Tracy Boyer
Ph.D. Candidate
Applied Economics, University of Minnesota
BOYE0044@UMN.EDU
Presented at the 2 nd World Congress of Environmental and Resource Economics
Monterey, CA (June 2002)
Preliminary Paper: Please do not quote without author’s permission

1. Introduction
In the last three decades renewed interest in wetlands as unique and useful ecosystems, rather than disease-ridden swamps, has resulted in a progressive reversal of more than a century of government policy favoring draining, filling, and conversion of wetlands. Over half of the wetlands in the lower 48 states were drained or filled between the late 1700’s and the mid 1980’s for agriculture or urban development (Dahl and Johnson, 1991). Wetlands have multiple important functions and values as habitats for wildlife, means of erosion and flood control, water filtration, and aesthetic value, among others. Wetlands directly benefit humans by providing recreation, water storage, filtration of pollutants from agriculture and urban runoff, and dissipation of the effects of floods and storms.
Today, the goal of federal policy is one of “no net loss of wetlands” which has led to a profusion of regulations and technological innovations to restore and mitigate the loss of wetlands under the increasing pressures of development. There are two levels of protection for wetland ecosystems, state and federal regulations. Among the ideas gaining favor among states is wetland mitigation banking (WMB), which is widely touted as a more market-oriented policy for decision-making in wetlands permitting and reviled by environmentalist’s as a vehicle to enable increased development of wetlands.
The idea of wetlands mitigation banking arose in the 1980’s as a way to ease the burdens of on-site mitigation and reduce the transaction costs for the developer and the Army Corps of
Engineers (ACOE). Wetland banking mitigation is a scheme by which an agency or developer typically restores a large tract a land to wetland which is valued as a “bank” in the form of credits to be sold to offset future wetland conversion. In the Midwest, wetland banks, such as

those created by the Wetlands Foundation in Ohio, primarily consist of agricultural land that is restored to its original wetland state by re-establishing water flow and replanting appropriate vegetation. The large “restored” wetland bank must meet a set of regulatory guidelines before credits may be legitimately sold in lieu of a permit. The credits for mitigation may be purchased or debited by the agency or another developer to mitigate or offset an unavoidable loss of wetlands elsewhere. The individual state authority stipulates the minimum compensation ratio and types of wetlands that are eligible for the banking program. Wetlands which are eligible for mitigation in the banking program usually must have exhausted the possibilities to mitigate on site or develop elsewhere according to the Army Corps of Engineers guidelines for permitting wetland development. In other words, a developer may purchase banked credits to offset development of a wetland only if given permission to participate in the market by the ACOE or the appropriate state or federal regulatory agency.
The Institute for Water Resources at the US Army Corps of Engineers estimates that at least 230 banks have been established, though were not necessarily under construction, as of
January 2000(Institute of Water Resources, 2000). If counted individually within state programs there are approximately 370-400 banks, of which approximately half have been developed as third party commercial banks.
In theory, WMB banking, nationally and in Minnesota, assures that functions of drained wetlands are replaced in kind in the same watershed through the purchase of credits. Compared to on-site mitigation, WMB is touted as more ecologically economically efficient. A mitigation bank allows the developer to create one large wetland instead of creating or mitigating at the site of the impact on an existing wetland. Ecologically, larger and less fragmented wetlands may

provide better habitat for sustainable ecosystems.
1
Mitigation through the purchase of banked credits is also more likely to be successful than on-site mitigation since values and functions can be evaluated before the wetland impacts occur elsewhere. Economically, restoration of larger sites allows developers to exploit economies of scale in restoration costs. King and Bohlen
(1994) estimate that there is a 31% decline in costs per acre for every 10% increase in project size. In a perfectly competitive market, the cost of a credit to mitigate for wetland loss reflects both restoration costs and the opportunity cost of land. WMB allows developers to exploit differences in land costs by purchasing mitigation credits on less expensive land in comparison to the expense of acquiring additional land to mitigate on site. Furthermore, the transaction costs of the permitting process are lessened for the credit purchaser since he or she does not need to acquire the scientific expertise and financial resources to personally mitigate wetland impacts.
In reality, wetlands are hard to build and failure is common. Wetland restoration is usually more successful than creation because the proper soils and hydrology mechanisms are already in place or can be repaired. Evidence suggests that certain types of wetlands are better suited for mitigation than others, i.e., marshlands rather than forested lowlands (Dennison and
Berry, 1993). Furthermore, some wetlands are “valued” more highly than others for recreation, water filtration, and habitat, etc. Improper design or construction of mitigation sites causes the greatest number of restoration failures. Three basic issues, hydrology, soils, and vegetation must be considered when restoring a wetland. However, even in the long term, restored wetlands are harder to maintain than natural wetlands. Even a “successful” wetland is vulnerable to degradation by future events such as storms, adjacent land uses, and takeover by invasive species such as purple loosestrife.
1 However, a diversity of sizes and types of wetlands remains important for providing a network of habitats.

