لخّصلي

Trends in both the magnitude and timing of streamflow are of principal interest to water resource managers (Lins and Slack, 1999; McCabe and Wolock, 2002). The magnitude of runoff is important to assess water availability and reservoir storage; timing of runoff is important to assess flood control regulations, hydropower generation, and irrigation demands. Due to increasing temperatures, research has shown a trend toward earlier spring runoff in both observed data (e.g., Cayan et al., 2001; Mauget, 2003; Regonda et al., 2005; Stewart et al., 2005; Kalra et al., 2008; Miller and Piechota, 2008) and modeled data (e.g., Hamlet et al., 2005, 2007; Maurer and Duffy, 2005). Miller and Piechota (2008) have previously shown a shift in the timing of naturalized flow over the Colorado River Basin (Figure 1). Trends in streamflow have decreased during the traditional peak runoff season (April through July) and increased in fall and winter months. Miller and Piechota (2008) hypothesized that increasing streamflow trends in the fall and winter were due to more frequent winter rain events and less frequent snow events; less snow events may contribute to decreased mountain snowpack and may result in reduced spring snowmelt runoff.

Details are in the caption following the image
Figure 1
Open in figure viewer
PowerPoint
The Colorado River Basin Is Divided Into the Upper and Lower Colorado Regions Within the U.S. The Colorado River and major tributaries, along with delineated major subbasins, are also indicated. The Dolores and Dirty Devil River Basins are not considered in this study because the Dolores River Basin is not a headwater river basin, and available data within the Dirty Devil River Basin are limited. For reference, Denver, Colorado is located approximately at 39.75°N, 104.87°W (1 kilometer is approximately 0.6 miles).

Trends in precipitation and snowpack characteristics have been the subject of interest to water resource managers, particularly with regards to the snowmelt driven hydrology of the Upper Colorado River Basin. Decreasing trends in snow water equivalent (SWE) have been noted in the Western United States (U.S.) and Colorado River Basin (e.g., Feng and Hu, 2007; Kalra et al., 2008). More interestingly, decreasing trends in SWE have correlated with the changing character of precipitation; that is, changes in the frequency and magnitude of rainfall and snowfall events (Trenberth et al., 2003). Knowles et al. (2006) noted a reduction in the ratio between the winter SWE and total winter precipitation between water year 1949 and 2004 that correlated with changing temperature trends over the Western U.S. Knowles et al. (2006) further found the largest changes to winter precipitation typically occurred in March, supporting other studies indicating a shift in changing character of precipitation (e.g., Serreze et al., 1999; Mote, 2003, 2006; Mote et al., 2005).

It is clear that streamflow and snowpack are vitally important to water resource availability in the Western U.S., particularly in snowmelt driven basins such as the Colorado River Basin. Research has indicated that climate change may significantly impact snowpack and streamflow in snowmelt dominated basins (e.g., Cooley, 1990; Salathé, 2005; Parry and Intergovernmental Panel on Climate Change, Working Group II, 2007) and are indicative of drought and arid conditions in the American Southwest (Seager et al., 2007; Timilsena and Piechota, 2008). Water supply forecasts within the Colorado River Basin are driven by observed snowpack conditions and historical hydroclimatic conditions over the 1971-2000 time period. As the character of precipitation changes in response to climate change, the past may no longer be representative of the future and, as such, historical observations of hydroclimatology may not be as useful to water supply forecasters. It is important then to understand the changing relationship between precipitation and streamflow trends.

While current research has focused on the identification of seasonal trends in either observed streamflow or snowpack, there has been significantly less investigation of the impact of observed trends that are coincident in both daily SWE and streamflow, particularly over the Colorado River Basin. Presumably, this is due to the relatively short period of record of daily snowpack observations in the mountainous Western U.S. In this study, trends in daily measurements of SWE in the Western U.S. are investigated. Furthermore, these trends are compared to observed streamflow trends over the Colorado River Basin in an attempt to quantify the impact of changing precipitation characteristics to streamflow in the basin and to improve understanding of the linkage between snowpack and streamflow over the basin. In this paper, trends in precipitation from 398 snowpack telemetry (SNOTEL) stations spanning the Western U.S. are investigated. Then, trends in USGS streamflow stations spanning the Colorado River Basin are then considered. Finally, corresponding trends in streamflow over Colorado River headwater basins are related to regional SNOTEL stations within the headwater basins to better understand the linkages between SWE and streamflow.

