Link to CIG Homepage

Hydrology and Water Resources

Theme page links

Link to Integrated Assessment Theme Page Integrated Assessment
Link to Climate Dynamics Theme Page Climate Dynamics
Link to Salmon Theme Page Salmon
Link to Forests Theme Page Forests
Link to Coasts Theme Page Coasts
Link to Human Dimensions Theme Page Human Dimensions

Major Topics:

  1. Introduction
  2. Overview of Hydrologic Simulation Tools
  3. Impacts of Climate Variability on the Columbia River Basin
  4. Effects of Climate Variability on Columbia River Streamflows
  5. Assessment of Climate Impacts to Columbia River Water Resources Using Hydrologic Simulation Models
  6. Application of Streamflow Forecasts to Columbia Basin Hydropower
  7. Applications of Long-Lead Streamflow Forecasts to Instream Flow Management
  8. Impacts of Climate Change
  9. An Overview of Water Management Adaptation Pathways to Cope with Climate Variability and Change


1. Introduction

Research in the Hydrology and Water resources Sector of the Climate Impacts Group has focused on three main topics:

  1. The impacts of interannual and interdecadal climate variability on Pacific Northwest streamflow and water resources

  2. Development of long-lead interannual streamflow forecasting techniques based on long-lead climate forecasts, and development of related applications of these forecasts to water management in the PNW

  3. Assessment of the potential impacts of climate change on PNW water resources, and investigation of some potential long-range adaptation pathways.

Each of these primary topics will be discussed in detail in subsequent sections.

1.1 Areas We Study
Much of the primary research that has been conducted by the CIG since 1995 has focused on the Columbia River basin (Fig.1). There were several reasons for this choice. First, the Columbia River is the predominant regional water resources system in the PNW, providing more than 70 percent of the region's electrical energy, and extensive irrigation of farm land east of the Cascades. The river basin and its forests, fish, and wildlife are also an important economic and cultural entities for people living in the PNW. Secondly, the Columbia is a regional-scale basin and tends to integrate winter climate signals over the PNW in a very useful manner, avoiding the confounding element of local variances that may be present at more smaller spatial scales. As we shall see in the subsequent discussion, the response of Columbia to interannual climate variability has strong connections to ENSO and PDO, a characteristic that has proven very useful for the development of streamflow forecasting techniques.

Although the investigation of Columbia as a whole has proven to be a very useful focus for the CIG's research, there are also important issues concerning urban water supply in the populous area west of the Cascades that are significantly different than issues in the Columbia basin. Starting in 1999 specific investigations of these smaller water resources systems west of the Cascades and their management has been added to the research agenda.

A more detailed investigation of energy and irrigation in Idaho is currently in progress as well.

1.2 Hydrology of PNW Rivers
A basic overview of the hydrology of PNW rivers is very helpful in understanding some of the underlying physical mechanisms that link climate to streamflow. Most of the hydrologically significant precipitation we get in the PNW comes in the winter months from October-March. At other times of the year there is much less precipitation, and what precipitation does fall is predominantly returned back to the atmosphere through evapotranspiration of moisture through plant root and leaf systems. So the winter months are crucial for determining what happens to PNW streamflow in any given water year. Different kinds of rivers, however, respond differently to this winter inflow of moisture to the region. Rivers at low elevation tend to respond quickly and directly to the precipitation that falls on the basin, since the basin temperatures are typically above freezing, and all the precipitation falls as rain. Rivers of this type are called rain-dominated and show a characteristic winter peak flow in their annual hydrograph (Fig.2). Low-lying coastal rivers such as the Chehalis River are rain dominated.

Many rivers on the slopes of the Cascades, like the Cedar River that supplies Seattle's water, are at moderate elevation and have a portion of the basin called the transient snow zone. The transient snow zone is an area in the basin where precipitation frequently falls as snow but then melts a few days or weeks later, a cycle that is typically repeated many times each winter. This transient snow zone can contribute to flooding if heavy rain and warm temperatures occur simultaneously when snow has accumulated (so called "rain on snow" events). Rivers of this type show both a peak in the winter and a peak in the spring and early summer (Fig.2).

Snow melt dominated catchments are generally at higher elevations where temperatures are below freezing for most of the winter. In these basins, the winter precipitation falls predominantly as snow, where it is stored until the spring melt. Rivers of this type show a characteristic low flow period in the winter months, and a large peak flow in spring and early summer as the accumulated snow melts (Fig.2). The Columbia and Snake Rivers are examples of snow melt dominated rivers.

Temperature and precipitation are the primary variables that determine the annual water cycle in the PNW. Precipitation in conjunction with antecedent soil moisture storage largely determines the volume of runoff on an annual basis. Temperature primarily affects the timing of runoff in transient and snow melt dominated basins, and is also associated with secondary changes to the annual runoff volume associated with changes in summer evapotranspiration (direct evaporation from soil or other surfaces, and transpiration of moisture from plants).

1.3 Overview Columbia River Basin and Regional Water Management
The Columbia River basin (Fig.1), on which we will focus much of our subsequent discussion, is one of the largest in North America. It drains much of the land area of the Pacific Northwest (PNW)1 and has an annual flow at its mouth of 7,785 cms (275,000 cfs), second in the United States only to the Missouri-Mississippi River (BPA, 1991). The Columbia River dam system, with more than 250 large reservoirs and more than 100 large hydroelectric projects, is one of the most highly developed in the world and has little room for significant future expansion or development. Despite the large number of dams, available water storage is only about 30% of annual flow (BPA et al. 1991), which has important consequences during periods of extended low flow.

The Columbia River water resources system is managed for electric power generation, flood control, fish migration, fish and wildlife habitat protection, water supply and water-quality maintenance, irrigation, navigation, and recreation by a variety of agencies and public and private utilities. The largest share of water withdrawn from the Columbia is used for agriculture, but there is increasing demand from other human uses, particularly municipal and industrial water supply. Recently, use of water resources for fish and wildlife protection, maintenance of water quality, and recreation have become more important as the region has grown and environmental values have changed (BPA et al., 1995). Specifically, the Pacific Northwest Electric Power Planning and Conservation Act of 1980 called for fish to receive equitable treatment in comparison to hydropower (Wood 1993; Callahan et al. 1999). In addition, the listing of several salmon stocks as endangered or threatened under the Endangered Species Act (ESA) during the 1990s has given anadromous fish stocks high priority for management agencies. The ESA prohibits any federal agency (indeed, any entity) from actions that would jeopardize these species or their critical habitat. Various agreements have established instream flow targets for fisheries protection. The 1988 Vernita Bar Agreement aims to protect fall chinook salmon spawning grounds below Priest Rapids Dam in the Hanford Reach, and the 1995 National Marine Fisheries Service and U.S. Fish and Wildlife Service Biological Opinions (NMFS, 1995) established system-wide flow targets (most significant in summer) for the overall protection of anadromous2 fish in the Columbia and Snake Rivers.

Historically, the management of the major Columbia River storage reservoirs has been dominated by the objectives of flood control and winter hydropower production, which tend to complement each other. To provide protection against spring snowmelt floods, the storage reservoirs must be partially evacuated during the winter, coincident with the period of peak demand for electricity in the PNW3. Irrigation is also an important use of water in the Columbia Basin, with most of the irrigation concentrated in the mid-Columbia area in eastern Washington, in the Yakima River Valley, and in the Snake River basin. Irrigation withdrawals account for about 6% of the annual average flow of the Columbia (BPA 1991), but a much larger fraction of the flow from the Yakima and Snake Rivers, which are tributaries within the larger Columbia basin.

An extensive and relatively centralized institutional infrastructure has developed to manage flood control (coordinated by the U.S. Army Corps of Engineers), hydropower production (coordinated primarily by the Bonneville Power Administration and the Northwest Power Planning Council), and irrigation (coordinated primarily by the U.S. Bureau of Reclamation) in the Columbia Basin. Management for other uses of water in the system tends to be quite decentralized, as may also be said with regard to smaller irrigation districts within the basin. These characteristics have important implications with regard to the ability of the Columbia management institutions to adapt effectively to conditions such as low streamflow increasing human population, changing management priorities, or changes in climate.

In the Columbia's management system, there are two primary planning periods during the operational water year, which runs from August to July. In the "fixed" period from August through December, operations are guided by critical period analysis4 and are essentially unaffected by forecast information (Fig.3). Fixed rule curves5 derived from the critical period analysisare designed to provide adequate flood storage in the fall and early winter; to restrict hydropower operation so as to ensure a high probability of reservoir refill by July; and to prevent early season use of storage that could threaten late-season hydropower production in the case of drought. In the variable period from January to July, reservoir operations are guided both by critical period analysis and forecasts of spring runoff based on measurements of snow pack (Fig.4). These forecasts are used to create rule curves for hydropower and flood control that reflect the conditions in the basin during each water year. During the variable period, reservoir evacuation for flood control operations tends to dominate other considerations, except in conditions of very low flow (BPA, 1991).

