MODELING THE EFFECTS OF VEGETATION ON HEAVY METALS CONTAINMENT

R. Green¹ , L.E. Erickson¹ , R. Govindaraju², P. Kalita³, and G. Pierzynski⁴

¹Department of Chemical Engineering , ²Department of Civil Engineering, ³Department of Biological and Agricultural Engineering, and ⁴Department of Agronomy, Kansas State University, Manhattan, KS 66506, Phone: 913-532-5584, Fax: 913-532-7372

Abstract

Soil and water contamination of lead, cadmium, and zinc are of concern in Southeast Kansas, where mining activities occurred until the middle of this century. Sediment erosion from the remnant piles of chat, an aftermath of the mining activity, is responsible for the increasing metal-contaminant concentrations in nearby farmland. Vegetation is being examined as a means of controlling the further spread of the metals. One current program used in watershed modeling, the Agricultural Non-Point Source Pollution Model, or AGNPS, is examined to determine the role that vegetation plays in controlling metal contamination from an 800-acre watershed, containing such chat piles, near Galena, Kansas.

Keywords: vegetation, metals, contamination, sediment, modeling, erosion, watershed

Introduction:

The Environmental Protection Agency, as well as state and local agencies and organizations throughout the United States, has made tremendous progress over the last twenty-five years in the reduction of water pollution due to point sources. Now, the contribution of non-point source pollutants is being addressed. Nonpoint source pollution occurs when runoff, from excess precipitation, extends the presence of a contaminant beyond its previous boundaries; the seriousness of this pollution was addressed in section 319 of the Clean Water Act, with Congress instructing that the best means be found to solve these problems (U.S. EPA, 1985; Novotny and Olem, 1994). Many researchers sought to create or employ simulation models to predict transport of contaminants from field to waterway and its subsequent transport downstream (Hamlett, et al., 1984). Simulation models can help in determining the best scenario or most effective means of controlling the contaminant.

In Southeast Kansas, where soil and water contamination of lead, cadmium, and zinc are of serious concerns, vegetation is being examined as a means of controlling the spread of soil containing elevated levels of metals. With mining ending in the middle of the twentieth century, many regions in Southeast Kansas are now experiencing elevated concentrations of heavy metals in the soil and surface water. Farmlands, located downstream of the exposed chat piles, are reportedly slowly being poisoned with increasing levels of these contaminants that are washed down with eroding sediment during periods of prolonged or intense rainfall. These sediments are often deposited in the flood plain (Pierzynski and Schwab, 1991). In this study, an 800-acre watershed was selected to examine the effects of best management practices in controlling sediment loss through simulation modeling. Since the majority of metal contaminant spread occurs from adsorption onto moving sediment, minimizing the loss of soil would reduce the amount of metals leaving the site.

The watershed under examination is located in Cherokee County, Kansas, immediately west of the town of Galena. This watershed was selected because it contains numerous mine tailings, or chat soils, and the size of the watershed will allow an examination of the effect of vegetation on sediment containment without significant interference from other uses of the land. Some urbanization, including both businesses and homes, is present. The outline of the watershed is presented in Figure 1.

The model being examined is the Agricultural Non-Point Source Pollution model, version 5.0, or AGNPS. Developed by the Agricultural Research Service (USDA-ARS), the Minnesota Pollution Control Agency and the Soil Conservation Service, AGNPS is an event-based model where the watershed is divided into square cells and sediment or runoff volumes and rates are calculated as the water is routed from cell to cell based on the defined water paths (Young, et al., 1989; AGNPS. 1994). Soil characteristics, topographic information, land use, and cover are input parameters in the program. In addition to development of the model, the Agricultural Research Service has collected data on more than 360 watersheds and has amassed the information into the ARS Watershed Database (Thurman and Roberts, 1995). This collection of rainfall and runoff data, spanning over 50 years, is available via the Internet. However, no records are available on the intermittent stream of the watershed under study.

The Agricultural Non-Point Source Pollution model consists of four components: erosion, hydrology, sediment transport, and chemical transport (Young, et al., 1989). The four components work together to determine the sediment generation, deposition and yield, runoff volume and rate, and concentrations of chemicals and pesticides after a specified storm event. The erosion component of the model employs the revised universal soil loss equation to calculate sediment erosion:

SL = EI K LS C P SSF ( 1 )

where

SL is the soil loss; the amount of sediment generated

EI is the product of storm kinetic energy and intensity

K is the soil erodibility factor

LS is the topographic factor

C is the cover factor

P is the supporting practice factor

SSF is the slope shape factor.

