Demand for water for environmental uses, such as to provide critical habitat for endangered species, has increased competition for agricultural water supplies. In the western United States, a significant portion of these water demands is to increase in-stream flows. Given that Endangered Species Act (ESA) requirements supersede prior appropriation rights, ESA water demands have the potential to reduce agricultural diversions, particularly in times of drought. This situation occurred in 2001 in the Klamath Basin of southern Oregon and northern California: an ESA-related judicial ruling on the needs of several endangered fish species resulted in a major reduction in water diversions to the Klamath Reclamation Project. Using the Klamath Basin as an empirical backdrop, this study examines the potential benefits and challenges of water banks to mitigate damages to appropriative water rights holders and to provide water for environmental purposes. Results from this case study indicate that water banks are a potentially cost effective way to meet environmental needs. This study, however, illustrated several of the challenges of implementing a water bank given that modifications to the proposed bank are needed to achieve cost efficiency. Specifically, expanded trading is needed, both intraproject and interproject, to achieve the objectives of providing environmental water at minimum cost to society.
 In many areas of the western United States, increasing demand for water for environmental uses, such as provision of habitat for endangered fish and wildlife, has decreased water available for agriculture and other consumptive uses. One of the most dramatic examples of this occurred in the Klamath Basin of southern Oregon and northern California in 2001, when enforcement of provisions of the Endangered Species Act (ESA), coupled with a drought, resulted in a major reduction of irrigation water to the Klamath Reclamation Project, one of the oldest federal irrigation projects in the Untied States. The ensuing economic effects on farmers and local communities and the subsequent political controversy that arose focused attention on ways of meeting ESA needs for endangered species while minimizing economic and social disruptions to rural communities that are dependent on federal irrigation projects.
 One commonly suggested approach to providing water for environmental uses is the creation of a water bank. In its 2002 biological assessment concerning management of the Klamath Project the U.S. Bureau of Reclamation (BOR) proposed creation of a water bank in 2003 to provide ∼50,000 acre-feet (1 acre-foot = 1234 m3) of water to meet ESA needs for three endangered fish species in the Klamath River Basin. (Of this total, 30,000 acre-feet were to come from idling of land. The remaining 20,000 was to be from groundwater.) The water is to be purchased, on a temporary basis, by the BOR from willing sellers in the project. Ultimately, it is envisioned that the bank could be expanded to 100,000 acre feet in future years. During the 2003 irrigation season the BOR was able to purchase water for this bank via a combination of land idling and groundwater pumping. Preliminary water prices were ∼$75 per acre-foot. Although the BOR has begun operating the bank, procedures for improving the efficiency of the bank in future years, including the geographic scope of the bank and how bids will be accepted, are still being explored. Also, it is not clear how such a water bank, if enacted on an annual basis, will affect on-farm and project revenues and hence local communities as well as actual water consumption within the project. Additional uncertainty arises since land idling and changes in water consumption (crop evapotranspiration) by project irrigators can have significant third-party effects, depending on the spatial pattern of water trading.
 The overall objective of this paper is to explore the efficiency of a water bank in the Klamath Project, including changes in total project revenues and actual water consumption. Specific objectives are to (1) review the use of water banks for environmental purposes in this region, (2) estimate the quantity of water to be acquired in the Klamath Project under a range of alternative price conditions, (3) measure how the presence of a bank will affect total and per acre water consumption patterns in the project, and (4) compare the present bank with a more comprehensive intraproject water bank that allows farmers to trade among themselves.
2. Empirical Backdrop
 The Klamath River originates in south central Oregon at elevations of 7000–9000 feet (2133.6–2743.2 m) above sea level. The headwaters descend to form the Sprague, Williamson, and Wood Rivers. These rivers and smaller tributaries drain into Upper Klamath Lake, a large, shallow natural water body that is regulated by the U.S. Bureau of Reclamation for water supply purposes through its Klamath Project. Water is diverted from the southern end of the lake to serve a large portion of the irrigated lands in the Klamath Project. Water flowing from Upper Klamath Lake creates the main stem of the Klamath River, which is impounded by dams to form five major reservoirs that provide hydroelectric power and regulate flow to the lower Klamath River. Iron Gate Dam on the Klamath main stem regulates flow from the main stem reservoirs to the Klamath River below, which receives water from four major tributaries as it flows 190 miles (305.71 km) to its estuary in northern California [National Research Council (NRC), 2002].