Because the Columbia River system is so highly developed, the observed streamflow in the main channel and major tributaries is heavily affected by human management of the system (Fig.5). However, it is possible to account for the effects of changing diversions (for irrigation and municipal and industrial use), storage in reservoirs, and increased evaporation (due to increased reservoir surface area). In the following discussion of the effects of climate on streamflow, the data have had such corrections applied and will be called "naturalized" flow.6



2. Overview of Hydrologic Simulation Tools

A broad overview of the hydrologic simulation tools developed to assess the linkages between climate, hydrology, and water resources is available at the following site {}. The two main components of the system of models developed for the CIG are the hydrologic model [in this case the Variable Infiltration Capacity (VIC) macro-scale model developed at the UW and Princeton (Liang et al., 1994)], and the ColSim (for Columbia Simulation) water management model (Hamlet and Lettenmaier, 1999b). The VIC hydrologic model translates climate information into natural streamflow, and other terms (e.g. snow accumulation, soil moisture, evapotranspiration) in the seasonal water balance of the Columbia River basin. The current Columbia basin VIC model is implemented at a spatial resolution of 1/8 degree latitude/longitude (Fig.6). A daily timestep meteorological driving data set for the model has been prepared from 1948 to 1997, and is used both for streamflow forecasting and climate change investigations described below7. The VIC model has been extensively implemented in large river basins in North America, Europe, and Asia (see, e.g., Abdulla et al. 1996; Nijssen et al. 1997; Lohmann et al. 1998; Matheusen et al. 2000; Nijssen et al. 2000). The Columbia River VIC model does quite well at reproducing naturalized streamflow at The Dalles, OR (Fig.7)

The ColSim reservoir model simulates the effects of dams and reservoirs and the management policies that govern their operation. ColSim is a monthly time step reservoir model that incorporates the major projects (dams and reservoirs) and operational features of the Columbia basin, and was constructed as a research tool for the experiments described here (see Hamlet and Lettenmaier 1999b). The model's domain ranges from Mica Dam in British Columbia, near the headwaters of the Columbia, to Bonneville Dam, near the mouth of the river, and includes many of the major tributaries: the Kootenai, Pend Oreille, Clark Fork, and Snake River systems (Fig.8). While inflows from smaller tributaries are included, dams on these streams are not simulated. The input to ColSim is monthly average naturalized streamflow, derived either from observations (Crook,1993) or from output from the VIC hydrologic model, which may in turn be driven by present climate simulations or future scenarios. ColSim can thus be used to explore the reliability of achieving various system objectives under current patterns of climate variability or hypothetical conditions, such as the climate of the 2040s (see Section 8). The primary outputs of the ColSim model are regulated streamflow throughout the basin, reservoir storage volumes and elevations, energy production from the dams, and the reliability8 of meeting the objectives of the water resources system described in Table 1. More details on the ColSim model can be found in Hamlet and Lettenmaier (1999b) and Miles et al. (2000).

The use of these two simulation models is also discussed in the context of specific research projects in various sections below.



3. Impacts of Climate Variability on the Columbia River Basin

As mentioned earlier, climate affects hydrology and water resources primarily through variation in temperature and precipitation. Winter climate conditions strongly influence PNW hydrology via the changes in snowpack that largely determine the subsequent spring and summer streamflows (Hamlet and Lettenmaier 1999a); cool/wet winters produce a large mountain snowpack and abundant streamflow, while warm/dry winters produce a small snowpack and below average streamflow (dell'Arciprete et al. 1996). This is the major pattern of variability in PNW climate, and is strongly associated with the El Niņo/Southern Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO) (Mantua et al., 1997). We will use ENSO and PDO to categorize climate variability because this offers direct linkages to existing seasonal climate forecasts that are useful for streamflow forecasting. This in turn has value in the context of PNW water resources management (see Hamlet and Lettenmaier 1999a and Hamlet et al. 2000a). Because of its persistence over long time scales (several decades), the PDO also offers a useful benchmark in the PNW for comparing the relative impacts of future human caused climate changes to observed variability over the past 100 years or so.




4. Effects of Climate Variability on Columbia River Streamflows

A long time series of naturalized streamflow data for the Columbia River at The Dalles, OR (the most downstream point in the river that we will examine) is available from 1879-1999, with extended records of peak stage going back to 1858. For the period from about 1900 to the present, estimates of historical ocean conditions allow us to retrospectively categorize each water year into climate categories based on retrospective observations of winter PDO and ENSO state.

To illustrate these connections between climate and streamflow, we examine variations in regional streamflow associated with these retrospective PDO and ENSO categories. Naturalized Columbia river streamflow at The Dalles tends to be higher during the cool phase of ENSO than during the warm phase, and the largest differences occur during the peak flow months, when snow pack contributes most significantly to streamflow. Yearly total flow is about 20% higher on average in cool ENSO, compared with warm ENSO years (Fig.9). The effect of the PDO on streamflow in the Columbia is similar in magnitude and timing to the effect of ENSO (Fig.10), but has a completely different timescale, tending to persist mostly in one phase for several decades (Fig.11). Note also the positive reinforcement between the effects of PDO and ENSO when they are in phase (i.e. warm PDO/warm ENSO and cool PDO/cool ENSO).

Snow melt and transient river basins across the PNW exhibit similar streamflow responses to ENSO and PDO, due to the tendency for winter precipitation and temperature in the PNW to be regionally cohesive. For example, the Cedar River in Washington, a small transient snow watershed, shows the same characteristic response to PDO and ENSO as does the Snake River in Idaho, a larger snow melt dominated river, and the Columbia at The Dalles (Hamlet and Lettenmaier, 2000a). The effect on streamflow associated with different combinations of PDO and ENSO is non-linear (Fig.12). The two phenomena strongly reinforce each other when in phase, i.e., warm PDO/warm ENSO and cool PDO/cool ENSO. Average seasonal flows under cool PDO/ENSO neutral conditions are similar to those under cool PDO/cool ENSO conditions (especially during spring peak flows); all other combinations of ENSO and PDO have composite mean hydrographs that are similar to the long-term mean on average, and the variability in each of these categories is also similar. We exploit these characteristic patterns of variability in making streamflow forecasts for the Columbia based on long-lead climate forecasts.

High flows9 are more likely when both ENSO and PDO are in the cool phase than in any other combination of cool PDO and ENSO, and have never occurred in the historic record (1900-1999) during the warm phase of the PDO. Low flows10 are much more likely when PDO or ENSO is in the warm phase, but also infrequently occur under the conditions of cool PDO/warm ENSO (water years 1924 and 1973) (Hamlet and Lettenmaier 1999a; Miles et al. 2000). Not only do ENSO and PDO influence the probability of high and low flow events, they also affect the likelihood of extremes; at least four11 of the top five highest-flow years in the Columbia basin occurred when the PDO was in its cool phase; three of these occurred when ENSO was also in its cool phase. Likewise, all five lowest-flow years occurred when the PDO was in its warm phase; in four of those years ENSO was also in its warm phase.

4.1 Floods
Most regulated rivers in the PNW are managed - at least in part - to prevent flooding. For smaller, unregulated watersheds, however, flooding may be a concern. In this section, we evaluate the connection between PNW climate variability and flooding in unregulated rivers.

Flooding may be broadly defined as river flow that is sufficiently large to overflow the normal river channel and inundate surrounding land. This generally occurs when the river's flow exceeds the channel capacity, or bank-full flow. For most PNW streams, bank-full flow corresponds approximately to the mean annual flood12 (Dunne and Leopold, 1978). The atmospheric conditions that lead to flooding differ for snowmelt dominant, transient, and rain-dominant river basins. In snowmelt-dominated rivers, flooding usually occurs in spring and is often caused by rapid warming, accompanied by intense rain falling on snow-covered ground. Flooding in rain-dominated basins usually occurs in fall or winter and is predominantly caused by extreme precipitation events in which intense rain falls on saturated soil. Transient basins can experience either spring flooding from rapid snowmelt and heavy rain, or fall or winter flooding from heavy rainfall, with or without a snow melt component. In transient basins on the west slope of the Cascades, most floods occur due to the latter conditions, primarily in November and December.

Other factors that influence the likelihood of flooding include basin topography, channel characteristics (e.g. channel aggregation), and the extent of human development of the flood plain. In parts of some river channels, low banks and adjacent low-lying areas lead to flooding for flows that are only moderately high, while river channels with high, steep sides experience flooding only with the most extreme flows. Low-lying areas may be protected by levies or dikes, or the channel configuration may have been intentionally altered to increase the flow rate.

