F. Spurlock, C. Garretson, and J. Troiano

Environmental Protection Agency 
Department of Pesticide Regulation 
Environmental Monitoring and Pest Management Branch 
Environmental Hazards Assessment Program 
Sacramento, California 95814-5624 

EH 97-02


This study was partially supported by Loveland Industries who contributed to analytical expenses and supplied the organosilicon surfactant. 


The mention of commercial products, their source, or use in connection with material reported herein is not to be construed as an actual or implied endorsement of such product. 


Abstract                                                                 ii 

Acknowledgements                                               iii 

Table of Contents                                                 iv 

List of Tables                                                        vi 

List of Figures                                                       vii 

Background                                                            1 

Study Objectives                                                   2 

Chemical Incorporation                                        2 

Herbicide Substitution                                          3 

Materials and Methods                                        4 

Site Description                                                    4 

Plot Preparation                                                   4 

Treatment Applications                                       5 

Sampling                                                               5 

        deposition sampling 

        runoff water sampling 

        soil sampling 

Chemical Analysis                                                6 

Mass Balance                                                       7 

Data Analysis                                                       7 

Results                                                                  8 

Application                                                            8 

Evaluation of Surfactant Effects                        8 

Herbicide Comparisons                                        9 

Mass Balance                                                       10 

Previous Studies                                                   10 

Mathematical Description of Runoff                  11 

Conclusions                                                           12 

Literature Cited                                                   14 


Appendix A.             Soil Texture Data 

Appendix B.             Catch Can Water Application Data 

Appendix C.             Herbicide Deposition (Application) Data 

Appendix D.             Spike Recovery Data 

Appendix E.             Repeated Measures ANOVA/Multiple Comparisons 

Appendix F.             Runoff Data: Sampling Intervals and Concentrations 

Appendix G.             Post-runoff Soil Sample Data 

Appendix H.             Model Fits to Runoff Data 


Table 1. Selected herbicide properties                                                                 16 

Table 2. Plot furrows and middles soil characteristics                                         16 

Table 3. Summary QA/QC analytical data: control limits(%), reporting limits, and spike recoveries (%).                                                                                                         17 

Table 4. Data distribution characteristics.                                                             18 

Table 5. Summary of RM-ANOVAs and multiple comparisons                            19 

Table 6. Application deposition summary data                                                       20 

Table 7. Summary mass balance data for runoff experiments 

            a) mass recovered                                                                                         21 

            b) fraction of measured application                                                             22 

            c) fraction of theoretical application                                                           23 


Figure 1. Plot layout schematic.                                                                               24 

Figure 2. Structure of organosilicon surfactant.                                                     25 

Figure 3. Runoff profiles for simazine, diuron, bromacil from surfactant plots.   26 

Figure 4. Runoff profiles for simazine, diuron, and bromacil from nonsurfactant plots. 27 

Figure 5. Average runoff data and least-squares minimization fit of Eq. 2.         28 


Runoff from citrus middles can cause surface and ground water contamination problems. A study was conducted to (1) compare the differences in runoff of three herbicides: simazine, diuron, and bromacil; and (2) evaluate the effect of a commercial organosilicon surfactant on reducing runoff of three soil applied preemergent herbicides from citrus orchard middles. No effect on herbicide runoff concentrations or herbicide remaining in the middles was observed due to surfactant. Approximately 2.2 cm, or about 2/3, of the 3.3 cm 1.5 h simulated rainfall water application was recovered from the plots as runoff, emphasizing the significance of low soil infiltration rates in driving herbicide runoff. Herbicide concentrations in the runoff samples ranged from about 50 to 1,500 ug L-1, comparable to field samples of actual wintertime rain runoff in and around citrus orchards in Fresno and Tulare Counties. Although less bromacil moved off-site than simazine or diuron, the peak bromacil runoff concentrations and the fraction of applied bromacil recovered in runoff were still of practical environmental significance, ranging from 400 to 1500 µg L-1 and 5% to 16 % , respectively. These data indicate that herbicide substitution based on physical properties is not a viable approach to mitigating off-site movement from soils where permeability is limited by low infiltration rates. Across all herbicides, the total fraction of measured application recovered in the surface runoff varied from 4 to 21 percent, with an overall mean of 10 percent. An emprical mathematical relationship between herbicide concentrations and cumulative runoff volume was derived that may have potential as an analysis tool in future studies.


