLSS is a township, which consists of 36 sections, or 9.654 x 9.654 km (6 x 6 mi). A township is referenced by its meridian base (M), township direction and value (T), and range direction and value (R). A section (S) in a township is numbered 1 through 36, depending on its position in the township (Fig. 3)
. For this reason, a section is annotated as MTRS in the PLSS geocoding system. In a pesticide use report (PUR), the location of a treated field is represented by a section, or MTRS, although the treated area is much smaller than a section. Therefore, location of a fumigation field is only an approximation, not an exact site.
View larger version (68K):
[in this window]
[in a new window]
|
Fig. 1. Location of monitoring sites in Kern County in 2000, and distribution of methyl bromide (MeBr) use during the air-monitoring period (16 July–31 Aug. 2000). Each cell on the use map represents a section, approximately 1.609 x 1.609 km (or 1 x 1 mi).
|
|
View larger version (65K):
[in this window]
[in a new window]
|
Fig. 2. Location of monitoring sites in Monterey and Santa Cruz Counties in 2000, and distribution of methyl bromide (MeBr) use during the air-monitoring period (8 Sept.–2 Nov. 2000). Each cell on the use map represents a section, approximately 1.609 x 1.609 km (or 1 x 1 mi).
|
|
View larger version (67K):
[in this window]
[in a new window]
|
Fig. 3. Section, township, and use area in relation to the monitoring site. A section is approximately 1.609 x 1.609 km (or 1 x 1 mi), which is numbered 1 through 36. A township consists of 36 sections, and is approximately 9.654 x 9.654 km (or 6 x 6 mi). The 5 x 5 and 7 x 7 use areas contain 25 and 49 sections, respectively, around the monitoring site.
|
|
Methyl bromide use surrounding a monitoring site was quantified over various areas (Fig. 3). The MeBr-use area was assumed a square in shape because the location of MeBr application was reported by section. The center section contained the monitoring site, and use areas were configured around the central section (Fig. 3). A MeBr-use area was measured in the number of sections it contained. For example, an area of 5 x 5 contained 25 sections centered on a section containing a monitoring site. Weekly use in each section was determined by querying the PUR database. Thus, weekly use for any given area, made up of a matrix of township sections, can be determined (i.e., weekly use mass over areas of 1 x 1, 3 x 3, ... 15 x 15). There was one site (MET) in Kern County where the size and arrangement of nearby sections was irregular (Fig. 1). Therefore, the MET site was dropped from the analysis.
The emission of MeBr from soil could last from two to several days (Yates et al., 1996, 1997). Therefore, the use week relevant to a weekly average concentration of 2000 was defined from Friday of the previous week to Thursday of the current monitoring week. The use week included three more days before the air monitoring to capture applications that may later affect concentrations.
Methods to Relate the Air Concentration to the Methyl Bromide Use
According to the Gaussian equation, air concentration is proportional to the emission rate under fixed soil status and weather conditions. When considering a large area and over a long period, it is assumed that this linear proportionality can be extended to the relationship between air concentration and the amount of MeBr used in the area. A linear regression model was used to relate the air concentration to the MeBr use:
 | [1] |
where Y is the average air concentration (µg/m3) over a certain period (1, 3–4, and 7–8 wk), X is the weekly average MeBr use (kg/wk) over various areas in that period, and a and b are regression coefficients, representing the intercept and slope of the regression line.
Mean square error (MSE) and R2 measure the fitness of the linear regression model. The term R2 represents the percentage of variation of the dependent variable that is explained by the independent variable, and it is often referred to as the coefficient of determination. Mean square error is the average squared residual error not explained by the model, which is defined as:
 | [2] |
where n is the number of samples and Yi and 
i are measured and corresponding regression-estimated air concentrations, respectively. Higher R2 and lower MSE values correspond to better comparison between Eq. [1] and measured data, and thus are ultimately used to determine over what period the MeBr use is most closely related to air concentration.
Only air-monitoring data of 2000 were used for model construction. The least squares method was used to estimate regression coefficients a and b. Confidence intervals for a, b, and R2 were also calculated (Agresti and Finlay, 1986).
Model Validation
An independent data set from the 2001 air monitoring was used for model validation. The best regression model determined in the above section was reconstructed using the air-monitoring data and use data of 2001. If regression coefficients using 2001 data are within the 95% confidence intervals of regression coefficients using 2000 data, then the model from 2001 data is not statistically different from the 2000 model. Moreover, the monitored air concentrations of 2001 were compared with the estimated air concentrations with the 2000 model to determine the magnitude of errors.
|
RESULTS
|
|---|
Air Concentration and Methyl Bromide Use
The weekly average air concentration of MeBr illustrated both site-to-site changes and week-to-week changes (Table 3). In Kern County, CRS had higher concentrations than other sites in all seven weeks. The highest concentration (60.46 µg/m3) was observed at PMS site in Monterey County in Week 5.