In order to determine the influence of climate variability on flooding in the PNW, we examined long (57-65 years) streamflow records from 26 uncontrolled basins throughout the region (Fig.13). These basins were chosen to relect a range of hydrologic types and regional topographical and geographical features. Most of the station records examined were from mixed rain and snow basins on the western slopes of the Cascades and snowmelt-dominated rivers east of the Cascades; a few are from coastal, rain dominated basins. For each basin, we calculated the observed probability of flooding for each of the six climate categories defined by the phases of ENSO and PDO. The probability of flooding was defined as the probability of observed daily streamflow exceeding the mean annual flood at least once during the year. Physical anomalies, such as those described above that alter the relationship between natural flow and flood risk, were ignored in our analysis.For transient and snow melt dominated rivers, there is a statistically significant difference in the mean conditional probability of flooding for the climate categories in which ENSO and PDO are in phase at the 99 % confidence level (Fig.14 far left and far right). For transient basins the difference in the mean probability in these two opposing climate categories is 0.23 with 99 percent confidence (Fig.15), whereas for snow melt dominated basins the difference in the mean likelihood of flooding is 0.39 with 99 percent confidence (Fig.16). Smaller differences are observed for comparisons between other climate categories at the same level of confidence, and for some there is no statistically significant separation in the mean. For the snowmelt dominated basins, however, the mean probabilities for the warm PDO/ warm ENSO category are lower than all other categories with 99 percent confidence. It should be noted that these comparisons are for the means of all stations of a particular hydrologic type, and the general tendency does not necessarily hold for each individual basin examined. The rain-dominated basins seem to show less sensitivity to climate variation, although the results are probably not statistically significant because of the small sample size (n=3). Transient snow basins appear to show sensitivity to ENSO but less to PDO. Snow-dominated basins are most sensitive to both ENSO and PDO, with the greatest separation between categories where ENSO and PDO are in phase.

Differences in the basin characteristics and the various causes of floods discussed above help explain the influence of climate variations on the likelihood of flooding. Whereas flooding in rain dominated basins results from individual storms, flooding in snow melt basins depends on both the weather over the entire winter and spring (i.e., the period of mountain snow accumulation), as well as on individual storms. Since wintertime climate is more sensitive to the state of PDO or ENSO than is a single intense precipitation event, the likelihood of flooding in snowmelt-dominated basins is more strongly associated with PDO or ENSO than in rain or transient basins.

Although the probability of flooding is associated with ENSO and/or PDO for most basins, the average severity of flooding (the amount of daily streamflow above the flood threshold, not shown) does not. It is likely, therefore, that flood severity is determined more by individual weather events than by average regional and seasonal climate characteristics. This result has important implications for seasonal forecasting of floods: it suggests that the probability of floods occurring during a given season can be estimated with long lead times based on ENSO forecasts and persistence of the PDO. Such forecasts could provide managers concerned about flooding with an estimate of the odds of flooding in a given year - in the summer before the flood season13. However, this approach could not be used to predict either the timing or severity of these events; such details can be predicted at most a week in advance, a time scale that is constrained by the theoretical limits of numerical weather prediction.

4.2 Droughts
The effects of droughts on water resources are specific to particular regions and river basins, and are a complex function of climate, hydrologic response, physical characteristics of the existing dams and reservoirs (e.g., amount of storage available), uses of water in the basin, and the reservoir operating policies in use at any given time. Because of these complexities, it is unlikely that any single definition of drought will be acceptable for every situation, or remain stationary in time. To explore the occurrence of drought in the Columbia River Basin, and its connection with patterns of PNW climate variability, we examine the historical streamflow record for periods of low streamflow that would likely have socioeconomic impacts if they occurred today, given today's dam and reservoir system and reservoir operating procedures.

First, the ColSim reservoir model was used to identify periods within the 1931-1988 streamflow record during which system storage would have been significantly depleted. To extend the analysis to years when the detailed streamflow data needed to run the reservoir model were not available (i.e., before 1931 and after 1988), we developed a quantitative definition of drought based on the identified drought periods from 1931-1988. Droughts were defined using a threshold of 0.9 standard deviations below the long-term (1900-1997) mean of naturalized streamflow at The Dalles, using monthly-averaged data. Droughts are those periods for which the streamflow was below this threshold for at least six months and did not exceed the threshold for more than three months out of 12. This definition is somewhat subjective, and is dependent on the simulated uses of the reservoir system and its current operating policies, implying that these drought periods may not have been perceived as droughts at the time. Nonetheless, the formula is useful for defining a group of low-flow sequences that may be considered the most severe multi-season droughts in the Columbia Basin. Using this definition, the identified drought periods from 1900-1997 are:

  • Feb 1905-Jun 1906 (17 months)
  • Dec 1928-Feb 1932 (39 months)
  • Oct 1935-Aug 1937 (23 months)
  • Jan 1944-Aug 1945 (20 months)
  • Jun 1987-Sept 1988 (16 months)
  • May 1992-Oct 1995 (42 months)

Until recently, a 42-month period from 1928-1932 was the critical period used for energy planning in the Columbia. Recent changes in the reservoir operating system to provide more spring and summer streamflow for salmon (see Miles et al. 2000 for more details) have moved the critical period to a 9-month streamflow sequence in the period from 1936-1937, confirming that the identified droughts are among the most severe in the historic record. Severe multi-season drought sequences such as these typically include several winter low-flow months during which reservoir storage is depleted in order to generate energy, followed by summer low-flow conditions that are inadequate to provide reservoir refill. Reservoir storage is further depleted in the following winter in order to meet energy requirements, especially if this second winter is also very dry. These droughts are therefore primarily caused by abnormally low winter precipitation, which is strongly influenced by PDO and ENSO. Five out of six events listed above occurred in warm PDO epochs, four out of six events contained multiple warm ENSO events, and three out of six contained back-to-back warm ENSO events. In addition, the set of years when both ENSO and PDO were in the warm phase contains several years with very low annual flows, including water year 1977, which had the lowest flow on record for the Columbia Basin. However, the impacts that would be experienced today in this extreme single year drought are not as severe as those that would be experienced during the multi-year droughts identified above, since reservoir storage was sufficient to cope with the severe but short-term reduction in streamflow.

A complex tangle of international, federal, regional, state, tribal, and local entities have competing jurisdictions over a variety of managerial aspects of the Columbia River system. Different kinds of climate variations pose different kinds of stresses on the system, and the institutional responses to stress are different for different aspects of the system. As discussed above, the Army Corps of Engineers has clear authority to coordinate operations when floods threaten. Droughts, however, expose the conflicts among various entities that assert competing claims to water. These conflicts over limited water supplies have been exacerbated by recent listings of several salmon species in the Columbia basin under the Endangered Species Act, which have elevated the priority of instream flow with respect to other uses without increasing the aggregate supply of water (Callahan et al., 1999). Without a primary basin-wide decision-maker, the system remains disjointed and managers are further constrained from action by the complicated decision-making process, legal agreements, and unresolved conflicts among water users (Miles et al., 2000). These characteristics of the Columbia basin management system also have important implications in the context of adaptation to climate change, as will be discussed below.



5. Assessment of Climate Impacts to Columbia River Water Resources Using Hydrologic Simulation Models

Because the Columbia River is so heavily managed, patterns of natural river flow associated with variations in climate cannot be used directly to assess the implications to the managed system. In order to quantify the impacts of climate variability and change on human uses of water in the Columbia basin, we use a ColSim reservoir simulation model described above. Using the ColSim model driven by naturalized flows (with irrigation diversions removed) for the period 1931-1989, we examine how system reliability depends on the phases of PDO and ENSO, given the current operating policies in the Columbia basin (Fig.17) (Miles et al. 2000). Firm energy production is essentially isolated from climate variability, while other uses that depend on summer streamflow (e.g., instream flow, irrigation, and recreation) typically have declining reliability in dry conditions (e.g., warm phases of PDO and ENSO) and increasing reliability in wet conditions (e.g., cold phases of PDO and ENSO). As a result of its longer persistence (25-30 years), PDO typically has larger impacts on water resources objectives than does ENSO. Flood control and desirable navigation conditions (not shown) both tend to be more reliable in dry conditions and less reliable in wet conditions. These changes in reliability are relatively small, however, showing that these particular objectives are largely insulated from climate variability. It should be noted that the reported reliabilities are based upon the streamflow variability observed over about 60 years of the historic record, which may not be fully representative of the true variability of past - or future - streamflow.

Given the current structure and objectives of the Columbia River water resources system, only one of the two competing objectives of firm energy production and instream flow protection can be fully protected from climate variability. To demonstrate this, the ColSim model was used to evaluate a hypothetical alternative reservoir operating policy in which all major system storage (including Canadian treaty storage) was used to meet the instream flow targets set for McNary (main stem) and Lower Granite (Snake River) dams. Rule curves and system objectives for all other water uses, including firm energy production and flood control, were unchanged. While the mean reliability (for 1931-1989) of meeting the McNary flow target increased from 86 percent, for the "status quo" operating system discussed above, to 99 percent for the alternate operating system, this improvement was accompanied by corresponding reductions in the reliability of current levels of hydropower production and of other uses in the system (Fig.18)(Miles et al. 2000). Because of the conflicting timing for hydropower production and major instream flow targets, it is not possible to insulate both uses from climate variability within the current framework of water resource objectives.

To highlight further the dependence of system objectives on climate (given current operating system design), we performed a simple threshold analysis to determine how far annual average streamflow has to deviate from the long-term (1900-1997) mean for the reliability of meeting the objective to drop to 85% or less. The standard deviation of mean flow is used to describe streamflow anomalies; one standard deviation (1s) is about 18% of annual flow.

