Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 73
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability 3 Climate and Hydrology of the Colorado River Basin Region The Colorado River basin contains climate zones ranging from alpine to desert and exhibits significant climate variability on a variety of time scales. These variations have important implications for snowmelt and river hydrology and are thus of interest to both scientists and water managers in the Colorado River region. Scientific research on the Colorado River basin’s climate and hydrologic systems has included measurements of the river’s flow, long-term studies of climate and river hydrology, reviews of statistics associated with temperature and precipitation extremes, and studies of connections to regional and global climate systems. In the 20th century, long-term water management and planning in the region generally relied upon the gaged record of Colorado River flows; specifically, great reliance was placed on measurements made at Lees Ferry, supplemented by data recorded at other stations on the mainstem and on tributary streams. Some of these gaged streamflow records for the Colorado River date back to the late 19th century, but most began during the 20th century. Although a time frame of over 100 years may appear to offer an extensive record of climate and streamflow variability, in fact it represents a relatively short period in terms of geologic history of the region. In recent years, the once-prevailing view of climate as static and unchanging on time scales important to river managers has given way to a new understanding that the gaged record represents only a small temporal window of the variability characteristics encompassing many centuries of Colorado River hydroclimate. River management decisions are inherently forward looking and rely heavily on forecasts. These forecasts typically assume that past properties of the river system, as revealed through observations, will be replicated in
OCR for page 74
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability future conditions. However, the prospect of changing states of atmospheric conditions and climate behavior, associated with anthropogenic emissions of greenhouse gases, calls this assumption into question. As a result, many water managers today are exploring ways of adjusting water planning and management strategies. The study of climates that occurred before direct measurements of weather and climate data—paleoclimatology—can serve as part of the hydroclimatic information considered in water management decisions. This field of study draws upon indirect, or proxy, information about past climate conditions obtained from evidence contained in glacial ice, landscape features, sediment deposits in ancient lakes, pollen, species distributions, preserved organisms (e.g., mollusks), and middens. The science of dendrochronology, or the study of the sequences of annual growth layers (rings) of coniferous trees, is particularly relevant in the Colorado River basin. For several decades, cores from coniferous trees in the western United States have been analyzed to enhance understanding of past climate. Recent tree-ring analyses have incorporated updated chronologies and longer calibration periods to estimate annual Colorado River flows over the past several centuries. These new dendrochronological reconstructions have stimulated heightened interest in questions regarding the rarity and recurrence of drought conditions across the region. This chapter discusses fundamental features and dynamics of Colorado River basin climate (including climate trends and future climate scenarios), the gaged record of Colorado River streamflow, and tree-ring studies of past Colorado River region streamflow. The concluding Commentary section discusses implications of this hydrologic and climatic information for water resources planning and decisions. FEATURES AND DYNAMICS OF COLORADO RIVER BASIN CLIMATE Precipitation Patterns and Sources The Colorado River is primarily a snowmelt-driven system, with most precipitation in the basin falling as winter snowfall in higher
OCR for page 75
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability elevations of Colorado, Utah, and Wyoming. In the upper Colorado River basin, approximately 20 percent of the basin’s precipitation falls in the highest 10 percent of the basin, and roughly 40 percent of the basin’s precipitation falls in the highest 20 percent of the basin. Cold temperatures at high elevation cause precipitation to occur mainly as snow and to remain frozen during the winter months. This “white reservoir” drapes the mountain terrain during winter months and survives into summer at the highest locations. Some of the water in this snowpack is lost to the atmosphere through sublimation (a phase change from solid to vapor) during the cool season. Most remains, however, and as the snowpack warms, or “ripens,” in the spring, meltwater is steadily metered into the soil. This process extends for several weeks to months at higher elevations, and melting occurs slowly enough to recharge the soil and allow water to enter the myriad channels that feed the Green and Colorado rivers. For these reasons, winter precipitation over the high-elevation portion of the upper basin plays an important large role in generating runoff and streamflow. Warm season precipitation plays a different role in the basin’s hydrology. During warmer months precipitation falls more intensely, often in localized, convective thunderstorms. Plants are photosynthetically active at all elevations and utilize some of this water immediately. Furthermore, almost all the summer precipitation intercepted by vegetation canopies evaporates directly to the atmosphere. Much of the remainder of summer precipitation that infiltrates into the soil column is transpired by plants or (in the case of bare ground) evaporates, aided by warm soil. A relatively small fraction of summer precipitation makes its way into aquifers and streams. In the basin’s high-elevation headwaters, summer precipitation amounts are generally less than winter values. The high-elevation winter dominance of annual precipitation is more pronounced in the Green River drainage than in the Colorado River headwaters in central Colorado. In the basin’s lower and drier reaches, summer precipitation can account for a larger share of annual total precipitation, but because of higher evaporation and transpiration rates, this moisture is less effective in contributing to streamflow. In the hottest and lowest portions of the basin, summer precipitation matters greatly to local vegetation and to small runoff channels, but hardly at all to the mainstem Colorado and its major tributaries.
