Figure 25. Idealized relationship between northern Rockies snowpack and streamflow anomalies with associated Pacific SSTs, atmospheric circulation, and surface feedbacks. Source: Pederson et al. submitted.
Central U.S. Rocky Mountains and the Greater Yellowstone Area
Temperatures for the CR-GYA have increased 1–2C (2–4°F) during the last century, with the greatest increases occurring in the second half of the 20th century (Vose et al. 2005; Bonfils et al. 2008; Mote 2006, 2003). This rate of increase is slightly higher than in the Southwest and slightly lower than in the northern Rockies, following a pattern of more pronounced temperature increases at higher latitudes (Cayan et al 2001; Ray et al. 2008). Increasing winter and spring temperatures have resulted in reduced snowpack, earlier spring snowmelt and peak flows, and, in some cases, lower summer flows for major basins (Mote 2006; McCabe and Clark 2005; Stewart et al. 2004; Hidalgo et al. 2008).
CR-GYA precipitation records show highly variable patterns across gradients in elevation, latitude, and longitude. No long-term trends are evident over the last century; reconstructions of hydrology from tree-ring records indicate interannual, decadal, and multidecadal variation (fig. 26; Watson et al. 2009; Graumlich et al. 2003). A greater proportion of precipitation is likely falling as rain rather than snow in this region but the impacts are less pronounced than in other parts of the western United States (Knowles et al. 2006). In many parts of the CR-GYA, the 1930s and 1950s were significantly drier than average and the 1940s wetter, although sub-regional variation is high, likely because the region is located between atmospheric circulation patterns, as discussed further below (Watson et al. 2009; Gray et al. 2007, 2004, 2003; Graumlich 2003).
Figure 26. (a) Observed annual (previous July through current year June) precipitation for Wyoming Climate Division 1 (gray line) compared to precipitation estimates based on the stepwise regression model (black line). (b) The full stepwise version of reconstructed annual precipitation (black line) for 1173 to 1998. The horizontal line (solid gray) near 400 mm represents the series mean, and the vertical line (dotted gray) at 1258 divides the well-replicated portion of the record from reconstructed values in the earlier, less replicated years. Source: Gray et al. 2007; reprinted with permission.
The influence of ocean-atmosphere interactions on decadal, multidecadal, and interannual variation in climatic conditions is more spatially variable in the CR-GYA than in the other regions because it falls in a transition area between northwestern and southwestern U.S. circulation patterns (Gray et al. 2007, 2004; Graumlich et al. 2003), where variations in ocean-atmosphere interactions, topography, latitude, and longitude often result in opposite trends in climatic conditions at sites within the same region (Gray et al. 2004).
High-elevation snow basins within the western GYA typically respond to large-scale climate forcing in a manner similar to that of the Pacific Northwest, where the cool-phase PDO results in cool, wetter than average winters and the warm-phase PDO brings warmer and drier than average winters (Gray et al. 2007, 2004; Graumlich et al. 2003; Dettinger et al. 1998). Similar to the Pacific Northwest, these areas experience increased precipitation during La Niña events and decreased precipitation during El Niño events, and the ENSO seems to be linked to the magnitude of PDO anomalies, especially during the winter (Gray et al. 2007). Alternatively, lower elevation sites and eastern portions of the GYA respond more like the Southwest or show a variable response to ENSO that depends heavily on the strength of event and interactions with other climate drivers (Gray et al. 2004; McCabe et al. 2007). Years with strong El Niño SSTs have increased winter precipitation and La Niña events bring drier conditions. This difference between high and low elevation precipitation regimes is common throughout the central Rockies, complicating predictions of future precipitation in the region (McCabe et al. 2007).
Southern U.S. Rocky Mountains
Temperatures have increased 0.5–1˚C ()0.9–1.8°F) throughout the southern Rockies during the last 30 years. The north central mountains of Colorado warmed the most (~1.35˚C, 2.43°F) and high elevations may be warming more quickly than lower elevations in some regions (Pepin and Lundquist 2008; Diaz and Eischeid 2007). Warming is evident at almost all locations, but temperatures have increased the most in the north central mountains and the least in the San Juan Mountains of southwestern Colorado (Ray et al. 2008). Only the Arkansas River Valley in southeastern Colorado shows a slight cooling trend during the 20th century; no trend is evident in this area for the second half of the century (Ray et al. 2008).