Measuring “success” of wetland restoration poses another challenge since measuring loss and replacement of functions and values is difficult. If we assume that no net loss is the ultimate goal, then a 1:1 ratio of restored to destroyed original wetland area is unlikely to achieve no net loss in function even if replacement was “in-kind” by type. Across types of wetlands, measuring tradeoffs between high and low value wetland types is infinitely more difficult. Although
BWSR, the bank regulator in Minnesota, has chosen replacement ratios under its new schemes to favor replacement in-kind and within the watershed, the science of whether any compensation ratio across wetland types can adequately trade off between functions and values is uncertain.
Minnesota Wetland Mitigation Bank’s trading ratios are designed to favor 2:1 replaced to destroyed wetland acreage, if the replacement is not in kind or in the same watershed; However, there is no explicit attempt to replace functional values. The Minnesota WMB also does not use any other ecosystem or habitat evaluation measures to assign relative value to credits.
In California, Florida, and Louisiana, state conservation agencies determine sites for restoration through WMB through “advanced identification” (Fernandez, 1999). Using advanced identification criteria serves to designate priority restoration sites within the watershed by assessing potential wetland functions and values. An advanced identification system can identify the most degraded wetlands for future development while selecting others for priority restoration. Unfortunately, such identification necessitates a great deal of public investment. At present, Minnesota does not have any such system to identify wetlands that are most favorable for restoration according to the bank administrator, Bruce Sandstrom (personal communication,
2001).
In this paper, I analyze the problem of selecting potential wetland restoration sites in the seven counties of the Minneapolis-St. Paul Metropolitan area. The goal is to demonstrate the

economic and ecological tradeoffs of advance identification of potential restoration sites in pursuit of maximizing the quality of wetland acreage restored. A binary integer programming model uses data derived from a GIS mapping of the twin cities’ region to generate site selection outcomes for the creation of wetland mitgation banks. Five different cost curves and spatial outcomes under five policy scenarios are generated: the private investor’s (banker’s) decision to maximize the number of acres restored subject to his or her budget, the environmental planner’s goal to maximize the number of high quality acres given the need to compensate for a target acreage of wetland development, and finally, the environmental planner’s choice to restore the maximum number of quality of acres at least cost under three priority weightings to trade off between nitrogen retention and habitat quality.
In the next section, I will review the literature that forms the theoretical basis for the restoration site selection model used. In Section 3, I describe the derivation of the data used to characterize potential restoration sites in the twin cities. The binary integer programming models for the five policy scenarios are also given in section 3. The cost curves under the five scenarios and illustrative spatial outcomes are given in section 4. Finally, the results are discussed and future data and modeling objectives outlined in section 5.
2. Literature Review
The restoration site selection model originates from the conservation biology literature aimed at setting aside undeveloped land for species preservation. Multiple methods have been used to contend with the issue of selecting conservation sites for species, known as the “reserve site selection problem” in conservation biology or the “maximum coverage location problem
(MCLP)” in the operations research literature. The simplest model selects the smallest set of

sites in which each species is represented at least once, the set covering model (Underhill, 1994).
In the MCLP, a set of n sites is chosen to maximize the coverage of species (Church et al,
1996a). This approach implies that acquisition of all sites is equally costly since no budget constraint is included.
The maximum coverage problem was first formulated as an integer programming problem by Church and Revelle (1974). The formulation is as follows (Camm et al, 1996):
Max i


1 y i subject to : j


N i x j
 y i for all i

I (2)
(1) j


J x j
 k (3) y i

(0,1) for all i

I x j

(0,1) for all j

J.
(4)
(5)
I
J where



 i j
|
| i j


1,...,
1,..., n m

 is is the index the index set of candidate set of species reserves
N i is a containing subset of species j, is i k is the maximum the number subset of sites of candidate to be chosen reserves
The variable x j
represents each site being considered as a reserve. The x j variable equals one if site j is selected and zero otherwise. The first constraint (2), defines whether a given species i is present among chosen sites, i.e., it ensures that species i is not counted as protected if none of the sites in which it occurs is selected. Constraint two (3) limits the number of sites selected to a number, k.

In reality, the selection of which sites to conserve is both an economic and biological problem since large differences in land prices exist, particularly in growing urban regions where development threatens conservation. By including a budget constraint in the reserve site selection model, others including Ando et al, (1998), Camm, JD, (1996), and Polasky et al,
(2001), show the set of feasible conservation sites which maximize the number of species represented within the given budget. To modify the above problem to include the cost of acquiring sites (the budget-constrained MCLP), the second constraint (3) is replaced by a budget constraint wherein each site has a cost of c j
and B is the amount of the entire budget. Thus, equation 3 becomes:
 j

