Climate Methods in Brief

Climate Model Evaluation

The World Climate Research Programme’s Coupled Model Intercomparison Project (CMIP) is a worldwide effort to establish a set of standard experimental protocols for the use of general circulation computer models (also called global climate models), or GCMs, in the development of climate scenarios. Essentially, the CMIP project is an attempt by the world’s climate modelers to improve the performance of GCMs, standardize methods of model evaluation, and make GCM outputs directly comparable.

Results from the CMIP project’s latest phase (Phase 5 or CMIP5) began to be available around the time of the launch of WW2100. CMIP5 ushered in a new wave of GCMs and resulting climate data, and was seen by the WW2100’s climate modeling team as an opportunity to both use the new CMIP5 GCMs in WW2100’s work and assess the performance of the CMIP5 models for the United States Pacific Northwest as a whole.

To that end, the WW2100 climate team assessed 41 GCMs from CMIP5 for their ability to simulate various aspects of climate in the Pacific Northwest. The team’s results were published in the Geophysical Research Letters: Atmospheres, where the researchers’ full results and methods can be reviewed. What follows is a summary of the methods employed and how it aided WW2100.

As the WW2100 team stated in their paper, their goal in evaluating the CMIP5 models was “to evaluate model performance in order to make informed recommendations to those who may use these model outputs” (Rupp et al., 2013). The researchers determined these “downstream” users, including resource managers and other scientists assessing climate impacts, would be best served by GCMs that gave the best statistical fit to the observed climate of the Pacific Northwest. (The team defined the Pacific Northwest as the area within longitude 124.5° and 110.5° W and in latitude 41.5° and 49.5° N, or roughly Oregon, Washington, Idaho, and western Montana.)

To find the best fit, the team evaluated the CMIP5 GCMs according to their ability to re-create in computer simulations the observed historical climate of the 20th century. This hindcasting ran from 1850-2005 and focused on temperature and precipitation. Observed climate data were taken from five gridded datasets of monthly means. A suite of statistics, or metrics, were calculated from both the hindcasts and observations and then compared. These metrics included mean seasonal values, interannual variability, amplitude of the seasonal cycle, consistency in spatial patterns, and sensitivity to the El Niño Southern Oscillation, among others. The researchers used two methods for ranking the performance of the GCMs based on these metrics: The first method assigned equal weight to each of the metrics. The second method excluded those metrics that were not considered robust. This involved ranking the metrics under the assumption that certain metrics might be more important in the assessment process. It also included an attempt to avoid redundancy, given that not all metrics are independent of one another.

The result was a ranking of the models according to metrics that led to a subset of GCMs that the team’s methods determined were the best statistical fit for the climate of the Pacific Northwest.

Emissions Scenarios Selection

The biggest certainty in climate science is that increasing greenhouse gas (GHG) concentrations, especially carbon dioxide, are heating the Earth’s atmosphere. Precisely how much regional climate temperatures will increase in response to a given rise in GHGs is not known, which is why climate researchers examine more than one GCM. However, the biggest uncertainty about future climate by the end of the 21st century stems from not knowing just how much GHG human industry will continue to emit.

With their chosen models, the WW2100’s climate team used GCM output (available from the CMIP5 project) for two emission scenarios to incorporate a range of GHG concentration uncertainty. These emission scenarios are known as Representative Concentration Pathways, or RCPs, a category created by researchers convened to support the work of the United Nation’s Intergovernmental Panel on Climate Change (IPCC).

RCPs are the new standard for modeling emission uncertainty. There are four RCPs representing different concentrations of greenhouse gases, or different possible futures based on how much GHGs human industry might emit. These scenarios are: RCP 8.5, RCP 6, RCP 4.5, and RCP 2.6. Here, higher numbers represent a greater degree of radiative forcing (in terms of W m2 over preindustrial levels; e.g., 8.5 = 8.5 W m2) that the scenarios are expected to produce by 2100 (RCP 8.5 is bigger than RCP 6 and so on) and, hence, represents a future scenario with more emissions.

For WW2100,  we employed two RCPs: RCP 8.5 (the high emissions scenario that assumes human industry will continue to emit greenhouse gases at a growing rate) and RCP 4.5 (a middle-of-the-road scenario in which emissions will be curbed starting in the middle of this century).

Selecting Representative Climate Scenarios Using a Sensitivity Approach

Resource managers and other downstream users of GCM data often lack the ability to process data from multiple climate models and scenarios, which requires considerable computing resources. This is especially true when considering other types of future scenarios in their impacts work, such as economics, demographics, and land uses, which require still more computing resources. What is often done instead is to select a subset of representative models and scenarios. For WW2100,  we selected a subset of three “representative” scenarios, that is scenarios that are representative of GCMs run with the RCPs, the data from which could then be fed into WW2100’s latter modeling.

To find their subset of scenarios, the team conducted a sensitivity analysis in the Willamette River Basin that allowed them to select three representative GCMs from the 33 CMIP5 climate models for which future climate scenarios were available (from the 41 GCMs evaluated in the first step).

The sensitivity analysis used simple perturbation experiments to estimate how changes in temperature and precipitation affect summertime streamflow in the Willamette River Basin. More specifically, using the Variable Infiltration Capacity (VIC) Macroscale Hydrologic Model, we simulated streamflow using historical weather data from 1975-2004. Then, we ran a perturbation: Using the same set up, we ran the hydrologic model again, but with incremental increases in temperature. The percent in which summertime streamflow changes in these perturbation experiments provides an estimate for how sensitive it will be in a warmer climate. The same type of perturbation experiments were then repeated, but this time for incremental precipitation increases and decreases.