Precipitation records for the southern Rockies for the last century indicate highly variable annual amounts and no long-term trends (Ray et al. 2008; Dettinger 2005). Like elsewhere in the interior West, a greater proportion of precipitation is falling as rain rather than snow than in the past, but these changes are less pronounced than in the northern Rockies (Knowles et al. 2006). Decadal variability is evident in records of precipitation and surface flows and is linked to variability in ocean-atmosphere and land-surface interactions (fig. 27; Stewart et al. 2005).
Figure 27. Observed time series (1895–2007) of annually averaged precipitation departures area-averaged over the Upper Colorado drainage basin (top) and annual Colorado River natural flow departures at Lees Ferry in million acre-feet (bottom). The precipitation data are based on 4-km gridded PRISM data. Colorado River natural flow data from the Bureau of Reclamation. (Source: Ray et al. 2008; reprinted with permission.)
Similar to trends evident throughout the interior West, more precipitation is falling as rain than snow in the southern Rockies, spring snowpack is decreasing, especially at elevations below 2500 meters, and peak streamflows are occurring earlier because of warmer spring temperatures (Knowles et al. 2006; Bales et al. 2006; Stewart et al. 2005; Hamlet et al. 2005; Clow 2007; Mote 2006, 2003). Summer flows are typically lower and annual flows show high variability but no significant trends in most locations (Ray et al. 2008).
Like the CR-GYA, the Colorado River Basin spans a transition area where the influence of Pacific Northwest and southwestern circulation patterns show opposite trends (Gray et al. 2007; Clark et al. 2001). During El Niño years, northern parts of this region experience drier than average conditions while the southern portions experience wetter than average conditions. The opposite conditions occur during La Niña years, and anomalies tend to be more pronounced in spring in southern portions. Long-term droughts are linked more closely to low-frequency oscillations in PDO and AMO, and are most commonly associated with the interaction between a cool-phase PDO and warm-phase AMO (McCabe et al. 2004).
Upper Columbia Basin
For most of the Upper Columbia Basin, average annual temperatures increased 0.7–0.8ºC from 1920 to 2003, and the warmest decade was the 1990s (fig. 28; Littell et al. 2009b). Average temperatures have increased as much as 2ºC (4°F) in parts of the region, and increases have been more pronounced at higher elevations (Mote 2006, 2003). During the mid-20th century, average and daily minimum temperatures increased more in the winter and spring than in other seasons and more than maximum temperatures. During the second half of the 20th century, minimum and maximum temperatures increased at similar rates (Watson et al. 2009; Dettinger et al. 1994; Karl et al. 1993: Lettenmaier et al. 1994: Littell et al. 2009b).
Figure 28. Trends in average annual Pacific Northwest temperature, 1920–2000. Increases (decreases) are indicated with red (blue) (red) dots, and the size of the dot corresponds to the magnitude of the change. Pluses and minuses indicate increases or decreases, respectively, that are less than the given scale. Source: Climate Impacts Group, University of Washington.
Upper Columbia Basin precipitation trends are less clear than temperature trends, and observations indicate high decadal variability. Precipitation increased 14% for the entire northwestern United States, (1930–1995), and increases ranged between 13% and 38% within the region (fig. 29; Mote 2003), but these trends are often not statistically significant, depending on the area and time interval measured. Similar to much of the interior West, variability in winter precipitation has increased since 1973 (Hamlet and Lettenmaier 2007).
Figure 29. Trends in average annual precipitation, 1920–2000. Increases (decreases) are indicated with blue (red) dots, and the size of the dot corresponds to the magnitude of change. Source: Climate Impacts Group, University of Washington.