OCR for page 76
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability The main source of summer moisture is the North American monsoon, which transports moisture into the region from sources in the subtropical Pacific and Gulf of Mexico. This annual phenomenon brings drama to the southwestern desert skies, but only occasionally does it provide enough precipitation to contribute appreciably to hydrologic supplies. For the mainstem Colorado River and its major tributaries, the bulk of the precipitation that contributes to water supply falls during the winter months, primarily in the form of snows at high elevation. Summer months comprise the period of higher water demands and, except in extreme weather years, will provide at best only modest additions to mainstem reservoir water supplies. If a season of winter precipitation and water storage is “lost” because of drought conditions, there will be little opportunity to replenish supplies until the following winter. The Tropical Pacific and ENSO Ocean temperature patterns that have the greatest influence on Colorado River basin climate are in the tropical Pacific in a band that straddles the equator between Peru and the International Date Line. At irregular intervals of typically 2-7 years, sea surface temperatures (SSTs) in this region warm above climatological averages.1 This phenomenon, called El Niño, is part of a complex ocean-atmosphere oscillation. El Niño has a climatic counterpart called La Niña that is characterized by below-average SSTs (La Niña events usually have smaller departures from average SST than do El Niño events). The terms El Niño and La Niña refer only and exclusively to ocean temperatures in this geographic domain and not to their effects elsewhere. Another atmospheric feature relates to barometric pressure gradients in the South Pacific. In the 1920s, British meteorologist Sir Gilbert Walker published his seminal work describing the inverse relationship in atmospheric surface pressure between Tahiti and Easter Island in the tropical Pacific, and over Darwin in northern Australia (Walker, 1925). That is, when atmospheric pressure is high in one of these locations it tends to be low in the other region, and vice versa. Walker termed this phenomenon the Southern Oscillation. It refers only to the atmosphere. The Darwin-Tahiti pressure difference (nor- 1 Tropical Pacific SSTs are 1-3°C above average in modest El Niño events, 3-5°C above average in major episodes.
OCR for page 77
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability malized for variability over the past century) is the basis of the Southern Oscillation Index (SOI). Furthermore, when Tahiti has lower than average pressure and Darwin has higher than average pressure (negative SOI), a strong tendency exists for El Niño to be present. Conversely, there is a tendency for La Niña conditions to exist with higher pressure in Tahiti and lower pressure in Darwin. The oceanic (SST) and atmospheric (SOI) measures are usually highly correlated and these terms are sometimes used interchangeably (McCabe and Dettinger, 1999). For historical reasons these phenomena are often lumped together and referred to (although somewhat asymmetrically) as El Niño-Southern Oscillation, or ENSO. The ENSO phenomenon owes its existence to coupled ocean-atmosphere interactions over the equatorial Pacific and is an important contributor to interannual global climate variability. The ENSO cycle has impacts on climate over large areas of both the tropics and extratropics. Jerome Namias was the first to investigate extensively the possible relationship between SST and North American atmospheric circulation. Jacob Bjerknes identified the equatorial Pacific as the source of climate variability associated with the Southern Oscillation. The winter storm track over the eastern Pacific Ocean shifts southward during El Niño episodes, often causing wet winters in the southwestern United States and dry winters in the Pacific Northwest and northern Rockies. La Niña winters tend to bring the opposite pattern, and moderately positive values of the SOI in the prior summer/autumn nearly guarantee a dry winter in the southwestern United States—it is the most dependable predictive climate relationship in the United States (Redmond and Koch, 1991). In Arizona and New Mexico, and extending into the San Juan Mountains of southwestern Colorado, El Niño winters are generally wetter than normal, but not always, and a few are extremely dry. Moreover, the likelihood of an extremely wet winter is much higher during El Niño winters and there are few wet winters when El Niño conditions are not present (Redmond and Koch, 1991). These patterns are accentuated in streamflow, particularly in extreme high and low streamflow (Cayan et al., 1999). Precipitation patterns in the western United States vary considerably among different El Niño events. These differences appear to depend on the particular spatial pattern of warm ocean temperatures, the magnitude of warming, and the particular months of the year when these patterns occur. Accurate forecasting of these ocean
OCR for page 78
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability features and their North American effects represents one of today’s principal ENSO-related forecasting challenges. Within the Colorado River basin, ENSO effects are more pronounced in the lower basin than in the upper basin. The San Juan River shows the same strong relationship to ENSO as does Arizona. By contrast, the headwaters of the Green River (in Wyoming’s Wind River Mountains) tend to be slightly more influenced by the northern pole (centered over the Columbia River basin) of this winter dipole pattern (Redmond and Koch, 1991). The main source regions of Colorado River basin precipitation and streamflow—the mountains of Colorado, Wyoming, and northeastern Utah—are not greatly impacted by ENSO events. Because roughly 90 percent of the river’s flows originate in mountain headwater regions with limited connection to ENSO, better forecasts of ENSO and its effects are not likely to greatly improve upper basin mainstem streamflow forecasts. Other Ocean