Shifts in historical streamflow extremes in the Colorado River Basin

Journal of Hydrology: Regional Studies 12 (2017) 363–377

Contents lists available at ScienceDirect

Journal of Hydrology: Regional Studies journal homepage: www.elsevier.com/locate/ejrh

Shifts in historical streamflow extremes in the Colorado River Basin ⁎

MARK

Kurt C. Solander , Katrina E. Bennett, Richard S. Middleton Earth and Environmental Science Division, Los Alamos National Laboratory, Los Alamos, NM 87545, United States

AR TI CLE I NF O

AB S T R A CT

Keywords: Streamflow Extremes Mann-Kendall trend analysis Generalized Extreme Value Snowmelt Colorado River Basin

The global phenomenon of climate change-induced shifts in precipitation leading to “wet regions getting wetter” and “dry regions getting drier” has been widely studied. However, the propagation of these changes in atmospheric moisture within stream channels is not a direct relationship due to differences in the timing of how changing precipitation patterns interact with various land surfaces. Streamflow is of particular interest in the Colorado River Basin (CRB) due to the region’s rapidly growing population, projected temperature increases that are expected to be higher than elsewhere in the contiguous United States, and subsequent climate-driven disturbances including drought, vegetation mortality, and wildfire, which makes the region more vulnerable to changes in hydrologic extremes. Here, we determine how streamflow extremes have shifted in the CRB using two statistical methods—the Mann-Kendall trend detection analysis and Generalized Extreme Value (GEV) theorem. We evaluate these changes in the context of key flow metrics that include high and low flow percentiles, maximum and minimum 7-day flows, and the center timing of streamflow using historical gage records representative of natural flows. Monthly results indicate declines of up to 41% for high and low flows during the June to July peak runoff season, while increases of up to 24% were observed earlier from March to April. Our results highlight a key threshold elevation and latitude of 2300 m and 39° North, respectively, where there is a distinct shift in the trend. The spatiotemporal patterns observed are indicative of changing snowmelt patterns as a primary cause of the shifts. Identification of how this change varies spatially has consequences for improved land management strategies, as specific regions most vulnerable to threats can be prioritized for mitigation or adaptation as the climate warms.

1. Introduction Increased greenhouse gas emissions have caused a worldwide rise in both temperature means and extremes (Easterling et al., 2000). Global temperatures have risen by an average of 1 °C since the preindustrial age (Hansen et al., 2006), while increases in extremes have resulted in the maximum (minimum) historical means from the early 20th century occurring at greater (less) regularity during the latter part of the 20th century (Alexander et al., 2006). Given the expected increase in greenhouse gas emissions over the next century, such temperature trends are only anticipated to continue or even accelerate (Meinshausen et al., 2011). Indeed, comparisons of extreme temperatures from Global Climate Models (GCMs) to observed temperature data over the latter half of the 20th century support this conclusion (Zwiers et al., 2011). Understanding the direction, magnitude, and seasonality of impacts from these temperature changes on hydrology and associated ecosystems is thus paramount to mitigating risk from climate change (McDowell et al., 2017, in review). The reported increases in mean and extreme temperatures have led to substantial changes in hydrology, resulting in a loss of

⁎

Corresponding author. E-mail address: [email protected] (K.C. Solander).

http://dx.doi.org/10.1016/j.ejrh.2017.05.004 Received 27 January 2017; Received in revised form 12 May 2017; Accepted 13 May 2017 2214-5818/ © 2017 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

