targeting nonpoint source pollution control - Wiley Online Library

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION VOL. 38, NO. 2

AMERICAN WATER RESOURCES ASSOCIATION

APRIL 2002

TARGETING NONPOINT SOURCE POLLUTION CONTROL: PHOSPHORUS IN THE MINNESOTA RiVER BASIN' John V Westra, K William Easter, and Kent D. Olson2

ABSTRACT: The state of Minnesota seeks to reduce phosphorus loading to the Minnesota River by 40 percent from current levels. Looking at one major watershed in the river basin, we examined

A range of alternative policies and practices could be used to reach this goal. Among others, practices

the cost effectiveness of targeting versus not targeting specific prac-

manure application rates, and changes in fertilizer or manure application methods. Policies may include land retirement or restrictions in farming practices by

tices or regions within a watershed for controlling nonpoint phos-

phorus pollution from agriculture. Integrating biophysical

simulation results from current and alternative farming systems with production cost and return estimates enabled us to analyze this policy. Our results indicated it is more cost effective to reduce nonpoint pollution by targeting particular regions or practices in a watershed compared to not targeting. Specifically, producers farming on cropland susceptible to erosion in close proximity to water

will appreciably reduce phosphorus nonpoint pollution loading potential by switching from conventional tillage to conservation tillage and by reducing phosphorus fertilization levels to those recommended by the state extension service. Efforts to target those

producers in the Minnesota River Basin could reduce potential transaction costs and compensation from "takings" by approximately $50 million (74 percent) over not targeting. (KEY TERMS: economic analysis; simulation; mathematical programming; transaction costs; takings.)

INTRODUCTION

Despite decades of regulation and management, many water bodies remain in poor quality. Initial

include conservation tillage, lower fertilizer or

location, taxes or restrictions on fertilizer inputs, effluent taxes, or a subsidy for pollution reducing practices. The policies may be mandated in a uniform

or nontargeted manner, or a targeted approach may be used. Each of these policies and the associated change in farming systems will affect producer income and the local economy. Before settling on a particular policy or set of policies to reduce nonpoint pollution, policy makers need to know what impacts each might have on farm income or the local economy.

To model these impacts requires an integrated approach that focuses on the differences in biophysical and socioeconomic conditions within the watershed.

If the environmental problem results from nonpoint sources that are heterogeneous in nature, one must consider the spatial distribution of the pollutant within the landscape. Using an approach that integrates socioeconomic elements and biophysical factors in a spatially heterogeneous manner, we can improve

cleanup efforts focused on point sources, even though nonpoint sources are included in the Clean Water Act of 1972. Poor water quality is particularly evident in

our understanding of the intended and unintended repercussions of policies for reducing agricultural

agricultural basins such as the Minnesota River,

nonpoint phosphorus pollution. By evaluating policies

One key pollutant in the river is phosphorus. To

within such a framework, policy makers can rank policies by factors or metrics considered critical to

improve water quality, the state and federal governments set a goal of reducing the river's phosphorus

society (such as environmental effects, agency budget impact, producer income, and local economic impacts).

which has 92 percent of its land in agricultural use.

load by 40 percent from 1980 levels (Frost and

The Minnesota River originates along the border between Minnesota and South Dakota and flows for

Schwanke, 1992).

'Paper No. 01104 of the Journal of the American Water Resources Association. Discussions are open until December 1,2002. 2Respectively, Assistant Professor, Department of Agricultural Economics and Agribusiness, Louisiana State University, 131 Agricultural Administration Building, Baton Rouge, Louisiana 70803; and Professors, Department of Applied Economics, University of Minnesota, 1994 Buford Avenue, St. Paul, Minnesota 55108 (E-MaillWestra: [email protected]). JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

493

JAWRA

Westra, Easter, and Olson

540 kilometers before joining the Mississippi River in

Minnesota River Basin (17 percent) (MWCC, 1994). Like the river basin, the Le Sueur River watershed is

St. Paul, Minnesota. Consisting of 12 major watersheds, it drains approximately 44,000 square kilometers or 4 million hectares in Minnesota, Iowa, and

dominated by agriculture. Finally, data for the Le Sueur River watershed were available, and a physical

South Dakota (Figure 1) (MPCA, 1992, 1994). Agricul-

ture within the river basin accounts for two-fifth's of the state's corn production and more than one-half of its soybean output. Considerable livestock production also occurs within the Minnesota River Basin. More

model had been calibrated to this area. Therefore, with the watershed as an example, we estimate the benefits of targeting and illustrate how they may be extended to the larger basin.

than one-fifth of Minnesota's beef production and two-

fifths of its hog output occurs in the basin (MPCA, 1994:1-11). Agriculture's prevalence and a large human population within the basin help explain the

THE PROBLEM: PHOSPHORUS POLLUTION

poor water quality. Because the scale of the basin is so large, we selected the Le Sueur River watershed in the river basin to

In appropriate quantities, phosphorus not only is beneficial, it is critical to production agriculture and, by extension, to society. Phosphorus is essential for terrestrial and aquatic plant growth. When phospho-

examine the effects of targeting efforts to control nonpoint phosphorus pollution. The Le Sueur River

rus is available in sufficient quantity for plant

is a major contributor of phosphorus load to the

uptake, it stimulates early plant growth and root

MINNESOTA RIVER BASIN MMOR WATERSHEDS AND COUNTIES MAJOR WATERSHED NUMBERS AND MAJOR WATERSHED NPMES 22

Uppor Minnesota River Major Watershed

23 Fbmme De Terre River MorWatershJ 24

Lac Oui Parb River Major Watershed

25 Hawk k - 'tIIcv Medicine River Mor WaIQrehi 26 Chippgn River Major Watershed 27 Redwood River Major Watershed 29 Middte Minnesota River Major Watershed 29 Cottonwood River Major Watershed

30 Bua Earth River Mapr Vtershed 31

Watonwan River Major Watershed

32 Le Susur River Mapr Watershed 33

Lower Minnesota Rwer Major Watershed

Major Watershed Boundary

County undary

Preparedfor: The Minne3ola Aker Banin

Joü Por Boor

Prepared r: Man kato Slate UnNerahy

ter Reiourcee Cenlar Certoqraphic and 515 Section.