Connections Another pattern of Pacific regional scale climate variability related to SST variations is the Pacific Decadal Oscillation (PDO). The term was coined in 1996 by fisheries scientist Steven Hare while he was studying connections between the Alaska salmon production cycle and Pacific climate (http://jisao.washington/edu/pdo). The PDO describes joint variations in SST, atmospheric pressure, and wind in the central and eastern Pacific poleward of 20°N (Mantua et al., 1997). The warm and cool phases of the PDO each historically have lasted two to three decades, for a total period of about a half-century. An abrupt jump in Pacific-wide environmental conditions known as the “1976 shift” (Ebbesmeyer et al., 1991; Trenberth and Hurrell, 1994) was identified retrospectively and helped lead to identification of the PDO. This pattern appears to alternately accentuate and counteract the effects of ENSO in the Pacific Northwest and the south-western United States and is expressed most strongly in winter. The origin of this oscillation has not been definitively determined. It is linked to periods of greater and lesser frequency of El Niño and La Niña at equatorial latitudes, even though the PDO index has only a modest correlation with the SOI (Mantua et al., 1997). Although there are intriguing statistical relationships associated with the PDO, the physical mechanisms that underlie the PDO behavior itself, and
OCR for page 79
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability that lead to its expression within the Colorado River basin (and primarily in the lower basin, as is the case with ENSO), have not been fully explained. In recent years another pattern has been identified that appears to have ties to the Colorado River basin. Atlantic Ocean SSTs exhibit a mode of variability that has similar departures from average for one to two decades over an area spanning low to high latitudes; this feature is known as the Atlantic Multidecadal Oscillation (AMO). That the AMO has effects on climate and streamflow in the eastern United States (Sutton and Hodson, 2005) is understandable; however, additional studies have shown some surprising results. Notably, when the North Atlantic is warm for a decade or longer, streamflow in the upper Colorado River basin tends to be lower than average, and vice versa (Gray et al., 2004; McCabe and Palecki, 2006; McCabe et al., 2004). This headwaters streamflow is largely governed by winter precipitation. The physical mechanism by which the Atlantic could influence mountain winter precipitation in Colorado and Wyoming, which are upstream in the atmospheric winter flow pattern, remains a puzzle. The evidence so far is statistical and largely dependent on just a few AMO cycles. Theory and models are just beginning to address this potential link (Delworth and Mann, 2000; Knight et al., 2005) and observational studies are continuing. For example, during warm Atlantic phases, moisture delivery to the conterminous United States is diminished (Schubert et al., 2004a). Diagnostic studies of the global pattern of ENSO cycle variability clearly have revealed that the atmosphere acts as a bridge linking SST anomalies in the equatorial Pacific to yet larger patterns of atmospheric and ocean variability. Variations in SSTs in the tropical Pacific may herald changes in jet stream patterns, strength and track of Pacific winter storms, and future water supply conditions across the Colorado River basin. Different patterns may accentuate or counteract each other. For example, the effect of Indian Ocean temperatures acting in concert with La Niña has been demonstrated as helping produce “the perfect ocean for drought” in the southwestern United States (Hoerling and Kumar, 2003). Research has shown that the American Dust Bowl of the 1930s was in part caused by tropical ocean temperature departures from normal (Schubert et al., 2004b; Seager, 2006; Seager et al., 2005). Other western droughts, such as the droughts during the Civil War era and in the 1890s, may have similar explanations (Seager, 2006; Seager et al., 2005). Linkages
OCR for page 80
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability among these patterns suggest modest predictability, enough that they may merit consideration in water supply planning across the western United States. CLIMATE TRENDS AND PROJECTIONS Climate Records and Past Trends In a previous era of Colorado River water management there was an implicit assumption that the main features of future climate states would closely resemble those of the past century. Over time, additional research has enhanced understanding of the variability of past climate over longer time scales. Moreover, increasing levels of atmospheric greenhouse gases and steadily increasing global mean surface temperatures have heightened awareness of the potential for human activities to impact the global climate system (Houghton, 2004). The assumption that future climate conditions will largely replicate past conditions is now frequently being called into question. Variations in precipitation and water supply have long been of interest to water managers for daily, monthly, and annual operations. Less widely appreciated are the impacts that temperature has on water availability, through effects on both supply and demand. Temperature affects the quantity of and timing of snowmelt runoff in spring and summer, the occurrence of large floods, and rates of evapotranspiration. Anything that affects basin temperatures in a long-term, systematic way thus should be of considerable interest, regardless of its origin. The observed time series of basin-averaged precipitation and temperature are important for assessing regional impacts of global climate change and are discussed in the following section. Precipitation Colorado River basin precipitation exhibits high year-to-year (interannual) variability. Figure 3-1 shows interannual precipitation variability across the upper Colorado River basin, spatially averaged over the basin upstream of Lees Ferry and aggregated to annual resolution (Kittel et al., 1997; updated data from ftp://ftp.ncdc.noaa.gov/pub/data/prism100).
