Agrochemical Fate and Movement - American Chemical Society

Chapter 11

Runoff Losses of Suspended Sediment and Herbicides: Comparison of Results from 0·2- and 4-ha Plots 1

2

3

1

1

Downloaded by NORTH CAROLINA STATE UNIV on September 20, 2012 | http://pubs.acs.org Publication Date: April 3, 2000 | doi: 10.1021/bk-2000-0751.ch011

L. M. Southwick , D. W. Meek , R. L. Bengtson , J. L. Fouss , and G. H. Willis 1

Soil and Water Research Unit, Agricultural Research Service, U.S. Department of Agriculture, 4115 Gourrier Avenue, Baton Rouge, LA 70808 National Soil Tilth Laboratory, Agricultural Research Service, U.S. Department of Agriculture, 2150 Pammel Drive, Ames, IA 50011 3Department of Biological and Agricultural Engineering, Louisiana State University Agricultural Center, Baton Rouge, LA 70803

2

Over the last dozen years we have conductedfieldstudies of runofflosses of herbicides and suspended sediment. In this paper we investigate the possible influence of plot size on our results by comparing data from plots of 0.21 and 4.4 ha. At the areal extent of our studies, this assessment fails to reveal a consistent plot size influence on runofflosses of atrazine and metolachlor, chemicals of moderate to high water solubility. The data may indicate more efficient extraction from soil into runoff of these herbicides in the larger plots. The runoff results for trifluralin and pendimethalin, of low water solubility, may indicate a plot size influence but also can be explained by a year effect related to application differences. Sediment yield in runoff has been sensitive to the plot size differences in our work;fromthe smaller plots we have observed three times larger sediment yields.

In investigations and descriptions of the contribution of runoff from agricultural fields to nonpoint source pollution problems, the effect of study area size on the applicability of results at the watershed and basin scale becomes an important concern. Spatial variability of rainfall and other weather factors, soil characteristics, topography, geology, vegetation, and drainage patterns lead to uncertainties when applying results fromfieldand small plot studies to larger geographical areas (1, 2) In studies of pesticide concentration patterns in the Lake Erie Basin starting in 1981, Baker and coworkers (3) have made the following scale effect observations: peak observed concentrations increase as watershed size decreases; and the average length of time during which intermediate pesticide concentrations are continuously exceeded tends to increase with watershed size. Richards and Baker state (3, p. 21) that scale effects "can be expected to be present all the way down to the plot scale."

© 2000 American Chemical Society

In Agrochemical Fate and Movement; Steinheimer, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2000.

159


160 An additional scale effect that has been reported is that sediment yield (sediment/unit area) decreases as watershed size increases (4, 5). Since 1985 we have been conductingfieldstudies of runoff losses of various herbicides and insecticides, nitrate, and suspended sediment. Our early work was performed on 2- to 4-ha plots (6); since 1994 we have carried out investigations on 0.2-ha study areas (7). The earlier larger plot work was done on Commerce clay loam (fine-silty, mixed, nonacid, thermic Aerie Fluvaquents), a Mississippi River alluvial soil graded to 0.1% slope; the later smaller plot work was performed 0.2-0.5 km away on Commerce silt loam (fine-silty, mixed, nonacid, thermic, Aerie Fluvaquents) graded to 0.2 % slope. Corn was the crop in both cases. In this paper we compare/contrast runoff results of herbicides (application rates are listed in Table I) and suspended sediment in order to test for trends that might be assignable to the differences in plot sizes in ourfieldwork. We look at results for atrazine and metolachlor in runofffrom4.4-ha plots in 1987 (6) andfrom0.21-ha plots in 1995 (8) and 1996. In addition, we compare/contrast runoff of trifluralinfromthe 4.4-ha plots in 1992 (9) with results for pendimethalinfromthe 0.21-ha plots in 1996. We also discuss the suspended sediment losses in runoff observed in these studies (8,70), Table 1. Herbicide Application Rates Herbicide

Year

Rate*

Atrazine

1987 1995 1996 1987 1995 1996 1992 1996

1.63 0.80* 1.49 2.16 1.00* 1.91* 1.12* 0.92*

Metolachlor Trifluralin Pendimethalin

t

f

f

* kg/ha; * unincorporated; * incorporated Table II. Herbicide Properties* Herbicide

s

K

33 530 0.3 0.3

100 200 5000 8000

w

Atrazine Metolachlor Pendimethalin Trifluralin

oc

7

2.9 χ 10" 3.1 xlO" 9.4 χ 10" 1.1 χ 10"

5

6

4

tin

Ref.