Firm energy production was 100% reliable over the 1931-1989 period almost by definition, since firm energy targets for the hydropower system are determined based on the most adverse low flow sequence in the historic record. Fish-flow targets at Priest Rapids and Columbia Falls were also 100% reliable in the model simulations because sufficient storage exists upstream of these locations to ensure flow even under severe drought conditions. Other uses impacted by low-flow conditions showed higher sensitivity. The flow thresholds corresponding to 85% reliability (arranged in order of increasing sensitivity) were -1.5s for non-firm energy, -0.25s for Lake Roosevelt Recreation, and only -0.1s for middle Snake River irrigation and the McNary fish-flow target. In other words, even under conditions of relatively normal streamflow, these last three objectives have a significant chance of not being met.

For uses negatively affected by high flow conditions, desirable navigation conditions (threshold at 0.7s) is of lower priority (that is, is impacted more strongly by the same level of flow anomaly) than flood control at The Dalles (threshold at 2s). The fact that flooding is 85% reliable at the 2s threshold, a level of flow rarely observed, shows that the Columbia River reservoir operating policies strongly protect against flooding. The approximate operational priority of the various objectives within the Columbia River water resources system, based on the threshold analysis, is shown in Figure 18b. Further details of this analysis are reported by Miles et al. (2000).

Under the existing operating policies, the performance of the Columbia River water resources management system does not fully reflect official policy regarding the priority of different uses. Specifically, although hydropower production and fisheries protection are of equal priority under federal law (the Northwest Power Planning Act of 1980), it is apparent that equality of the two management objectives has not been achieved in the context of instream flow, despite recent changes in the operating system designed to provide greater spring and summer flows for salmon. This is because the portion of available system storage allocated to energy production remains much larger than the portion allocated to maintain instream flow targets for fisheries protection. In periods of below average flow, the storage allocated for meeting instream flows is quickly exhausted, and streamflow targets are not met in late summer (Miles et al., 2000). Despite the clear disparity in priority between fish flows and hydropower production in the Columbia Basin, however, it should be noted that other efforts undertaken to mitigate impacts to salmon have affected hydropower production and hydropower revenues significantly (Miles et al., 2000).

5.1 Long-Lead Streamflow Forecasting
Operational streamflow forecasts for the Columbia River basin are currently available starting in mid winter (January 1) and are based on statistical relationships between observations of descriptive variables (e.g., snowpack) and summer runoff volumes. The primary limitation of these methods is that their earliest useful skill for water planning is provided in January (Lettenmaier and Garen, 1979). Recent advances in experimental streamflow forecasting techniques, based on long lead climate forecasts, which we will describe here, provide opportunities to extend the lead time of streamflow forecasts by roughly six months, providing useful forecast skill before any winter snowpack measurements are available (i.e. in June preceding the water year). In this section (condensed from Hamlet and Lettenmaier, 1999a), we describe the new forecast technique. In the next section we examine the potential utility and economic value of this type of long-lead streamflow forecast for the production of non-firm hydropower in the Columbia River basin (Hamlet and Lettenmaier, 2000b).

As described above, PNW winter climate and the subsequent summer streamflow from snowmelt are linked to the Pacific Decadal Oscillation (PDO) and the El Niņo/Southern Oscillation (ENSO). Forecasts of these recurrent patterns of climate variability can be exploited to provide summer streamflow forecasts for the Columbia River basin with lead times of about 12 months (Hamlet and Lettenmaier 1999a). These methods are summarized below and are discussed in more detail at

The forecasting procedure begins by assigning the upcoming water year to one of six climate categories, based on forecasted ENSO conditions (El Niņo, La Niņa, or ENSO neutral) and assumed PDO conditions (cool or warm PDO, generally based on assumed PDO persistence). An ensemble of streamflow forecasts is developed using the VIC hydrology model, estimated June-September streamflows that are used to initialize the hydrologic model, and meteorological data for the subsequent months (October-September) taken from water years of the climate category as that forecast for the coming water year using data from 1948-1995. These types of probabilistic streamflow forecasts therefore provide information about future streamflows within the context of observed variability over the past 45 years or so. The forecast for water year 2001, for example, shows a high likelihood of average flows, with a low likelihood of either strongly above or below average flows. This ensemble streamflow forecast for the Columbia River at The Dalles was created using meteorological data selected from cool PDO/ENSO neutral years in the historic record, and was made available on an experimental basis June 9 2000. The assignment of the climate category of cool PDO/ENSO neutral to water year 2001 was based on the 2001 ENSO forecast for neutral ENSO conditions and an assumption of cool phase PDO, based on heuristic methods of identifying PDO transitions (Hamlet and Lettenmaier, 1999a).

Figure 19 shows a timeline for these types of streamflow forecasts. An ENSO forecast becomes available in June, a single meteorological time series is used to initialize the hydrologic model for the period from June-September, and then multiple realizations (ensemble "members") are produced from October-September using the driving data extracted from water years in a particular climate category as discussed above to develop simulations of streamflow for the coming water year. Hamlet and Lettenmaier (1999a) discuss a number of potential uses of these kinds of forecasts, among them potential improvements to fall non-firm hydropower production, which we will quantify in the following section.




6. Application of Streamflow Forecasts to Columbia Basin Hydropower

The strong seasonal differences in river flow in the snowmelt dominated Columbia River present challenges to the hydropower industry. Snowmelt occurs predominantly between April and July, resulting in high streamflows in the spring and summer. This timing is unfortunate for hydroelectric generation, because the spring period experiences the lowest seasonal power demand in the PNW. Storage reservoirs help to alleviate this difficulty by shifting some flow from spring to fall and winter, but storage on the Columbia accounts for only 30% of average annual flow, which is relatively low in comparison with other large rivers in the U.S.14 In addition to producing hydropower, the Columbia water system must simultaneously provide flood control, water supply for irrigation, lake and river recreation opportunities, navigation, and flow enhancement for the protection of riverine ecosystems, as described above. Addressing these multiple objectives prevents operation of the power system to maximize hydroelectric potential. Flow enhancement for the protection of salmon, for example, requires significant releases from storage to maintain more natural instream flows during the late summer and early fall, times when hydroelectric producers might otherwise retain reservoir storage for winter energy production.

As discussed in previous sections the reservoir operating system for the Columbia River Basin has evolved to make use of streamflow forecasts that are based on observed snowpack and its statistical relationship to spring and summer streamflow. These forecasts become available starting January 1 and guide water management decisions from January-July. In the period from August-December, however, reservoir operations are managed without any forecast information. Our discussion will focus here on the Energy Content Curve (ECC), the primary reservoir rule curve15 guiding hydropower production. In simple terms, the ECC restricts the use of reservoir storage for hydropower production according to the following rules: (1) The reservoir operators are not permitted to draft below the ECC for firm energy production unless all reservoirs do so in a balanced manner; and (2) operators are not permitted to draw reservoirs down below the ECC to generate non-firm energy (BPA et al., 1991). To provide background needed for the subsequent discussion, we now describe in more detail how the ECC is constructed.

The Energy Content Curve is a composite rule curve constructed using three reservoir rule curves: the flood evacuation curve, the "assured refill" curve, and the "critical" curve, each of which is a set of guidelines for reservoir operations designed to ensure that certain objectives are met. The flood evacuation curve is designed to ensure that sufficient reservoir storage space is available for catching and retaining the high flows of April-August to prevent (or protect against) flooding. Flood evacuation requirements are fixed from August-December, and are based on streamflow forecasts in January-July. Figure 20 shows a flood storage evacuation diagram (for Libby Dam) that is typical for the major federal storage projects in the basin; note that the family of possible flood evacuation curves splits in January when the forecasts become available. The assured refill curve tends to guide operations in August-December and is designed to ensure a high likelihood of reservoir refill in the following summer. It is constructed using the third lowest streamflow sequence on record. The critical curve is designed to protect the system from fall and early winter overdraft during droughts. This curve is constructed by simulating the amount of reservoir drawdown that would occur if firm energy requirements were satisfied during the most adverse streamflow sequence in the historic record (the "critical period", currently a portion of 1936-37).

From August-December, and during reservoir refill in May and June, the ECC is generally determined by the higher (i.e. more restrictive of use of storage) of the assured refill curve and the critical curve. In the period from January-April, the ECC is generally determined by the flood evacuation requirements. Figure 21 shows the construction of the ECC from the three primary rule curves for a wet year with significant winter flood evacuation requirements.

The way the ECC is determined in the fall is very cautious, because an assumption of drought conditions for the following summer is built into to the way the assured refill and critical curves are constructed. It is this aspect of the Columbia's operating plan that will be altered, in the modeling study described below, to provide an input pathway for long-lead streamflow forecasts.

The improvements in Columbia River basin streamflow forecasting described above provide opportunities to improve the decision processes associated with the marketing of non-firm energy from the Columbia River hydroelectric system in the fall and early winter. The streamflow (resource) forecasts are used here as input to a complex decision-making process, which in this case is the reservoir operating plan for the Columbia basin.