OCR for page 81
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability For example, after a period of less variability for several decades in the mid-20th century, there has been a tendency toward greater variability in the latter decades of the 20th century. The past 30 years of data include the highest and lowest annual precipitation in the 100-year record, and there has been a tendency toward multiyear episodes of both wet and dry conditions. Some years in the early and mid-1980s were at least as wet as the period that preceded the signing of the Colorado River Compact. Prior to the early 21st century drought, the driest comparable 5-year consecutive interval was the 1950s drought. The only other comparable 5-year dry period was at the end of the 19th and beginning of the 20th century. Despite these variations, there is no significant trend in interannual variability of precipitation over the past 110 years. FIGURE 3-1 Annual precipitation for the Colorado River basin above Lees Ferry, 1895-2005. NOTE: Red: annual values. Blue: 11-year running mean. SOURCE: Western Regional Climate Center.
OCR for page 82
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability Temperature Figure 3-2 shows annual mean temperatures for the entire Colorado River basin from 1895 to 2000 (adjusted for variations in elevation using the same method as for precipitation in Figure 3-1). Upper and lower basin temperature trends are similar and bear a strong resemblance to the history of temperature across the entire western United States (Redmond, in press), as well as to mean global surface temperature trends. Figure 3-2 shows that since the late 1970s the Colorado River region has exhibited a steady upward trend in surface temperatures. The most recent 11-year average exceeds any previous values in the over 100 years of instrumental records. One striking aspect of Figure 3-2 is how much warmer the region has been in the drought of the early 2000s as compared to previous droughts. For example, temperatures across the basin today are at least 1.5°F warmer than during the 1950s drought. Increasing FIGURE 3-2 Annual average surface air temperature for entire Colorado River basin, 1895-2005. NOTE: Red: annual values. Blue: 11-year running mean. SOURCE: Western Regional Climate Center.
OCR for page 83
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability temperatures in the region have many important hydrologic implications, including the impacts of drought. For example, the drought of the early 2000s has taken place in particularly warm conditions. Figure 3-3 shows temperature departures for that 6-year period (2000-2005) as compared to 1895-2000 averages. Both in terms of absolute degrees and in terms of annual standard deviation, the Colorado River basin has warmed more than any region of the United States—a fact that should be of great interest throughout the region. This trend continued through the first half of 2006. This warming is well grounded in measured climatic data, corroborated by independent data sets, and widely recognized by climate scientists throughout the West. The trend of increasing temperatures in the western United States also is seen in larger, global temperature trends. For example, a 2005 paper on western mountain snowpack trends notes that “increases in temperature over the West are consistent with rising greenhouse gases, and will almost certainly continue” (Mote et al., 2005). And in a recent review of global surface temperature records of the past 2,000 years, a committee of the National Research Council (NRC, 2006) concluded that It can be said with a high degree of confidence that global mean surface temperature was higher during the last few decades of the 20th century than during any comparable period during the preceding four centuries. This statement is justified by the consistency of the evidence from a wide variety of geographically diverse proxies (NRC, 2006). Key manifestations of warmer temperatures in western North America are a shift in the peak seasonal runoff (driven by snowmelt) to earlier in the year, increased evaporation, and correspondingly less runoff. In fact, many of these changes have been documented: Winter and spring temperatures have increased in western North America during the twentieth century (e.g., Folland et al. 2001) and there is ample evidence that this widespread warming has produced changes in hydrology and plants…. The timing of spring snowmelt-driven streamflow has shifted earlier in the year (Cayan et al. 2001; Regonda et al. 2004 [corr Regonda et al., 2005]; Stewart et al. 2005), as is expected in a warmer climate (Hamlet and Lettenmaier 1999a) (Mote et al., 2005).