From this, we then used these derived sensitivities to draw contours of constant summertime streamflow change on a scatter plot of temperature and precipitation changes in GCM output. With the contours as guides, the team selected GCMs and accompanying RCPs with the objective of spanning a wide range of warming (high, middle, and low) while also spanning a wide range of hydrological impact. Where multiple models were available to choose from in each category (high, middle, low), the team chose one of the better performing GCMs according to the model ranking discussed above.

From this analysis, three scenarios, now a combination of GCMs that had been run with the two separate RCPs, were ultimately selected.

The final representative selections are: the High Climate Change scenario (the HadGEM2−ES climate model run with RCP 8.5); the Reference Case scenario (the MIROC5 climate model run with RCP 8.5); and the Low Climate Change scenario (the GFDL−ESM2M climate model run with RCP 4.5). Note: GFDL-ESM2M was not ranked highly by performance, but it was selected nonetheless to represent the Low Climate Change scenario, as none of the lowest-warming models were ranked highly. Hereafter the scenarios will be referred to as HighClim, Reference, and LowClim.

The LowClim scenario represents a small temperature increase and small decrease in summertime streamflow (lowest impact). The HighClim scenario represents a large temperature increase and large decrease in summertime streamflows (highest projected impacts). The Reference scenario lies between the two extremes.

It’s worth noting that an additional constraint was placed on the selection process: that all requisite data for a given GCM was available for the downscaling procedure (see section on downscaling below). This constraint ultimately limited the team to choosing from among 20 GCMs.

Downscaling

In order to feed data from the HighClim, Reference, and LowClim scenarios into WW2100’s later modeling work, the team performed a process called downscaling. Downscaling is a means to convert or translate the coarse resolution of GCM grids (which are as large as 375 km, roughly 233 miles, to a side) down to a finer resolution, which for WW2100 was about 4 km, roughly 2.5 miles. This is done to account for the details of local topography and local climate. This adjustment is needed in the Pacific Northwest, which has a complex mountainous topography that does not appear in a detailed form in the GCMs.

For the WW2100, we downscaled data from the HighClim, Reference, and LowClim scenarios for the Willamette Valley. The method used was the Multivariate Adaptive Constructed Analogs (MACA), a downscaling method developed by University of Idaho (UI) researcher John Abatzoglou. Resulting data from the downscaling was then fed into the Envision model’s component models.

Land Use Change Modeling in Brief

The land-use team simulated how land moves among agricultural, forest, and urban uses. These transitions are a function of economic returns to alternative uses, which are determined themselves by site characteristics such as farm rents, distances to cities, and population and income of cities. Returns to land and land-use transitions also are influenced by urban growth boundaries (UGBs), the land use planning lines that restrict development. Here we provide a brief explanation of land-use change modeling in Willamette Envision. For a more detailed explanation refer to Jaeger et al. (2016) and Bigelow (2015).

During a model run, Willamette Envision simulates land-use changes annually at the scale of parcels (referred to as Integrated Decision Units or IDUs within Willamette Envision). Growth in population and income increase the returns to developed uses relative to forest and farm uses, while agricultural land values are influenced by the availability of water for irrigation. We derived functions that estimate the economic returns to each land use from historical data for the Willamette Valley (Bigelow, 2015). In response to changes in the relative returns to different uses, IDUs may shift among agricultural, forest, and urban uses as the model runs. Land-use changes in the model are governed by probabilistic equations estimated with historical data from the National Resources Inventory (NRI).

To account for zoning rules under Oregon’s land-use planning system, the model treats land inside and outside of UGBs differently. Land outside of UGBs can move between undeveloped uses (i.e., ag-to-forest and forest-to-ag transitions are allowed), but transitions to developed use are not allowed. For IDUs inside of UGBs, all of the transitions are allowed, but development is treated as irreversible (i.e., once a forest or agricultural IDU is developed, it cannot return to its original use). 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 allowed for UGBs to expand once the developed share became sufficiently large. Once a specified threshold is exceeded, a UGB expansion will be triggered. UGB expansions occurred in a way that approximated rules under the land-use planning system. For example, land zoned for Exclusive Farm Use and Forest Conservation will be brought inside of UGBs only when other opportunities are exhausted. In the case of the Portland Metropolitan Area, priority is given to areas designated as urban reserves.

Population and Income Growth

Willamette Envision treats population and income growth as external drivers, forces that are determined outside of the model and can be varied for different modeling scenarios. The population and income projections adopted for the Reference Case and many other WW2100 modeling scenarios, were based on forecasts from the Oregon Office of Economic Analysis (2011; OEA) and Woods and Poole Economics, Inc. (2011). The OEA forecasts population for each county to 2050, while Woods and Poole forecast mean household total personal income to 2040 (in inflation-adjusted dollars). We used linear extrapolation to 2100, which implied diminishing growth rates over time for both variables (Fig.1). Projected county population was allocated to areas within UGBs and rural residential zones based on population percentages in the 2010 Census. Rural residential zones were allowed population growth until they reached a density of one household for every two acres of land. Once these areas were full, all new population was added within UGBs. All cities within a county were assumed to have the projected mean household income.