As described above, the ECC rule curve currently restricts fall hydropower production as though the reservoir system were experiencing a critical drought. With improved long-lead streamflow forecasts, however, these restrictions can sometimes be relaxed in an objective manner, permitting increased non-firm generation in years that are likely to be wet. This energy could be generated during the fall (August-December), a time when prices are typically higher than during the spring and early summer period when non-firm energy has traditionally been marketed. As will be shown, it is possible to relax these constraints on production of non-firm energy without affecting other uses of the system, while simultaneously increasing long-term revenue from spot market energy sales. These revenue increases are therefore directly derived from use of the new forecast information, rather than from altered tradeoffs between different system uses.

To relax these fall and early winter constraints in a consistent and systematic manner, a new method of constructing the Energy Content Curve is defined, using the long-lead ensemble streamflow forecasts discussed above. A flood evacuation requirement based on the lowest simulated streamflow in the forecast may be taken as an estimate of the least flood evacuation expected for that forecast. A new rule curve that ensures the reservoir will refill to this least flood evacuation requirement given the lowest ensemble streamflow sequence is then constructed. This new rule curve, called the Refill to Least Flood Curve (RLFC) (Fig.22) ConstructionRLFC) then replaces the status quo ECC for August-December. The goal here is to provide more water for energy production when streamflow is likely to be high in the subsequent summer, and less when conditions are dry, while simultaneously ensuring a high likelihood of refill to actual flood evacuation targets in spring. This is achieved by the changes in the ECC for a wet year (Fig.23) (1972 cool PDO/La Niņa) and a dry year (Fig.24) (1987 warm PDO/El Niņo) at Libby Dam. More water is made available for energy production in the wet year compared to the status quo, and less storage is made available in the dry year than in the status quo.

To simulate the economic effects of these changes, retrospective streamflow forecasts for water years 1931-1987 were used to construct a new ECC for each water year, and the monthly time step ColSim reservoir model was used to estimate energy production and revenue for the new versus status quo ECCs over this time period. (The ColSim model simulates energy production from the major storage and run of river dams in the Columbia basin, which accounts for about 55% of the total basin energy production.) Several alternative energy targets for fall and early winter were used in these simulations (Fig.25); alternative 1 is associated with the least aggressive spot marketing in fall, and alternative 5 with the most aggressive marketing targets. These alternative marketing strategies essentially trade reliability of non-firm energy production in spring for long-term economic benefits.

In addition, the monthly non-firm energy targets for August-January were scaled based on the forecasted climate categories. The scaling factors were constructed so that the overall likelihood of meeting non-firm energy targets was between 90% and 95% for each climate category. The rationale here is that greater or lesser available water in August-December associated with changes in the ECC should be accompanied by a corresponding increase or decrease in energy marketing targets.

The energy targets for February-July were always fixed as in the status quo in the simulations, since energy production capability in this time period is not determined by the new forecasts. All other settings in the model were unchanged from the status quo, thus focusing the analysis on the value of forecasts for fall and early winter non-firm energy revenue. In particular, firm energy production targets, which are important because of energy capacity considerations, remain unchanged, and are met with 100% reliability in both the status quo and all alternative formulations.

6.1 Evaluation of Economic Benefits
In each retrospective water year of the ColSim model simulation, the ECC and energy scaling factors change (based on the climate and streamflow forecasts), and the model attempts to meet the resultant non-firm energy targets without drafting the storage reservoirs below the ECC, while simultaneously attempting to meet all other system objectives. The revenue generated is based on estimated monthly average spot market prices (BPA, forecasts of monthly average prices for 2002-2006 in 1998 dollars) (Fig.26), which are assumed to remain constant for the entire simulation period.These prices are also assumed to be unaffected by changing non-firm marketing practices in the Columbia, a reasonable assumption since the simulated changes are on the order of 0.5 percent of the total load for the western power grid upon which prices depend. Hydroelectric energy production is also assumed to have associated costs of $4.0 per MW-hr (personal communication, John Fazio, Northwest Power Planning Council). Using long-lead streamflow forecasts to guide the construction of the ECC would result in long-term average increases in spot market revenue on the order of $40 million per year for the least aggressive fall marketing strategy (alt 1) and approximately $150 million per year for the most aggressive fall marketing strategy (alt 5) (Fig.27). These simulated increases are achieved by moving non-firm generation from the spring to the fall months when energy generation is more valuable, and also by reducing non-power producing spill from reservoirs during the spring in wet years. The reliability of the major Columbia River water resources system objectives under these different modeled alternatives are shown in Figure 28. For alternative 1, the performance of other system objectives is almost identical to the status quo, while under alternative 5, some reductions in system storage have minor negative impacts on Lake Roosevelt recreation conditions and the reliability of the McNary fish flow target (for salmon protection). The reliability of non-firm energy production declines as the energy is more aggressively marketed in fall (compare alternative 1 to 5), because the status quo spring non-firm energy targets (applied uniformly to all years in the simulation as described above) cannot be met under all conditions in the simulation when more energy production occurs in the fall. These results demonstrate that the increases in revenues reported are almost completely associated with the systematic use of new information and the revised reservoir operating plan, as opposed to resulting from altered tradeoffs between different system objectives. The results probably significantly underestimate the actual potential economic benefits associated with shorter time scale spot marketing strategies, since monthly average prices were used in this analysis, and actual marketing decision processes function on much shorter time scales.



7. Applications of Long-Lead Streamflow Forecasts to Instream Flow Management

Another potential application of these kinds of long-lead streamflow forecasts involves decision processes regarding releases from reservoir storage to enhance river flows to aid seasonal anadramous fish migration (either passage of smolts, or returning adult spawners). In irrigation water supply systems east of the Cascades (e.g. Yakima River basin) and urban water supply systems west of the Cascades (e.g. the Cedar River basin), there is frequently remaining storage at the end of the summer (the period of peak demand) that could be used to enhance instream flow in the early fall. Water managers, however, are unwilling to release extra water from storage at this time, because there is currently no operational forecast of the upcoming spring inflows at this time, and if a drought were to occur, water supply reliability could be compromised. These spring streamflow forecasts become available in January, but this is too late to benefit the fish.

In a manner analogous to the non-firm hydropower application above, an objective decision process can be developed to determine a "safe" level of fall release based on an ensemble streamflow forecast. In years likely to be wet, extra releases of storage for fish would probably result in no impacts to water supply reliability, whereas in years likely to be dry a very cautious approach might be most appropriate, preserving water for subsequent spring fish flows and summer water supply.

Hansen and Palmer (1999) have conducted a study based on these ideas for the Cedar River basin, and concluded that instream flows could be increased overall without impacting water supply reliability by increasing instream flows within water years falling in certain forecast climate categories (e.g. cool PDO/La Niña).




8. Impacts of Climate Change

In this section we examine the effects of several climate change scenarios on the hydrology and water resources of the PNW. The analysis is divided into three topics:

  1. The potential effects of climate change on the hydrologic response of the Columbia River
  2. The impacts of these hydrologic changes on the current water resources system
  3. The ability of current management systems to adapt to climate change

8.1 Quantifying the hydrologic implications of climate change
To evaluate the potential impacts of climate change on the water resources of the PNW, we use a chain of numerical simulation models for the Columbia River basin representing, in turn, the changes to PNW climate, the hydrologic response of the river, and the implications for the water resources system.

Several issues associated with state-of-the-art global climate models (GCMs) must be resolved in order to implement such a modeling structure. The horizontal resolution of GCMs is still insufficient to resolve mountain ranges whose horizontal extent is smaller than the Rockies. For this reason, climate model output in its raw form is not suitable for evaluating hydrological changes at the regional scale in the PNW. Important features may be missing entirely, like the difference in climate between the west and east sides of the Cascades. To translate climate model output to the regional scale, a number of approaches are possible. The approach used by Hamlet and Lettenmaier (1999b), is an empirical method that applies projected monthly mean changes in regional temperature and precipitation from GCM simulations to observed regional climate data.

In this method, the GCM output (with a finest resolution of 2.8 degrees (latitude/longtitude), see Chapter 4) was translated to the finer grid of the regional hydrology model (1/8 degree latitude/longitude) as follows. First, GCM data is regridded to 1-degree resolution over the Columbia Basin and the average temperature change (DT, calculated as a difference from the long-term climate model control runs) and average precipitation change (Dp, calculated as a ratio) for each month are calculated over this approximate geographic area. These changes are then applied uniformly to gridded temperature and precipitation records based on observations (1961-1997), which are then used to drive the VIC hydrology model for the Columbia Basin. The simulated streamflows from the hydrology model are used to drive the ColSim reservoir model for a simulated base case (current climate) and a group of climate change scenarios in order to evaluate climate change impacts on water resource objectives. In this way we obtain projected changes in streamflow and reliabilities of water resource objectives that are consistent with both the spatial and temporal variability of observed PNW climate and the changes in monthly means projected by the global climate models (Fig.29). For further details regarding these methods, see Hamlet and Lettenmaier (1999b).

At the time of this writing climate change simulations from eight GCMs were available. To limit the amount of computational time in making hydrologic simulations, the four GCMs with the highest spatial resolution and the most sophisticated land surface schemes were selected: the ECHAM4, HadCM2, HadCM3, and PCM3 models. Characteristics of these models and the simulated temperature and precipitation changes over the Columbia basin associated with increasing concentrations of CO2 are shown in Figure 30 and Figure 31. Two climate change scenarios were constructed from each model, i.e., projected changes in regional temperature and precipitation over the Columbia Basin for the decades of the 2020s and the 2040s (see Nijssen (2000) for more details).