hydrologic stationarity (Milly et al., 2008). This has manifested itself in all aspects of the water cycle above and below the ground surface, causing shifts in the mean and extreme values of precipitation (Karl et al., 1995; Karl and Knight, 1998; Asadieh and Krakauer, 2015), runoff and streamflow (Lins and Slack, 1999; Dai et al., 2009), evapotranspiration (Szilagyi et al., 2001; Szilagyi et al., 2001; Jung et al., 2010), water vapor (Seager et al., 2010), and soil moisture (Andreadis and Lettenmaier, 2006; Albergel et al., 2013). Groundwater is similarly affected, although the quantification of associated impacts are less clear due to the difficulty in isolating the competing impacts from land use change, such as the intensification of groundwater use for irrigation (Taylor et al., 2013). Given the motivation of understanding how water and energy supplies as well as associated critical infrastructure might be affected—as these are particularly vulnerable to floods and droughts—we focus our efforts here in understanding how historic streamflow extremes have already shifted. Ultimately, the results of these efforts will be used to compare with how streamflow extremes will change in the future. Trends in extreme streamflow are often not as pronounced as extreme precipitation in areas where runoff is dominated by snowmelt. This is due to the offsetting effects of simultaneous declines in snowpack, which have been observed in areas such as the mountainous western United States (U.S.) (Mote, 2006). For example, there is a strong connection between high streamflow and the incidence of heavy precipitation events where they are becoming more common in the eastern U.S. (Groisman et al., 2001). However, due to the concurrent decreases in snowpack and snow cover extent in the more mountainous west, no such pattern exists in this part of the country in spite of the increase in heavy precipitation episodes also occurring in the region (Prein et al., 2016). The strong association with elevation and observed shifts in western U.S. streamflow center timing further indicates that snowmelt changes are largely contributing to this transformation (McCabe and Clark 2005; Regonda and Rajagopalan, 2004; Stewart et al., 2005; Mote, 2006; Clow et al., 2010; Dudley et al., 2017), likely offsetting any appreciable gains to high streamflow events from an increase in the frequency of heavy precipitation. In spite of some evidence for weaker extreme streamflow trends due to the observed competing changes in snow, broad spatial and temporal patterns have still been recognized across North America. For example, increases in streamflow were generally reported for minimum flow percentiles over much of the conterminous U.S. (Lins and Slack, 1999) or centered in the eastern U.S. (McCabe and Wolock, 2002; Ahn and Palmer, 2015), while decreases were observed in the Pacific Northwest (Ahn and Palmer, 2015; Kormos et al., 2016). Such trends were more variable for maximum flow percentiles, although a tendency towards increased variability in high flows was noted for the continental U.S. (Ahn and Palmer, 2015). By contrast, in Canada a greater number of decreases in streamflow maximums were observed, while the direction of trends were more variable for streamflow minimums (Zhang et al., 2001; Burn et al., 2010). The reduction in maximum flows is again related to decreases in or earlier peak snowmelt, which leads to a more gradual rise in peak streamflow and dampening of the overall hydrograph (Zhang et al., 2001). Similar to what was found across Canada, within boreal basins of Alaska and the Yukon Territory decreases in annual maximum streamflow were particularly pronounced in snowmelt-dominant basins, while more statistically significant increases were observed in minimum flows (Bennett et al., 2015). The Colorado River Basin (CRB) is located within the intermountain region of the western United States and serves as an excellent setting to analyze shifts in hydrologic extremes for a variety of factors. The region recently underwent one of the largest droughts observed in the historical record (Piechota et al., 2004) and based on future temperature increases simulated in Global Climate Models (GCMs), more severe droughts than what have been observed in the paleoclimate record are anticipated by 2100 (Woodhouse et al., 2010). Other climate change impact studies indicate a CRB-wide reduction of 8–11% in streamflow due to changes in climate (Christensen and Lettenmaier, 2007), as well as a 6–11% decline in annual streamflow by 2100 from vegetation and climate disturbances in the San Juan sub-basin of the CRB (Bennett et al., 2017, in review). Even higher reductions were projected using a range of emission scenarios for other CRB sub-basins (Ficklin et al., 2013). Additional impacts include a ∼30% decline in runoff (Barnett and Pierce, 2009), and longer-lasting, higher-severity droughts (Cayan et al., 2010; Woodhouse et al., 2016). Moreover, the high rate of population growth coupled with the projected temperature increases are expected to further strain water supplies and make critical water and energy supply infrastructure particularly vulnerable to floods and droughts (MacDonald, 2010; Stewart et al., 2015). Two leading statistical methods used to understand changes in hydrologic extremes include the non-parametric Mann-Kendall trend analysis (Mann, 1945; Kendall, 1975) and the Generalized Extreme Value (GEV) theory (Coles, 2001). The Mann-Kendall trend analysis has been applied in previous studies to understand shifts in hydrologic extremes (Lins and Slack, 1999; Zhang et al., 2001; Miller and Piechota, 2008; Villarini et al., 2009; Burn et al., 2010; Cheng et al., 2014; Sagarika et al., 2014; Bennett et al., 2015). However, as this statistical method was not specifically developed for analysis of extremes and provides no means of determining changes in the probability of different flow magnitudes—both of which are important for quantifying risk in water resource management and engineering design—the GEV analysis offers a good compliment to this approach. GEV theory can be applied to determine changes in extreme streamflow metrics in either a stationary or nonstationary setting by using time-dependent parameters in the GEV distribution (Katz, 2013). As such, similar to the Mann-Kendall analysis, the GEV is also used to detect trends in extremes. In this study, we evaluate changes in extreme streamflow in the CRB using the Mann-Kendall and GEV statistical approaches. The Mann-Kendall trend analysis was applied to extreme streamflow metrics and the center timing of streamflow, which were derived from historical daily discharge to determine the size and direction of changes at annual, seasonal, and monthly timescales. Next, the Generalized Extreme Value (GEV) analytical approach was used to estimate how the high- and low-flow tails of streamflow maxima and minima are changing in the region and to assess changes in extreme flow return intervals (Zhang and Zweirs, 2013). The MannKendall and GEV results were both evaluated against the respective elevation and latitude of the streamflow gages to identify shifts associated with changing snow and snowmelt patterns due to their strong spatial dependency. We examined changes in extremes using exclusively unimpaired gages to isolate the source of the change to climatic factors alone due to the high vulnerability to impacts from changing climate within the CRB region. This study differs from previous studies by providing the first investigation of streamflow extreme shifts and the drivers of change 364