J c j x j

B ( 6 )
Such a budget-constrained reserve site selection model is applied in Ando et al. (1998) using information on endangered species and average land value by county for the United States.
They find that the cost of achieving the same number of species preserved was far lower in the budget constrained than the site constrained approach, i.e., choosing those sites rich with species first without regard to cost. Polasky et al. (2001), use more detailed data for individual 635 km cells of land with heterogeneous land values in Oregon. They too find that covering species using a budget-constrained approach is cheaper than the site-constrained (no budget constraint) solution by roughly 10% for coverage up to 350 species (Polasky et al, 2001).
The basic MLCP has been modified to include more complexity, such as constraining the location of sites, weighting site quality (Church et al, 2000), or allowing for varying probability that vegetation communities will be represented in a reserve network in Superior National Forest
(Haight et al, 2000). Arthur, Haight, Montgomery, and Polasky (2000) use two operations research approaches to the maximal coverage problem for reserve site selection for species

preservation. Using data from Oregon where occurrence of species is probabilistic, they compare the expected coverage approach using Monte Carlo simulations and the threshold approach that uses integer programming. Although this model does not consider the spatial makeup of selected reserve sites, the incorporation of uncertainty of whether a species will be present in a given area lends greater realism to the model as a potential conservation tool.
Binary integer programming is a special case of linear programming that is particularly suited to problems involving interrelated “yes” or “no” decisions. In spatial planning, binary integer programming (BIP) models can be used to represent whether a location site has been selected for conservation. Decision variables are restricted to two binary variables or zero-one variables. The j th yes or no decision is represented by a variable x j
, such that: x j

1 ,

0 , if if decision decision j is yes j is no
Although integer programming problems ostensibly appear easier to solve, unreasonably large numbers of finite feasible solutions may exist. For instance, if there are n variables, 2 n
solutions must be considered. Increasing n by one doubles the number of solutions, i.e., exponential growth. Furthermore, removing the divisibility assumption of the solution by restricting it to integer values means that there may not necessarily be an optimal corner-point feasible solution for the overall problem. Optimization differs from heuristic approaches in that it identifies and evaluates a set of sites for a given selection criteria, rather than selecting sites sequentially based on characteristics. Unlike heuristic approaches, optimization is not sensitive to the starting point.
However, as outlined above, binary integer problems can “blow up” in size and prove intractable to solve. (Hillier and Lieberman, 1986).
Geographical Information Software (GIS) and mathematical programming have been used extensively in combination for optimizing land use problems such as habitat protection, timber

management and landfill site selection (Campbell et al, 1992; Kao and Lin, 1996). Carver
(1991) uses a multi-criteria decision-making for optimal location of radioactive waste in UK.
Carver’s analysis has two stages. First he uses basic criteria to identify potential sites using GIS and secondly, he implements a MCDM to weight secondary objectives and political tradeoffs.
Increased availability of digitized land use data in GIS allows us to more accurately characterize data for multi-criteria site selection problems that involve tradeoffs between ecological and economic criteria.
This paper uses two environmental quality indicators to inform economic management of a resource in an interdisciplinary approach to wetlands economics and policy that is currently lacking in the environmental economics literature. Historically, the analysis of wetlands in economics has focused on economic valuation of their use and non-use benefits using hedonic analysis and contingent valuation (Taff and Doss, 1997, Bateman et al, 1995; Spash, 2000;
Oglethorpe and Miliadou, 2000, Earnhart, 2001). Fernandez and Karp (1998) and Fernandez
(1999) have developed a dynamic optimal control model for characterizing the banker’s revenue maximization problem, the only rigorous economic analysis of mitigation banking in the literature. Fernandez and Karp (1998) develop a stochastic optimal control model that incorporates the ecological uncertainty of wetland restoration success to illustrate conditions that favor increased restoration by bankers. They find that the entrepreneur’s investment in restoration increases with a reduction in restoration costs, an increase in biological uncertainty or an increase in the value of wetlands credits. The article addresses the question of whether WMB achieves the goal of increased economic and ecological efficiency in wetland mitigation by evaluating the subsidy and the change in costs of achieving different levels of wetland functions using indicator species. Although the article addresses wetland growth under uncertainty,

Fernandez and Karp only examine the effects at one WMB site in California. Using a site selection model with actual data mapped and characterized in GIS allows me to examine the potential for location bias and loss of ecological function under wetland mitigation at a landscape scale when site-specific ecological criteria are imposed under the different policy objectives.
3.
Wetland Restoration Site Data in Minnesota and Policy Scenario Models
Data on 7870 potential wetland restoration sites in the seven county metropolitan area of
Minneapolis-St. Paul is analyzed to illustrate the relative cost and spatial makeup of wetland restoration in the landscape through the selection of a set of restoration sites under five policy scenarios. The cost of achieving a given level of ecological quality as measured by the sites’ potential for water quality improvement through nitrogen abatement and habitat improvement as measured by minimizing the distance to another potential site is compared. The data for the analysis was derived from GIS data on hydric soils, the national wetlands inventory, land use, and land parcel value (MetroGis, 1972,1977, NWI, 1994, MetCouncil, 1997, MetCouncil 2002).
To be considered a potential restoration site, the land parcel is situated on land that is currently undeveloped agricultural or vacant land, is currently hydric soil, and not presently wetland (Total
Potential Wetland Restoration Sites by county are shown in Map I). Only hydric soils were considered potentially viable for wetland restoration since their presence indicates the hydrology and topography of former wetlands.
In this paper, the cost of restoring wetland is assumed to be the opportunity cost of purchasing the land, according to the estimated land value.
2
The purchase cost of the land reflects the net present value of land to the owner. The value of the land lies in its potential to be
2 In future work, the cost of restoring land per acre will also be included in the price of acquiring a site in a parcel, but the econometric analysis of survey data on the restoration data in Minnesota not yet complete .