View this table:
[in this window]
[in a new window]
|
Table 3. Weekly average air concentrations of ambient air monitoring for Kern, Monterey, and Santa Cruz Counties in 2000.
|
|
Air concentrations over a longer period such as 3 to 4 and 7 to 8 wk were also estimated (Table 4) using the weekly average air concentrations. Again, CRS site and PMS site had the highest average air concentrations over all three periods in Kern and Monterey counties, respectively. Visual inspection of MeBr-use maps indicates that the CRS and PMS sites were in the near vicinity of the most intensive use areas (Fig. 1 and 2). The air concentrations over the period of 7 to 8 wk agreed well with the calculated MeBr use in most areas (Table 5).
View this table:
[in this window]
[in a new window]
|
Table 4. Average air concentrations over 3 to 4 and 7 to 8 wk during the air-monitoring period for Kern, Monterey, and Santa Cruz Counties in 2000.
|
|
View this table:
[in this window]
[in a new window]
|
Table 5. Monitored average air concentrations and reported average weekly methyl bromide (MeBr) uses over the period of 7 to 8 wk for Kern, Monterey, and Santa Cruz Counties (2000 data).
|
|
Effects of Temporal Scales
The correlation coefficient between air concentration and MeBr use was significant over many areas and time periods (Table 6). The R2 values are higher over longer periods, although the threshold for a significant R2 value also increased when the number of samples decreased. For most areas, the MSE declined with longer periods. The regression model using average data over 7 to 8 wk generated the lowest MSE values. Therefore, lengthening the averaging period appeared to reduce the noise in the concentration–use relationship. Because the 7- to 8-wk averaging period yielded the highest correlation and the lowest MSE values, and because the main concern of this study was subchronic effects, analyses in the following paragraphs are based on the 7- to 8-wk average data.
View this table:
[in this window]
[in a new window]
|
Table 6. Coefficient of determination (R2) values between average air concentration (µg/m3) and average methyl bromide (MeBr) usage (kg/wk) over various areas and periods (2000 data).
|
|
Effect of Spatial Scales
Linear regression between the air concentration and MeBr use was conducted over various areas (Fig. 4)
. The R2 value increased in going from small to mid-sized areas. It peaked at the 7 x 7 area, and declined at 9 x 9 and above. The MSE showed a corresponding pattern (Table 6). Dispersion of MeBr may reach several miles away from the application sites. However, the air concentration was better correlated to MeBr use in certain sized areas around the monitoring sites.
View larger version (43K):
[in this window]
[in a new window]
|
Fig. 4. Regression models between average of weekly average air concentration and average of weekly methyl bromide (MeBr) use over various areas. Averages of air concentration and use were taken over the 7- to 8-wk period.
|
|
Although the regression coefficients and correlation coefficients differ with the size of areas, statistically, the regression lines are probably not different from each other. As shown in Table 7, only the additive constant for the 1 x 1 area was significantly different from zero. While none of the remaining additive constant estimates of a were significantly different from zero, the confidence interval shifted from the positive side to the negative side when the size of MeBr-use area increased. For the 7 x 7 MeBr-use area, the confidence interval was the narrowest, and there seemed no significant contribution to the monitored air concentration from use beyond the 7 x 7 area.
Sensitivity to PMS Data
The highest 7- to 8-wk monitoring results occurred at PMS in Monterey, which might exert heavy influence on the regression. To examine the sensitivity of regression to this data point, the regression equation for the 7 x 7 area was recalculated after omitting the PMS data point. The results were similar to the original results with all of the data points. The additive constant a changed from 0.46 to 0.93, while the multiplicative constant b changed from 0.00120 to 0.00103. The R2 value lowered from 0.95 to 0.84. The regression remained highly significant (p < 0.001).
Model Validation
In the above analysis and model construction, only air-monitoring data of 2000 were used. The air-monitoring data of 2001 and corresponding use data were employed to validate the best-fit model, which was the model over 7 x 7 area and averaged over the 7- to 8-wk period. The regression model obtained from the 2001 data was:
 | [3] |
The intercept and slope of the above regression line are both in the 95% confidence intervals of the regression coefficients based on 2000 data (Table 7). Therefore, the model from the 2001 data is not statistically different from the model based on 2000 data.