Data
Precipitation and Snow Water Equivalent
Daily snowpack and precipitation observations are obtained from the Natural Resource Conservation Service (NRCS) SNOTEL network. NRCS SNOTEL stations subjected to analysis span the western U.S. and Alaska (Figure 2a). For inclusion in this study, SNOTEL stations are required to have a period of record of at least 20 years, and at least 50% complete over any given year. These data requirements are slightly more stringent than those described in Huntington et al. (2004) and Knowles et al. (2006).

Details are in the caption following the image
Figure 2
Open in figure viewer
PowerPoint
The Spatial Distribution of (a) SNOTEL Gages Over the Western U.S. and (b) USGS HCDN and Reclamation Natural Flow Stations Over the Colorado River Basin Included in This Study. It is important to note that analyses over SNOTEL gages in Alaska are considered, though they are not pictured.

Currently, the NRCS operates 761 SNOTEL stations in 13 states, the farthest east being located in South Dakota and the farthest west located in Alaska. Of the 761 total SNOTEL stations, 398 stations met the aforementioned completeness criteria to be included in this study. With the exception of the Bettles Field Station located in Alaska, the data for each station go through the end water year 2008. The Bettles Field Station contains 25 years of data spanning 1980-2005. Of the stations considered in this study, the least amount of years considered was 21, spanning water years 1987 through 2008. The oldest station considered spans 45 water years (1963-2008). Figure 3 presents a histogram of the period of record for the stations considered here. The majority of stations span 25-30 water years.

Details are in the caption following the image
Figure 3
Open in figure viewer
PowerPoint
The Period of Record for Each Station Is Variable. With the exception of the Bettles Field Station in Alaska, the data for each station considered in this study span through water year 2008.

Of the stations included in this study, the furthest east is located in southern Utah, and the furthest west is located in central Alaska. Of interest, studies have noted a relationship between elevation and snowpack. Mote (2003) found that below 1,800 meters (5,900 feet), declining SWE observations coincide with increasing temperature trends. Of the stations included in this study, the lowest is located at 144 meters (375 feet) in Alaska, and the highest is located at 3,536 meters (11,600 feet) in Colorado. In the continental U.S., the lowest station is located at 792 meters (2,600 feet) in Oregon.

Streamflow
Daily streamflow data from the USGS are used in this study, provided the gage is part of the Hydro-Climatic Data Network (HCDN) and located within the Colorado River Basin (Figure 2b). USGS HCDN gages are selected because they are predominantly free of anthropogenic influence and have a length of record >30 years (Slack et al., 1993). It is important to note that while these gages are considered to be predominantly free of anthropogenic influence, it does not mean that flow measured at these gages are completely free of anthropogenic influence. Water use over the Colorado River Basin has changed appreciably over the last 30 years; as such, it is acknowledged that there is some degree of anthropogenic influence within the HCDN gage record. When investigating seasonal trends in streamflow, it is recommended by the authors that naturalized flow published by the Department of the Interior, U.S. Bureau of Reclamation (Reclamation) be used (Prairie and Callejo, 2005).

Data Limitations
Data collected from the NRCS and the USGS are at the daily timestep. Daily gage data are considered from the inception of the gage through water year 2008. For USGS data, the gage record, on average, extends back to the 1930s. The shortest record begins in 1967 at the Mogollon Creek near the Cliff, New Mexico station. The longest record considered in this study extends back to 1894 at the Green River near the Green River, Utah station. Although each gage considered in this study is part of the USGS HCDN, the gage record utilized in this study may extend past or previous to the record described in the USGS HCDN.