35 23 55 31

15, 16 15, 16 8, 15 9, 15

* S , water solubility, mg/L; K , soil organic carbon sorption coefficient, mL/g; V , vapor pressure, mm Hg; t , soil half life, days. w

oc

p

1/2


161


Although this work has been conducted on plots with and without subsurface drains, we consider in this paper results onlyfromplots without subsurface drains. An additional interest in our comparison is an effort to assess the size of an "elemental area" (11) in thesefieldstudies. We and others have reported work on smaller field plots (12, 0.0013 ha; 13,14, 0.00070 ha). Table II lists relevant properties of the chemicals treated in this paper.

Concentrations of Atrazine and Metolachlor in Runoff Disappearance of pesticidesfromsoil normally shows an exponential decrease with time and often approximates tofirstorder kinetics, even though the mechanism of disappearance is a complicated mixture of physical and chemical processes such as irreversible sorption to the soil, volatilization, leaching, runoff, and chemical/microbiological decompositon (17,18). Similarly, concentrations of pesticides in runoff also diminish exponentially with time, reflecting the corresponding decrease in concentration in the runoff-active zone of the soil (19). Atrazine. Ourfieldstudies with atrazine have routinely shown that the concentration of this herbicide in runoff quickly drops with increase in elapsed time after application (Figure 1). This disappearance of atrazine in runoff closelyfitsmodifiedfirstorder

1987,4.4 ha 1995, 0.21 ha 1996,0.21 ha 1987: C=1+923exp(-0.111t) 1995: O1+206exp(-0.111t) 1996: O1+397exp(-0.193t)

0

10

20

30

40

50

Days after Application Figure 1. Concentration of atrazine in runoff. In Agrochemical Fate and Movement; Steinheimer, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2000.

60

162 decay curves (Table III). [All reported regression results are based on the means for treatment or day. The generalized least squares (GLS) method (20) was used for the regressions of Table ΠΙ. This table reports results for one model for all years, starting with all parameters different for each year and simplifying based on testing corresponding parameters for equality. The resultant weight models were generally consistent with a Poisson or lognormal error structure.] These equations predict that at t = 0, the initial concentration is (a + b) ^g/L; k is thefirstorder rate constant; and as t increases, these equations predict that C approaches (a) Mg/L. The equations allow calculation of DT s (50% disappearance times) for the chemical in runoff by solving for t when C = 0.5(C ). Runoff DT values in Table III correspond to soil DT5 s of 35 (16, 1987), 18 (β, 1995), and 15 (27,1996) days. In ourfieldwork, we routinely observe that DT (runoff cone.) < DT (soil cone). This observation is reasonably due to rapid leaching of the runoff-available residue to just below the runoff extraction zone (22-24) but not as quickly below the zone removed in soil sampling procedures (in our case, usually the top 2.5 cm soil layer). The analyses reported in Table III yield identical rate constants k for 1987 and 1995; consequently, the DT s are the same for these two study seasons. The plot size differences between these two studies did not influence the observed persistence of atrazine in runoff


ro

50

ro

ro t = 0

50

0

50

50

50

Table III. Atrazine Concentration in Runoff Model: C = a + bexp(-kt) Parameters, (T Values), and [95% Confidence Intervals] ro

Year

a

b

k

R

2

DT , days* 50

1987

4.4 ha 1.00(4.41) 923 (55.1) 0.111 (22.6) 0.986 [0.50, 1.50] [705, 1210] [0.101, 0.122]

6.23 [5.68, 6.89]

1995

0.21 ha 1.00(4.41) 206 (50.7) 0.111(22.6) 0.986 [0.50, 1.50] [164, 259] [0.101,0.122]

6.23 [5.68, 6.89]

1.00(4.41) 397 (77.5) 0.193 (23.8) 0.986 [0.50, 1.50] [336, 470] [0.175, 0.211]

3.59 [3.29, 3.95]

