Journal or Conference Title:
Journal of Environmental Science and Health, Part B (Marcel Dekker, Inc.)
Volume: 38
3
Issue:
Year:
2003
Pages: 257-273
Title of Article: Pesticide runoff model (PERM): An integrated modeling system for simulation of pesticide runoff losses from agricultural lands
PESTICIDE RUNOFF MODEL (PERM): AN INTEGRATED MODELING SYSTEM FOR SIMULATION OF PESTICIDE RUNOFF LOSSES FROM AGRICULTURAL LANDS
Yue-Ren Li1, Yi-Fan Li2 *, John Struger3, Bing Chen1, and Guo H. Huang1 1
Environmental Systems Engineering Program, Faculty of Engineering, University of Regina, Regina, Sask S4S 0A2, Canada 2
Meteorological Service of Canada, Environment Canada, Toronto, Ont., Canada
3
Environmental Conservation Branch, Ontario Region, Environment Canada, Burlington, Ont., Canada
* Corresponding author; e-mail:
[email protected]
Abbreviated Title:
Pesticide Runoff Model
2
ABSTRACT
An integrated model, the Pesticide Runoff Model (PeRM), has been developed to predict runoff losses of pesticides from agricultural lands. The model is an integration of a mathematical model, a relational database system, and a Geographic Information System. Information on soil types, land use, slope, watershed boundaries, precipitation, pesticide usage, as well as physical and chemical properties of pesticides have been digitized and gridded by using GIS, managed within a database, and made ready for simulation by the mathematical model. The developed modeling system can simulate pesticide losses due to runoff by considering the emission, degradation, adsorption and desorption of pesticides, as well as their movement in dissolved and adsorbed phases. Results from the model are in turn entered in the database, such that runoff patterns along with pesticide losses could be further simulated by using a powerful database management system. The final results were displayed and visualized through GIS. The developed modeling system has been used to calculate the losses of atrazine from agricultural lands in the Kintore Creek Watershed, Ontario, Canada between 1988 and 1992. The modeling outputs have been verified against actual monitoring data, which were obtained from a water quality monitoring project carried out in the same watershed over the same period of time. The results indicate that the model provides an effective means for forecasting pesticide runoff from agricultural lands.
Keywords: agriculture, environment, modeling, monitoring, nonpoint source pollution, pesticide runoff, watershed.
3
INTRODUCTION
Pesticides have been widely used in agriculture to control unwanted plants and destructive insects. There have been growing concerns about the fate and transport of pesticides in the past decades Due to their adverse impacts on the environment and human health (1, 2, 3, 4, 5, 6). Runoff losses of pesticides to surface water and their transport in streams are one of the principal processes leading to their widespread dispersion in the environment, especially to our drinking water (7, 8). To determine the extent of surface water pollution by pesticides, several pesticide monitoring programs throughout the North America were used to quantify pesticide runoff losses (9). Those studies have indicated that pesticide loss from runoff has ranged from 1% to 5% of the pesticide applied.
A joint project between Environment Canada and the University of Regina, Canada was started in 2000 to create an integrated modeling system for simulating pesticide runoff losses on a watershed scale. The Pesticide Runoff Model (PeRM) can integrate a mathematical pesticide runoff module with a Geographical Information System (GIS) and a related database management system into one dynamic system, so that the simulations are physically correct and geographically accurate. All the spatial data will be organized in GIS and converted into the database system that can automatically estimate model parameters and prepare input files for the mathematical module. The output data from the mathematical model are also stored in the database and displayed by using GIS. The model can account for the significance of spatial characteristics and the effects of such structures on runoff losses of pesticides. This model should be complete enough to integrate the mathematical model, GIS, and database management, yet simple enough to be used on a PC platform.
4
THE ARCHITECTURE OF PERM The architecture of PeRM is shown in Figure 1. It consists of three major parts: a mathematical module, the GIS, and a database.
These three parts are manipulated by a control panel.
Information on soil type, land use, slope, watershed boundaries, precipitation, and pesticide usage have been digitized and gridded by using GIS, and inputted along with physical and chemical properties of pesticides into the model. The input data are stored and managed within the database, and used by the mathematical module. The developed modeling system can simulate pesticide losses due to runoff by considering the emission, degradation, adsorption and desorption of pesticide, as well as their movement in dissolved and adsorbed phases. Results from the module are in turn entered in the database, such that runoff patterns along with pesticide losses could be further simulated by using a powerful database management system. The final results were displayed and visualized through GIS.