Although the earliest SNOTEL record considered in this study begins in October of 1963, most SNOTEL records do not begin prior to the mid to late 1970s. The average first year of each SNOTEL station considered in this study is 1981. It is acknowledged that the length of this record is relatively limited compared to snow records used in other studies (e.g., Knowles et al., 2006), and that it is often desirable to have longer periods of record to evaluate statistically significant trends. One of the main goals of this study is to examine the correspondence between daily SWE and streamflow over the Colorado River Basin; as such, this precludes the use of longer SWE records such as those measured at snow course sites. Whereas snow course sites tend to be more representative of regional snowpack, SWE observations from SNOTEL sites correlate well with those observations made at snow course sites without consistent bias (Dressler et al., 2006).

Through the use of point observations at USGS and SNOTEL gaging stations, it is the intent of this research to illustrate trends over the Western U.S. and Colorado River Basin. While the temporal scale over which each gage is not uniform, the intent is not to define regional trends over any specific time period; rather, the intent is to present trends such that snowpack and streamflow data may be compared over the available data record.

Methodology
Trends in hydroclimatic observations from SNOTEL and USGS stations are investigated using Kendall’s tau (τ) nonparametric test for monotonic trend with a correction for ties. Kendall’s τ is well-suited for applications to water resources, as it is a rank-based procedure that is resistant to outliers in time series (Helsel and Hirsch, 1992). Kendall’s τ-test has successfully been used in previous research investigating the trends in precipitation and streamflow observations (e.g., Huntington et al., 2004; Rood et al., 2005; Knowles et al., 2006). The significance of monotonic trends detected in SWE observations are calculated through comparison of the Kendall’s τ-test statistic to a standard two-sided Student’s t table. For completeness and to ensure the accuracy and applicability of Kendall’s τ to the collected observations, Spearman’s rho (ρ) test for monotonic trend is also applied to hydroclimatic time series investigated in this study. Like Kendall’s τ, Spearman’s ρ is a rank-based statistical test; however, Spearman’s ρ weights the magnitude of differences in time series ordinates more heavily. Essentially, Kendall’s τ and Spearman’s ρ measure the same correlation at different scales of magnitude (Helsel and Hirsch, 1992).

The interaction between changing snowpack characteristics and streamflow is investigated over headwater basins within the Colorado River Basin. That is, the interdependency of snowpack and streamflow is examined such that observations from SNOTEL stations are compared only to streamflow within the same geographic headwater basin within the Colorado River Basin. Trend tests utilized in this study and associated equations are described and presented more fully in Miller and Piechota (2008) and Maidment (1993), among others.

Trends in Western U.S. Precipitation
Daily SNOTEL Trends
Trends in cumulative water year precipitation are investigated using Kendall’s τ over standardized observations. Miller and Piechota (2008) previously noted the lack of statistically significant precipitation trends over the Colorado River Basin using climate division data. The vast majority of stations (342 stations or 86%) exhibit a decreasing linear trend in water year precipitation over a broad elevation range (Figure 4). Due to the short length of the SNOTEL record, the presence of persistent drought over the majority of most stations’ records, and temporal variability over the gage records, there is a lack of statistical significance in the daily records of cumulative water year precipitation. Cumulative precipitation trends over the Western U.S. are similar to those observed over the Colorado River Basin (Table 1). The average change in cumulative water year precipitation over the basin is approximately −3.6 millimeters (−0.14 inches) per year. This decreasing trend in cumulative precipitation does not appear to correlate with the length of a station’s period of record Figure 5.

Details are in the caption following the image
Figure 4
Open in figure viewer
PowerPoint
For Each of the 398 SNOTEL Stations Included in This Study, the Linear Trend in Cumulative Water Year Precipitation Over the Station’s Period of Record Is Plotted Against the Elevation at the Station Location. Dark circles are indicative of a gage located within the Colorado River Basin; open circles indicate a station outside of the Colorado River Basin (1 millimeter is approximately 0.04 inches).