1996

* C , cone, £*g/L; t, days after application. * 50% Disappearance time. ro

Metolachlor. As with atrazine, ourfieldwork has also shown rapid decreases of metolachlor runoff concentration with time after application (Figure 2). The


163 500 • Ο •

400

ι Ο 300 c

ο

1987,4.4 ha 1995,0.21 ha 1996,0.21 ha 1987:O1944exp(-0.136t) 1995:C=14.7+1944exp(-0.356t) 1996:C=14.7+232exp(-0.136t)


ο

%

200

\

\

c

100

ο

\

\

î

2

IT

0

10

20

30

40

50

Days after Application Figure 2. Concentration of metolachlor in runoff.

disappearance curves of Figure 2fitmodifiedfirstorder equations (Table IV) that provide DT s that are shorter than the respective soil DT values. [The results of Table IV were obtained by the statistical procedures of Table III] The corresponding metolachlor soil DT s were 20 (16,1987), 29 (8,1995), and 17 (21,1996) days. As with the results for atrazine in Table III, the metolachlor results of Table IV show a similarity in rate constant k (and therefore in persistence in runoff) across plot size (1987 and 1996). so

50

50

Concentrations of Atrazine and Metolachlor in Runoff as Functions of Soil Concentrations 12

2

Leonard et al. (19) developed a power equation, Y = 0.05X , R = 0.86, to describe the relation between runoff concentrations (Y) of various herbicides transported in the water phase and their respective soil surface concentrations (X). These investigators


164 Table IV. Metolachlor Concentration in Runoff* Model: C = a + bexp(-kt) Parameters, (T Values), and [95% Confidence Intervals]


ro

Year

a

1987

0

1995 1996

b

k

R*

f

DT , days 50

1944(84.2) 0.136(22.3) 0.988 [1598, 2364] [0.122, 0.149]

5.12 [4.66, 5.67]

0.21 ha 14.7(5.74) 1944(84.2) 0.356(17.2) 0.988 [9.14,20.32] [1598,2364] [0.311,0.400]

1.95 [1.73,2.23]

14.7(5.74) [9.14, 20.32]

5.12 [4.66, 5.67]

232(94.4) 0.136(22.3) 0.988 [205, 263] [0.122, 0.149]

*C , cone, Mg/L; t, days after application. *50% Disappearance time. ro

Table V, Atrazine and Metolachlor Concentration in Soil and Runoff* Model:C = a + bC Parameters, (T Values), and [95% Confidence Intervals] 3

ro

Year

s

2

h?

â

b

R

1987

4.4

0

Atrazine 0.0626(8.69) [0.045, 0.080]

0.986

1995

0.21

0

0.0874(8.15) [0.0532, 0.122]

0.985

1996

0.21

0

0.0369(17.1) [0.0317, 0.0422]

0.990

1987

Metolachlor 4.4 -0.0326(5.34) 0.00650(18.7) [-0.0521, -0.0132][0.0054, 0.0076]

0.987

1995, 0.21 0.0159(2.83) 0.00260(19.6) 1996 [0.0034, 0.0284] [0.00230, 0.00289]

0.990

f

* C , runoff cone, Mg/L; C , soil cone, Mg/kg. Plot size, ha. ro

8



165

1

2

Atrazine Soil Cone, ppm Figure 3. Concentration of atrazine in soil and runoff. considered the coefficient to be an "extraction coefficient" and suggested that greater distancefromlinearity shown by the exponent to reflect lower extraction efficiency with increasing time after application. That is, runoff extraction was more efficient early after application when soil concentrations were at their highest. If aged soil residues become more tightly bound or degraded, extraction becomes less efficient and hysteresis is observed (25, pp. 65-66; 26). From ourfielddata sets we have related the runoff concentrations of atrazine (Figure 3) and metolachlor (Figure 4) to their respective soil surface concentrations and have developed power regression equations (Table V) to describe these relationships. The analyses of Table V reveal a trend toward statistically significant differences (the 1995 atrazine data are not consistent here) between the extraction coefficients b with respect to plot size and/or year. [Table V results were based on linear measurement error regression (27).] For both herbicides the extraction coefficient is lower with the smaller plot size. For atrazine, b = 0.59b (1996 results compared to 1987); for metolachlor, b = 0.40b (1995 and 1996 data compared to 1987). If these trends are due to the plot size differences, the longer runs of the larger plots before the samples passed through the sampling flumes may have led to increased extraction efficiencyfromthe runoff-active zone of the soil. On the larger plots there is greater time for the herbicides to desorbfromthe soil. Higher runoff concentrations of atrazine and metolachlor from the larger plots as a function of time after application are indicated in Figures 1 and 2. Additionally, the extraction coefficients are consistently higher for atrazine than for metolachlor: the 10-fold 02l

021

44

44



166

0

1

2

3

4

Metolachlor Soil Cone, ppm Figure 4. Concentration of metolachlor in soil and runoff.