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

while focusing on the CRB. Unlike previous examinations of streamflow extremes elsewhere, we also incorporated all of the following methods: 1) the analysis was conducted for multiple extreme streamflow metrics as well as the center timing of streamflow; 2) trends were analyzed at multiple temporal scales and related to elevation and latitude to assess which part of the annual hydrograph is contributing most to the observed changes and provide multiple lines of evidence to demonstrate how and when these changes are occurring; and 3) gage stations used in the analysis were grouped according to record length and four groups were used to minimize potential bias from single climatic events on the overall results. Application of this suite of methods is vital for gaining a robust understanding of historic extreme streamflow changes in the CRB so that we may ultimately examine how water and energy supplies, as well as critical infrastructure, will be affected under future climate conditions by running simulations in hydrology models and GCMs. 2. Methods 2.1. Data Historical daily streamflow gage records were obtained using only sites listed in the United States Geological Survey (USGS) Geospatial Attributes of Gages for Evaluating Streamflow version II database (GAGESII, Falcone et al., 2010). GAGESII consists of a subset of USGS national streamflow gage records that were targeted based on the relatively long length of record and minimal influence from upstream water resources management. Only those gages listed as ‘reference’ gages with negligible upstream flow alterations were used to increase the likelihood that reported shifts in hydrologic extremes could be attributed to changes in climate. Furthermore, only gages with a minimum of 30-years of data and less than 25% missing values were selected to ensure long records were used and consistency was maintained so comparison of results among each group could be justified. The Mann-Kendall trend analysis was run without data gap filling to avoid bias. However, gaps were filled for the GEV analysis using the mean of two adjacent data points, as the GEV method applied necessitated the use of continuous records. As such, gages with two or more consecutive records missing were omitted from the study and fewer gages were used in the GEV with respect to the Mann-Kendall trend analysis. Gages were further filtered to include only those situated within our study region; the upper and lower CRB (located within 2-digit, USGS Hydrologic Unit Classification (HUC) basins 14 and 15). USGS streamflow data is available for download at waterdata.usgs.gov. Gage station records were grouped according to record length (Fig. 1) and results presented in terms of the mean changes observed per group. Four record length groups were used, which consisted of 30, 40, 50, and 60 years (hereafter referred to as station groups) with each terminating at the end of 2014. This technique was implemented to remove the effects of single climatic (e.g. El Niño Southern Oscillation) or storm events from having too much of an influence on the overall results, which would have been more of an issue if the analysis was conducted over a single period (Hannaford et al., 2013; Salas and Obeysekera, 2014). Summary spatial statistics of the data and the number of gage stations used per station group and streamflow metrics analyzed are presented in Table 1. 2.2. Mann-Kendall trend analysis The Mann-Kendall trend analysis was used to identify trends for three different flow indices: percentiles [maximum (Q100), minimum (Q0), Q95, and Q05], 7-day flows, and streamflow center timing. Due to the high degree of similarity between the high flow (Q100 and Q95) and low flow (Q05 and Q05) percentiles, we only show trend results for the Q100 and Q0 flows. Trend analysis of percentiles was conducted at annual, seasonal, and monthly timescales. Pre-whitening measures were not employed to avoid dampening the signal of trends, as the trend magnitude was high enough or was derived from sufficient observation quantities to make autocorrelation effects negligible (Bayazit and Önöz, 2007). For the purposes of this study, the seasons are defined as the Northern Hemisphere winter (JFM), spring (AMJ), summer (JAS), and fall (OND). The 7-day maximum and minimum flows that had a 1:10 (7Q10) and 1:2 (7Q2) probability of occurrence were also evaluated using the Mann-Kendall trend analysis. Of these, the 7Q10 minimum flow is the most widely reported due to its relevance for the survival of fish species (Richter et al., 1996). Moreover, this metric has significance for water quality, as it is used to establish pollutant discharge limits in the U.S. by the Environmental Protection Agency (Kormos et al., 2016). The 7Q10 maximum flow was included for its importance to different engineering design criteria (Mishra et al., 2011). The 7Q2 minimum flow was also tested due to its application in Mantua et al. (2010) and its maximum flow counterpart was incorporated as well to maintain consistency. Maximum and minimum 7Q10 and 7Q2 flows were calculated based on the log-Pearson Type III probability distribution of annual 7day flows over 10 years using a 10-year moving window. The number of observations used in the calculation of the 7-day flow trends was necessarily reduced by nine relative to that of the percentiles due to application of the 10-year moving window. Seasonal 7Q2 flows were omitted from the results due to similarities with the 7Q10 flows. The Mann-Kendall trend analysis was also applied to the center timing of streamflow, which represents the Julian Day of each year when half of the total annual streamflow has passed the gage. Unlike the other metrics, the streamflow center timing was applied to daily data arranged according to the water year rather than calendar year due to its strong relation to peak snowmelt timing (Stewart et al., 2005; Clow, 2010). Although not considered to be an extreme streamflow metric, the center timing was still analyzed due to its strong connection to snowmelt and to assess how overall shifts in annual flow towards earlier or later in the season may have contributed to a higher or lower incidence of extreme seasonal flow events. Two sets of plots were used for presentation of the Mann Kendall results. Annual, seasonal, and monthly percentile, as well 365

