US Army Corps of Engineers
Walla Walla District

Acknowledgements

This research was supported by funding from the U.S. Army Corps of Engineers, and benefited from the numerous suggestions of Phil Benge, Dennis Wagner, Bill Hansen, and Brad Fowler. Phil Meyer of Meyer Resources, and Daniel Huppert of the University of Washington provided valuable comments. Bill Spencer, John McKean, and Hugh McKean of Agricultural Enterprises, Inc., provided many helpful suggestions and extensive logistical support on the focus groups, survey printing, mailing, and data entry. Opinions expressed in this report are those of the author.

Executive Summary

Need For The Study

Analytical Approaches

Recreation Use Value

Passive-Use Values

Sampling Strategy

Description of Likely Effects With Natural River

Using a Range of Salmon and Steelhead Figures to Deal With Uncertainty

Contingent Recreation Behavior Model for River Recreation

Focus Groups

Survey Pretesting

Mailing Procedures

Survey Response Rate

Descriptive Statistics

Contingent Behavior Modeling Results

Passive-Use Value Benefit Transfer Analysis

Conclusion

References

Appendix: Copy of Survey Instrument

EXECUTIVE SUMMARY

Recreation Use and Benefits

Using a survey of households in the Pacific Northwest and California (n=4,780 completed surveys and a response rate of 54% overall), the potential participation of anglers and other visitors that would occur if the four Lower Snake River dams were removed and the river is restored are estimated. Travel cost demand models were also estimated for anglers and non-anglers using intended trips to a free-flowing Lower Snake River to calculate benefits. For both user groups we estimated benefits based on reported trip cost and based on travel cost only, using a cost per mile consistent with the estimate for the existing recreation use of the reservoirs. A policy simulation of the relationship between number of angler trips and salmon catch rates is provided to identify the biological linkages between chinook salmon population estimates and catch rates that will be needed to apply our results in the Environmental Impact Statement.

Five estimates of angler demand and benefits are provided that range from a low estimate (using just households that indicated they would definitely visit with dam removal and assuming zero visitation from survey non-respondents) to an upper bound based on households that indicated they definitely or probably would visit and assuming that survey non-respondents would visit at the same rate as survey respondents. Middle demand estimates are provided by assuming that households that did not respond to the survey would visit at the same rate as households that did respond to the survey but applying this assumption only to the fraction of the population that were definitely sure they would visit. Lastly, an middle-high demand estimate is provided which uses the households that indicated they definitely or probably would visit but assuming that survey non-respondents would not visit.

The low estimate of angler demand is 1.3 million angler days that would increase to 2.7 million angler days over time when chinook salmon populations would increase to the point that salmon fishing would be allowed and fishing quality would be one salmon caught per 35 hours of fishing. With the salmon population increases sufficient to allow fishing, the Travel Cost Method demand model indicated angler benefits (at $76 per day) would increase from $98.3 million to $206 million. Small increases in angler days would occur with subsequent increases in salmon populations and salmon fishing quality to 20 hours per salmon caught (new total angler days would be 2.81 million angler days) and 10 hours per salmon caught (2.87 million angler days).

A low middle range estimate of angler demand uses only those that would definitely visit, but assumes some proportion (around 20%) of the survey non-respondents would visit. This is based on the observation that if we would have used a financial incentive some of the households that were non-respondents, would have been survey respondents reporting some positive number of trips. This low middle estimate yields 1.8 million angler days increasing to 3.85 angler days demanded if salmon recovery would allow for salmon fishing.

A middle range estimate of angler demand uses only those that would definitely visit, but assumes that all households would visit at the same rate as households that responded to the survey. Using this assumption estimated angler demand would increase from 2.78 million angler days to 5.95 million angler days over time when salmon fishing would be allowed. Corresponding fishing benefits would increase from $210 million to $452 million. Of course, the full amount of these use and benefit estimates will not occur for several years into the future, after sufficient salmon recovery to allow delisting and re-establishment of a recreational fishery. Even then, there would be a stair-step increase as salmon recovery of recreational fishing over time to the numbers estimated here. An approach for scaling the angler demand to salmon population levels, eventually to be provided by PATH, is discussed in the main body of the report.

The middle high and upper bound demand estimates are provided in the summary table below and the calculations provided in the main body of the report. The upper bound demand estimate is so large, that biologists should review it to determine if even a recovered chinook salmon fish stock could biologically support such a high level of fishing pressure.

The low, low-middle and middle estimate of angler use seems plausible, given there were about 40 million angler days of freshwater fishing reported by anglers from California, Idaho, Oregon, and Washington in the 1991 U.S. Fish and Wildlife Service's National Survey of Hunting, Fishing and Wildlife Associated Recreation.

The salmon fishing demand and benefits relationships estimated in this report could likely be applied to increases in salmon estimated by biologists to occur from improved transport and other dam improvements. However, our demand estimates may overestimate the number of salmon fishing trips and benefits associated with dam improvements, if anglers strongly prefer fishing for salmon in a free-flowing river (as described in the survey) as compared to reservoirs. Of course, the non-angling river recreation benefits would not occur without dam removal. However, the existing reservoir recreation estimated by McKean (1998b) would likely remain and could be used to estimate the non-fishing benefits of dam improvements.

These angler demand and benefit estimates can be compared to the current reservoir angler use and benefits estimated by McKean (1998a). In McKean's report he uses the Normandeau Associates fishing use estimates collected by David Bennett's group at the University of Idaho. This yields about 3,305 anglers, fishing 87,050 days. This yields annual reservoir angling benefits of $1.956 million. Thus current angler use and benefits are less than 10% of what is projected to occur if the dams are removed and the anadromous fishing improved.

Using the same ranges of assumptions about survey responses (assume or not assume non-respondents would visit at the same rate as respondents) and respondent certainty (definitely yes vs sum of definitely yes and probably yes) the same four range of river visitor demand and benefit estimates were calculated for non-fishing river visitors.

The low estimate for non-angler river demand is estimated at 1.688 million visitor days once the Lower Snake River is restored to conditions described in the survey. The travel cost method demand model estimates these river recreation benefits at $190 million annually once the river is restored. The low-middle estimate assuming some of the non-respondents would visit yields an estimated demand of 2.85 million days. The middle demand estimate (definitely yes visitors assuming non-respondents would visit at same rate as respondents) is 4.85 million visitor days with an economic value of $548 million annually. This visitor demand estimate and the high-middle and high estimate provided in this report need to be reviewed by recreation planners to determine if sufficient carrying capacity exists at the campgrounds and parking lots to accommodate these high estimates of visitor demand. If not, either demand estimates would need to be capped at facility carrying capacity or costs for expanding facilities included if the middle and higher visitor demand estimates are relied upon in the EIS.

Current non-angler use is estimated by using COE reported visitor hours of 9.2 million hours and average length of stay per day. Estimated visitor days consistent with McKean's (1998b) estimate of total non-angler recreation benefits of $31.578 million is 890,000 visitor days. Thus current non-angler use is about one-half to one-fifth of the projected visitation if the dams are removed, depending on the scenario.

The table above summarizes the recreation demand analysis. Detailed tables documenting angler and non-angler river demand calculations and disaggregated by the counties nearby the Lower Snake River, rest of Pacific Northwest and California are provided in tables 6-8. The reader is encouraged to read the detailed explanation of the procedures used in the report before jumping to conclusions about the results. In particular, these demand estimates and benefits will not be immediately realized following dam removal but will occurring gradually over several years.

Passive Use Values

Using a benefits-transfer approach the passive use value of various increases in salmon population are estimated. A passive use value function is provided that allows analysts to calculate the change in annual total passive use values with different levels of salmon recovery. Application of this function for an increase in chinook salmon populations from their current levels (around 10,000 returning adults) to levels necessary for removal from the Endangered Species list (around 32,500 returning adults) has a minimum estimated annual value of $56 million if the gain in Snake River chinook and sockeye salmon is considered equivalent to other wild salmon and the increment is calculated from estimated numbers of returning adult salmon in the Pacific Northwest. If Snake River chinook and sockeye are treated as they under the Endangered Species Act as distinct species, and the increment is calculated starting from current populations of these two species, then the incremental passive use value is $2.9 billion annually to households in the Pacific Northwest and California.

Also based on the existing literature there appears to be a passive use value of $420 million annually for returning the Lower Snake River to a free-flowing condition, independent of any effect on salmon populations.

NEED FOR THE STUDY

Four dams were built on the Lower Snake River between the Tri-Cities area at the mouth of the Columbia and Snake Rivers and Lewiston, Idaho from a period of 1960 to 1974. These dams provide hydroelectricity and allow Lewiston to essentially be an inland port for grain and lumber.

Despite fish ladders for ascending adult salmon and barge transport of smolts, populations of chinook and sockeye salmon have plummeted over the three decades since the first dam, Ice Harbor was completed in 1962. Populations of returning adult chinook salmon have fallen from around 100,000 fish during the early 1960's to around 9,000 today. In some years the number of sockeye salmon returning to Redfish Lake in Idaho have been in the single digits. Snake River sockeye salmon are listed as an Endangered species, while chinook salmon are listed as a Threatened Species under the Endangered Species Act.

National Marine Fisheries Service (NMFS) has requested that the U.S. Army Corps of Engineers (COE) evaluate removal of the four lower Snake River dams as part of NMFS recovery planning effort. Dam removal has both direct costs to remove the dams as well as large opportunity costs in terms of foregone hydropower and barging. Several groups of fisheries biologists have evaluated the likely gains from dam removal allowing for quicker migration of smolts downstream to the ocean. While there is some debate regarding the net effect of dam removal versus improved barging, dam removal is considered by federal and state agency biologists to have nearly twice the probability of recovering chinook salmon as dam improvements (Marmorek et al., 1998).