5

167 differences (within same year, atrazine > metolachlor) seen in Table V are consistent with the two-fold differences in K s (Table II, metolachlor > atrazine). w


Trifluralin and Pendimethalin Yield in Runoff In 1992, we investigated runoff of the dinitroaniline herbicide, trifluralin. Our 1996 work included runoff of another dinitroaniline, pendimethalin. Since both water solubilities (0.3 ppm) and K^s (8000 mL/g, trifluralin; 5000 mL/g, pendimethalin) are the same or similar for the two compounds (Table II), we have compared the yield in runoff as a function of flow and have developed regression equations for these observed relationships (Table VI). [The statistical treatment for Table VI was similar to that for Table V.] Yield is expressed as percent of application since the dinitroaniline application rates for the two years were slightly different (Table I). Table VI. Yields of Trifluralin (1992) and Pendimethalin (1996) in Runoff* Model: Y = a + bV Parameters, (T Values), and [95% Confidence Intervals] 1/2

R

2

Year

ha

1992

4.4

0.00412 (15.7) -0.0104(7.12) [-0.0150, -0.0058] [0.00329, 0.00495]

0.987

1996

0.21

-0.0140 (2.97) [-0.141,-0.0139]

0.00765 (7.72) [0.00522, 0.0101]

0.826

a

b

* Y, herbicide yield, % of appl.; V, runoffflow,mm. Theflowcoefficient b (Table VI) is statistically higher for the 1996 (0.21 ha) data than for the 1992 (4.4 ha) results (b i 1.8*4,4iJhigher coefficient from the smaller plot size is consistent with the application difference between the two years-incorporated in 1992 but not in 1996 (Table I). Trifluralin and pendimethalin show water solubilities, and K^s (Table II) that encourage association with sediment in runoff (22, Fig. 9-8; 24); the greater coefficient b from the smaller plots is also consistent with the larger sediment yieldsfromthese plots (see below). =

T n i s

02 h a

Sediment Yield in Runoff From our 4.4 ha plots we have nine years of monthly sediment yields in runoff. For a 9-year period, 1988-1996, we have calculated means and standard errors of the means (7, yields normalized with respect toflow,kg/ha/mm) for each of the months April, May, June (Figure 5). We have also calculated the respective means and standard errors of sediment yieldsfromthe 0.21 ha plots for the 1995 season (mostly in May) and the 1996 season (mostly in April). Comparative box plots (Figure 5) show the differences in the sediment yield/runoff*flowdata via a graphical method of exploratory data analysis (28). Figure 5 is based on Friendly's adaptation (29, pp. 285-317) of McGill et al. (30). In Agrochemical Fate and Movement; Steinheimer, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2000.

168


*

4. 40

0.21

Plot

Size,

ho

Figure 5. Comparative variable width notched box plots of the sediment yield/runoff for the two plot sizes. Each box represents the middle 50% of the data. The width of each box is proportional to the square root of the number of observations. The asterisks are the respective arithmetic means, the lines within the notches are the medians, and the notches are approximate 95% confidence intervals for the medians. The upper and lower whiskers represent the range of the data for each plot size distribution with the maximum extent not exceeding the median value ±1.5 interquartile range (the length of the box). Points that exceed that limit, the diamonds above the top whiskers, can be considered outliers.

To formally test for a plot size difference in the Y data, an ANOVA was performed. In order to match the method of collection for the larger plot data, the smaller plot yield and runoff data were totaled 30-day periods in just 1995 and 1996 because these were the only years with matching months during the period of record. That is, the appropriate data of 1995 and 1996fromboth studies were matched. A traditional completely randomized model was used; the treatment structure was oneway with just two levels, the plot size. The matched values were samples in time resulting in eight data points, four for each plot size. The plot size means were 13.0 kg/ha/mm for the 4.4 ha plots and 41.3 kg/ha/mm for the 0.21 ha plots. The two means are significantly different (P^O.016) and give the ratio of mean values: ^0.21 ha ^ 4 . 4 ha

3.2.