developed or farmed; a choice, which the bank creator relinquishes irreversibly by law after the land, is restored to wetland. Therefore, market price should approximate opportunity cost to the owner if we assume the market is perfectly competitive. For each potential wetland restoration site, a weighted average of estimated land value was derived from county parcel land value data for 2002 tax rolls using the percentage of each site that overlapped known parcel values
(Hennepin county data, available for only 2001, was adjusted using the CPI for the first quarter of 2002)(Metcouncil, 2002). Because many publicly held lands are valued at zero marginal value by their county assessors, there are sites that can be obtained in the program at zero cost.
These “free” sites were left in the data set to allow for the case where county governments restore publicly held land for county road impacts. When restoration costs are included in the cost of site acquisition and restoration at a later date, these public lands while cheaper than private lands, will not longer be “free.”
When sites for restoration may be chosen freely based on cost alone, I expect that the cheapest lands will be restored first cheaper areas on the urban fringe of Scott, Carver, and
Hennepin Counties. Furthermore, when sites are chosen in any model with regard to costs, I expect the higher cost sites in Ramsey and Washington counties to be chosen last not only because of their relatively high cost, but also because of they have fewer available sites. In general, referring to Map 1, the most concentrated development is in the urban core around
Minneapolis in Hennepin county and St. Paul in Ramsey county. Scott, Dakota, and Washington counties are made up of rapidly growing second tier suburbs that are projected to have increased impact on wetlands due to housing and road construction. Summary statistics on potential restoration sites by county are included in Table 1.

The private investor’s (banker’s) problem is to maximize the area of wetland restored
(i.e., potential credits to be sold) subject to a budget constraint. The dual problem, to minimize cost of acquiring acres given an acreage constraint, is equivalent. The integer programming formulation for the private investor’s problem is as follows:
Scenario 1: Private Investor’s (Banker’s) Acreage Maximization Problem (PIAM)
Max
 i

I a i x i
( 7 ) subject
 i

I c i a i x i to :
 b ( 8 )
I where

 i | i x i

1 if

1 ,..., n
 is site i is the set of chosen , 0 if candidate not , i sites

1 ,...., n , c i

0 , opportunit y cos t of acquiring site i to per a i
 area of site i in acres b
 maximum bud get allowed for all be selected acre restoratio n .
In the private investor’s problem above, the problem is formulated as one representative entrepreneurial banker doing all the restoration allowed in the budget since they are assumed to be homogeneous here.
As an environmental planner, the wetland bank administrator is concerned with replacing impacts to wetlands with the highest quality of banked wetland acres that are available for sale as credits. The issue of how to assign weights is tricky due to the complexity of assigning functions and values to existing wetlands, much less potential wetlands (Mitsch and Gosselink, 2000). In addition to leading to increased wetland conversion, urbanization often contributes to loss of hydrological function, fragmentation, pollution, and habitat degradation of wetlands.

In this example, placing a restored wetland in a location where it might decrease nitrogen pollution in the groundwater and locating a wetland closer to another restoration project to decrease fragmentation are considered as a potential improvement in the quality of the restored landscape. Using the presence of hydric, mineral soils as an indicator of higher potential for water quality improvement, 6855 potential sites were rated of high denitrification value (n i equals one, otherwise a site was assigned a zero value). In the upper Midwest, the creation of prairie potholes (seasonal, depressional wetlands) resulted in a distinct spatial distribution of soils that allows for the identification of wetland type through soil maps (Richardson et al, 1990).
These ephemeral wetlands are better suited for denitrification of agricultural and organic pollution runoff carried in wetlands which recharge groundwater. Lehtinen, Galatowitsch, and
Tester (1999) found that fragmentation led to a negative relationship with species richness and that road density and urban land use, were negatively correlated with the number of amphibians in wetlands. As a measure of potential reversal of habitat fragmentation, sites which were located closer to the next nearest restoration site were assigned a proximity index d i
which was a normalized value of the site’s distance to the other wetland compared to the entire set of sites.
This index was calculated as follows:
0
 d

1 , d i

1 if adjacent to another potential site d i

( Di )