Spring snowpack and SWE declined throughout the Upper Columbia Basin in the second half of the 20th century. The decline was most pronounced at low and mid-elevations, and declines of more than 40% were recorded for some parts of the region (fig. 30; Hamlet et al. 2005, Mote 2006, 2003). Declines in snowpack and SWE are associated with increased temperatures and declines in precipitation during the same period (Mote et al. 2005; Mote 2003). The timing of peak runoff shifted 2–3 weeks earlier for much of the region during the second half of the 20th century (Stewart et al. 2004), and the greatest shifts occurred in the mountain plateaus of Washington, Oregon, and western Idaho (Hamlet et al. 2007). Because Upper Columbia Basin ecosystems rely on the release of moisture from snowpack, these shifts are significantly impacting plant species, causing some to bloom and leaf out earlier in the spring (Mote et al. 2005; Cayan et al. 2001; Schwartz and Reiter 2000).
Figure 30. Trends in April 1 SWE, 1950–2000. Increases (decreases) are indicated with blue (red) dots, and the size of the dot corresponds to the magnitude of the change. Pluses and minuses indicate increases or decreases, respectively, that are less than the given scale. Source: Climate Impacts Group, University of Washington.
Variations in Upper Columbia Basin climatic conditions are related to ocean-atmosphere and land-surface interactions, namely the ENSO and PDO phenomena. In their warm phases, both the ENSO and PDO increase the chance for a warmer winter and spring in the Upper Columbia Basin and decrease the chance that winter precipitation will reach historical averages. The opposite tendencies are true during a cool-phase ENSO and PDO: they increase the odds that Upper Columbia Basin winters will be cooler and wetter than average (Clark et al. 2001). While strong El Niño years are typically warmer than average, SWE anomalies are often less pronounced then, and winter precipitations are commonly close to historical averages (Clark et al. 2001). Clark et al. (2001) suggested that El Niño circulation anomalies are centered more in the interior West than are La Niña circulation anomalies and are most evident in mid-winter.
What can we learn from 20th century observations?
Small changes can have large impacts
Changes in the distribution of minimum temperatures and frost-free days illustrate how small changes in temperature (1–2ºC, 2–4°F) may result in large changes to surface hydrology (Barnett et al. 2004, 2005) as they contribute to earlier melt-off and diminished spring snowpack (Pederson et al. 2010, 2009; Mote 2006; Barnett et al. 2008; Stewart et al. 2004; McCabe and Clark 2005), increases in the proportion of winter precipitation as rain rather than snow (Knowles et al. 2006; Bales et al. 2006), decreased snow season length at most elevations (Bales et al. 2006), and lower summer flows (Barnett et al. 2008). Evidence from a number of studies suggests that even small temperature increases can have dramatic impacts on water availability for much of the western United States. Along with changes in snowpack and earlier spring runoff, the predicted temperature increases will likely contribute to increased drought severity, duration, and frequency (fig. 31; Hoerling and Eischeid 2007; Barnett and Pierce 2009). Increased winter precipitation predicted for central and northern regions of the study area will likely be inadequate to offset the increased evaporation and plant water use driven by rising temperatures (Gray and McCabe 2010; Hoerling and Eischeid 2007; Seager et al. 2007).
Figure 31. Modeled changes in annual mean precipitation minus evaporation (P–E) over the Southwest (125°W to 95°W and 25°N to 40°N, land areas only), averaged over ensemble members for each of the 19 climate models participating in the Fourth Assessment Report (AR4) of the IPCC (Solomon et al. 2007). The historical period used known and estimated climate forcings; the projections used the SResA1B emissions scenario. The median (red line) and 25th and 75th percentiles (pink shading) of the P–E distribution among the 19 models are shown, as are the ensemble medians of P (blue line) and E (green line) for the period common to all models (1900–2098). Anomalies (Anom) for each model are relative to that model's climatology from 1950 to 2000. Results have been 6-year, low-pass, Butterworth-filtered to emphasize low-frequency variability that is of most consequence for water resources. The model ensemble mean P–E in this region is ca. 0.3 mm/day. Source: Seager et al. 2007; reprinted with permission.