 Two endemic fishes of the upper Klamath Basin, the shortnose sucker (Chasmistes brevirostris) and the lost river sucker (Deltistes luxatus), were listed as endangered under the Federal Endangered Species Act (ESA) in 1988 by the U.S. Fish and Wildlife Service (USFWS). The USFWS cited overfishing, water management, habitat alteration, nonnative species, poor water quality, and several other factors as likely contributors to the decline of these once abundant fishes. In addition, the southern Oregon-northern California coast evolutionary significant unit of coho salmon, which is native to the Klamath Basin and several adjacent drainages, was listed by the National Marine Fisheries Service (NMFS) in 1997 as threatened under the ESA. The NMFS cited water management, water quality, loss of habitat, overfishing, and several other potential causes of decline for the coho salmon [NRC, 2002].
 Irrigators, Native American tribes, the threatened and endangered species and six federal wildlife refuges compete for Upper Klamath Basin water. The bureau's Klamath Project encompasses ∼220,000 acres of land in the south part of the basin. Initiated in 1905, the Klamath Project is one of the first federal irrigation projects constructed by the U.S. Bureau of Reclamation. In 1996, irrigators in the project generated $107 million in revenue by farming 190,000 acres of land. Production is primarily barley, alfalfa hay, and irrigated pasture, with smaller acreages devoted to higher-valued potatoes and onions (M. Green, USBR, personal communication, 2001).
 The waters of the basin support substantial wetland habitats, which are essential for migratory waterfowl in the Pacific Flyway. The Klamath River itself is one of the Pacific coast's most important rivers for the production of salmon and steelhead trout. Currently at issue in the basin is the historic allocation of water between agricultural and environmental uses. Biological opinions issued by the USFWS and by NOAA Fisheries in 2003 found that the operations of the project pose a threat (“jeopardy”) to the listed species.
 In addition to their effect on fish habitat, project operations also affect wetland habitat for resident and migratory waterfowl and other wildlife species. This is because agricultural drainwater and return flows from the project are important water sources for two of the federal wildlife refuges, Tule Lake National Wildlife Refuge (TLNWR) and Lower Klamath National Wildlife Refuge (LKLNWR). The proximity of farming and wildlife in the Klamath Project is striking. Farm fields border on or are located within the TLNWR and the LKLNWR, and water flows directly into the marshes from irrigation drainage channels. Within the Klamath Project the distinction between canal and drainwater becomes blurred as conveyance channels carry a mixture of source water and irrigation return flows. Legally, the LKLNWR and TLNWR hold junior water rights to agriculture, so they receive water only after all other demands are met. Exacerbating the shortages caused by holding junior rights is the fact that most of the water the refuges receive for wetlands is agricultural return flow, which is severely reduced in years when the bureau reduces deliveries to farmers [Burke, 1999]. For example, during the water curtailment of 2001, major portions of the wildlife refuges were dewatered by the loss of agricultural return flows. This resulted in a substantial loss of wetland habitat for resident and migratory waterfowl.
 In short, between Upper Klamath Lake and the Klamath River, both geographically and politically, lies the BOR's Klamath Project. The bureau must balance the tradeoffs between water level requirements in Upper Klamath Lake, streamflow requirements in the Klamath River, and water deliveries to agriculture. Agricultural deliveries, in turn, determine the quantity of water available to the national wildlife refuges.