( Dis tan ce of site x i
( Di ) to its nearest neighbor )
( 9 ) where
D i

Greatest dis tan ce from any site in set and its nearest neighbor
Thus a potential wetland site, x i,
which is adjacent to another potential site, will have a value of d i
equal to 1. Any site that is located at some distance up to 2 miles away from its nearest

neighbor will have a proportional value on a scale of zero to one. This d i
value in future analyses will be tied to the actual choice of the adjacent cell, but in this analysis it can be construed as an option value for choosing a nearest neighbor in the next restoration period.
First we consider the environmentalist’s objective to maximize the quality-adjusted acres of restored land in order to reach a target objective to obtain a certain number of acres (k), the environmental planner’s problem with no budget constraint . A restored wetland is potentially of higher quality if it is located on a site of high value for denitrication and in close proximity to another restoration site. The quality characteristics n i and d i
are used to weight the objective function toward the selection of the highest quality sites.
Scenario 2:Environmental Planner’s Problem (No Budget Constraint) (EPNBC)
Max
 i

I
( n i a i x i
 d i a i x i
) ( 10 ) s .
t .
 i

I a i x i
 k ( 11 )
I where

 i | i x i

1 if

1 ,..., n
 is site i is the set chosen , 0 of if candidate not , i sites

1 ,...., n , a n i i
| n area i of

1 if site the i in site acres has high to be selected potential for denitrific ation , 0 if not .
k target level for restored actual acres a i d i area of site i in acres index of relative distance from next potential x i
, 1
 adjacent, 0 if the most isolated
The environmental planner’s non-budget constrained problem could be seen as the equivalent of picking the highest quality sites first regardless of cost. In a policy setting, this might occur if an the agency which oversaw the bank permitting process identified sites in advance with an actual

acreage target for wetland acreage restoration and a requirement that bankers restore the sites which are of highest potential for improving water through denitrification and most proximate to other sites to decrease habitat fragmentation.
As discussed in the MCLP in the species reserve site selection model, we are concerned with the budgetary consequences of maximizing the quality of acres restored. The environmental agency or planner may want to restore as much quality wetland acreage as possible, but would like to achieve that quality at the least cost since public funds are often limited. In the environmental planner’s problem
, the objective is to maximize the quality weighted acres of chosen sites subject to a given budget constraint.
Scenarios 3-5: Environmental Planner’s Budget Constrained Problem (EPBC)
Max
 i

I
( p n n i a i x i
 p h d i a i x i
) ( 12 ) s .
t .
 i

I c i a i x i
 b ( 13 )
I where
  i | i x i

1 if

1 ,..., n
 is site i is the set of chosen , 0 if candidate not , i sites

1 ,...., n , to be selected a n i d i i area of site i in acres
| n i

1 if index of the site relative has high distance potential from for next denitrific potential ation , 0 if not .
x i
, 1
 adjacent, 0 if retention the most and habitat , respective ly , 0
 p n isolated.

1 , 0
 p h p n
, p h c i priority weight for nitrogen p n  p h

0 , opportunit

1 y cos t of acquiring site i per acre

1 a i b
 area
0 is of the site i in acres max imum budget allowed for restoring all acres .

Each of the 7870 potential wetland restoration sites had an associated n i
and d i
value to represent its potential for reducing groundwater pollution and decreasing fragmentation of the restored habitat. For the Environmental Planner’s budget constrained problem, 3 policy weightings were analyzed to show equal weighting of the nitrogen objectives (p n
=.5 and p h
=.5), priority for the nitrogen only (p n
=1, p h
=0) and priority for habitat by valuing proximity to adjacent sites only
(p n
=0, p h
=1). The five scenarios were solved using branch and bound mixed integer programming in GAMS 20.7.11, OSL Solver (2000).
4. Results
The five maximization problems facing the private investor, the quality-adjusted acreage optimizing environmental planner with no budget constraint, and the policy-weighted scenarios for the budget-constrained environmental planner are presented here. The five scenarios were solved using a range of constraints on the budget for the banker and environmental planner problems and on acreage for the environmentalist’s problem. The results of the solutions to the five problems are summarized in Figure 1, which plots the 5 cost curves under the different scenarios. The total number of quality-adjusted acres restored is measured on the horizontal axis
(quality adjusted acreage=
 i

I
( n i a i x i
 d i a i x i
) ) at each solution under a given budget constraint.
The solutions for the private investor’s scenario were converted to the quality-adjusted acre scale by calculating the number of quality acres represented by the sites contained in the solution to the banker’s problem under the complete range of possible budget constraints. The total dollar cost of reaching a given level of quality-weighted acreage restored is measured on the y-axis in
2002 dollars.