Fresno and Tulare are the two California counties with the largest number of confirmed premeergent herbicide detections in well water. Statistical analysis of associated soil variables indicates that herbicide detections in these counties are typically found in either of two general types of soil conditions. One condition is characterized by coarse soil types where it has been hypothesized that leaching is the predominant mechanism for pesticide movement to ground water. A second soil condition, where more than 240 wells have been confirmed positive, is characterized by the presence of hardpan soils. In many of the hard pan soil areas, an important pathway for herbicide movement to ground water appears to be direct transport. In this mechanism herbicide-containing surface runoff enters drainage structures, subsequently moving directly to deeper, more permeable subsurface regions below shallow hard pan layers. This direct transport process effectively bypasses the root zone where organic carbon, microbial activity, and herbicide breakdown rates are highest. Dry wells are commonly used to dispose of excess surface water in hard pan areas in certain areas of Fresno and Tulare Counties; concentrations of simazine, diuron, and bromacil ranging up to 1100 ppb have been detected in runoff water entering dry wells during winter rain events (Braun and Hawkins, 1991).

The most common preemergent herbicide residues found in ground water of Fresno and Tulare Counties are simazine, diuron, and bromacil. Citrus accounts for approximately 75 percent of use of these herbicides in the hardpan soil areas of Fresno and Tulare Counties. Citrus growers favor bare soil conditions in winter to enhance frost protection, so that more than half of the annual preemergent herbicide applications in citrus occur in late fall. Furthermore, many users of residual herbicides rely on natural precipitation to incorporate these materials into the soil surface for activation. Bare citrus orchard middles are often highly compacted, with correspondingly low infiltration rates. These soil characteristics inhibit downward water movement, limiting incorporation at the point of application. As a result, heavy rainfall events following herbicide applications can move unincorporated preemergent residues off-site in surface runoff where they may utimately enter the subsurface via dry wells.

A second potentially important situation where off-site movement of herbicides in surface runoff may be important to ground water quality is preemergent herbicide applications to rights-of-way (ROW). Simmons and Leyva (1994) sampled roadway storm water runoff from infiltration basin inflows, basin storage, and basin dry wells in San Joaquin County, and they reported residues of five preemergent herbicides in runoff at levels ranging up to ~ 80 µg L-1. Powell et al. (1996) conducted a study of highway runoff in Glenn County where concentrations of simazine and diuron ranging up to 570 and 2800 µg L-1, respectively, were observed in simulated and actual rainfall runoff from highway shoulder plots. ROW applications are a potentially important source of preemergent herbicides in ground water because ROW applications account for about 15 percent of simazine and diuron use in Fresno and Tulare Counties, maximum use rates for ROWs are generally 3-8 times greater than crop application rates, and ROW applications are often made in fall months prior to winter rains.

Shallow mechanical incorporation has been shown to be effective in reducing the total mass of herbicide moving off-site in runoff from middles of citrus orchards (Troiano and Garretson, 1997). However, many growers are reluctant to disturb orchard middles due to concerns over crop health and/or damaging feeder roots. In the case of ROW applications, mechanical incorporation may be physically impossible or economically prohibitive. Therefore, alternate strategies to reduce runoff of soil applied preemergent herbicides are needed.


Chemical Incorporation. One goal of this study was to evaluate an alternative method of herbicide incorporation: use of nonionic surfactant applied concomitantly with residual herbicides. While nonionic surfactants have traditionally been used to improve leaf wetting, plant uptake and rainfastness of contact herbicides (e.g., Roggenbeck et al, 1993), experimental data for surfactant effects on efficacy and off-site movement of preemergent herbicides are sparse. A study evaluating the effect of organosilicon surfactants on trifluralin residual efficacy in sugarcane suggested that nonionic surfactants improved shallow incorporation of trifluralin (Loveland Industries, unpublished data). They found that residual weed control after 90 days was effective in both shallow incorporation plots and those where nonionic surfactants had been used, whereas residual weed control in the control plots (no incorporation and no surfactant) was markedly poorer. Other studies have found that application of nonionic polymers to some soils can aid in maintaining high infiltration rates under simulated rainfall conditions (Ben-Hur et al., 1989; Letey et al., 1961), however, effects on compacted soils were not evaluated. For a soil applied herbicide, improved infiltration suggests that increased incorporation may be observed, and therefore reduced herbicide movement off-site in surface runoff. A study of nonionic organosilicon surfactants in Florida found greatly enhanced herbicidal activity of diuron and norflurazon in greenhouse pots when the surfactants are simulataneously applied with residual herbicides (Tan and Singh, 1996). The authors concluded that much lower use rates of these herbicides may be possible when used in conjunction with surfactants.