3. Water Markets and Water Banks
 Water markets and water banks are being developed in the western United States as cost effective and equitable tools to reallocate water, either between relatively lower valued uses to higher-valued uses or for use in the environment (see Saliba and Bush , NRC , or Getches  for detailed discussions of the history, development and application of water markets). The terms “water markets” and “water banks” have varying meanings in the literature and in application. Generally, water markets refer to the temporary or permanent transfer of a water right or contract entitlement for the use of water. For example, a temporary water market was created during the 2001 drought to allow irrigation districts in the BOR Yakima Project in Washington to sell water between districts within the project. During years of full water supply, water is not traded. Conversely, water markets in Colorado have resulted in permanent shifts in water from agriculture to municipal use.
 Water banks can take a variety of forms. In Kern County, California, the Kern County Water Bank provides underground storage of water in wet years, recovering the water in dry years. In the Klamath Project in 2003, the water bank referred to compensating growers for a voluntary reduction in their historical delivery quantity, primarily through land idling [U.S. BOR, 2003]. Although the forms of water market and water bank programs vary, they each face similar obstacles to actualizing transfers. These obstacles frequently keep transaction costs high and the volume of water transferred lower than expected. For example, the Quantification Settlement Agreement (QSA) which would transfer water from the Imperial Irrigation District (IID) to the city of San Diego and the Metropolitan Water District of Southern California (MWD) has proven difficult to complete despite the potentially serious consequences to MWD and IID of losing a portion of their Colorado River water supplies. In March 2003 the U.S. District Court granted a preliminary injunction returning irrigation water to farmers in the Imperial Irrigation District. State, federal, and local agencies continue to work towards implementing the QSA. A brief review is presented below of programs under development in Oregon and California designed to transfer water to environmental purposes and the obstacles to transfers encountered by these efforts.
4. Programs and Projects to Transfer Water to the Environment
 Four programs exist or are being developed in California to reallocate water to environmental uses. These programs are state or federally funded. They are the Environmental Water Account (EWA), the Environmental Water Program (EWP), the Water Acquisition Program (WAP) and the Drought Water Bank (DWB). The EWA and EWP are part of the CALFED Bay-Delta Program. CALFED is a cooperative effort involving state and federal agencies with management and regulatory responsibilities in the San Francisco Bay/Sacramento-San Joaquin Delta Estuary (the Bay-Delta). Among its various goals, CALFED aims to provide a long-term solution to problems affecting the ecological health of Bay-Delta. The WAP, formed under the authority of the Central Valley Project Improvement Act (CVPIA), is a joint U.S. Department of the Interior program of the BOR and the U.S. Fish and Wildlife Service. The WAP acquires water for protecting, restoring, and enhancing fish and wildlife populations to meet the goals of the CVPIA. The state's Drought Planning Program would build upon the experience gained with the DWB, which is intended to address immediate water shortage problems in California until the CALFED long-term solution for water supply reliability is in place. Table 1 shows the water acquired between 1993 and 2001 by the WAP. As Table 1 indicates, substantial amounts of water have been transferred for fish and wildlife purposes at a wide range of prices.
Table 1. Summary of WAP Water Transactions
Total Water Acquired, acre-feet
Price Range, $ per acre-foot
 Oregon's experience with water banks is much more limited than that California's, both in terms of number of trades and quantity of water traded. Relatively small quantities of water have been leased, on a seasonal basis, for the benefit of in-stream uses. For example, the Oregon Water Trust has secured small quantities of water (typically <5 ft3 s−1 (14.1585 × 10−2 m3 s−1)) for in-stream uses in the Deschutes River of central Oregon. Additionally, small flow augmentations have been achieved via leasing in the Rogue River valley. However, the market is sufficiently thin to question whether the value (prices) attached to such leases reflects the marginal value of water [Stevens et al., 2000].
5. Obstacles to Transfers
 Both water markets and water banks face obstacles to transfers. Obvious conditions that must be met for water transfers to be feasible are differences in value across uses and well-defined property rights. Several unique characteristics of water also make it difficult to transfer water from the original place and use to an alternate site (conveyance systems, such as canals or natural water courses, are needed to move water to another site). Third-party effects exist in the form of impacts on regional economies of exporting regions along with potential for conflict over the use of return flows. In addition, there are issues of water rights law, hydrology, and engineering to consider in any transfer. In short, a transfer must be both feasible (legally, technically, and economically) and appropriate [Haddad, 1999].