Table 1. SNOTEL Station Results for the Entire Western U.S. and Colorado River Basin.
SNOTEL Station Characteristics Western U.S. Colorado River Basin
398 Stations 79 Stations
Decreasing water year precipitation 342 (86) 69 (87)
Increasing water year precipitation 56 (14) 10 (13)
Decreasing peak SWE/earlier peak 227 (57) 46 (58)
Decreasing peak SWE/later peak 69 (17) 18 (23)
Increasing peak SWE/earlier peak 57 (14) 10 (13)
Increasing peak SWE/later peak 45 (11) 4 (5)
Earlier snow/earlier melt 179 (45) 41 (52)
Later snow/earlier melt 119 (30) 30 (38)
Earlier snow/later melt 68 (17) 5 (6)
Later snow/later melt 32 (8) 3 (4)
Notes: The number of stations exhibiting a particular behavior is presented with the percentage with respect to the total number of gages presented in parenthesis. SWE, snow water equivalent.
Details are in the caption following the image
Figure 5
Open in figure viewer
PowerPoint
For Each of the 398 SNOTEL Stations Included in This Study, the Linear Trend in Cumulative Water Year Precipitation Over the Station’s Period of Record Is Plotted Against the Duration of Each Station’s Record. Circles are as described in Figure 4.

Accurate trend analysis is dependent on data completeness, and research has indicated that initial values within a dataset may influence trend test results. Using rank-based nonparametric tests, such as Kendall’s τ and Spearman’s ρ, mitigates the impact of outlying values at the beginning of a time series. Nonparametric trend tests are more robust against outliers and departures from normal (Villarini et al., 2011). Table 2 presents the results of trend analysis performed over water year precipitation data collected from SNOTEL stations located within the state of Colorado. It is interesting to note that of this subset of data, the range of trend magnitude varies between gages with identical periods of record. This suggests that over the state of Colorado, trend results may be more dependent on the geographic location of the SNOTEL station, rather than any potentially outlying observations at the beginning of the time series.

Table 2. SNOTEL Sites Utilized in This Study Typically Contained Very Complete Datasets.
SNOTEL Site Name Site ID Latitude Longitude Elevation (m) Cumm. Precip. Trend (mm) Completeness (%) Length of Record (years)
Columbine Pass 08L02S 38°25′ −108°22′ 15 0.29 99.73 22
Crosho 07J04S 40°10′ −107°3′ 373 0.04 99.73 22
Mesa Lakes 08K04S 39°3′ −108°3′ 114 −0.50 99.73 22
Rabbit Ears 06J09S 40°22′ −106°44′ 533 3.29 99.71 22
Ripple Creek 07J05S 40°7′ −107°17′ 2,781 −4.29 99.70 22
Spud Mountain 07M11S 37°41′ −107°46′ 2,195 0.81 99.73 22
Stump Lakes 07M34S 37°28′ −107°37′ 2,749 −7.81 99.73 22
Vallecito 07M31S 37°29′ −107°30′ 2,591 −4.74 99.73 22
Upper Rio Grande 07M16S 37°43′ −107°15′ 2,329 −4.01 99.73 22
Wolf Creek Summit 06M17S 37°28′ −106°48′ 2,966 −1.52 99.73 22
Bison Lake 07K12S 39°45′ −107°21′ 30 0.79 99.73 23
Nast Lake 06K06S 39°17′ −106°36′ 15 −0.17 99.73 23
North Lost Trail 07K01S 39°4′ −107°9′ 259 2.87 99.73 23
Schofield Pass 07K11S 39°1′ −107°3′ 2,560 0.16 99.73 23
Beartown 07M32S 37°42′ −107°30′ 15 −4.20 99.73 26
Butte 06L11S 38°53′ −106°57′ 76 −3.66 99.71 27
Dry Lake 06J01S 40°32′ −106°46′ 6 −1.29 99.73 28
Independence Pass 06K04S 39°4′ −106°36′ 9 −2.55 99.73 28
Lynx Pass 06J06S 40°5′ −106°40′ 427 −3.77 99.07 28
Phantom Valley 05J04S 40°23′ −105°50′ 869 −2.69 99.73 28
Red Mountain Pass 07M33S 37°53′ −107°42′ 2,225 −4.98 99.73 28
Bear Lake 05J39S 40°18′ −105°38′ 945 −4.97 99.69 28
Niwot 05J42S 40°2′ −105°32′ 500 −5.69 99.73 28
Roach 06J12S 40°52′ −106°3′ 2,435 −3.14 99.73 28
Whiskey Creek 05M14S 37°12′ −105°7′ 2,591 −6.03 99.73 28
Culebra #2 05M03S 37°12′ −105°11′ 76 −7.78 99.71 29
Tower 06J29S 40°32′ −106°40′ 2,326 −1.36 99.73 29
Arrow 05K06S 39°54′ −105°45′ 853 −4.67 99.06 30
Cascade 07M05S 37°39′ −107°48′ 91 −3.81 99.00 30
Copper Mountain 06K24S 39°29′ −106°10′ 9 0.23 99.73 30
Elk River 06J15S 40°50′ −106°58′ 8 −0.86 99.73 30
Mineral Creek 07M14S 37°50′ −107°43′ 15 −5.94 96.40 30
Park Reservoir 07K06S 39°2′ −107°52′ 573 −7.11 98.94 30
Porphyry Creek 06L03S 38°29′ −106°20′ 808 −4.63 99.73 30
Upper San Juan 06M03S 37°29′ −106°50′ 2,804 −11.19 99.71 30
Vail Mountain 06K39S 39°37′ −106°22′ 2,286 −1.82 99.73 30
Willow Creek Pass 06J05S 40°20′ −106°6′ 2,417 −1.26 99.73 30
Berthoud Summit 05K14S 39°48′ −105°46′ 381 −0.80 99.71 30
Deadman Hill 05J06S 40°48′ −105°46′ 305 −2.66 99.73 30
Joe Wright 05J37S 40°31′ −105°53′ 9 −4.50 99.71 30
Lake Eldora 05J41S 39°56′ −105°35′ 168 2.27 99.73 30
Notes: This table presents the results of trend analysis performed over cumulative water year precipitation grouped by length of record. Precipitation trends are varied within groups, and trends may be more a function of geographic location.
For the stations selected for this study, daily data reported regularly and consistently; most of the stations considered had 0.0 millimeters indicates the presence of snowpack at a particular station; as such, the first indication of snowpack during the water year is interpreted from reported SNOTEL measurements. Similarly, the first day after April 1 when a reported SWE measurement of 0.0 millimeters is observed signals the “melt day,” or the end of the snow season during the course of a water year. Additionally, for each water year, the peak SWE and days since the beginning of the water year (October 1) to reach peak SWE are investigated for linear trends. Table 1 summarizes the results of this analysis. Most stations tend to show declining peak SWE observations and occur at an earlier date in the water year. Table 1 also summarizes the amount of stations that exhibit earlier or later initial starts to the snow season and those stations that exhibit earlier or later melt days.