In California, anecdotal information indicates that organosilicon surfactants improve residual efficacy of soil-applied preemergent herbicides in ROW applications (Paul Washburn, Washburn Agricultural Applicators, personal communication). Improved residual control suggests that the surfactant may help maintain herbicides on-site - potentially due to improved herbicide incorporation. A possible reason for this apparent phenomena is the exceptionally low surface tensions associated with organosilicone surfactant solutions relative to other common surfactants. Low surface tensions improve spreading upon application, and entry and movement of low surface tension liquids in capillary pores occurs at substantially reduced pressures relative to solutions of higher surface tension (Hiemenz, 1986). This effect is also observed for aqueous solutions infiltrating into capillary soil pores (Taylor and Ashcroft, 1972). Thus, any significant increase in water infiltration rates will improve rainfall incorporation of a herbicide into the soil. However, the question of whether these effects are significant at economically realistic surfactant application rates has not been investigated.

Herbicide Substitution. An alternative proposal for reducing herbicide movement off-site in surface runoff is herbicide substitution: the use of substitute herbicides which, by virtue of their physicochemical properties, may have a lesser potential for off-site movement. There are little data to support or refute this idea - especially under the conditions of interest here - compacted, low permeability soils subjected to wintertime rainfall. An additional objective of this study is to compare the runoff of simazine, diuron, and bromacil under simulated rainfall conditions. Table 1 lists the most important characteristics of simazine, diuron, and bromacil that determine herbicide runoff potential: aqueous solubility and soil sorption.


Site Description. The study was conducted in a mature citrus grove located on the California State University campus in Fresno, California. The soil is a Hanford sandy loam, a coarse-loamy, mixed, nonacid,thermic, Typic Xerorthorent (USDA-SCS, 1971). Summary surface soil characteristics of the twelve experimental plots are reported in Table 2; raw data for the individual sites are provided in appendix A. National Resource Conservation Service tabulated properties for a Hanford sandy loam are bulk density ranging from 1.5 to 1.6 g cm3, and a moderately rapid permeability class of 5.1 to 15.3 cm h-1 (USDA-SCS, 1971). The measured data demonstrate a highly compacted low infiltration surface condition, with an average bulk density of 1.7 g cm-3 and infiltration rate of 0.9 cm h-1 (Troiano and Garretson, 1997). Soil analysis of cores taken in previous studies in this citrus grove show a low permeability layer at approximately the 100 cm depth where silt and clay content increase sharply (Troiano and Garretson, 1997).

Plot Preparation. The study was conducted on twelve 5.5 m by 3 m plots (e.g., fig. 1). About two weeks prior to the study, glyphosate was applied to each plot to control weeds. Row middles were leveled by hand using a shovel in cases where extreme uneveness was evident. Trees bordering the plots were pruned of any foliage extending over furrows or middles to reduce interference with the simulated rainfall event. Soil berms were built across the ends of each plot to contain the simulated rainfall water. Runoff was collected by inserting short lengths of PVC pipe into the berms at the downstream end of each furrow; the pipes were placed through the furrow berms such that runoff water could be directed into 5 gallon sample collection buckets placed in open holes immediately outside the berms.

The rainfall simulations occurred approximately 24 hours after herbicide application to each plot. Simulated rainfall was delivered using two Rainbird impact sprinklers (Model 2045PJ Maxi-Bird) placed at a 1.8 m height on metal stakes at diagonally opposite corners of the plots. Sprinkler heads were adjusted to turn in a 90 arc, allowing full coverage of the entire plot with the simulated rainfall. Water pressure was adjusted to 25 psi which allowed for a maximum radius of throw of several inches beyond the sprinkler head in the opposing corner. The total water delivery of 1.0 m3 (1000 L) to the sprinklers was measured using an in-line flow meter. Catch cans were used to measure actual water depths applied to the plots (Appendix B). Approximately 3.3 cm of water (~ 550 Liters) was intercepted by each plot during the 90 minute simulation, corresponding to a water application rate of about 2.2 cm hr-1 