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Fig. 1. Study site map showing USGS gage station locations along with corresponding record length and elevations within the Colorado River Basin (CRB).

streamflow center timing trends were presented in the form of box plots for each station group to provide a general overview of changes occurring across the whole CRB, rather than specific sites, which is one of our aims of the study. Annual and seasonal Q0 and Q100 streamflow changes were also presented with respect to elevation and latitude for each individual gage station. This was done for the 40-year station group, which consisted of the most gage stations and was representative of the changes observed among other station groups. Following application of the Mann-Kendall trend analysis, the magnitude of trend is estimated (Sen, 1968). Trends were deemed statistically significant if p < 0.33, given the likely existence of a trend below this threshold (Hirsch et al., 2015). The percentage of significant trends for each station group is reported to provide an overall indication of statistical robustness for the given trend.

2.3. Generalized extreme value analysis Upon completion of the Mann-Kendall trend analysis, the GEV analysis was used to detect stationary and non-stationary changes in high and low streamflow at the annual and seasonal timescales. With some notable exceptions, the approach outlined in Bennett et al. (2015) was applied to calculate the GEV distribution using the R-project GEVcdn package explained in Cannon (2010, 2011). 366

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Table 1 Summary spatial statistics per analysis for different station groupings. Station Group

Number Stations

Latitude Range (dec deg)

Longitude Range (dec deg)

Elevation Range (m)

Analysis

1985

n = 41 n = 26 n = 31 n = 41 n = 42 n = 30 n = 33 n = 42 n = 23 n = 27 n = 20 n = 14 n = 15 n = 20 n=8 n=6 n=6 n=8

31.36 31.36 31.36 31.36 31.36 31.36 31.36 31.36 33.06 33.06 31.36 31.36 31.36 31.36 31.36 31.36 31.36 31.36

−115.30 −115.30 −112.60 −115.30 −115.30 −112.60 −112.60 −115.30 −112.60 −115.30 −115.30 −115.30 −111.54 −111.78 −110.81 −110.71 −110.71 −110.81

428 to 3179 536 to 3179 536 to 3179 428 to 3179 536 to 3179 536 to 3179 536 to 3179 536 to 3179 536 to 3179 536 to 3179 536 to 2850 536 to 2850 536 to 2850 536 to 2850 829 to 2789 1408 to 2789 1408 to 2789 829 to 2789

Mann-Kendall Mann-Kendall Mann-Kendall Mann-Kendall Mann-Kendall Mann-Kendall Mann-Kendall Mann-Kendall GEV – Q05 GEV – Q95 Mann-Kendall Mann-Kendall Mann-Kendall Mann-Kendall Mann-Kendall Mann-Kendall Mann-Kendall Mann-Kendall

1975

1965

1955

to to to to to to to to to to to to to to to to to to

43.03 42.11 42.11 43.03 43.03 42.11 42.11 43.03 42.11 42.11 43.03 43.03 42.11 42.11 43.03 42.11 42.11 43.03

to to to to to to to to to to to to to to to to to to

−105.91 −105.91 −105.91 −105.91 −105.91 −105.91 −105.91 −105.91 −105.91 −105.91 −105.97 −105.97 −105.97 −105.97 −106.27 −106.27 −106.27 −106.26

– – – – – – – –

Percentiles Min 7Q2 & 7Q10 Max 7Q2 & 7Q10 Center Timing Percentiles Min 7Q2 & 7Q10 Max 7Q2 & 7Q10 Center Timing

– – – – – – – –

Percentiles Min 7Q2 & 7Q10 Center Timing Max 7Q2 & 7Q10 Percentiles Min 7Q2 & 7Q10 Max 7Q2 & 7Q10 Center Timing

Here, we summarize some of the main points of this method, which are described in Bennett et al. (2015) in greater detail. The GEV is based on the premise that samples of block maxima from a large enough sample size will approach the GEV distribution, which can be used to obtain the frequency and magnitude of high and low streamflow (Coles, 2001). The GEV distribution is defined by the cumulative distribution function, shown in Eqs. (1) and (2): 1

⎡ (y − μ) ⎫ κ ⎤ y−μ F (y; μ, α, κ ) = exp ⎢−⎧1 − κ >0 ⎥, κ ≠ 0,1 − κ ⎬ ⎨ α α ⎭ ⎩ ⎥ ⎢ ⎦ ⎣

F (y; μ, α ) = exp ⎡−exp ⎧ − ⎢ ⎨ ⎩ ⎣

(y − μ) ⎫ ⎤ ,κ=0 α ⎬ ⎭⎥ ⎦

(1)