Forest Modeling in Brief

The forest modeling team developed component models for Willamette Envision that simulate how upland forests will age and change through time, given forest type, climate conditions, and disturbance by wildfire and harvest. Here we provide a brief explanation of forest modeling in Willamette Envision. For full details on methods and results from WW2100 forest modeling studies, refer to Turner et al. (2015, 2016).

On an annual basis, Willamette Envision determines the forest type and age in each modeling polygon based on information from models of forest growth and succession (called forest state-and-transition models, STMs). The STMs determine the sequence of forest types that occur over time. In the simplest case, the forest progresses from new growth following a disturbance, through different successional stages, and ultimately to an old growth forest state. Depending on the type and timing of disturbances, forest growth and succession can follow alternate pathways specified by different STMs. When a disturbance occurs, the forest can also “reset” to a new forest type better suited to the current climate conditions. These new “potential” vegetation types were determined for the three WW2100 climate scenarios using offline runs of a dynamic global vegetation model called MC2. MC2 simulates wildfire occurrence and simulates the type of vegetation best suited to grow at a location based on climate, soil, elevation, and latitude. 

Willamette Envision simulates forest harvest on the landscape, according to criteria defined for each scenario. For example, a scenario can specify a harvest rate (the total area of forest harvested each year) for specific forest age classes and land ownership categories (e.g., private lands with forests older than 40 years). Over the simulation, Willamette Envision randomly selects modeling polygons that meet the criteria for harvest. Users can also prescribe the extent of wildfire (the forest area burned per year), and Envision places fires randomly on the landscape. In WW2100, the extent of wildfire was determined from historical observations and the offline runs of MC2 for the three WW2100 climate scenarios. MC2 takes into account factors such as air temperature, relative humidity, and ensuing forest moisture conditions, and determines the area of forests that burn each year. Hotter, drier conditions lead to more extensive wildfires.

Additional Modeling Details:

• The initial condition of the landscape, which classifies different species of vegetation, and state and transition models (STMs) were based on work from the Integrated Landscape Assessment Project (ILAP) (Halofsky et al., 2014; INR, 2013). Boundaries for land ownership and protection status were from the US Geological Survey (GAP, 2014).

• Harvest rates in the Reference Case scenario were based on the observed harvest rate from 1986-2010 in the Willamette River Basin (Kennedy et al., 2010; Kennedy et al., 2012). These Landsat-based observations suggested a harvest rate of 1.3% per year across all private forestland, equivalent to ~11,740 hectares per year (29,000 acres) and 0.5% per year on public lands, equivalent to ~3,240 hectares per year (8,006 acres). Harvest on public lands was limited to unreserved stands with ages between 40-80 years, since older forests are largely conserved for wildlife on public lands.

• The initial probability of fire in the Reference Case scenario were based on observations of the Landsat record (Kennedy et al., 2012), and we did not stratify by ownership class. Over the 1986-2010 period, 0.2% per year of forestland area was burned in the Willamette River Basin. The future extent of fire was based on the MC2 results, thus capturing the increasing incidence of fire associated with a warming climate. Annual area burned was input to Willamette Envision, and fires were placed randomly on the landscape. Fire size was 22,500 ha except when only a fraction of that was needed to reach the prescribed total area burned.

• We assigned values for Leaf Area Index (LAI), a measure of forest canopy cover, for each forest type and stand age class based on off-line runs of the Biome-BGC productivity model (Thornton et al., 2002). LAI is used to estimate forest evapotranspiration in hydrologic modeling.

Snow Modeling in Brief

In the Willamette River Basin (WRB), most annual precipitation falls between November and March, with snowfall occurring mainly at elevations above about 1200 m (3937 ft). Sub-basins such as the McKenzie, North Santiam, and Middle Fork, whose headwaters contain substantial area above 1200 m (3937 ft), have more snow than those in the Coast range and elsewhere in the WRB.

We simulated the seasonal evolution of the mountain snowpack at a daily timestep within the Willamette Hydrology Model (WHM) portion of Willamette Envision. WHM is a modified version of the HBV model (Seibert, 1997) in which snow is computed using a degree-day model. Melt rate is governed by air temperature in excess of 0° C and moderated by a melt factor. Precipitation is partitioned into rain and snow by incorporating a temperature-based transition “ramp” rather than a single fixed temperature threshold thus allowing for mixed rain-snow events. The model accounts for the effects of canopy interception and snow sublimation from the canopy as a function of leaf area index (LAI), a measure of canopy cover. Last, we included a radiant energy term as a function of LAI, where both convective heat (as a function of air temperature) and radiant energy can affect snowmelt. Snowmelt is constrained by the amount of actual snow available in each IDU, and when snowmelt exceeds the water-holding capacity of the snowpack, melt is routed into the soil.