Figure 4 shows that any given level of quality-weighted acres is cheaper under the budget constrained approaches than under the environmental planner’s acreage target scenario without a budget constraint. To restore roughly 1500 quality adjusted acres under the quality acreage maximization problem without a budget constraint is 4.4 million dollars as compared to roughly
$200,000 under the budget constrained quality maximization approach, twenty two times more expensive. The EPBC problem at 1500 quality-adjusted acres is 4.5% of the cost of acquiring the same level of quality under the EP acreage non-budget-constrained problem. Even the most expensive of the budget constrained approaches, scenario 4,which maximizes nitrogen capacity of acreage chosen, represents only 5% of the cost using the un-budget constrained approach at the quality level of 1500 acres.
Table 2: Summary of Scenarios for Achieving Roughly 1500 quality-adjusted acres
Quality Adjusted Acreage*
Expenditure
Sites
Environmental
Planner
(Budget
Constrained)
1506
$199,704
Environmental
Planner (No
Budget
Constraint)
Private
Investor
1500 1498
$4,406,345 $204,729
Water Quality
Maximizer
(Pn=1,Ph=0 )
1501.6
$223,990
Habitat
Maximizer
(Pn=0, Ph=1)
1500
$210,340
107
809
33
750
110
821
92
761
108
816 Actual Acres
* Q-adjusted acres= n i a i x i
+ d i a i x i
*Because of the integer choice of sites, achieving an exact equivalence of quality adjusted acres is not possible.
The large gap between the costs of the area constrained versus the budget constrained environmental planner’s problems shows that restoring quality lands without regard to cost to fit the target of restored actual acres will rapidly deplete any restoration budget. If the willingness to pay for restoration in the year were $200,000 and the environmental planner (EP) had two restoration scenarios to maximize quality acreage under the acreage vs. budget constrained scenarios, the spatial outcome would be drastically different as shown in Map 2. To clarify, note

that the scenario is used to choose the advanced identification of sites to be restored, and then credits are sold theoretically in a competitive market according to the cost of restoration
(acquiring land here). In Map 2, only four sites have been restored under the EPNBC problem because the best quality sites were chosen without regard to cost (achieving on 59.2 quality acres and 29 actual acres). In Map 3, the Environmental Planner under a budget constraint restores
107 sites achieves greater quality acreage (1506) and actual acreage (808 acres) at the same cost.
None of the sites are held in common between these two scenarios.
Figure 2 shows that if a given level of quality restoration is desired, the environmental planner is able to achieve it more efficiently than the private investor who does not consider quality in the selection of acres to restore. Referring to Figure 2, a detail of the lower cost curve for the five scenarios, the budget constrained scenarios in decreasing order of cost are Nitrogen retention, Habitat Quality, Private Investor, and Environmental Planner with budget constraint.
The lack of a distinct difference in costs among the budget constrained approaches is not surprising because of the high number of choices among the sites and the high incidence of cheaper sites in the outlying counties. Examining the denitrification value as the strongest indicator of environmental quality in this model, I find that the water quality priority scenario is
12% more expensive at 1500 quality adjusted acres than the environmental planner’s maximization problem and 8.5% more expensive than the private investor’s scenario. By comparing the outcomes for private investor and nitrogen maximized EP problem on the high quality nitrogen abating acres, we see that the private investor achieves only 704 acres of high nitrogen value sites and while the EP achieves 761 acres. Thus allowing the private investor to choose sites freely results in an implicit subsidy to the private banker in cost savings in achieving

a given level of nitrogen abatement when quality is not used to value trades between credits and destroyed wetlands.
At the level of 230,000 quality-adjusted acres shown in Figure 1, the cost curves for all of the budget-constrained scenarios (1 and 3-5) increase dramatically. At the level of 230,000 quality adjusted acres you can achieve 84% of the total acreage to be restored at 22% of the entire cost of restoring all available acres under the environmental planner’s budget constrained maximization problem. This hockey-stick shape of the cost curve in the budget-constrained scenarios is a result of the cheapest high quality sites being chosen first. As the budget increases, the marginal cost of an additional site rises substantially for many of the sites in Washington and
Ramsey County. For instance, one prominent industrial firm holds a 5-acre potential site in
Ramsey County near downtown St. Paul at an estimated land value of 276 million dollars on drained, potentially restorable wetland. The results of this “80-20” finding suggest that the policy maker could achieve the vast majority of restoration relatively cheaply as long as the unlikely goal of complete restoration is not pursued.
Spatial differences in the outcomes of these models show similar restoration mosaics to that shown for the Environmental Planner’s budget constrained model in Map 2. On average, the number of sites that are not held in common between any budget-constrained model and those chosen under the environmental planner’s budget constrained model represent 5% of the total area chosen by the EPBC problem. In future analyses where actual adjacency and watershed restrictions are added to the scenarios above, landscape indices of fragmentation and an accounting of restoration by watershed will shed better light on the spatial outcomes of different regulatory approaches. For this paper, little information is added by showing each map in detail.