The hydrologic effects to the Columbia basin due to changes in temperature and precipitation are dominated by changes in snow accumulation and melt. Changes in snow extent within the Columbia Basin that would result from the climate changes projected by the ECHAM4 model are shown in Figure 32. With warming, low-elevation areas lose their snow in the hydrologic simulations, and the most obvious changes in area covered by snow are in the lower part of the Columbia basin. Note that the upper part of the basin (at relatively high elevation in the Canadian Rockies) remains covered with snow in June even for much warmer climates. Results from the HadCM2 (shown in Hamlet and Lettenmaier (1999b)), HadCM3, and PCM3 models (not shown) are similar in character.

Figure 33 and Figure 34 show changes to long-term monthly average natural streamflow for the 2020s and 2040s at four different locations in the Columbia basin for the ECHAM4 and HadCM2 models only: The Dalles, the most downstream river location simulated; Corra Linn, Chief Joseph, and Ice Harbor, dams within three sub-basins at different elevations, all of which are snow-dominated in today's climate. Note the significant differences between different locations in the basin. Although the results are presented as averages, keep in mind that each time period discussed - base case (current climate), 2020s, and the 2040s - was constructed using a thirty-six year climate record.

At The Dalles, the flow in the Columbia River is strongly snow melt dominated, and the predicted changes in snow accumulation and melt are primarily responsible for the simulated changes in streamflow under climate change. The climate model scenarios for the 2040s lead to increased winter streamflow, decreased summer streamflow, and a timing shift in the period of peak flow of two weeks to one month earlier in the year. Increased winter flow results from a combination of increased precipitation and higher temperatures that raise the snow level so that more precipitation falls as rain over the basin and is not stored as snow. The summer decreases occur largely because of higher temperatures, which decrease total spring snowpack, increase evapotranspiration16, and cause snow to melt earlier. Flows at Corra Linn and Chief Joseph, which are at high altitude and have cold winter temperatures over much of the contributing basin area even under climate change, are much less affected by warmer temperatures than are the flows at Ice Harbor, whose contributing basin is at lower elevation and is warmer in the winter. These differing sensitivities are a general characteristic of the response of transient and snow dominant river basins to warming. These results suggest that river basins whose present-day winter temperatures are close to freezing level (e.g., most of the transient snow water supply basins on the west side of the Cascades) are the most vulnerable to climate change due to the dramatic shifts in streamflow timing that would result from relatively small increases in wintertime temperatures.

The net effect of seasonal changes in temperature and precipitation for the 2020s ranges from a reduction in annual flow volume of about 3% for the ECHAM4 model to an increase of 22% for the HADCM2 model. For the 2040s, the net effect ranges from a decrease of 16% for the ECHAM4 model to an increase of 4% for the HADCM3model. While the changes in annual volume vary significantly for different models, increases in winter flow and decreases in summer flow are consistently projected by all models and would have important consequences, as will be discussed in the following section.

Projections of temperature changes, both global and regional, are made with higher confidence than projections of precipitation changes. It would be useful to know the degree to which the projected changes in streamflow depend on the precipitation changes, which are less certain. To elucidate the separate roles of changes in temperature and precipitation in altering the region's hydrology, we have run the VIC hydrology model using projected climate change for the 2040s from the HADCM2 model, holding one variable fixed while changing the other (Fig.35). If precipitation changes but temperatures remain as observed, increases in streamflow occur in all months but the timing of the peak flow does not change. If temperatures change but precipitation remains as observed, winter flows increase owing to the greater fraction of precipitation falling as rain, and summer flows decrease, primarily because of reduced snowpack. Thus it is clear that temperature changes alone, which can be projected with greater confidence than precipitation changes, have a substantial impact on summer streamflow. These changes in streamflow timing associated with rising temperatures explain the increases in winter streamflow and decreases in summer streamflow resulting from all of the climate change scenarios despite the significant differences in precipitation changes in each scenario.

Although we have not simulated the impacts of climate change on transient and rain dominated basins on the west side of the Cascades, the implications for these basins are relatively clear from the results of the Columbia study. Transient basins would shift towards rain dominant behavior, with increases in flood risk in late winter, due to increases in both temperature and precipitation, and significant reductions in summer streamflows, primarily due to reductions in spring snowpack. Rain dominant basins would likely show increased winter streamflows resulting from increased precipitation, with less pronounced decreases in streamflow from temperature increases via increases in evapotranspiration. These changes in the fundamental seasonal patterns of river flow would have important consequences that could eventually mandate changes in the ways the rivers are managed in the PNW.

8.2 Impacts of Climate Change on Water Resources in the Columbia Basin
The actual future impacts of climate change on water resources in the Columbia Basin will be a very complex function of geophysical responses (such as changes in the timing and quantity of streamflow), changing water or energy demands (due, for example, to rising human populations and higher temperatures), changing economic conditions (e.g., the future demand and markets for energy and water), and human adaptations (e.g., changes in the way water is managed in response to the other changes). In order to limit the degrees of freedom in the problem we take the approach of evaluating the sensitivity of current management systems and water uses to future climate change in the context of altered streamflow. In the following analysis, some aspects of the water resources system and its management policies are found to be relatively robust to the projected changes in streamflow. Others are found to be sensitive to these changes, suggesting that future research to identify appropriate adaptations should be focused in these areas.

The response of the Columbia basin water resources system to observed climate variability {Link back} already gives us several clues about future climate conditions to which the Columbia basin is likely to be sensitive. As mentioned earlier, the system for managing water resources in the Columbia River Basin is relatively effective at dealing with high flows, but low flows expose the management system's weaknesses and activate a number of important impact pathways for even small reductions in flow. The system copes least effectively with long-term droughts. The climate change scenarios discussed above suggest that once anthropogenic climate change emerges above the "noise" of observed inter-annual variability, the changes in temperature and precipitation will be in the "wrong" direction, i.e., toward lower flow in summer, compounding the conflicts generated by other factors such as the rapidly growing population and the recent requirements to maintain higher instream flows for threatened fish runs. The projected increases in winter streamflow may present less of a challenge to existing management institutions, although operational changes may be required to cope with changes in quantity and timing of peak streamflows in spring.

We now consider how reliability of Columbia River water resources system objectives could change as climate and seasonal patterns of streamflow change. Using the projections of future streamflow discussed above in the ColSim reservoir model, we evaluated the impacts of climate change for the 2020s (Fig.36) and the 2040s (Fig.37) (Hamlet and Lettenmaier 1999b). In most of the scenarios examined, increases in winter flow help assure that firm energy is not significantly altered from the base case. The exception is the ECHAM4 2040s scenario, which is unusually dry and warm in comparison with the other GCM scenarios. Some objectives (e.g. non-firm energy, irrigation, recreation, and instream flows at McNary and Lower Granite) are significantly impacted by lower summer streamflows, particularly in the 2040's scenarios.

Water demand is also sensitive to climate. A study conducted by Hossein Parandvash of the Portland Water Bureau has evaluated some potential changes in urban water demand in the Portland area associated with climate change scenarios {Link **Not Done Yet**}. These kinds of changes in seasonal water demand have potentially significant implications for water resources management, particularly on the west side of the cascades.

8.3 Adaptability of the Columbia Basin Management System to Climate Change
The adaptability of management systems and institutions in the event of significant changes in climate may be examined (after Miles et al. 2000) in the context of institutional design and past adaptability to climate variability. Miles et al. (2000) conducted an integrated assessment of the Columbia basin and its management system based on a characterization of the current institutional frameworks and an evaluation of the observed ability of the system to cope with climate variations. The study arrived at the following conclusions:

  1. The Columbia River is essentially a fully developed resource, and no significant opportunities for increasing reservoir storage are likely to emerge, regardless of the actual or perceived need.
  2. The Columbia basin and its management system is adaptable to changing flood risks due to a relatively centralized management system, clear authority of the institutions involved (primarily the US Army Corps of Engineers) to take action to protect this objective, and a high priority for flood control within the existing management framework.
  3. The Columbia basin and its management system are vulnerable to increasing risk of summer low flows, because the management system is fragmented and because the existing water resources are insufficient to meet the needs of all water users in moderately low flow years. Furthermore, any attempts to change existing policy will attempt to reallocate resources without resolving the inherent conflicts, a situation which will tend to polarize potential winners and losers in the courts and political arena. Lastly, increasing population and inflexible water laws in the west will also probably contribute to the region's vulnerability to drought irrespective of any changes in climate.

In a case study of the Yakima River basin irrigation and the Seattle water supply system, Gray (1999) also showed that institutional design of water resources management systems can contribute to increased vulnerability to climate variability and change {Link **not done yet**}. These characteristics of the Columbia basin and its management institutions suggest that the basin will adapt readily to potential increases in winter high flows associated with climate change, but that without radical changes in the way the system is managed for low flows, the existing fragmented system of conflicting water policies will not be able to adapt effectively to increasing stress from growing human populations, increasing summer water demand due to rising temperatures, and the likely reductions in summer streamflows described in the preceding sections. These considerations also apply to the heavily populated areas west of Cascades, whose water supply systems are increasingly stressed by human populations and environmental needs, and are vulnerable to relatively small increases in temperature17.