OCR for page 101
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability sensitive species (see Meko et al.  and Meko  for more detailed discussions on methods for assessing relationships between annual tree-ring growth and streamflow). To generate streamflow reconstructions, trees are sampled with an increment borer at collection sites based on the factors described above that affect tree ring growth. Sample replication at individual sites is important, and two cores are collected from each of 15-40 different trees per site. Cores from each site are cross-dated, measured, standardized (e.g., the size/age trend is removed), and combined into tree-ring site chronologies (Cook and Kairiukstis, 1990; Stokes and Smiley, 1968), which are the basis of a streamflow reconstruction. Tree-ring chronologies are calibrated with gage data to develop a reconstruction model. Several statistical approaches, typically based on multiple linear regression, have been employed to develop these models (Loaciga et al.  provide a review of these approaches). Reconstruction models are evaluated with a suite of statistics that quantifies the variance explained in the gaged record by the reconstruction, and the uncertainty related to the unexplained variance. Reconstructions are validated by testing the model on data not used in the calibration, to ensure that the model is not tuned specifically to the calibration data, but performs well on independent data as well (Cook and Kairiukstis, 1990; Fritts, 1976; Loaciga et al., 1993). The model is applied to the full length of the chronologies to generate an extended record of flow. In applying these models back in time, the assumption is made that the relationship between tree growth and climate in the calibration period also existed in the past, while recognizing that conditions of the past were not necessarily the same as in the instrumental period (Fritts, 1976). Uncertainties in Streamflow Reconstructions Considering that reconstructions are only estimates of flow, uncertainties in these reconstructions derive from several different sources. The fact that trees are imperfect recorders of hydrologic variability is an inherent source of uncertainty and is reflected in the inability of tree ring-based models to account for 100 percent of the variance in the gaged record (e.g., Brockway and Bradley, 1995). This also makes a direct comparison between reconstructed and gaged values inappropriate unless this uncertainty is considered. The preci-
OCR for page 102
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability sion with which tree-growth rings can be used to estimate past flows is quantified by the statistical model generated in the calibration, and a measure of the error in the reconstruction model can be used to describe model confidence intervals. However, this is only one source of uncertainty. Other sources include changes in tree-ring sample numbers over time, which affects the strength of the common (hydroclimatic) signal in the reconstruction (Cook and Kairuikstis, 1990; Wigley et al., 1984). Uncertainties can also derive from the preservation of low-frequency (multidecadal to centennial scale) information in the tree-ring data, which is limited by the lengths of the individual tree-ring series and how these series were standardized to remove the biological growth curve (Cook et al., 1995). There is also some degree of uncertainty because of the quality of the gage record used for the calibration and how that may vary over time. In addition, most reconstructions better replicate dry extremes than wet extremes (Michaelsen, 1987). Reconstructed flows that are higher or lower than the range of values in the gage record often reflect tree-ring variations beyond the range of variations in the calibration period, and may be less reliable than indicated by regression results (Graumlich and Brubaker, 1986; Meko and Graybill, 1995; Meko et al., 1995). Dendrochronologists have long acknowledged and reported the model error in reconstructions, although error bars have not typically been presented with reconstructions, which would reinforce the probabilistic nature of the reconstruction values. A variety of techniques are used, with some currently under development, to identify and quantify other sources of uncertainty (Meko et al., 2001; Woodhouse and Meko, 2007). An approach to systematically quantify the amount of error attributed to each of these sources, however, has not yet been developed. Reconstructions of Colorado River Flows at Lees Ferry, Arizona As methods for tree-ring-based reconstructions have evolved, the set of streamflow data from the Lees Ferry gage has been a focus of reconstruction studies. Several reconstructions for Lees Ferry flow have been generated, first by Stockton and Jacoby (1976), followed by Michaelsen et al. (1990), Hidalgo et al. (2000), and Woodhouse et al. (2006). Stockton and Jacoby (1976), Michaelsen et al. (1990), and
OCR for page 103
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability Hidalgo et al. (2000) used similar networks of tree-ring data, with at least 30 percent of the chronologies shared and with a common end date of 1963. Woodhouse et al. (2006) used a new network of tree-ring data, ending in 1995. All four studies used different gage data for calibration, and Stockton and Jacoby (1976) used two different sources of gage data, illustrating the difference the gage records can make in the final reconstruction. The number of years for calibration also varied from 49 to 90 years. The reconstructions also included some differences in the statistical treatment of the tree-ring data and statistical approaches to the calibration (see Table 3-2). The resulting reconstructions differ in some respects. Given that these studies employed different data sets, assumptions, and methods, some differences across results are to be expected. All these reconstructions, however, share similar key features with respect to the timing and duration of major wet and dry periods. These reconstructions, as depicted in Table 3-2 and shown in Figure 3-6, support the following points: Long-term Colorado River mean flows calculated over these periods of hundreds of years are significantly lower than both the mean of the Lees Ferry gage record upon which the Colorado River Compact was based and the full 20th century gage record (Woodhouse et al., 2006). High flow conditions in the early decades of the 20th century were one of the wettest in the entire reconstruction. The longer reconstructed record provides a richer basis from which to assess the range of drought characteristics that have been experienced in the past, revealing that considerably longer droughts have occurred prior to the 20th century. These three points have important implications for water management decisions for the Colorado River basin and are revisited in the Commentary section at the end of this chapter.