The monitored air concentrations of 2001 were compared with those estimated with the 2000 model by use in a 7 x 7 area (Fig. 5) . The order of estimated air concentrations at various sites was close to that of measured ones. The difference between estimated and measured concentrations was generally within twofold of the actual air concentration level. The 2000 air-monitoring data were better represented by the regression than the 2001 data (Fig. 5).
View larger version (15K):
[in this window]
[in a new window]
|
Fig. 5. Predicted vs. measured average air concentrations over the monitoring period of 2000 and 2001. The model use for prediction was calibrated using the 2000 air-monitoring data and use data.
|
|
|
DISCUSSION
|
|---|
This analysis examined the relationship between MeBr use, area, and time period, and measured air concentrations during the intense use period of 2000–2001. Besides use, air concentration was affected by many other factors. For example, wind direction might play an important role in determining daily or weekly average air concentration. However, over a 7- to 8-wk interval, wind direction became more random compared with a short period, and its importance in determining average air concentration decreased. The uncertainty introduced by the wind direction is largely reduced if both air concentration and use are averaged over longer time intervals. This might be one of reasons that subchronic air concentrations can be predicted with reasonable accuracy by use alone.
In the linear model Eq. [1], regression coefficients a and b have a clear meaning: a represents the predicted air concentration when there is no MeBr use in the considered area during the 7- to 8-wk period, and b represents the predicted increase of air concentration resulting from one unit increased MeBr use in the specified area and period. The coefficient b decreased exponentially when the area increased (Table 7). The coefficient a could also be interpreted as the local background concentration because it represents the estimated concentration when no MeBr is applied in the MeBr-use area defined for any particular regression. A background air concentration should be a small positive constant that is independent of the size of MeBr-use area. In other words, when the right size of MeBr-use area is reached, the background concentration (coefficient a) should be a small positive value, and it will not change significantly when the size of MeBr-use area further increases. There was a general decrease in magnitude for coefficient a as the size of the base area increased (Table 7). The coefficient a decreased when the MeBr-use area increased from 1 x 1 to 7 x 7, and its value stabilized when the MeBr-use area changed from 7 x 7 to 15 x 15 (Table 7). This suggested that MeBr use beyond the 7 x 7 area exerted very little influence on the air concentration at the center of the 7 x 7 area. This supports choosing the 7 x 7 area as the best MeBr-use area for the regression model. The local background concentration estimated in this study (0.46 µg/m3) was much higher than that of the global background concentration, which is 0.038 µg/m3, or 9.8 ppt (Lobert et al., 1995). Intensive field fumigation in the study areas led to a higher local background concentration. Therefore, background concentration is only relative to the spatiotemporal scale of study, and is subject to changes with location and time.
There are several caveats to this analysis. First, this analysis only includes pesticide use data from field fumigations. Pesticide use data for structural, commodity, and other types of MeBr fumigations were not amenable to this type of analysis because they do not include information on specific location or date. Structural or commodity fumigations may have occurred during the monitoring, but there is no way to take their contribution to the air concentrations into account. However, these effects were probably minor, based on the strength of the statistical relationships determined in the analysis and the relative amount applied for structural or commodity fumigations and soil fumigations in historical data. Also, omitting the high value from the 7 x 7 regression did not make a substantial change to the regression result. Second, while there are significant differences in emission rates between methods over a 24-h period, it is likely that there is little difference between methods in emissions over several weeks. Adjustments for method differences do not appear to be necessary for subchronic exposure mitigation. However, additional monitoring is needed to verify this assumption. Third, weekly average air concentrations were estimated from 4-d monitoring values, which might be different from those estimated from 7-d monitoring data. Nevertheless, analysis of MeBr use among weekdays and weekends did not show a biased distribution. Finally, other factors might also play a role in air concentration. Such factors may include weather conditions, use configurations within the use area, use frequency and patterns, topographical characteristics near the monitoring sites, and changes in fumigation practices as required by new regulations.
|
CONCLUSIONS
|
|---|
There were significant regression relationships between MeBr use and measured ambient air concentrations for differing time periods and differing area sizes, underscoring a fundamental positive relationship between level of MeBr use and measured air concentration. The ambient air concentration of MeBr can be estimated with reasonable accuracy and confidence from nearby use amount if the time frame and the size of MeBr-use area are appropriately chosen. The 8-wk average air concentration was affected by field fumigations that are within 4.8 to 6.4 km (3 to 4 mi) from the monitoring site. The best model was Y = 0.46 + 0.00120X (R2 = 0.95), where Y was the 8-wk average ambient air concentration (µg/m3), and X was the weekly average use (kg/wk) over an area of 11.3 x 11.3 km (7 x 7 mi). The model was validated with independent air-monitoring data and use data of a different year. The model was based on air-monitoring data in both coastal and valley areas of California, and it should be applicable to broad regions with similar weather conditions, soil properties, and application methods.
|
REFERENCES
|
|---|
- Agresti, A., and B. Finlay. 1986. Statistical methods for the social sciences. 2nd ed. Dellen Publ., San Francisco.