Being biologically effective, is not the same as being economically effective. If dam removal would result in quicker recovery of the chinook salmon runs in the Snake River, one question may be whether the added direct and opportunity cost of dam removal would warrant the greater eventual recovery. Further, a powerful political obstacle to dam removal in the region could request a "God Squad" exemption from the Endangered Species Act under the guise that recovery of the Lower Snake River stock of chinook salmon, is prohibitively costly. In addition, the COE is required under the National Environmental Policy Act of 1969 to prepare an Environmental Impact Statement that must address the social and economic consequences of such a significant Federal action.

ANALYTICAL APPROACHES

Recreation Use Value

Since natural river conditions do not exist in the Lower Snake River, one cannot survey existing users and directly apply standard methods such as the Travel Cost Model (TCM) to estimate the value of river recreation with dam removal. Since this is becoming a common problem, a hybrid approach known as "contingent behavior" (CB) has been developed. This approach involves: 1) describing the new recreation conditions (e.g., natural river scenario); 2) asking whether the person would visit and if so, how many times per year; and 3) asking the distance and travel time to their most likely spot on the river they would visit.

From this information, a travel cost model using expected trips as the dependent variable and round trip distance and time as independent variables is estimated. From this model, use and benefits can be calculated. To incorporate the sensitivity of expected trips to anadromous fish catch, we would also vary catch of salmon and steelhead in our contingent behavior scenario.

This approach has been used successfully for river rafting (Ward, 1988) and fishing (Layman, Boyce, and Criddle, 1996). Loomis (1993) has shown the method to be reliable in a test-rest study as well as valid. The validity test compared individual's actual visitation rates to contingent visitation rates for the same lake level.

Passive-Use Values

Resources that are unique and irreplaceable have long been known to provide existence and bequest values to citizens (Krutilla, 1967). Existence values are defined as the benefit received from simply knowing the resource exists even if no use is made of it. Free-flowing rivers, were one of the first examples of such resources with existence values (Krutilla and Fisher, 1975). Essentially people who never planned to visit, raft, or fish these rivers may still pay something to have a free flowing nature. Avoiding extinction of endangered species was quickly recognized as another source of existence and bequest values (Meyer, 1974; Randall and Stoll, 1983; Stoll and Johnson, 1984). Wild stocks of Snake River Sockeye and Chinook Salmon clearly fit into this picture. As noted by Olsen et al., in his existence value of salmon study "Existence values as the value an individual (or society) places on the knowledge that a resource exists in a certain state is theoretically sound and can be measured for assessment within the resource decision making arena." Non-use value are also public goods, in that these benefits can be simultaneously enjoyed by millions of people all across the region and the country (Loomis, 1996a).

At present the only method generally recognized as being capable of measuring existence values is the Contingent Valuation Method (CVM). This method involves describing the resource to be valued, including the consequences of paying or not paying. The individual is then asked to indicate how they would vote if it cost their household a particular dollar amount. Since different groups of households are asked to pay different dollar amounts, the analyst can estimate a demand like relationship and calculate mean WTP. This referendum approach is recommended by NOAA's Blue Ribbon Panel on CVM (Arrow et al.). Using CVM to measure passive use value has been controversial (Arrow et al.). CVM has proven to be reliable in measuring passive use values in several test-retest studies (Loomis, 1990; Carson et al.). The criterion validity of CVM to elicit passive use values is less encouraging, with stated WTP exceeding cash WTP, except when stated WTP is rescaled by respondent certainty (Champ et al., 1997). However, it is difficult to develop a cash instrument for a public good that does not encourage free-riding, making comparisons with cash and stated WTP problematic (Randall, 1998).

Incorporating non-use values has become fairly routine in many Federal benefit-cost analyses for critical habitat designations of endangered species. Examples include benefit-cost analyses for the Northern Spotted Owl, Mexican Spotted Owl, and Wolf Recovery program. The USDA Economic Research Service's economic analysis of salmon recovery efforts on the Snake River included estimates of non-use values drawn from the existing literature (Aillery et al., 1996).

While the possibility exists that constructed objects such as dams or development may have existence value, economic theory and empirical evidence to date suggests this is likely to be small. As noted above, scarcity and uniqueness are major determinants of the size of non-use values. Dams and reservoirs are not scarce on rivers in the Pacific Northwest. Most of the value of development such as dams or barge transport comes from the market outputs created or the non-market recreation use values. Most public support for the dams can likely be traced to commodities produced or the recreation benefits provided by reservoir. These are being measured as part of the overall economic analysis and therefore are already reflected in the opportunity cost estimates. The other source of public support for projects is due to the local economic activity supported by the projects. While very important at the local level, the economic activity will be relocated elsewhere in the U.S. if the dams are removed. Thus from a benefit-cost standpoint, we cannot count jobs and local incomes that would be lost in the region as these will be gained in other regions of the U.S.

At the request of the COE, non-use values will be approximated based on existing non-use value estimates (e.g., using a benefit-transfer approach) rather than an original study.

SAMPLING STRATEGY

The sampling strategy is driven by the primary objective of this study, to estimate recreation use and benefits with natural river conditions. Efficient sampling requires concentrating sampling effort in the geographic region where a majority of the potential visitors may come from. This is a challenge here, since natural river conditions do not exist. In addition, there are no established lists of rafters, hikers, etc., like there are for anglers. Since recreation use is driven largely by distance to the recreation site and site attributes, it is difficult to use data on other rivers. Thus, to determine how broad a geographic area to concentrate on we relied upon two surveys of current users of the Lower Snake River projects. The first was conducted in 1985 by the Walla Walla District of the U.S. Army Corps of Engineers. Those surveys found that between 80% and 89% of use came from within 100 miles of the sites. The second survey was conducted during the summer of 1998 by AEI/Normandeau Associates. Using just the zip codes of visitors contacted at boat ramps, we found that half the water dependent recreation came from seven cities in the area (Tri-Cities of Kennewick, Richland and Pasco along with Clarkston, Lewiston, Pullman and Walla Walla). Spokane and Yakima were the next largest contributors of visitation. Therefore the main strata for contingent behavior surveying is designated "local counties" made up of:

Idaho Counties (Benewah, Latah, Lewis, Nez Perce, and Idaho County);

Oregon Counties (Morrow, Umatilla, and Wallowa); and

Washington Counties (Adams, Asotin, Benton, Columbia, Garfield, Franklin, Spokane, Walla Walla, Whitman, and Yakima).

At the direction of the review team this was supplemented by surveys of households in the rest of Idaho, Oregon, Washington and all of California and Montana. This second strata was sampled more lightly. Nonetheless, this strata is important since the upriver fishing survey by McKean (1998c) indicated that fishing for steelhead on free-flowing sections of the Snake River attracted anglers from great distances.

The sample size was based on standard statistical criteria such as the variance of the population, allowable error and desired confidence level. We obtained estimates of the variance and the prevalence of river recreation activities such as rafting and canoeing in the population from the existing literature. To provide 90% confidence that our sample estimated trips per person would be within 10% of the population mean for river-based activities would require a sample size of 8,600 surveys assuming a 50% response rate of deliverable surveys. Given higher than average non-deliverables associated with rural addresses, we recommended this sample size be increased to 9,000. The 9,000 surveys are allocated as follows: 1) 6,000 to the 18 counties within 150 miles of the Lower Snake River based on population; and 2) 3,000 to the rest of Idaho, Oregon and Washington, and all of California and Montana.

DESCRIPTION OF LIKELY EFFECTS WITH NATURAL RIVER

Based on discussions with COE staff as well as information in a wide variety of documents, we sketched a description of the likely effects of natural river conditions. Figure 1 below presents what the key contingent behavior information presented to the respondents was in the final mail survey.

Using A Range of Salmon and Steelhead Figures to Deal With Uncertainty Regarding Anadromous Fish Response

Since we were unable to obtain specific information from the Federal agency biologists on likely changes in salmon populations or salmon/steelhead fish catch rates with dam removal or improved dams, the XX, HH, and DD on the previous page represents different catch rates and season lengths in the different survey versions. These changes in fish catch are used as variables in the travel cost method demand equation. Using information on existing catch data from University of Idaho, current catch rates of steelhead at the reservoir were set as the baseline catch rate (one fish every 24 hours of fishing). Steelhead catch rates in the free-flowing section were used as guidance for dam removal catch rates. Information on catch rates from Idaho Fish and Game's 1997 special season hatchery salmon fishery and information provided by Bert Bowler of Idaho Fish and Game based on the chinook salmon fishery in the Hanford Reach of the Columbia River were used to set the range of catch rates for chinook salmon. Catch rates were expressed as 1 fish per X hours of fishing, where X varied with salmon population levels used in that survey version. While not a perfect measure, discussions with Idaho Fish and Game as well as University of Idaho biologists studying steelhead fishing, suggested this is the way fishing success rate data is frequently reported in the local newspapers. Therefore, anglers were used to judging the quality of fishing in terms of the hours it takes to harvest one steelhead. We used the same measure for salmon fishing and supplemented it with an associated season length to further define the relative scarcity or abundance of salmon. The range includes a low end which involves small increases in steelhead fishing but no salmon fishing to an upper end of fully fishable estimate. To reduce multicolinearity between steelhead and salmon catch rates in order to estimate separate coefficients, the 16 different versions were mixed in an off-setting pattern.