In smaller watersheds within the Columbia basin that do not have any significant storage, the opportunities for adaptation are essentially reversed from the situation described above. Although water use restrictions may be very contentious to implement, managers can potentially adapt to altered climate and likely reductions in summer streamflows by appropriately restricting use of surface and groundwater resources in summer (i.e., via the permitting process). High flows are essentially uncontrolled in these unregulated basins, and aside from the proper design and construction of engineering structures like culverts and retention ponds, there is no adaptation pathway available.

Over the twentieth century, decadal-scale patterns of streamflow variability associated with the PDO have been characterized by significant long-term shifts in the average annual streamflow within a few years time. This suggests that the PDO may have an important effect on the monitoring of and adaptation to climate change, by creating strong changes in mean conditions within a very short time frame that could mitigate or exacerbate the effects of warming. One possible future scenario is a return to a cool phase of the PDO for the next 25 years or so (see e.g. Minobe, 1997; and Ware and Thompson, 1999), which could mitigate the impacts of warming to some extent, followed by a return to warm PDO, potentially causing severe drought conditions in conjunction with the warming that would have already occurred due to climate change. If monitoring and planning efforts fail to anticipate these kinds of rapid changes in "average" climate, human institutions may have difficulty adapting quickly enough to avoid impacts.

Despite these rather daunting prospects for the region's water resources, innovative management alternatives may be able to reduce or eliminate some of the potential impacts identified here.




9. An Overview of Water Management Adaptation Pathways to Cope with Climate Variability and Change

The problem of establishing effective and equitable regional scale water resources management in the changing PNW is formidable. In the Columbia River basin, for example, there is not enough water available in periods of even moderately low flow to meet all objectives simultaneously, and the institutions that manage the Columbia's water resources are fragmented and frequently ineffective at reconciling the conflicts that arise. By extension, these structural weaknesses in the Columbia Basin management system also highlight a likely inability to deal effectively with significant variations in supply associated with climate variability (e.g. decadal-scale variability like that associated with the PDO) and long-term climate change, which will result in the same kinds of conflicts between human and ecological needs for water (Miles et al., 2000). The management framework for the Columbia basin is further complicated by the presence of state and international boundaries within the basin, which are the source of political, social, legal, and economic conflicts and constraints that are not easily resolved.

Although the amount of hydropower produced by the Columbia is enormous in comparison with most water resources systems in the U.S., the management problems in the Columbia River Basin are not unique to the PNW, but are a subset of the water management problems that are facing most large river basins in United States, perhaps most notably the Colorado River Basin. How best to facilitate changes in water resources management to reduce the existing vulnerabilities to climate variability and prepare for changing climate remains an open question. Central to this issue is the complex institutional problem of controlling, and equitably distributing, impacts to various conflicting human users of water, and to ecosystems, under adverse conditions such as drought.

Even if future climate were not expected to change, it appears likely that the PNW region is likely to encounter severe difficulty in meeting water resources objectives during the next century as a result of potential increases in demand for water and hydropower (expected because of rapid population growth--see Chapter 1) and other potential changes such as the expansion of irrigated farmland, mandates for increased instream flows to protect ecosystems, increased demand for consistent reservoir elevations for recreation. These likely increases in demand for water in the PNW imply bitter conflicts between irrigated agriculture, fish protection, municipal and industrial supply, recreation, and hydropower interests. Add to these stresses those generated by climate change. Warmer, wetter winters and hotter summers will likely reduce winter snowpack, increase winter runoff and flooding in lowland areas, decrease the volume of the spring freshet, and reduce summer water supply and water quality. These changes in the timing and quantity of runoff and water supply will increase the region's vulnerability to climate and add to the existing conflicts between water resources objectives.

In attempting to design better management systems that can cope effectively with conflicting demands for water and potential changes in climate, it is appealing to attempt to construct a management system that clears away the constraints of historical development and existing institutions and functions by a single guiding principal (e.g. based solely on economic metrics, see e.g. Maas et al., 1962). In practice, however, such a simple policy can rarely cope effectively with the wide range of conflicting economic, social, political, ecological, and legal constraints that are embedded in real water resources management systems. The question of how to manage a river like the Columbia simultaneously for salmon (relatively low economic value, but high ecological, social, political, and legal value) and for irrigation and hydropower (high economic value, which defines its social and political value) when the three objectives compete for the same resource highlights the difficulties involved in constructing effective multi-use water resources management policies. Even for a single objective, the managerial goals cannot necessarily be clearly defined across the board. The objective of fish protection in the Columbia Basin, for example, cannot be readily separated from the international boundary issues inherent in the Columbia River management problem. From Canada's perspective the protection of anadromous fish runs in the Lower Columbia is directly in conflict with the goal of providing consistent lake levels to protect important fisheries habitat in the Upper Columbia (which no longer has any anadromous fish runs due to dam blockage). Thus in attempting to design policies to protect instream flow for salmon and other fish, Canada and the U.S. are unlikely to agree even about the objectives of such a policy. How best then to approach the problem of reducing regional vulnerability of water resources systems to climate variability and change?

A number of coping mechanisms for dealing effectively with climate variability and change in conjunction with increasing human population have been proposed and are discussed in the following sections. Many of the strategies discussed in the next section fall into the category of "no regrets" pathways, in the sense that they would probably produce benefits regardless of the accuracy of any predictions of change in the region (e.g. increases in population, associated changes in water demand, climate change, etc.). This is not to say that there will not be costs associated with attempting to implement change. Many of these coping mechanisms must be evaluated, planned for, designed, and/or implemented substantially ahead of the time, regardless of whether they are ultimately adopted or produce actual benefits to the region, and there is no guarantee that these costs can ultimately be recovered.

The coping mechanisms that we will examine are grouped here according to three broad categories:

  1. managing demand, e.g., by increasing the efficiency of water use
  2. increasing the aggregate supply of water, and
  3. facilitating institutional and operational flexibility by clarifying how organizational units relate to each other, providing centralized authority, and developing an ongoing and objective mechanism for matching decision rules to changing conditions via flexible operating rules based on real time information and forecasts.

9.1 Strategies for Controlling and Managing Demand
Encouraging conservation of water resources has been identified as an effective means for controlling demand. A number of opportunities for controlling and managing demand have been identified in the 1993 study by the Office of Technology Assessment (OTA) of the U.S. Congress (1993). These focus on how the federal government could encourage conservation, without explicitly directing how the conservation would be achieved.

  • Revise the tax code and create state revolving-loan funds to facilitate conservation investment.
  • Tie funding of state water projects to improved efficiency in management and consumption.
  • Encourage adoption of risk management and risk minimization practices to mitigate drought effects.
  • Encourage water conservation in federal and state facilities.
  • Require demand management via modifying rate structures, reducing use of water for landscape irrigation, modifying plumbing and irrigation systems to increase efficiency, educational programs, and metering.

Other specific possibilities for reducing demand have been identified in the OSTP Climate Change Workshop (OSTP Climate Change Report, 1997) including:

  • developing new technology for more efficient application methods of irrigated water, which could decrease water use per irrigated acre;
  • reducing irrigated acreage by increasing crop yields;
  • adopting agricultural and land management practices that reduce soil moisture loss,
  • developing new technology that would allow for increased water use efficiency (e.g. high efficiency plumbing fixtures and appliances for household use), and providing incentives for their use;
  • educating the public about the social and environmental benefits of efficient water use;
  • creating water banks to trade irrigation rights between users; and
  • introducing market forces in water management systems to adjust demand to supply and to distribute impacts during periods of water shortage.

Most of the coping mechanisms listed above are straight-forward economic incentives, public education campaigns, or engineering solutions designed to increase the efficiency of water use. Such practices have been shown to be effective in the past, and can probably be successfully extended into the future. Water banks and the introduction of market forces into water management practices, on the other hand, represent a significant departure from traditional water management practices, are largely untested in practice, and have profound social, political, and legal implications that are not immediately apparent.

9.2 Strategies for Increasing the Aggregate Supply of Water
The options that are listed below were suggested by regional water managers and researchers who attended the OSTP/USGCRP workshop on climate change.