OCR for page 104
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability TABLE 3-2 Lees Ferry Reconstructions Reconstruction Calibration Years Source of Gage Data Chronology Type c Regression Approach d Variance Explained Reconstruction Years Long-Term Mean e (MAF) Stockton and Jacoby (1976) a. 1899-1961 b. 1914-1961 c. 1914-1961 Average of a and b Hely (1969) Hely (1969) UCRSFIG (1971) Standard Standard Standard PCA with lagged predictors 0.75 0.78 0.87 1512-1961 1512-1961 1511-1961 1520-1961 14.15 13.9 13.0 13.4 Michaelsen et al. (1990) 1906-1962 Simulated flowsa Residual Best subsets 0.83 1568-1962 13.8 Hidalgo et al. (2000) 1914-1962 USBR, see Hidlago et al. (2000) Standard Alt. PCA with lagged predictors 0.82 1493-1962 13.0 Woodhouse et al. (2006) USBRb Lees-A Lees-B Lees-C Lees-D 1906-1995 1906-1995 1906-1995 1906-1995 Residual Standard Residual Standard Stepwise Stepwise PCA PCA 0.81 0.84 0.72 0.77 1490-1997 1490-1998 1490-1997 1490-1998 14.7 14.5 14.6 14.1 a Simulated flows developed from the U. S. Bureau of Reclamation (USBR) Colorado River Simulation System. b J. Prairie, USBR, personal communication, 2004. c Standard chronologies contain low-order autocorrelation related to biological persistence; residual chronologies contain no low order autocorrelation. d Regression approach: PCA is principal components analysis (regression). Best subsets is multiple linear regression, using Mallow’s Cp to select best subset. Alternative PCA used an algorithm find the best subset of predictors on which to perform PCA for regression. Stepwise is forward stepwise regression. e Long-term mean based on 1568-1961 except for Michaelsen et al. (1990), which is based on 1568-1962.
OCR for page 105
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability FIGURE 3-6 Colorado River annual streamflow reconstructions, Lees Ferry, AZ (smoothed with a 20-year running mean). NOTE: Year plotted is the last year in the 20-year mean. SOURCE: Lees-B (standard chronologies, stepwise regression) from Woodhouse et al. (2006); Hidalgo et al. (2000); S&J from Stockton and Jacoby (1976; average of two models); Lees Gage is gage record, 1906-1995, J. Prairie, USBR, personal communication, 2004. Differences among the Reconstructions The most obvious difference among the reconstructions is the long-term mean, a measure with implications for long-term water allocation decisions. The reconstructions based on the calibration periods that end in 1961 or 1962 generally have lower long-term means than more recent reconstructions with a calibration period that ends in 1995 (Table 3-2). A second noticeable difference is the magnitude of the high and low flow periods, which vary between all reconstructions to some degree. Some differences in the Lees Ferry reconstructions may be attributed to the tree-ring and gage data, including the length of the calibration period. Differences may also result from choices made in statistical methods when processing tree-ring data, which can affect the characteristics of the chronology and, in turn, affect the reconstruction (see Meko et al.  and Woodhouse and Meko  for details on the treatment of tree-ring data). In the Lees Ferry reconstructions, Stockton and Jacoby (1976) and Hidalgo et al. (2000) used chronologies that retained the biological persistence (standard chronologies), which is the tendency for a tree’s growth in one year to be associated with growth in a following year due to biological processes. In contrast, Michaelsen et al. (1990) used chronologies in which this bio-
OCR for page 106
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability logical persistence was statistically removed. Woodhouse et al. (2006) tested models using both types of chronologies. Different results may also arise from the statistical approach used in the calibration process and can stem from the inclusion of “lagged predictors” (tree-ring chronologies lagged forward and backward several years relative to the gage record) and details of regression methods used (see Woodhouse et al.  for more information on statistical methods used in dendrochronology research). The sensitivity of the resulting reconstructions to some of these statistical treatments and approaches has been tested for Lees Ferry (Woodhouse et al., 2006). Results from this study indicate that different types of chronologies (standard versus residual) can have an influence on the skill of the reconstruction in replicating some of the time-series characteristics of the gage record, and persistence of low flows may be heightened with standard chronologies (Woodhouse et al., 2006). The use of different modeling approaches in model calibration was not an obvious source of differences. In addition, the length of the calibration period did not appear to be critical, as calibrating a model on a shorter time period (1914-1961 versus 19061995) resulted in a similar reconstruction (Woodhouse et al., 2006). In summary, differences among Lees Ferry reconstructions can likely be attributed to several factors. There are some indications that periods of persistent low flows may be accentuated using standard chronologies and/or lagged predictors, but the sources of the differences in long-term mean are not yet clear. Additional studies will be needed to more accurately assess the impact of the different sets of chronologies and gage records on the final reconstructions. As to which reconstruction might be the most accurate or “best,” reconstructions with the longest calibration period are statistically more robust (i.e., exhibiting similar results when tested with different models), particularly considering that the recently recalibrated gage record from 1906-1995 is assumed to be the most accurately estimated natural flow data. Within the set of reconstructions calibrated on the longest period, however, there is no clearly superior solution, with each reconstruction containing strengths and weaknesses (e.g., match in persistence in the gage record, over/underestimation of decadal-scale low flows; Woodhouse et al., 2006).