 The types of characteristics that need to be analyzed prior to completing a transfer in California are described in the Guide to Water Transfers published by the California State Water Resources Control Board. The process is sufficiently complex and, some would argue, onerous to discourage transfers. In the Oregon portion of the Klamath Basin, water rights have not been adjudicated. Until these rights, which include Tribal Trust and ESA provisions, are defined by Oregon courts, permanent water sales or transfers between, say, irrigators above Upper Klamath Lake and those in the project, will not be institutionally feasible (although such transfers are technically and economically feasible). Despite such obstacles, water transfers for environmental purposes are occurring with greater frequency. One of the fascinating developments of these nascent water markets is the development of the price of water. Water pricing during the BOR's pilot water bank in the Klamath Project reveals some of the challenges in “getting the prices right.”
6. Establishing the Price
 The complications that surround estimating an offer price for water by the BOR or any other agency are many. Water has not historically been considered a commodity. However, increases in demand, combined with increasing difficulty in developing new structural solutions that would increase water supply, have encouraged the use of market-like transfers to reallocate existing water supply. These market-like transfers depend, in part, on the negotiation of an offer price.
 As with most commodities an offer price is expected to be linked to the quantity offered for sale, or supplied. The higher the offer price the higher the quantity offered for sale. An example of this occurred in the 1991 Drought Water Bank in California. For a brief write-up on the California Emergency Drought Water Bank, see http://www.westlandswater.org/wtrsupply/droughtwb.htm, where a reprint of a Water Education Foundation report is provided. A panel of experts was assembled to determine what the offer price should be to transfer water from willing sellers (presumably agricultural users) to buyers (presumably urban users in southern California). The state decided to offer a price of $125 per acre-foot, which was based on estimates of what farmers would receive from growing relatively low-value crops plus an amount added as an incentive to sell water. At that price the state received offers from willing sellers for a total of 820,000 acre-feet of water. However, the demand for water (the amount the state wished to purchase) in 1991 was only 400,000 acre-feet. (The state did purchase all the offered water. The state carried over the remaining 380,000 acre-feet, which they then sold in 1992.) One reason that the state acquired more water than they needed in 1991 is that the offer price was set higher than necessary. The following year, in 1992, the state reduced its offer price to $75 per acre-foot. The reduction in offer price occurred in part to balance the quantity of water offered for sale with expected demand (decrease the quantity of water that would be offered for sale).
 Other factors besides the quantity of water to be purchased affect the offer price for water. The hydrologic year type, in particular, also plays a large role in determining the offer price. As an example, the EWA in California has historically paid between $50 per acre-foot and $125 per acre-foot to northern California agricultural users, depending on the hydrologic year type (prices vary with the magnitude of water shortages).
 The needs of the BOR with regard to the Klamath Project are somewhat different than the goals of the California programs. The California Drought Water Bank buys and sells water to willing participants. The BOR is not attempting to buy water from the project water users in order to sell it to any other user. Rather, the BOR is attempting to reduce irrigation water demand and shift this water savings to environmental uses for the benefit of the endangered fish species. Despite this dissimilarity, the lessons learned through the implementation of such programs as the Drought Water Bank (as well as the other programs listed above) are applicable to the Klamath Project, particularly when it comes to establishing an offer price. The following quantitative analysis focuses on the possibility of using the offering price as a tool to elicit varying quantities of water offered for sale by Klamath Project growers. The analysis estimates the value farmers would receive from growing crops and then adds an amount as an incentive for farmers to sell water. This information can then be used to estimate a supply curve for water (the amount of water sellers are likely to offer at varying prices).