How do we know if observed changes are related to human-caused climate change?
The Intergovernmental Panel on Climate Change (IPCC, Solomon et al. 2007) included studies to determine whether a detected climate change is significantly different from natural variations of the climate system. Attribution studies seek to establish the principal causes for observed climate phenomena, including trends in temperature and extreme climate events, and whether they are related to human activities. In order to attribute a detected change, scientists must demonstrate that the change is consistent with an identified anthropogenic cause and inconsistent with any alternative, physically plausible explanation that excludes anthropogenic causes (Houghton et al. 2001). If attribution is established, the IPCC may assign a likelihood for the probability that the identified cause resulted in the observed conditions.
Attribution studies use empirical analyses of past climate relationships and evaluate cause-and-effect relationships with climate models. Model simulations of past climate are compared with the observed record using statistical analysis, including estimates of natural variability and trends from climate models, historical observations, or paleoclimate reconstructions. “Fingerprint” methods seek the unique signature of climate change by simultaneously looking at changes in many variables. In studies conducted to determine the cause of the observed warming of temperatures in western and northern North America over the last half-century, annually averaged North American surface temperatures from 1950 to 2007 were computed using the IPCC (CMIP3) models forced with the observed record of greenhouse gases, volcanic aerosols, and solar forcing during 1950 to 1999 and subsequently (2000–2007) with the A1B scenario of greenhouse gas emissions. (The A1B scenario = rapid economic growth, global population that peaks in mid-century and declines thereafter, and the rapid introduction of new and more efficient technologies). Comparison of these modeled temperatures (fig. 1, top panel) with observations (fig. 19) suggests that anthropogenic greenhouse gas emissions have contributed about 1°C (1.8°F) of the observed warming in the last 30 years. Similarities between the modeled climate and the observed trends provide the best available evidence for external (rather than natural variability) forcing of surface temperature change by anthropogenic greenhouse gases because the bulk of the warming occurs after about 1970 in both time series and the externally forced warming of about 1°C (1.8°F) since 1950 is close to the observed rate.
Figure 1. Annual average North American surface temperature (1950–2007) from 22 IPCC model simulations with greenhouse gas, aerosol, solar, and volcanic forcing from 1950 to 1999, and the A1B emissions scenario from 2000 to 2007 (top panel). Annual anomaly values of surface temperature averaged over the whole of North America (compared to 1971–2000 average, lower panel). The smoothed curve highlights multi-year variations. (Source: Ray et al. 2008, adapted from CCSP 1.3, figure 3.3; reprinted with permission.)
A series of recent studies sought to detect and attribute climate change in the western United States (Bonfils et al. 2008; Pierce et al. 2008; Hidalgo et al. 2009; Das et al. 2009) using the same downscaled projections and PRISM data as the westwide projections shown below (figs. 35, 36). Bonfils and colleagues conducted a detailed analysis of models thought to best simulate the climate of the western United States, and using these models, found that natural variability is insufficient to explain the increase in daily minimum and maximum temperatures, a sharp decline in frost days, a rise in degree days above 0°C (32°F), and a decline in snowpack at low and mid-elevations. They ruled out solar variability and volcanic forcing as a cause. They found that the anthropogenic signal is detectable by the mid-1980s in a signal-noise ratio of minimum temperature. Other attribution papers focus on streamflow (Hidalgo et al. 2009), snowpack (Pierce et al. 2008), and the structure and detectability of hydrological variables (Das et al. 2009). These studies have estimated that up to about half of the trends in temperature and associated hydrologic variables can be attributed to anthropogenic causes (Barnett et al. 2008; Pierce et al. 2008).