Analysis Approach

We selected three climate scenarios to examine the effects of climate and changing forest cover on snowpack. As described in the Climate section, the scenarios representing high, medium, and low levels of climate warming are referred to as HighClim, Reference, and LowClim, respectively. For each climate scenario, the WW2100 team examined snowpack across three elevation zones: 500 m (1640 ft) and below, 500-1200 m (1640-3937 ft), and 1200 m (3937 ft) and above. While April 1 SWE is a traditional metric used by streamflow forecasters, we found that under climate warming this metric does not account for snow water equivalent that melts prior to that date. A more hydrologically appropriate metric is the maximum winter snow water equivalent (MaxSWE). We computed MaxSWE for the January-April time period for elevation zones 500-1200 m (1640-3937 ft, or “low elevation snow”) and 1200 m (3937 ft) and above (“high elevation snow”). Because snow cover below 500 m (1640 ft) is only ephemeral and insignificant, we did not analyze changes in snowpack in that zone. Snowpack in individual winters typically varies from year to year, so our temporal analysis examined decadal and 30-year averages of SWE over the 90-year model run. We also analyzed correlations between snowpack, winter temperature, and winter precipitation on a yearly basis for each of the climate scenarios. In addition to the elevation zone analyses of SWE, we examined decadal-averaged SWE for sub-watersheds (HUC-12 level of detail).

Urban Water Use Modeling in Brief

Water in urban areas of the Willamette Basin is put to residential, commercial, and industrial uses. The amount used will depend on a range of factors including population, price, and income, as well as urban population density. This web page provides a brief introduction to urban water demand modeling in Willamette Envision; for a more detailed description, refer to Jaeger et al., 2016.

The urban water demand component of Willamette Envision consists of models of residential and nonresidential urban water demand for the Portland Metropolitan Area, Salem, Corvallis, and Eugene/Springfield, as well as a separate model for smaller urban areas. We selected model variables based on a review of the economics literature on urban water demand and on the need to use variables that could be forecasted over the entire study period as exogenous drivers (income and population growth) or as variables generated within the Willamette Envision framework (population density). The economics literature suggests that a water demand function must include marginal price of water, pricing structure, and income (Olmstead et al., 2007; Olmstead, 2009; Olmstead, 2010; Bell & Griffin, 2011; Mansur & Olmstead, 2012). Given the specific forecasting needs of the urban water component for WW2100, we also included population and population density in the model. We collected the most current information available (at the time the project was begun) on each of these variables for Portland, Salem, Corvallis, Eugene, and Springfield. We used coefficients from the literature 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. Finally, we adjusted the demand models to reflect seasonal variations in demand. Our baseline scenario, called the Reference Case (1) assumes initial prices that are commensurate with the basin’s major cities in 2010, (2) includes price increases comparable to what occurred on average from 2010-2015, and (3) assumes a 1.5% annual increase from 2016-2025 (in real, inflation-adjusted dollars), in recognition of the existing backlog of infrastructure needs and system maintenance and upgrades expected for western Oregon. After the year 2025, the Reference Case assumes prices remain constant in inflation-adjusted terms for a given size city. For more information about urban demand modeling in Willamette Envision refer to Jaeger et al. (2016).

Additional Modeling Details:

• Many Willamette Basin metropolitan areas have multiple water providers that divert water from multiple sources. Many water providers also buy and sell water between municipalities. Willamette Envision does not model these complex arrangements and instead models water demand for each metro area in aggregate. We used water-use reports from recent years to apportion urban water demand among water rights that have been used most by each metro area. As demand grows, the model allows for additional water to be made available by diverting water from the Willamette mainstem.

• Water supply for the Portland Metro area also includes water sources that are outside of the Willamette Basin including Bull Run (the major source for the city of Portland) and Barney and Scoggins Reservoirs (that serves Metro area customers in the Tualatin River Basin). In the case of Bull Run, we have restricted our model so that no more than two-thirds of total Portland Metro water demand comes from this dominant water right (based on recent water use data). However, the Bull Run water right has a maximum legal rate of 636 cfs, which is 50% above the highest rate of withdrawal in the reference run model (in 2100) with this restriction in place. In addition, our model does not include the mid-Willamette water supply source currently under development by the Tualatin Valley Water District. That source, to be completed in 2026, will have a capacity of 100 million gallons per day, or more than 36,000 acre-feet during the four peak summer months.

• 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 Portland Water Bureau. Total predicted yearly water demand from above is divided by 365 to obtain daily use, and then multiplied by indoor and outdoor fractions to reflect seasonality. Water demands in rural residential zones, relying on groundwater, is also included in our model. It is predicted using the cost of pumping as a measure of the price of water, the population of the rural-residential area, income per household, and population density.

Agricultural Land and Water Use Methods in Brief

Agricultural water use depends on a range of factors. These include (1) the land area on which agriculture is practiced, (2) the choice of crops, (3) whether irrigation water rights are held, the availability of water from a given source, and the usage rates of those irrigation water rights. Water is consumed by plants on a daily basis, whether irrigated or not. For rain-fed agriculture crops, water demand will reduce the amount of water in soils, in groundwater reserves, and the amounts seeping to streams; for surface irrigation, the crop water demand will involve diverting water from surface flows when soil moisture is insufficient to meet crop water needs.

Agricultural Modeling

The agricultural water use models consist of several interconnected economic models including the limitations imposed by water rights and dynamic models of agricultural evapotranspiration and evolving soil moisture.  The models operate at the scale of the Willamette Envision’s map polygons, which are called Integrated Decision Units (IDUs). The four economic models determine: (1) which lands are put to agricultural uses, (2) which crops are grown on a given parcel of land, (3) whether the parcel of land has or will have an irrigation water right, and (4) whether the irrigation water right is used in a given year. These four models interact with the agricultural evapotranspiration model (crop water demand) that simulates daily evapotranspiration (ET) as a dynamic function of climate, land cover, soil water, and growth stage of each crop, from planting date, to ‘greening up,’ to harvest and dormancy. Soil water will vary as a function of precipitation, crop cover, irrigation, and seepage. These four models also interact with water rights that may impose limits on the timing and quantities of water available for irrigation. Farmland transitioned in or out of agriculture versus developed or forest land uses is described in the land use change section. This page provides a brief overview of agricultural modeling in Willamette Envision.  For a more complete explanation, refer to refer to Kalinin (2013) and Jaeger et al. (in prep).