7. Analytical Framework
 We use an integrated approach that includes models of both the hydrology and the economic features of the project to evaluate the effects of a water bank on Klamath Project irrigators and to estimate the underlying supply curve area for water. Specifically, this analysis links an on-farm economic decision model with a basin-wide hydrologic model. The economic and hydrologic models are linked by irrigation efficiency to form the Klamath Basin model (see Burke  for more details). A simplified hydrologic model of return flow in the basin is
where RF is return flow, I is inflow (or agricultural deliveries), and IE is the basin irrigation efficiency rate. Recall that IE = ETc/I, so equation (1) reduces to RF = I − ETc, which states that return flow is the remainder of inflow, I, less crop evapotranspiration (ETc). Recall that in this simplified example, deep percolation is ignored.
 The farmer's goal is assumed to be profit maximization, and the “choice” decision is assumed to be the allocation of resources, including land, applied water, variable costs, labor, and capital dollars (represented as amortized capital investment), across crops, which will maximize profits. It is assumed that farmers choose IE rates indirectly by determining the optimal combination of applied water, capital, and labor costs required to deliver the yield maximizing quantity of crop evapotranspiration (ETc) to the crop. In addition to the standard requirement of quasi-concavity the production function should possess two properties to accurately model agricultural production: (1) input substitution between applied water and the combination of labor and capital costs and (2) the ability to account for land heterogeneity. Each of these properties is discussed below.
 Substitution between labor, capital, and applied water is represented by the farmer's choice of irrigation technology and management. For example, flood irrigation requires less capital but significantly more applied water than drip irrigation, yet both technologies can supply the yield-maximizing quantity of ETc. The practice of delivering less ETc than is yield maximizing is referred to as stress irrigation. On the basis of reports and personal conversations with Agricultural Experiment Station staff working in the Klamath Project area, stress irrigation is not practiced on the crops grown within the project.
 For simplicity, two inputs, labor and capital, are combined into one composite good called technology, making ETc a function of applied water and expenditures on technology. Technology is denominated in dollars, representing increases in labor (as per hour labor dollars) and capital (as annual depreciation costs of capital investment). An increase in irrigation efficiency requires an investment in either (or both) capital or labor.
 Irrigation efficiency is defined at a water district level on the basis of an aggregation of field-level efficiency for various lands within the district. Specifically, the production isoquant in the model represents an aggregate ETc function for a water district as an envelope of functions reflecting field-level changes in either management practices or capital investment. These assumptions allow for the use of a continuous function to represent “production” of ETc and the IE rate implied by the function.
 A second component of the model accounts for land heterogeneity, measured as changes in average crop yield per acre. The underlying assumption is that farmers introduce land into production based on land quality. The highest-quality land comes into production first, followed by continually decreasing quality until the last unit of land placed into production has a marginal return of zero. Conversely, as land is forced from production by a reduction in other inputs, the least productive land is the first to exit. The result of this decrease (increase) in land in production is increasing (decreasing) average crop yield per acre.
 A functional form that possesses the above two properties for crop yield, input substitution and land heterogeneity, can be created be combining three common yield functions. First, a Leontief yield function (where the variables are acres of land and the land implied by choosing the yield maximizing ETc) is combined with a functional form that allows substitution between applied water and technology to “produce” ETc. This combined function is shown in equation (2a). Using the Leontief alone does not allow for substitution between inputs. However, allowing for substitution between land, applied water, and technology is not correct because the substitution of land for applied water in the production of a crop does not occur. The result of combining these two yield functions for one crop is presented in equations (2a)–(2c).
the minimum of hectares implied by ETc;
e, l, t, w
subscripts for the resources ETc, land, technology, and applied water, respectively;
coefficient of yield maximizing Etc per hectare;
input of the j resources, land, technology and applied water;
marginal rate of technical substitution.
 If γ = min xl, , then γ is the amount of land that can be irrigated with the allocations of technology and water. Given this specification, there is no need to have a specific land resource constraint (Leontief inequality) in the formal model because the land being irrigated (γ) is a function of technology (xt) and water (xw). Thus the applied water restriction constrains land in production. It follows that equation (2a) can be rewritten
Equations (2b) and (2c) define the diminishing marginal rate of technical substitution (MRS) between applied water and technology in the production of ETc. This analysis used a constant elasticity of substitution (CES) function to model ETc.