Shifting distributions and new norms
Many parts of the study region are vulnerable to small changes in temperature because the overall climate is arid to semi-arid to begin with, and the water available in these areas depends heavily on the mountain snowpack dynamics (Gray and Anderson 2010). While ecosystems are adapted to natural variations in water availability, a shift in drought frequency and magnitude, or even the occurrence of an especially severe and prolonged dry event, could result in regional ecosystems reaching a tipping point whereby a major redistribution of vegetative communities ensues (Gray et al. 2006; Jackson et al. 2009b). Rapid changes in surface hydrology related to altered snowpack dynamics could bring similar impacts to aquatic ecosystems that have developed under distinctly different hydroclimatic regimes (fig. 32).
Figure 32. Relationships between climate change, coping range, vulnerability thresholds, and adaptation. Idealized version of a coping range, showing the relationship between climate change and threshold exceedance and how adaptation can establish a new critical threshold, reducing vulnerability to climate change (modified from Jones and Mearns 2005). Source: IPCC AR4 WGI 2007; reprinted with permission.
Small increases in temperature (e.g., 1–2ºC, 2–4°F) will result in greater evaporative losses from lakes, streams, wetlands and terrestrial ecosystems, and it is likely that this enhanced evaporation will lead to significant ecosystem and water management impacts (Arnell 1999; Gray and Andersen 2010). In the central and southern portions of the study area, the increases in winter precipitation predicted by some models are not expected to offset this increased evaporation and transpiration. For example, models from Gray and McCabe (2010) estimate a 15–25% decrease in average Yellowstone River flows from a 1.5–3ºC (2.7–5.4°F) temperature increase, and it would require the equivalent of the wettest years in the last millennium to offset the impacts of increased evapotranspiration on this system (Gray and McCabe 2010). As seen in many recent observational records, seemingly small changes in mean conditions can lead to an increased frequency of hot weather relative to historical conditions and extreme precipitation events (Karl et al. 2009; Solomon et al. 2007; Groisman et al. 2005; Kunkel et al. 2003; Madsen and Figdor 2007; fig. 33). Multiple assessments point to a potential shift in precipitation such that storms will become more intense but less frequent (Groisman et al. 2005; Kunkel et al. 2003; Madsen and Figdor 2007). This, in turn, would increase the number of dry days between precipitation events, while also altering runoff, infiltration, and erosion rates.
Figure 33. Schematic for a normal temperature distribution showing the effect on extreme temperatures when the mean temperature increases. Source: IPCC AR4 WGI 2007; reprinted with permission.
What can we expect in the future?
Many of the trends in climate evident in the last century are expected to continue in the future. Projections shown in this report are based on the global climate model (GCM) projections done for the IPCC (Solomon et al. 2007), a coordinated large set of climate model runs known as the Coupled Model Intercomparison Project, Phase 3 (CMIP3), performed at modeling centers worldwide using 22 global climate models. Output of most of these models is at large resolution, often a 200-kilometer grid. Although it was common in past climate impact studies to present the results of only one or two global climate models, research now suggests that the average of multiple models provides a better approximation, and use of ensembles is made possible by increasing computing capacity and technical abilities for analyzing multiple model simulations (Salathé et al. 2010). The CMIP3 models and knowledge from these comparisons are the current state of the art in climate modeling and assessments.
GCM projections for North America
These global models project broad-scale increases in temperature in North America through the mid-21st century. Projected changes compared to a recent baseline (1950–1999 average) through mid-century (2040–2060 average) are shown in figure 34. For much of the interior western United States, the multi-model average projects an annual mean warming of about 2°C (4°F, in orange) by 2050. Individual global models also show a broad-scale pattern of warming, though of different magnitudes across models. The range of individual GCM projections (10th and 90th percentiles of the model projections) is from about +2.5F to +5.5°F (1.4–3.1°C). GCM projections show summers warming by about +5°F (range: 3–7°F, 2–4°C) and winters by about +3°F (range: 2–5°F, 1–3°C) (fig. 34, top row). The multi-model average and many individual global models show less warming within several hundred kilometers of the Pacific coast. This feature may be a result of the inability of the global models to simulate the effects of the coastal mountain ranges and hence the moderating coastal influence penetrates too far inland. Regional climate modeling studies corroborate this (Salathé et al. 2010), showing large values of summertime warming much closer to the coast than for the global models.