Treatment Applications. Six plots were randomly assigned to each of two treatment application groups: (1) simultaneous application of simazine, diuron, and bromacil plus organosilicon surfactant, and (2) simultaneous application of simazine, diuron, and bromacil without surfactant. All twelve plots received a 20 to 30 minute pre-irrigation one day prior to herbicide application to minimize plot variabilities in antecedent soil water content for the runoff experiments. All simulated rainfall events were performed approximately 24 hours after application. Nominal herbicide target rates were 2 kg formulated product ha-1 for each herbicide. The surfactant treatment group also included application of a polyalkane ether heptamethyltrisiloxane surfactant (fig. 2) at approximately 330 mL ha-1, corresponding to about 5 oz acre-1. Current weed control costs in citrus are about $25 - $40 acre-1; the application would therefore represent about an additional $7-$8 acre-1 at current surfactant costs of roughly $1.60 oz-1. The formulations of simazine, diuron and bromacil used in the application consisted of 90%, 90%, and 80% active ingredient, respectively. The total application solution volume for each plot was approximately 0.05 Liters, and the surfactant concentration in the surfactant treatment spray solution of 0.13% (w/w) exceeded the critical micelle concentration (CMC) of the surfactant by a factor of ~15. At concentrations equal to or exceeding the CMC, surface tension remains at a minimum of approximately 22 dynes cm-1 (Knoche et al., 1991), as compared to 73 dynes cm-1 for pure water.

The treatment solutions were applied only to the plot middles using a CO2 pressurized backpack sprayer with a handheld spray boom consisting of 4 nozzles spaced 48.3 cm apart (1.5 m swath width). Furrows were covered with a plastic tarp during application. The boom nozzles were equipped with Teejet 8002 VS Driftguard Flat spray tips with 50 mesh screens. Pressure at the boom was 40 psi and an application speed of approximately 3.2 km h-1 was used to deliver the nominal solution application rate of 30 L ha-1.


Deposition Samples. Three open 7.6 cm diameter jars containing exactly 50g of air-dried seived herbicide free soil were randomly placed in each plot during herbicide application to determine deposition rate. They were immediately removed from the plot, capped, the exterior wiped clean and stored at -4 C. The raw deposition data are reported in appendix C.

Runoff water samples. Runoff water samples were collected in 8 gallon increments. As previously described, runoff water from the plot middles flowed to the furrows and was collected in calibrated 5 gallon buckets placed at the end of each furrow. Each bucket was removed when 4 gallons were collected and replaced with an empty bucket. Each set of two full buckets was combined in a large container and vigorously stirred. One L subsamples were then taken by immersing pre-labeled sample bottles into the mixed solution. Water samples were immediately refrigerated at 4C until analysis.

Soil samples 

Background soil samples. A furrow background and a middle background soil sample from each plot was submitted for analysis of simazine, diuron and bromacil. The plot middle sample was a composite of 3 individual 10 cm soil cores, while the furrow sample was a composite of 4 soil cores - 2 taken from each furrow. All furrow background samples showed nondetectable levels of the three herbicides (reporting limit of 12 ug kg-1 for simazine, diuron and bromacil); however, four plots were rejected due to relatively high levels of bromacil in the plot middles (240 - 800 ug kg-1). Substitute plots were selected as replacements after further background sampling.

Post rainfall soil samples. Immediately after all water had drained from the plot, soil samples were taken from the plot middles and plot furrows. A 10 cm long by 7.5 cm diameter bucket auger was used in 2 steps to take 15 cm deep cores from both middles and furrows. The middles soil sample for each plot was a composite of 3 individual 15 cm cores, while the furrow sample was a composite of 4 cores, 2 taken from each furrow.

Chemical Analysis. Herbicide analyses were performed by APPL laboratories using an HPLC/UV-VIS method previously developed for simazine, diuron and bromacil in water (APPL SOP# ORG038) and soil (APPL SOP# ORG039). Each extraction set included a QC blank and two matrix spike samples for each analyte. Reporting limits, control limits, spike levels and mean spike recoveries for the soil background and post rainfall simulation plot soils, soil deposition, and aqueous runoff samples are summarized in Table 3. All spike recovery data are reported in appendix D. All QC blank analyses yielded nondetectable concentrations of herbicide. Analytical recoveries for the matrix spike samples were generally low and somewhat variable, especially for the deposition and water samples (Table 3).