The length of the snow season at a particular SNOTEL gage is defined as the duration of time, in days, since the first observation of SWE after the beginning of the water year to the first observation of 0.0 millimeters of SWE after April 1 (i.e., the melt day). Of the SNOTEL gages included in this study, approximately 60% (238 stations) of the gages exhibited a decreasing linear trend in the length of the snow season (Figure 7). Of those gages located within the Colorado River Basin, 66% (52 stations) exhibited a decreasing linear trend in the length of the snow season. The Cascade station in Colorado lost approximately 1.4 days of its snow season over the course of its gage record; conversely, the Hams Fork station in southern Wyoming gained approximately 1.3 days to its snow season over the course of its gage record. The median loss to the length of the snow season over each gage in the Colorado River Basin is approximately 0.2 days.

Details are in the caption following the image
Figure 7
Open in figure viewer
PowerPoint
Changes to the Length of the Snow Season Over the Course of the Station Record at Each SNOTEL Station. Dark circles indicate a decrease in the length of the snow season over the period of record for the station. White circles indicate longer snow seasons over the period of record for the station.

Table 1 summarizes the potential shift in timing of the snow season throughout the Western U.S. and Colorado River Basin. Those stations showing earlier starts to the snow season and a later melt day also tend to trend toward higher peak annual SWE; conversely, those showing later starts to the snow season and an earlier melt day also tend to trend toward lower peak annual SWE. Most SNOTEL stations in the Western U.S. (235 or 60%) and including those in the Colorado River Basin (49 or 62%) reporting an earlier melt day also show a trend toward earlier peak annual SWE as well. Half of the 238 stations reporting an earlier melt day also report a later start to the snowpack season.