OCR for page 107
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability Analyzing Reconstructed Colorado River Flow The extended record of streamflow provided by the tree-ring reconstructions is useful for assessing the characteristics of the gage record in a long-term context and for examining low-frequency (multidecadal-scale) behavior of flow, which is not possible with the shorter gage record. Questions relevant to drought and management in the upper Colorado River basin that can be addressed are: How does the drought of the early 2000s compare to other past droughts of similar duration? Have longer periods of drought occurred? and What is the character of decadal-scale variability over time compared to the 20th century? When early 2000 drought conditions are assessed as a 5-year (2000-2004) mean value, the reconstruction indicates one period—1844-1848—with a lower mean value, but several additional periods with a fairly high probability of being lower as well (Woodhouse et al., 2006). The Lees Ferry gage record contains no periods of below median flow that lasted more than 5 consecutive years. In the Lees Ferry tree-ring-based reconstruction, however, longer periods of below-median flow have occurred, including periods of up to 10 and 11 years. The reconstruction also reflects the nonstationarity—or changes in the values of decadal-scale means—of flow over decadal time scales (Figure 3-6). Colorado River Sub-Basin Relationships and Circulation In addition to the record of upper Colorado River flow at Lees Ferry, reconstructions can provide information about long-term hydroclimatic variability within Colorado River sub-basins. Along with Lees Ferry, flow records at gages on major tributaries of the upper Colorado River—the Green River, the San Juan River, and Colorado River mainstem (i.e., before it joins the Green and San Juan rivers, which was historically known as the Grand River)—have been reconstructed (Woodhouse et al., 2006). A comparison of reconstructions for these tributaries suggests that major multiyear droughts and multidecadal dry periods impact the entire basin, although the relative
OCR for page 108
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability magnitude may vary spatially. Similarly, research that examined reconstructions of several tributaries of the lower Colorado River basin in Arizona—in the Salt and Verde River basins—found droughts (and wet events) in the upper Colorado and Salt-Verde River basins to be concurrent more often than not (Hirschboeck and Meko, 2005). Details of the primary mechanisms that influence upper Colorado River basin climate and hydrology at multidecadal time scales are not yet clear. Studies of extended periods of streamflow, however, considered along with other high-resolution climate reconstructions, have the potential to increase scientific understanding of the links between ocean/atmosphere circulation and Colorado River basin water supply. COMMENTARY A steady warming trend of about 2°F has been under way over the past three decades across the Colorado River basin. Results from several different climate modeling experiments indicate that future temperatures will continue to rise across the Colorado River basin. Projections of annual precipitation changes from these same models exhibit a range of results, most of them approximately centering around present values. The models project a tendency for increases in winter precipitation of about the same magnitude as decreases in summer precipitation. Higher temperatures will cause higher evaporative losses from snowpack, surface reservoirs, irrigated land, and land cover surfaces across the river basin. Hydrologic modeling studies of future Colorado River runoff exhibit a variety of results, and such forecasts always reflect some degree of uncertainty. Collectively, however, these studies indicate that future Colorado River streamflow will decrease with increasing future temperatures. The 20th century saw a trend of increasing mean temperatures across the Colorado River basin that has continued into the early 21st century. There is no evidence that this warming trend will dissipate in the coming decades; many different climate model projections point to a warmer future for the Colorado River region. Modeling results show less consensus regarding future trends in precipitation. Several hydroclimatic studies project that significant decreases in runoff and streamflow will accompany increasing temperatures. Other studies, however, suggest increas-
OCR for page 109
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability ing future flows, highlighting the uncertainty attached to future runoff and streamflow projections. Based on analysis of many recent climate model simulations, the preponderance of scientific evidence suggests that warmer future temperatures will reduce future Colorado River streamflow and water supplies. Reduced streamflow would also contribute to increasing severity, frequency, and duration of future droughts. In the context of multidecadal and multicentury hydroclimatic patterns across the Colorado River region, the Lees Ferry gaged record represents a chronologically limited sliver of information. Paleoclimate-based reconstructions of Colorado River streamflow have become of great interest to water managers across the region because, instead of 100 years of Colorado River flows, the reconstructions provide estimates of hundreds of years of flows. The first tree-ring-based reconstruction