In order to use PeRM, a grid system must be created according to the watershed under study. An agricultural watershed is usually composed of fields of different crops, soil types, and slopes, leading to various ways for pesticide runoff to occur. A grid system is a discrete representation of a terrain, made up of spatially distributed parameters, with each of them being uniform with respect to crop species, soil type, hydrological conditions, and topographical characteristics. Thus, each grid cell can be treated as a homogenous unit for pesticide movement with inflows from and outflows to other grid cells.
5
Output from the grid cell at the top of the watershed is routed to cells below them and/or to the channel, and finally to the watershed outlet. Water from the cells not receiving pesticide application dilute the resulting concentrations and change the distribution of pesticide between the solid and solution phases.
The following assumptions were made in the development of the PeRM: 1) the sediment transport of a pesticide was excluded from the calculation since dissolved pesticide runoff losses generally exceed solid-phase losses greatly (10, 11); 2) a thin layer of surface soil (less than 1 cm) mixes completely and/or uniformly with rainwater in the process of pesticide transfer from soil to runoff (12, 13, 14); and 3) a unique slope direction and magnitude must be assigned within a single grid cell to avoid ambiguous flow directions.
THE MATHEMATICAL MODULE
Hydrological Component
A model developed by the US Soil Conservation Service (SCS) is used in our model to deal with the hydrological component. This SCS model is utilized to estimate runoff during rainfall events with a known precipitation volume. The volume of runoff (Q) depends on the volume of precipitation (P) and the initial abstraction (I). The initial abstraction (I) is a fraction of total rainfall that does not appear as runoff (15): (P − I ) 2 Q= (P − I ) + S
(1)
6
The relation between I and S is given by I = 0.2 S
(2)
where S=
1000 − 10 CN
(3)
The curve number (CN) expresses the combined effects of soil hydrologic characteristics, land use and antecedent moisture conditions on the runoff volume (15).
Pesticide Component
The amount of the pesticide lost to runoff during rainfall depends on the amount of the pesticide left on the farmland when rain starts. Residues of the pesticide are calculated by equations (16) Rt = ∑U i (1 − Fi )e − βt i
(4a)
i
β=
ln 2 t1 / 2
(4b)
where Rt is the residue of the pesticide in time t when the rainfall starts; Ui is the total amount of the pesticide during the ith application; Fi is the emission factor of the pesticide during the ith application; ti is the duration from the ith application of the pesticide to the time when the rain fall starts, and t1/2 is the half-life of the pesticide in the soil.
To calculate the solution phase pesticide concentration Cr in the runoff, the following equation is used (17):
7
Cr =
(5)
Cs K e 1 + Ke Kd
where Cs is is the runoff-available pesticide concentration in the surface soil layer when rainfall begins, Ke is the extraction coefficient, Kd is the soil/water partitioning coefficient that can be derived from the linear adsorption coefficient for organic carbon Koc. The relation between Kd and Koc is given for different contents of organic carbon (OC) as: K d = K oc
(6)
OC 100
Leonard et al. (17) developed a functional relationship to relate Ke to Kd:
K e = 0.5,
for K d ≤ 1.0
K e = 0.7 − 0.2 K d ,
for 1.0 < K d ≤ 3.0
K e = 0.1,
for K d > 3.0
(7)
Routing Component
The routing of pesticide runoff is carried out using a mass balance technique based on the continuity equation. This simple method is used here as a first approximation for the transport module to create a complete system. In general terms, the process is summarized as: 1) the amount of mass generated on each cell is calculated for each time step ∆t, which is the time needed for the all runoff in the cell to enter the following cell; and 2) all the amount of mass generated on each cell plus all the amount of mass entering the cell from other cells at time step ∆t will enter the following cell at the next ∆t.
The continuity equation for the pesticide runoff can be written as:
8
M out = M in + M generated
(8)
where M out is the amount of pesticide leaving the cell in runoff, M in is the sum of all the amount of pesticide entering the cell, and M generated the amount generated within the cell.