Trends in Snowfall and Rainfall Frequency
Most SNOTEL stations record SWE and total precipitation daily, regardless of whether that precipitation occurs as snow or rain. The assumption was made that the recording of a precipitation event coupled with an increase to or stationary SWE observation would indicate a snow event, whereas a recording of precipitation coupled with a decrease to the station’s SWE observation indicate a rain or rain-on-snow event. Miller and Piechota (2008) hypothesized that rain events over the Colorado River Basin region has increased due to increasing temperature trends in the basin; in turn, a corresponding decrease in snowfall frequency would also be apparent. However, the results of this study do not confirm that hypothesis with any statistical significance using SNOTEL stations and this methodology. A relatively sparse meteorological monitoring network such as the SNOTEL network may not be dense enough to capture potentially spatially localized rainfall events, particularly over the Colorado River Basin.

Statistically significant seasonal trends in the frequency of rainfall and snowfall events were not apparent. Despite the lack of statistical significance, it may be worth noting that at the annual time scale, moderate increases in rainfall frequency were observed, as approximately 74% of SNOTEL stations showed an increasing trend (67% of SNOTEL stations located in the Colorado River Basin). The average increase in rainfall frequency was approximately 0.1 days per water year. No consistent trends in snowfall frequency were observed throughout the dataset, although decreasing trends were detected in eastern Utah just inside the Lower Green Headwater Basin on the Wasatch Front Range. The Daniels-Strawberry station at the mouth of the Strawberry River showed a decrease of approximately 1.6 days per water year and contributes to flow in the Green River, a major tributary to the Colorado River.

While the results of the current study do not confirm the hypothesis proposed by Miller and Piechota (2008), the results do support those proposed by Huntington et al. (2004) and others regarding hydrologic intensification. The results of the current study do support that the volume of inflow as precipitation over the Western U.S. and Colorado River Basin has decreased over approximately the last 25 years.

Trends in Colorado River Basin Streamflow
USGS HCDN Streamflow Observations
The USGS currently operates 43 stations within the Colorado River Basin that are within the HCDN as described by Slack et al. (1993). It is important to note that while Slack et al. (1993) identified periods of the streamflow record as minimally affected by anthropogenic factors, this study uses the entire period of record at each of these stations. Applying the Kendall’s τ statistical test to daily USGS HCDN time-series data revealed interesting trends throughout the Upper Colorado River Basin (Figure 8). Gages in the northern area of the basin located within the Upper Green, Lower Green, and Yampa subbasins yielded frequent decreasing trends at the 99% confidence interval. However, a small cluster of gages in the Gunnison and northern portion of the San Juan subbasins yielded frequent increasing trends at the 99% confidence interval.

Details are in the caption following the image
Figure 8
Open in figure viewer
PowerPoint
Trend Results and Confidence Intervals From Kendall’s τ Statistical Test Applied Over the Period of Record of Daily Streamflow at Each USGS HCDN Gage.

Daily Streamflow Trends
Daily time series are investigated over the operational record of the gage for linear trends in annual water year flow volume. Of the 43 stations investigated, 29 (67%) exhibit a decreasing trend in water year flow volume. While the magnitude of decreasing volume range between approximately 4,900 cubic meters (4 acre-feet) and 25.040 million cubic meters (20,300 acre-feet), the average decrease in flow relative to each station is approximately 0.3% per year. Over the Colorado River Basin, 34 stations (79%) exhibited decreasing linear trends in April through July runoff. Again, the average decrease in April through July runoff relative to each station is relatively small and is approximately 0.5% per year.

Table 3 shows that most stations (67%) in the Colorado River Basin exhibit both decreasing April through July runoff in conjunction with decreasing water year runoff. Of the 14 stations with increasing trends in water year runoff volume, 9 stations also exhibit increasing April through July runoff; over the Colorado River Basin, the majority of annual runoff has traditionally been observed during these months. There are no stations within the Colorado River Basin that exhibit increasing April through July runoff and decreasing water year runoff.