 The third common yield function used to reflect diminishing returns to land is the quadratic function. Equation (2d) shows how the CES function and the quadratic function, are combined to form a hybrid yield function with properties of each, i.e., substitution and diminishing yield:
where Y is total crop yield for one crop, α is average yield intercept parameter, γ is defined in equation (2d) above, and δ is the average yield slope parameter.
 Assuming costs are linear, the above functional form can be used to model a profit, maximizing farmers' decisions of whether to utilize water for production or sell water to a water bank. Specifically,
input cost per acre of the j resources;
the fixed price of water paid by the bank;
the quantity of water sold to the bank;
coefficient of per acre ETc requirement for land
technology (t), applied water (w) and land (l);
CES technology parameter;
CES share parameter;
function of the elasticity of substitution, σ (σ = );
total endowment of resource k.
 The CES function for ETc is represented by equation (3c), where ETc is defined in terms of applied water and technology. Equation (3d) constrains the number of acres in production. Equations (3e) and (3f) are resource constraints. Specifically, equation (3f) constrains the sum of water used in production xw and water sold in the bank s to be less than or equal to Xw. Equation (3g) is a constant returns to scale condition on the CES function.
 Examination of the first-order conditions reveals intuitive marginal relationships. Notice that equation (3f) binds with equality because any water that is not used in production will always be allocated to the water bank through the variables as an increase to the profit function. Let μ1 be the Lagrange multiplier on ETc, equation (3b), and let μ2 be the Lagrange multiplier on equation (3e).
 The first-order conditions for applied water and technology are shown below. Both equations contain the expression , which represents the cost of land, since land does not enter explicitly. The expression is the portion of land cost associated with the marginal product of the resources used for production of ETc. If the resource constraint binds, so that μ1 equals zero, equation (4) establishes the profit maximization quantity of water where the value of the marginal product of water (VMPw) equals not only its cost but also the water bank price W and the cost of land attributable to the marginal product of water.
The first-order condition for technology (equation (5)) is similar to the first-order condition for water, except there is no water bank price. Therefore the resource constraint may not bind, in which case μ2 may have a value greater than zero.
 Examination of the first-order conditions associated with this problem show the relationship between the land, applied water, and technology resources and ETc. Specifically, the optimal quantity of land in production is determined by equating the value of the marginal product of land (VMPl) to the price of land plus a term representing the value of ETc. This term is μRe, where the shadow value of ETc, μ1, is multiplied by the per acre requirement of ETc, Re. As a shadow value, μ1 represents the increase in the objective function that would result from an increase in ETc. Similarly, the first-order conditions generated by differentiating with respect to applied water and technology equate the respective input price to the shadow value of ETc multiplied by the respective input's marginal product. The first-order conditions of applied water and technology are analogous to the familiar first-order conditions that equate the value of the marginal product of an input to its price. Here the first-order conditions equate the shadow value of the marginal product of inputs to their prices.
 The parameters of the model are developed using positive mathematical programming [Howitt, 1995a, 1995b] and a cross section of data. The year 1995 was chosen from a data set ranging from 1980 and 1996 as the base case year because there were full agricultural water deliveries to the project as a result of above average snowpack and rainfall. Crop prices were obtained from California and Oregon County Agricultural Commissioner's annual reports. A weighted average of the 3 years preceding 1995 was used as a proxy for expected prices. Production costs were obtained from Oregon State University Extension Enterprise budgets. Costs of irrigation technology were obtained from Whittlesey . The elasticity of substitution used in the ETc CES function was verified using the Central Valley Improvement Act Draft Preliminary Environmental Impact Statement (volume 8, 1998). (Part of the verification process included personal conversations with Steve Hatchett of CH2M Hill.) Historic crop acreage was obtained from the BOR staff in the Klamath Falls, Oregon, office.
 The parameterized economic model is run assuming a reduction in full agricultural water allocation. The economic model's predicted IE rates are then used to calculate changes in return flow and resulting shortages of water supply to users dependent on return flow.