The eroded soil is broken down into the percentages belonging to five class sizes: large aggregates, small aggregates, sand, silt, and clay. The type and placement of vegetative cover affects the value of both the cover factor and the supporting practice factor. Within the hydrology component of the program, the runoff volume, based on the SCS curve number method, and the peak runoff rate are calculated. The equation for the peak runoff rate is the same as that used in the CREAMS model (Young, et al., 1989); it is

Q_p = 3.79 A^0.7 CS^0.16 (RO/25.4)^B ( 2 )

LW^0.19

where

Q_p is the peak runoff rate, m³/s

A is drainage area, km²

CS is channel slope, m/km

RO is runoff volume per unit surface area, mm

B equals 0.903 A ^0.017

LW is watershed length-width ratio.

The runoff volume is obtained using the SCS curve number method:

Q = ( P - 0.2 S )² / ( P + 0.8 S ) ( 3 )

S = ( 25400 / CN ) - 254 ( 4 )

where

Q is the volume per unit area of runoff, mm

P is the precipitation depth, mm

S is the storage depth, mm

CN is the curve number, and depends on land use, soil type, and hydrologic soil conditions.

Vegetation has an effect on the value of the curve number, lowering the value as the degree of cover increases. As overland flow becomes significant, the suspension and transportation of surface sediment begins. As previously stated, the detached sediment is routed from cell to cell in the model. As reported by Young, et al. (1989), the sediment transport is governed by the basic routing equation:

( 5 )

Q_S(x) is the sediment discharge at the downstream channel end, kg/s

Q_S(o) is the sediment discharge into the upstream channel end, kg/s

Q_SL is the lateral sediment inflow rate, kg/s

x is the downstream distance, m

L is the reach length, m

w is the channel width, m

D(x) is the deposition rate, kg/m²s

which is equal to

D(x) = ( V_SS / q(x) ) ( q_S(x) - g_S(x) ) ( 6 )

V_SS is the particle fall velocity, m/s

q(x) is the discharge per unit width, m³/m-s

q_S(x) is sediment load per unit width, kg/m-s

g_S(x) is the effective transport capacity, kg/m-s

which is given by

g_S(x) = nkTV² / V_SS ( 7 )

n is an effective transport factor

k is the transport capacity factor

T is the shear stress, Pa

V is the average channel flow velocity, m/s

which is determined from

V = R^2/3 S_L^1/2 / n ( 8 )

where

R is the hydraulic radius of the channel, m

S_L is the slope of the channel, m/m

n is the manning's roughness coefficient.

Values for the overland Manning's coefficient are reported for barren soil as well as grass and various crop covers. An increase in the vegetative cover results in an increase in the coefficient's value.

The dissolved chemical transport component of the model, which was developed for nitrogen and phosphorus, is presented below. At this time, the author is unaware of any modifications or improvements being done to AGNPS, but one may be able to modify the model by providing a database of soil-water partitioning coefficients for a variety of substances, including heavy metals, so that the transport of metal contamination can be examined in the aqueous phase in addition to the adsorbed phase. Young et al. (1989) report the modeled soluble transport as follows:

NUT_(sol) = C_(nut)NUT_(ext)Q ( 9 )

where

NUT_(sol) is the amount of soluble chemical, g

C_(nut) is the mean soluble chemical concentration at the surface, g/m³

NUT_(ext) is an extraction coefficient for movement into runoff

Q is the total runoff, m³.

Sediment adsorbed transport is given by

NUT_(sed) = NUT_(f) Q_S(x) ER ( 10 )

where

NUT_(sed) is the chemical transported by the sediment

NUT_(f) is the concentration of the chemical in the field

Q_S(x) is the sediment yield

ER is an enrichment ratio, given by

ER = 7.4 Q_S(x) ^-0.2 T_f ( 11 )

where

T_f is a correction factor for soil texture.

Although some question the validity of AGNPS, being empirical in nature and described as "not adequate to simulate the complex dynamic processes of runoff, soil erosion due to raindrop impact, soil detachment by runoff water, and sediment transport and deposition by runoff water in watersheds" (Borah, 1989), it is relatively easy to use, runs on a personal computer, and is used "most appropriately to compare the impacts of alternative land management strategies on surface water quality" (Panuska, et al, 1991). Thus, AGNPS gives managers a means by which to make qualitative judgments on the effectiveness and best placement of vegetative buffers or covers to control the loss of contaminated soils.