Auila,ra: Cia rg and John Rongalod Eata..bnuarytS. 1556 ncool Tio,,,uoo Mc,oo,2.aan t S

Data for thia rmp wee prepared wir care ar compoaad

laoJ,,1 ,neo ID

ID 2

0

20 20

40

arerdtwwa

40n4L,

using PRCeINFO SISCPLOT.

GOkIcn,dc,,

Figure 1. Minnesota River Basin (Minnesota River Basin Data Center).

JAWRA

494

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

Targeting Nonpoint Source Pollution Control: Phosphorus in the Minnesota River Basin

development, facilitates fruit and seed production, and accelerates plant maturity. As crops take up

load. Though the Le Sueur River watershed constitutes only 9 percent of the surface area of the Minnesota River basin, it contributes 17 percent of total phosphorus load. Nine watersheds in the basin drain three-fourths of the total basin but generate only onethird of the total sediment and phosphorus loads. Previous research of sediment and phosphorus in the Minnesota River Basin shows significant increases in both loads and yields going from the western to the eastern part of the basin (MPCA, 1994). Two primary reasons for this increase are:

phosphorus in soil solution, the concentration of phosphorus in solution decreases. This causes phosphorus

from the active phosphorus pool to be released into the soil solution to reestablish a chemical equilibrium. As the amount of phosphate in solution decreases the amount of phosphate absorbed by soil decreases (and vice versa) (Busman et at., 1997). This explains why soil particles serve potentially as either a source of or

sink for phosphate to the surrounding water. When soils with high levels of phosphate (like most soils within the Minnesota River basin) erode into a water body with relatively low levels of phosphate, phosphates are released from the soil particles into the water. When phosphorus is released into water bodies with adequate nitrogen available, the biological activity of surface water increases (eutrophication). Accelerated or cultural eutrophication of surface waters caused by nutrient inputs such as phosphorus stimulates algal and rooted aquatic plant growth (Sharpley et al., 1994). As these plants expire and decompose, oxygen levels in the water may decrease and produce deleterious conditions for other aquatic life. In addition to these negative ecosystem effects, cultural eutrophication impairs amenity and recreational uses (fishing, boating, and swimming, among others), as well as industrial and municipal uses, which can have

• Mean annual precipitation increases from 56 cm on the western side to 81 cm on the eastern side of the

basin. Consequently, mean annual runoff increases from less than 5 cm on the western side to 20 cm eastern side of the basin.

• Steeper landscape and a wetter climate results in soils being more erodible in the eastern part of the basin than in the western part (MRBIP, 1999).

Significant agricultural and non-agricultural

sources of pollution degrade water quality in the Minnesota River basin. For example, it is estimated that

municipal and industrial wastewater discharges account for about 10 percent of the loading during high flow years and for about 65 percent of the total phosphorus loading in the Minnesota River during

negative local and regional economic effects. The Minnesota Pollution Control Agency (MPCA)

low flow years (MPCA, 1994). Indirect measurements suggest that nonagricultural sources of sediment and

documented frequent violations of federal or state standards for bacteria, phosphorus, turbidity, and dis-

phosphorus, such as stream bank erosion and construction sites, account for about 25 percent of the

solved oxygen at several monitoring stations in the

river's total loading (MRBIP, 1999). This implies that

Minnesota River basin in a report entitled Minnesota

agriculture may account for 10 to 65 percent of the

River Assessment Project (MRAP) (MPCA, 1994). MRAP suggested both nonpoint and point sources of pollution are responsible for degrading the river. Potential sources for these pollutants included feedlots, septic systems, wastewater treatment plants, stream and ditch erosion, and runoff or erosion from agricultural lands. During spring and summer especially, water quality in the Minnesota River can be

sediment and phosphorus loading in the river. The U.S. Environmental Protection Agency (EPA), the Minnesota Pollution Control Agency (MPCA), and

severely impacted by nonpoint pollution.

include programs to reduce the contribution of agriculture and other nonpoint sources of phosphorus poi-

the Minnesota River Basin Joint Powers Board recommended that sediment and phosphorus pollution entering the Minnesota River be reduced by 40 per-

cent from pre-1980 levels (Frost and Schwanke, 1992). Efforts to achieve the goal necessarily will

The nature and extent of the phosphorus problem is demonstrated by water samples taken over a 15year period in St. Paul, Minnesota. These samples indicate 1,450 metric tons of total phosphorus (TP) flow from the Minnesota to the Mississippi each year (MWCC, 1994:141). The average concentration of total phosphorus (0.394 mgIL) in samples was suffi-

lution in the Minnesota River. However, the key question is how nonpoint pollution reductions can be achieved cost effectively.

STUDY AREA: LE SUEUR RIVER WATERSHED

cient to cause eutrophication (MWCC, 1994).

The problem of phosphorus nonpoint pollution is spatially heterogeneous. Of the 12 major watersheds in the Minnesota River basin, the three closest to its mouth account for two-thirds of the total phosphorus JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

A major watershed of the Minnesota River and the study area for this paper is the Le Sueur River water-

shed. As in the Minnesota River Basin, intensive 495

JAWRA


agricultural production occurs in this watershed, with

and Lee, 1997). When agricultural phosphorus pollution has been analyzed, it has been an ancillary issue with sedimentation reduction analysis (such as Setia

almost all of the cropland (95 percent) planted to either corn or soybeans (USDC, 1999). Over 80 percent of the surface area of the watershed is in some type of farming system, with approximately 40 percent of the cropland under some type of conservation tillage system (USDC, 1999; CTIC, 1999). Considerable livestock production occurs in this watershed, as it does in the Minnesota River basin. Located in south central Minnesota, this watershed (No. 32 in Figure 1) is one of the 12 major watersheds of the Minnesota

and Magleby, 1987; Vatn et al., 1996a, 1996b, 1997) or

the focus of pollution reduction from livestock (Rorstad and Vatn, 1996).

In the integrated analysis we conducted, we first focused on the two agroecoregions of the watershed.