Mass Balance Calculations. Total herbicide mass recovered from each plot was calculated as the sum of herbicide recovered in runoff water, furrows, and plot middles. Herbicide recovered in runoff water from each plot was calculated by summing herbicide mass recovered from each sampling interval (= concentration x interval volume) and summing over all sampling intervals in a plot. Equation 1 was used to calculate post runoff herbicide mass in the furrows and middles under the assumption that the herbicides had moved no deeper than the soil coring sampling depth, z.

[1] Mass herbicide = [C ( 1+ )] x b x A x z 

where C = laboratory reported analytical herbicide concentration in furrow or middles soil samples (mass/mass, wet wt. basis), = water content of soil samples (mass water/mass dry soil), b = soil bulk density (mass soil/volume soil, Table 2), A = area of furrow or plot middles, and z = sampling depth of soil cores. Values for furrow and plot middle areas were 5.06 m2 and 16.7 m2, respectively, and post runoff soil cores were 0.152 m in length (= z in eq. [1]).

Data Analysis. The application day deposition, herbicide in runoff, total herbicide off-site (runoff plus furrow), and post runoff middles herbicide data were analyzed using two-way analysis of variance with repeated measures (RM-ANOVA) on one factor (Zar, 1996; Glantz and Slinker, 1990). The individual plots were the subjects, surfactant was a nonrepeated factor at two levels, and herbicide was a repeated, or within subject, factor at three levels (simazine, diuron, and bromacil). Multiple comparisons were conducted using the Tukey multiple comparison test.

Theoretical and measured herbicide application rates showed substantial differences, recoveries from the QC spike samples were low and variable, and herbicide mass balances were generally low and variable between plots. Therefore statistical analyses were conducted on the post-runoff analytical herbicide data expressed in three forms: using concentrations expressed on a (1) mass basis, (2) fraction of measured application basis, and (3) fraction of theoretical application basis. The three sets of statistical analyses yielded generally consistent conclusions.

Formal assumptions of between-cell homogeniety of variances and within-cell normality were evaluated using the Levene test for homogeniety of variances and the Anderson-Darling normality test, respectively (Minitab statistical software, v. 10.2). There were indications that selected group data may be non-normally distributed (e.g., fraction of measured bromacil application recovered in runoff, Table 4). The null hypothesis - that there was no difference in off-site movement of the 3 herbicides - was evaluated using a "weight-of evidence approach" based on 9 RM-ANOVAs (II, III, IV in Table 5) because nonparametric analogues for two way RM-ANOVA are not available, and ANOVA is generally considered robust to nonnormality (Zar, 1996).

In RM-ANOVA an additional condition - sufficient but not necessary - for the F test statistic confidence levels to coincide with those of the theoretical F distribution is compound symmetry (Glantz and Slinker, 1990). Compound symmetry is met when group data are equally correlated across the subjects and display equal equal variances (Zar, 1996; Glantz and Slinker, 1990). Pair-wise correlation coefficients among the three repeated measures groups (e.g., simazine, diuron, and bromacil) are reported along with results from the Levene and Anderson-Darling tests in Table 4. RM-ANOVA results are summarized in Table 5; full ANOVA tables are given in appendix E.


Applications. Deposition results are summarized in Table 6. While all individual herbicide mean depositions for the surfactant treatments were lower than the nonsurfactant treatments, differences in mean total herbicide deposition to the two surfactant treatment groups was not highly significant (p = 0.09, Table 5). Across all 12 plots, mean simazine deposition was greater than either diuron or bromacil (p < 0.01), the latter two herbicides not being significantly different at the p = 0.05 level (Table 5). The percent active ingredient in the formulations were 90, 90, and 80 percent for simazine, diuron, and bromacil, respectively. The theoretical application rates of 3370 mg/plot (simazine, diuron) and 3000 mg/plot (bromacil) corresponded to application of 2 lb formulated herbicide per acre. The 12 plot mean measured applications were 71%, 64%, and 73% of theoretical application for simazine, diuron, and bromacil, respectively.

Evaluation of Surfactant Effects. The runoff profiles of the herbicides from the plots are shown in figure 3 (surfactant treatment) and figure 4 (nonsurfactant treatment). Table 7 summarizes the runoff and soil data. All raw runoff water and post-runoff soil concentrations are reported in appendices F and G. Surfactant did not significantly effect volume of runoff water collected from the plots, herbicide mass recovered in runoff, total herbicide mass recovered off-site (= runoff + furrows), or herbicide mass remaining in the middles after runoff (Table 7). It might be argued that the water application was large enough to mask any surfactant effect over the time course of the experiment due to dilution, however, there was no significant difference in herbicide recovered in the first sampling interval between the surfactant and nonsurfactant treatments (p = 0.69, 0.23, 0.68 for herbicide mass, fraction of measured application, and fraction of theoretical application basis, respectively; Table 5). Practically speaking, this suggests that any surfactant effect would probably also be minimal for shorter rainfall events of comparable intensity as the rainfall simulation.