CONTINGENT RECREATION BEHAVIOR MODEL FOR RIVER RECREATION

For any one indicating that salmon or steelhead fishing would be an activity, we statistically estimated the probability of visiting and the number of visits as a function of travel cost to the site, expected hours to catch a salmon and steelhead as well as respondent demographics. We initially estimated a two-stage model of the form:

1. Probability of visitation = f(Milesi, HoursSteelhead, HoursSalmon, Demographicsi); and

2. Number of tripsi = f(TCi, TTi, HoursSteelhead, HoursSalmon, Income, Recreation Days, SubstituteCost, Demographicsi)

   Where:

   Miles - mileage from respondents place of residence to the area on the Lower Snake River they would most likely visit.

   TCi - is round trip cost of respondent. This was measured as reported trip cost of respondents or for consistency with McKean's (1998a,b) of reservoir recreation, as travel costs.

   TTi - is travel time from respondent's home to the area on the Lower Snake River they would most likely visit.

   HoursSteelhead and HoursSalmon - are the average number of hours it takes to catch steelhead and salmon. This number varied from respondent to respondent in the survey to allow estimation of a coefficient on fish catch. There were eight different levels of hours to catch a steelhead and six different levels for salmon (essentially two salmon versions with zero catch). These variables are key policy variables that can be changed by feasibility study alternative.

   Income: The household income is tested to see if the income constraint or money budget influences the number of trips.

   Recreation Days: The number of days available for recreation is tested to see if the time constraint or time budget influences the number of trips (see Bockstael et al.; McKean et al., 1995, for the theoretical consistency with disequilibrium labor markets).

   Substitute River Recreation Cost: For those respondents that currently visit other rivers, the travel costs to those rivers.

   Demographics: This would include boat ownership, number of recreation activities they would participate in, boat ownership, age, etc.

This model formulation allows for river restoration to:

1. induce non-visiting anglers to the area to begin to visit for anadromous fishing or other river recreation. This is modeled in equation (1), using a Probit model since the dependent variable is zero (no visit) or one (visit); and

2. anglers to increase the number of trips in response to better anadromous fishing quality or riverine conditions. This is the Travel Cost Method demand equation specified in (2).

For those respondents who indicate they would not fish, the same basic two step Travel Cost Model will be tested without including the Steelhead and Salmon variables.

Since the number of trips taken is a non-negative integer, a count data model is an appropriate statistical model to estimate the Travel Cost Method (TCM) demand equation (Creel and Loomis, 1990; McKean, 1998). If the mean of the dependent variable is not equal to the variance, then overdispersion is present. This condition requires use of a negative binomial count data formulation rather than the more standard Poisson model (Creel and Loomis, 1990). The negative binomial count data TCM is the same approach used by McKean (1998) to estimate the economic value of reservoir recreation at the existing Lower Snake River dams. Thus there is consistency in valuation methodology between non-market benefits being lost and gained.

FOCUS GROUPS

To check for understandability and internal validity of the survey, portions of the survey were made into one page worksheets that were distributed at the focus groups. The individual was then asked to review the description of what the river would be like with dam removal. They were asked to point out any words that were not clear. They were asked to identify any additional information that they felt was needed before they could make an informed decision about whether to visit the restored Lower Snake River. The responses on the written worksheets were supplemented by discussions with the focus group. These were recorded on flip charts. After each focus group, changes were made to materials for the following focus group. Focus groups were conducted during the month of January 1998, in the cities of Seattle, Kennewick, Boise, Spokane, Lewiston, and San Jose, California. There were 9 to 11 people at each focus group.

SURVEY PRETESTING

The revised survey was printed as a booklet with color cover and color recreation map insert. The pretest involved mailing five versions of this booklet. The household was then phoned and their answers read back to interviewers. In addition, a set of check questions were included to probe for lack of clarity on any informational materials or specific survey questions. Individuals were then asked to mail their booklets back. A total of 45 surveys were completed over the months of February and March. Individuals interviews conducted over the phone gave us several useful insights for revising the survey. The results of the pretest also aided in refining the design of the bid amounts to be used in the survey.

Numerous revisions to the survey were made during the months of April and May. At the end of July a decision was made to go forward with a survey that would only address recreation use with and without the dams. Other than omission of the non-use values, the revised recreation portions were only slightly different than the previous version that was pretested. Nonetheless, to insure that the changed context and wording changes made during the intervening four months had not altered the internal validity of the survey questions, the survey was again pretested. This time, 11 pretests were conducted in-person in Spokane, Washington, on August 28. In addition, 100 surveys were priority mailed to Snake River recreation users to solicit their feedback. A total of 36 the 100 surveys were returned completed by the cut-off date for survey printing, with an additional five being returned undeliverable. A complete copy of the 8-page survey instrument and map is found in the appendix. We were often restricted from using questions that current economic theory suggested would be important, because such questions had not been included in the previously approved set of questions from OMB.

MAILING PROCEDURES

The 10,000 surveys were not sent out until October 9, 1998. The survey package was sent first class and included a personalized cover letter, stamped return envelope and the survey with recreation map insert. A reminder postcard was sent to all households. The second mailing was sent out four weeks later with a new more emphatic cover letter that attempted to address any concerns individuals might have for not returning the survey and stressing the importance of completing the survey. A replacement survey was included. A third mailing was sent by U.S. Post Office Priority Mail with a cover letter stressing the importance of the survey and urging them to complete it. This survey was sent out December 10th. This is not an ideal time to survey due to the holidays.

SURVEY RESPONSE RATE

The three mailings of the survey resulted in 4,780 surveys being returned by respondents and another 1073 being undeliverable surveys sent back by the Post Office. There were also 83 survey packages returned by spouses or relatives indicating the addressee was deceased. There were 108 people who returned the survey package, refusing to answer it and six people moved out of the sampling area. Thus there were a net deliverable surveys of 8838 (10,000 minus 1162) yielding a response rate of 54% This is fairly good given the requirement not to use any financial incentive such as a $1 bill and the timing of the third mailing around the holidays. Many general household surveys without financial incentives are often in the mid to low 40% range. Household surveys with financial incentives can attain responses rates of local populations in the range of 68% to 77% and 55% for U.S. households (Loomis, 1996). Our response rate in this survey varied from a low of 28% in California to 56% for areas around the Lower Snake River such as Spokane, Lewiston and the Tri-Cities. To calculate the low-middle estimate we assumed this higher response rate of 68% response rate would have resulted in some Pacific Northwest non-respondents to our survey, having been respondents, and therefore reported some positive number of trips. For California, we used the 55% response rate in the same way to estimate the low-middle estimate of recreation demand.

DESCRIPTIVE STATISTICS

Table 1 summarizes respondents basic attitudes toward the relative importance of different recreational uses of the Lower Snake River and whether they current visit the Lower Snake River.

Table 2 summarizes the probability that individuals from each of the sampling areas would visit the Lower Snake River if the dams were removed.

Comparing the responses in Table 1 and 2 with regard to actual versus intended visitation is reassuring. With exception of those in the Local Counties, about the same percentage of people that currently visit from each area, would definitely visit if the dams were removed. Thus the definitely visit percentages appear to be grounded in reality. If the dams are removed, another 15%-28% would likely visit the Lower Snake River. Adding together the Definitely and Probably Yes responses yields a percentage visitation from Local Counties about equal to their current visitation. Again, this is reassuring regarding the realism of the intended visitation percentages.

Table 3 presents information on average recreation expenditures of people that would visit the Lower Snake River if the dams were removed.

The pattern of expenditures appear reasonable, as more distant users have higher expenses per trip and stay longer per trip, than nearby users, who can make more frequent weekend trips. Anglers from the Pacific Northwest tend to stay longer than other river users. The cost per day per person is also quite reasonable given that it includes travel costs and any boat gas, camping fees or lodging.

The cost per angler trip averaged over all four areas of residence is $197.30, averaging $21.43 per angler per day of the trip. The cost per non-angler river visitor averaged over all four areas of residence is $148.60 which yields $18.40 per day.

The overall average household income reported in the survey was $55,400. This compares the 1990 Census estimate (updated to 1996 dollars) of $52,526. Thus, the survey incomes are within 10%, and part of the difference may be due to the survey being in 1998 dollars.

CONTINGENT BEHAVIOR MODELLING RESULTS

Anglers

Probability of Visiting the Free-Flowing Lower Snake River

Respondents were asked in the survey whether they would visit the Lower Snake River if the dams were removed and the river restored to conditions described in their version of the survey (e.g., number of hours to catch a steelhead and salmon). Individuals could respond with definitely yes, probably yes, probably no and definitely no.

We estimated a probit model on the definitely yes responses and the sum of the definitely yes plus probably yes responses as a function of their driving distance to the Lower Snake River, whether they owned a boat, number of river based recreation activities they would participate in beside fishing at the Lower Snake River and the importance of river based recreation (from Section I of the survey). These variables were statistically significant and had a sign consistent with economic theory. However, hours to catch a salmon and hours to catch a steelhead had little systematic influence on the probability of visiting the Lower Snake River in either the definitely yes probit model or the summed definitely yes/probably yes probit model. Therefore it appears that steelhead or salmon fishing quality is not a strong predictor of the probability a household would visit the Lower Snake River if the dams were removed. Therefore, we do not present the probit equations, as we simply use the point estimate for that region and user group of the probability of visiting the Lower Snake River in the analysis below. By using separate visitation rates for the three main geographic regions, we have utilized a spatially varying visitation rate, similar to what would have occurred had the probit models been applied, but with much less complications.

To facilitate sensitivity analysis, we calculate five different estimates of the number of anglers that would visit. These range from a low estimate, a low-middle, two middle estimates and one high estimate. The low estimate uses only the definitely yes responses to calculate the probability (or percentage of that region's respective population) visiting the Lower Snake River and then adjusts downward for the survey response rate from that region. The use of definitely yes responses is consistent with a criterion validity study of Champ et al. (1996), which showed a close match between actual behavior and intended behavior of the persons that were definitely sure of their answers. By itself the visitation rate associated with the definitely yes will be one of our middle estimates. However, to make this the low estimate in the table below, we adjusted downward by the mailing survey response rate. In effect, we are assuming those with insufficient interest to return the survey, would not visit the Lower Snake River if the dams were removed. The low-middle estimates incorporates two insights: 1) some of the non-respondents to the survey might actually visit the river if the dams were removed; and 2) with a higher response rate that would have occurred with the use of financial incentives, some non-respondents would have become respondents and reported some positive visitation. For the low-middle estimate we continue to use the definitely yes responses, but allow for some of the non-respondents to visit, but using a higher survey response rate adjustment of 68% for the Pacific Northwest and 55% for California. The other middle estimate uses the combined proportion definitely and probably yes responses adjusted for the sample response rate from that region. The inclusion of probably yes responses recognizes that some of those anglers will likely visit the Lower Snake River each year. The high estimate uses the combined definitely and probably yes and does not adjust for the survey response rate. Again, we allow for the possibility that some people not returning the survey, would still visit the Lower Snake River if the dams were removed.

Angler Travel Cost Method Demand Model

Given the probit model did not find a systematic effect between probability of participation and anadromous fishing quality, there was not much advantage in linking the probit model to the travel cost model. Avoiding the link greatly simplifies the application of the Travel Cost Model to evaluating EIS alternatives. As explained in more detail in McKean's analysis of reservoir recreation (McKean, 1998a, 1998b) and in Loomis and Creel (1990) if the mean of the dependent variable does not equal the variance, a negative binomial count data model rather than a Poisson count data model is appropriate. The negative binomial count data model is more general and includes the Poisson count data model as a special case.

Two negative binomial count data models are reported in Table 4. The first model uses the reported trip cost per person (RTRCPP) from our survey. To allow greater comparability of benefit estimates with McKean's (1998a) estimate of reservoir angling we estimated a second model using the same cost per mile as McKean (1998a, 19 cents).

In both angler TCM models, the crucial cost coefficient is negative and highly significant (p >.01 for 1% level). Travel time (RIVTRTIME) is negative in both models but statistically significant only in the survey based trip cost model. The lack of significance of travel time in the second model is likely due to multicolinearity caused by scaling driving distance by a single cost per mile. The total days available for recreation (RECDAYS) is positive and statistically significant at the .01 level indicating those with more days available for recreation would take more trips to the Lower Snake River. Age has a negative and statistically significant influence on the number of trips to the free-flowing river. Income was insignificant in both models, suggesting income is not a determining factor in the number of visits an angler would make to the Lower Snake River.

In terms of the key policy variable, salmon fishing quality, hours to catch a salmon (HRSAL) has the correct sign, negative and is significant at the 5% level (p=.0118 and p=.02 in models 1 and 2). Negative is the correct sign since the longer it takes to catch a salmon, the fewer trips anglers will take. Thus, if salmon populations rise with dam removal, this will reduce the number of hours it takes to catch a salmon and will increase the number of trips taken by anglers. For example, with only steelhead fishing and no salmon fishing, the model predicts a typical anadromous fishing angler would take 2.05 trips each year. Once sufficient salmon recovery occurs that salmon fishing could begin with an estimated 35 hours to catch one salmon, then fishing trips per angler would double to 4.13 trips per year. Once salmon fishing is introduced, additional improvements in salmon catch rates have smaller effects on trips (as would be expected with diminishing marginal utility). Increasing the salmon populations sufficiently to reduce hours to catch a salmon from 35 hours to 20 hours per fish, would increase trips by .13 per angler. An increase in salmon populations in the future to reduce hours to 10 hours per salmon caught yields an increase of .3 trips per angler more. These incremental effects will be used to illustrate increases in angler trips associated with a range of anadromous fish increases ranging from small near-term effects to larger long-term effects.

These salmon fishing responses could likely be applied to increases in salmon from improved transport and other dam improvements. Our demand estimates may over-estimate the number of salmon fishing trips and benefits associated with dam improvements, if anglers strongly prefer fishing for salmon in a free-flowing river (as described in the survey) as compared to reservoirs, however.

The TCM model based on respondent's reported cost is robust with respect to checks for potential multi-destination trips, potential outliers and inclusion of a substitute variable. In particular, including the price of (cost of travelling to) substitute river recreation had little effect on the demand function and benefit estimate. This is part was due to cost of travelling to other river sites having a negative sign, indicating other rivers have a complementary relationship with the Lower Snake River, rather than a positive sign indicating a substitute relationship. A dummy variable for whether the angler visited other rivers was insignificant and made no change in trip benefits. We also screened trips for one-way travel distances in excess of 1,000 miles as these long trips could end up being multidestination. Dropping those nine observations had no effect on average consumer surplus.

Angler Benefit Estimates

The net willingness to pay (WTP) or consumer surplus for a salmon/steelhead fishing trip to the free-flowing Lower Snake River is calculated from the count data regression equations in Table 4 based on the trip cost coefficient. Since the trip cost coefficient is per person, the resulting value per trip is $256 per person. These trips average 3.36 days in length, yielding a value of $76 per day. As mentioned above this value per day is quite robust to issues of potential multidestination trips or possible outliers. If one uses just the transportation costs at the cost per mile of reservoir users that was applied by McKean in his reservoir recreation travel cost model (19 cents per mile), the value per person per trip is $131.80 with a per day value of $39. This will be referred to as the Low Recreation Value. Generally speaking, the Low Recreation Value travel cost demand model was not statistically significant on the travel time variable but otherwise was quite similar, particularly with respect to the coefficient on hours to catch a salmon.

These values compare to a value of $88 per trip from McKean's (1998c) estimate for steelhead fishing on the Snake River upriver from Lewiston, Idaho. Dividing McKean's (1998c) reported angler days per year (14.82) by reported annual number of trips per year (12.33) yields 1.2 days per trip as an average. Therefore, the equivalent value per day is $73 per day, nearly identical to our estimate of $76 per day. Our estimate also compares to $110 estimate (updated for inflation) from Olsen et al. (1991) for steelhead fishing in the Columbia River. On a per day basis this is about $55 per day. This value is reported as the value for the last trip, so given the downward slope of the demand curve, this would be an estimate less than our average value over the season. Walsh et al. (1992), report an average (adjusted for inflation) of $75 per day for values of steelhead/salmon fishing in Idaho, Washington and Oregon from 1967 to 1982. While these older benefit estimates suffer from methodological limitations associated with travel cost models from that time period, the $76 per day estimate based on respondents reported cost is quite similar to these past studies, while the $39 per day estimate based 19 cent per mile is substantially below estimates in the literature.

Non-Angling Visitors

Respondents were asked in the survey whether they would visit the Lower Snake River if the dams were removed and the river restored to conditions described in the survey. To avoid additional complexity (there were already numerous survey versions due to differences in fishing conditions) none of the non-fishing, river-based attributes were varied. Given the description of the restored river, individuals could indicate whether they would definitely visit, probably visit, probably not visit or definitely not visit.

To facilitate sensitivity analysis, we calculate five different estimates of the number of non-fishing river recreationists that would visit. These range from a low estimate, two middle estimates and one high estimate. The low estimate uses only the definitely yes responses to calculate the probability (or percentage of that region's respective population) visiting the Lower Snake River and then adjusts downward for the survey response rate from that region. The use of definitely yes responses is consistent with a criterion validity study of Champ et al. (1996), which showed a close match between actual behavior and intended behavior of the persons that were definitely sure of their answers. By itself the visitation rate associated with the definitely yes will be one of our middle estimates. However, to make this a lower bound in the table below, we adjusted downward by the mailing survey response rate. In effect, we are assuming those with insufficient interest to return the survey, would not visit the Lower Snake River if the dams were removed. As with anglers, the low-middle estimates incorporates the observation that some of the non-respondents to the survey might actually visit the river if the dams were removed and a higher response rate that would have occurred with the use of financial incentives, some non-respondents would have become respondents and reported some positive visitation. For the low-middle estimate we continue to use the definitely yes responses, but allow for some of the non-respondents to visit, but using a higher survey response rate adjustment. The other middle estimate uses the combined proportion definitely and probably yes responses adjusted for the sample response rate from that region. The inclusion of probably yes responses recognizes that some of those households will likely visit the Lower Snake River each year. The high estimate uses the combined definitely and probably yes and does not adjust for the survey response rate. Again, we allow for the possibility that some people not returning the survey, would still visit the Lower Snake River if the dams were removed. This is the same procedure used for anglers.

Non-Angler Travel Cost Method Demand Model

Table 5 presents the negative binomial count data models for non-fishing visits to the free-flowing Lower Snake River. The reported trip cost per person (RTRCPP) is negative and significant at the .01 level. Travel time (RIVTRTIME) is also negative and statistically significant at .01 level. The total days available for recreation (RECDAYS) is positive and statistically significant at the .01 level. Income is negative and significant at the 5% level. The number of recreation activities the respondent would participate in has a positive and statistically significant effect on the number of trips the individual would take. Those that own a boat also would take more trips. We attempted to use a cost per mile consistent with McKean's (1998b) estimate for general recreation at the Lower Snake River reservoirs (20.2 cents per mile). However, transformation of round trip distance into travel costs using this mileage figure resulted in the travel cost coefficient being insignificant at conventional levels (p=.2). Since the benefit estimate depends directly on this coefficient, this was unacceptable. An alternative approach to calculate travel cost consistent with McKean's (1998b) model was used. In this alternative approach, we relied upon his estimate of the percentage of total trip cost that was spent just on car transportation (e.g., spending at service stations). This percentage was multiplied by our reported total trip cost. This estimated coefficient was statistically significant at the .05 level, and the remaining variables had the same sign and similar significance as in the reported cost model.

Non-Angler Benefit Estimates

Using the TCM demand equation estimated from respondent reported cost in Table 5, the net WTP or average consumer surplus is $297 per individual trip. Given the 2.62 day duration of trips, the average value per day would be $113. The travel cost method estimate based on McKean's estimate transportation cost spending is $71.36 per trip or $27.23 per day. These estimates compare to McKean's (1998a) travel cost method estimate of $71.31 per trip or $29-35 per day for general reservoir recreation on the Lower Snake River reservoirs. There does not appear to be similar river-based recreation studies to what the Lower Snake River will be like. Thus comparing our values is difficult.

Application of Models to Estimate Total Visitation To and Benefits From a Free-Flowing Lower Snake River

Since the rural counties surrounding the Lower Snake River were sampled at a higher rate than the remainder of the states in the Pacific Northwest and they are located closer to the Snake River, it is appropriate to apply their probability of visiting to their respective populations. Further, since anglers are likely to have a different propensity to visit a free-flowing Lower Snake River than non-anglers, it is appropriate to keep these groups separate. Further, California's population is so large and its response rate was much lower than for the Pacific Northwest, it was decided to split this geographic area out separately. Thus, to calculate the number of potential visitors we essentially have a six components that go into any estimate of overall recreation use and benefits. To estimate angler and non-angler recreation use, households in a region were split into these two mutually exclusive groups based on the USFWS National Survey of Fishing, Hunting and Wildlife Associated Recreation. This is a conservative approach, which assumes that angler households do not make non-angling recreation trips. We repeat all calculations five times, using a different assumption about probability of visitation (i.e., using only households definitely certain they would visit or using households that are definitely and probably certain they would visit) and whether those not returning the survey would visit or not. Further, we compute recreation benefits using the TCM demand model based on reported costs as well as the travel cost only estimate for consistency with McKean (1998a,b). Nonetheless, these variety of estimates provides a broad basis for a sensitivity analysis as part of the NED or benefit-cost analysis in the EIS.

River Recreation Use Estimates

The upper portion of Table 6 estimates river recreation use relying upon only those households that indicated they would definitely visit the Lower Snake River if the dams were removed. About 7% of residents of the Idaho, Oregon and Washington counties adjacent to or nearby the Lower Snake River would definitely visit the free-flowing Lower Snake River for river recreation. About 4% of those living in the more distant urban areas of the Pacific Northwest would definitely visit the free-flowing Lower Snake River. Only about 2% of California residents would definitely visit the free-flowing Lower Snake River. Visitors from the counties surrounding the river would take an average of four visits per year while those living in the remainder of the Pacific Northwest would take an average of 1.46 trips per year. Those in California would take about one trip per year. Multiplying the percentage of the population that would definitely visit, times the number of households in the region times the number of trips per household and days per trip, yields the column labeled Total Days, for the Middle Estimate or 4.8 million visitor days annually. This assumes that households that did not return the survey would visit at the same rate as those that did return the survey. (Recreation planners should verify that there is or would be sufficient recreation carrying capacity and facilities to accommodate 4.8 million visitor days annually). Two different NED benefit estimates are reported. The column labeled NED benefits uses the value per day from the Travel Cost Method (TCM) demand model estimated using survey respondents reported trip costs as the price variable ($113 per day), while the Low NED Benefits ($27.23 per day) uses the TCM demand model reflecting just the proportion of trip costs attributable to travel to be consistent with McKean's (1998b) TCM of reservoir recreation use. Using the two estimates the annual non-angling river recreation use benefits range from $132 million to $548.4 million.

To calculate the row labeled Low Estimate, we assume that those not returning the survey would not visit and hence take zero trips. Thus, we multiple the number of households in the row labeled Sample Est by the response rate for Rural ID,OR, WA (56%), the Pacific Northwest (51%) and for California (28%) to yield the low estimate of number of households. All other calculations in this row are otherwise similar. Using the row labeled Low Estimate, there would be about 1.688 million visitor days. This has an economic value of $190.8 million using the higher consumer surplus estimate per day and $45.9 million for the lower consumer surplus estimate, respectively, each year after the Lower Snake River is restored to the free-flowing condition described in the survey. We calculate the Low-Middle Estimate in Table 6 assuming that some proportion of the non-responding households would visit the lower Snake River if the dams were removed. We calculated this proportion using a higher response rate which would have made some of the current non-respondents into respondents that would have reported some positive visitation. Thus the Middle Estimate in the Pacific Northwest and California are multiplied by 68% and 55%, respectively, to arrive at the Low-Middle estimate.

The lower portion of Table 6 reflects our Middle-high and High estimates that use visitors that would definitely and probably visit the Lower Snake River if the dams were removed. The High estimate assumes the proportion of households that did not respond to the survey would visit at the same rate as households that did respond to the survey. The Middle-high assumes that households that did not respond to the survey would not visit. The Middle-high estimate of visitor days is 9 million, while the High estimate is 28 million visitor days. Clearly, recreation planners should verify that there is or would be sufficient recreation carrying capacity and facilities to accommodate 9 or 28 million visitor days annually before such estimates should be used in the benefit-cost or EIS sensitivity analysis. It may be that the High estimate needs to be capped or limited to recreation area carrying capacity or the cost of building additional facilities included as a cost in the benefit-cost analysis.

Estimated visitor days consistent with McKean (1998b) annual benefits of $31.6 million is 890,000 visitor days. Thus current non-angler use is about one-third of the projected visitation if the dams are removed.

Angler Recreation Use Estimates

Table 7 provides Low, Low-Middle and Middle estimates of angler use and benefits, while Table 8 provides Middle-High and High estimates of angler use and benefits. Since salmon fishing is a significant variable in the Travel Cost Method demand equation, we calculate how angler trips changes when salmon populations increase sufficiently to allow salmon fishing, and then with an improvement in salmon fishing catch rates (i.e., less time to harvest one salmon).

As shown in Table 7, about 20% of angler households living in counties surrounding the Lower Snake River would definitely fish at the free-flowing Lower Snake River. About 13% of anglers living in the remainder of the Pacific Northwest would definitely visit while 4% of California anglers would visit the Lower Snake River to fish if the dams were removed. Assuming households not responding to the survey would not visit, yields the full adjustment for response rate estimate of the number of anglers for each of the geographic areas. The next to last right hand column totals this up across the three geographic areas and yields a low estimate of 99,809 anglers would visit the Lower Snake River if the dams were removed. The survey data indicates that residents of the surrounding counties would take 6.12 trips per year, while rest of Pacific Northwest residents would take 1.41 trips per year with a free-flowing Lower Snake River and steelhead fishing but no salmon fishing. Combining the number of anglers, number of trips times the average length of stay, yields about 1.29 million angler days if the low response rate adjustment is used. Using the estimate of average consumer surplus per day from the angler Travel Cost Method demand equation ($76 per day) yields an annual economic value of $98.3 million with the low estimate of angler use for just steelhead fishing.

Since hours to catch a salmon had the correct sign and was statistically significant, the remainder of the table simulates how the addition of salmon fishing and increased salmon fishing catch rates can be used with biologists' estimates of changes in salmon populations to estimate the incremental benefits over time as the fisheries increase with habitat restoration. The first scenario involves the benefits occurring when chinook salmon populations consistently surpass the level where the Snake River chinook salmon would be delisted (estimated at 32,500 returning adult chinook salmon by National Marine Fisheries Service) and a limited chinook salmon recreational fishery could be allowed involving an average of 35 hours to catch one salmon. While the exact mapping of chinook salmon populations to alternative catch rates is not known with certainty, discussions with Idaho Fish and Game suggests this might occur at 150% of delisting levels or roughly 50,000 returning adults. Thus, setting HRSAL to 35, the Travel Cost Method demand equation estimates angler trips would increase by 2.088 per angler. Applying this new level of trips to estimated number of anglers, yields an additional 1.425 million angler days using the low approach. The additional NED benefits of these increases in angler days is $108 million annually. Of course the present value of this benefit will depend upon how far into the future chinook salmon fishing will be initiated on the Lower Snake River. The angler response to opening of salmon fishing on the Lower Snake River has the biggest effect, and as shown in the next simulations, increases in catch rates have a much smaller incremental effect on angler trips. This is sensible given diminishing marginal utility of additional units of a good. If we assume that anglers that did not respond to the survey would still visit the Lower Snake River if chinook salmon populations increased to allow salmon fishing, total angler days would be 5.9 million angler days with the increment due to opening up of salmon fishing being 3.178 million angler days. The low-middle estimate assumes that some proportion of anglers that were non-respondents to our survey, would visit. This proportion is by adjusting the for the survey response rate using a higher survey response rate that would have resulted in some of the non-responding anglers, responding and therefore reporting positive trips. This procedure yields an estimate of 1.8 million angler days prior to salmon fishing being allowed, and 2.7 million angler days once habitat recovery is sufficient to allow salmon fishing.

Before any of these demand estimates should be used in the sensitivity analysis in the benefit-cost or EIS it must be determined by biologists whether such a recently recovered fishery could withstand the estimated level of angler pressure. It may be that this demand exceeds what a recently recovered salmon fishery could support and angler use would have to rationed using drawings to the amount of angler use biologically sustainable. The determination of the biologically sustainable use level at 35 hours per salmon harvested should be made to determine what proportion of the 5.9 million angler days could be supported once the chinook salmon populations recover to allow limited recreational fishing.

In Table 7 (and 8), we provide line below our angler estimates to suggest what the adult salmon catch would have to be to accommodate the predicted level of demand. If the PATH prediction of adult salmon numbers are less than this, then the increase in salmon angler days can be scaled back proportionately. In addition, the time line of salmon population increases can be applied to the demand estimates to develop the time profile of increases in angler demand that can be accommodated as salmon populations increase over time. However, at time we are finalizing this report, the estimated adult salmon catch figures are not available to us.

Besides the biological capacity, other simple tests of reasonableness can be performed to evaluate which of the five different estimates are most plausible. The low and low-middle estimates of 2.7 to 3.85 million angler days seem plausible, given there were about 40 million angler days of freshwater fishing reported by anglers from California, Idaho, Oregon and Washington in the 1991 U.S. Fish and Wildlife Service's National Survey of Hunting, Fishing and Wildlife Associated Recreation. Thus less than 10% of freshwater angler days in the region would be involved, and the 140 miles of river is a significant addition of free-flowing salmon fishing opportunities in the region. Further, our survey does not account for any angler days that might be taken by people living outside the survey region (e.g., in nearby states of Wyoming, Utah, Nevada and Colorado) which have substantial populations of anglers and no anadromous fishing opportunities. The low and middle estimate of angler days also yield estimates of about 100 angler days per mile of river over a 180 day season. While this is a season average, photographs of peak steelhead and salmon fishing days on other rivers appear to be substantially higher than this.

These angler demand and benefit estimates can be compared to the current reservoir angler use and benefits estimated by McKean (1998a). In McKean's report, he uses the Normandeau Associates fishing use estimates collected by David Bennett's group at the University of Idaho. This yields about 3,305 anglers, fishing 87,050 days. This yields annual reservoir angling benefits of $1.956 million. Thus current angler use and benefits are less than 10% of what is projected to occur in the future if the dams are removed and the anadromous fishing improved. Of course, there will be several years between dam removal and salmon recovery where there will be reduced recreational fishery, even below current levels. Given the interest rate, each dollar of these near term losses weigh more heavily than a dollar of future gain.

The lower portion of Table 7 analyzes a further improvement in chinook salmon fishing quality to one salmon caught 20 hours of fishing. Changing HRSAL in the TCM demand equation we estimate angler trips would increase by an additional .1343 trips per angler per year. Applying this increase yields an estimated additional 91,674 angler days, in total, from all groups of anglers (e.g., local and rest of Pacific Northwest and California). This has an incremental value of $6.9 million annually. Once again, the present value of this would need to be calculated accounting for the likelihood that it may take several decades before salmon populations would increase to these levels (possibly around 140,000 returning adult chinook salmon). Clearly, the biologists will need to map from returning adult chinook salmon into hours to catch a salmon to apply our approach for calculating the angler benefits over time and alternatives. The last policy simulation in Table 7 calculates future angler use and benefits if salmon fishing quality improved to 10 hours per salmon caught. The low estimate is an additional 62,764 trips worth an additional $4.77 million annually.

The salmon fishing benefits could likely be applied to increases in salmon from improved transport and other dam improvements. Our demand estimates may over-estimate the number of salmon fishing trips and benefits associated with dam improvements, if anglers strongly prefer fishing for salmon in a free-flowing river (as described in the survey) as compared to reservoirs, however. Of course, the non-angling river recreation benefits would not occur without dam removal. However, the existing reservoir recreation estimated by McKean (1998b) would likely remain and could be used to estimate the non-fishing benefits of dam improvements.

Table 8 repeats the analysis and policy simulations at the Middle-high and High estimates of angler visitation rates. The Middle-high and High estimates use the sum of anglers that definitely and probably would visit if the dams were removed. The Middle-high estimate assumes anglers that did not respond to the survey would not visit the Lower Snake River if salmon fishing were made available. The resulting demand estimates of 11-27 million angler days need to be evaluated by fisheries biologists to determine what portion of this angler demand could be accommodated by a newly recovered chinook salmon fishery. It may be that once full recovery occurs in 3-4 decades, that this level of angler demand might be accommodated on the Lower Snake River.

PASSIVE-USE VALUE BENEFIT TRANSFER ANALYSIS

Salmon Passive Use Value Literature

Before reviewing the specific studies on the economic value of salmon and free-flowing rivers, we wish to define benefit transfer. There are several closely related definitions of benefit transfer (Brookshire and Neill, 1992). A commonly referred to definition is provided by Boyle and Bergstrom (1992:657) "...benefit transfer is defined as the transfer of existing estimates of nonmarket values to a new study [site] which is different from the study for which the values were originally estimated." Desvousges et al., characterize benefit transfer as application of economic values from a previously studied site to a new policy site under evaluation.

A review of two large computerized economic databases (American Economic Association's EconLit and Environment Canada's recently developed Environmental Values Reference Inventory or EVRI) yielded four published studies, but only three which presented original passive use values for salmon. These three are: 1) Olsen, Richards and Scott's article published in Rivers on existence values for doubling the size of Columbia River Basin salmon and steelhead runs; 2) Loomis article in Water Resources Research on the economic benefits of increased salmon from removing the dams on the Elwha River; and 3) Hanemann, Loomis and Kanninen's article published in American Journal of Agricultural Economics on benefits of increasing chinook salmon populations in the San Joaquin River.

The key characteristics of these three studies are described in Table 9. The Olsen et al., study involved a telephone interview of Pacific Northwest households, using an open-ended WTP question format. Their means of payment (i.e., payment vehicle) was an increase in household electric bill. The change in salmon was a doubling from 2.5 million to 5 million salmon, for a net change of 2.5 million salmon. The response rate on the phone interviews was quite good at 72%. The original study existence value is $26.52 per household amounting to $32.52 in 1996 dollars.

The Loomis (1996a) Elwha study used a mail questionnaire and a dichotomous choice willingness to pay question format. An increase in federal taxes was the payment vehicle. The mail survey had a response rate of 68% for Washington residents and 55% for rest of U.S. residents. Respondents were shown a bar chart in the survey indicating the increase in salmon population due to dam removal was approximately 300,000 fish. The Washington residents value was $73 in 1994 dollars or $76.46 in 1996 dollars. The value to the rest of U.S. residents was quite similar at $68 in 1994 dollars or $71.24 in 1996 dollars.

The Hanemann et al. (1990), study involved a combination telephone contact, mail survey booklet and then phone interview of respondents using the survey booklet. A dichotomous choice willingness to pay question was used, with taxes as the means of payment. The resource was an increase from 100 chinook salmon to 15,000 chinook salmon for a net increase of 14,900 in the San Joaquin River. The combination phone-mail-phone survey had a 51% response rate. The value per household was $181 in 1989 dollars or $222 in 1996 dollars. Using Pate and Loomis' reanalysis of this data to relate it to distance from the resource, we are able to approximate the value a household in the mid-western U.S. and east coast of the U.S. would have for protecting salmon. This value is about 75% of the value per household on the west coast.

Key characteristics of each study for the benefit transfer are shown in Table 9. More details on each study can be found in the respective journal article.

Criteria for Benefit Transfer

Boyle and Bergstrom (1992:659) "...propose the following idealistic technical criteria:

1) the nonmarket commodity valued at the study site must be identical to the nonmarket commodity to be valued at the policy site;

2) the populations affected by the nonmarket commodity at the study site and the policy site have identical characteristics;

3) assignment of property rights at both sites must lead to the same theoretically appropriate welfare measures (e.g., willingness to pay versus willingness to accept compensation)."

In reviewing existing empirical studies on the non-use value of salmon, criteria 1) and 3) are met quite well. With respect to criteria 3), we have obtained estimates of willingness to pay (WTP) from all of the studies and this is value measure of value required under the Principles and Guidelines (U.S. Water Resources Council, 1983). The studies we have measure the value of salmon or salmon and steelhead. These are the general category of fish species we are concerned with in the Lower Snake River study so criteria #1 is satisfied, with two minor exceptions. Our policy involves increasing populations of officially listed Threatened and Endangered salmon, while the existing studies value increases in salmon that are not threatened with extinction. The second subtle issue has to do with valuation of wild stocks versus hatchery stocks. One study below, deals primarily with wild stocks, whereas the other two studies deal with a mix of wild and hatchery stocks. The species of greatest policy relevance in the Lower Snake River are the recovery of wild salmon and steelhead populations. On both these issues, the values in the existing literature will tend to understate the values the public would likely place on increasing officially listed wild stocks at stake in the Lower Snake River. Some evidence for the fact that existing scarcity influences WTP may be found in comparing the WTP estimates from three studies in relation to baseline populations and target populations. The Olsen et al., study has the highest baseline population (2.5 million fish) and highest increase (another 2.5 million fish for a total of 5 million). Loomis (1996) presents the baseline at about 50,000 fish with an increase to 350,000, indicating a net gain of over 300,000 fish. Finally, the San Joaquin Valley study has a baseline of 100 salmon versus the program target of 15,000. The marginal values per fish in Table 9 follow the relative scarcity.

Criteria 2) is met very well for the Olsen et al. study, as their sampling frame is nearly identical to our study (e.g., the Pacific Northwest states of Idaho, Montana, Oregon, and Washington). The Loomis (1996a) study on the Elwha matches fairly well, as the State of Washington is by far the most populous of the Pacific Northwest states affected by the removal of the Lower Snake River dams. Only the study on the economic value of salmon in the San Joaquin River in California, is weak on criteria 2) (since the majority of the sample is California residents, with only a portion of the sample being Oregon and Washington residents). Although the species match is good, the San Joaquin transfer is weaker than the rest in that the river is located in central California, rather than the Pacific Northwest.

Nonetheless, compared to many benefit transfers illustrated in the special section of Water Resources Research and attempted by others, our benefit transfer adheres fairly closely to the technical criteria for an ideal benefit transfer.

A further factor suggesting the merit of the benefit transfer relates to the quality of the studies used in the analysis. All of the studies had large samples and reasonably high response rates. More importantly, the studies of Olsen et al., Loomis (1996a) and Hanemann et al., and Pate-Loomis (1997) were all published in peer reviewed scientific journals.

Generalizing the Value per Household to the Region

As noted by Gardner Brown and Jason Shogren (1998) the benefits of threatened, endangered or rare species extend far beyond the local area where the critical habitat is located. Since non-use values are public goods which can be received by all households in the U.S. simultaneously, the potential market area over which to vertically sum the individual values could be nationwide. Statistical analysis using the San Joaquin Valley data (Pate and Loomis, 1997) indicates that the non-use values for salmon did not fall off rapidly with distance from the river where the salmon were being protected. Statistical analysis of using the Elwha data (Loomis, 1996b) found only a very slight decline in WTP with distance, falling from $78 to $68 per household as distance went from 100 miles to 2,500 miles from the river. Residents of both the east and west coasts in the United States valued protection of California and Northern Spotted Owl habitat (Loomis and Caban, 1996).

Besides statistical criteria is the fact that this salmon species being recovered in the Lower Snake River are federally listed species. The recommendation on dam removal are being made by Federal agencies. It is likely some Federal tax dollars would be used to pay for any dam removal and river restoration. Finally, the U.S. Water Resources Council Principles and Guidelines requires a national accounting stance when performing benefit cost analysis. Thus, one end of the non-use value spectrum will be to include distance adjusted values per household to households in the rest of the U.S.

At the other end of the spectrum is to use just households in the Pacific Northwest region of Idaho, Montana, Oregon and Washington. This would be a lower bound estimate of the non-use values, as it presumes no other households in the rest of the U.S. receives any benefits. As noted above, empirical evidence from two different studies suggests this is would result in a substantial underestimate of the non-use values.

In designing the original survey which was to include passive-use values, it was decided by the interagency team to use Pacific Northwest households plus California as the relevant population. California was added as that state is tied into the electricity market for power from the hydropower system. Northern California is sometimes considered part of the Pacific Northwest. Consistent with Olsen et al. (1990), we calculated a lower bound estimate using only households that did not include an angler as a non-user or existence value household. This is a low approach as anglers can also hold non-use value (Loomis, 1988), but it definitely eliminates any possible concern with double counting of recreation and non-use values. Information from the U.S. Fish and Wildlife Service's National Survey of Fishing, Hunting and Wildlife Associated Recreation was used to calculate the number of angler households in the states of California, Idaho, Montana, Oregon and Washington. As noted above, since users can still hold existence values, a less conservative approach expands the existence values per household to all households in the Pacific Northwest and California. This results in a small increase in non-use values per salmon and total non-use values in this region.

Estimating Benefit Transfer Functions

While the range of value per household and value per salmon in Table 9 may appear large, the pattern of values actually correspond with economic principles of diminishing marginal value. Specifically, the Olsen et al., study offers a gain of 2.5 million salmon from a base of 2.5 million salmon, and thus the value per additional salmon is lower than in the Hanemann et al., where they were offered an increase of 14,900 salmon on a base of 100 salmon. When you have very little of something and are offered a small increase, each additional unit is very valuable.

To incorporate the value estimates for all three studies and provide a WTP function that could be applied to the change in number of salmon for a wide range of policy alternatives, the non-use value per salmon was regressed on the number of salmon offered in the program. The regression equation for non-angler Pacific Northwest households is:

Natural Log($NUV per Salmon)= 25.353 -1.37358(Natural Log of Number of Salmon)
R² = .99

T-Statistic	(-102.32)

The regression equation for all Pacific Northwest households is:

Natural Log($NUV per Salmon)= 25.357 -1.37315(Natural Log of Number of Salmon)
R² = .99

T-Statistic	(-104.73)

As is evident from the regression equation it is nearly a perfect fit and number of salmon is highly significant, even adjusted for the number of observations. This equation attempts to approximate a marginal benefits function, estimating the marginal non-use value of another salmon in the Lower Snake River. From this marginal function, the total non-use value is constructed. A summary of the annual values are presented in Table 10. From this table simple calculation of the net difference due to dam removal or alternative transport policies can be measured.

This calculation can be performed from two different perspectives, that yield different starting and ending points on the total benefits curve. If one applies the benefit function just to the Lower Snake River, then the starting point is the current number of returning adult chinook salmon has averaged between 8,000 and 10,000 over the last decade. If 10,000 is expected to continue, baseline passive-use value would be roughly $7.8 billion annually. If dam removal would boost returning adult salmon populations in 20 years by 50,000 salmon to 60,000 salmon, total passive-use value would rise to $11.6 billion annually for a gain of $3.8 billion each year (this incremental passive use value represents less than one-half of one percent of regional income, so is quite reasonable). However, as suggested by one reviewer, this benefit function could be interpreted as a Columbia River basin or even an entire Pacific Northwest salmon benefit function. In this case the starting point would be around 260,000 returning adult salmon in the Columbia River plus the number of returning adults in other coastal streams in the Pacific Northwest. To illustrate this other perspective, if there are 350,000 returning adults this would be the baseline for the Lower Snake River analysis. To continue the example, the value of the same 50,000 adult salmon increment would now be $103 million annually ($13.723 billion minus $13.620 billion). Which of these perspectives the DREW and the COE wishes to adopt is in part a policy question and in part a question of what is the baseline for the future without. To a certain extent the EIS is addressing solely changes in the operation of the Lower Snake River dams, not the entire Columbia River system, so one might argue the Lower Snake River starting point of 10,000 adult salmon is appropriate. This view is reinforced by the fact the chinook and sockeye salmon being evaluated in the Lower Snake River are listed as separate species under ESA. Alternatively, to if all wild salmon are treated equally and to ensure this analysis is a marginal analysis reflecting current amounts of salmon in the Pacific Northwest would be used as the baseline. The COE and DREW will need to decide whether to use the Lower Snake River, Columbia River Basin, or the entire Pacific Northwest as the baseline to use.

By using the values in Table 10, the analyst can calculate the annual change in passive use values to Pacific Northwest household for changes in salmon (linear interpolation would be needed for changes intermediate between our increments or the analyst could use the regression equation to estimate changes in passive use value for finer increments). Note that even using a total passive use value interpretation of Table 10, the existence value of 400,000 salmon represents about 1% of regional income. This seems quite reasonable given the high profile nature of endangered salmon in the region.

To calculate the passive use value for rest of U.S. households, we only had two data points: 1) the rest of U.S. benefit estimate from Loomis (1996) Elwha study; and 2) Pate-Loomis (1997) estimated willingness to pay for San Joaquin River salmon as a function of distance. Using these two data points a rest of the U.S. WTP equation was calculated. Because there are two points, the equation is a straight line of the form:

($NUV per Salmon)= 948.024 -1.60232*(Number of Salmon)

Using this equation, the marginal and total passive use value to rest of U.S. households was calculated. The annual values are displayed in Table 11. Given the large number of households in the mid-west and east coast, the passive use values are quite large, nearly ten times that of the Pacific Northwest and California. As before the analyst can use the table or the equation to calculate the change in passive use values with changes in salmon populations in each year of recovery.

Passive Use Value of Free-Flowing Rivers

Besides the existence and bequest values of the salmon themselves, is the existence value of having the Lower Snake River as a free-flowing river once again. This is the value of restoring the canyon back to its natural, pre-dam condition. Like the estimating the non-use value of salmon, we were asked by the Corps of Engineers to make a rough estimate of this value using existing studies.

A mail survey of WTP to preserve free-flowing rivers was performed by Sanders, Walsh, and Loomis (1990) of Colorado households statewide. The mail survey had a 51% response rate of deliverable surveys. The annual WTP per household for option, existence and bequest value was $77 in 1983 dollars or $116 in 1996 dollars. Dividing this by the 555 miles being valued yields a value of 21 cents per mile. Multiplying this by the 140 miles of the Lower Snake River yields a value per household of $29.40 per year per household. The rivers included in this list are all within Colorado and include the Yampa, Dolores and Green River.

Another study was a contingent valuation method estimate of preserving the Black Canyon of the Upper Snake River from development. This study was performed by Scott and Wandschneider (1993). The University of Idaho conducted telephone interviews of residents of the four counties in Southeastern Idaho surrounding this section of the river. The study was conducted for the Bureau of Land Management. The survey had a response rate of 76% and a sample size of nearly 350.