However, we found that neither the watershed nor the agroecoregions provided sufficient detail to make targeting conservation practices in critical areas effective. Therefore, the two agroecoregions were disaggregated into six major soil associations (three for each agroecoregion). Physical, chemical, and topological characteristics of the three predominant soils in each soil association were used in the biophysical simulation of all farming systems. Each of the six soil associations in the watershed was divided into areas within 90 meters of water bodies ("close") and areas not within 90 meters of water ("distant"). Thus our analysis represented differences in soil erodibility and sedi-

River basin and contains approximately 285,000 hectares. Another method for delineating regions within a river basin is to use agroecoregions. The Minnesota River Basin has 13 unique agroecoregions that are distinguished primarily by differences in soil types and geologic parent material, slope steepness, internal drainage (natural and artificial), erosion poten-

tial, and climatic factors that influence crop

productivity. Agroecoregions are zones with unique soil, landscape, and climatic characteristics. These characteristics help define the types of crop and ani-

ment and phosphorus delivery ratios to the water body.

mal production that occur in that region. Each

Fourteen producers representative of typical farming systems occurring in the six soil associations with-

agroecoregion contains unique physiographic factors that influence the potential for production of nonpoint

in the Le Sueur River watershed were surveyed. For this study, a farming system was the combination of a producer and one of the soil associations on which that producer farmed. One producer's management practices would correspond to and be reflected in several farming systems. Thus, a farming system would associate the impact on production and the environment with the unique management practices of a producer and the soil associations in which the producer farmed. If each producer farmed on every soil association, there would have been 84 farming systems on land close to water and 84 farming systems on land distant from water, for a total of 168 farming systems. However, not every producer farmed in each soil association. Therefore, only 98 current farming systems were simulated (49 systems on land close to water and the same 49 systems on land distant from water). To represent current conditions in the watershed, we simu-

source pollution and the potential for adoption of farming systems (MRBIP, 1999).

Each of the 12 major watersheds in the basin has

from two to six agroecoregions. One must consider the

variability in soils and landscapes within a watershed, as illustrated by agroecoregions, to understand the sources of nonpoint pollution and the potential for management systems to reduce phosphorus pollution.

Of the two agroecoregions in the Le Sueur River watershed, the "less steep moraine" on the eastern side of the watershed, with steeper slopes, has much higher erosion potential than the relatively flat "wetter clays and silts" agroecoregion. This suggests that targeting efforts would need to begin in the steeper region first.

METHOD: INTEGRATED ANALYSIS

lated these current farming systems on the soil association where they occurred. To allow for changes in farming systems under var-

Analyses integrating biophysical and economic policy models have included nitrates in ground water or surface water (such as Mapp et al., 1994; Helfand and House, 1995; Larson et al., 1996; Johnson et al., 1991; Wu et al., 1996), pesticides in ground water or surface water (such as Bouzaher and Shogren, 1997; Bouzaher et al., 1992), sediments in surface water (such as Braden et al., 1989; Prato and Wu 1991; Braden et al., 1985), and combinations of these (such as Randhir JAWRA

ious policies, we examined alternative management

practices for each current farming system. These alternative systems consisted of changes in tillage (conventional to conservation, for producers currently using conventional tillage), reduction in phosphorus

fertilizer application rate (from a producer's current rate to 16.8 kilograms per hectare at planting of corn),

change in phosphorus application method (from 496



broadcast only to broadcast and incorporate, for producers currently broadcasting phosphorus in the fall),

and combinations of these. Given the number of current systems (98), potential tillage systems (2), phosphorus application rates (2), and phosphorus application methods (3), there were 1,176 possible alternatives. However, because each current farming system limited the choice of changes to that system, only 270 alternative systems were simulated. Consider, for example, a farming system that currently had

to conservation tillage obtained from ADAPT simulation conformed well to observed data from Minnesota (Randall et al., 1996). Because producers identified

field locations as well as their farming systems, we represented tillage and nutrient practices (and associated sediment and nutrient effluent) spatially in the watershed.

We estimated production costs from information producers provided in our farmer survey and from the

applied only in the spring. The options or alternatives

local South Central Farm Business Management Association for the pertinent crops (corn and soybeans) (Jackson, 1999). We calculated production

available for this system would be limited to one -

costs for alternative systems based on changes in pro-

conservation tillage and had phosphorus being reducing the phosphorus application to the "low rate."

Another farming system using conventional tillage and a high rate of phosphorus fertilizer applied in the fall would have several changes to the current system

duction (equipment use and fertilizer input levels) appropriate to the systems analyzed. Our estimates of production costs for the alternative systems indicated that for most producers these

Any associated changes in costs resulting from

systems were marginally less costly than their current farming system. For example, producers switching to conservation tillage from conventional tillage generally could reduce production costs by $0.50 to $1.50 per hectare. This was consistent with estimates of cost of converting to conservation tillage in other parts of Minnesota (Olson and Senjem, 1996). Addi-

changes in tillage, phosphorus application rates, and application methods were incorporated into the estimates of the respective farming system's production costs. Likewise, net returns were adjusted because changing the farming system changed crop yields. We used producer management information to con-

tional cost reductions (with no yield reduction) occurred for systems using phosphorus application rates that were consistent with University of Minnesota Extension Service recommendations. As a result, many alternative systems were more profitable than systems currently being used by the pro-

struct representative systems that were simulated

ducer. By not selecting a more profitable choice, producers

available. There could be: (1) only a change in tillage;

(2) a change in tillage and a reduced phosphorus application rate; (3) a change in tillage and a change in phosphorus application methods; (4) a change in tillage, application rate, and application method, and so on.