In a given year where a particular land parcel is assigned to the agricultural land use, farmer decisions are modeled to simulate crop and irrigation decisions. Irrigation is only possible 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 (discussed below).

The combination of decisions, choices, actions, and responses to other factors produces a unique pattern of crop water use, irrigation diversions, soil moisture, and groundwater contributions. It also influences economic returns to farming (annual farmland rent) at the parcel level. To the extent that irrigation water is shut off by regulators, current and expected future annual farmland rent is reduced.

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 varying attributes, 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 (i.e., a crop rotation schedule) were found in the data or in interviews with farmers or agricultural extension personnel. The resulting modeled values are interpreted as the probabilities for each crop to be grown. For perennial crops (orchards, vineyards, tree crops), a fixed set of IDUs is permanently assigned.

The model of irrigation decisions is based on a detailed farmer survey conducted for WW2100 by the USDA National Agricultural Statistics Service (Kalinin, 2013). Data on a six-year history of irrigation and cropping practices for a sample of fields from 530 randomly selected farmers was collected. From this, an irrigation decision model was estimated to represent the probability of irrigating a specific parcel as a function of parcel attributes (e.g., soil type, elevation), and seasonal factors (e.g., June precipitation).

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 estimate of farmland rent takes a “Ricardian” approach that is common in models of the economic returns to agriculture (Mendelsohn et al., 1994). 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 rents associated with factors affecting agricultural productivity (see Kalinin, 2013 for more detail).

Agricultural lands in the WRB 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 US 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 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 Bureau of Reclamation 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 a new irrigation water right is attractive to a farmer depends on its profitability. Farmlands without irrigation water rights are given the opportunity to acquire new irrigation water rights based on stored water in the “New Irrigation Scenario,” provided that there is an economic justification for doing so.

Water Rights Modeling

Irrigation water demands compete with other water uses, include instream water rights. Because irrigation and instream water rights have the greatest potential to compete directly, we include a description of these water rights and our modeling of them here.

Water is allocated in the WRB according to Oregon water law, which operates according to the “prior appropriations doctrine” used in Oregon and most western states (Getches et al., 2015). The water rights system allocates water according to water right priority date (first date of use historically). Under Oregon law, all water is publicly owned. Water rights certified by the state are defined in terms of the timing of use, the maximum rate of diversion, and the annual volume allowed under the water right. When conflicts arise due to shortage, the more senior water right is given priority, while more junior water rights are required to curtail their water use if it conflicts with the 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 (Amos, 2008), for example an irrigation right transferred from one farm to another.

Willamette Envision mimics this process: it takes account of the demand or request for water at a given point of diversion (POD) on a given day (from a farm, city, rural residential water user, or instream flow water right), and it evaluates the availability of water from the relevant streams and groundwater source. If there is sufficient water available, it withdraws water to satisfy the demand. If there is insufficient water to meet the needs of an existing water right, the request is denied. At the same time the model determines whether there is a junior water right in the same river reach or any upstream reaches that could be curtailed to make additional water available to satisfy the senior water right. A similar procedure is followed to satisfy instream water rights by protecting flows in streams where such water rights exist. The model includes instream water rights implemented as of 2010. When more than one instream water right applies at the same time to the same reach, the water rights model applies both water requirements. If an instream water right is “senior” to an irrigation water right, the irrigator may be shut off is there is insufficient water to meet both demands. Willamette Envision includes more than 15,000 irrigation water rights, 1,000 municipal water rights, and 90 instream water rights.

Hydrology Modeling in Brief

Within Willamette Envision, we model the storage and flux of water from rain or snow, into soils, groundwater, and streams using a sub-model developed for the project called the Willamette Hydrology Model (WHM). WHM is, in part, based on the widely-used rainfall-runoff model called HBV (Bergström and Singh 1995; Bergstrom et al., 2001; Seibert, 1997). For a more detailed explanation of the hydrologic modeling framework concept, refer to Vache et al. (in review).

WHM translates daily values of meteorological input (including precipitation, air temperature, wind speed, and radiation) into a spatially distributed estimate of water storage and release. It is run on a daily timestep over the full 90-year WW2100 modeling timeframe, and results in a dynamic estimate of the response of the hydrology to the evolving landscape. The hydrologic model is comprised of four overall elements describing key features of the system.

• The mountain snowpack - Each winter, snow accumulates in the higher elevations of the Willamette River Basin (WRB). This natural reservoir serves to store a proportion of the winter precipitation, releasing into streams (and reservoirs) during the spring. We simulate the seasonal evolution of the mountain snowpack with a hybrid approach that includes the influence of both air temperature and radiation. The temperature- and radiation-driven melt depends directly on vegetation, so that the snowpack changes as forest disturbance and growth occurs over the scenario timeframe.