was developed in the 1970s and has been followed by several other studies using similar tree-ring data. Although the various reconstructions are not perfectly congruent, this is not unexpected given that the reconstructions were independently developed by scientists using different data sets and relying upon differing assumptions and statistical methods. Nonetheless, the reconstructions exhibit broad agreement in several important respects: they replicate similar past wet and dry periods; they suggest that the Colorado River’s long-term mean annual flow is less—ranging from 13 to 14.7 million acre-feet—than 15 million acre-feet (the mean annual value based on the Lees Ferry gaged record); they show that the 1905-1920 period was one of the wettest periods in the past several centuries; and, they indicate multiple drought periods that were more persistent and severe than droughts reflected in the gaged record. Past climates may not necessarily be replicated in the future but reconstructions of past flows provide information that, when used in concert with projections of future climate, can offer valuable guidance to aid future water resources planning. Although much remains to be learned about the drivers of hydroclimatic variability in the basin—particularly those that operate at multidecadal and longer time scales—the scientific foundation underlying contemporary understanding of Colorado River streamflow patterns has evolved markedly during the past 50 years. Whereas in the mid-1950s that foundation relied almost solely upon gaged flow records, today it consists of a more sophisticated understanding and modeling of the global climate system, better temperature data from
OCR for page 110
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability the Colorado River region and across the world, paleoclimate studies and streamflow reconstructions, and a longer record of gaged river flows. Assessed collectively, this body of knowledge invalidates any assumption that Colorado River flows vary around an annual mean value that is static and unchanging. For many years, scientific understanding of Colorado River flows was based primarily on gaged streamflow records that covered several decades. Recent studies based on tree-ring data, covering hundreds of years, have transformed the paradigm governing understanding of the river’s long-term behavior and mean flows. These studies affirm year-to-year variations in the gaged records. They also demonstrate that the river’s mean annual flow—over multidecadal and centennial time scales, as shown in multiple and independent reconstructions of Colorado River flows—is itself subject to fluctuations. Given both natural and human-induced climate changes, fluctuations in Colorado River mean flows over long-range time scales are likely to continue into the future. The paleoclimate record reveals several past periods in which Colorado River flows were considerably lower than flows reflected in the Lees Ferry gaged record, and that were considerably lower than flows assumed in the 1922 Colorado River Compact allocations. Multicentury, tree-ring-based reconstructions of Colorado River flow indicate that extended drought episodes are a recurrent and integral feature of the basin’s climate. Moreover, the range of natural variability present in the streamflow reconstructions reveals greater hydrologic variability than that reflected in the gaged record, particularly with regard to drought. These reconstructions, along with temperature trends and projections for the region, suggest that future droughts will recur and that they may exceed the severity of droughts of historical experience, such as the drought of the late 1990s and early 2000s. Data from the gaging station at Lees Ferry, Arizona, represent the best-known Colorado River measured flow record. As flow data accumulated over time at Lees Ferry, it became clear that 1905-1920—the period upon which Colorado River allocations were ascribed—was significantly wetter than average. It has also become evident that the river’s average annual flow is considerably less than the approximately 16.4 million acre-feet figure used by Colorado River Compact
OCR for page 111
Colorado River Basin Water Management: Evaluating and Adjusting to Hydroclimatic Variability negotiators. For many years the 20th century gage record of Colorado River flows represented the best understanding of the river’s year-to-year hydrologic variability. Despite the value of data from these gage records—especially from sites that have accumulated data for several decades—support for the USGS system of stream gages has not always been steady and has seen some past periods of decline. Today, science-based knowledge of the river’s flows and the basin’s climate systems has become more sophisticated. Nevertheless, the gage record of river flows will remain an important source of information for scientists and water resources planners. Measured values of streamflow of the Colorado River and its tributaries provide essential information for sound water management decisions. Loss of continuity in this gaged record would greatly diminish the overall value of the existing flow data set, and once such data are lost they cannot be regained. The executive and legislative branches of the U.S. federal government should cooperate to ensure that resources are available for the USGS to maintain the Colorado River basin gaging system and, where possible, expand it.
Representative terms from entire chapter: