Water, economics and climate change in Oregon's Willamette Basin

Executive summary

Climate change, population growth and income growth have the potential to significantly affect the availability and use of water in the Willamette River Basin. How these changes will affect water scarcity is uncertain. Individuals, communities and governments across the WRB need to better understand how the supply and demand for water will evolve and vary across space and time in coming decades. The Willamette Water 2100 project was motivated in response to that need. The project was a six-year research effort aimed at developing a computer simulation model that represents the important processes of both the natural and human systems related to water supply and demand. Our understanding of water supply relies heavily on hydrology and related natural sciences, whereas our understanding of the demand for, and allocation of, water comes from economics, law, engineering and related social sciences. The linkages, interconnections and feedbacks in complex systems of this kind are very difficult to fully understand without a detailed model.

The WW2100 model allows us to project into the future the ways that changes in climate, population and income from 2010 to 2100 will alter the supply, demand, allocation and scarcity of water. This report describes results from the WW2100 model, with a focus on its economic dimensions, that is, the impacts on people who live in the WRB.

Water supply

On the water supply side, the severe decline in snowpack in the next 80 years will reduce the amount of snowmelt runoff from April to June. The projected reduction in average available snowmelt (as of April 1 each year) represents a decline of about 600,000 acre-feet of stored water. However, precipitation plays a far greater role than snowmelt in determining spring streamflows in the WRB. Indeed, the projected decline in snowmelt is only one-tenth of average April–July precipitation (nearly 7 million acre-feet). As a result, unlike arid basins in eastern Oregon or neighboring states, the loss of snowmelt will have a relatively small impact on water availability in the lower elevations of the WRB in spring and summer. Nevertheless, reduced snowpack, when combined with higher summer temperatures, is projected to increase stress on upland forests and, as a result, increase the risk of wildfire.

Water use

In general, water use is influenced by both the demand (willingness to pay) for water and the cost of transporting, storing, or transforming water to make it available for a given use. The importance of cost considerations is critical. Demand for transported water depends on the value of water for a specific purpose relative to conveyance costs. For example, water is transported up to 25 miles from outside the WRB (often aided by gravity) to serve urban users. In contrast, we estimate that a quarter mile of horizontal or uphill conveyance can be costly enough to make delivery of water for irrigation uneconomic on most currently unirrigated agricultural lands in the Basin. Indeed, economic considerations explain why one-third of irrigable farmland (parcels with irrigation water rights) goes unirrigated each year. Irrigation involves costs and benefits, and, in some years and on some lands, the costs outweigh the benefits.

Key findings related to water use are as follows:

• Water use for irrigated agriculture fluctuates from year to year, but has exhibited no significant upward trend in recent decades. The per-acre amount of water required for irrigation is expected to remain relatively stable. However, seasonal patterns of irrigation are likely to shift about two weeks earlier in response to earlier planting dates resulting from climate change. The potential use of stored water to expand irrigation to farmlands that currently do not have irrigation water rights is limited by economic realities; conveyance costs are high relative to the economic gain from irrigating.
• Urban water use is projected to rise significantly by 2100, due primarily to population growth, but also to rising income. The growth in demand will be tempered to some degree by recent and near-term price increases related to cost recovery for infrastructure investments. The burden of these increases on WRB water supplies will be limited by several factors. First, most urban water is used indoors and is nonconsumptive, that is, it is returned to the surface-water source from which it originated. Second, a large fraction of urban water supplies in the WRB come from sources outside the Basin. Thus, consumptive use from in-basin surface water sources represents only about 7% of total urban water deliveries and is projected to increase by only 16,000 acre-feet by 2100. Urban consumptive use of water from in-basin sources is small compared to other uses. Agricultural use (475,000 acre-feet) is 25 times greater, and regulatory minimum flows in the Willamette River (3.5 to 4 million acre-feet at Salem) are 200 times greater.
• Protection of in-stream flows (by federal and state law) is the largest allocation of water in the Basin under human influence or control. These flows serve multiple purposes, but are determined largely by habitat requirements of native fish. These minimum streamflows result from both federal requirements under the Endangered Species Act and state-mandated perennial minimum flows protected by in-stream water rights.
• The 13 federal storage reservoirs in the Basin produce enormous social value by reducing the risk of flood events. This benefit has been estimated at more than $1 billion per year. As urban areas expand, the value of potential damages during a flood will rise. Thus, the economic benefits of flood damage reduction will increase. To the extent that climate change leads to increases in high flow events, these benefits will become even more valuable.

Water scarcity

In some parts of the WRB and at some times of year, water is scarce, and that scarcity is likely to increase in the future. The potential for increased water scarcity will be location- and time-specific. Our model results suggest the following:

• The municipal water rights currently relied upon may reach capacity in the Metro area (in 30 years) and in Salem (in 60 years). However, when the model accounts for currently underutilized water rights and those under development, urban water rights appear to be capable of meeting the overall growth in urban water demand.
• Climate change is projected to result in earlier planting for agriculture. Earlier planting will lead to more crop growth during the months when temperatures are cooler and soil moisture is more available. Earlier planting will also lead to an earlier start, and completion, of irrigation. In the future, more farmers will have finished irrigating by the time the threat of a shutoff arises, according to the model results. As a result, the model shows a slight decrease in irrigation shutoffs. Climate models differ, however, in terms of whether precipitation is predicted to increase or decrease overall, although most models suggest somewhat wetter winters and drier summers.
• Implementation of all of the “unconverted” in-stream water rights intended to protect perennial flows would represent a significant increase in the amount of water allocated under state law to environmental values. Overall, however, our results suggest that flow requirements to protect salmon and steelhead can be met, based on 10-year average flows. Exceptions are likely to occur in drought years.
• The effects of changes in forest wildfires and fire suppression policies could have a larger effect on water supply in the Valley than all of the changes in human water use combined. If forest cover is dramatically reduced, the resulting decrease in forest evapotranspiration will increase streamflows and make more water available for human use.

Introduction

Climate change, population growth and income growth have the potential to significantly affect the availability and use of water in the Willamette River Basin (WRB). Human decisions will contribute to these potential changes. Indirectly, changes in land use will affect where and when water is used for various purposes. Legal and regulatory changes, such as modifications to augment required in-stream flows for salmon and steelhead, will also influence the supply and demand for water.

Uncertainty about future supply of and demand for water across space and time created the motivation for this report. Individuals, communities and governments across the WRB have a strong interest in gaining a better understanding of how changes in land use, climate, population, income and regulatory requirements will affect the supply, demand and scarcity of water. Policy makers, government agencies and private interests need to better understand what kinds of changes are likely to occur so that they can act to mitigate or adapt.

Predicting the magnitude, timing and location where water scarcity will arise requires a detailed understanding of the region’s biophysical and human systems — often referred to as a “coupled human–natural system.” Economic and biophysical components interact across space and time in complex ways.

The linkages, interconnections and feedbacks in complex systems of this kind are very difficult to fully understand without a detailed model. Such a model was constructed by Willamette Water 2100 (WW2100), a six-year research project funded by the National Science Foundation (EAR 1039192, 1038925 and 1038899). The WW2100 project involved dozens of researchers from Oregon State University, Portland State University and the University of Oregon, bringing expertise in biophysical sciences, economics, engineering, planning and law. The centerpiece of the project was the development of a large simulation model integrating all of these dimensions.

Creating a model of the human system for this region required a broad economics perspective of human actions and interactions. The model includes a system-wide representation of how humans interact with the natural system individually and collectively, as reflected in laws, regulations, property rights and other institutions that guide and constrain our choices. These individual and public aspects of the human system manifest themselves in land use and land-use change; water supply and demand for urban, agricultural, reservoir and in-stream flow allocations; and ecological water use by forests and other vegetation.

The main goals of the WW2100 study were twofold: (1) to project where, when and under what institutional conditions (laws, regulations and rights) water scarcity might increase in the WRB, and (2) to consider what kinds of policies and other actions might be warranted to prepare for, mitigate, or adapt to changes in water scarcity.

This report describes results from the WW2100 model, with a focus on its economic dimensions, that is, the impacts on people who live in the WRB. Those impacts include two types:

• Changes in the availability of resources that are priced and exchanged in markets, such as land and urban water supplies
• Changes in nonmarket resources valued by society, including in-stream flows, forest health, flood risk protection and recreational opportunities

The body of this report does the following:

• Describes the WW2100 model and provides an understanding of how the model and its components represent processes, interactions and changes in the way people and resources interact
• Describes the main findings of the model’s simulations
• Highlights the economic aspects of expected changes in the Basin as they relate to water supply, demand and availability for specific uses

Much of the technical background about the model components is included in the Appendix.

It is important to recognize the kinds of questions that our research, and this report, are not intended or able to answer. This work does not provide specific guidance to individual local governments, for example, regarding decisions to invest in or finance water supply infrastructure. Nor does our model provide advice to individual farms about adopting new irrigation methods or other farm-level investments. Although many characteristics and processes in the WRB have been modeled at high spatial and temporal detail, many other important dimensions cannot be described at the level of detail needed for decision making by individuals, farms and local governments.

Moreover, it is important to acknowledge that results from most models, especially ones that make projections decades into the future, should be interpreted not as precise predictions, but as likely trajectories of change. Models are based on our best understanding of biophysical and human processes and how they are likely to interact in the future. In this way, a model is a tool that describes future conditions under a range of alternative assumptions and scenarios.

Water allocation and scarcity

Economics is the study of the allocation of scarce resources — how people use and value resources such as water and land. Some resources are exchanged in markets. In this case, prices reflect their marginal (incremental) value, provide incentives and encourage efficient use. For other resources, including water, some other allocation mechanism is often used.

Allocating water is complicated for a variety of reasons. Water is essential to all living things; as a result, most water is left in situ, or in the environment. Water has been described as a “fugitive” resource due to how it moves around, flows, seeps, evaporates and is transpired by plants. In this and other ways, water poses challenges for individuals and society. For example:

• The availability of water varies and is uncertain across space and time.
• The cost of transporting or storing large amounts of water is high due to its high volume and weight relative to its value per unit.
• Water plays a key role in complex natural–human systems and is strongly related to public goods, such as ecosystems, fish and wildlife habitat and recreation.
• Water plays a role in quasi-public goods, such as municipal water delivery, flood control and power generation.

“Water scarcity” can be understood as the cost (or benefit) of having one less (or more) unit of water. It is therefore the marginal, or incremental, cost or value of water. Another way to look at this is as the cost of a decrease in water availability or the benefit of an increase in water supply.

The Willamette River Basin and the WW2100 model

The Willamette River Basin encompasses 11,500 square miles in northwest Oregon, or 12% of Oregon’s land area (Figure 1). The Coast Fork of the Willamette River (originating in the Coast Range), the Middle Fork of the Willamette River and the McKenzie River (both originating in the Cascade Range) join near Eugene, Oregon to form the mainstem Willamette River. From this confluence, the Willamette flows north between the Cascades and the Coast Range for 283 river kilometers (176 miles) and enters the Columbia River near Portland, Oregon. The Willamette Valley, an agriculturally intensive region, contains Oregon’s three largest cities — Portland, Salem and Eugene — home to over 60% of the state’s population.

The Basin’s underlying geology creates a setting in which various drivers of landscape change may impact hydrological processes. Thus, the Basin provides a rich and varied environment for studying the potential impacts of climate and human change on water resources.

The Cascade Mountains form the eastern border of the WRB. In this region, glaciers and extensive areas of mid-elevation snowpack are underlain by layers of low- and high-permeability volcanic bedrock. On the western border of the WRB, the Coast Range includes steep slopes underlain by low-permeability sedimentary and volcanic rock. This region receives more than 2,500 millimeters (98 inches) of rain per year, but little snowfall.

Tributaries of the Willamette traverse the region, providing the opportunity to examine streamflow response across a range of climatic, geologic and ecological gradients and boundaries. After the McKenzie River, the largest tributaries are the Clackamas River, the North Santiam River and the South Santiam River.

The WW2100 project developed a computer-based model of the Basin’s human and natural system components across time and space. (Willamette Water 2100 Project involved more than 40 researchers; see a list of the entire project team.) The large number of detailed relationships and processes makes this model innovative and exceptional. The model includes physically and empirically based submodels of the biophysical system, as well as empirically based economic submodels of the human system. These submodels are linked using a simulation software platform, Envision.

The model’s main components and linkages are represented in Figure 2. Processes determined outside the model (exogenous to the model) include daily temperature, precipitation, humidity, wind and radiation (derived from downscaled regional climate data) and annual changes in population and income. Downscaling is a procedure whereby climate model projections for a large area (say 100 square miles) are converted into sets of estimates at a higher resolution (for example, each 10-square mile area within the 100-square mile region).

The model’s economic components incorporate human behavior. The use of land and water for different purposes reflects the multitude of decisions that farmers, firms and households make on a daily basis. Thus, the economic models incorporate choices and responses to prices in order to predict how agricultural and urban water withdrawals will vary by year and by season.

In the case of agriculture, evolving temperature and precipitation conditions affect plant growth, daily evapotranspiration (ET) and soil moisture. In turn, these factors affect decisions about crop choices, planting dates and irrigation. In urban areas, population and income growth lead to increased water use, while other changes, such as higher water prices or increased urban density, limit the rate of increase.

The WW2100 climate projections are based on three General Circulation Models (GCMs) downscaled to 4-kilometers resolution. The model’s reference scenario uses MIROC5. The “high climate change” scenario uses HadGEM, and the “low climate change” scenario uses GFDL.

The Basin’s hydrology is represented by stocks and/or flows of rain, snow, soil moisture, groundwater and streamflow in each of 160,000 landscape polygons, an overlapping network of river reaches and nodes and 13 U.S. government reservoirs (11 of which store significant volumes of water). Daily temperature, precipitation and atmospheric humidity determine surface hydrology, including snowmelt, ET, streamflows and water temperatures.

The relationships between these hydrologic factors generate daily average streamflow and reservoir levels throughout the Basin’s network of stream reaches and reservoirs, all the way to the Columbia River. A reservoir model simulates reservoir fill and discharge to meet federal flood control, storage and streamflow targets. All irrigation, municipal and in-stream water rights are fully represented by a detailed submodel reflecting their point of use, point of diversion, priority date, maximum rate, duty (maximum total annual diversion) and beginning and end date.

Forest water use is modeled on a daily basis according to estimated ET rates, which vary with evolving forest characteristics (stand age, species type) and meteorological conditions. Forest fire and forest harvest models introduce disturbances to forest land cover and stand age.

At a fine spatial scale, the model simulates the period 2010 to 2100, with some processes adjusting annually and others taking daily time steps. Model processes that follow an annual time step include forest growth, harvest and wildfires; the determination of land values and land-use changes; regulatory adjustments to urban growth boundaries; crop choices; and irrigation decisions. Daily time-step processes include routing of surface hydrology throughout the Basin’s stream network; water use (ET) by forests and other vegetation; and timing of crop planting, crop growth, irrigation diversions, urban water use, soil moisture, groundwater flows and reservoir management.

Projected changes in climate, population and land use, 2010–2100

The WW2100 model, as described above, can simulate current or recent patterns of water supply and use. However, its main purpose is to simulate the future. In our model, future years will differ from current years due to three factors: changes in climate, changes in population and changes in household income.

The model’s reference scenario generates trajectories of change for 2010–2100 in response to changes in climate, population and income. The reference scenario is intended to reflect the most likely trajectory of change. It is based on “business-as-usual” assumptions about the existing system, including midrange estimates for exogenous drivers such as climate change, population growth and rising incomes.

The reference scenario is one of more than 20 scenarios (see Appendix). Other scenarios include five different types of alternative scenarios developed for different purposes.

Alternative scenarios can be very useful in helping to attribute outcomes to specific factors. They are also useful for policy analysis. The comparison of a reference scenario to a scenario that represents a policy change can provide insights into “what-if” questions about the costs, benefits, or effectiveness of public policies. See the Appendix for more details.

Natural system changes

The climate models used in WW2100 generate daily temperature, precipitation and other variables from 2010 to 2100, based on major General Circulation Model (GCM) outputs. The WW2100 model includes three different climate scenarios. Results indicate that by the year 2100, temperatures in the WRB will rise by 1° C to 7° C (2° F to 13° F). Summer temperatures are projected to warm about 2° C (3.6° F) more than winter temperatures.

In the case of precipitation, the three climate scenarios indicate that winters will become slightly wetter and summers slightly drier (Figure 3) . However, there is no consensus, based on examination of more than 40 climate models, about whether the Basin’s climate will become wetter or drier overall.

Where streamflow depends on snowmelt, this loss of “natural” water storage will result in flows that are smaller and come earlier in spring and summer than has been the case historically (Figures 5 and 6). However, spring precipitation plays a much larger role than snowpack in determining spring and summer flows in the WRB. Thus, the reduction in snowpack likely will have little effect on the supply of water for human uses in the lower basin. In Figure 7, we compare the 10-year-average midsummer (July and August) discharge for each subbasin. In both the reference scenario and the high climate change scenario, we see no significant decline in average summer flows.

County-level population projections to 2050 are taken from the Oregon Office of Economic Analysis. Income projections to 2040 are from Woods & Poole Economics, Inc.. Beyond 2050 and 2040, population and income are assumed to increase at the historical average annual rate. Income figures represent mean total personal household income expressed in “real” (inflation-adjusted) 2005 dollars. Projected trends are described in more detail and compared to historical trends in the Appendix.

The land-use model is based on historical data from the National Resources Inventory (NRI) and county assessors’ offices. The land-use models incorporate the economic returns to forest, developed, and agricultural uses for each parcel, or IDU (Integrated Decision Unit). Relevant characteristics of IDUs include distance to the nearest city, population density and household income in the nearest city, and soil quality. The calculated value for each use changes over time with evolving IDU characteristics. For example, growth in population and income increases the returns to developed uses of land, relative to forest and farm uses.

The functions that estimate the economic returns to each land use were derived from historical data for the Willamette Valley. In the same way that population, income and climate are treated as exogenous, we assume that prices for agricultural and forest commodities, which influence forest and agricultural land values, are exogenous, since they are determined in global markets. We assume that they will remain at current levels in real, inflation-adjusted terms, since we have no specific basis for assuming that they will rise faster or slower than other prices or wages.

During a model run, land-use changes may occur annually at the IDU level; in response to changes in the relative returns to different uses, IDUs may shift among agricultural, forest and urban uses. Urban expansion is regulated by Oregon’s statewide land-use planning system, which requires the use of urban growth boundaries (UGBs) to guide the location of urban development. It is assumed for the reference scenario that future land development will take place inside UGBs, although UGBs can expand (that is, IDUs are added to a UGB area) when the amount of developed land within the UGB exceeds a specified percentage.

Population density is an important determinant of developed land values. In many of the major cities in the Basin, population density is projected to increase under the reference scenario, although in some cases it remains relatively constant (Figure 12). Increasing urban populations and population density will lead to rising values for developed land (Figures 13 and 14), thus increasing the conversion of land to developed uses. Other contributing factors to land conversion are income growth, which raises per-capita consumption of land for housing, and expansion of UGBs, which makes more land available for development.

Urban water use and pricing

Water in urban areas is put to residential, commercial and industrial uses. The amount used depends on a range of factors, including population, price of water, income, population density and the use of water-conserving technologies such as low-flow toilets. Many empirical studies based on household, city-level, or national data have estimated urban water demand and identified the most important factors affecting municipal water use (see Appendix).

The WW2100 model’s water demand relationships are used to generate estimates of the quantity of water expected to be used in each urban area in future decades. It is important to recognize that these results reflect underlying assumptions about population, income and water price trends. Because multiple changes may affect the economics and demographics of urban water demand, these projections can be interpreted only as suggestive of a plausible future path. For example, if urban water prices rise faster than model assumptions, urban water use will be lower than the model predicts. In this case, water supply capacity might be more than sufficient to meet demand.

Urban water pricing

Unlike competitive markets, urban water prices are regulated by local governments. Urban water utilities set water prices to achieve multiple goals. These goals include generating revenue to cover costs, assuring affordability, providing stability in revenue and achieving an allocation of cost that is considered “fair” to various types of ratepayers. Utilities may wish to fully cover all costs, while at the same time providing customers with efficient and transparent incentives to conserve water.

Urban water delivery systems are highly capital intensive, with a need for large investments in infrastructure (building, maintenance, or replacement) on an intermittent basis, sometimes decades apart. In a typical year, customers may not be aware of these capital costs, making it difficult to set prices so that long-term costs are covered. As a result, water prices tend to be somewhat lower than long-run average or marginal costs, leading to financial deficits and delays in infrastructure investment, repair and replacement.

This situation is well documented in historical data, surveys and engineering analyses. U.S. Environmental Protection Agency (EPA) survey data, for example, indicate that average water prices are frequently more than 20% below long-run average cost. The gap between average price and average cost has fluctuated over time and across cities and states.

Water prices in major U.S. cities have at times risen more slowly than inflation. In the past 20 years, however, they have risen faster than inflation. Nationwide, an extended period of declining urban water prices (inflation-adjusted) lasted until about the 1980s. Since that time, rising urban water prices have been observed.

Since 2010, urban water prices have increased in the WRB. Increases have been due in part to the need to expand or upgrade infrastructure. In some cases, utilities face new requirements to implement seismic risk reduction upgrades. As a result, urban water prices in some parts of the Basin increased more than 30% between 2010 and 2015.

In a separate survey, the EPA has documented the backlog of infrastructure needs for drinking water systems nationwide. Nationwide, the “20-year need” totaled $376 billion in 2011. This backlog has fluctuated, but has been rising, from $843 per person in 1995 to $1,205 per person in 2011.

The backlog of infrastructure needs in Oregon has generally been higher than the national average, rising from $1,108 per person in 1995 to $1,442 in 2011. Since most urban populations in Oregon live in the WRB, the model uses Oregon-wide data as a reasonable indicator of the situation in the Basin. The exception was in 2007, when Oregon’s per-capita need was $845, compared to the national average of $1,220. The reduced backlog of infrastructure needs in Oregon between 2003 and 2007 followed a 53% increase in average prices in Portland between 1999 and 2007, which likely financed significant infrastructure improvements.

Predicting urban water price trajectories is complicated not only by long-term infrastructure investment costs, but also by the regulatory and political factors involved in setting water prices, which often cause price increases to lag behind cost increases. Furthermore, when prices are raised to cover higher costs, the effect can be counterproductive to some extent; higher prices may lead to reduced water use, thus reducing the anticipated increase in revenues.

Our reference scenario assumes initial prices that reflect prices in the Basin’s major cities in 2010. Future price trends reflect observed recent price increases in many cities (as of 2015), as well as the fact that the backlog of infrastructure needs in Oregon is relatively high. To represent observed price increases, the model implements annual price increases of 6% from 2011 to 2015 (in real, inflation-adjusted dollars). To reduce estimated system needs over the next 20 years, from $1,442 per person to $1,050 (the national average observed since 1995), annual per-person revenues would have to increase by an additional $40, or more than 25%. Thus, average water prices are assumed to increase 1.5% per year from 2016 to 2025. The result is a cumulative price increase between 2010 and 2025 of 55%. After 2025, urban water prices (in inflation-adjusted dollars) are assumed to change only in proportion to changing costs. Figure 17 shows these price trends for the nine largest urban areas.

Total urban water use

Population growth is one of the main drivers of increasing urban water use. Another factor that affects urban water use is income growth. As incomes rise, people tend to use somewhat more water (for example, with bigger houses, yards and gardens). The model’s water demand relationships are based on peer-reviewed economics research, including more than 100 published studies of urban water demand.

Although it is sometimes thought that urban water use is not responsive to changes in price, dozens of economic studies have shown that long-run responsiveness to price is substantial. Indeed, on average, a 40% increase in the price of water can be expected to result in approximately a 24% decrease in water consumption. See the Appendix for more detail on the impact of price and other factors on water demand.

For four major urban areas (Portland Metro, Salem, Corvallis and Eugene-Springfield), the WW2100 model consists of separate models for residential and nonresidential urban water demand. For other cities, residential and nonresidential demand are combined.

For 2015, the model estimates total annual urban water withdrawals of about 330,000 hundred cubic feet per day (272 million gallons), or 305,000 acre-feet per year. Our projections show this total rising in coming decades for the Basin as a whole, especially for the Portland Metro area, mainly due to population growth (Figures 21, 22 and 23).

Consumptive use of WRB surface water

Consumptive use of surface water refers to water that is not returned to its source; it is lost to ET, evaporation, or groundwater. In contrast, nonconsumptive use refers to water that is returned via wastewater treatment facilities to streams. Three important factors affect urban consumptive use of surface water in the WRB.

First, urban water use varies significantly between indoor and outdoor uses. Outdoor use is largely consumptive. By contrast, indoor water use is mostly nonconsumptive; the amount of water returned to streams is roughly equal to the amount diverted. Thus, the net use of water in urban areas is significantly less than total diversions (Figure 27).

Second, urban water use shows seasonal patterns that reflect the rise and decline of outdoor water use. Whereas indoor water use is assumed to be evenly distributed throughout the year, outdoor water use is assumed to begin at low levels in April, peak in July and decline to zero in October (Figure 28, 29 and 30). In the summer, about 40% of urban water use is outdoors (consumptive). The seasonal pattern of urban outdoor water use is based on 24 years of data from the Portland Water Bureau. It is assumed that the seasonal distribution of total demand will not change in future years. For the six largest metropolitan areas in the Basin, our model predicts an increase of 36,800 acre-feet per year in summer outdoor (consumptive) use (Table 1).

Finally, it is important to recognize that nearly half of urban water demand in the WRB is met by water imported from out-of-basin surface-water sources. Most of this water comes from the Bull Run Watershed, located 25 miles east of downtown Portland in the Sandy River Basin on the Mt. Hood National Forest. Bull Run supplies all of the drinking water for the City of Portland and some water to other cities in the Portland Metro area. From June through October, additional water is supplied to the western Metro area from two other out-of-basin sources, Barney Reservoir and Scoggins Reservoir, both on the western side of the Coast Range. The WW2100 model does not include the climate and hydrology of these out-of-basin areas. Our implicit assumption is that these sources will continue to provide the quantities of water authorized by the corresponding municipal water rights.

As a result, consumptive urban water use from in-basin surface-water sources is less than 10% of total urban water deliveries. These different measures of urban water use are reflected in Figures 27 through 30.

Effects of climate change on urban water demand

Climate change could affect urban water demand, primarily due to the effects of changing temperatures and precipitation on lawn and garden irrigation, as warmer temperatures may increase the ET rate of grass, flowers and trees.

Our urban water demand model does not include any adjustment for the direct impacts of climate change on urban vegetation and outdoor water use. However, the agricultural growing season (March–September) corresponds to the seasonal increase in urban water use; thus, we can make inferences based on the effects of climate change on agricultural ET and water use. The model includes ET for grass seed; orchards, vineyards and tree crops; and a broad category of “other crops.”

In the reference scenario, the trend in “maximum ET” (the total seasonal ET that occurs if plants always have adequate soil moisture) for agricultural crops is flat, fluctuating slightly around an average of 425 millimeters (16.7 inches). In the high climate change scenario, maximum ET for agriculture increases about 2% between the beginning and end of the century. For grass seed and “other crops,” maximum ET decreases slightly. By contrast, the model shows a 24% increase in maximum ET for orchards, vineyards and tree crops.

Assuming that ET for urban vegetation is similar to that of crops, these estimates suggest that climate change will not lead to a significant increase in urban water use in the WRB by the year 2100. Data suggests that the effects of climate change on urban water use would be no more than 1 or 2% by late in the 21st century.

Furthermore, the increase in ET for orchards and trees, if comparable to urban tree cover, could be partially offset by planting species with lower water requirements.

Effects of reduced agricultural irrigation on urban water supply

Urban growth will to some extent displace agriculture, including some irrigated lands. The model assumes that irrigation water rights are relinquished or converted to municipal water rights when farmland is converted to urban development. These water rights will help meet the growing demand of cities.

The model projects that displacement of surface-irrigated farmland could offset about one-third of the increase in urban consumptive (outdoor) water use, reducing the net increase to an estimated 24,000 acre-feet.

This effect varies significantly among cities, depending on the extent and direction of urban expansion, as well as on the proximity of the city to surface-irrigated farmland. In Albany, for example, the offset is only 20%. In McMinnville, reductions in surface irrigation will more than offset increased urban water use (see Table 1).

Scarcity in urban areas

City governments are understandably concerned about how the growing demand for water will be met. “Live” or natural surface-water flows are already fully appropriated, and currently federal reservoir storage can be allocated only to agriculture. Cities will also compete with in-stream water rights and regulatory flows established under the Endangered Species Act. These requirements represent a large proportion of summer flows and may increase (see “In-stream flows”).

The results presented in Figure 31 suggest that currently utilized municipal water rights may reach capacity in the Metro area in 30 years. Demand (primarily in summer months) may exceed existing capacity by about 12,000 acre-feet annually by the end of the century. Our model suggests that Salem may reach the capacity of currently utilized water rights in 60 years and may require an additional 3,000 acre-feet per year by the end of the century.

Nevertheless, it appears that municipal water rights will be adequate to meet nearly all of the increased water demand expected through the year 2100 (see Appendix for details). One factor is that many cities rely on out-of-basin water and/or have multiple water rights, some of which are not currently utilized. Unutilized or underutilized sources may include surface water, groundwater and aquifer storage and recovery. (Cities also plan strategically to have extra or redundant water rights for unexpected circumstances). Not all of these water rights are included in our model. Furthermore, the Tualatin Valley Water District is constructing a large new water supply system, which will draw water from the Willamette mainstem. This water right is not included in our model. Scheduled to be completed in 2026, this source will be able to deliver 100 million gallons (more than 36,000 acre-feet) per day during the four peak summer months. It is designed to serve more than 300,000 residents in Beaverton, Hillsboro and Tigard.

Conditions vary across cities and towns in the WRB and will often differ from the overall or average results summarized here.

Water use in agriculture

More than 800 significant crops are grown in the WRB by some OSU estimates. Two crops stand out in terms of land area — grass seed, which is grown on about one-third of farmland, and pasture, which accounts for one-sixth of farmland. The dominance of grass seed on an acreage basis has been relatively constant for more than 50 years. No other crop is grown on more than 5% of the Basin’s acreage. Crop choices have changed in the Basin and are likely to continue to do so. Nursery crops have become more important, and, as of 2016, significant new hazelnut acreage is being planted.

In Oregon, as elsewhere in the western U.S., agriculture is the largest out-of-stream human use of water, accounting for about 80% of out-of-stream use. Unlike many parts of the western U.S., however, most WRB farmland is not irrigated. A majority of the 1.5 million acres of farmland in the WRB rely directly on precipitation, rather than irrigation. Most of the major crops are grown both with and without irrigation. Exceptions include corn, which is always irrigated, and winter wheat, which is almost never irrigated. About a quarter of grass seed is irrigated, as is 10 to 20% of pasture.

About one-third of WRB farmland has irrigation water rights. Slightly more than half of these are surface-water rights; the rest are groundwater rights. Growth in the acreage of farmland with irrigation water rights began in the 1940s and leveled off in the 1990s (Figure 32). The location of farmlands with irrigation water rights (both surface and groundwater) is shown in Figure 33.

Irrigation practices in the WRB are unusual, compared to other irrigated areas in the West, in that in any given year only about two-thirds of irrigation water rights are utilized. Of the total acres with water rights, only 60% or less are irrigated in a given year. See Figure 32.

Over the past 20 years, irrigated acreage in the WRB has been flat. The area irrigated varies from year to year, but averages less than 300,000 acres, or 20% of the region’s farmland (Figure 32). This stability is not surprising, given the dominance of grass seed and pasture. No other crop is planted on enough acreage to have a significant effect on overall crop water use.

Crop choice

On a parcel of land with an irrigation water right, the choice of crop and whether to irrigate is based on several factors (see Appendix). Some factors do not vary from year to year (for example, soil type, elevation and average precipitation and temperature). Other factors are unpredictable, such as crop prices, costs of fertilizer and energy and spring rains. As a result, the number of acres planted to a given crop and the number of irrigated acres varies slightly from year to year.

Our modeling of crop choice and irrigation decisions is based on established economic theory, empirical data and a detailed farmer survey conducted in the WRB (see Appendix). Although the model allows acres planted to most crops to fluctuate from year to year, land planted to tree crops (Christmas trees), orchards and vineyards is not allowed to vary.

Given the WRB’s strong climatic advantages for growing seed crops, the most likely future scenario is for grass seed and pasture to continue to dominate acreage, with other crops entering and exiting the crop mix. The projected pattern of crop mix for all agriculture, and for irrigated lands specifically, is shown in Figures 34 and 35.

Irrigation water use

The amount of water required for irrigation is expected to be relatively stable, with a slight decline in both surface and groundwater irrigation. The projected decline is due to land-use change resulting from urban expansion and the corresponding loss of irrigated acreage.

Even if new crops enter the crop mix, they are unlikely to have more than a small effect on average ET basin-wide. To produce a significant change, multiple new crops would have to have extremely high or low ET and displace current crops on a large number of acres.

One important question is whether climate change will increase irrigation water demand due to the effects of warmer temperatures on crop ET. At a given crop development stage, higher daytime temperatures lead to higher ET. However, our model indicates that warmer springtime temperatures will allow farmers to plant earlier, irrigate earlier and harvest earlier. Thus, more plant growth will take place during months with relatively cooler temperatures, more precipitation and higher levels of soil moisture. As a result, the positive effect of warmer temperatures on ET may be offset by the opposite effect from earlier planting, thus reducing, rather than increasing, irrigation.

What results are indicated by our model? The model simulates daily crop water demand from planting to harvest. As a result, it can capture the direct and indirect ways that climate change will affect crop water use and irrigation demand. Even in the high climate change scenario, the model shows only a 2% increase in maximum ET from 2010 to 2100.

There is, of course, the potential for expansion of irrigation onto currently unirrigated lands. However, “live flow” surface-water sources are already fully appropriated in the Basin. Moreover, some parts of the Basin have seen declining groundwater levels and face limitations on groundwater withdrawals. One area in southeastern Washington County has been designated a “critical” groundwater area.

The projected reduction in irrigation does not take account of the possible addition of irrigation water rights from federal reservoirs. To date, farmers have contracted for only 80,000 acre-feet, or about 5% of this stored water. The potential for use of federally stored water is discussed below (see “Stored water”).

Agricultural water supply and demand in alternative scenarios

Water use in agriculture may vary from the levels suggested by the reference scenario. If changes in prices, crops, or technology make irrigation more profitable, the share of irrigation water rights that goes unused each year could decline. An increase in contracts for stored water from reservoirs might also result in increased irrigation. Conversely, higher energy costs could make irrigation less attractive, or lower density urban expansion could displace more irrigated farmlands.

Stored water

The potential for using some of the 1.6 million acre-feet of water stored in federal reservoirs for new irrigation has been a topic of interest since the 1990s. Farmers may apply for a contract to use this water, provided that the water can be diverted downstream and transported to a farmer’s field for irrigation purposes. Both an Oregon water right and a contract with the U.S. Bureau of Reclamation (USBR) are required. Currently, contracts cover only about 80,000 acre-feet of stored water. The amount authorized for potential use under such contracts is currently limited by Biological Opinions under the Endangered Species Act to a maximum of 95,000 acre-feet, as discussed in the Appendix 6.

To evaluate this potential, a scenario was developed that makes new contracted water available for irrigation on lands that currently do not have irrigation water rights. The model introduces an annual probability of acquiring a water right for stored water, where the probability is a function of the marginal value of irrigation (which varies by soil class) and the cost of conveyance to the field in question.

Estimated conveyance costs (infrastructure and pumping) are based on distance and lift from the nearest point on the relevant river. (See Appendix for details.) Conveyance costs are high for moving water to a farmer’s field, and economic returns to doing so are relatively low. Thus, the addition of irrigated lands is limited to those areas in close proximity to streams below federal reservoirs. Estimated conveyance costs for agricultural lands, and the areas that are able to profitably adopt new irrigation water rights, are shown in Figures 38 and 39. The model indicates that only 7,200 additional acres are likely to add irrigation water rights from stored water. This acreage is more than offset by the reduction in surface-irrigated acres resulting from urban expansion under the reference scenario. As a result, total irrigated acreage would change only slightly.

If we modify the scenario to reflect optimistic assumptions about the costs of conveyance, and if we eliminate the price irrigators must pay to the USBR for water contracts, the number of irrigated acres would increase by 27,400, representing about 55,000 acre-feet of stored water.

Profitability of irrigation

In one scenario (high irrigation), an increase in the profitability of irrigation reduces by half the number of acres that go unirrigated in a typical year. This change would increase irrigated acres by 86,000, or 32%. Because of this increase in demand, competition among irrigators would be expected to increase somewhat, with a modest increase in the frequency of irrigation shutoffs (see “Water scarcity in agriculture”).

Fallow

All agriculture in the WRB uses significant amounts of water. Since 80% of agricultural land is not irrigated in a given year, the amount of water used in rainfed farming is important, since its use by crops makes it unavailable for irrigation or other uses. In fact, total annual ET per acre for nonirrigated crops is, on average, slightly higher than that of irrigated crops, likely owing to crop choice and ET associated with winter groundcover.

However, a decline in rainfed farming would not necessarily result in an increase in streamflows and water available for other uses. The reason is that fallowed land will have vegetation on it. The ET for this vegetation is not significantly different on average, in our analysis, from that of crops such as grass seed and pasture. Even when all agricultural land is constrained to be fallowed, the total ET for agriculture in our model does not change significantly.

If ET did decline with increased land fallowing, streamflows would increase slightly. However, some of the moisture that would have been consumed by a crop would remain in the ground for a period of time. For this reason, it might not contribute additional water to surface flows at the time water is needed.

Water scarcity also depends on whether a farmer possesses a water right and whether that water right has “priority” over other water rights that may compete for the same water. Under Oregon water law, irrigation water rights are based on the seniority system common to most western states, known as “the prior appropriations doctrine.” Irrigation water rights are tied to a specific parcel of land on which the water can be put to a “beneficial use.” Each water right has a priority date corresponding to the first use of water on that land. The most senior water rights predate 1900. See Appendix 6 for more details on water rights.

In a time of water shortage, a senior water right holder (a person having a right with an older priority date) can “make a call,” requiring relatively junior water right holders to stop diverting water from a common stream so that the remaining water will reach the diversion point for the senior user.

A junior water right holder who is “regulated off” usually cannot irrigate for the remainder of the year. If crops have been planted that require irrigation, the loss of irrigation leads to economic losses that exceed what would have been gained by completing the irrigation season. For example, costs for planting, fertilizing and partially irrigating the field will already have been incurred.

In-stream flows

Under Oregon state law, all water rights must serve a beneficial use. In Oregon and some other western states, beneficial use includes in-stream flows to protect habitat for salmon, steelhead and other native fish species, as well as for ecosystem services and other purposes.

In the WRB, in-stream water rights were established on various stream segments in the 1960s in order to provide minimum recommended perennial streamflows to protect fish habitat. Some of these water rights have been implemented or “converted” to certified water rights, that is, put into place and enforced. Our reference scenario includes 93 certified in-stream water rights.

Many of these water rights, however, have not yet been converted to certified water rights. The OWRD has initiated a process for converting the remaining water rights. These additional regulatory allocations of water for in-stream flows represent a significant “new” use of water for many of the Basin’s main tributaries. In some cases, conversion involves an increase in the minimum flows associated with existing in-stream water rights.

An alternative scenario (“new in-stream”) assumes the conversion of in-stream water rights and gives them priority dates corresponding to their creation, typically in the 1960s. The implementation of these unconverted in-stream water rights is consistent with the Instream Water Rights Act. Full implementation represents an additional commitment of surface water from April through August of 1.1 million acre-feet. This total is derived by summing the flow requirements at the outlet of each main tributary, some of which would increase.

Some of these streamflows, however, are already protected by existing in-stream water rights under state law. In other cases, the minimum flows coincide with the operations of federal dams, where flow targets are already mandated under Endangered Species Act (ESA) Biological Opinions (BiOps). (See Amos, 2013, for details about relevant state and federal laws.)

The 2008 final BiOps for ESA-listed salmon and steelhead in the Willamette Basin establish minimum in-stream flows below federal reservoirs in the WRB from April through October (NMFS, 2008). Required flows are highest from April through June. These flows have built-in flexibility, whereby required flows are reduced in years considered to be “deficit” or “insufficient” water years. This determination is based on reservoir fill levels.

The BiOp flow requirements are tied to downstream control points (see Appendix). Existing required BiOp flows are higher than those that would be required under state law by conversion of all remaining in-stream water rights. Although the addition of new in-stream water rights and continued BiOp flow requirements represent challenges for federal and state water managers, our model results suggest that these flow requirements can be met, based on average 10-year flows. Exceptions may occur in severe drought years.

The system of 13 reservoirs that comprise the United States Army Corps of Engineers (USACE) Willamette Project is one of the primary mechanisms used to mitigate water scarcity in large parts of the WRB. Although the Willamette Project reservoirs were built primarily for flood control, they fortuitously have a large capacity (1.6 million acre-feet in total) to store water from abundant winter and spring streamflows. This stored water is available for use during the summer, when natural flows are low.

Flood damage reduction remains the priority authorized use of these reservoirs. Nonetheless, stored water uses have become increasingly important. These uses include reservoir recreation and the augmentation of downstream flows for endangered species and irrigated agriculture. By increasing mainstem flows, these releases also indirectly contribute to urban water supplies.

When water is released to maintain reservoir capacity to buffer storm events, that water is not available for later use. Thus, flood mitigation and water storage become competing objectives as reservoirs fill during the transition from the wet to the dry season.

Lost recreational benefits across the Willamette Project reservoirs are estimated to remain relatively stable until late in the 21st century. By the 2080s and 2090s, these foregone benefits are projected to increase, to more than $12 million per year in the reference scenario. Under the high climate change scenario, they exceed $13 million per year (Figure 46). These losses represent a 5 or 6% decline in total recreation visits, respectively.

Given the tradeoff involved in managing reservoirs for both flood reduction and stored water uses, these lost recreational benefits should be considered in light of the estimated value of flood damage reduction. The USACE estimates that, as of June 2015, the Willamette Project reservoirs have prevented more than $23 billion in flood damages since their completion in 1969. Based on analyses undertaken to complement the WW2100 effort, the current annual value of reservoir buffering capacity is estimated at more than $1 billion. This estimate is based on avoided flood damages to downstream developed land, buildings and their contents (Figure 47). It reflects population and economic growth, but does not account for changes in flood risk due to climate change.

With the levels of economic growth and urban expansion projected under the reference scenario, this benefit is expected to triple by 2100. Under the high population scenario (Figure 47), it increases more than fivefold.

Subbasin differences

The 11 WRB subbasins differ in size, acreage of farmland, developed land area (Table 2), precipitation patterns, changes in forest water use due to harvest and wildfire and other ways. Thus, patterns of water supply and use vary considerably and will continue to do so in the future. This section examines the differences across subbasins, how those differences are likely to change in future decades and what implications they may have for water scarcity at the subbasin scale. For this discussion, we are omitting the Willamette mainstem as a subbasin.

It is important to keep in mind that the hydrology of subbasins in the eastern half of the WRB differs significantly from that in western subbasins. In the eastern half of the WRB, snowmelt, reservoir storage and summer in-stream flows are greater than in the western subbasins. As indicated in Table 2, several of the eastern tributaries have minimum summer flow requirements of 520 to 2,100 cubic feet per second, while several western tributaries require flows of only 30 to 150 cubic feet per second.

The scale of agricultural land and water use differs considerably among the 11 subbasins. The largest agricultural acreage, as a percentage of total land area, is in the Yamhill and Molalla subbasins, followed by the Tualatin and Long Tom subbasins. By contrast, very little agriculture is found in the McKenzie, Middle Fork Willamette, or Coast Fork Willamette subbasins (Table 2).

We are especially interested in identifying which subbasins have the most irrigation and in particular those with large-scale surface irrigation, since surface-water irrigation represents the largest out-of-stream human consumptive use of water during the summer when water may be scarce. Table 2 reveals that the Yamhill subbasin has the most surface irrigation, followed by the Tualatin, Molalla and Long Tom subbasins.

Developed urban land is greatest in the Tualatin subbasin, followed by the Molalla and Long Tom subbasins. A large fraction of developed land is not represented in subbasin data, however, because developed lands are concentrated along the mainstem (for example, portions of the Portland Metro area and Salem). The area designated as the Willamette mainstem subbasin also includes substantial areas of farmland and irrigation.

The period when water may become scarce is in spring (when dams are being filled and in-stream BiOp flow requirements are highest) and summer (when flows are lowest and out-of-stream use by agriculture and urban populations is highest). Ideally we would like to compare spring and summer flows in each subbasin to current and future demand for water. However, estimating urban water use by subbasin is difficult, as some major urban areas straddle multiple subbasins. In some cases, cities draw water from only one of these subbasins or from outside the WRB (for example, the City of Portland and some cities on the west side of the Portland Metro area). Moreover, because much of the water used in urban areas is returned to its original surface-water system, quantifying the net consumptive use from a given subbasin’s water sources is not straightforward.

In the case of irrigation, the number of acres irrigated provides a rough idea of the demands placed on subbasin water sources. Each acre of irrigated land diverts 1.5 to 2 acre-feet of water during a season, typically between April and August.

In the reference scenario, subbasins vary greatly in projected changes in April–August streamflows. For some subbasins, April–August flows are projected to decline by as much as 7.3%; for others, the model results indicate increased flows of up to 17% (Table 2). In the case of the Tualatin, discharge from urban water systems that draw on out-of-basin reserves (Barney and Scoggins Reservoirs) is expected to increase. It is noteworthy that all of the subbasins showing decreased flows have their headwaters in the Cascades, whereas nearly all of those with increased flows are on the west side of the Basin.

Another way to look at changing subbasin water supplies is in terms of the changes in stream flow during spring (April–June), shown in Figure 48; and in summer (July–August), shown in Figure 49. We see small decreases in spring flows in several subbasins, although the pattern varies greatly.

The largest commitment of surface water in most subbasins is the allocation of water for in-stream water rights under state law (separate from the similarly large quantities of water allocated under federal BiOp flows). In-stream water rights are largest at the confluence of the Middle and Coast Forks of the Willamette River, both for the April–September irrigation season and for the July–August low-flow period. The McKenzie has the second highest total in-stream water rights. Based on unconverted in-stream water rights, the North and South Santiam rivers follow.

One of the largest factors affecting future water scarcity in some subbasins will be the conversion of perennial minimum streamflows into binding in-stream water rights. The largest increases are for the confluence of the Coast and Middle Forks of the Willamette and the McKenzie, each with increases of more than 600,000 acre-feet from April through August. An important factor for those tributaries with federal dams is the ability to control flows by modifying management rules for dam releases.

Conclusions

Our understanding of water supply relies heavily on hydrology and related natural sciences, whereas our understanding of the demand for, and allocation of, water comes from economics, law, engineering and related social sciences. A hydro-economic model like the one presented here makes it possible to represent the important processes of both the natural and human systems. The WW2100 model sheds light on how these components interact and helps us understand where, when and why we may see more water scarcity in the future.

The projections in this report for both urban water use and agricultural water use are based on the set of behavioral economic models described here and elsewhere. These models reflect and are derived from economic theory. They are spatially and temporally explicit and take account of many factors, including the following: water price, household income, population, population density, water delivery costs, land values and farm profits, land-use change, crop choice, planting date, water availability across space and time shifts in seasonality of crop growth due to climate change, daily determination of crop ET, and utilization rates for irrigation water rights. The 2015 Statewide Long-Term Water Demand Forecast Report, prepared by the consulting firm MWH for Oregon’s Water Resources Department, also estimates future water demand in Oregon. Their model takes account of a far more limited set of factors than the WW2100 model: (1) In the case of agriculture, the MWH report draws on USGS estimates (which in turn are based on USDA Census of Agriculture data) for irrigated acres by county and by crop. Irrigation water demand is estimated based on average net irrigation water requirements, which are then adjusted to reflect the effects of climate change. (2) In the case of urban water demand, MWH relies on existing Water Management and Conservation Plans (WMCP), developed by various city governments. These plans were then adjusted in proportion to estimated population growth only. Changes in per-capita demand were estimated from 50 of the most recent WMCPs from communities across Oregon.

Changing supply

On the supply side, the severe decline in snowpack in the next 80 years will reduce the amount of snowmelt runoff from April to June. The projected reduction in average available snowmelt (as of April 1 each year) represents a decline of about 600,000 acre-feet of stored water. However, precipitation plays a far greater role than snowmelt in determining spring streamflows in the WRB. Indeed, the projected decline in snowmelt is only one-tenth of average April–July precipitation (nearly 7 million acre-feet).

As a result, unlike arid basins in eastern Oregon or neighboring western states, the loss of snowmelt will have a relatively small impact on water availability in the lower elevations of the WRB in spring and summer. Nevertheless, reduced snowpack, when combined with higher summer temperatures, is projected to increase stress on upland forests and, as a result, increase the risk of wildfire.

To the extent that drier forests result in more frequent wildfires, the consequent changes in forest cover will reduce forest water use (ET) and allow more surface water to flow into the Willamette Valley. Depending on the extent of wildfires and on fire suppression policies, these changes in forest cover may have a greater influence on water supply to the lower Basin than the projected reduction in snowmelt.

Changing demand

The economic forces related to where and when water is used or not used may at times be overlooked, in part because they are not visible or immediately recognizable. The locations of various types of economic activity are influenced by land markets, and these individual and societal choices in turn influence the demand for water at each location. Our model projections find that from 2010 to 2100 developed land as a percentage of total land area in the WRB will rise from 4.7% to about 7.2%, a 54% increase. Agricultural land is projected to decline from almost 22% to 20.2%, and forest land from 70.6% to 69.7%. The decline in agricultural lands is about 7.5% for farmland overall and about 5% for irrigated lands.

At each location, given a particular land use, water use is influenced by both the demand (willingness to pay) for water and the cost of transporting, storing, or transforming water to make it available for a given use. The importance of cost considerations is illustrated by the case of water conveyance costs. Demand for transported water depends on the value of water for a specific purpose relative to conveyance costs. For example, water is transported up to 25 miles from outside the WRB (often aided by gravity) to serve urban users. In contrast, we estimate that a quarter mile of horizontal or uphill conveyance can be costly enough to make delivery of water for irrigation uneconomic on most agricultural lands in the Basin. Indeed, economic considerations explain why one-third of irrigable farmland (parcels with irrigation water rights) goes unirrigated each year. Irrigation involves costs and benefits, and, in some years, and on some lands, the costs outweigh the benefits.

Urban demand

Urban water use is projected to rise significantly by 2100, due primarily to population growth, but also to rising income. The growth in demand will be tempered to some degree by recent and near-term price increases related to cost recovery for infrastructure investments.

However, most urban water is used indoors and is nonconsumptive, that is, it is returned to the surface-water source from which it originated. Consumptive use (outdoor use that is not returned to streams via wastewater infrastructure) is projected to increase by only about 37,000 acre-feet per year by 2100. (Our model suggests that climate change will not lead to a significant direct increase in consumptive use.) The displacement of surface-irrigated farmland near expanding cities may offset as much as one-third of the increase in urban consumptive water use.

Moreover, a large fraction of urban water supplies in the WRB come from sources outside the Basin (primarily the Bull Run watershed on the slopes of Mt. Hood, but also reservoirs on the west side of the Coast Range). Thus, consumptive use from in-basin surface-water sources is projected to increase from 18,000 to 34,000 acre-feet per year, representing only about 7% of total urban water deliveries.

It is important to note that, compared to other uses, urban consumptive use of water from in-basin sources is small. Agricultural use (475,000 acre-feet) is 25 times greater, and regulatory minimum flows in the Willamette River (3.5 to 4 million acre-feet at Salem) are 200 times greater.

Irrigation

Water use for irrigated agriculture fluctuates from year to year, but has exhibited no significant upward trend in recent decades. The per-acre amount of water required for irrigation is expected to remain relatively stable. However, seasonal patterns of irrigation are likely to shift about two weeks earlier in response to earlier planting dates resulting from climate change.

The potential use of stored water to expand irrigation to farmlands that currently do not have irrigation water rights is limited by economic realities: conveyance costs are high relative to the economic gain from irrigating.

In-stream flows

The largest allocation of water in the Basin under human influence or control is the protection of in-stream flows. These flows serve multiple purposes, but are determined largely by habitat requirements of native fish. These minimum streamflows result from both federal BiOp requirements, which are tied to the Endangered Species Act, and state-mandated perennial minimum flows protected by in-stream water rights. Indeed, when completed, the implementation of currently unconverted in-stream water rights represents a significant increase in water allocation in the near term.

Reservoirs

The 13 federal storage reservoirs in the Basin produce enormous social value by reducing the risk of flood events. This benefit has been estimated at more than $1 billion per year. As urban areas expand, the value of potential damages during a flood will rise. Thus, the economic benefits of flood damage reduction will increase. To the extent that climate change leads to increases in high flow events, these benefits will become even more valuable.

Economics frames choices involving tradeoffs such as that between (a) the value of keeping reservoirs empty (for flood risk reduction) and (b) the value of filling reservoirs (for multiple summer uses). Since reservoirs cannot simultaneously be kept empty and full, the timing of refill during the spring is a critical economic decision. To maximize social benefits, the choice of reservoir fill level on any given date should balance the expected benefits of the competing uses (flood risk reduction versus storing water for summer uses).

Potential water scarcity

Based on results from the WW2100 model, the potential for increased water scarcity is likely to be location- and time-specific.

• Currently utilized municipal water rights may reach capacity in the Metro area (in 30 years) and in Salem (in 60 years). However, when the model accounts for currently underutilized water rights and those under development, urban water rights appear to be capable of meeting the overall growth in urban water demand.
• Our model shows a decrease in irrigation shutoffs of 10 to 30% under the reference scenario and the high climate change scenario. This result is due to earlier planting, which in turn leads to an earlier start, and completion, of irrigation. In the future, more farmers will have finished irrigating by the time the threat of a shutoff arises.
• The model results indicate increasing but modest shortfalls in reservoir fill levels during summer, resulting in some loss of recreational benefits, on the order of 5%.
• Implementation of all of the “unconverted” in-stream water rights intended to protect perennial flows would represent a significant increase in the amount of water allocated to environmental values. The effect, based on our model results, would be a small increase (5%) in the number of irrigation shutoffs. Overall, however, our results suggest that flow requirements to protect salmon and steelhead can be met, based on 10-year average flows. Exceptions are likely to occur in drought years.
• The effects of changes in forest wildfires and fire suppression policies could have a larger effect on water supply in the Valley than all of the changes in human water use combined. If forest cover is dramatically reduced, the resulting decrease in forest ET will increase streamflows and make more water available for human use, reducing the likelihood of scarcity.

The high spatial resolution of the WW2100 model makes it possible to identify specific land-use changes, such as those that displace irrigated agriculture, and to estimate the distances over which water conveyance is likely to be profitable for farmers. Models of this kind have important strengths but, like any model, also have limitations. Model projections should not be interpreted as precise quantitative predictions, but rather as indicators of the kinds of change that can be anticipated, based on what is known about the system being described.

References

Amos, A. 2013. Developing the law of the river: The integration of law and policy into hydrologic and socio-economic modeling efforts in the Willamette River Basin. University of Kansas Law Review 62:1091.

Bigelow, D. 2015. How Do Population Growth, Land-use Regulations, and Precipitation Patterns Affect Water Use? A Fine-scale Empirical Analysis of Landscape Change. Doctoral dissertation, Oregon State University.

Chang, H., S. Praskievicz, and H. Parandvash. 2014. Sensitivity of urban water consumption to weather and climate variability at multiple temporal scales: The case of Portland, Oregon. International Journal of Geospatial and Environmental Research 1(1).

Fisher, D. 1972. An Economic Analysis of Farms Producing Grass Seed in the Willamette Valley, with Special Attention to the Cultural Practice of Field Burning. Doctoral dissertation, Oregon State University.

Herrera, N., E. Burns, and T. Conlon. 2014. Simulation of Groundwater Flow and the Interaction of Groundwater and Surface Water in the Willamette Basin and Central Willamette Subbasin, Oregon. U.S. Geological Survey Scientific Investigations Report No. 2014-5136.

Jaeger, W., A. Plantinga, H. Chang, K. Dello, G. Grant, D. Hulse, et al. 2013. Toward a formal definition of water scarcity in natural–human systems. Water Resources Research 4506–4517.

Kalinin, A. 2013. Right as Rain? The Value of Water in Willamette Valley Agriculture. Master’s thesis, Oregon State University.

Lewis, E. and G. Peri. 2015. Immigration and the economy of cities and regions. In Handbook of Regional and Urban Economics, volume 5. G. Duranton, V. Henderson, and W. Strange (eds.). North-Holland, Amsterdam.

Moore, K. 2015. Optimizing Reservoir Operations to Adapt to 21st Century Expectations of Climate and Social Change in the Willamette River Basin, Oregon. Doctoral dissertation, Oregon State University. (accessed November 7, 2016).

National Marine Fisheries Service (NMFS). 2008. NMFS Willamette Project Biological Opinion. (accessed November 29, 2016).

Turner, D., D. Conklin, and J. Bolte. 2015. Projected climate change impacts on forest land cover and land use over the Willamette River Basin, Oregon, USA. Climatic Change 133:335–348.

U.S. Army Corps of Engineers (USACE). 2007. HEC-ResSim Reservoir System Simulation User’s Manual, version 3.0. USACE, Davis, CA. (accessed November 7, 2016).

U.S. Army Corps of Engineers (USACE). 2011. Willamette Valley: Operating the Reservoirs: Introducing the Rule Curve. USACE, Portland District.

U.S. Army Corps of Engineers (USACE). 2012. Willamette Basin Guide Standard Operating Procedures (SOP).

U.S. Army Corps of Engineers (USACE). 2015. A Primer on Flood Risk Management. (accessed August 5, 2016).

U.S. Army Corps of Engineers (USACE) and Oregon Water Resources Department (OWRD). 2011. Small-scale Water Supply Allocation Process: Willamette River Basin. USACE, Portland District.

U.S. Environmental Protection Agency (USEPA). 2009. 2006 Community Water System Survey. (accessed November 29, 2016).

U.S. Environmental Protection Agency (USEPA). 2013. Drinking Water Infrastructure Needs Survey and Assessment: Fifth Report to Congress.

U.S. Fish and Wildlife Service (USFWS). 2008. Final Biological Opinion on the Willamette River Basin Flood Control Project.

Appendix 1. Land use and land-use change models

The land-use change component of the WW2100 model simulates decisions to change land between agricultural, forest and urban uses as a function of economic returns to alternative uses. These economic returns are estimated on the basis of site characteristics, such as farm rents (annual net returns to farming), distance to cities and population and income of nearby cities. Returns to land and land-use transitions also are influenced by urban growth boundaries (UGBs). The land-use component includes a mechanism by which UGBs expand over time (described under “Zoning and UGB expansion”).

Changes between forest, agricultural and urban land uses comprise the focus of the WW2100 land-use change model, since these are the major uses, and sources of land-use change, that occur within the Willamette River Basin (WRB) study area (USGS, 2012). In addition, the property value data used to parameterize the economic returns to different land uses are not readily available for more disaggregated land-use categories (for example, specific crops or forest types). Differences within each of these broad land-use categories are controlled for by including separate sets of fine-scale land characteristics in each land value equation. Across the WW2100 study area, the initial land use on each Integrated Decision Unit (IDU) is determined by overlaying a set of spatial data layers, which include data from the National Land Cover Data (NLCD) data set and Cropland Data Layer (CDL).

The scope of the land-use transitions model is limited to privately owned lands, as the bulk of land-use change in our study area occurs as the result of decisions made by private actors. Wetland areas are not allowed to change uses, in accordance with local and federal planning rules and regulations that discourage wetland conversion. In addition, it is difficult to determine the market value of wetlands, since they do not easily fit into any one of the three categories we consider, and much of their value stems from recreation and nonmarket ecosystem service provision (for example, water filtration). Bidirectional transitions are allowed between privately owned forest and agricultural land anywhere in the “low-land” portion of the study area. Land development, on the other hand, or the conversion of forest or agricultural land to an urban use, is restricted to occur within UGBs, which, as mentioned above, are allowed to expand over the course of the model simulations. Development is also considered to be irreversible for the purposes of the WW2100 model (that is, transitions from urban to agriculture or forest are not allowed).

Economic returns to different land uses

Parcel-level data were collected from county assessors in four counties (Benton, Lane, Marion and Washington) in the WRB. Data included the real market value (RMV) of developed, agricultural and forest land; parcel size; and value of improvements on the property (for example, structures). These four sample counties were selected to represent the major urban areas within the WRB (Corvallis, Eugene-Springfield, Salem and Portland Metro), which have historically been the predominant drivers of land-use change in our study area, and to cover the geographic extent of the WRB.

The parcels included from each county were stratified according to three broad land use categories: agriculture, forest and residential development. County-specific random samples were then drawn from each category, with samples for Benton (430), Lane (1,134), Marion (1,576) and Washington (1,599) counties. The relative size of the county samples is proportional to the size of the urban areas they represent. Observations were obtained for the years 1973, 1980, 1986, 1992 and 2000, making it possible to use panel data methods to estimate a hedonic land value model for each of the three major land uses considered. The set of variables, definitions and data sources are described in Table A1.

The model estimation is summarized in Table A2 for developed, agricultural and forest land values. Developed land values were estimated using a Hausman-Taylor model, which allows for certain covariates in the model to be treated as endogenous and does not require the use of external instruments. The time-invariant effect of a parcel being located within a UGB was treated as endogenous. The area covered by a given municipality’s UGB is determined by city planners who, when setting the UGB, would likely target land parcels with characteristics that make them suitable for urban-oriented land uses. If unobserved factors influence both the decision to place a parcel within a UGB and the subsequent value of that land, the UGB effect will not be properly identified.

The following equation illustrates the general form of the hedonic models that were estimated:

ln(V_it^j) = β0_it^j + Σ_k=1^K_j β_kit^j X_kit^j + μ_iⁱ + ε_it^j

where the dependent variable, ln(V_it^j), denotes the natural log of the per-acre land value for parcel I in use j at time t. Additional terms denote the use-specific constant term (β0_it^j); use-specific covariates (X_kit^j); and their associated parameters (β_kit^j), where the number of covariates (K_j) varies by land use; a parcel-specific error component (μ_iⁱ ); and a standard idiosyncratic error term (ε_it^j).

In modeling the land values for forest and agricultural uses, a general method was used to control for unobserved parcel heterogeneity that could influence land values. Specifically, the correlated random effects (CRE) estimator was used. This approach entails including the parcel means of the time-varying explanatory factors in the model as additional regressors and provides a means to generate parameter estimates for time-invariant land characteristics, such as soil quality and topography, which are important in estimating the value of undeveloped land.

Initial estimates of agricultural and forest land values included several factors that convey a parcel’s potential for development (for example, population density, household income). For purposes of predicting land use transitions as a function of the relative economic returns to developed, agricultural and forest uses, the values of forest and agricultural lands should not be confounded by the capitalization of expected future development rents. To accomplish this, the effects of population density and household income were set to zero prior to using the estimated forest and agricultural land value estimates. The UGB distance variable was also set to zero for predictions in the agricultural land value equation for similar reasons.

Additional adjustments were made prior to using the hedonic model estimates for prediction in the WW2100 simulations. Each model in Table A2 contains an inverse Mill’s ratio variable (IMR) and, in the case of forest and agriculture equations, its mean. This variable is derived from a set of panel sample selection models, estimated with the approach given in Wooldridge (1996), which describe the attrition that is present in the land value panel. For the purposes of using the models for prediction in the WW2100 simulations, each of the use-specific IMR variables is set to one. Additionally, there is a large discrepancy between the developed land parcels used in the developed land hedonic equation and the size of the WW2100 simulation model IDUs. In order to assure that development decisions are made at the average lot size observed in the underlying hedonic model data, the average observed developed parcel size (0.39 acre) was used to compute developed land returns. Not doing so would substantially reduce the developed land values that drive land-use changes. Last, the estimated hedonic models also contain variables representing the value of improvements on each parcel. Since land improvements (for example, houses and barns) are not accounted for in the WW2100 simulations, the average improvement value for each land use is set to its average value observed in the underlying hedonic model input data.

The dependent variable in each land value equation is the logged per-acre value of land. To obtain the unlogged predictions, the exponent of each predicted value is computed and multiplied by a correction factor listed at the bottom of Table A2. For example, the developed land value prediction is given by 1.035 e^ln(V_idD). This adjustment is required to remove any bias stemming from the standard assumption in linear regression models that the disturbances are normally distributed, which does not necessarily apply in our case due to the log-transformation on the hedonic dependent variables.

The estimated hedonic model for each land use contains a set of county dummy variables to control for unobserved county-level heterogeneity. In using these models for prediction in the WW2100 simulation, the county dummy coefficients were applied to the four counties that comprised the samples used to estimate the hedonic models and to the other counties in the Basin, based on their proximity to the four original study counties. Predictions with the county dummy variables in non-sample counties were made as follows: the Washington County coefficient was applied to IDUs in Clackamas, Columbia, Multnomah and Yamhill counties; the Lane County coefficient was applied to IDUs in Douglas County; the Marion County coefficient was applied to IDUs in Linn and Polk counties.

Land-use change model

In the WW2100 model, each IDU is described by a set of attributes, including land cover. The initial land use on each IDU is determined by overlaying a set of spatial data layers, which include data from the U.S. Geological Survey National Land Cover Data, or NLCD, and the USDA Cropland Data Layer, or CDC. Starting from an initial land cover map for 2006, we model changes in land use as a function of the land values described above. This component of the model (as well as the components related to zoning and UGB expansions) is concerned only with privately owned land and considers land in only three uses: agriculture, forest and developed. Land in public uses or in uses other than agriculture, forest, or developed is assumed to remain in the same use.

With three uses, there are a total of nine possible transitions (counting transitions in which land does not change use; for example, forest-to-forest). This number is reduced to six because land that is in developed use can be assumed to remain in that use indefinitely. Thus, we are concerned with the six transitions shown in Table A3.

In particular, we model the probability that each transition takes place, where P_jk in the above matrix indicates the probability that land moves from use j to use k (for example, P_af is the probability that land moves from j = agriculture to k = forest). Because land must end up in one of the three categories, it follows that P_aa + P_af + P_ad = 1 and that P_ff + P_fa + P_fd = 1.

For those IDUs that begin 2006 in agriculture or forest and are categorized as privately owned, we define the six probabilities indicated above as logistic functions of economic returns to agriculture, forest and development. Formulas given below are used to compute transition probabilities for a five-year time period for each IDU. Calculation of these transition probabilities requires three steps. First, we use the equations described above, together with data on IDU characteristics, to calculate dev_val, ag_val and for_val for each IDU. Note that some of the IDU attributes are fixed over time (for example, slope), but others, such as population and income of the nearest city, change according to the procedure described under “Population and income projections.” Second, the computed economic returns (dev_val, ag_val and for_val) are scaled using formulas given in Table A4.

The scaling was necessary because the parameters in the transition probability formulas were obtained from another study (Lewis et al., 2011), which used different measures of economic returns from those used here. Lewis, et al. (2011) estimated a model of land-use transitions for western Oregon and Washington using land-use data from the National Resources Inventory and measures of economic returns from Lubowski, et al. (2006).

Thus, if an IDU begins in forest, the economic return to agriculture on that IDU is found by first computing ag_val, multiplying ag_val by 0.01, and then (if necessary) constraining the result to lie between 0 and 2,000. We refer to the scaled/constrained values of the economic returns as AR_a, Fr_a, and Dr_a when the starting use is agriculture, and ARf_, FR_f, and DR_f when the starting use is forest. Because DR_f = DR_a, we simply refer to this value as DR.

Once the values of AR_a, FR_a, and DR or AR_f, FR_f, and DR are determined for each IDU, the third step is to plug them into formulas that determine the five-year transition probabilities. As mentioned above, the parameters of these functions were obtained from Lewis, et al. (2011). Specifically, for IDUs starting in agriculture, the five-year transition probabilities are given by:

P_aa = e^0.00489AR_a / e^0.00489AR_a + e^{-7.799_0.00242FR_a} + e^{-4.788+0.00013DR}

P_af = e^{-7.799_0.00242FR_a} / e^0.00489AR_a + e^{-7.799_0.00242FR_a} + e^{-4.788+0.00013DR}

P_ad = 1 - P_aa = P_af

For parcels starting in forest, the five-year transition probabilities are given by:

P_ff = e^0.00489FR_f / e^0.00489FR_f + e^{-7.799_0.00242AR_f} + e^{-4.788+0.00013DR}

P_fa = e^{-7.799_0.00242AR_f} / e^0.00489FR_f + e^{-7.799_0.00242AR_f} + e^{-4.788+0.00013DR}

P_fd = 1 - P_ff = P_fa

With the WW2100 model, we wish to represent land-use change on an annual basis. To this end, the five-year probabilities are converted to equivalent annual probabilities using the following formula: if P_jk is the five-year transition probability, then the corresponding annual transition probability (AP_jk) is given by

AP_jk = 1 – (1 – P_jk)^0.2

When the model is run, the annual probabilities are used for a five-year period and then updated.

Given sets of values of AP_jk for each IDU, the final step is to determine whether or not land-use changes occur on an IDU. For this, we will use a random number generator. Suppose that a given IDU is currently in agriculture and has an 80% probability of remaining in agriculture (that is, AP_aa = 0.80), a 10% probability of switching to forest (AP_af = 0.10), and a 10% probability of switching to developed use (AP_ad = 0.10). We draw a random variable r from a uniform distribution defined on the unit interval. The IDU remains in agriculture if 0.8 > r ≥ 0, changes to forest if 0.9 > r ≥ 0.8, and changes to developed use if 1.0 ≥ r ≥ 0.9. This procedure is repeated for each IDU using a newly drawn random variable. As the end of this process, a new land-use map is produced.

Zoning and UGB expansion

To account for zoning rules under Oregon’s land-use planning system, we treat land inside and outside of UGBs differently. Land outside of UGBs can move between undeveloped uses (that is, agriculture-to-forest and forest-to-agriculture transitions are allowed), but transitions to developed use are not allowed. To account for this restriction on development, the transition probabilities need to be adjusted. For IDUs outside of UGBs, the probability associated with the transition to developed use is added to the probability associated with the land remaining in the same use. Thus, if the initial use is agriculture, P_aa(new) = P_aa(old) + Paa. If the initial use is forest, P_ff(new) = P_ff (old) + P_fd. These restrictions on development are also applied to areas zoned as rural residential (which are outside of UGBs), since only a small portion of the lot is allowed to be developed.

For IDUs inside of UGBs, all of the transitions are allowed, except for transitions out of developed land. Given the irreversibility of development, it follows that, over time, the share of developed land within each UGB will increase. To mimic the land-use planning process, we allow for UGBs to expand once the developed share becomes sufficiently large. The developed share is defined as the ratio of the area of private land within a UGB that is developed to the area of private land within a UGB that is developable. Developable land includes all private land that is in developed, agriculture, other vegetation and forest categories. It excludes land in the barren, wetlands and water/snow/ice categories. As long as the developed percentage of the UGB is below a specified threshold (in the reference scenario, the threshold is 80%), the existing UGB remains as it is. (In the reference scenario, the developed threshold for Eugene-Springfield is 70%. This change was made in recognition of the ongoing UGB expansion process during our model’s development in 2010–2015.)

However, once the threshold is exceeded, a UGB expansion is triggered. We assume that the need for a UGB expansion is evaluated every five years, coinciding with the updating of the five-year transition probabilities.

A UGB expansion involves adding new IDUs to the existing UGB area until the specified threshold is no longer exceeded. Which IDUs to add is determined by the following criteria:

1. Select only IDUs that are adjacent to the existing (or expanded) UGB area.
2. Select only IDUs that are privately owned and developable (that is, in developed, agriculture, other vegetation, or forest categories).
3. Do not select IDUs that are already inside another UGB.
4. Select IDUs in order of least equally weighted distance from: (a) the centroid of the IDU to the center of the UGB area, and (b) the centroid of the IDU to the nearest major road.
5. Do not select IDUs that are zoned for exclusive farm use (EFU) or forest conservation (FC), unless there are no other IDUs that satisfy criteria 1–3 and the developed percentage is still higher than the prescribed threshold. Once the non-EFU and non-FC IDUs that satisfy criteria 1–3 are exhausted, continue selecting IDUs that are zoned as EFU or FC using criteria 4 until the developed land percentage is again below the prescribed threshold.

These criteria do not apply to expansions in the Portland Metro UGB. In this case, the regional planning authority (Metro) has designated areas called urban reserves that indicate where future expansions will take place until 2060. For expansions in the Portland Metro UGB prior to the year 2060, the criteria are as follows:

1. Select only IDUs that are adjacent to the existing (or expanded) UGB area.
2. Select only areas within designated urban reserves.
3. Select IDUs in order of least equally weighted distance from: (a) the centroid of the IDU to the center of the UGB area, and (b) the centroid of the IDU to the nearest major road.

After 2060, the Portland Metro UGB is treated like all other UGBs in the WW2100 study area, and expansions take place in accordance with criteria 1–5 listed above.

Population and income projections

The developed land value equation includes variables for the population density and household income of the nearest city. As the WW2100 model progresses through time, we allow population and income to increase. Future population and income are assumed to be exogenous and, as such, are external drivers in the same way as climate. County-level population projections to 2050 and real (2005 dollars) mean household total personal income projections to 2040 are taken from the Oregon Office of Economic Analysis and Woods and Poole, respectively. Linear extrapolation is used to obtain projections to 2100. Actual, forecasted and extrapolated population and income by county for the period 1970–2100 are reported in Tables A5 and A6, respectively. For the 10 WRB counties, population increases at an average annual rate of 2.1% over the 1970–2100 period. The rate of increase is 1.2% over the period 2010–2050 and 0.8% over the period 2050–2100. For the 10 counties, mean house-hold total personal income increased at an annual rate of 1.9% over the period 1970–2010 and is projected to increase at a rate of 1.7% over the period 2010–2040 and 1.4% over the period 2040–2100.

As the WW2100 model runs, population is allocated on a five-year basis to UGBs and areas zoned for rural residential (RR) use. No population is allocated to areas outside UGBs and RR areas. For a given increase in a county’s population, we use the following procedure to determine the allocation of population to the UGB and RR areas within the county. We begin with the Census block data for 2010 from the U.S. Bureau of the Census. These data provide a population count for each block, allowing us to mine the initial spatial distribution of population within each county. The block-level estimates are aggregated into a single population count for each UGB and RR area within each county. (Note that a UGB may span several counties. For example, the Salem-Keizer UGB includes land in both Marion and Polk counties. In such cases, we treat a county’s portion of a UGB as a separate area for the purpose of allocating population.)

Weights are then determined for each UGB and RR area within a county according to their respective share of the 2010 population, disregarding the residual population outside of UGB and RR areas. As the model runs, the increase in a county’s population is allocated to the UGB and RR areas according to these weights.

In allocating population to RR areas, we impose a maximum density of one household for every 2 acres of land to remain consistent with the existing rules governing rural land development. When an RR area as a whole meets this prescribed density threshold, it is shut off from future population growth and its population weight is reallocated proportionally to UGBs and other RR areas that are still eligible to receive population. We allow household size to vary through time using county-level forecasts in the Woods and Poole data for 2010–2040 and linear extrapolation to 2100.

Appendix 2. Urban water use models

The urban water demand component developed for WW2100 consists of models of residential and nonresidential urban water demand for the Portland metropolitan area, Salem-Keizer, Corvallis and Eugene-Springfield, as well as a separate model for smaller urban areas. These models were developed to predict water consumption for urban areas in the WRB. The variables included in these models were selected on the basis of a review of the economics literature on urban water demand. An additional consideration was the need to use variables that could be forecasted over the entire study period, either as exogenous drivers (such as income and population growth) or as variables generated by other models within the Envision framework (such as population density).

Data and methods

The economics literature that focuses on estimation of water demand suggests that a water demand function must include marginal price of water, pricing structure and income and must control for seasons and weather. Given the specific forecasting needs of the urban water component for WW2100, we also included population and population density in the model.

Despite anecdotal evidence of the effect of non-price water demand management policies (for example, installing low-flow appliances), there is very little empirical evidence in the economics literature of the impact of such measures on water demand. A literature review yielded only one peer-reviewed paper that systematically measured the effect of non-price management policies on water demand; this paper was based on data from 1989–1996. Furthermore, the water demand literature emphasizes the fact that, while the demand reductions attainable from non-price management options are clear from a technical perspective, once consumer behavior is taken into account the net effects are not well understood. For example, there is anecdotal evidence of “double-flushing” low-flow toilets, of consumers altering low-flow devices to function like traditional fixtures (see Timmins, 2003, and references cited therein), and of households increasing the frequency of clothes washing when using more efficient front-loading clothes washers (a “rebound effect” from increased efficiency).

The magnitude of these effects, and hence the net impact of conservation policies to promote adoption of more efficient appliances, has not been sufficiently studied in the economics literature and hence is not as well understood and quantified as the effect of price and income. Because of this lack of reliably measured impacts, and given the difficulty of projecting conservation attitudes or development of new technologies over the study period (up to the year 2100), we opted to omit the non-price management effect from the water demand model. To the extent that this component would be expected to have a negative impact on demand (assuming rebound effects are smaller than direct impacts), we are being conservative by slightly over-predicting residential demand for water.

The model of residential demand therefore predicts total daily water used by residential customers in hundreds of cubic feet as a function of price (dollars per hundred cubic feet), pricing structure (increasing block rate (IBR) or flat rate), city population, median house-hold income and population density (persons per square mile). An increasing block rate pricing structure charges different prices for incremental volumes consumed each month. One price is set for a given base volume, and a higher rate applies for quantities above that volume. In some cases, another threshold triggers a third “block rate.”

The model of nonresidential water demand predicts total daily water used by nonresidential customers (in hundred cubic feet) as a function of price, total city industrial (manufacturing) income, total city commercial income and population.

The values of the response parameters for these variables were obtained from economics literature cited in Appendix References:

• Price: Long-term price elasticity of demand = - 0.6
• Income: Income elasticity = 0.13 for flat rate pricing, 0.18 for increasing block rate pricing
• Population: 1.0
• Population density: -0.048
• Industrial (manufacturing) income: 0.11
• Commercial income: 0.04

We collected the most current information available (at the time the project was started) on each of these variables for Portland, Salem-Keizer, Corvallis, Eugene and Springfield. Information on water rates, price structure and water use was obtained from Water Management and Conservation Plans for Portland (2010), Salem-Keizer (2009), Corvallis (2005) and Eugene (2012), and through personal communication for Springfield (2012). Information on personal, manufacturing and commercial income comes from the Bureau of Economic Analysis. Population and population density information is from the U.S. Census.

We used the coefficients specified above and the averages of water quantity, price, income, population and density for the five cities to calibrate a log-linear model and calculate the intercept term corresponding to the baseline averages. Because the intercept varies with the pricing structure, we calculated two intercepts, using the corresponding averages for cities with flat rate and IBR. Then we calculated the necessary coefficient for the IBR indicator variable to give the correct intercept when IBR is set to 0 or 1. Finally, the demand models are adjusted to reflect seasonal variations in demand. Thus, note that the model is not intended to predict water use for any particular city in the WRB. Rather, its purpose is to generate predictions for the Basin’s urban areas in general on the basis of a limited number of variables that are forecasted to the year 2100 as part of the WW2100 project.

The specific models for the four main cities (Portland Metro, Salem-Keizer, Corvallis, Eugene-Springfield) are as follows:

We adjust residential demand for seasonality by decomposing daily water use into outdoor and indoor use components, based on 24 years of daily data from the Portland Water Bureau. Total predicted yearly water demand is divided by 365 to obtain daily use, and then multiplied by indoor and outdoor fractions to reflect seasonality.

Baseline prices for the four main cities are as follows:

Residential:

• Portland (all of metro area): $2.44 per hundred cubic feet
• Salem-Keizer: $2.04 per hundred cubic feet
• Corvallis: $1.93 per hundred cubic feet
• Eugene-Springfield: $2.00 per hundred cubic feet

Nonresidential:

• Portland (all of metro area): $2.44 per hundred cubic feet
• Salem-Keizer: $1.50 per hundred cubic feet
• Corvallis: $2.11 per hundred cubic feet
• Eugene-Springfield: $2.00 per hundred cubic feet

IBR = 1 for Corvallis, Eugene-Springfield; IBR = 0 for Portland, Salem. Use IBR = 0 for other urban areas.

Baseline (initial) manufacturing income ($1,000s):

• Portland: 9,851,720
• Salem-Keizer: 651,857
• Corvallis: 461,476
• Eugene-Springfield: 723,165

Baseline (initial) commercial income ($1,000s):

• Portland: 58,292,148
• Salem-Keizer: 7,350,692
• Corvallis: 1,598,343
• Eugene-Springfield: 6,565,399

Price adjustments

Urban water utilities set prices to achieve multiple goals. These include generating revenues so that the water utility can cover its costs. Providing a sufficient and stable source of revenues is important, but rate structures should not be overly complex. Water providers want to achieve a fair allocation of costs, and they may have a goal of allocating costs among different types of uses and users. They wish to fully allocate private and social costs, and at the same time provide incentives for conservation to customers, and do so in a way that is efficient and transparent to customers. Because urban water delivery systems are highly capital intensive, with a need to make large capital investments for infrastructure building, maintenance, or replacement on an intermittent basis (sometimes decades apart), in a typical year these capital costs are not in ratepayers’ minds, so it is frequently difficult to set prices in such a way that long-term average or marginal costs are covered, as opposed to short-run (current-year) average costs. This situation is endemic to urban water supply systems and is well established and widely recognized in economics research.

As a result of these characteristics of urban water systems, water prices tend to be somewhat lower than long-run average or marginal costs, leading to financial deficits and delays in infrastructure investments, repairs and replacements. It is common that when the severity of financial needs leads water providers to raise prices, the result may be a worsening financial situation, as price increases discourage consumption and thus reduce the anticipated increase in revenues.

These phenomena are well documented in historical data, surveys and engineering analyses. U.S. Environmental Protection Agency (EPA) survey data, for example, indicate that average water prices are frequently more than 20% below long-run aver-age cost (EPA, 2009). In a separate survey, the EPA has documented the resulting backlog of infrastructure needs for drinking water systems nationwide (EPA, 2013). The nationwide total “20-year need” reported to EPA in 2011 was $376 billion, which amounts to more than $1,200 per capita.

The gap between average price and average cost has fluctuated over time and across cities and states for a variety of reasons. Nationwide, we have observed rising (inflation-adjusted) urban water prices since about the mid-1980s. Prior to that time, there was an extended period of declining urban water prices. The backlog of infrastructure needs, as estimated from EPA surveys, has also fluctuated, but has been rising nationwide, from $843 per person in 1995 to $1,205 per person in 2011. The backlog of infrastructure needs in Oregon has generally been higher than the national average, rising from $1,108 per person in 1995 to $1,442 per person in 2011. The exception was in 2007, when Oregon’s per-capita need was only $845, compared to the national average of $1,220. The reduced backlog of infrastructure needs in Oregon between 2003 and 2007 came following a 53% increase in average prices in Portland between 1999 and 2007, which likely contributed to financing significant infrastructure improvements.

For our modeling purposes, we want to project future price trends that recognize the relatively large backlog of infrastructure needs in Oregon in the most recent EPA survey, as compared to national levels and historic levels in Oregon. Since most urban populations in Oregon live in the WRB, we use Oregon-wide data as a reasonable indicator of the situation in our study area. In order to finance infrastructure needs over the next 20 years, additional increases in water prices will be needed. To reduce the level of these system needs over the next 20 years from the 2011 Oregon estimate ($1,442 per person) to the national average observed since 1995 ($1,050 per person) would require additional revenues of $40 per person per year — more than a 25% increase in average per-capita water payments. To recognize the unusually high level of infrastructure needs currently estimated for Oregon, our model includes an increase in average water prices of 1.5% per year for the first 15 years of the modeled scenario, resulting in a cumulative rise in price of 25%. After that point, prices are held constant in real terms for a given population served.

Further changes in price beyond 2025 occur as a function of average current cost, which is a function of population served. These relationships were estimated based on national EPA data. The average cost is estimated as

AC_t = 0.748•exp{9.93 – 0.355•ln(Pop(t))+ 0.030•(ln (Pop(t)))2 – 0.001•(ln Pop(t))3}

We assume that prices will increase at the same rate as average cost:

P_t-P_t=0/P_t=0 = AC_t-AC_t=0/AC_t=0, which implies that

Pt = AC_t-AC_t=0 • P_t=0

For other cities, demand is also a function of population, income and price (dollars per hundred cubic feet). Other cities do not have separate nonresidential demand functions, nor do they use IBR pricing.

Prices are based on the water delivery average cost. Demand is based on assuming price equals AC from the relationship above, so that

Ln Q_t = -2.16432007 – (0.6•ln(p_t)) + ln(Pop_t) + (0.13•ln(I_t)) – (0.048•ln(D_t))

and pt = i. Demand is adjusted for seasonality as described above.

Rural residential demand is

ln(Q_RR) = -3.55 – (0.6•ln(PC)) + ln(Pop) + (0.13•ln(I)) – (0.048•ln(D))

where PC = “price” for water (dollars per hundred cubic feet) (cost of pumping, $0.30/ccf)

Pop = the population of the rural residential IDU
I = income per household ($/household), average for relevant county
D = population density (people per square mile) (assumed to be 768 per square mile in rural areas, or 2 acres per household)
Demand (Q_RR) is in ccf/day

The model that determines price and demand also tracks the evolution of estimated long-run average cost (LRAC) (dollars per hundred cubic feet) as a function of changing population (but holding other variables constant). Long-run average cost, including capital costs, is estimated with national EPA data as follows:

LRAC = (0.748/1,000)•exp{13.39 – 1.246•(ln Pop) + 0.117•(ln Pop)² – 0.004•(ln Pop)³}

Validation

Table A7 shows the predicted residential and total per-capita water consumption calculated from output generated by the WW2100 model (April, 2015) for the first forecast year (2010) for the four major urban areas. For reference, Table A7 also shows values reported by the corresponding city utilities.

Table A7 suggests that the model predicts per-capita water use for 2011 reasonably well, both on average and for individual urban areas. The model estimate for per-capita consumption for 2011 is somewhat lower than observed levels. This reflects in part the rising prices and declining trend in per-capita consumption observed during this period.

As an additional validation exercise, we used the urban water demand model and data on water prices and water consumption for Portland from 1995 to 2011 (from Portland Water Bureau) to generate a “precast” of water consumption in Portland during that period. The objective is to compare water consumption predicted by the model with observed data. As shown in Figure A1, prices increased (in real terms) during this period. The average yearly increase was 5.3%.

Table A8 and Figure A2 show the water use predicted for Portland by the model, along with observed consumption amounts.

Per-capita water use decreased by 39.83 gallons per day (26%) between 1994 and 2012. For that period, our model predicts a total decrease of 39.84 gallons per day (31%).

Hence, this “precasting” exercise suggests the urban water model is capable of replicating the trend in per-capita water consumption in response to changes in water price fairly well.

Note that the model is not intended to predict water consumption in Portland specifically, as it is calibrated using data for the five major urban areas in the Basin.

Appendix 3. Agricultural crop choice and irrigation

In each year that an IDU is assigned to agricultural land use, farmer decisions are modeled to simulate crop choices and irrigation decisions. Irrigation is possible only on IDUs with existing irrigation water rights. These initial decisions are then followed by daily decisions related to planting and harvesting, and (possibly) applying irrigation water. The availability of irrigation water is also subject to regulatory shutoffs in accordance with the prior appropriations seniority system under state law (described under Appendix 6). These combinations of decisions, choices, actions and responses to exogenous factors produce a unique pattern of crop water use, irrigation diversions, soil moisture and ground-water contributions and net farm income (annual land rent) at the IDU level. To the extent that irrigation water is shut off by regulators, short-run annual land rent is reduced, as is long-run (expected) land rent.

Crop water use in agriculture is estimated on a daily basis reflecting the crop or vegetative cover of each IDU and ET as a function of meteorological factors, crop type and growth stage. The model also estimates crop planting and harvest dates.

Crop choice

The crop choice model estimates the probability of growing each of seven crop types or groups for the modeled year. The empirical model is estimated at the parcel level based on observed cropping patterns in recent years. The model estimates the crop observed as a function of IDU characteristics, including soil quality (land capability class), elevation and the presence of an irrigation water right, as well as variable attributes, including crop prices and expected water availability (for those IDUs with irrigation water rights). Given the estimated probabilities for each IDU, the simulation models determine the crop for each IDU in each year with a random draw reflecting these estimated probabilities. No evidence of crop choices being correlated across years (that is, a crop rotation schedule) was found in the data or in interviews with farmers or agricultural Extension personnel.

The crop choice model is estimated as a hedonic relationship based on parcel-level GIS data for 100,555 parcels over a six-year period. Crops were identified using USDA CropScape agricultural land cover data. Parcel data included land characteristics (land capability class (LCC), elevation, slope, field size and water rights), climatic characteristics (average precipitation and minimum growing season temperature) and crop prices.

The model was estimated using ordinary least-squares as follows:

P_j = β₀ + Σ_i=1¹⁴β_i

for crops j = 1 to 8, where 8 is “other crops,” where P_j is the probability of planting crop j in a given year, and independent variables 1–14 are as follows:

• LCC₁ to LCC₇ = the dominant soil type (land capability class) in the IDU
• EL = elevation (in meters, demeaned, where E̅L̅ = 97.2)
• SL = slope (percent, demeaned, where S̅L̅ = 3.704)
• PR = precipitation, April–October (inches, demeaned, where P̅R̅ = 13.63)
• MT = minimum temperature, April–October (degrees C, demeaned, where M̅T̅ = 8.556)
• PG = price of grass seed (average over period = $64 per ton)
• PW = price of wheat (average over period = $5 per bushel)
• IR = existence of irrigation water right (1 if a water right exists; otherwise 0)

The estimated coefficients are indicated in Table A9.

The resulting modeled values are interpreted as the probabilities for each crop to be grown. The model is implemented as a “random draw” to deter-mine which crop is grown each year at a given IDU location. The effect of the existence of an irrigation water right on crop choice will be moderated to the extent that a water right shutoff is anticipated in a given year. To represent this circumstance, the implemented crop choice model included an additional term, IR•SE, the interaction of IR as defined above with SE (Expected Snow), where (a) SE = 1 if the model estimate of April 1 snowpack is greater than or equal to the average April 1 snowpack in the previous 10 years, or (b) SE = (1 – SH) if the April 1 snowpack measure is less than the average snowpack in the previous 10 years, and where SH is the frequency in the previous 10 years that the water right experienced a regulatory shutoff (0 < SH < 1, and its initial value is 0 for all IDUs). Snowpack is estimated in the model on a daily basis at the IDU level, based on the interactions of precipitation, temperature and other factors. Data on snowpack for calibration purposes come from the Natural Resources Conservation Service’s data collection service, SNOWTEL.

For perennial crops (orchards, vineyards and tree crops), a fixed set of IDUs is permanently assigned. These areas represent a relatively small proportion of farmland in the Basin. These crops are relatively stable in area and location, and changes would be difficult to predict.

Irrigation decisions

Farmers in the WRB typically own and cultivate multiple fields and each year make decisions about which crops to grow on each field, whether to irrigate those fields with irrigation water rights, what equipment to use on which fields, etc. We model these decisions at the parcel (or IDU) level as representative of farm-level decisions. A survey was conducted in the fall of 2012 to collect data from farmers about their irrigation and crop choice decisions for a sample of parcels and over the previous six years for those individual fields. The survey had a unique design in that farmers were asked to identify their irrigated fields on a map, so that their responses could be matched to land quality and climate data. Integration of survey responses with spatial data allowed for development of an irrigation decision model that could explain why a large percentage of existing water rights are not used in a given year.

A sample of 530 farmers was surveyed, and data were collected on up to three fields for each farmer over a six-year period. A discrete choice irrigation decision model was estimated based on these data. The logit model estimated takes a form such that the probability P_IRR of irrigating a given field is estimated. Using a panel of 4,409 observations, the following relationship was estimated:

ln(P_1RR/1-P_1RR) = β₀ + Σ_I=1¹⁷ β_ix_i

where the variables and their estimated coefficients are as indicated in Table A10.

The spatial data sources include PRISM data for monthly precipitation from April to August (2007–2012), as well as for 30-year monthly averages (1980–2010). The soils gridded data (10 meters) and elevation data came from USDA GeoSpatial Gateway. Groundwater data was provided by Roy Haggerty, Oregon State University. Water rights were from the Oregon Water Resources Department (OWRD). Distances to major cities and to the Willamette River were computed using Oregon GIS clearinghouse data on streams and city boundaries. Soil water-holding capacity comes from the Arc GIS SSURGO layer.

Irrigation efficiency is assumed to be 0.8 for all irrigation in the Basin (meaning that 1.25 acre-feet must be applied to make 1 acre-foot available to plants). This is a midrange assumption for sprinkler irrigation, which is by far the dominant technique used in the Willamette Valley. Efficiency losses are assumed to include conveyance losses.

Crop choice and irrigation decisions are inter-dependent. Farmers may make crop choices and irrigation decisions simultaneously or, if sequentially, in either order. In the WRB, most major crops are sometimes grown nonirrigated and sometimes irrigated. The correlations in these decisions are reflected in the crop cover data, irrigation survey data and crop choice probabilities. In addition, farmland rent reflects profits from farming and irrigation decisions. In general, farmers will not plant or irrigate land that is expected to generate no profit or rent. Reflecting this, the following rules were introduced in the simulations:

• If farmland rent < $1, crop choice is fallow and probability of irrigating = 0.
• If the irrigation decision is “yes,” the probability of fallow is zero.
• If the irrigation decision is “yes,” the probability of wheat is zero.
• If the irrigation decision is “no,” the probability of corn is zero.

As a result of these adjustments for fallow, wheat and corn, the probability of “other crops” (j = 8) is adjusted for consistency so that

P₈ = 1 – Σ_j⁷=₁P_j

Irrigation can increase profits through higher yields and by expanding the range of crops that can be grown. However, it is also costly in time, energy costs and capital costs. Since farmers are heterogeneous in their production skills and in the attributes of their land, the additional benefits will justify irrigation for some farmers but not for others, and in some years but not in others.

Farmland rent

The economic rent or annual profit from farming a given piece of land can play an important role in farm decisions to plant a crop, irrigate, or transition out of farming. Our estimation of farmland rent takes a Ricardian approach that is common in models of the economic returns to agriculture. Land value is assumed to equal the net present value of future rents from putting the land to its highest value use; as a result, we expect to see variation in land values and annual rents due to characteristics of the land that would influence agricultural productivity, such as soil quality and precipitation or irrigation water rights. Similar to the hedonic model of crop choice, here we decompose the farmland rent associated with factors affecting agricultural productivity.

The source data on farmland values and rents originate from data collected by county assessors, a process required in Oregon to monitor levels and trends in both real market land values and assessed values. Drawing on land sales, land rentals, surveys and expert analysis, county assessors produce estimates of average farmland rents within a county by soil type (LCC) and zone for parcels with and without irrigation water rights (Table A11). In the absence of an adequate sample of data on individual farmland sales, these semiaggregated data reflected sufficient variation across zones and soils that hedonic analysis could be used to infer the contribution of other covariates such as elevation, precipitation, growing season minimum temperature, etc. The first step was to assign a rent to each parcel across zones, soils and water right assignments, according to the county-estimated real market value for those locations and characteristics. These became the dependent variables for a hedonic model estimation that included two variables that determined the assigned rent level (LCC and existence of irrigation water rights), as well as other characteristics of the parcels (elevation, temperature, precipitation, etc.). By regressing these variables on the farmland rent estimate, we are able to recover, for example, the marginal value of higher summer temperature or lower elevation, independent of soil class.

To compile a data set of agricultural tax lots spanning the extent of the Willamette Valley, cadastral and zoning tax lot data were collected from all counties, identifying the tax lots zoned for agricultural use. Small tax lots of fewer than 10 acres were excluded. The hedonic estimation includes the following variables: soil classes LCC 1–4, 6 and 7 independently and, in the case of LCC 1–4, interacted with a dummy variable for existence of an irrigation water right (IRR); IDU elevation; historical average growing season precipitation (demeaned using the basinwide mean); and historical average growing season minimum temperature (demeaned using the basinwide mean). (There were insufficient observations of LCC5 land in the data to estimate a coefficient. As a result, we are assuming LCC5 lands are treated as if they have the same value as LCC6 lands.)

In addition, parcel size in acres was interacted with each of the variables elevation, precipitation and temperature. For the IDU values of farm-land rent, mean parcel size was assumed.

Using these data, the rent for each parcel was estimated with the following form:

R=β₀+Σ₁¹⁶ β_jX_j+ε

where R is the parcel rent (per acre per year), β₀ is the intercept, X_j represents the variables in Table A12, and β_j is the coefficient on X_j.

Validation: Crops, irrigation, land use and farmland values

The combination of model results for crop choice, irrigation decisions and farm rents are compared here with data from the USDA Census of Agriculture, 2012. The main crop categories among harvested cropland acreage in the USDA data are grass seed (53%), hay (15%), orchards (5.6%), vegetables (included in Table A13 as “other”) (5.4%), field crops (4.5%) and nursery crops (4%), also included as “other.” This pattern has been relatively stable for many years for the major crops; grass seed has been the dominant crop by acreage for more than 50 years.

Harvested cropland has averaged about 900,000 acres over the past decade, declining from a high of about 1 million acres in the 1980s. About 267,000 of these acres, or about 30%, are irrigated in any given year (USDA Census of Agriculture data, 1997–2012). Of these, just over half (about 140,000 acres, or 52%) are irrigated with surface-water rights. The break-down between surface and groundwater irrigation is based on our own farm survey (discussed under “Irrigation decisions”).

The acreage of irrigated farmland in the Basin has remained relatively stable since the mid-1990s. Prior to that time, irrigated acres rose gradually, reflecting the acquisition of new irrigation water rights. Since the available live flow water rights were fully allocated as of the 1990s, no new irrigation water rights have been approved by the OWRD.