- California Air Resources Board. 2001a. Ambient air monitoring for methyl bromide and 1,3-dichloropropene in Kern County—Summer 2000. Project C00-028 (final report). Testing Section, Eng. and Certification Branch, Monitoring and Laboratory Div., California Air Resour. Board, Sacramento.
- California Air Resources Board. 2001b. Ambient air monitoring for methyl bromide and 1,3-dichloropropene in Monterey and Santa Cruz Counties—Fall 2000. Project C00-028 (final report). Testing Section, Eng. and Certification Branch, Monitoring and Laboratory Div., California Air Resour. Board, Sacramento.
- California Air Resources Board. 2002a. Ambient air monitoring for methyl bromide and 1,3-dichloropropene in Kern County—Summer 2001. Project P01-004 (final report). Testing Section, Eng. and Certification Branch, Monitoring and Laboratory Div., California Air Resour. Board, Sacramento.
- California Air Resources Board. 2002b. Ambient air monitoring for methyl bromide and 1,3-dichloropropene in Monterey and Santa Cruz Counties—Fall 2001. Project P01-004 (final report). Testing Section, Eng. and Certification Branch, Monitoring and Laboratory Div., California Air Resour. Board, Sacramento.
- California Department of Pesticide Regulation. 2000. Summary of pesticide use report data in 1999: Indexed by chemical. California DPR, Sacramento.
- Honaganahalli, P.S., and J.N. Seiber. 2000. Measured and predicted airshed concentrations of MeBr in an agricultural valley and applications to exposure assessment. Atmos. Environ. 34:3511–3523.
- Lobert, J.M., J.H. Butler, S.A. Montzka, L.S. Geller, R.C. Myers, and J.W. Elkins. 1995. A net sink for atmospheric CH3Br in the east Pacific Ocean. Science (Washington, DC) 267:1002–1005.[Abstract/Free Full Text]
- Ross, L.J., B. Johnson, K.D. Kim, and J. Hsu. 1996. Prediction of MeBr flux from area sources using the ISCST Model. J. Environ. Qual. 25:885–891.[Abstract/Free Full Text]
- USDA. 1993. The biologic and economic assessment of methyl bromide. USDA Natl. Agric. Pesticide Impact Assessment Program, Washington, DC.
- USEPA. 1995. User's guide for the industrial source complex (ISC3) dispersion models. Vol. 1: User instructions. Vol. 2: Description of model algorithms. USEPA Office of Air Quality and Standards, Emission, Monitoring, and Analysis Div., Research Triangle Park, NC.
- Williams, J., N.Y. Wang, and R.J. Cicerone. 1999. Methyl bromide emissions from agricultural field fumigations in California. J. Geophys. Res. [Atmos.] 104:30087–30096.
- Yagi, K., J. Williams, N.Y. Wang, and R.J. Cicerone. 1995. Atmospheric methyl bromide (CH3Br) from agricultural soil fumigations. Science (Washington, DC) 267:1979–1981.[Abstract/Free Full Text]
- Yates, S.R., F.F. Ernst, J. Gan, F. Gao, and M.V. Yates. 1996. Methyl bromide emissions from a covered field: II. Volatilization. J. Environ. Qual. 25:192–202.[Abstract/Free Full Text]
- Yates, S.R., D. Wang, F.F. Ernst, and J. Gan. 1997. Methyl bromide emissions from agricultural fields: Bare-soil, deep injection. Environ. Sci. Technol. 31:1136–1143.
Related articles in JEQ:
- This Issue in Journal of Environmental Quality
JEQ 2005 34: 403-407.
[Full Text]
This article has been cited by other articles:
|
|
|
|
 
S. A. Cryer
Predicting Soil Fumigant Air Concentrations under Regional and Diverse Agronomic Conditions
J. Environ. Qual.,
November 7, 2005;
34(6):
2197 - 2207.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