 Taking the total differential of the return flow (equation (1)) reveals the impact on return flow from (1) changes to inflow and (2) changes in irrigation efficiency. The differential is
 The first expression on the right hand side of equation (6) is the planned reduction in agricultural diversions, or inflow, I, times the portion of inflow that becomes return flow, (1-IE). This expression represents the policy maker's choice of the reduction in return flows. The second expression on the right-hand side of equation (6) is the change in return flow resulting exclusively from changes in irrigation water application. This resulting change, IΔIE, is more difficult for a policy maker to plan as it is generated from the profit-maximizing irrigation technology choices of the farmers, conditioned by the reduction of agricultural diversions.
8. Model Results and Implications
 The agricultural production model described above calculates the returns to various crops using cost data developed by University Extension reports under a range of prices and water supplies. This information can be used to estimate a producer's willingness to sell water (i.e., participate in the water bank). Since decreases in the amount of irrigation water available to an irrigator increases the value of the remaining water in crop production, it is expected that higher offer prices are needed to elicit additional water for the bank; that is, the supply curve is positively sloped.
Figure 1 shows two estimates of supply curves for water from project irrigators. The bottom curve is estimated using output from the agricultural production model. The model estimates returns associated with the acreage of both high- and low-valued crops grown in the project. The lower-valued crops include small grains, alfalfa hay, and irrigated pasture, while the higher-valued crops include such crops as onions, potatoes, and mint. The components of the water supply curve are fixed costs and variable returns associated with water applied to a profit-maximizing crop mix or sold to the water bank.
 Fixed costs include items such as property insurance and taxes that are paid regardless of whether the land is farmed. The changes in variable returns were estimated in the model to reflect farm production decisions to either utilize water in production or sell some portion of the water at a fixed price to the bank. The bottom curve in Figure 1 represents these two water supply function components.
 The top supply curve was constructed from the bottom curve by shifting it upward to include the price point observed in the 2003 Water Bank Program (of $75 per acre foot). This shift could be considered a “participation factor,” i.e., an incentive growers would need to change their behavior from the point of indifference between using or selling water, as reflected in the bottom curve. Using the estimated supply curve from the production model and the actual data point from the 2003 Water Bank Program allows us to draw the higher supply curve. The implied participation factor is $15 per acre-foot or ∼25–40% of the sum of the fixed costs and variable returns used to estimate the lower curve in Figure 1.
 The estimated supply curves shown in Figure 1 are relatively flat, meaning that a change in offer price produces a relatively large change in the quantity of water offered for sale. This flat shape is due to the fact that the value of the marginal product of water is relatively constant over lands within the project since for most districts the percent of high-value crops grown in the project is relatively small, as a percentage of total acreage.
 The implication of these flat supply curves is that per acre-foot water price may not provide the BOR with a particularly good tool for meeting a specific volumetric water banking goal since a small change in offer price results in a large change in the volume of water offered for sale. For example, the estimated supply curve suggests that 10,000 acre-feet of water could be obtained at ∼$50 per acre-foot, while the 2003 water bank goal of 30,000 acre-feet (from land idling) would be achieved by offering a price of ∼$73 per acre-foot. After that, however, substantial quantities of water could be obtained by small increases in price. This is what actually occurred in the water bank in 2003: an offer price of $75 per acre-foot resulted in 60,000 acre-feet of water offered for sale. Thus a $2 difference between the estimated offer price of $73 per acre-foot and the actual offer price of $75 per acre-foot resulted in a 30,000 acre-foot difference in the water offered for sale. When the BOR found itself in a situation where it had offers for more water than it was attempting to reallocate, it had to find an equitable method for determining which offers to accept.
 One possible longer-term solution to effectively use price as a discriminator for meeting a volumetric water goal would be to expand the water bank to areas outside the Klamath Project, where crop mixes are, on average, of lower value, primarily consisting of irrigated pasture. The effects of such a geographic expansion on crop mix are described by Jaeger . However, such an expansion cannot occur until water rights in the upper basin are adjudicated. Another possible solution to reallocate water to the environment would be to consider an intraproject market. The agricultural production model can be used to evaluate an intraproject water market that could be used in conjunction with a bank or as a “stand-alone” alternative. The benefits of such an intraproject market, relative to the current bank, are examined in section 9.