Methods

Modeling of the watershed begins with defining the boundaries of the catchment. The Kansas State Geological Survey, Lawrence, Kansas, provided several geologic survey maps for the Baxter Springs Quadrangle, in Cherokee County, Kansas. About one third of the input parameters for the model are "terrain based" and can be determined directly from these maps (Panuka, et al., 1991). Following the boundary definition, the appropriate model cell size was selected. Prior experience indicates that cell sizes usually vary between 0.01 ha and 160 ha. The larger the cell size, the less data required for entry. However, if too large a cell size is chosen, the accuracy of the values entered for the watershed characteristics, and thus the modeling process, is questionable. For the watershed under study, a 10-acre (4 ha) cell size was chosen.

Following size selection, the shape of the watershed was entered into the model by assigning numbers to each of the cells, defining which cells were receiving input flow from other cells, and entering flow directions. The variations in altitude within the catchment were modeled by specifying the slope, shape of slope, and flow length for each cell. Figure 2, shows the simulated watershed broken into cells. Arrows in the figure show the water flow path within the catchment. Mine chat is present in cells which are shaded. A permanent structure is present in the region where "S" appears.

Other input data required by the model are ground type and cover; SCS curve number; overland Manning's coefficient; K, C, and P factors in the revised Universal Soil Loss Equation; surface condition constant; and the chemical oxygen demand constant. Values for the watershed input parameters were estimated as accurately as possible, using the AGNPS User's Guide, Novotny and Olem (1994), the United States Department of Agriculture, Soil Conversation Service's Soil Survey of Cherokee County, Kansas (1985), Rainfall Intensity Tables for Counties in Kansas published by the Kansas Department of Transportation (1991), topographic maps of the Cherokee County - Baxter Springs quadrangle from the Kansas Geological Survey, and visitations to the site. Areas containing mine chat were modeled as barren or fallow areas with a silt loam soil. Information on the storm event, including the intensity and duration of rainfall, is entered as initial data for the program. The model also has the option of selecting either a rectangular or triangular channel. Calculations of the channel's widths, depths, and lengths were determined by geomorphic relationships. Much of the creek was not easily accessible and precise water channel measurements are unknown, thus this option was chosen.

It has been reported (Hamlett, et al, 1984) that the "factors found to be most closely related to soil loss and runoff include plant cover and mulches, cropping practices and tillage, rainfall variables, and slope parameters"; these were the parameters selected for variation. Using rainfall data provided by the Kansas Department of Transportation for Cherokee County, Kansas, the model was used to simulate for two-year, ten-year, and fifty-year storm events, each having a duration of 30 minutes, corresponding to rainfall rates of 2.66 iph, 3.75 iph, and 4.91 iph; or 6.76 cm/hr, 9.53 cm/hr, and 12.47 cm/hr, respectively.

The effects of vegetation on sediment erosion were examined by varying the type and percent of vegetative cover on the cells containing the chat material. Percent covers varied in 25% increments from barren soil to complete vegetative cover. Modeled vegetative types included grass and trees. Additionally, the strategic planting of vegetative buffer strips and their effectiveness in comparison to partial and complete coverage was examined. Finally, the benefits of slope reduction by terracing, as modeled by the program, were examined. Those cells which contain chat material and whose slopes exceeded 1.1% were reduced to 1%. Comparisons were also made on the combined effects of vegetation and slope reduction.

Figure 3 shows the watershed as simulated by the model. Cells containing chat material, and thus undergoing varying levels of vegetative cover, are shaded, as are the darkened regions where the vegetative buffer is placed.

Results and Discussion

To examine the effects of rainfall events and vegetation on erosional output, three different precipitation rates, corresponding to two-year, five-year, and ten-year storm return periods, were selected. For each of the separate rainfalls, the overall sediment yield, given in tons per rainfall event, was predicted for the watershed with those cells containing chat material being barren or having either a grass or tree buffer, as displayed in Figure 3. The results of each of the trials is shown in Table 1.

As can be seen, with an increase in the rainfall rate, corresponding to a larger return period, the model predicts the percent reduction in sediment yield, the amount of erodible material leaving the watershed through the outlet, will increase as well. This relation was predicted for both grass and tree buffers, with AGNPS predicting that trees are slightly more effective than grass in yield reductions, typically on the order of two percent. When running the scenarios for vegetative buffers in Table 1, complete cover of the revegetated regions was assumed. To examine the effects of incomplete cover, a range of coverage was selected, from no coverage to complete, with 25% increments in between. Table 2 shows AGNPS' predictions for those scenarios. Each of the regions containing chat material was modeled as being revegetated with 25%, 50%, 75%, and 100% survival rates of the cover planted over the entire cell.