using ADAPT (Agricultural Drainage And Pesticide Transport) (Desmond and Ward, 1996). ADAPT is a field scale water table management model that combines GLEAMS (Leonard et al., 1987) and DRAINMOD (Chung et al., 1992). This model was selected primarily for two reasons. First, ADAPT is able to model crop fields that have artificial drainage — a dominant feature of fields in this watershed. Second, ADAPT was calibrated to several years of nutrient loss data from several experimental plots at the University of Minnesota Experiment Station located in

are demonstrating some risk aversion. A risk-averse person is one who values a certain income more than

an equal amount of income that involves risk or uncertainty. Producers who are risk-averse would be willing to pay a risk premium to avoid situations that

may have uncertainty (such as changes in farming

the Le Sueur River watershed (Davis, 1998). The

systems). We used a method described by Olson and Eidman (1992) to estimate a risk aversion coefficient (X) for each of the alternative management decisions producers faced in reducing phosphorus load. These included: change to conservation tillage, change to

experimental plots reflected similar farming systems to those present in the watershed and those simulated for this research and analysis. Many of the soils on which crops were simulated for this study, including

change in fertilizer application method, and change in fertilizer application method combined with a change

reduced fertilizer rate, change to reduced fertilizer rate combined with a change to conservation tillage,

the major soils in the watershed, were the same as

to conservation tillage. First we estimated the cer-

those used in the calibration exercise of Davis (1998). Therefore, ADAPT was used to simulate how varia-

tainty equivalent of each current systems using:

tions in farming systems and locations within the

YCE = Ely]

watershed affected sediment and nutrient output. We also used ADAPT to estimate how crop yields changed if producers switched to conservation tillage practices. Estimated yield reductions (1 percent) from switching JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

—

(AJ2)*(2y)

(1)

where YCE is the certainty equivalent of net returns

of a system and E[y] is the expected value of net returns for that system (essentially our estimate 497

JAWRA


the certainty equivalent of net returns for that alternative system (RP = YcE - ycE)

of net returns). To estimate YCE the Pratt-Arrow absolute risk aversion coefficient (A) was assumed to

be 0.0001 per dollar for all current systems. This value was within a range of estimates for producers (Olson and Eidman, 1992). Combining information on yield variation across soils with 50 years of physical simulation with ADAPT, the variation of net returns for each system (2) was determined. Variation of net

returns (2) was the variation in crop yield multiplied by crop price, assuming production costs

remained constant for each system. With this information a certainty equivalent of net returns (.cE) for each current system was calculated. In the second step, a risk aversion coefficient (A*) for each alternative system was estimated. On each area of the farm, each producer has a current system in use and at least one alternative system from which to choose. The risk aversion coefficient (A*) for each alternative system, was estimated by substituting the certainty equivalent (cE) of the current system associated with that alternative system, estimates of net returns for that alternative system, and variations of net returns for that alternative system into equation 4.1 and solving for (A*). This method was used for each alternative system. Third, because multiple alternatives were available for some producers, the maximum risk aversion coefficients (A*max) for each type of production decision for each producer (e.g., reduction in phosphorus applica-

tion rate) was used to estimate a certainty equivalent (y*CE) for each alternative system. This step was taken because it was assumed a producer would be constant in their risk aversion across soils when considering the same type of production decision. Recall that a given producer raised crops in several soil associations but in the same manner in each soil association. For example, if a producer had a corn-soybean rotation with conventional tillage that occurred on four distinct soil associations this represented four distinct systems in the model. Risk aversion coefficients (A*) were calculated for each of these alterna-

tive systems. It was assumed a producer would actually have only one risk aversion coefficient for a change in the current farming system (e.g., changing from conventional to conservation tillage). Therefore, the maximum of all risk aversion coefficients (A*max),

geneous quality of land and livestock" (Howitt, 1995:329). The PMP approach is useful when the empirical constraint set does not reproduce base-year

results. Essentially, observed acreages and outputs are used in the PMP model to infer marginal costs for observed crop allocations. This information is included in the objective function to adjust the yield or production function to create a model that self-calibrates to baseline levels. With a PMP approach, initially a linear program model is defined that constrains farming systems to levels currently observed at the field scale. In the next step of this method (calibration), the

marginal values on the appropriate land constraints from the baseline model are used to create a nonlinear production function for the crops in each system. Essentially, the marginal values of binding land constraints (set at currently observed levels for each system) are used to adjust crop yields for all farming systems — creating new intercept and slope coefficients for the crop yield function. These adjustments to the crop yields allow the model to solve exactly to the baseline levels of cropland for each system, with-

out the cropland constraints for each system. With land constraints present only for each region, as opposed to each field, the model can select the combi-

nation of farming systems in response to the policy modeled. Traditionally, researchers established upper and lower acreage constraints for each system in linear programming models to control the response of the model to policy scenarios. Compared to a linear programming model, the PMP model can respond in a

across all soil associations on which a producer

more realistic or smooth manner to shocks from

farmed, was selected for each type of decision a pro-

changes in policy (Howitt, 1995). Additionally, the PMP model responds to policy shocks in a manner

ducer faced.

In the fourth and final step, the risk premium (RP) was estimated for each alternative production decision. The risk premium was the difference between the certainty equivalent of net returns for the current system (associated with that alternative system) and

JAWRA

These estimated risk premiums were included as part of the cost of each alternative system. Including these risk premiums allowed the model to estimate more accurately the income penalties inflicted by forcing producers to change or the incentive needed to encourage producers to switch to an alternative system. To analyze policies we developed a positive mathematical programming (PMP) model (Howitt, 1995). "The PMP approach uses the farmer's crop allocation in the base year to generate self-calibrating models of agricultural production and resource use, consistent with microeconomic theory, that accommodate hetero-

more consistent with economic theory The objective function of the linear economic model

was to maximize net farm income in the watershed (Equation 2).

498



E SeçFs(çiC Maxfl(t,m;f,s,e)= j_ iPc +GPise This objective function was subject to the following set of constraints: X1,

Aise*(l.025) Vf,s,e

Xfse

Afse *.(o.975) Vf,s,e

1Xfse i'SAfse * Vs,e

SeFs x1 0

ySeFs *.(Ø975) Ve

Vf,s,e

(3)

rnfsewn

RPise) Xfse

(2)

than 97.5 percent of observed levels for that portion of

the watershed (n = 2). Equation (7) constrained all systems to nonnegative levels. During the initial solution to the linear version of the economic model, only the 98 systems currently being used by surveyed producers were included in the analysis.