SYSTEM IMPLEMENTATION
The data required to run the model are summarized in Table 1. The spatial data include land use information, soil types and topography. These data sources consist of a number of polygons with attached attributes. In the case of soil type, the attribute of interest is soil texture classification. Physical-chemical pesticide properties are readily available from a number of reference works (18). Information about pesticide applications should be obtained from field surveys. Climate data are available from local meteorological stations.
All the input data need to be digitized and localized in each cell in the grid system by using MapInfo, a GIS software. The overlay of the obtained spatial data is realized through MapInfo and Map Basic, a programming language for MapInfo. The results will be stored in a database managed by an interface coded with Visual Basic. By using the input parameters the control panel can run the mathematical model, with the outputs being managed by the database and graphically presented by the MapInfo.
CASE STUDY
9
A surface water quality monitoring project was carried out from 1988 to 1992 in two subwatersheds near the town of Kintore, Ontario, Canada by Environment Canada and the Upper Thames River Conservation Authority (UTRCA) (19, 20, 21). The main purpose of the project was to collect environmental fate data on atrazine and metolachlor – two herbicides used to control broadleaf and grassy weeds in corn and soybeans. By using the data collected from the surface water quality monitoring project, the PeRM was used to calculate the losses of atrazine from agricultural lands in Kintore Creek Watershed, Ontario between 1988 and 1992 due to runoff. The modeling outputs are verified by using monitoring data, which was obtained from the same project.
Overview of Study Area
Figure 2 indicates the location of the Kintore Creek Watersheds, which is located northeast of London, Ontario, Canada. The two drains join southeast of Kintore to form Kintore Creek, which flows southward to join the Middle Branch of the Thames River. In the Kintore Creek Paired Watersheds, thirty landowners manage dairy, beef, swine and cash crop operations. The primary crop in both east and west subwatersheds is historically corn, but soybeans and grains are also produced in both watersheds. The detailed Kintore Creek sub-watershed characteristics are given in Table 2.
Data Acquisition
A. Spatial database
10
The spatial database for the Kintore Creek watershed was developed to support simulation modeling of pesticide runoff losses from agricultural lands. Figure 3a-c indicate the topography, soil, and landuse in the two Kintore watersheds. Within the watershed there are four major soil associations which directly influence the fate and transport of the pesticides. The first association represents soils with high infiltration rates, including Fox sandy loam, Huron clay loam, and Guelph loam. The second association includes soils with moderate infiltration rates, such as Embro silt loam and Tavistock silt loam. Soils of this association have moderate rates of water transmission, with moderately fine to moderately coarse textures. The third association represents soils with slow infiltration rates, such as Maplewood silt and Crombie silt loam, and the fourth association includes those with very slow infiltration rates (e.g. Muck soil). Agricultural practices occupy the majority of lands in the watershed, with soybeans and corn being the main crops.
B. Pesticide data
All landowners in the watershed were interviewed for information regarding pesticide application. Information collected included crop types, number of acres allocated for each crop, type of pesticide, number of acres applied with the pesticide, and rate and time of pesticide application. Table 3 shows parts of the pesticide-application information. Figure 4 shows the atrazine application pattern in 1990.
C. Climate data
11
The climate of the watershed is dry continental with an annual average precipitation of nearly 50 cm in watershed, and is unevenly distributed throughout the year. The precipitation data from the station located in London was used, since London is the nearest station where precipitation data are available. The Kintore Creek watershed is approximately 20 km away from London.
The daily precipitation data for London, Ontario in 1990 are given in Figure 5a and the hourly precipitation data for London, Ontario on August 28, 1990 are given in Figure 5b. Figure 6 shows the precipitation data for the Station of London, Ontario and the water discharge measured at the sampling site of the west sub-watershed started at 2:00 pm, August 28, 1990. In this figure, we assume that the storm event and the water discharge curves started at the same time. It turns out that similar storm events also happened in the Kintore watersheds as those happened at the Station of London. It seems, however, the last storm event did not occur in the Kintore watersheds, or it occurred, but was not strong enough to an effect on water discharge.
System Implementation
The procedures for model application are outlined here. First, a fine grid system for the Kintore Creek watersheds was created. The grid system is comprised of 165 computational elements (cells), each representing approximately 11.3 hectares (see Figure 7). Spatial data including soil type, land use, slopes, pesticide application patterns, etc. were localized on the grid system by using MapInfo. Results were stored in a database. The input data were organized in the database according to the required format for the mathematical simulation module. All the database processes were controlled by the control panel coded with Visual Basic. As an example, Figure 8 shows the flow path of the streams generated by GIS and database management.