(2)

where μ is the location parameter, a is the scale parameter defining the deviation around μ; and κ is the shape parameter that determines the tail of the distribution (also shown in Bennett et al., 2015). These parameters are estimated using the maximum likelihood (ML) measure, which is described using the probability density function (pdf) calculated in Eqs. (3) and (4) (also shown in Bennett et al., 2015):

f (y; μ, α, κ ) =

(y − μ) ⎫(1 − κ )/ κ (y − μ) ⎫1/ κ ⎤ 1⎧ 1−κ exp ⎡ ,κ≠0 − ⎧1 − κ ⎢ ⎨ α ⎬ α ⎬ α⎨ ⎭ ⎥ ⎩ ⎭ ⎦ ⎣ ⎩

f (y; μ, α ) = exp ⎡− ⎣

(y − μ) ⎤ (y − μ) ⎫ ⎤ exp ⎡−exp ⎧ − ,κ=0 ⎢ ⎨ α ⎦ α ⎬ ⎩ ⎭⎥ ⎣ ⎦

(3)

(4)

The shape parameter is estimated using a penalized version of ML that uses a prior distribution based on conditions observed in nature (Cannon, 2010; Bennett et al., 2015). In this study, the Q05 and Q95 flows with ten-year return periods at the annual and seasonal timescales were employed as block maxima and minima. The Q05 (Q95) flows were used over the Q0 (Q100) flow percentiles and 7Q10 minimum (maximum) flows because streamflow in some of the Lower CRB gages is representative of ephemeral conditions and zero-flow events make GEV parameter estimation impossible (Katz, 2010). Note that other studies of climate extremes also consider more moderate extreme values than the maxima (Mannshardt-Shamseldin et al., 2010). GEV model selection was limited to three types: stationary (S), non-linear stationary (NLS), and non-linear non-stationary (NLNS). Selection of the stationary model assumed model parameters were independent with time, while parameter values vary linearly and log-linearly for the NLS model. Additional parameters were used if the NLNS model was selected to constrain the magnitude of nonlinearity. For example, scale parameters had to be above 15% of the range of streamflow values to prevent over-fitting to outliers (Cannon, 2012). To best-fit model parameters to the optimum model, the Akaike Information Criterion corrected (AICc) for small sample sizes was applied (Burnham et al., 2011). Selected model types along with the magnitude and direction of change were plotted against both latitude and elevation to investigate potential causal factors of the change. 3. Results The Mann-Kendall trend analysis showed the majority of annual Q100 flows as well as the maximum 7Q10 flow metrics are decreasing (Fig. 2), indicating a trend towards lower water availability in the CRB. The median and majority of the interquartile 367

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Fig. 2. Box plots of annual percentile and 7Q10 streamflow Mann Kendall trend results for each station group. The percent of statistically significant (p < 0.33) stations is also shown.

368

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Fig. 3. Box plots of seasonal percentile and 7Q10 streamflow Mann Kendall trend results for each station group. The percent of statistically significant (p < 0.33) stations is also shown.

range (IQR) of changes were negative for these flow types among the four station groups. In contrast, the annual low flow metrics exhibited weaker changes overall, as the median IQR of changes were much closer to zero. An exception is the 30-year station group of the low flow metrics, as the IQR of changes were consistently negative among all annual minimum flow metrics. However, we note that much stronger decreases for the 30-year station group was not observed for the high flow metrics, meaning that the cause was specific to low flows. The percentage of statistically significant trends tended to be higher overall for the 7Q10 flow changes and Q0 over the Q100 flows. At the seasonal scale, results indicate the majority of Q100 flows increased in three of four station groups during the winter, but were more stable or decreased among the four station groups during the other seasons (Fig. 3). In contrast, the majority of Q0 flows increased in three of four station groups during the spring, but decreased or displayed an overall more neutral change among the four station groups during the other seasons. Similar to the annual results, the 30-year station group showed a stronger decreasing trend most pronounced during the winter for both high and low flow metrics. Thus, the same factors listed as a potential cause of this pattern for the annual flow trends are also applicable here during the winter. Similar to the annual results, the levels of statistical significance were greater for the 7Q10 than the other flow metrics, which is likely caused by a difference in how the 7Q10 flows were calculated and the fewer observations used to compute each trend. The Q100 trend results at the monthly scale reveal majority increases in three of the four station groups during March and April by up to 21% and 14%, respectively, for the median values (Fig. 4). Stronger majority decreases were observed for Q100 flows in June and July in all station groups by up to 26% and 41%, respectively for the median values. Majority increases and decreases were also observed in three of four station groups during January and August, respectively, but these changes were weaker in magnitude and at a much lower level of statistical significance overall. Q100 changes were weaker or more neutral during the other months. The Q0 trend results revealed majority increases of up to 19% in the median for three of four station groups during April, and decreases of up to 26% and 41%, respectively, in the median among all station groups in June and July. Decreases were also observed 369

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Fig. 4. Box plots of monthly percentile Mann Kendall trend results for each station group. For the Q0 results (top), the mean monthly percent of statistically significant (p < 0.33) stations ranged from a minimum of 42% in October to a maximum of 72% in April. For the Q0 results (top), the mean monthly percent of statistically significant (p < 0.33) stations ranged from a minimum of 31% in September to a maximum of 60% in April.