After the linear economic model was solved with this set of constraints, the calibration stage of the analysis began. First, optimal levels of the current systems were observed and recorded. Next, levels for (5)

(6)

the alternative systems were set at one acre each. Because producers were not using alternative systems, the current acreage for each of these systems was zero. However, for the PMP model to work prop-

erly, each alternative system had to be set at some positive level. One acre was selected for the alternative systems. The estimated acreage from linear

(7)

model solution and one acre for each alternative system defined Afse**. These acreage estimates were used in the calibration portion of the analysis. During

In these equations, for each system: t is tillage sys-

calibration, the objective function was the same

tem; rn is nutrient management (such as reduced

(Equation 2) but subject to constraints:

phosphorus application rates or change in application method); f is field or portion of the farm within the soil association-water proximity combination; s is soil

Xfse

association; e is proximity to water within the soil association; Xfse is acres of farming system; A18e* 5

Afse**+0.OOlVf,s,e

Xfse Afse

acres of farming system estimated to be present (from producer surveys); cfse is output c from each system; is price of output c; rnfse is variable input n used for each system; w is price of variable input n; GPise S government payment for each system; FCise is fixed costs for each system; and RPise is the risk premium

**

— 0.OOlVf,s,e

1Xfse A1

*

Vf,s,e

(8) (9) (10)

In Equation (8), land for each current and alternative system was constrained to no more than 0.001

(assumed equal to zero for all current farming systems). In the objective function, for a given farming system, total returns from per acre were

acres more than levels determined in the linear model or by assignment described above (n = 368). Land for

each system with negative net returns was constrained, with Equation (9), to no less than 0.001

and variable costs per acre were rnfsewn. To reflect current distribution of farming systems, the objective function (Equation 2) was subject to the following set of constraints. In Equation (3), land was constrained at the field level within each soil associa-

acres less than levels determined in the linear model or by assignment described above (n = 32). For Equation (10), cropland was constrained at the field level within each soil association-water proximity combination to no more than 100.0 percent of cropland limits in that area (n = 98).

tion-water proximity combination to no more than 102.5 percent of observed levels (n = 98). Land was constrained, with Equation (4), at the field level within each soil association-water proximity combination

In the calibration formulation observed average output is assumed to be

to no less than 97.5 percent of observed levels (n = 98). Equation (5) constrained the total cropland used in eac:h region (soil association-water proximity com-

qefse

bination) to no more than 100.0 percent of the total land available for cultivation in each region (n = 12). Land in conservation tillage was constrained, with Equation (6), at the water proximity level to no less JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

- FCise -

Icfse

—

5cfseXfse

(11)

is the slope where I3cfse is the intercept and for the marginal yield function of crop c (corn and 499

JAWRA


soybeans). With average crop yields assumed for the linear and calibration stages, this implies 6cfse must equal zero. Dual values (Yfse) were obtained from Equation (8) of the calibration model. Dual values reflect the opportunity cost or the imputed value of a resource such as agricultural land. They represent what the producer gives up (in terms of potential profits) for using the cropland for one farming system instead of another. These dual values were used to estimate unobserved intercept and slope coefficients for both crops (corn and soybeans): cfse = (Yfse05)/ (qciseAîse * *) I-'

cfse

(12)

= cfse + 6 cfseAfse *

(13)

Substituting Equations (12) and (13) into (11), these values are included in a new primal nonlinear model:

Max fJ(t,m; f, s,e) =

E Se Fs(C( * *cfse 6 *

*cfse

Xfse)Pc + GPfse -

'nfsen

- FC15e -

RPfse)xfse

(14)

The nonlinear model was only subject to cropland constraint 10 (n = 98). The solution to the primal nonlinear model had the same levels for farming systems observed in the calibration and linear models. The economic model was used to analyze a reduction standard or "command and control" effort to reduce phosphorus effluent (targeted and not targeted by location). To analyze the reduction standard, the objective function of the nonlinear programming model was Max fJ(t, m; f,s,e) =

E Se Fs(C( *

*cfse

6*

*cfse

This nonlinear model was subject to constraints: v Fs

/_df XfseAfse*Vf,s,e

xise)pc + GPfse -

'nfseWn

- FCfse -

RPfse)xfse

(15)

"hot spots" in effect were targeted. With this model formulation, we could evaluate the cost effectiveness of a regulated reduction in agricultural phosphorus pollution directed in a targeted or nontargeted man-

'16'

ner. (17)

In this analysis, the estimated portion of current phosphorus load attributable to agriculture would be reduced by 40 percent (Equation 18). There is dis-

(18)

agreement about how much phosphorus load is attributable to agriculture (30 to 65 percent in the

In these equations, Yc is phosphorus effluent per

Minnesota River basin), not to mention how much comes from crops versus livestock. Using the current

v Se 'ç Fs I-IS

I_u XfsefAfse*VS,e

Fs

Laf Ycfsefse(l_bWCp* Vf,s,e

distribution of crop acreage, total phosphorus loading associated with those systems was calculated for the watershed. The total phosphorus load from nonpoint

acre and b is the bound for phosphorus load reduction (0 to 0.4).

For a nontargeted or uniform reduction of phospho-

sources originating from farming systems in the

rus, Equation (16) was effective at the field level.

watershed was 30 percent of the total nonpoint load. The results reported pertain only to crop agriculture's portion of the load.

For example, if the watershed manager (state agency responsible for pollution reduction in the Minnesota River) reduced phosphorus pollution by 40 percent uniformly across the watershed, then each field had to reduce its baseline phosphorus load by 40 percent (n = 98). On the other hand, a targeted implementa-

RESULTS

tion of the policy allowed for more flexibility in achieving the desired reductions (Equation 17) (n = 12). Using an approach similar to Prato and Wu

To compare the effects on phosphorus load, cropland, and net farm income of nontargeted and targeted reductions of phosphorus nonpoint pollution from agriculture, we constrained the economic model at

(1991), a 40 percent reduction of phosphorus pollu-

tion was constrained at the watershed level and JAWRA

500



one of two scales as described above. Administering a

regulated reduction in phosphorus in either manner (targeted or nontargeted) resulted in the same level of total phosphorus load from agriculture in the watershed (Table 1). With targeted implementation the 40 percent reduction in phosphorus load is at the water-