The survey identified that slightly more than half of the sample were non-users (n=196) and they had an annual WTP of $58 for preservation of the Upper Snake River (as compared to users who had a WTP of $92). Dividing this value by the 63 miles protected yields a value of 92 cents per mile per household. This value is naturally higher since only individuals living in counties adjacent to the river were sampled. This value per household per mile is applied to the counties surrounding the Lower Snake River. Thus a value of 92 cents times 140 miles or $129 would be multiplied by non-use households in the counties surrounding the Lower Snake River. To calculate this we use the data in Table 6. We subtract out the number of non-angling households that would visit the Lower Snake River from the total non-angling households. This yields 305,467 non-angling, non-user households. Therefore $129 per household times 305,467 households yields our estimate of $39.4 million in passive-use value for restoring a free-flowing Lower Snake River. We use the Sanders et al., value of $29.40 for the rest of non-user households in Washington, Oregon, Idaho, Montana and California. Using data from Table 7 we estimate 12.95 million non-angling, non-user households. Multiplying by Sanders et al., value of $29.40 per household yields an estimated non-use value for restoring the Lower Snake River of $380.73 million. Thus aggregate passive use value for just the restoration of the free-flowing nature of the Lower Snake River is $420.13 million.

CONCLUSION

The removal of the four dams on the Lower Snake River and restoration of the river to a free-flowing condition appears to have substantial recreation use and passive use values to residents of the Pacific Northwest and California. The importance of the passive use values is evident as they are more than ten times the recreation benefits. Nonetheless, the river recreation benefits of a restored free-flowing river are estimated using the Low NED benefits for river recreation at least $144.2 million a year, increasing to $252 million annually once chinook salmon populations increase to the point where salmon fishing is allowed. Using the river recreation benefits based on river visitors' reported trip cost yield an estimated annual benefits of $288 million annually, increasing to $396 million annually once chinook salmon populations increase to the point where salmon fishing is allowed. These estimates of restored river recreation benefits are at least four times larger than current reservoir recreation benefits ($33.5 million) and may be as much as ten times larger than current reservoir recreation benefits, if all of the free-flowing river recreation users projected to visit the restored river can be accommodated at a restored lower Snake River.

REFERENCES

Aillery et al.

Salmon Recovery in the Pacific Northwest: Agricultural and Other Economic Effects. Report #727. USDA Economic Research Service, Washington DC.

Arrow, K., R. Solow, P.Portney, E.Leamer, R. Radner and H.Schuman., 1993.

Report of the NOAA Panel on Contingent Valuation. Federal Register 58(10):4602-14.

Bockstael, N., I. Strand and M. Hanemann. 1987.

Time and the Recreation Demand Model. American Journal of Agricultural Economics 69:293-302.

Boyle, Kevin and John Berstrom. 1992.

Benefit Transfer Studies: Myths, Pragmatism and Idealism. Water Resources Research 28(3):657-663.

Brookshire, David and Helen Neill. 1992.

Benefit-Transfers: Conceptual and Empirical Issues. Water Resources Research 28(3):651-655.

Brown, Gardner and Jason Shogren. 1998.

Economics of the Endangered Species Act. Journal of Economic Perspectives 12(3):3-20.

Callaway, John, Shannon Ragland, Salley Keefe, Trudy Cameron and Douglass Shaw. 1995.

Columbia River System Operation Review Recreation Impacts: Demand Model and Simulation Results. Appendix J-1, Final Environmental Impact Statement. DOE/EIS-0170, U.S. Army Corps of Engineers, North Pacific Division, Portland, OR.

Carson, R. and 7 others. 1997.

Temporal Reliability on Estimates from Contingent Valuation. Land Economics 73(2): 151-163.

Champ, P., R. Bishop, T. Brown and D. McCollum. 1997.

Using Donation Mechanisms to Value Nonuse Benefits from Public Goods. Journal of Environmental Economics and Management 33(2):151-162.

Creel, M. and J. Loomis. 1990.

Theoretical and Empirical Advantages of Truncated Count Data Estimators for Analysis of Deer Hunting in California. American Journal of Agricultural Economics 72(2):434-41.

Desvousges, William, Michael Naughton and George Parsons. 1992.

Benefit Transfer: Conceptual Problems in Estimating Water Quality Benefits Using Existing Studies. Water Resources Research 28(3): 675-683.

Hanemann, Michael, John Loomis and Barbara Kanninen. 1991.

Statistical Efficiency of Double Bounded Dichotomous Choice Contingent Valuation. American Journal of Agricultural Economics, Volume 73(4): 1255-1263.

Krutilla, John. 1967.

Conservation Reconsidered. American Economic Review 57:777-786.

Krutilla, John and Anthony Fisher. 1975.

The Economics of Natural Environments. Resources for the Future, Johns Hopkins University Press, Washington DC.

Layman, Chris, John Boyce and Keith Criddle. 1996.

Economic Valuation of Chinook Salmon Sport Fishery on the Gulkana River, Alaska under Current and Alternate Management Plans. Land Economics 72(1):113-128.

Lockwood, Michael, Terry DeLacy and John Loomis. 1994.

The Relative Unimportance of a Nonmarket Willingness to Pay for Timber Harvesting. Ecological Economics 9:145-152.

Loomis, John. 1988.

Broadening the Concept and Measurement of Existence Value. Northeastern Journal of Agricultural and Resource Economics, Volume 17(1), April.

Loomis, John. 1990.

Comparative Reliability of the Dichotomous Choice and Open Ended Contingent Valuation Techniques. Journal of Environmental Economics and Management, Volume 18(1): 78-85.

Loomis, J., B. Roach, F. Ward and R. Ready. 1995.

Testing Transferability of Recreation Demand Models Across Regions: A Study of Corps of Engineer Reservoirs. Water Resources Research 31(3):721-730.

Loomis, J. 1993.

An Investigation into the Reliability of Intended Visitation Behavior. Environmental and Resource Economics. 3:183-91.

Loomis, John. 1996a.

Measuring the Economic Benefits of Removing Dams and Restoring the Elwha River: Results of a Contingent Valuation Survey. Water Resources Research, 32(2):441-447.

Loomis, John. 1996b.

How Large is the Extent of the Market for Public Goods: Evidence from a Nationwide Contingent Valuation Survey. Applied Economics 28:779-782.

Loomis, John and Armando Gonzalez-Caban. 1996.

The Importance of the Market Area Determination for Estimating Aggregate Benefits of Public Goods: Testing Differences in Resident and Nonresident Willingness to Pay. Agricultural and Resource Economics Review 25(2):161-170.

Marmorek, D., C. Peters and I. Parnell, eds. 1998.

PATH Final Report for Fiscal Year 1998. ESSA Technologies Ltd. Vancouver, B.C., Canada.

McKean, J., D. Johnson and R. Walsh. 1995.

Valuing Time in Travel Cost Demand Analysis. Land Economics 71:96-105.

McKean, John. 1998a.

Sportfishing Use and Value on the Lower Snake River Reservoirs. Agricultural Enterprises Inc., Masonville, CO.

McKean, John. 1998b.

Outdoor Recreation Use and Value on Lower Snake River Reservoirs. Agricultural Enterprises Inc., Masonville, CO.

McKean, John. 1998c.

Sportfishing Use and Value on the Free Flowing Snake River Above Lewiston. Agricultural Enterprises Inc., Masonville, CO.

Meyer, Phillip. 1974.

Recreation and Preservation Values Associated With Salmon of the Frasier River. Environment Canada. PAC/IN-74-1, Vancouver, Canada.

Olsen, Darryll, Jack Richards and R.Douglas Scott. 1991.

Existence and Sport Values for Doubling the Size of Columbia River Basin Salmon and Steelhead Runs. Rivers 2(1):44-56.

Pate, Jennifer and John Loomis. 1997.

The Effect of Distance on Willingness to Pay Values: A Case Study of Wetlands and Salmon in California. Ecological Economics 20:199-207.

Randall, Alan and John Stoll. 1983.

Existence and Sport Values in a Total Valuation Framework. in R. Rowe and L. Chestnut, Managing Air Quality and Scenic Resources at National Parks and Wilderness Areas. Westview Press: Boulder, CO.

Randall, Alan. 1998.

Beyond the Crucial Experiment: Mapping the Performance Characteristics of Contingent Valuation. Resource and Energy Economics 20:197-206.

Sanders, L., R. Walsh and J. Loomis. 1990.

Toward Empirical Estimation of the Total Value of Protecting Rivers. Water Resources Research 26(7):1345-1357.

Scott, R.D. and P. Wandschneider. 1993.

A Hedonic Model of Preservation Value Components: A Contingent Valuation Study of the Black Canyon of the Upper Snake River. in Proceedings of the Twenty-Seventh Annual Pacific Northwest Regional Economic Conference. Kennewick, WA.

Stoll, John and Lee Ann Johnson. 1984.

Concepts of Value, Nonmarket Valuation and the Case of the Whooping Crane. Transactions of the 49th North American Wildlife and Natural Resources Conference, Wildlife Management Institute. Washington DC.

U.S. Water Resources Council.

Economic and Environmental Principles and Guidelines for Water and Related Land Resources Implementation Studies. U.S. Government Printing Office, Washington DC. March 10, 1983.

Walsh, R., D. Johnson and J. McKean. 1988.

Benefit Transfers of Outdoor Recreation Demand Studies: 1968-1988. Water Resources Research 28:707-713.

Ward, F. 1987.

Economics of Water Allocation to Instream Uses in a Fully Appropriated River Basin. Water Resources Research 23(3):381-392.

Return to the Products Page