among all station groups in August and September, but these changes were weaker in magnitude and at a much lower level of statistical significance overall. Higher levels of statistically significant trends for both high and low flows tended to occur during late winter to spring months. The 30-year station group again showed stronger decreasing trends for most of the months relative to other station groups. The same factors are likely causing this distinction as in the seasonal and annual trend results. Moreover, this highlights the importance of conducting the analysis for more than one period. Q0 changes were weaker or more neutral during the other months. Distinct spatial patterns exist with the annual and seasonal trend results presented in Figs. 5 and 6 . With the exception of one gage station, all gages exhibiting an increase in Q0 flows at all seasons and annually occurred at high elevations (> 2300 m). Decreasing and neutral flow trends dominated at low elevations (< 2300 m) at all seasons and annually, and even at high elevations during the summer. Similar patterns were observed for the Q100 flow changes, except decreases dominated at high elevations in spring with fewer decreases at lower elevations during this season. An equal number of statistically significant trends (p < 0.33) were distributed throughout the latitude and elevation ranges that were tested, although the number of statistically significant trends is slightly higher during the spring and summer than other seasons and annually. Trends in the annual center timing of streamflow reveal there is a tendency towards streamflow occurring earlier in the season among all station groups (Fig. 7). The majority of gage stations in each group showed a decreasing trend in center timing with the largest median reduction of 10 days (2.5 days/decade) occurring for the 40-year station group. The percent of stations showing statistically significant trends (p < 0.33) ranged from 50 to 62%. The AICc-minimized model along with the magnitude and direction of trends for the ten-year return periods are shown with respect to elevation and latitude for the GEV analysis of Q05 (Fig. 8) and Q95 (Fig. 9) seasonal and annual flows. Similar to the MannKendall results, the Q05 flow results show a distinct temporal pattern where flows shifted from mostly increasing or stable trends at upper elevation (> 2300 m) stations during the winter and spring to a higher prevalence of stationary or decreasing trends during the summer. Flow decreases and neutral changes are in the majority for lower elevation stations (< 2300 m) during the summer and fall, resulting in an overall decrease in flows at most lower elevation stations during the annual timescale. The Q95 GEV results showed a much stronger spatiotemporal pattern during the winter and spring, as winter flow increases dominated at high elevations (> 2300 m), while flow trends above this elevation in the spring were stable or decreased. Flow decreases were of higher magnitude and more prevalent at low elevations (< 2300 m) during the spring than any other season. Low elevation sites during the summer, fall, and annual timescales exhibited much less change overall, with most stations showing stationary trends. The spatiotemporal patterns for both high and low flows, particularly during the winter, spring, and summer, support a lot of the Mann-Kendall trend results shown in Figs. 5 and 6. The exception is that more neutral changes are shown at high elevations for high spring and summer flows in the GEV, while more decreases are shown in the Mann-Kendall trend analysis. 4. Discussion The results support strong temporal variability in streamflow changes where part of the year shows more increases and other parts show decreases. Such findings indicate a diminishing water supply at critical times for the region, which could result in shortages for the generation of power or regional water supplies. For example, the shift in peak runoff observed from the spring and summer towards the winter, particularly at high elevations (> 2300 m) shown in Fig. 6, suggest that reservoir water drawdown from annual maximum storage will occur earlier, increasing the likelihood of future water and energy shortages during the summer and fall—exactly when these resources are needed the most due to higher warm season energy demands for cooling and elevated water demands from agriculture. Indeed, reservoir water and energy delivery shortages from decreasing water availability have already been projected in the CRB (Christensen and Lettenmaier, 2007; Barnett and Pierce, 2009), leading to investigation of improved 370

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Fig. 5. 40-year seasonal and annual Mann-Kendall trend results with respect to elevation and latitude for Q0 minimum flows. Circle size and color correspond to magnitude and direction of trend, while closed circles represent statistically significant trend (p < 0.33).

operation strategies to mitigate water and energy supply risk (Rajagopalan et al., 2009). In addition, the falling low flow conditions occurring at all elevations in summer through the fall for low elevations (< 2300 m) will make it more difficult for reservoir operators to meet minimum low-flow requirements that are essential for sustaining downstream fisheries (Richter et al., 2003). Although not explicitly investigated here, the spatiotemporal patterns identified in these streamflow changes are indicative of the strong influence of changing snow and snowmelt patterns. The reduction in spring-summer flows and increase in winter flows at high elevations (> 2300 m) and latitudes (> 39°N) is suggestive of a shift in flows from the traditional peak runoff to the melt season of the CRB. Increased melting or more rain-on-snow events earlier in the year, leading to higher peak flows during winter and spring and lower flows during the summer, could cause such a shift. Such findings are bolstered by the overall negative trends observed in the center timing of streamflow, providing additional evidence of a shift in streamflow extremes towards earlier in the year. Note that 371

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Fig. 6. 40-year seasonal and annual Mann-Kendall trend results with respect to elevation and latitude for Q100 maximum flows. Circle size and color correspond to magnitude and direction of trend, while closed circles represent statistically significant trend (p < 0.33).