LC)(NCO

OO (N—ICC

ci cOON

C)Co(N

(DN'-I

(N(CcO ,-4

CO Co Co

(Q(NCO N (C CC

shed scale (one phosphorus reduction constraint). As a

LtDLtD

result, phosphorus load reductions varied by region

(N

CC

(water proximity-soil association) and effectively were

(NCOcO

CO(CCo

r-cq (0CO C)0 '-I

LtD (N N

LCD 0 (0

O(NCo COOCO

(N

LCD

targeted to regions where the most cost effective

cco 0C'1 COCC

reduction in phosphorus loads occurred. For example,

phosphorus load was reduced by only 7 percent on land distant to water in soil association MN16O (Fig-

.

ure 2). On the other hand, nonpoint phosphorus pollu-

(N-lCo

CO(0 OCOC 0(0(0 CO(N

ONN (CN Co(NLC NCO NCO

0)00 CO 0 CC

tion was eliminated completely (100 percent

CO CC 0

reduction) on land close to water in soil association MN163 (Figure 2). Although only 9 percent of cropland in the watershed was within 90 meters of water

(N CO —

C)c

(D(NN (N0(N

(N

c000 NcO(D

COOC)

N CO (0 CC (N

O)N(0 N CC —

C

(close), 58 percent of the phosphorus reduction occurred there, when reductions were targeted. When the policy was implemented in a nontargeted manner, each field within each water proximity/soil

0 CC CC

(0(NCO

C)CC N CC —

(N—

association combination in the watershed had to

N0) O)CON N0) ——

LCD

(NWCO

(N-c O)Co CoCO CC_

Z

reduce its phosphorus pollution load by 40 percent (i.e., 98 phosphorus reduction constraints were effective) (Table 1). If the results from the nontargeted 40 percent reduction were presented in Figure 2, every

N

-N

O ' O-ss CO CO LCD

LCD

C) 0) CO

C) ' ,-4

region would have 40 percent reductions in loading.

N

CISCONNC CO

(N

(N '-4 CCL

CC)

When the reduction standard was implemented in a targeted manner, only 6 percent of cropland was not farmed (2 percent of the cropland distant from water

.NC0 CC CO — CO

o CC) (0

CC 'l '

and 45 percent of cropland within 90 meters of water).

On the other hand, with a nontargeted 40 percent

reduction in phosphorus load from agriculture,

O (0 (0 N CO LCD

approximately 20 percent of cropland was not farmed (20 percent on both cropland distant from and close to water) (Table 1). Because net farm income is a function of crop production (which is a function of land planted to each

LCD

(N ' cC CC1''

C') CO ,

(N LCD CO

C)CC N CO (0 '

CO

.

0 (N

NON

000) 0)0)0) '-4

(N

0

'

CC)

C) 0) CO

0) (N

(0 ' CCCN LCD O)

COOC) LCD 0 LCD

LCD

LCD

-4'

crop), reductions in cropland reduced net farm

CN (0 CC C)

CO (N

income. When the phosphorus reduction standard was implemented in a targeted manner, net farm

CC 'C LCD

income was reduced 5 percent, to $50.2 million from $53.0 million .(Table 1). In the targeted example, net farm income on land close to water decreased by onethird, while on cropland distant from water, income declined by only 3 percent. How extensively net farm income changes in response to a targeted 40 percent

cC

Cl)

CC)

C) COOCO

O(N(N (N (N

COCC -4'

*C ..

(N

0) (00(0COC) CO

..

'-4

LCDC)

CO C

'

CC CC)

C)

OCo

reduction in phosphorus is seen in Figure 3. Net

income on farms with cropland distant from water in Soil Association 79 does not change at all. Conversely, net income is eliminated on land close to water in Soil

-u a CC)

1)

Association 163.

In contrast, a 40 percent nontargeted reduction in phosphorus load reduced farm income by almost 22 percent on land close to and distant from water. It is JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

4 501

JAWRA


rht,d II

I

ri- MI

I

all

I

Cnn,hinrltc7

('ln. 17

I

flitnt I67

1

Cnnthin-d I4

I

rln4c1

I

F)itsnt ç4

Cnnhin-d I CIn 151

flkI,nt 151

I

Cnnhinrl ISO

I

Cln ISO 160

Cnnhn.-d 087 CIn-OR7

flk5nI 087 Cnnhind 070 ('In (170

I

flit,t trio

I

•20°/

0%

-60%

-40%

-80%

-100%

Reduction in Phosphorus Load

Figure 2. Targeted 40 Percent Reduction in Phosphorus Load by Water Proximity and Soil Association.

iried All

CIn All stant Alt

2ombiiied 457 ]Distant 457

lonsbined 454

CIn-144 Distant 454

('nnhim-el 151

CIn 151

flkt,nt 151

I

rnnhii) I

I

I

Thstant 087

ornbiiied 079

CIn 070

I

Distant 079 0%

-20%

-40%

-60%

-80%

-100%

Reduction in Net Farm Income

Figure 3. Change in Net Farm Income from Targeted 40 Percent Reduction in Phosphorus Load.

JAWRA

502



noteworthy that retiring land from farming was the least costly alternative for many farmers faced with a

such water quality improvement programs would be

uniform phosphorus reduction requirement.

Reducing agricultural phosphorus load by 40 percent can have a major impact on cropland, production,

CONCLUSIONS

The physical analysis indicates that certain regions of the watershed contribute disproportionately more phosphorus than do other areas. This corresponds well with observed phenomena and conforms to intuition about the nature of nonpoint source pollution in a heterogeneous environment. Most watersheds have "hot spots" that contribute more of the nonpoint pollutant than others. These may be due to physical properties of the soil, location of the field being farmed,

the farming systems occurring on that soil, or the presence of artificial drainage. In the instance of the heavy soils in soil association MN163, which are tile drained, phosphorus loss from tile drains was highest. On the other hand, steeply sloped soils in MN087 had the highest phosphorus loss from runoff. One could imagine losses in farm income as "takings" of property or production rights the farmers enjoy currently. A producer creates an externality by producing nonpoint source phosphorus pollution from current farming systems. However, the current sys-

tems examined in this research were not illegal.