Table 3. Results of Linear Trend Analysis Applied Over Each USGS Station Considered.
Streamflow Station Characteristics USGS HCDN
43 Stations
Decreasing water year volume 29 (67)
Increasing water year volume 14 (33)
Decreasing April-July volume 34 (79)
Increasing April-July volume 9 (21)
Increasing water year/increasing April-July volume 9 (21)
Increasing water year/decreasing April-July volume 0 (0)
Decreasing water year/increasing April-July volume 5 (12)
Decreasing water year/decreasing April-July volume 29 (67)
Earlier peak flow/earlier date to 50% annual flow 32 (74)
Earlier peak flow/later date to 50% annual flow 1 (2)
Later peak flow/earlier date to 50% annual flow 3 (7)
Later peak flow/later date to 50% annual flow 7 (16)
Note: Values in parentheses are percentages.
Trends in the Timing of Daily Runoff
The timing of inflow in the Colorado River Basin is not only important to water resource managers but also to those who benefit from timely inflows impacting hydroelectric and environmental endeavors. For the purposes of this study, the maximum daily flow observed over the course of a water year is referred to as the “peak flow.” Also considered is the number of days since the beginning of the water year to reach half of that water year’s annual flow volume. Most stations over the Colorado River Basin tend to show trends toward earlier peak flows and also tend to show trends toward reaching 50% of the annual water year flow earlier. For both parameters, the average amount of days to reach each date decreased by approximately 0.1 days per year. Table 3 summarizes the number of stations experiencing changes to the timing of peak flows and changes to the timing of reaching 50% of the annual water year total. The majority of stations (74%) yield earlier peak flows and reach 50% of the annual flow earlier, which supports various other studies which have noted a trend toward earlier runoff in the Colorado River Basin (e.g., McCabe and Clark, 2005; Regonda et al., 2005; Stewart et al., 2005; Miller and Piechota, 2008).

Relationship Between Western U.S. Snowpack and Streamflow Within the Upper Colorado River Basin
Trends in snowpack and streamflow observed in this study tend to support other recently published work (e.g., Mote, 2003; Stewart et al., 2004; Regonda et al., 2005; Knowles et al., 2006). This study proposes to further address the correlation between observed snowpack and runoff within Colorado River Basin headwater basins and impacts to the magnitude and timing of flows. The Upper Green, Gunnison, and San Juan headwater river basins are presented here and are representative of the major subbasins of the Upper Colorado River.

Gunnison River Basin
The Gunnison River Basin is located in west central Colorado and contributes approximately 16% of the total annual runoff to the Colorado River from the Upper Colorado River Basin. Reclamation operates the Aspinall Unit (i.e., the system of three dams, Blue Mesa, Crystal, and Morrow Point and their associated reservoirs) within this subbasin to protect endangered fish species within the Gunnison River while also providing water for municipal and agricultural use in accordance with the Aspinall Unit Operations Draft Environmental Impact Statement (U.S. Department of the Interior, Bureau of Reclamation, Upper Colorado Region, 2009).

Figure 9 illustrates recent snowpack and streamflow characteristics over the Gunnison River Basin. The relationship between annual April through July runoff with observations of SWE is similar when SWE is compared to total annual runoff. Over the last 10 years, the average water year total runoff and April through July runoff have been below average consistently as the length of snow season has been below average. Since 1979, April through July runoff in the basin has consistently been above or below average with peak SWE observations from SNOTEL stations. Over the last decade, earlier dates to the timing of the end of the snowpack season have corresponded well with decreased average streamflow in the Gunnison River Basin.

Details are in the caption following the image
Figure 9
Open in figure viewer
PowerPoint
In Each Plot, the Years Are Indicated as Dots. On the x-axis, the average of standardized annual April through July runoff (annual April-July Q) from all considered daily USGS stations.

Upper Green River Basin
The Upper Green River Basin is the northernmost subbasin in the Upper Colorado River Basin, the bulk of which is located in southwest Wyoming. The subbasin contributes approximately 13.2% of the annual water year runoff to the Colorado River and is primarily regulated by Reclamation through the Fontenelle and Flaming Gorge Dams. Flaming Gorge Dam is operated in accordance with the EIS published by Reclamation (U.S. Department of the Interior, Bureau of Reclamation, Upper Colorado Region, 2005) in order to protect critical habitat for endangered fish species in the region while maintaining water use development goals under the Colorado River Storage Project.