these conclusions could not have been derived without conducting the analysis at multiple temporal scales for several high and low flow metrics and relating some of the observed changes to both elevation and latitude, which distinguishes this study of CRB streamflow extreme shifts from others already conducted in the basin. The idea that changing snow and snowmelt patterns are leading to changes in streamflow extremes we identified is supported by other climate trends reported in the literature. Strong reductions in snow-to-rain fractions were reported during March (Knowles et al., 2006), corroborating with our observation of higher peak streamflow during this month and the winter season. Increasing temperature and precipitation trends reported for the region were strongest for January to March (temperature) and February (precipitation) (Miller and Piechota, 2008), further promoting higher flows during this time of year. Declines in snow albedo caused by increased dust loading from agricultural development is also contributing to the shift, as the reduction in albedo has shortened the 372

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Fig. 7. Box plots of annual streamflow center timing Mann Kendall trend results. The percent of statistically significant (p < 0.33) stations for each station group are as follows: 1955 = 50%, 1965 = 50%, 1975 = 62%, 1985 = 50.

duration of snow cover in the Upper CRB by several weeks over the 20th century (Painter et al., 2010). Although not necessarily related to changes in snow, at least some of the decreasing trends observed in this study are likely related to overall declines in water availability in CRB streamflow due to warming (Woodhouse et al., 2016). The elevation threshold of 2300 m identified in this study that demarks a shift in the direction of streamflow change due to the stronger influence of changes in snow at higher elevations is also supported in the literature. This same elevation threshold is used to define the transition from rain-dominated to snow-dominated streamflow regimes (Kampf and Lefsky, 2016). Snow is reported to be the dominant form of precipitation in the CRB just above the elevation threshold at 2500 m (Clow, 2010). In addition, the transitional snow cover zone, or the region where snow cover is expected to experience the most variability during the winter and spring, is defined as being between 2550 and 3050 m (Richer et al., 2013). Thus, the persistence of snow and seasonal changes in its contribution to streamflow within this region is more vulnerable to temperature increases. This notion is supported by the winter and spring streamflow increases and summer decreases that were located within or just below this elevation band down to 2300 m. Moreover, reductions in the contribution of snow to streamflow were reported in this transition zone (Kampf and Lefsky, 2016), suggesting that more precipitation is occurring as rain and favoring higher peak runoff earlier in the year, which aligns with our results. Further providing support for our findings is the similarity to conclusions drawn from a number of comparable studies that evaluated streamflow changes in western North America. Reductions in annual maximum flow were also observed in boreal Canadian and Alaskan streams (Bennett et al., 2015), which, similar to here, are due to strong changes in snowmelt that accounts for a large proportion of the annual water budget in the two regions. Given the strong covariance between mean and extreme streamflow (McCabe and Wolock, 2014), good agreement was also found with the results of Miller and Piechota (2008), who observed increasing trends in mean CRB winter streamflow and a corresponding decline during the spring. Strong shifts towards earlier season snowmelt runoff and streamflow center timing were also observed in the upper CRB (McCabe and Clark, 2005; Clow 2010). Similar to the changes in streamflow noted here, the snowmelt shifts in those studies were strongly related to elevation and latitude. Even though the study returned results that are well-aligned with existing literature on CRB streamflow changes, it is important to highlight some caveats from our study. The number of statistically significant trends for the 7-day flows was greater than the flow percentiles or center timing results. Moreover, unlike the flow percentile results, most of the 7Q10 maximum and minimum flows exhibited more decreasing trends during the winter and 7Q10 minimum flows increased during the spring. These differences were likely due to the smaller number of stations employed or calculation of 7-day flows using ten-year moving windows, which reduced the number of observations used to calculate the trends relative to the flow percentile trends or GEV results. Thus, the 7-day flow trends should be treated separately from the percentile trends and results provided from the GEV. Even still, the importance of the 7day trends should not be underestimated due to the practical use of this metric to inform engineering design practices within floodplains, river flow management for fish species survival, and establishment of pollutant discharge limits by the EPA. As such, professionals using these metrics should take the changes presented here into account for future related planning purposes in their respective fields. Some of the annual, seasonal, and monthly trends were much more negative for the 30-year station groups. This dissimilarity could have resulted due to effects from more pronounced warming or extreme weather events affecting the region in the latter part of the record (Piechota et al., 2004; Kampf and Lefsky, 2016), differing influence from macroscale climate events (e.g. ENSO) (Salas and Obeysekera, 2014), or difference in the gage stations used relative to the other station groups. In contrast to the results shown or implied in this study, negligible advancement of CRB spring streamflows and variable changes in snowmelt were reported in Regonda and Rajagopalan (2004). This disparity occurred likely due to the difference in stations that were selected for analysis in that study. We also note that Murphy and Ellis (2014) observed no trends in mean streamflow within the 373

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Fig. 8. 40-year seasonal and annual minimized GEV models with respect to elevation and latitude for Q05 minimum flows. Circle size and color correspond to magnitude and direction of trend for flows with 1:10 probability of occurrence, respectively. Shape denotes favored minimized model type.

CRB. This discrepancy could be caused by the use of different stream gage data with longer record lengths or because 8 of 10 gage stations used in their analysis are listed as non-reference gages within the GAGES II database, whereas only reference gages were used in this study.