Therefore, if the government must compensate the farmers, one could consider the "takings" as the difference in watershed net farm income from the baseline to one of the two 40 percent reductions modeled. In the case of the targeted implementation, the annual compensation would be approximately $3 million, while the nontargeted case would require more than $11 million in annual compensation. Thus in this watershed the savings in reduced compensation costs for targeting would exceed $8 million. Because the watershed accounts for 17 percent of the phosphorus load and only 9 percent of land in the river basin,

$14 million to $39 million.

and net farm income, depending on how the regulation is implemented. To estimate how a less stringent reduction standard would affect net farm income and cropland production in the watershed, we evaluated a lower regulated reduction in phosphorus (30 percent). If targeted, this 30 percent reduction standard would decrease producer income by only 2.5 percent (less than $1.5 million) and keep most cropland in produc-

tion (96.5 percent) while achieving a substantial reduction in phosphorus pollution. If the analysis was expanded beyond the Le Sueur

River watershed to the entire river basin, the cost savings from targeting could be substantial. Because

the Minnesota River basin is more heterogeneous than the Le Sueur River watershed, the differences in

erosion potential are even greater at the river basin level than at the watershed scale. As a result, the cost

savings would be even greater at the larger scale. What would likely happen in the Minnesota River basin is that with a targeted approach, conservation practices (conservation tillage and reduced fertilizer rates) and land retirement would be concentrated in

the eastern portion of the basin in two or three agroecoregions.

Though monitoring and enforcement costs per acre may be greater under a targeted reduction standard, considerably fewer acres would need to be monitored. Therefore, targeting the control efforts could potentially reduce total transaction costs of implementing a program to cut phosphorus loading to the Minnesota River by 40 percent. An earlier study by McCann and

Easter (1999a) found that farmers in the Minnesota River basin favor programs targeting highly erodible soils. This can be critical in determining the total cost

of controlling nonpoint pollution. For example,

McCann and Easter (1999b) estimated the transaction costs of implementing nontargeted land retirement programs or conservation tillage programs at

annual savings for the entire river basin might approach $50 million using these model results.

$9.4 million and $7.9 million, respectively. Of transaction costs estimated, administrating, monitoring, and

1997. Using a survey of residents in a subset of counties in the basin, they estimated households would be willing to pay $14 to $39 per year for a hypothetical

prosecuting costs comprised of $2.8 million for land retirement programs and $7.4 million for conservation tillage programs (McCann and Easter, 1999b). If targeting allowed agencies to decrease their administration, monitoring, and prosecution costs by 50 percent, this potentially could reduce transaction costs by $1.4 million for land retirement programs and $3.7 million conservation tillage programs for the basin. Combining reduced transaction costs with reduced compensation from the "takings" estimated earlier, could save $51 million to $54 million by targeting

Another question often raised in relation to the costs of reducing nonpoint pollution is the benefits that society enjoys by have a cleaner environment. Greden et al. (1999) examined the value of reducing phosphorus pollution in the Minnesota River basin in

water quality improvement program that would reduce phosphorus pollution by 40 percent. With slightly more than one million households in the basin, they estimated the total willingness to pay for JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

503

JAWRA


nonpoint sources of phosphorus pollution in the Minnesota River Basin.

SUMMARY

In this study we used an integrated model to estimate the impacts of reducing phosphorus loading to the Minnesota River by 40 percent from 1980 levels. Using a biophysical process model, we simulated a set

of farming systems representative of the range currently observed in this watershed. Additionally, we simulated alternatives to current systems that could potentially reduce phosphorus loading potential. Estimates of phosphorus load from the simulations of current and alternative farming systems were combined

with production cost and return estimates in an economic model of the watershed.

Our results from this integrated model indicate that significant cost savings can be achieved in reduc-

ing nonpoint pollution by targeting particular practices or regions of a watershed. Specifically, producers

farming on cropland susceptible to erosion in close proximity to water will appreciably reduce phosphorus nonpoint pollution loading potential by switching from conventional tillage to conservation tillage and reducing phosphorus fertilization levels to those recommended by the state extension service. Efforts to target those producers could reduce potential transac-

tion costs and compensation from "takings" by

Chung, S. 0., A. D. Ward, and C. W. Schalk, 1992. Evaluation of the Hydrologic Component of the ADAPT Water Table Management

Model. Transactions of the American Society of Agricultural Engineers 35:571-579. Conservation Tillage Information Center (CTIC), 1999. Consei-vation Tillage Survey Data. Available at . Accessed in July 1999. Davis, D. M., 1998. Simulation of Tile Drainage and Nitrate Loss From a Clay Loam Soil. M.S. Thesis, University of Minnesota, St. Paul, Minnesota. Desmond, E. and A. D. Ward, 1996. ADAPT: Agricultural Drainage

and Pesticide Transport User Manual, Version 4.1. The Ohio

State University, Department of Agricultural Engineering, Columbus, Ohio.

Frost, J. and S. Schwanke, 1992. Interim Strategy to Reduce Nonpoint Source Pollution to the Minnesota River. Publication No. 640-92-038, Metropolitan Council, St. Paul, Minnesota.

Greden, L. M., F. R. Homans, and K. W. Easter, 1999. Reducing Phosphorus Pollution in the Minnesota River: How Much is it Worth? Staff Paper No. P99-4, University of Minnesota, Department of Applied Economics, St. Paul, Minnesota. Helfand, G. E. and B. W. House, 1995. Regulating Nonpoint Source Pollution Under Heterogeneous Conditions. American Journal of Agricultural Economics 77:1024-1032. Howitt, R. E., 1995. Positive Mathematical Programming. American Journal of Agricultural Economics 77:329-342. Jackson, D., 1999. 1998 Annual Report, South Central Minnesota. South Central Technical College, Mankato, Minnesota. Johnson, S. L., R. M. Adams, and G. M. Perry, 1991. The On-Farm Costs of Reducing Groundwater Pollution. American Journal of Agricultural Economics 73: 1063-1073.

Larson, D. M., G. E. Helfand, and B. W. House, 1996. Second-Best

Tax Policies to Reduce Nonpoint Source Pollution. American Journal of Agricultural Economics 78:1108-1117.

Leonard, R. A., W. G. K.nisel, and D.A. Still, 1987. GLEAMS: Groundwater Loading Effects of Agricultural Management Sys-

tems. Transactions of the American Society of Agricultural Engineers 30(5):1403-1418.

approximately $50 million (74 percent) over not tar-

Mapp, H. P., D. J. Bernardo, G. J. Sabbagh, S. Geleta, and K. B.

geting.