Differences between Herbicides. There was a statistically significant difference between the mean mass and mean fractional amounts of bromacil moving off-site relative to simazine and diuron (Table 5). This result is not entirely unexpected; the solubility of bromacil is much greater than either diuron or simazine (Table 1), suggesting a greater bromacil mass incorporated into the soil per depth of water infiltrated. However, from a practical standpoint the measured concentrations and fraction of bromacil application moving off-site are still important. Typical detection limits for bromacil in ground water samples are 0.05 - 0.1 ppb, approximately 10,000 times less than the mean initial bromacil runoff concentration (~ 1,000 ppb); even a relatively small runoff volume entering the subsurface may have a potentially large impact on ground water. The mean fraction of the bromacil applications recovered off-site was about six to eight percent, depending on assumptions of application rate. These data indicate that herbicide substitution for the purpose of mitigating off-site herbicide movement in surface runoff is probably an ineffective strategy, particulary over the range of KOC and solubility represented by simazine, diuron, and bromacil.

The net water infiltration was low in these plots; about 2/3 of the applied water was recovered as runoff (Table 7). It is apparent that the compacted soil in this orchard has a low infiltration rate. By difference between applied water and runoff water recovered, the maximum average infiltration rate observed here over the course of the simulated runoff events was about 0.7 cm hr-1. The actual average middles infiltration rate is somewhat less due to water infiltration in the furrows during the experiment. A low infiltration rate inhibits incorporation, so that even moderate rains could result in significant off-site movement in runoff - regardless of herbicide solubility. Infiltration rate is therefore a dominant factor in this soil - relative to herbicide properties - in determining the amount of herbicide runoff.

Mass Balance. Total herbicide recoveries were calculated by summing herbicide mass recovered in runoff water, post-runoff herbicide mass in the soil furrows, and post-runoff herbicide mass in the plot middles. The calculated recoveries were somewhat low, ranging from about 60 to 75 percent of measured application. Two factors that contribute to the low recoveries are (1) low analytical recoveries as indicated by the ongoing quality control soil and water matrix spikes (Table 3), and (2) inaccurate water content data for the soil samples. The raw soil herbicide analytical data were reported by the laboratory on a wet weight basis [ug herbicide (kg wet soil)-1 ] along with laboratory determined soil water content (mass water/mass dry soil), the latter used along with soil bulk density to calculate total herbicide mass recovered in soil (eq.1). The laboratory reported soil moisture content of middles soil samples were uniformly low, with a mean of about 8.9%. This average water content is unrealistically low for soil sampled immediately after a water application, and accounts for at least a portion of the low herbicide total recoveries.

Comparison to previous studies. The runoff herbicide concentrations illustrated in figures 3 and 4 ranged from about 50 to 1500 ppb, similar in magnitude to simazine and diuron concentrations reported by Bruan and Hawkins (1991) in their study of actual wintertime runoff in and around citrus groves. Their highest concentration samples were collected about 2 1/2 months after herbicide application; the simulated rainfall events in this study were conducted within 36 hours of herbicide application. The Braun and Hawkins wintertime study reported much lower bromacil concentrations than observed here due to the fact that most citrus applications of bromacil occur in the spring, simazine and diuron being the most common fall-applied citrus herbicides.

Infiltration rates are a dominant factor influencing the runoff process. In a previous study conducted in this orchard (Troiano and Garretson, 1997), a similar water application resulted in a smaller fraction of applied water running off the plots (~38% vs 68% observed here). The only known differences in conditions or procedures between the two experiments is that a longer pre-irrigation was conducted in the current study, 30 - 45 minutes vs. about 15 - 20 minutes in the previous study (C. Garretson, personal communication). Because infiltration rates are inversely related to soil moisture content, it is possible that a higher antecedent soil water content contributed to the greater fraction of applied water as runoff in the current study. A consequence is that a larger portion of applied herbicide was recovered in runoff in the present study; the average fraction of simazine recovered in runoff here was about 10% of applied (Table 7), as compared to 4.3% reported by Troiano and Garretson (1997).