169


Summary and Conclusion In this assessment of ourfielddata, with respect to plot size (0.21 and 4.4 ha), of runoff losses of herbicides (both of moderate and low water solubility) we have observed that this behavior shows trends variably assignable to plot size differences. From both plot areas of our studies, the water soluble compounds atrazine and metolachlor demonstrated rapid concentration declines with increasing time between application and the runoff event. When these runoff concentration data are regressed with time after application (Tables III and IV), the resulting modifiedfirstorder equations produce the same DT s across plot size. Regression of atrazine and metolachlor runoff concentrations against soil concentrations (Table V) revealed more effective desorptionfromthe soil in the larger plots. The herbicides of low water solubility (trifluralin and pendimethalin) also show trends that are consistent with a plot size influence. The probability of this effect is enhanced since the observed trend (Table VI) is in the same direction as that of sediment. But the issue is clouded by the fact that application differences (year effect) are also in line with the trends of Table VI. Sediment yield in runoff shows a definite trend attributable to plot size. In the analysis of these sediment runoff data, the three 4.4-ha data sets, each extending over a 9-year period, demonstrate a consistently lower runoff sediment yield compared to the two 0.21-ha studies (Figure 5). This difference stands when the 1995 0.21-ha data (collected mostly in May) are compared with the May, 1995 4.4-ha data and when the 1996 smaller plot data (mostlyfromApril) are compared with the larger plot data of April, 1996 (ANOVA). An elemental agricultural area (11, pp. 176-181) will show the following characteristics; uniform soil type, constant vegetation cover, constant overall slope, and uniform distribution of precipitation. In our studies, these conditions are generally met at both plot size levels. The lack of a consistent trend attributable to plot size in our runoff data for atrazine and metolachlor points to a general similarity of the above plot characteristics across the areas of our studies, as these characteristics affect runoff of these compounds of high water solubility. The data indicate that from soil of similar atrazine and metolachlor concentrations, larger concentrations result in a runoff event from the larger plot size. A general breakdown in elemental area across the plot sizes in the work described in this paper shows up in our sediment data and is suggested by the results for the herbicides of low water solubility. so

Acknowledgment The authors are grateful for the statistical assistance of Deborah L. Boykin, USDA, ARS, Stoneville, MS. Literature Cited 1.

Bailey, G. W.; Swank, R. R., Jr. In Agricultural Management and Water Quality; Schaller, F. W.; Bailey, G. W., Eds.; Iowa State University Press: Ames, Iowa, 1983, 27-47.


2.

3. 4.


5. 6. 7.

8.

9. 10.

11.

12. 13. 14. 15. 16. 17. 18.

19. 20.