5. Concluding Remarks
Although the number of wetlands restored under the WMB in Minnesota yearly would not significantly contribute to nonpoint pollution reduction, every restoration banking site could potentially serve as a compliment to other best management practices for improving water quality and maintaining connectivity of habitats in the landscape. There are many factors that are important in wetlands restoration site selection such as the habitat suitability for species, the probability of successful hydrological and vegetative restoration, the potential for water quality improvement through denitrification, and the diversity of wetlands in the entire landscape. Future versions of this model will include preferences for adjacent site selection and restrictions for compensation within watersheds to allow for greater ecological sensitivity of the restoration site selection model and policy relevance to changes in WMB regulation. The analysis will then explore how changing the probability of successful restoration and institutional rules, such as trading ratios and inter-watershed trading constraints affect the spatial arrangement, cost, amount, and quality of restoration.
In this paper, I have used a generalized model to integrate the issue of quality for nitrogen abatement and proximity of adjacent habitats and cost to show that the formulation of your optimization problem can greatly affect the planner’s ability to reach some given quality at a reasonable cost. These simple models suggest that if advanced identification is pursued as a policy to improve on private incentives to locate wetland banks freely where they are easiest to restore, a budget constrained approach will produce more cost-efficient outcomes than a naïve ranking approach. Furthermore, the relative cost of achieving restoration after the cheapest lands have been restored is prohibitively expensive, placing a potential choke price on permitted development of wetlands with no on-site mitigation options. As a cautionary note, these results

should not be interpreted as actual policy prescriptions, but instead should underline the importance of understanding the ecological and economic factors in the identification of potential sites for wetland restoration.

References
Ando, A., J. Camm, S. Polasky, and A. Solow. 1998. “Species Distributions, Land Values, and
Efficient Conservation.” Science 279 (27): 2126-28.
Bateman I. J., I. H. Langford, and A. Graham. 1995. A Survey of Non-users Willingness to Pay to
Prevent Saline Flooding in the Norfolk Borads. CSERGE Working Paper GEC 95-11.
Centre for Social Research on the Global Environment. School of Environmental Services
University of East Anglia, Norwich.
Camm, J. D., S. Polasky, A. Solow, and B. Csuti. 1996. “A Note on Optimal Algorithms For
Reserve Site Selection.” Biological Conservation 78 (3): 353-355.
Church, R. and C. Reveille. 1994. “The Maximal Coverage Location Problem.” Papers of the
Regional Science Association 32: 101-18.
Church, R., D. Stoms, and F. Davis. 1996a. “Reserve Selection as a Maximal Covering Location
Problem.” Biological Conservation 76 (2): 105-12.
Church, R., R. Gerrard, A. Hollander, and D. Stoms. 2000. “Understanding the Tradeoffs Between
Site Quality and Species Presence in Reserve Site Selection. Forest Science 46 (2): 157-67.
Dahl, T. E. and C. E. Johnson. 1991. Status and Trends of Wetlands in the Coterminous United
States, Mid-1970's - Mid-1980's. Washington, DC: U.S. Department of the Interior, Fish and Wildlife Service.
Dennison, M. S. and J. F. Berry. 1993. Wetlands: Guide to Science, Law, and Technology . Park
Ridge, NJ: Noyes Publications.
Doss, C. R. and S. J. Taff. 1993. “The Relationship of Property Values and Wetlands Proximity in
Ramsey County, Minnesota.” Economic Report 93-94 . Department of Agricultural and
Applied Economics, July.
Earnhart, Dietrich. 2001. “Combining Revealed and State Preferences Methods to Value
Environmental Amenities at Residential Locations.”
Land Economics 77 (1): 12-29.
Fahrig, L., J. H. Pedlar, S. E. Pop, P. D. Taylor, and J. F. Wegner. 1995. “Effect of Road Traffic on Amphibian Density.” Biological Conservation 73 (3): 177-82.
Fernandez, L. 1999. “An Analysis of Economic Incentives in Wetlands Policies Addressing
Biodiversity,” The Science of the Total Environmen t, 240: 107-22.
Fernandez, L. and L. Karp. 1998. “Restoring Wetlands Through Wetlands Mitigation Banks.”
Environmental and Resource Economics 12 (3): 323-44.