As expected, sediment yield decreases as the percent cover increases. However, with each incremental increase in vegetative cover, the benefits, in terms of reductions in sediment yield, diminish. Approximately a 30% reduction in erosional losses is predicted with the establishment of a 25% vegetative cover, which is a realistic vegetation thickness according to work done by Pierzynski and others. If an additional 25% of the vegetation was to survive, for a total of a 50% cover, the percent reduction in yield would increase another 20% only and additional covers have even less of an effect. As of June 1997, data taken at the Galena site shows a 70% grass cover in the experimental cells.

Table 3 compares the reduction in overall yield in terms of tons of soil per acre undergoing revegetation. The overall yield for the 800-acre watershed is shows in column one. The reduction in the overall yield for the watershed, noted in the second column, was figured by taking the difference between yields for barren conditions and vegetative cover and dividing by the area of ground treated. The grass and tree covers were treated as if the entire region underwent soil amending and seeding to establish the vegetative cover, while the buffers had much smaller zones that were worked on. Note that while the buffers were modeled as having a 100% cover where they were established, the projected effective reductions, in tons per acre vegetated, were still greater than that for the 100% vegetative cover of the entire cell. Thus, the model predicts that one can choose the best means of controlling erosional losses and use existing resources as wisely as possible.

Because the sediment yield for the watershed includes erosional losses from both contaminated and uncontaminated soils, the predicted reductions in sediment generation from those sections containing chat material were determined. Sediment generation, which differs from sediment yield, is the amount of material that will become detached from the soil matrix at some point during the rainfall event and subsequent runoff. Some of the sediment generated will again settle out and this is known as the deposition. The difference between sediment generation and deposition is the yield for that section of ground. Tables 4 and 5 show the predicted generation, as well as the reduction due to vegetative buffers and varying covering thicknesses for grasses and trees, respectively. The reduction in sediment generation follows trends similar to that of overall yield, with the reductions, in terms of tons per acre vegetated, being an order of magnitude greater.

The reduction in sediment generation was determined as the difference in sediment generation between the fallow and covered conditions per the area treated. Finally, the combined contribution of terracing and vegetative cover was examined. Terracing reduces the effective slope, which has a direct relation on soil loss through the universal soil loss equation. The effects of terracing were modeled by reducing the slope of those cells containing chat material to 1%. Tables 6 and 7 show the predictions made by AGNPS on sediment yield and generation, respectively. Terracing reduced the overall yield for the watershed by 26%. If terracing was combined with the establishment of a 50% vegetative cover, then sediment yield reductions increased to approximately 57%, compared to barren conditions. Decreases in sediment generation ranged from 2.5 to 4.5 tons per acre treated, again depending upon the presence of a vegetative cover. The establishment of a 25% vegetative cover over terraced chat material would reduce the amount of possible erodible soil by 35%.

Conclusions

Although the Agricultural Non-Point Source Pollution Model is best used on a qualitative scale for deciding the best management practices for a specific site, the results are still informative. The model predicts greater benefits in controlling sediment loss from vegetative buffers with more intense rainfall events. Increasing the thickness of vegetative covers helps reduce both the sediment yield and sediment generated in the watershed; however, the increased effectiveness of thicker covers diminishes. Sediment yield reductions, in terms of tons per acre vegetated, were greatest for the 100% cover and buffers. Buffers had the highest reduction per acre revegetated, indicating that strategic planning and placement of vegetation, while not as good as complete ground cover for the purpose of soil stabilization, is effective in offering some form of containment for the topsoil. If revegetation occurs over the entire region containing chat material, and 25% of the plants survive and flourish, reductions in sediment generation on those plots would decrease by one third. Finally, terracing those regions containing chat material was predicted by the model to be more effective in soil loss reductions in the presence of vegetation.

Acknowledgments

Although this work has been funded in part by the U.S. Environmental Protection Agency under assistance agreement R-819653, through the Great Plains/Rocky Mountain Hazardous Substance Research Center headquartered at Kansas State University, it has not been subjected to the agency's peer and administrative review and, therefore, may not necessarily reflect the views of the agency. No official endorsement should be inferred. The Center for Hazardous Substance Research provided partial support.

References

Agricultural Non-Point Source Pollution Model, Version 4.03, AGNPS User's Guide, 1994.

Borah, D.K., "Sediment Discharge Model for Small Watersheds," Transactions of the American Society of Agricultural Engineers, 1989, p 874.

Hamlett, J.M., J.L. Baker, S.C. Kimes and H.P. Johnson, "Runoff and Sediment Transport within and from Small Agricultural Watersheds," Transactions of the American Society of Agricultural Engineers, 1984, p. 1355.

Kansas Department of Transportation, Rainfall Intensity Tables for Counties in Kansas, 1991.

Novotny, Vladimir and H. Olem, Water Quality: Prevention, Identification, and Managment of Diffuse Pollution, Van Nostrand Reinhold, 1994.

Panuska, John C., I.D. Moore and L.A. Kramer, "Terrain Analysis: Integration into the Agricultural Nonpoint Source (AGNPS) Pollution Model," Journal of Soil and Water, 46(1), 1991, p. 59-63.

Pierzynski G.M. and A.P. Schwab, "Characterization of Lead and Zinc Mine Tailings and Heavy-Metal Contaminated Soils in Southeast Kansas," Proceedings of the Conference on Hazardous Waste Research, Kansas State University, Manhattan, Kansas, 1991, p. 511 - 520.

Soil Conservation Service, Soil Survey of Cherokee County, 1985.

Thurman, Jane L. and R.T. Roberts, "New Strategies for the Water Data Center," Journal of Soil and Water Conservation, 50(5), 1995, p. 530-531.

Water and Pesticide Division, United States Environmental Protection Agency, Region 7,

"Clean Water Act, Section 319 Nonpont Source Management, What Is It All About?",1985.

Table 1. Effects of rainfall amounts on the overall sediment yield from a watershed with vegetative buffers.

storm return period (years)	barren soil (tons sediment)	grass buffer (tons sediment)	reduction in yield %	tree buffer (tons sediment)	reduction in yield %
2 10 50	60.05 198.99 599.03	49.16 158.36 461.00	18.1 20.4 23.0	47.94 153.92 449.03	20.2 22.6 25.0

Table 2. Effects of ground cover on the overall sediment yield of the watershed with a two-year storm event.

Vegetative Cover

Cover Type

% covered	barren soil (tons sediment)	grass (tons sediment)	% reduction	trees (tons sediment)	% reduction
25 50 75 100	60.05 60.05 60.05 60.05	42.14 29.97 22.31 18.48	29.8 50.1 62.8 69.2	40.49 27.09 19.30 17.24	32.6 54.9 67.9 71.3

TABLE 3. Comparison of vegetative buffers versus general

cover in controlling sediment yield for a two-year storm event.

Type of Vegetative Cover	Overall Yield (tons sediment)	Reduction (tons/acre vegetated)
barren	60.5	0.00
25% grass 50% grass 75% grass 100% grass buffer	42.14 29.97 22.31 18.48 49.16	0.11 0.18 0.22 0.24 0.34
25% trees 50% trees 75% trees 100% trees buffer	40.49 27.09 19.30 17.24 47.94	0.12 0.19 0.24 0.25 0.37

TABLE 4. Effects of grass cover on sediment generation in contaminated regions.

Contaminated Soil

Sediment Generated Within Contaminated Regions for Various Grass Coverage

	fallow	buffer	25%	50%	75%	100%
sediment generation (tons)	951.07	773.43	588.81	315.31	118.47	5.69
reduction (tons/acre vegetated)	0.00	5.47	2.13	3.74	4.90	5.56

Table 5. Effects of tree cover on sediment generation in contaminated regions.

Contaminated Soil

Sediment Generated Within Contaminated Regions for Various Tree Coverage

	fallow	buffer	25%	50%	75%	100%
sediment generation (tons)	951.07	774.39	614.25	348.21	141.41	1.27
reduction (tons/acre vegetated)	0.00	5.44	1.98	3.55	4.76	5.59

TABLE 6. Effects of terracing, or effective slope reduction, on sediment yield.

Ground Cover Type	Yield with No Terracing (tons sediment)	Yield with 1% Slope (tons sediment)	% Reduction
barren 50% grass 50% trees	60.05 29.97 27.09	44.50 26.14 23.80	25.90 12.78 12.14

TABLE 7. Effects of terracing and vegetative cover on sediment generation

originating from contaminated regions.

Contaminated Soil

Sediment Generated within Contaminated

	unterraced fallow	terraced fallow	terraced 50% grass	terraced 50% trees
sediment generation (tons)	951.07	521.16	172.05	190.81
reduction (tons/acre vegetated)	0.00	2.53	4.58	4.47

watershed

Galena

Figure 1. Location of the watershed under study, near Galena, Kansas. Taken from a topographic map of the Baxter Springs Quadrangle, Kansas.

Figure 2. AGNPS simulation of the watershed under study. Shaded cells contain mine chat.

Figure 3. ANGPS simulation with vegetative buffers. Vegetated regions are darkened.