9. Mitigation With an Intraproject Water Market
 An intraproject water market could take several forms. The modeling results presented below assume that the BOR's involvement is to “rubber stamp” the trade and, where necessary, change the pattern of delivery. The analysis assumes that water is initially allocated to growers under the existing appropriative A/B/C water rights system (A has the highest priority, and C has the lowest). In water-short years the existing system would allocate irrigation deliveries first to the A users, then B users and finally C users.
 For an example of how the intraproject market would work, assume that an irrigation shortage resulted in B and C users receiving no water and A users receiving full supply. In such a case the intraproject market would allow B and C users to purchase water from an A user at a price that would be mutually beneficial to both buyer and seller. The price would be negotiated between the buyer and seller. The BOR would be involved only to the extent that a change in water releases was required.
 This intraproject water market is patterned after the existing water market in Westlands Water District in southern California. Westlands annual average shortage of irrigation water is 15% of water demand. For several years the district has assisted in intradistrict trades. The prices paid for these transfers are proprietary information between the buyer and seller.
Figure 2 shows how on-farm revenue is higher when growers trade among themselves as total irrigation water availability changes. When water deliveries are 100% of full year supplies, revenues are ∼1% higher when growers trade than when water is allocated using the A/B/C appropriation rights. As the percent of full water supply falls, so does revenue. However, the relationship is not linear, and the revenue under the trading allocation is higher than under the appropriative allocation. For example, if water availability is 50% of full supply, on-farm revenue is estimated to be slightly above 70% of normal when no trading occurs and 80% of normal with trading, i.e., 10% higher under intraproject trading. As noted above, if water trading could be extended to lands outside the project that are dominated by lower-valued crops, it is expected that the benefits would be substantially higher.
Figure 2 shows that revenue is nearly the same with trading versus no trading when water availability was 100% of full supply. On the surface, this seems to suggest that there is little benefit from trading when water is 100% of full supply. This is not the case, however. Table 2 shows the change in land use by crop and crop evapotransporation (ETc) under trading and no trading. While revenues are relatively unchanged (as shown in Figure 2), land in production falls by ∼8.2 thousand acres, and ETc falls by ∼22.6 thousand acre-feet.
Table 2. Changes in Land and Evapotransporation (ETc) Use Under Trading
 Land in production decreases since trading provides an incentive to maximize the returns on ETc across the entire project. Farmers increase the value of the ETc either by crop switching or choosing to vary their irrigation practices. Therefore the revenue per acre and revenue per acre-foot of ETc is greater than when there is no trading. The fact that trading can reduce water consumption but maintain project revenue is important for two reasons. First, reduced water use means that additional water is available for environmental uses. Second, by maintaining the revenue of agriculture producers at a higher level, the third-party effects on the community are reduced (although not eliminated), compared with a similar water allocation in the absence of trading.
 Water markets are viewed by many as a simple yet effective way of addressing water allocation problems, including the need to meet environmental demands. In some settings, such as California, water markets and banks have provided substantial quantities of water for environmental use and added needed flexibility to water allocation. A water bank in the Klamath Project, initiated in 2003, is being implemented for such purposes.
 The Klamath water bank highlights some of both the promise and the challenges of water banks. One problem of this bank for the management agency (BOR) is the similarity in marginal product of water values because of the relatively homogenous crop mixes within the project. This similarity makes it difficult to use “offer price” as a tool for achieving a specific reduction goal. The longer-term solution is to expand the bank to regions outside the project to take advantage of greater variability in crop returns and the subsequent variability in the value of water to agriculture. Until that is institutionally feasible, an intraproject market was shown to be a successful mitigation tool for reductions in water supplies, despite the similarities in the marginal values of water within the project.
 This work was funded in part by a USDA NRI grant, 96-35102-3363, and an Assistance Agreement with the Bureau of Reclamation, 99-FC-20-0052. The helpful comments of two anonymous reviewers and the Associate Editor are appreciated.