Mathematical description of the runoff process Chemical movement offsite during individual runoff events has been modeled as a function of time with some limited success (Ahuja and Lehman, 1983; Ahuja et al., 1981), however no time measurements were made during runoff in the current study. Alternately, the relationship between concentration and runoff volume was evaluated. An empirical first-order relationship was found between herbicide concentrations and the square root of runoff volume.


where V = measured cumulative runoff volume (L), C = measured concentration in runoff, Co is a parameter estimate of the initial herbicide runoff concentration at V = 0 (at the moment runoff just begins to leave the plot), and a concentration decline parameter k (L-1/2). The parameter k is analogous to a rate constant, describing the extent to which herbicide concentration in runoff decreases with increasing cumulative runoff volume. Large values of the decline constant k correspond to situations where rapid decreases in herbicide concentration are observed as the runoff process proceeds.

Figure 5 illustrates the observed 12 plot average data for each herbicide and the respective fits of the model based on nonlinear least squares minimization of eq. 2. Because the number of sampling intervals varied from plot to plot (i.e., total runoff volumes varied among the 12 plots), only sampling intervals for which at least half of the plots reported data were used to create the average runoff data set (see figure 5 for further discussion). A total of 138 of the 148 total runoff samples taken from the 12 plots are represented by the average data. The correlation coefficients for eq. 2 fit to the average data exceeded 0.99 for all 3 herbicides. Although eq. 2 was somewhat less succesful in describing the individual plot data (0.78 < r < 0.99), the majority of fits were reasonable as indicated by a mean correlation coefficient for all 36 (3 chemicals x 12 plots) runoff profiles of 0.93 (Table 8). In addition, no systematic deviations of the runoff profile data from eq. 2 were evident.

An additional comparison between simazine, diuron and bromacil runoff behavior was made using the nonlinear least-squares minimization best fit values for the "initial" runoff herbicide concentration C0 and the concentration decline constant k (L-1/2) from the 36 runoff profiles (Table 8). Plots of the observed data and individual fits are given in appendix H. The best fit C0 values for the 3 herbicides are not statistically different. However, the mean bromacil decline constants (k) were significantly different than those for simazine and diuron (P < 0.0001, Table 5), while the k values of the latter 2 herbicides were not significantly different at the P = 0.05 level. It is apparent that the more rapid decline in bromacil concentrations as the experiments progressed relative to simazine or diuron contributed to a smaller fraction of bromacil moving off-site in runoff previous discussed. However, the difference between herbicides may have only limited practical significance. Both observed concentrations and fraction of application moving off-site were substantial for all 3 herbicides. Under actual rainfall conditions in a given field, a herbicide's runoff tendency will be a function of both rainfall depth and duration. The water application rate of about 2.2 cm h-1 for 1.5 h simulated an intense rainfall event.

In spite of the relatively coarse surface soil in these plots (~ 75 percent sand), it is apparent that the very low infiltration rate is the dominant factor driving herbicide movement off-site in runoff. The low infiltration rate is probably related to compaction (surface bulk density ~ 1.7 g cm3, Table 2). Moreover, the cultural practices that lead to surface compaction and correspondingly low infiltration rates (complete weed control, no cultivation) are typical of citrus culture in Fresno and Tulare Counties. These observations further indicate that practically, in low permeability soils, herbicide substitution is not a satisfactory method for mitigating off-site movement.


1.The organosilicon surfactant applied at a nominal application rate of 5 oz acre-1 had no effect on the fraction of simazine, diuron, or bromacil recovered in runoff during a simulated rainfall event in a low permeability citrus orchard soil.

2. The percentage of herbicide recovered off-site was substantial under these conditions, ranging from about 4 to 21 percent of the initial herbicide application. The overall mean fraction of measured herbicide application recovered in runoff across all plots and all 3 herbicides was 10 percent. Runoff concentrations ranged from about 50 - 1500 µg L-1 in runoff water samples.

3. Bromacil had a lesser tendency to move off-site in runoff than simazine or diuron. This difference was attributable to bromacil's much greater solubility, providing a modicum of increased downward movement with simulated rainfall into the profile.

4. These results indicate that natural rainfall is a poor incorporation method in pan or compacted low permeability soils - regardless of the choice of herbicide. Substantial quantities of all 3 herbicides were recovered in the runoff water. Any approach to mitigating herbicide movement from low permeability soils in runoff should seek to either (i) significantly improve water infiltration rates into the soil, or (ii) mix the herbicide into the soil, rendering it unavailable to move with surface runoff, or (iii) apply sufficient water for complete incorporation at rates lower than the infiltration rate of the soil.