12
RESULTS AND DISCUSSION
After running the routing process, the time series outputs in the two outlets of watersheds were obtained and compared with the monitoring data. Figure 9 depicts the results of the predicted and measured (a) water discharge and (b) atrazine runoff loss from the Kintore Creek watersheds for August 28, 1990. Both figures show that the predicted results from the Kintore Creek watersheds from PeRM match the monitoring data quite well, particularly following the first precipitation event at London. The differences between model simulations and observations are also obvious, which indicates that the precipitation history at Kintore is quite different from that at London Station (See Figure 6).
The results of the total rainfall events between 1988 and 1990 are plotted in Figure 10. Linear regressions were generated (at the 95% confidence level) between predicted values and observed samples for peak discharge and peak pesticide loading. Figure 10a is a plot of the predicted and observed peak discharge for the total rainfall events. A high correlation coefficient (r =0.979) is generated from this regression. The coefficient of determination (r2=0.96) indicates that approximately 96 % of the variation in the peak discharge model output can be explained by the input parameters and data. Figure 10b shows a plot of predicted and observed peak pesticide loading producing a correlation coefficient of r =0.923 and a coefficient r2 =0.85.
The model shows great potential for predicting pesticide runoff losses from agricultural lands. As the results indicate (see Figure 10), correlation coefficients of r =0.98 between observed and predicted peak runoff discharge and r =0.92 between observed and predicted peak pesticide
13
loading. The unexplained variance between the observed and predicted values may be attributed to a number of causes. These causes include errors in input data (application rates, time of pesticide application, and rainfall intensity, for example), or incorrect estimates of parameter values (the curve number, for example). Obviously, rainfall intensity is rarely constant and plays an important role in the amount of runoff generated. The rainfall intensity data from London Station selected in this study may not be an accurate reflaction of what occurred in the watershed. This creates the largest error for the results predicted by the model.
CONCLUSIONS
The successful utilization of PeRM, an integration of mathematical simulation module with GIS and a relational database management in this study demonstrates the potential ability of this model. Using automation techniques employed by GIS and a relational database can minimize the actual number of required parameters and can improve the accuracy of variable assignment when a large number of cells are involved. Also GIS and Database System automation allows the user to spatially average input parameters for each of the grid cells. This feature is highly beneficial as it allows for a more accurate representation of the natural watershed conditions.
The developed modeling system has been used to calculate the losses of atrazine from agricultural lands in Kintore Creek Watershed, Ontario. A water quality monitoring project was carried out from 1988 to 1992 in the watershed to detect conditions of surface water quality due to the use of pesticides in this area. The modeling outputs are verified by using monitoring data, and demonstrating reasonable prediction accuracy. The result indicated that the model provides an effective means for forecasting pesticide runoff from agricultural lands into surface waters.
14
ACKNOWLEDGEMENTS
The financial support from the University of Regina and Environment Canada is highly appreciated; Special thanks to D. Li from LDC Consulting Inc. for his constant help in programming of the database and GIS; Thanks also go to A. Deptuch-Stapf, B. Phinney, K. Puckett and S. Venkatesh of Environment Canada for their help and encouragement.
REFERENCES
1. Baker J. L.; Laflen, J. M.; Johnson, H. P. Effect of tillage systems on runoff losses of pesticides, a rainfall simulation study, Trans ASAE, 1978, 21, 886-892. 2.Wauchope R. D.; Leonard, R. A. Maximum pesticide concentrations in agricultural runoff: a semi-empirical prediction formula, J Environ. Qual., 1980, 9, 665-670. 3. Wu, T. L.; Correll, D. L.; Remenapp, H. E. H. Herbicide runoff from experimental watersheds, J Environ. Qual., 1983, 12, 330-336. 4. Leonard R. A. Movement of pesticides into surface waters. In Pesticides in the soil environment, SSSA Book Series no. 2. Madison: Soil Science Society of America, 1990, 303-348. 5. Agassi, M.; Letey, J.; Farmer, W. J; Clark, P. Soil Erosion Contribution to Pesticide Transport by Furrow Irrigation. J Environ Qual, 1995, 24 (5), 892-896. 6. Ng, H. Y. F.; Clegg, S. B. Atrazine and metolachlor losses in runoff events from an agricultural watershed: the importance of runoff components. Sci Total Enviro., 1997, 193, 215-228. 7. Leonard, R. A. Pesticides in Surface Waters. Environment Chemistry of Pesticides. CRC, 1988, 46-82. 8. Burgoa, B.; Wauchope, R. D. Environmental Behaviour of Agrochemicals. Edited by Roberts, T.R. and Kearney, P.C., John Wiley & Sons Ltd., 1995. p223-349. 9. Baker, D. B. Sediment, Nutrients and Pesticide Transport in Selected Lower Great Lakes Tributaries. USEPA-905/4-88-001. Great Lakes National Program Office, Chicago,IL, 1988.