Smith, C. N.; Brown, D. S.; Dean, J. D.; Parrish, R. S.; Carsel, R. F.; Donigian, A. S. Jr. Field Agricultural Runoff Monitoring (FARM) Manual (EPA/600/3 85/043); U. S. Environmental Protection Agency: Athens, Georgia, 1985, 1. Richards, R. P.; Baker, D. B. Environ.Toxicol.Chem. 1993, 12, 13-26. Johnson, H. P.; Moldenhauer, W. C. In Agricultural Practices and Water Quality; Willrich, T. L.; Smith, G. E., Eds; Iowa State University Press: Ames, Iowa, 1970, 3-20. Gottschalk, L. C. In Handbook of Applied Hydrology; Chow, V. T., Ed; McGraw-Hill: New York, New York, 1964, 17-1 to 17-34. Southwick, L. M.; Willis, G. H.; Bengtson, R. L.; Lormand, T. J. Bull. Environ. Contam. Toxicol 1990, 45, 113-119. Willis, G. H.; Fouss, J. L.; Rogers, C. E.; Southwick, L. M . In Groundwater Residue Sampling Design; Nash, R. G.; Leslie, A. R., Eds; ACS Symposium Series 465, American Chemical Society: Washington, DC, 1991, 195-211. Southwick, L. M.; Willis, G. H.; Fouss, J. L.; Rogers, J. S.; Carter, C. E. In Proceedings of the Twenty-seventh Mississippi Water Resources Confe Daniel, B. J., Ed; Water Resources Research Inst.: Mississippi State, Mississippi, 1997, 239-246. Southwick, L. M.; Willis, G. H.; Mercado, Ο. Α.; Bengtson, R. L. Arch. Environ. Contam. Toxicol 1997, 32, 106-109. Bengtson, R. L.; Carter, C. E.; Morris, H. F. In Drainage in the 21st Century: Food Production and the Environment; Brown, L. C. Ed; American Society of Agricultural Engineers: St. Joseph, Michigan, 1998, 530-537. Huggins, L. F.; Burney, J. R. In Hydrologic Modeling of Small Watersheds; Haan, C. T.; Johnson, H. P.; Brakensiek, D. L., Eds; American Society of Agricultural Engineers, St. Joseph, Michigan, 1982, 169-225. Southwick, L. M.; Willis, G. H.; Reagan, T. Ε.; Rodriguez, L. M . Environ. Entomol. 1995, 24, 1013-1017. Baldwin, F. L.; Santelmann, P. W.; Davidson, J. M . Weed Science. 1975, 23, 285-288. Baldwin, F. L.; Santelmann, P. W.; Davidson, J. M . J. Environ. Qual. 1975, 4, 191-194. Hornsby, A. G.; Wauchope, R. D.; Herner, A. E. Pesticide Properties in the Environment. Springer-Verlag, Inc.: New York, New York, 1996. Southwick, L. M.; Willis, G. H.; Bengtson, R. L.; Lormand, T. J. J. Irr. Drain. Engin. 1990, 116, 16-23. Guenzi, W. D., Ed. Pesticides in Soil and Water. Soil Science Society of America, Inc.: Madison, Wisconsin, 1974. Sawhney, B. L.; Brown, Κ., Eds. Reactions and Movement of Organic Chemicals in Soils. Soil Science Society of America, Inc.: Madison, Wisconsin, 1989. Leonard, R. Α.; Langdale, G. W.; Fleming, W. G. J. Environ. Qual. 1979, 8, 223-229. Carroll, R. J., Ruppert, D. Transformations and Weighting in Regression:


171

21.


22.

23.

24. 25.

26. 27. 28.

29. 30.

Monographs on Statistics and Applied Probability. Chapman and Hall: New York, New York, 1988. Southwick, L. M.; Willis, G. H.; Fouss, J. L. In Drainage in the 21st Century: Food Production and the Environment. Brown, L. C., Ed; American Society of Agricultural Engineers: St. Joseph, Michigan, 1998, 668-675. Leonard, R. A. In Pesticides in the Soil Environment: Processes, Impacts, and Modeling. Cheng, H. H., Ed; Soil Science Society of America, Inc.: Madison, Wisconsin, 1990, 303-349. Leonard, R. Α.; Wauchope, R. D. In CREAMS: A Field-Scale Modelfor Chemicals, Runoff, and ErosionfromAgricultural Management Systems. Knisel, W. G., Ed; U. S. Dept. of Agriculture: Washington, DC, 1980, 88-112. Wauchope, R. D. J. Environ. Qual. 1978, 7, 459-472. Koskinen, W. C.; Harper, S. S. In Pesticides in the Soil Environment: Processes, Impacts, and Modeling. Cheng, Η. H., Ed; Soil Science Society of America, Inc.: Madison, Wisconsin, 1990, 51-77. Ma, L., Southwick, L. M., Willis, G. H., Selim, H. M. Weed Sci. 1993, 41, 627-633. Fuller, W. A. Measurement Error Models. John Wiley and Sons, New York, New York, 1987. Hoagland, D.C.;Mosteller, F.; Tukey, J. W. (Eds.). Understanding Robust and Exploratory Data Analysis. New York, New York: John Wiley and Sons, Inc., 1983. Friendly, M. SAS System for Statistical Graphics. SAS Institute: Cary, North Carolina, 1996. McGill, R.; Tukey, J. W.; Larsen, W. Amer. Stat. 1978. 32, 12-16.


Agrochemical Fate and Movement - American Chemical Society

Recommend Documents