• Watershed runoff - Incoming precipitation and snowmelt are stored within a set of conceptual reservoirs representing soil and groundwater (Fig. 1). WHM defines the spatial distribution of those reservoirs based primarily on a set of approximately 9000 sub-watersheds, defined by the National Hydrography Dataset (NHD). Water is released from the reservoirs based on algorithms developed as part of the HBV model (Bergström and Singh 1995; Bergstrom et al., 2001; Seibert, 1997). Biophysical processes such as evapotranspiration respond to changes in land cover simulated by the forest and human systems models.

• Instream routing and reservoir storage - WHM uses a kinematic wave approach to simulate the flow of water within the stream network. The spatial pattern of the network was taken from NHD, and allows the model to estimate stream discharge throughout the watershed. Manmade reservoirs are a key feature of the Willamette river system, and WHM simulates operation of the 13 largest reservoirs – those that are managed by the U.S. Army Corp of Engineers (USACE) as the Willamette Project. Refer to the reservoir operations page for more details about reservoir modeling.

• Water Allocation and water use - The use of water by human society is the fourth key feature of the hydrology of the Willamette. WHM models the movement of water through the stream network, while also allowing water to be added or removed at specific points in accordance with Oregon water law. Human water demand is estimated in the WW2100 economic models as the outcome of household and farm demand relationships. Refer to the agricultural land and water use and urban water use pages for more information about these models.

Figure 1. Hydrologic modeling within Willamette Envision involves movement of water through a set of conceptual reservoirs simulating soil and groundwater (diagram by K. Vache).

Evapotranspiration

Willamette Envision models both biophysical and human aspects of water use. The primary biophysical use modeled is evapotranspiration (ET) — water movement from soil and plants into the atmosphere. ET is calculated on a daily basis for all modeling polygons and responds to changes in landcover modeled by the forest and human systems modeling components. ET is modeled as follows:

• Forest (~70% of the basin land area in 2010 starting conditions) - For forested lands, WHM models ET using a Penman-Monteith expression that includes both transpiration from vegetation and soil evaporation. The Penman-Monteith expression includes vegetation canopy, represented by leaf area index (LAI), as an independent variable. Thus, forest areas respond to changes in canopy cover simulated by the forest dynamics models. For example, as the forest in a modeling polygon matures, leaf area index (LAI) rises, which leads to a higher ET. When a fire occurs, the LAI decreases to almost zero. Refer to Turner et al. (2016) for more details about modeled vegetation shifts and modeling ET in the Willamette upland forests. Some forest IDUs (about 2% of basin area) were left out of the state-and-transition model due to gaps in the initial condition data. Those forest IDUs are modeled for ET using Penman-Monteith, an LAI of 1, and a canopy height of 10 meters.
• Agricultural lands (~22% of the basin land area in 2010) - On agricultural lands, modeling of ET takes into account crop type and growth stage, and augmentation of soil moisture due to irrigation. The sequence for a modeling polygon is as follows. First, the human systems modeling components set the crop type and determine whether the land will be irrigated that year. Next, at a daily timestep, WHM determines crop growth stage and water requirements. The model then estimates actual crop water use (ET) by taking into account the amount of water available in the soil. Soil water can come from natural sources (e.g., precipitation) or it can be supplemented by irrigation if water is available (as determined by the water rights model). Modeling of crop water demand and ET is based on the crop cover approach as described in the Food and Agriculture Organization (FAO) Irrigation and Drainage Paper 56 (Allen, Pereira, Raes, & Smith, 1998) and further developed by Allen and Robison (2007). Refer to Jaeger et al. (2017) for more details. 
• Urban and other non-forested, non-agricultural lands (~8% of the basin land area in 2010) - Elsewhere, WHM approximates ET using a Penman-Monteith expression using a LAI of 1.0, and a canopy height of 10 meters, that does not change. In urban areas (~5% of the basin land area in 2010) modeling of ET also takes into account outdoor water use for lawns and gardens. Municipal water use is modeled by the human systems models as a unit for each urban growth area. A fraction of the water diverted for municipal use is added to ET. This fraction varies seasonally so that it is very small in winter and higher in summer when outdoor water use peaks. 

Model Calibration

The rainfall-runoff model within WHM has nine calibration parameters related to snowfall, runoff, and percolation. Calibration of these parameters affects the flux between the conceptual reservoirs shown in Figure 1. We used the Parameter Estimation program, PEST, to identify sets of the 9 parameter values which produce good correlations between modeled and measured datasets. The calibration period was 1980-1994 and we used the following measured data sets in the calibration process: USGS stream gage records, NRNI synthetic flow data, Oregon SNOTEL snow survey records, and measured inflow data for federal reservoirs. We ran the model for the calibration period using using meteorological forcings from MACA (Multivariate Adaptive Constructed Analogs) training data derived from observed weather station data (MACAv1-METDATA).

We divided the WRB into a number of sub-basins for calibration purposes, ultimately using 14 different sets of parameter values in different parts of the WRB. Thirteen of the parameter sets were produced by Oregon Freshwater Simulations in 2016; one parameter set was taken from prior calibration work by Eric Watson and Heejun Chang at Portland State University in 2015. Our use of PEST was an adaptation and refinement of the methods used by Watson and Chang (PSU) in 2015.

The main analysis steps within PEST are as follows: (1) PEST selects a set of parameter values, (2) it runs the WHM model, and (3) it calculates a value representing the divergence of the simulation results from the historical stream gage records. This process is repeated with different sets of parameter values in a systematic exploration of the parameter space. The process terminates when successive adjustments to the parameter values fail to reduce the divergence by a specified amount. The model may exhibit equifinality: different sets of parameter values may produce equally good simulation results.