10. Haith, D.A. A mathematical model for estimating pesticide losses in runoff. J.Environ. Qual., 1980. 7(1), 63-68. 11. Huber, A.; Bach, M.; Frede, H.G. Modeling pesticide losses with surface runoff in Germany. The sci. of Total Envion. 1998, 233, 177-191. 12. Donigian, A. S., Jr.; Beyerlein, D. C; Davis, H. H.; Crawford, N. H. Agricultural Runoff Management (ARM) model version II: refinement and testing. EPA 600/3-77098. Environmental Research Laboratory, U.S. EPA, Athens, GA. 1977. 13. Frere, M. H.; Ross, J. D.; Lane, L.J. The nutrient submodel. Chap.4. In Knisel, W.G. (ed.) CREAMS: A field scale model for chemicals, runoff, and erosion from agricultural management systems, U.S. Dept. Agr. Conservation Res. Rep, 26, 1980, 65-87. 14. Leonard, R.A.; Wauchope, R.D. The pesticide submodel. Chap.5, In Knisel, W.G. (ed.) CREAMS: A field scale model for chemicals, runoff, and erosion from agricultural management systems, U.S. Dept. Agr. Conservation Res. Rep, 26, 1980, 88-112. 15. McCuen R. H. A guide to hydrologic analysis using SCS methods. Engelwood Cliffs, Prentice Hall, 1981, 62. 16. Li, Y. F., Scholdz, M. T.; van Heyst, B. J. Global gridded emission inventory of αhexachlorocyclohexane, J. Geophys. Res., 2000, 105 (D5), 6621-6632. 17. Leonard, R. A.; Knisel, W. G.; Still, D. A. GLEAMS: Groundwater loading effects of agricultural management systems. Trans. ASAE, 1987, 30, 1403-1418.
18. Mackay, D.; Shiu, W. Y.; Ma, K. C. Illustrated handbook of physical-chemical properties and environmental fate for organic chemicals. Volume V., Lewis Publishers, New York, 1997, 697 pp. 19. Struger, J.; Fischer, J. D.; Wilcox, I.; Li, Y. R.; Huang, G. H; Li, Y. F. Monitoring Runoff Losses of Pesticides from Agricultural Wathersheds in the Kintore Creek, Ontario, Canada, 2002a, to be submitted. 20. Struger, J.; Fischer, J. D.; Wilcox, I.; Li, Y. R.; Huang, G. H; Li, Y. F. Continuous Monitoring of Agricultural Herbicides in Paired Subwatersheds: Conservation vs Conventional Tillage, 2002b, to be submitted. 21. Struger, J.; Fischer, J. D.; Li, Y. R. An Automatic Sampling Method to Continuously Monitor Atrazine and Metolachlor in Streams, 2002c, to be submitted.