April through July runoff in the Upper Green River Basin is representative of the total annual water year flow. Throughout the period of shared observations, there is good correspondence between observed aggregate average peak SWE and April through July runoff. The average aggregate date to the peak SWE observation also agrees with April through July runoff, where earlier (later) peak SWE observations typically indicate below (above) average aggregate April through July runoff. Again, average aggregate April through July runoff corresponds more strongly with the end of the snowpack season than with the beginning of the snowpack season (Figure 10).

Details are in the caption following the image
Figure 10
Open in figure viewer
PowerPoint
As Described in Figure 9 for the Upper Green River Basin.

San Juan River Basin
The San Juan River Basin contributes nearly 14.3% of the average annual runoff to the mainstem Colorado from the Upper Colorado River Basin. The San Juan River within the basin is regulated primarily by Reclamation through the operation of the Vallecito and Navajo Dams and reservoirs. The Navajo Reservoir is part of the aforementioned Colorado River Storage Project and is operated to aid the continued development of water resources in the Upper Colorado River Basin. The Navajo Reservoir is operated under accordance with Environmental Impact Statement published by Reclamation (2006) and in conjunction with the Fish and Wildlife Service’s (FWS) San Juan River Basin Recovery Implementation Program (U.S. Fish and Wildlife Service, 2006) in an effort to protect critical habitat to endangered fish species in the basin.

Average aggregate snowpack characteristics correspond with average aggregate April through July runoff characteristics in a similar fashion as previously discussed river basins. In particular, there is strong correspondence between average aggregate peak SWE and the timing of the end of snowpack season with average aggregate April through July runoff (Figure 11).

Details are in the caption following the image
Figure 11
Open in figure viewer
PowerPoint
As Described in Figure 9 for the San Juan River Basin.

Discussion
Basin scale hydroclimatology has become an important consideration of water resource managers, particularly as it relates to streamflow within a river system (e.g., Grantz et al., 2005). As such, consideration of hydroclimatic trends within the Colorado River Basin has become important, particularly in light of the recent historic drought. There is evidence to suggest significant decreasing trends in snowpack, particularly during the current drought period. In the snow driven hydrology of the Upper Colorado River Basin, this correlates well with decreasing trends in both observed and natural streamflow in the basin.

Based on daily SNOTEL observations, the length of snowpack season has shortened during this period of drought, and corresponds to below average aggregate April through July runoff in Colorado headwater river basins. Interestingly, there is a much stronger correspondence between runoff characteristics and the timing of the end of the snowpack season than correspondence between runoff characteristics and the timing of the beginning of the snowpack season. As snowpack and precipitation characteristics change in the Colorado River Basin in response to climate change, water supply forecasters in the region may no longer be able to rely on past observations of basin hydroclimatology to aid in the development of water supply projections. By improving the understanding of the changing relationship between precipitation and streamflow, improved forecasts may be attainable.

While these results agree and provide support for previous studies showing a shift in the timing and magnitude of runoff in the Colorado River Basin, this study does not confirm an earlier hypothesis suggesting that the timing of runoff in the Colorado River Basin is due to the changing characteristics of precipitation in the basin (Miller and Piechota, 2008). This study did not observe any significant trends in the frequency of snowfall and rainfall events. Investigation into the frequency of precipitation events with a more robust gaging network (e.g., COOP stations) in conjunction with temperature observations may provide improved insight as to the changing character of precipitation in the basin. Results do support that over this period of drought, the Colorado River Basin is experiencing decreased snowpack and shorter snowpack seasons due to earlier snowmelt.

The Colorado River Basin has experienced extreme drought in the past (e.g., Woodhouse et al., 2006; Meko et al., 2007; Timilsena et al., 2007) and may persist into the future as the impacts of climate change are realized. As lower water supply is projected to persist in the Colorado River Basin, water resource managers and forecasters should continue to expect shorter snowpack seasons and resultant decreased and earlier runoff in the basin. It is possible that earlier snowmelt runoff is more susceptible to infiltration and evaporative losses throughout the basin, as increasing temperatures may increase both potential and actual evaporation rates. With continued drought and decreased spring runoff, water resource managers and water supply forecasters must continue to investigate methods to improve projections of water supply as the climate changes in addition to continuing effective water management policies and conservation practices.