Data on existing irrigation water rights indicate about 462,000 acres with irrigation water rights from surface, groundwater, or stored water (federally contracted) sources. Given the annual irrigated acreage of 267,000 acres, this indicates a rate of utilization of irrigation water rights of 55 to 60%. The estimated frequency of utilization of irrigation water rights from our farm survey (and the basis for the modeled acreages irrigated) was 62%. An exact count of irrigation water rights is complicated by several factors: (a) Water rights can sometimes have multiple uses (irrigation, domestic and livestock); (b) A small fraction of irrigation water rights in the WRB are “supplemental water rights” rather than primary, and thus cannot be used unless the primary right is exhausted or unavailable (supplemental water rights were not included in this model); (c) The Oregon water rights database contains measurement errors, GIS errors and other imperfections; and (d) Some water rights may have been abandoned or are not known to the landowner.

These values are similar to data reported by the USDA Land Values 2012 Summary. That report estimates the average farm-land value to be $2,290 per acre in Oregon overall; average values for irrigated lands are $3,650 per acre, for nonirrigated lands $1,800 per acre, and for pasture lands $670 per acre. For irrigated lands, the USDA Oregon average is higher than for all but the best soil class in the WRB. Indeed, comparisons with data from other parts of Oregon where irrigation is prevalent suggest that the value of irrigable land in the WRB is somewhat lower than that of irrigated lands in the Deschutes, Klamath, or other basins in Oregon. In general, the estimated farmland values are very close to the averages reported by the USDA. Looking at farmland prices in years after 2010 (for example, 2016), modeled prices would be expected to diverge from those observed. The reason is that the model assumes prices to be inflation-adjusted so that a stable real (inflation-adjusted) price is constant in the model. Inflation has been about 3% per year; farmland prices in the WRB have risen faster than that during the 2010–2015 period.

The implicit farmland prices in our model, if adjusted to reflect inflation from 2010 to 2016, would range from $2,500 to $4,000 for irrigated land, and from $600 to $3,000 for nonirrigated land. Indeed, farmland sales reported as of June 2016 in several WRB counties average $3,100 per acre, with a high of $4,800 per acre. Lands with buildings and infrastructure, for example to support nursery production, have significantly higher market values due to these capital improvements. Sales prices will be higher for lands where there is an expectation that land development will be possible in the near future, for example where changes in land-use regulations, UGBs or both are anticipated.

Appendix 4. Conveyance cost estimation

Agricultural lands in the Willamette Basin that currently do not have irrigation water rights may benefit from opportunities to acquire new water rights under federal contracts for stored water at one of the U.S. Army Corps of Engineers reservoirs. The profitability of a new contract for stored water will depend on a comparison of the irrigation benefits (higher yields and a wider range of crop choices) and the additional costs (capital investments in infrastructure, labor and energy costs). For farmlands with existing irrigation water rights, these costs and benefits are already incorporated into the WW2100 estimates of farmland rent (annual profits) by soil class.

For new contract water rights, we would expect the irrigation premium to be the same as for existing irrigation water rights if the costs of irrigating are similar to the average costs for existing surface and groundwater rights. In the case of new water rights from stored water, we expect the costs to be somewhat higher due to (a) the fee paid to the U.S. Bureau of Reclamation (USBR) for the water contract, (b) the extra cost for mainline conveyance to bring the water from a below-reservoir tributary to the field and (c) the extra lift required. Whether or not a new irrigation water right will be attractive to a farmer will depend on whether farming is more profitable with irrigation than without.

The extra costs are estimated here. There are two components to these costs: (a) the infrastructure and installation capital costs, and (b) the additional energy costs. For both of these costs, we estimate average values for a representative irrigation operation of 120 acres (which could involve combining multiple fields via cooperation or land sales).

Capital costs for mainline conveyance infrastructure

Capital costs for mainline conveyance systems include the pipe and below-ground installation costs. The size of the pipe is tied to the flow rate needed to serve the irrigable area; larger pipes are more costly, but increase capacity and reduce friction losses. For a range of pipe diameters, the costs, capacity, friction losses and irrigable areas are shown in Table A15. Our analysis is based on use of an 8-inch pipe to serve a 120-acre farm. The annualized capital costs are estimated to be $0.59 per year per 100 feet of mainline (assuming 120 acres, or an average rate of 6.5 gallons per minute). For a mainline of 1,000 feet, this would mean an added cost of $5.90 per acre per year. In addition, the charge from the USBR averages about $9 per acre.

Variable cost: Per-acre energy costs

The additional energy costs for irrigation from stored water contracts will reflect the additional distance and lift along the mainline to bring water from the point of diversion to the irrigated field. The added cost is estimated as a function of the mainline length and the lift.

The energy cost, c, can be expressed as

c = p•E (1)

where p is the price of electricity ($ per kilowatt-hour), and E is the energy consumed in kilowatt-hours. We have

E = t•(output power) (2)

where t is time (hours), and output power is the power per unit time. The rate of energy use is

output power = q•TDH/3.960 (3)

where q is the pumping rate per hour, and TDH is the total dynamic head, or the sum of the lift, head losses, friction losses and the pressure at the pump (in pounds per square inch multiplied by 2.306 to get horsepower).

To convert output power from horsepower to kilowatts, we multiply by 0.746. To adjust for the overall pumping plant efficiency, E_plant, the expression is divided by this value, which can range from 0.7 to 0.8. A midrange value for pumping plant efficiency is 0.75, meaning that input power must be one-third higher than output power.

The hours of pumping, t, necessary to apply the required irrigation water (acre-inches), d, is computed as

t = (d•27,154)/(q•60) (4)

Combining the relationships in (2), (3) and (4) above and simplifying (for example, 27,154/(60•3,960) = 0.114285), we can write this as

c=p[d•TDH•0.114285•0.746/E_plant) (5)

For our benchmark assumptions, we use p = 0.06, d = 16 inches per acre, and E_plant = 0.75.

To estimate the incremental energy costs for the length and lift of additional mainline systems, we need to compute the additional TDH due to the additional mainline length and lift.

The Hazen-Williams formula for head loss in a pipe is

H_f = K•((Q/C)^1.852/D^4.87)•(Length) (6)

where H_f is friction head loss (feet/foot), Q is flow rate (gallons per minute), D is the inside diameter of the pipe (inches), K_f is a constant (1,046), length is the length of the pipe (feet) and C is a “roughness factor.”

Based on 11 years of visitor count data, the statistical model estimated indicates that on average for every foot drop in water elevation below full pool, the total number of reservoir visitors (V^TOT) declines by 0.3%:

V_rt^TOT = Σ_d=153²⁴⁴V_rdI_t(1 – 0.003(E_rf=E_rd))

That is, for each foot decline in the water elevation of the reservoir, on average, the number of visitors declines by 0.3% from the estimated base levels (for a given reservoir, in a given month).

Welfare losses ($) from foregone recreation due to a decline in visits, valued at $55 per visit, in year t are estimated as:

W_rt^REC=Σ_rΣ_d=153²⁴⁴ -0.165•V_rdI_t (E_rf-Erd)

The coefficient -0.165 equals the average value per visit ($55) times the visitor response to declining water levels (-0.003). The scarcity or marginal value of water ($ per acre-foot) for recreation at reservoir r, on day d, in year t, also reflects how water elevation changes with fill volume (Q) at each reservoir:

dW_rt/dQ_d = -0.165•V_rdI_th(E_rd) for E_rd < E_rf

dW_rt/dQ_d = 0 for E_rd = E_rf

The water level–volume relationship, h(E_rd) in feet per acre-foot, is computed each day, d = 153–244, at each reservoir based on a quadratic estimation of the inverse function of the storage volume, S = f(E), taking the derivative, and inverting the marginal value (slope) to get dE/dS, as shown in Table A17. (For Fern Ridge, the estimation is based on a second-order polynomial).

Appendix 6. Water rights model

Under Oregon law, all water is publicly owned. A permit from OWRD is required for irrigation, municipal water use and other uses.

The water allocation module in WW2100 simulates the seniority system for water allocation that is used in Oregon and other western states. This system allocates water according to water right priority date (first date of use historically); it is based on western water law known as the prior appropriation doctrine — “first in time, first in right.” Under Oregon law, surface or groundwater must be put to a “beneficial purpose without waste.” Final certificates of water rights define the timing of use, the maximum rate of diversion, and the annual volume (duty) allowed under the water right. When conflicts arise due to shortage, the more senior water right is given priority, and more junior water rights are required to curtail their water use if it conflicts with the appropriation of water by a senior water right holder. Water rights may be transferred between points of use under Oregon law when transactions are arranged by parties and approved by the OWRD.

In cases of regional shortage, when available water supplies are insufficient to meet all needs (such as during drought), Oregon’s statutes give priorities for domestic uses. In this situation, preference is given in the following order: human consumption, livestock consumption and all other beneficial purposes. Oregon water law also provides protections for minimum perennial streamflows for the environment.

Water rights model

Water is allocated in our model with a water rights and allocation model (AltWaterMaster) that simulates an approximation of the process that occurs in reality. The model takes account of the demand or request for water at a given point of diversion on a given day (from a farm, city, rural residential water user, or in-stream flow water right). It evaluates the availability of water from the relevant streams and groundwater source and, if sufficient water is available, it withdraws water to satisfy the demand. If there is insufficient water to meet the needs of an existing water right, the algorithm will determine whether a junior water right in the same river reach or any upstream reach could be curtailed in order to make water available to satisfy the senior water right.

In the case of an IDU with an irrigation water right, there are three steps: (1) crop choice, (2) irrigation decision and (3) biophysical water requirements. Depending on the irrigation decision and crop choice outcomes, soil moisture and crop ET are estimated for each day from planting to harvest. If soil moisture falls below a level adequate to meet the needs of the crop at a particular stage of growth, there will be an “irrigation request” for an amount of water needed to satisfy crop growth requirements. The computation of the amount of water requested, therefore, depends on human decisions, crop development, temperature, precipitation and soil moisture. In the case of urban water rights, urban water demand is estimated as described in Appendix 2.

In the case of an in-stream water right, the amount of water “requested” is the amount provided for as minimum flow for the in-stream water right. Under Oregon water law, in-stream flows are counted as a beneficial use so that in-stream water rights are protected from competing uses in the same way that out-of-stream beneficial uses are protected. State-sponsored studies in the early 1960s recommended in-stream flow levels needed to support native fishes in major streams, based on a series of Basin-specific reports. These reports recommended specific in-stream flows, month by month, needed to support anadromous salmonid species. These recommendations were then used by the OWRD to set minimum perennial streamflows throughout Oregon by administrative rule (ODFW, 1997). Figure A4 shows the sum of both existing and unconverted in-stream water rights at the outlets of the WRB tributaries (all of which flow into the Willamette mainstem). Figures A5–A8 show these water rights individually for each tributary.

Many of the flows adopted were set at levels well below those recommended, especially during summer months. In 1987, the Oregon Legislature supplemented the perennial flow law with SB 140 (the Instream Water Right Act). Legislators sought to maintain water levels that support public uses within natural streams or lakes; these in-stream water rights are held by the state in trust to support public uses such as recreation, pollution abatement, navigation and maintenance and enhancement of fish and wildlife and their habitats. Under Oregon law, water rights such as irrigation water rights may be temporarily leased or permanently transferred to in-stream use on a voluntary basis.

In-stream flows in the WRB are maintained at prescribed levels in two main ways. State law, as described here, is one way. The other mechanism comes from Biological Opinions for threatened or endangered species of anadromous fish under the Endangered Species Act. In the WRB, there are several such species, and minimum flows to support their habitat are the basis for flow requirements on the mainstem and major tributaries. The BiOp flow requirements are tied to downstream control points. The control points at Salem are represented in Figure A9. Figures A10–A12 show required flows below each reservoir.

In general, these flows are maintained by way of dam operations above the control points; reservoir management rules require the release of water to satisfy these BiOps.

Irrigation and in-stream rights are satisfied in order of priority date for the whole Basin, with the senior appropriated rights satisfied first. If water cannot be diverted from the stream reach without dropping the water level below the minimum in-stream flow for the reach, and if this occurs for seven consecutive days, then the right is shut off for the remainder of the season. In the case of urban water rights, the prioritization for large cities occurs based on a range of factors other than the priority date of the water right.

The USBR has authority to enter into contracts with irrigators in the Basin to supply water from storage in the federal USACE reservoirs. Any contract for stored water must also be accompanied by an Oregon water right indicating the corresponding point of diversion, place of use, use, maximum rate and duty. Currently USBR is authorized to offer water contracts only to agricultural users.

The total amount of stored water in years when the reservoirs are filled is approximately 1.6 million acre-feet. However, current ESA-related requirements for April–October minimum flows have resulted in a cap on the stored water contracts allowed from federal dams at 95,000 acre-feet of the total 1.6 million acre-feet (NOAA, 2008b). This level was increased in 2003 from 85,000 acre-feet. Currently, approximately 80,000 acre-feet are allocated under service contracts for irrigation in the Basin.

Initially, Congress authorized USBR to issue contracts for stored water for agricultural uses only. However, the Flood Control Act of 1950 expanded the authority of the USACE to (potentially) include municipal and industrial water supply as an intended and authorized project purpose. Currently the USACE has not issued any contracts to municipal or industrial users, but reallocation of stored water to these uses is currently under review as part of a multiyear process to assess the existing restrictions.

Water rights input data

The water rights model requires a detailed input data set to represent the irrigation, municipal and in-stream water rights in the Basin, with specific details about the locations (point of use and point of diversion), allowed use, priority date, maximum rates and duties. The input data for the model were based on OWRD GIS files for points of use (POUs), points of diversion (PODs) and tabular data. These data were intersected with the model’s IDU parcels and stream layer. Overlaps between POUs and PODs were not exact, but were approximated to achieve a high correspondence between the modeled and the actual existing water rights in terms of size, location and other characteristics. The input data set includes 15,413 irrigation water rights, 1,024 municipal water rights and 93 in-stream water rights. Of the irrigation water rights, 8,678 are surface water (232,720 acres) and 6,735 are groundwater (228,800 acres). A parcel of farmland cannot have more than one “primary” irrigation water right, which can be used if sufficient water is available. In some cases, farmers have a “supplemental” water right, usually a groundwater right, which can be used only if the primary water right is unavailable. Given the relatively small number of supplemental water rights in the WRB, these water rights are not included in our model. We also omitted some minor and nonconsumptive water right types such as hydropower and fluming.

Because the vast majority of irrigation water rights in the WRB have identical start dates (March 1), end dates (October 31), maximum rates (1/80 cubic feet per second per acre) and duties (2.5 acre-feet per acre), these values were applied uniformly in the model. For municipal and in-stream water rights, specific rates (maximum or minimum) were applied to each water right.

In the case of municipal water rights, most large cities have multiple water rights, including surface-water and groundwater rights, utilizing more than one point of diversion. Use of these water rights

is prioritized based on a variety of considerations, including costs, water quality, seasonal availability, etc. It is therefore impossible to predict which water right will be used, or used first, based on the water right priority date. To simulate the sources preferred by the large cities in the Basin, water use reports for recent years were used to apportion the urban water demand to the water rights that have been used the most by each city. See Table A18.

The water right sources, types of storage and forms of conveyance for municipal water supply represent a complex capital-intensive supply system that would be very difficult to accurately model and predict in terms of specifics decades into the future. For example, neighboring cities often buy and sell water from each other. Portland’s main water supply is an out-of-basin source at Bull Run. Portland currently sells about one-third of Bull Run diverted water to other municipalities within the Metro area. Our model reflects an assumption that cities will buy and sell water as needed within the Metro area and will continue to rely on the major surface-water sources used currently. As demand grows, the model allows for additional water to be made available by utilizing Willamette mainstem sources of water, which is in fact what is currently being developed by the Tualatin Valley Water District as the largest new water supply in the Basin.

Appendix 7. Modeling of scenarios

This section describes some of the rationale for the set of scenarios developed for WW2100. Research studies such as WW2100 often use simulation models that compare a base case or “reference scenario” to a set of alternative scenarios (see Figure A13 and Table 19).

One of the most common uses of this kind of approach is for policy analysis, which is described as asking “What if?” questions (for example, what if policy X were implemented?). By comparing a reference, or “business-as-usual,” scenario to an alternative scenario in which a policy change has been introduced into the simulation, the impact of the policy change can be evaluated in a way that is impossible to do in the real world (comparing the outcomes in a world “with” the policy change to what the world would have been like “without” the policy change). Simulations of this kind represent one of the most widely used operations research methods, one that is widely applied for public policy analysis. The rationales for other scenarios described below include an effort to address the uncertainty surrounding some of the assumptions by running scenarios to do “sensitivity analysis.”