Table 1. Parameters to be inputted to the PeRM. Data type
Description
Extent
Land use
Crop type
Spatial data
Crop area
Spatial data
Soil type
Spatial data
Soil hydrological characteristics
Spatial data
Soil bulk density
Spatial data
Soil organic matter
Spatial data
Pesticide type
Spatial data
Pesticide application area
Spatial data
Pesticide application rate
Spatial data
Pesticide sorption coefficient
Single data
Pesticide half life
Single data
Pesticide application time
Spatial data
Soil moisture condition
Spatial data
Rainfall time
Spatial data
Rainfall intensity
Spatial data
Rainfall duration
Spatial data
Watershed area
Spatial data
Watershed slope
Spatial data
Watershed inter boundary
Spatial data
Soils
Pesticides
Climate
Watershed
Table 2. Kintore Creek sub-watersheds characteristics (19, 20). West Kintore Creek
East Kintore Creek
Size of sub watershed
6.61 km2
6.42 km2
Soil types
silt loam
silt loam, sandy loam & muck
Soil erosion potential
medium to high
medium to high
Area under study
653 ha
635 ha
# of major landowners
15
13
Area tile drained
55
30
Total forest cover
78 ha
175 ha
Total crop area
473 ha
333 ha
Table 3. Atrazine application rates for Kintore Creek sub-watersheds in 1990 (19). West Watershed
East Watershed
Amount applied
213 kg
218 kg
Area treated
184 ha
153 ha
Average application rate
1.16 kg/ha
1.42 kg/ha
List of Figure Captions Figure 1. Architecture of the integrated modeling system. Figure 2. Location of the Kintore Creek Watersheds. Figure 3. Map of the Kintore Creek watershed (a) slope, (b) soil type, and (c) land use (18). Figure 4. Pesticide application pattern in 1990. Figure 5. Rainfall in 1990 for the Station of London, Ontario, Canada. (a) Daily rainfall in 1990 (35.2 mm on August 28, 1990); and (b) hourly rainfall on August 28, 1990, which indicates that there were 5 rain events on that day, and hourly rain falls for each event were 2.6, 4.8, 25.8, 0.3, and 4.3 mm, respectively. Figure 6. Precipitation data for the Station of London, Ontario and water discharge measured data from the west sub-watershed on August 28, 1990. Figure 7. Grid system for the Kintore Creek watersheds. Figure 8. Flow path of the Kintore Creek watersheds. The two circles in red indicate the outlet of the each watershed, or sampling sites. Figure 9. The Predicted and measured (a) flow discharge and (b) pesticide load from August 28, 1990. Figure 10. The predicted and observed (a) peak discharge and (b) peak pesticide load.
Figure 1. Architecture of the integrated modeling system.
West watershed
East watershed
Sampling sites Kintore
Figure 2. Location of the Kintore Creek Watersheds.
(a)
(b)
(c) Figure 3. Map of the Kintore Creek watershed (a) slope, (b) soil type, and (c) land use (20).
Figure 4. Pesticide application pattern in 1990.
600
Daily rainfall (0.1mm)
500
Aug. 28, 1990
400 300 200 100 0 120
140
160
180
200
220
240
260
280
300
320
Julian day
(a)
Precipitation (0.1mm
300.0 250.0 200.0 150.0 100.0 50.0 0.0 1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time
(b) Figure 5. Rainfall in 1990 for the Station of London, Ontario, Canada. (a) Daily rainfall in 1990 (35.2 mm on August 28, 1990); and (b) hourly rainfall on August 28, 1990, which indicates that there were 5 rain events on that day, and hourly rain falls for each event were 2.6, 4.8, 25.8, 0.3, and 4.3 mm, respectively.
1.6 Precipitation
Precipitation (0.1mm)
250
Discharge
200
1.4 1.2 1 0.8
150
0.6
100
0.4 50
Discharge (m^3/s)
300
0.2
0
0 0
60
120 180 240 300 360 420 480 540 600 660 Time (minutes)
Figure 6. Precipitation data for the Station of London, Ontario and water discharge measured data from the west sub-watershed on August 28, 1990.
Figure 7. Grid system for the Kintore Creek watersheds.
Figure 8. Flow path of the Kintore Creek watersheds. The two circles in red indicate the outlet of the each watershed, or sampling sites.
Water discharge (m^3/s)
1.6
Simulation
1.4
Monitoring
1.2 1 0.8 0.6 0.4 0.2 0 0
100
200
300
400
500
600
700
800
Time (minites)
(a) 6000
Pesticide loading (ug/s)
Simulation 5000
Monitoring
4000 3000 2000 1000 0 0
100
200
300
400
500
600
700
800
Time (minites)
(b)
Figure 9. The Predicted and measured (a) flow discharge and (b) pesticide load from August 28, 1990.
(a)
(b) Figure 10. The predicted and observed (a) peak discharge and (b) peak pesticide load.