6. An empirical relationship was found between runoff concentration and cumulative runoff volumes under the conditions studied - low permeability soil exposed to short-term constant intensity simulated rainfall. Concentrations were found to decrease exponentially with the square root of runoff volume, the decline characterized by a depletion constant. This relationship may provide an additional evaluation tool in future studies.


Ahuja, L.R., A.N. Sharpley, M. Yamamoto, and R.G. Menzel. 1981. The Depth of Rainfall-Runoff-Soil Interaction as Determined by 32P. Water Res. Research 17:969-974.

Ahuja, L.R. and O.R. Lehman. 1983. The Extent and Nature of Rainfall-soil Interaction in the Release of Soluble Chemicals to Runoff. J. Environ. Qual. 12:34.

Ben-Hur, M., J. Faris,. M. Malik, and J. Letey. 1989. Polymers as soil conditioners under consecutive irrigations and rainfall. Soil Sci. Soc. Am. J. 53:1173.

Braun, A.L. and L.S. Hawkins. 1991. Presence of Bromacil, Diuron, and Simazine in Surface Water Runoff From Agricultural Fields and Non-Crop Sites in Tulare County, California. report PM91-1. Environmental Monitoring and Pest Management Branch, California Department of Pesticide Regulation.

Glantz, S.A., and B.J. Slinker. 1990. Primer of Applied Regression and Analysis of Variance. McGraw-Hill, New York.

Hiemenz, P.C. 1986. Principles of Colloid and Surface Chemistry, 2nd Ed. p. 338. Marcel Dekker, New York.

Letey, J., R.E. Pelshik and J. Osborne. 1961. Wetting Agents. Calif. Agric. October 1961, p. 8.

Powell, S, R. Neal, and J. Leyva. 1996. Runoff and Leaching of Simazine and Diuron used on Highway Rights-Of-Way. report EH96-03.Environmental Monitoring and Pest Management Branch, California Department of Pesticide Regulation.

Rao, P.S.C., A.G. Hornsby, and R.E. Jessup. 1985. Indices for Ranking the Potential for Pesticide Contamination of Groundwater. Proc. Crop and Soil Sci. Soc. Florida, 44:1

Roggenbuck, F.C., D. Penner, R.F. Burrow, and Bryan Thomas. 1993. Study of the enhancement of herbicide activity and rainfastness by an organosilicon adjuvant utilizing radiolabelled herbicide and adjuvant. Pest. Sci. 37:121.

Simmons, S.E. and J.J. Leyva. 1994. Presence of Soil-Applied Herbicides in Three Rights-Of-Way Basins in San Joaquin County, California. report EH 94-01. Environmental Monitoring and Pest Management Branch, California Department of Pesticide Regulation.

Spurlock, Frank C. and J.W. Biggar. 1994. Thermodynamics of organic chemical partition in soil: 2. Nonlinear partition of substituted phenylureas from aqueous solution. Environmental Science and Technology 28:996-1002.

Spurlock, Frank C. and J.W. Biggar. 1990. Effect of naturally occurring soluble organic matter on the adsorption and movement of simazine [2-chloro-4,6-bis(ethylamino)-s-triazine] in Hanford sandy loam. Environmental Science and Technology 24:736-741.

Tan, S., and M. Singh. 1996. Weed control efficacy of diuron and norflurazon as affected by adjuvants. submitted to Florida Agricultural Experimental Station Journal.

Taylor, S., and G. Ashcroft. 1972. Physical Edaphology: The physics of irrigated and nonirrigated soils. p. 342. W.H. Freeman and Co., San Francisco.

Troiano, J., B. Johnson, S. Powell, and S. Schoenig. 1994. Use of cluster and principal component analyses to profile areas in California where ground water has been contaminated by pesticides in California. Environmental Monitoring and Assessment, 32:269-288.

U.S. Department of Agriculture-Soil Conservation Service. 1971. Soil Survey: Eastern Fresno Area, California. USDA in cooperation with California Agric. Exp. St. Superintendent of Documents, U.S. Gov. Printing Office, Washington, D.C.

Zar, J.H. 1996. Biostatistical Analysis, 3rd ed. Prentice Hall, New Jersey.