Accurate simulation of the inflows to the WRB reservoirs is important to the success of the WW2100 model as a whole, so we identified nine reservoir drainages to be individually calibrated using PEST as the first step in our process. Parameter values for the remaining parts of the WRB were selected using a variety of methods in subsequent steps. For a more detailed description of model calibration, refer to Jaeger et al. (2017).

Because our focus in WW2100 was on water scarcity, calibration focused on low flows and reservoir inflows. The model was less able to accurately predict high flows, and users should use caution when making interpretations about flood events and winter peak flows.

Model Limitations

As of summer 2016, Willamette Envision does not include a detailed groundwater model and instead relies on the simpler approach of a conceptual groundwater reservoir that is part of HBV, and includes a calibration parameter, which represents groundwater recharge. Because of the simplified groundwater model, we do not model water scarcity for groundwater withdrawals. We simulate water withdrawals from groundwater, but the model does not impose a specific limit on them.

In addition, we do not model potential future changes in groundwater supplied by mountain springs in the High Cascades. Instead, we simulate springflow by adding constant discharge to High Cascades catchments that reflect measured spring flows (Jefferson et al., 2007; Grant & Lewis, 2016). The addition of High Cascades groundwater is held constant throughout the simulations. This assumption is based on measured flows in spring-fed mountain sub-basins.

As of summer 2016, Willamette Envision does not model water quality or water temperature. We are developing an energy balance stream temperature model and hope to incorporate it into future versions of Willamette Envision. WW2100 did include some offline field measurements and modeling of the potential effects of increased water temperature on mainstem fish population. Read about that analysis on the fish and stream temperature section of this website.

Water Budget Calculations

One of our goals for the hydrologic analysis was to determine how the flow of water into and out of the WRB might change in response to changing climate and water demands. WHM allowed us to calculate a water budget for the WRB and estimate how the inputs and outputs to that budget might change over the 21st century. Here we describe the calculations we made to determine the water budget using output from Willamette Envision.

All calculations were performed using daily values that were then averaged over each month of the year. All values were calculated and reported as fluxes (cm3/y per cm2 of the WRB, so cm/y). Most of us are familiar with precipitation as cm/y, but all other inputs, stores, and outputs of water can be reported in the same way. This allowed us to easily compare basinwide precipitation, storage, human diversions of water, and so forth. Monthly, seasonal, and annual fluxes are all reported in the same units of cm/y. Therefore, the annual precipitation can be calculated simply as the sum of the monthly precipitation.

Precipitation is the sum of all rain and snow in the basin divided by the basin area. Evapotranspiration (ET) is all ET in the WRB. The ET related to human uses (municipal and agriculture) is included in the basinwide number and indicates the fraction of total ET based on human uses. Environmental flows are those in the Biological Opinion [U.S. Fish and Wildlife Service, 2008, Appendix B, p. 12, Table 9.2-1] for the Willamette River at Salem. These are the flow objectives for "adequate" and "abundant" water years. We did not include upstream environmental flows because these would be duplicates for the purpose of the water budget. Reservoir storage was calculated as the difference between inflow and outflow each month. Snow was calculated as the difference between snow water equivalent (SWE) on the first of each month and the first of the previous month. Municipal water use was calculated endogenously within Willamette Envision. The ET for municipal uses was calculated as the total water use each month minus the municipal indoor water use, which we estimated each year from Jan. 3 to Feb. 3. Agriculture ET was calculated within Willamette Envision. Storage of water in soil moisture and groundwater was calculated to balance the monthly water budget of the basin. The model does not include evaporation from open water such as reservoirs and lakes.

The seasons “Summer” and “Winter” were chosen to be of the same length of time, six months. No single six-month period uniquely qualifies as “Winter” or “Summer”, but we chose Winter to be defined in the water budget as November 1 through April 30. This aligns approximately with the historical snow storage season. We defined “Summer” as the other six months of the year. This seasonal definition is equivalent to “dry” and “wet” season in the WRB (Chang and Jung, 2010).

Reservoir Modeling in Brief

Reservoir modeling within Willamette Envision duplicates the rules of the US Army Corps of Engineers’ 2012 ResSim model for the Willamette River Basin (WRB). Each reservoir has operational zones, which are based on pool elevation on a particular date (Fig. 1) and rules sets associated with each zone. These rule sets are prioritized within each zone and across reservoirs, such that the model calculates release quantities and locations (e.g., Powerhouse, Regulating Outlet, Spillway) at each time step to meet the highest priority rule. Example rules are those that establish minimum or maximum flows or water elevations, such as the minimum flow targets established by the 2009 Biological Opinion (NMFS, 2008) for April to October, which vary based on the type of water year. These rules are often evaluated at important downstream locations, called control points, which are used to establish releases at reservoirs upstream. Many reservoirs can be influenced by a single control point (e.g., Salem), and many control points can be influenced by a single reservoir. For more information about reservoir modeling in Willamette Envision, refer to the supplemental materials from Jaeger et al. (2016).

Evaluation of Predicted Reservoir Inflows

We compared simulated reservoir inflows to those observed historically to evaluate the reliability of the hydrologic model in replicating the key hydrologic processes of the catchment. The three measures and their interpretation are summarized below.