GAMS Integrated Development Environment Software. 2000. Version 2.0.7.11, Solver Module
OSL.
Haight, R. G., C. S. Revelle, and S. Snyder. 2000. “An Integer Optimization Approach to a
Probabilistic Reserve Site Selection Problem.” Operations Research 48 (5): 697-708.
Hillier, F. S. and G. J. Lieberman. 1986. Introduction to Operations Research (Fourth Edition).
Oakland, CA: Holden-Day, Inc.
Institute for Water Resources, US Army Corp of Engineers. 2000. “Existing Wetland Mitigation
Bank Inventory, Spring 2000, Sponsor and Location.” Http: www.wrsc.usace.army.mil/iwr
(Retrieved Oct 2, 2000).
King, D.M. and C.C. Bohlen. 1994. “Technical Summary of Wetland Restoration Costs in the
Continental United States,” University of Maryland, Technical Report UMCEES-CBL-94-
048.
Lehtinen, R. M., S. M. Galatowitsch, and J. R. Tester. 1999. “Consequences of Habitat Loss and
Fragmentation for Wetland Amphibian Assemblages.” Wetlands 19 (1): 1-12.
MetroGIS. 1977,72. Digital Soil Surveys for Anoka and Ramsey. Content digitized from NRCS mylar maps. Obtained from: http://www.datafinder.org
.
Metropolitan Council. 1997. Generalized Land Use 1997 for the Twin Cities.
Metropolitan Area, ArcView Polygon Coverage. Content date 04/14/1997.
Metropolitan Council. 2002. Regional Parcel Data Set—Academic Version. ArcView Polygon
Coverage of Hennepin, Anoka, Carver, Dakota, Washington, Scott, and Ramsey Counties.
Content date: 4/30/2002.
Mitsch, William, and J.G. Gosselink. 2000. Wetlands.
Third Edition. John Wiley & Sons: New
York.
National Wetlands Inventory. 1994. Minnesota, Content date 1979-88. Obtained from: http://deli.dnr.state.mn.us
.
Oglethorpe, D. R. and M. Despina. “Economic Valuation of the Non-use Attributes of a Wetland:
A Case-Study for Lake Kerkini.” Journal of Environmental Planning and Management 43
(6): 755-67.
Polasky S., J. D. Camm, and B. Garber-Yonts. 2001. “Selecting Biological Reserves Cost-
Effectively: An Application to Terrestrial Vertebrate Conservation in Oregon.” Land
Economics 77 (1): 68-78.

Richardson, J.L , L.P. Wilding, and R.B. Daniels, 1990. “Recharge and Discharge of Groundwater in the Aquic Moisture Regime, Illustrated with Flownet Analysis . Proceedings of the
Eighth International Soil Correlation Meeting : 212-218.
Spash, C. L. 2000. “Ecosystems, Contingent Valuation and Ethics: The Case of Wetland Recreation.” Ecological Economics 34 (2): 195-215.
Underhill, L. G. 1994. “Optimal and Suboptimal Reserve Selection Algorithms.”
Biological
Conservation 70 (1): 85-87.

COUNTY
SITES
ACREAGE
Average
Total
Minimum
Maximum
St. Deviation
Miles to Next Nearest
Potential Site
Average
Minimum
Maximum
HIGH N VALUE
Number
LAND VALUE/ACRE
Average
Minimum
Maximum
St. Deviation
TOTAL VALUE/SITE
Average
County Total
Minimum
Maximum
St. Deviation
ANOKA
1089
17.38
18,926
2.00
704
38.77
CARVER
2270
19.49
44,242
2.02
668
33.25
DAKOTA HENNEPIN
902 1376
23.43 13.69
21,138
2.00
1111
59.99
18,843
2.00
302
20.92
RAMSEY
128
5.31
680
2.01
71
7.60
SCOTT WASHINGTON ALL 7 COUNTIES
1522 583 7870
18.75 12.45
28,538
2.05
888
41.84
7,258
2.00
558
27.09
17.74
139,626
2.00
1111
37.70
0.06
0.00
1.23
904
0.02
0.00
0.78
1880
0.07
0.00
1.85
879
0.05
0.00
1.85
1342
0.16
0.00
0.85
76
0.04
0.00
1.03
1270
$13,237
$0
$437,364
$6,326
$0
$496,811
$13,011
$0
$298,348
$22,297
$0
$1,496,867
$0
$841,738 $69,389,300
$5,961
$397
$152,197
$52,608
$58
$1,179,007
$24,479
$190,255
$19,989
$80,063
$28,387
$163,366
$52,268
$220,129
$7,983,271
$6,158,263
$10,080
$98,555
$126,408
$422,561
$206,997,384 $181,742,671 $147,192,392 $302,897,055 $788,257,651 $150,000,385 $246,352,837
$0
$19,965,275
$0 $0 $0 $0
$5,547,594 $7,776,689 $17,653,206 $276,337,688
$956
$3,498,546
$125
$8,024,036
0.11
0.00
1.68
504
$730,175 $222,901 $465,424 $711,277 $29,898,110 $214,662 $882,710
0.05
0.00
1.85
6855
$38,448
$0
$69,389,300
$1,032,551
$257,173
$2,023,440,375
$0
$276,337,688
$3,909,306
3

2,000,000
1,500,000
1,000,000
500,000
0
0 50,000 100,000 150,000
Quality Adjusted Acreage
200,000 250,000
S2: Env. Plan. (No BC)
S3:Env. Plan. W/BC
S1:Private Investor
S5: Priority Habitat
S4: Priority N.
300,000

500
450
400
350
300
250
200
150
100
50
0
0
500 1,000 1,500
Quality Adjusted Acres
S2: Env. Plan. (No BC)
S3:Env. Plan. W/BC
S1:Private Investor
S5: Priority Habitat
S4: Priority N.
2,000 2,500