Modeling challenges

Systems linking people and nature, often referred to as social-ecological systems, are recognized to be complex, adaptive systems with feedbacks, strategic interactions, variation across space and varying time scales. Such systems pose substantial challenges for modeling. The features and dimensions of these kinds of systems must be studied and understood in an integrated way, because we cannot know in advance which features will be important for the outcome of a given policy question or potential management intervention.

Various kinds of models are used to represent complex social-ecological systems at the regional or global scale. These models are often referred to as “integrated assessment” models (IA). Integrated assessment has been defined by the Intergovernmental Panel on Climate Change (IPCC) as “an interdisciplinary process that combines, interprets and communicates knowledge from diverse scientific disciplines from the natural and social sciences to investigate and understand causal relationships within and between complicated systems.” Integrated assessment studies using spatial-temporal simulation models like the one developed in this project are an important tool to meet the needs of decision makers faced with policy and management questions. Given the complex nature of social-natural systems, the broad objective of these kinds of integrated models is to understand the direction and magnitudes of change in relation to specific management interventions so as to be able to differentiate between associated outcome sets.

Key elements of multidisciplinary, integrated frameworks of this kind include an interdisciplinary and participatory process of combining, interpreting and communicating knowledge from diverse scientific disciplines to achieve a better understanding of a system and its future trajectory. Best practices include making assumptions transparent in both natural and social science components, and also including decision makers and other stakeholders in the model development process.

Reference case or “baseline” scenario

Model scenarios should contain as many of the elements that shape a society as possible and should aim to be a coherent, internally consistent and plausible description of likely future states. Such a model can be used to predict future trends and inform policy makers of potential decisions and their consequences. With limited knowledge of processes determining both social change and biophysical change, multiple scenarios are needed to represent a plausible range of alternative outcomes. This kind of sensitivity analysis explicitly acknowledges the inherent uncertainty in a model and its projections. One category of scenarios described below represents sensitivity analysis.

The most common application of integrated models is as a tool to aid policy makers and public sector agency managers. The first step is to choose a “reference scenario,” also called a “business-as-usual,” or baseline, scenario. The idea is to include in this scenario what is believed to be the most likely future trajectory in the system under study, in the absence of any unexpected intervention. This base case then becomes a reference point against which alternative scenarios can be modeled and compared to the reference case.

The reference scenario, by including as many relevant elements of the social-ecological system as possible, makes it possible to project a trajectory of future change that provides evidence of changes in resource supply and demand, scarcity and value. The economic components of the model not only project how people will use resources, but also how they will respond to changes in resource availability. These feedbacks in the human components of the model are critical to understanding how people will respond to changes in scarcity. Importantly, where these models include economic indicators and metrics, the trajectory in the reference scenario (especially when compared to alternative scenarios) generates output that allows policy makers and stakeholders to explicitly see evidence of costs, consequences and tradeoffs involved in different courses of action. A strength of the WW2100 model is the ability to estimate the empirical magnitude of key tradeoffs so that decision makers are better informed about the true costs and benefits of alternative policies.

The reference scenario addresses the first two of the three objectives of the project: (1) to identify and quantify the linkages and feedbacks among human, hydrologic and ecologic dimensions of the water system, and (2) to make projections about where and when human activities and climate change will impact future water scarcities and to evaluate how biophysical and human system uncertainties affect those projections. The third objective of the project is addressed with the alternative scenarios described below and compared to the reference scenario: (3) to create “alternative scenarios” in which one or more policy levers or other interventions have been introduced into the model and to evaluate how these interventions affect future water scarcities (relative to the reference scenario).

Alternative scenarios

In this study, we include five types of alternative scenarios for different purposes. First, we have scenarios that represent “sensitivity analysis.” These scenarios are intended to evaluate how sensitive the model projections are to assumptions about specific parameters or assumptions. Specifically, we have varied our assumptions about “external drivers” — those aspects of the system that are essentially beyond the control of individuals or policy makers in the region. We include here our assumptions about how climate, population and income will change in the future.

Second, we have a range of policy analysis scenarios. These scenarios are intended, as described above, to understand the direction and magnitude of change in relation to specific management interventions so as to be able to differentiate between associated outcome sets and their costs and consequences. Changes in water prices, water law and reservoir management rules would come under this category of alternative scenario. These scenarios are integrative because a change in one part of the model has linkages and feedback effects elsewhere in the model. For example, a change in rules for UGB expansion will have implications for both municipal and agricultural land use in specific areas, which in turn will affect in-stream flows and reservoir management decisions. One advantage of running a policy scenario with one change or intervention relative to the reference scenario is that it allows us to attribute changes in outcomes to that individual change in policy or other model modification. Often policy analysis combines policy interventions in the same scenario. In some cases, this strategy is intended to evaluate whether a second intervention can offset the adverse side effects of the first intervention — ones that were identified when the first intervention was evaluated individually in a prior alternative scenario.

Two of the alternative scenarios included here involve changes to water rights: one for new irrigation water rights tied to stored reservoir water and one for in-stream water rights created in the 1960s but currently “unconverted.”

Third, we have several scenarios described as “counterfactual” scenarios. The term “counterfactual” refers to the idea that these scenarios intentionally model a situation that will not occur, that is, that is counter to the facts. These scenarios can help us measure the impact of changes in the model that have occurred, or will occur, by comparing one scenario intended to represent what we expect to happen (that is, the reference scenario) with a scenario that omits one source of change, for example. We have included one scenario of this kind, in which climate does not change (but population grows), and one scenario in which there is no change in population or personal income (but where climate change does occur). Comparing these counterfactual scenarios to the reference scenario provides insights into the magnitude of the effects on future water scarcity due to one of these elements versus the other. In one other case, we have modeled a scenario with no agricultural production (all fallow). This is clearly counter to the facts; however, a comparison of this scenario to the reference scenario or other alternative scenarios will provide evidence of the impact of agriculture on water use and streamflows.

Fourth, we have modeled historical scenarios. Two scenarios covering a 60-year time span from the past (1950–2010) are used to compare modeled historical conditions with modeled future conditions.

Fifth, additional scenarios, referred to as “integrated scenarios,” combine changes in multiple scenario elements and provide a way to evaluate the combined effect of multiple, simultaneous changes to the regional system. These scenarios were developed in collaboration with a group of interested stakeholders, the Technical Advisory Group (TAG), a group with diverse expertise in WRB land and water use and management. One scenario developed by the TAG was dubbed the “Extreme scenario” because it combines high change climate with high population growth and other model settings designed to maximize resource use and water demand for cities and agriculture. One other integrated scenario developed by the economics researchers was the

“Worst case” scenario. This scenario combined mid-range climate change with high population growth, increased fire suppression, high utilization of irrigation and the certification of all in-stream water rights. It also allowed for new irrigation at half the costs estimated in the New Irrigation scenario.

Not all scenario results are reported here. See additional information about the WW2100 project.

Table A19. Descriptions of selected reference case and alternative scenario components - Model component and descriptions

Climate

Reference case: Climate inputs are from the regionally downscaled projections from the MIROC5 global climate model with the RCP 8.5 emissions scenario. Projections are in the middle of the range of possible changes predicted by a suite of global climate models that perform well for the Pacific Northwest. Annual mean temperature in the Willamette River Basin (WRB) increases about 4° C (7.5° F) over the century.

Alternative scenario: Low Climate Change — GFDL Model (LowClim). Projections are from the GFDL-ESM2M model, RCP 4.5. WRB annual mean temperature increases about 1° C (2° F) over the century.

Alternative scenario: High Climate Change — HadGEM Model (HighClim). Projections are from the HadGEM2–ES model, RCP 8.5. WRB annual mean temperature increases about 6° C (10.5° F) over the century.

Population and Income growth

Alternative scenario: High Population Growth (HighPop). Population growth rates within UGBs are doubled relative to the Reference Case. Basinwide population in 2100 = 8.25 million; population in Portland UGB in 2100 = 4.89 million.

Alternative scenario: Zero Population Growth (NoPopGrowth). Population remains at 2011 levels throughout the century; income rises as in the Reference Case.

Alternative scenario: Zero Income Growth (NoIncGrowth). Income remains at 2011 levels throughout the century; population rises as in the Reference Case.

Alternative scenario: Zero Population & Income Growth (NoGrow). Population and household income remain at 2011 levels throughout the century.

Reference case: The level of wildfire suppression is held at historical rates throughout the simulation. However, because of changing climate and forest conditions, the area of forest burned per year rises over the simulation from 0.2%/year in 2010 to 0.6%/year in 2100. Harvest by clear-cut is maintained at historical rates (8,000 acres/year on public lands + 29,000 acres/year on private lands). There is no harvest of protected areas. Harvest occurs only in stands older than 40 years on private lands, and between 40 and 80 years old on public lands.

Alternative scenario: Upland Wildfire Suppression (FireSuppress). Fire suppression efforts are increased to hold the area burned annually to historical rates.

Urban Development

Alternative scenario: Relaxed Urban Expansion (UrbExpand). UGBs expand when 70% of the land within the UGB is developed; no urban reserves.

Agricultural water demand

Alternative scenario: Limited Irrigation Rates & Duties (LowIrrig). The legal maximum irrigation rate is reduced from 1/80 cfs/acre to 1/100 cfs/acre. The duty is reduced from 2.5 to 2.0 acre-feet/acre.

Alternative scenario: Higher Irrigation Usage (HighIrrig). The average fraction of irrigation rights utilized in a given year is increased from two-thirds (Reference Case) to five-sixths.

Alternative scenario: All Fallow (AllFallow). Crop choice is set to "fallow" for all agricultural lands (including trees and orchards); no irrigation.

Water rights

Alternative scenario: New Irrigation Rights (NewIrrig). New irrigation contracts for stored water in the Willamette Project and related rights are introduced 2015–2044. The probability of adopting irrigation depends on the profitability of irrigating, given increased revenue and increased costs for irrigation infrastructure and pumping, as well as contract fees (about $9/acre).

Alternative scenario: New Instream Water Rights (NewInstream). New water rights are introduced in the model in 2010 (with priority dates for 1960s), reflecting recommended minimum flows established by the state.

References

• Abbott, J. and H. Klaiber. 2011. An embarrassment of riches: Confronting omitted variable bias and multiscale capitalization in hedonic price models. The Review of Economics and Statistics 93(4):1331–1342.
• Amos, A. 2014. Developing the law of the river: The integration of law and policy into hydrologic and socio-economic modeling efforts in the Willamette River Basin. University of Kansas Law Review 62:1091.
• Bell, D. and R. Griffin. 2011. Urban water demand with periodic error correction. Land Economics 87(3):528–544.
• Bigelow, D. 2015. How Do Population Growth, Land-use Regulations, and Precipitation Patterns Affect Water Use? A Fine-scale Empirical Analysis of Landscape Change. Doctoral dissertation, Oregon State University.
• English, M. 2015. Irrigation and Pumping Textbook (unpublished).
• Fipps, G. 1995. Calculating Horsepower Requirements and Sizing Irrigation Supply Pipelines. Texas Agricultural Extension Service. Publication B-6011, Texas A&M University.
• Fischer, F. and G.J. Miller (eds). 2006. Handbook of Public Policy Analysis: Theory, Politics, and Methods. CRC Press.
• Fisher, D. 1972. An economic analysis of farms producing grass seed in the Willamette Valley, with special attention to the cultural practice of field burning. Doctoral thesis, Oregon State University.
• Gaudin, S. 2006. Effect of price information on residential water demand. Applied Economics 38:383–393.
• Hausman, J. and W. Taylor. 1981. Panel data and unobservable individual effects. Econometrica 49(6):1377 –1398.
• Holman, I., et al. 2005. A regional, multi-sectoral and integrated assessment of the impacts of climate and socio-economic change in the UK. i 71(1-2):9–41.
• Intergovernmental Panel on Climate Change (IPCC). 2001, Technical Summary. Climate Change 2001: Impacts, Adaptation, and Vulnerability. A Report of Working Group II of Intergovernmental Panel on Climate Change. Cambridge University Press, New York.
• Jakeman, A. and R. Letcher. 2003. Integrated assessment and modelling: Features, principles and examples for catchment management. Environmental Modelling & Software 18(6):491–501.
• Kalinin, A. 2013. Right as Rain? The Value of Water in Willamette Valley Agriculture. Master’s thesis, Oregon State University.
• Levin, S., et al. 2013. Social-ecological systems as complex adaptive systems: Modeling and policy implications. Environment and Development Economics 18(02):111–132.
• Lewis, D., A. Plantinga, E. Nelson and S. Polasky. 2011. The efficiency of voluntary incentive policies for preventing biodiversity loss. Resource and Energy Economics 33:192–211.
• Loomis, J. 2005. Updated Outdoor Recreation Use Values on National Forests and Other Public Lands. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station.
• Lubowski, R., A. Plantinga and R. Stavins. 2006. Land-use change and carbon sinks: Econometric estimation of the carbon sequestration supply function. Journal of Environmental Economics and Management 51:135–152.
• Mansur, E. and S. Olmstead. 2012. The value of scarce water: Measuring the inefficiency of municipal regulations. Journal of Urban Economics 71:332–346.
• Mendelsohn, R., W. Nordhaus and D. Shaw. 1994. The impact of global warming on agriculture: A Ricardian analysis. The American Economic Review 84(4):753–771.
• Moore, K. 2015. Optimizing Reservoir Operations to Adapt to 21st Century Expectations of Climate and Social Change in the Willamette River Basin, Oregon. Doctoral dissertation, Oregon State University. (accessed November 7, 2016).
• Mundlak, Y. 1978. On the pooling of time series and cross section data. Econometrica 46(1):69–85.
• National Oceanic and Atmospheric Administration (NOAA). 2008a. 2008–2023 Willamette River Basin Project Biological Opinion. NOAA Fisheries Log Number F/NWR/2000/02117. NOAA, National Marine Fisheries Service, Northwest Region, Seattle.
• National Oceanic and Atmospheric Administration (NOAA). 2008b. Consultation on the Willamette River Basin Flood Control Project. NOAA Fisheries Log Number F/NWR/2000/02117. NOAA, National Marine Fisheries Service, Northwest Region, Seattle.
• Nolin, A. 2012. Perspectives on climate change, mountain hydrology, and water resources in the Oregon Cascades, USA. Mountain Research and Development 32:S35–S46. 10.1659/mrd-jour-nal-d-11-00038.s1.
• Olmstead, S., W. Hanemann and R. Stavins. 2007. Water demand under alternative price strucures. Journal of Environmental Economics and Management 54:181–198.
• Olmstead, S. 2009. Reduced-form vs. structural models of water demand under non-linear prices. Journal of Business and Economic Statistics 27(1):84–94.
• Olmstead, S. 2010. The economics of managing scarce water resources. Review of Environmental Economics and Policy 4(2):179–198.
• Oregon Department of Fish and Wildlife (ODFW). 1997. Backgrounder: Instream Water Rights. (Accessed December 12, 2016).
• Portland Water Bureau. 2010. Water Management and Conservation Plan.
• Portland Water Bureau. 2013. Conservation Rate Structure Review.
• Renwick, M. and R. Green. 2000. Do residential water demand side management policies measure up? An analysis of eight California water agencies. Journal of Environmental Economics and Management 40(1):37–55.
• Rotmans, J. and M. Van Asselt. 1996. Integrated assessment: A growing child on its way to maturity. Climatic Change 34(3-4):327–336.
• Timmins, C. 2003. Demand-side technology standards under inefficient pricing regimes: Are they effective water conservation tools in the long run? Environmental and Resource Economics 26:107–124.
• Turner, R. 2000. Integrating natural and socio-economic science in coastal management. Journal of Marine Systems 25(3):447–460.
• U.S. Environmental Protection Agency (EPA). 2009. 2006 Community Water System Survey.
• U.S. Environmental Protection Agency (EPA). 2013. Drinking Water Infrastructure Needs Survey and Assessment. Fifth Report to Congress. EPA 816-R-13-006. (Accessed December 14, 2016).
• U.S. Geological Survey (USGS). 2012. Land Cover Trends. (Updated December 13, 2012; accessed March 10, 2013).
• Wooldridge, J. 1996. Selection corrections for panel data models under conditional mean independence assumptions. Journal of Econometrics 68:115–132.
• Wooldridge, J. 2003. Introductory Econometrics: A Modern Approach, 2nd edition. Southwestern, Mason, OH.