Watkins, 1994. Economic and Environmental Impacts of Limiting Nitrogen Use to Protect Water Quality: A Stochastic Region-

al Analysis. American Journal of Agricultural Economics 76:889-903.

LITERATURE CITED

McCann, L. M. and K. W. Easter, 1999a. Differences Between Farmer and Agency Attitudes Regarding Policies to Reduce

Braden, J. B., G. V. Johnson, and D. G. Martin, 1985. Efficient Control of Agricultural Sediment Deposition in Water Courses. In: Options for Reaching Water Quality Goals, T. M. Schad (Editor). Tech. Pub. Series 84-2, American Water Resources Association, Middleburg, Virginia, pp. 69-76. Braden, J. B., G. V. Johnson, A. Bouzaher, and D. Miltz, 1989. Opti-

Phosphorus Pollution in the Minnesota River Basin. Review of Agricultural Economics 2 1(1): 189-207.

McCann, L. M. and K. W. Easter, 1999b. Transaction Costs of Poli-

cies to Reduce Agricultural Phosphorus Pollution in the Minnesota River. Land Economics 75:402-414. Minnesota Pollution Control Agency (MPCA), 1992. The Minnesota River. Reclaim a Legacy. Minnesota Pollution Control Agency, St. Paul, Minnesota.

ma! Spatial Management of Agricultural Pollution. American Journal of Agricultural Economics 71:404-413. Bouzaher, A. and J. Shogren, 1997. Modeling Nonpoint Source Pol-

Minnesota Pollution Control Agency (MPCA), 1994. Minnesota River Assessment Project: Project Summary. In: Minnesota River Assessment Project Report. Minnesota Pollution Control

lution in an Integrated System. In: Modeling Environmental Policy, W. E. Martin and L. A. McDonald (Editors).Kiuwer Academic Publishers, Boston, Massachusetts, pp. 7-42. Bouzaher, A., D. Archer, R. Cabe, A. Carriquiry, and J. F. Shogren,

Agency, St. Paul, Minnesota, Vol. 1.

Minnesota River Basin Data Center, 2001. Minnesota State

1992. Effects of Environmental Policy on Trade-Offs in AgriChemical Management. Journal of Environmental Management

University, Mankato, Minnesota. Available at . Accessed in November 2001. Minnesota River Basin Information Page (MRBIP), 1999. Depart-

36:69-80.

Busman, L., J. Lamb, G. Randall, C. Rehm, and M. Schmitt, 1997.

ment of Soil, Water, and Climate, University of Minnesota. September 1999 Available at . Accessed in September 1999.

504


Targeting Nonpoint Source Pollution Control: Phosphorus in the Minnesota River Basin Minnesota Waste Control Commission (MWCC), 1994. Water Quali-

ty Analysis of the Lower Minnesota River and Selected Tributaries: River (1976-1991) and Nonpoint Source (1989-1992) Monitoring. Volume 1, Report No. QC-93-267. In: Minnesota River Assessment Project Report, Volume II. Minnesota Pollution Control Agency, St. Paul, Minnesota. Olson, K. D. and V. R. Eidman, 1992. A Farmer's Choice of Weed Control Method and the Impacts of Policy and Risk. Review of Agricultural Economics 14: 125-137.

Olson, K. D and N. Senjem, 1996. Economic Comparison of Incre-

mental Changes in Tillage Systems in the Minnesota River Basin. Minnesota Extension Service FO-6675-C, University of Minnesota, St. Paul, Minnesota. Prato, T. and S. Wu, 1991. Erosion, Sediment, and Economic Effects of Conservation Compliance in an Agricultural Watershed. Journal of Soil and Water Conservation 46:211-214.

Randall, G. W., W. E. Lueschen, S. D. Evans, and J.F. Moncrief, 1996. Tillage Best Management Practices for Corn-Soybean Rotations in the Minnesota River Basin. Minnesota Extension Service FO-6676-C, University of Minnesota, St. Paul, Minneso-

ta. Randhir, T. 0. and J. G. Lee, 1997. Economic and Water Quality Impacts of Reducing Nitrogen and Pesticide Use in Agriculture. Agricultural and Resource Economics Review, pp. 39-5 1.

Rorstad, P. K. and A. Vatn, 1996. Environmental Policy Measures for Livestock Production: An Integrated Economics and Natural Science Analysis. lOS Discussion Paper No. D-6/1996. Agricultural University of Norway, Norway. Setia, P. and R. Magleby, 1987. An Economic Analysis of Agricultur-

al Nonpoint Pollution Control Alternatives. Journal of Soil and Water Conservation 42:427-431. Sharpley, A. N., S. C. Chapra, R. Wedepohl, J. T. Sims, T. C. Daniel,

and K. R. Reddy, 1994. Managing Agricultural Phosphorus for Protection of Surface Waters: Issues and Options. Journal of Environmental Quality 23:437-451.

U.S. Department of Commerce (USDC), Bureau of the Census, 1999. 1997 Census of Agriculture. U.S. Government Printing Office, Washington, D.C. Available at . Accessed in July 1999. Vatn, A., L. Bakken, P. Botterweg, and E. Romstad, 1996a. Eco-Eco

Modeling: A Complex Game that Pays. lOS Discussion Paper No. D-7/1996, Agricultural University of Norway, Norway.

Vatn, A., L. Bakken, P. Botterweg, H. Lundeby, E. Romstad, P. K. Rorstad, and A. Vold, 1996b. Regulating Nonpoint-Source Pollution from Agriculture: An Integrated Modeling Analysis. lOS Discussion Paper No. D-8/1996, Agricultural University of Norway, Norway.

Vatn, A., L. R. Bakken, H. Lundeby, E. Romstad, P. K. Rorstad, A. Vold, and P. Botterweg, 1997. Regulating Nonpoint-Source Pollution from Agriculture: An Integrated Modeling Analysis. European Review of Agricultural Economics 24(2):207-229. Wu, U., H. P. Mapp, and D. J. Bernardo, 1996. Integrating Economic and Physical Models for Analyzing Water Quality Impacts of Agricultural Policies in the High Plains. Review of Agricultural Economics 18:353-357.


505

JAWRA