  • encourage innovative methods of increasing water storage, including groundwater recharge schemes in which surface water (raw or treated) is pumped into the ground during times of high runoff; new (or higher) dams could also potentially increase water storage, but there are few potential dam sites left in the Columbia basin, and dams pose additional problems for salmon recovery
  • seek new sustainable sources of water, e.g., groundwaterd
  • develop strategies to encourage optimal use of existing water supplies of differing quality, for example, delivery of non-potable supplies (such as reclaimed water) for some uses such as landscape watering
  • manage water resources more effectively at the watershed level, making use of seasonal and long-term forecasts
  • reduce non-essential use of water (e.g. replace some hydropower capacity with alternative electrical plants)
  • increase flexibility, cooperation, coordination, and information sharing among users to allow increased effectiveness of response to currently unknown climate effects
  • improve the robustness and flexibility of water resources systems by connecting water supply systems with different characteristics (e.g., the proposed intertie between Tacoma and Seattle water supplies) or by developing cooperative agreements within large basins (e.g. negotiate with Canada to increase use of storage in British Columbia with the PNW region as the prime customer for this water)
  • build desalination plants to augment drinking water supply system yields during droughts (although currently expensive, cost effectiveness is likely to improve, as discussed below)

Although some of these options are more innovative than others, all are traditional water resources engineering solutions (or extensions of them) designed to increase the effective supply available for meeting demand. Most water managers are familiar with these types of coping mechanisms.

The topic of high-grade waste water treatment and water reuse is an important one that shows promise for radically increasing water use efficiency in some applications. The city of Yelm, Washington, is a planned community that will implement treatment of 100% of its waste water to grade-A quality18. Some of this reclaimed water will be used for watering landscaping and golf courses, strongly reducing demand that must be supplied by the primary resource (ground water) in summer, and the rest will be applied to wetlands where it will return to the aquifer. Direct waste water returns to the Nisqually River will be completely eliminated, which is expected to improve surface water quality and impacts to the Nisqually River estuary associated with point source nutrient loadings.

Reverse osmosis water treatment (which is capable of treating waste water or sea water to drinking water standards by forcing water through permeable filtration membranes) currently costs about $1.00 per cubic meter19 to treat and deliver (Shamir, 2000) and is relatively energy intensive in comparison with other waste water treatment methods, but is becoming cheaper and more energy efficient with increased development of membrane technology, and may ultimately become an attractive drought relief option20.



9.3 Changes to Water Resources Management to Reduce Climate Vulnerability: Institutional Redesign vs. Revised Reservoir Operations

The social and political influences on water resources that have created the institutional and operational water management problems discussed above are unlikely to disappear, and it is doubtful that the process by which changes to water management institutions occur can be divorced from their political and social origins either. Major restructuring of PNW water resources institutions and/or better coordination between institutions has been proposed as one pathway to reducing the PNW's vulnerability to climate variability and change (see e.g. Callahan et al.,1999), and there appears to be a growing awareness in the water management community that water resources needs to take a regional or basin-wide view, and that better coordination between institutions is needed. While major structural changes are to PNW water management institutions are attractive, and could significantly benefit the region, such changes also appear to be very difficult to implement on the time scales needed to effectively address the problems occurring in the PNW now, problems that we believe are likely to get worse and expand in scope as regional population grows and the climate changes over the next 20-40 years. In particular, neither the existing PNW institutions themselves nor the state legislatures in the PNW have embraced the kinds of changes that would need to be made to significantly alter the existing institutional framework. This suggests that impetus for an institutional redesign would have to be Federal in nature, a circumstance which introduces additional problems associated with the Federal government coordinating effectively with the states.

Working to change reservoir operating policies within the existing institutional and management framework may therefore be argued to be the most expeditious route to adapting regional water resources management to current and future stresses (e.g. the need to maintain flows for fisheries protection and to minimize the conflicts that arise during conditions of low flow). Recent work suggests that the effectiveness of water resources management in the Columbia River Basin (under the existing institutional framework) could be improved by using improved climate, streamflow, and demand forecasts to guide operational decisions (see e.g. Hamlet and Lettenmaier, 1999a, 2000). As we describe above, managers could increase the adaptability of hydropower operations to year-to-year changes in streamflow conditions by adapting the Columbia River reservoir operations decision process to include improved streamflow forecasting information. This kind of change, i.e., adapting existing institutions and decision processes to make better use of forecast information, would avoid the issue of a complete institutional redesign. Less extensive institutional changes would be required, and as the ongoing forecast information improved, so would the performance of the water management system. After the initial changes to link forecasts to existing water management decision processes, the system would simply respond on a year-to-year basis to the changing forecasts; no ongoing structural changes to the management system would be required, and institutional coordination would be focused on the objective application of the forecasts.

This type of approach may be particularly useful for coping with relatively rapid changes in climate. Climate change may alter the timing and quantity of regional streamflows to such a degree that the historic record will no longer be useful for assessing risk to water resources objectives. If water resources systems have adapted to make use of forecasts on a year-to-year basis, using forecasts that include the changes in climate (derived from global climate models) may offer an alternate method of effectively operating water resources systems as the climate changes (Hamlet and Lettenmaier, 2000a).

The potential benefits of implementing better forecasts in existing institutional decision processes does not imply that the process of making changes in the existing reservoir operating policies is a simple matter. On the contrary, the inherent conflicts within the Columbia basin may result in polarization of stakeholders in the legal or political arena when specific changes in management practice threaten to create a risk-based management structure where such structures do not currently exist. Entrenched institutions or economic interests may oppose change of any kind, and any changes would have to be approved by a large number of institutions, which itself requires a coordinated institutional effort. Still, these kinds of obstacles may be easier to surmount than the hurdles associated with a complete redesign of water resources management institutions in the PNW, which adds yet another layer of political and institutional conflicts to be overcome.



1. The Columbia River basin contains portions of Washington, Oregon, Idaho, Montana, Nevada, Utah, Wyoming, and British Columbia.

2. Anadromous fish are those, such as salmon, that spend part of their lives in freshwater and part in saltwater.

3. The Columbia Basin Treaty with Canada (ratified in 1964) dramatically increased the reservoir storage in the Canadian portion of the Columbia basin, and facilitated the use of Canadian projects for U.S. flood control in return for the marketing, by the Bonneville Power Administration, of energy produced in Canada. This simultaneously increased both the amount of storage available for flood control and the hydropower capacity of the Columbia River water resources system, and provided a specific mechanism for basin-wide coordination between flood control and hydropower production.

4. A critical period analysis uses an observed adverse (i.e., low) streamflow sequence from the historic record to represent the greatest stress - resulting from streamflow variability - that the management system is likely to face. To ensure that water management objectives such as reservoir refill, flood control, maintenance of energy capacity, etc. are met most of the time, operating policies (such as reservoir rule curves) are constructed based on this flow sequence.

5. A reservoir rule curve is series of seasonal (minimum and/or maximum) reservoir storage levels that is used to guide dam operators at each water storage reservoir in the system. A flood rule curve, for example, will specify for each month the maximum reservoir storage permitted by the dam operating plan.

6. The time series of natural flows was derived by the Climate Impacts Group based on natural flow estimates for 1928-1988 at The Dalles, Oregon from the BPA (Nancy Stephan, BPA, personal communication). Naturalized flows for 1989 to present and for prior to 1928 were calculated based on methods developed by the BPA to convert gage (observed) flows to naturalized flows.

7. The fine-scale meteorological data used to drive the VIC model are derived from climate monitoring station observations, interpolated to a resolution of 1/8 degree (latitude x longitude), and then rescaled to correct for topographic variation at finer scales (see Daly et al. 1984)

8. Reliability is defined as the observed probability of meeting a particular objective. For example, an objective with 90% reliability will be met in 90% of the months of the simulation.

9. High flow is defined here as more than 1.5 standard deviations above the long-term mean.

10. Low flow is defined here as more than 1.5 standard deviations below the long term mean.

11. The possible exception to the pattern was water year 1997, the second-wettest year on record, which may have occured in a warm PDO cycle. However, because this event exceeded the greatest flow observed for a warm-phase PDO year by such a wide margin, it suggests that the PDO may have shifted back to the cool phase prior to 1997, in which case all five highest-flow years would have occurred in the cool phase.

12. The mean annual flood is defined as the long-term average of the largest daily flow occurring each year.

13. The Washington Department of Transportation, for example, requested information from the Climate Impacts Group in the fall of 1997 about the risk of roadway flooding in the fall and winter of 1997-1998.

14. Compared to other rivers, the Columbia has a low ratio of storage to average run-off. On the Missouri river, dams hold up to two or three times the annual average runoff, thus allowing for greater control. See BPA et al. (1991) and (1993).

15. A reservoir rule curve is series of seasonal reservoir storage levels that is used to guide dam operators. A flood rule curve, for example, will specify for each month the maximum reservoir storage permitted by the dam operating plan.

16. Hamlet and Lettenmaier (1999b) show that changes in evapotranspiration (ET) are most pronounced in June, when ET during snow melt increases strongly at moderate elevations. In other times of the year, the changes in ET and soil moisture are variable between models and are influenced by changes in both temperature and precipitation, which determine trade-offs between available energy for evaporation and available moisture.

17. Most of the west side water supply reservoirs are fed by small transient snow river basins. As discussed above, these are the types of basins most vulnerable to dramatic streamflow timing shifts associated with warming.

18. Grade A treated water is suitable for all non-potable water uses, such as lawn watering in parks where there is public use.

19. For reference, Seattle municipal water customers pay about $0.89 per cubic meter for water in summer.

20. The California Department of Water Resources, for example, lists six operational reverse osmosis desalination plants with capacities in excess of 2 million gallons per day (Bulletin 160-98: California Water Plan, California Department of Water



Back to top | CIG Home