• Root mean square error (RMSE) is an absolute measure of model fit, meaning that the units are directly interpretable, such that higher values represent higher errors. RMSE is calculated as the square root of the variance of the residual. There is no absolute value for a reliable RMSE, as it must be interpreted with respect to the range of predicted values and practical application.
• Normalized RMSE (NRMSE) is a nondimensional form of the RMSE, normalized to the mean of the observed data, which is useful for comparing between basins and across seasons with different ranges of flows. Expressed as a percentage, the NRMSE is interpreted as the error relative to the mean.
• Nash Sutcliff Efficiency (NSE) measures the signal-to-noise ratio of hydrologic models by comparing the magnitude of model residuals to the variance in the observed data. A value of 1 represents perfect fit between observed and simulated inflows, a value of 0 represents a model that is equally well represented by the mean, and a value less than 0 indicates a model of questionable value. The reliability of NSE is compromised when extreme values are present in the dataset.

Analysis of Reservoir Operational Performance

The 13 federal reservoirs in the WRB are operated with a primary objective of flood regulation, while power production, recreation, fish and wildlife conservation, irrigation, and water quality regulation provide ancillary benefits. In this analysis, we focused on the potential effect of climate change on two high priority operational objectives: flood regulation and fish and wildlife conservation (low flow targets). We evaluated the ability of reservoirs to meet these operating objectives based on the concept of operational reliability, where reliability is defined (e.g., Hashimoto, 1982; Ray et al., 2010) as the likelihood of achieving a flow target or level of flood protection.

Reservoir Economics Methods in Brief

We estimated the value of stored water for recreation at the Willamette Basin federal reservoirs using monthly visitor count data obtained from the US Army Corp of Engineers (USACE) for the years 2001-2011 (period available at time of analysis), and an expected average willingness-to-pay per visitor day (Loomis, 2005). Lowered water levels can impact recreational use in various ways, including loss of boat ramp access and compromised aesthetics such as ‘bathtub rings.' Estimated losses in recreational benefits were based on WW2100 projected shortfalls in summer storage, and empirical evidence that fewer people visit the reservoirs when water levels are reduced (Moore, 2015).

We then estimated the value of reservoir capacity for flood risk reduction based on WW2100 projected land use within the Willamette River floodplain, the probability of flood events, and expected flood inundation. The extent of the floodplain was delineated according to the SLICES data layer developed by the Pacific Northwest Ecosystem Research Consortium (Hulse et al., 2002). We used a ‘bathtub’ model along with high-resolution topographic information (LiDAR) to estimate inundation associated with flood stage. Flood frequency and timing was assessed using the stream gauge record at Salem. Expected flood damages and the related value of reservoir capacity for flood risk reduction were estimated by integrating the spatially explicit estimates of structural value in the Willamette River floodplain, the modeled flood inundation, and assessed flood frequency distributions (Moore, 2015).

Fisheries Methods in Brief

Field Sampling Methods

We divided the mainstem Willamette River into three sections from its mouth, upstream 301 km to the confluence of the Coast and Middle Fork Willamette, based on analysis of river geomorphology. The SLICES framework served as the geomorphic basis of randomized site selection for fish community assessment. Each 1-km slice of the mainstem river or floodplain slough within a river slice represented a single sampling location. In each 1-km slice in 2011-2013, we captured fish with standardized boat and backpack electrofishing.  Sampling locations included 96 mainstem reaches and 71 sloughs. We measured environmental and habitat characteristics at each site.

We developed a Willamette River Fish Database (http://gis.nacse.org/wrfish/) to make spatially explicit data on fish communities in the mainstem Willamette River publicly available. Watershed councils, land trusts, NGOs, and state and federal agencies have full access to the data for designing projects to conserve or restore the aquatic ecosystems and floodplains of the Willamette River. The information also contributes to the collective development of a guiding vision of a future Willamette River and its floodplain for all partners.

Modeling Approach

We related possible temperature increases in the mainstem river to the likelihood of capturing various fish species in a standard sampling effort (e.g., spring Chinook salmon, coastal cutthroat trout, common carp). Specifically, we describe the consequence of a hypothetical increase (2° C or 3.6° F) in stream temperature on the likelihood of capturing representative fish species in our standard sampling protocol, based on the temperatures at which we observed them in 2011-2013. Such an increase in stream temperature is highly likely in the future during normal flow years.

The fisheries analysis is limited to the mainstem Willamette River, downstream of the major reservoirs, and was not coupled to Willamette Envision, the project's integrative water system model.

Survey Methods in Brief

We used property tax records to create a sample of landowners for the survey. We used a geographic information system to stratify the sample in three ways:

• Location in the Willamette Valley,

• Residential versus agricultural property, and

• Inside versus outside of the Urban Growth Boundaries (UGBs).

In winter and spring 2013, we sent questionnaires to 1,600 landowners in each of Lane, Marion, and Washington-Yamhill Counties. From this effort, 1,402 surveys were completed and returned. There were 430 surveys returned as undeliverable and 56 refusals. Therefore, the overall response rate was approximately 32%.

Table 1. Survey respondents by stratum (n=1,402).

A follow-up questionnaire was sent to all landowners from the sample who did not answer the original survey. Approximately 400 questionnaires were completed. The most commonly reported reasons for not completing the survey were lack of interest in completing surveys, perceived lack of knowledge about water management, the survey was too long or complicated, and personal reasons.