374

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Fig. 9. 40-year seasonal and annual minimized GEV models with respect to elevation and latitude for Q95 minimum flows. Circle size and color correspond to magnitude and direction of trend for flows with 1:10 probability of occurrence, respectively. Shape denotes favored minimized model type.

5. Conclusions and recommendations Our aim with this work was to evaluate the extent and magnitude of historical shifts in streamflow extremes in the CRB at multiple temporal scales. Annually, most of the maximum and minimum flow trends for the different metrics among the station groups indicate reduced flows, which could lead to shortages at critical times of the year. Seasonal and monthly results indicate 375

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

temporal heterogeneities exist, with stronger basin-wide decreases in summer (and spring for low flows) and increases during winter (and spring for high flows). Monthly results indicate changes were strongest and more statistically robust in the snowmelt season, as the basin-wide median of maximum and minimum streamflow was observed to decrease by up to 41% during the traditional peak runoff months of June and July and increase by up to 24% during March and April. The 2300 m elevation and 39°N latitude thresholds of winter and spring streamflow changes identified in both the Mann-Kendall and GEV analysis indicate that changing snow and snowmelt patterns are the likely primary cause of changes during these periods. The strong spatial link of our results to both latitude and elevation and findings from other studies suggests warming-induced accelerated snowmelt or increases in the quantity of rain with respect to snow as well as overall increases in February precipitation are the likely primary causes of these changes. Given the prognosis of future climate impacts on hydrologic extremes, more research is necessary to assess how these changes will further contribute to shifts in streamflow extremes and affect associated water and energy supplies in the CRB. Causes of streamflow extreme shifts should be quantified for different elevations and latitudes and related to changes in snow. In addition, the causes as well as changes in extremes should be simulated in GCMs and hydrology models to estimate future water and energy resource vulnerabilities from changes in the timing and magnitude of high and low flows. Such work is critical to prioritize locations for better management of water resources, energy, and critical infrastructure so appropriate steps are taken to avoid major economic losses under a warmer regime. Acknowledgements This project was supported by the Los Alamos National Laboratory’s Laboratory Directed Research and Development (LDRD) program under the Critical Watersheds project (20150397DR). References Ahn, K., Palmer, R.N., 2015. Trend and variability in observed hydrological extremes in the United States. J. Hydrol. Eng. 21 (2), 4015061. Albergel, C., Dorigo, W., Reichle, R.H., Balsamo, G., de Rosnay, P., Muñoz-Sabater, J., Isaksen, L., de Jeu, R., Wagner, W., 2013. Skill and global trend analysis of soil moisture from reanalyses and microwave remote sensing. J. Hydrometeorol. 14 (4), 1259–1277. Alexander, L.V., Zhang, X., Peterson, T.C., Caesar, J., Gleason, B., Klein Tank, A.M.G., Haylock, M., Collins, D., Trwein, B., Rahimzadeh, F., Tagipour, A., Rupa Kuma, K., Revadekar, J., Griffiths, G., Vincent, L., Stephenson, D.B., Burn, J., Aguilar, E., Brunet, M., Taylor, M., New, M., Zhai, P., Rusticucci, M., Vazquez-Aguirre, J.L., 2006. Global observed changes in daily climate extremes of temperature and precipitation. J. Geophys. Res. Atmos. 111 (5), 1–22. Andreadis, K.M., Lettenmaier, D.P., 2006. Trends in 20th century drought over the continental United States. Geophys. Res. Lett. 33 (10), 1–4. Asadieh, B., Krakauer, N.Y., 2015. Global trends in extreme precipitation: climate models versus observations. Hydrol. Earth Syst. Sci. 19 (2), 877–891. Barnett, T.P., Pierce, D.W., 2009. Sustainable water deliveries from the Colorado River in a changing climate. Proc. Natl. Acad. Sci. 18, 7334–7338. Bayazit, M., Önöz, B., 2007. To prewhiten or not to prewhiten in trend analysis. Hydrol. Sci. 52 (4), 611–624. Bennett, K.E., Cannon, A.J., Hinzman, L., 2015. Historical trends and extremes in boreal Alaska river basins. J. Hydrol. 527, 590–607. Bennett, K.E., Bohn, T., Solander, K., McDowell, N.G., Xu, C., Vivoni, E., Middleton, R.S., 2017. Climate change and climate-driven disturbances in the San Juan River sub-basin of the Colorado River. Hydrol. Earth Syst. Sci. Discuss (in Review). Burn, D.H., Sharif, M., Zhang, K., 2010. Detection of trends in hydrological extremes for Canadian watersheds. Hydrol. Process. 24 (13), 1781–1790. Burnham, K.P., Anderson, D.R., Huyvaert, K.P., 2011. AIC model selection and multimodel inference in behavioral ecology: some background, observations, and comparisons. Behav. Ecol. Sociobiol. 65 (1), 23–35. Cannon, A.J., 2010. A flexible nonlinear modeling framework for nonstationary generalized extreme value analysis in hydroclimatology. Hydrol. Process. 24 (6), 673–685. Cannon, A.J., 2011. GEVcdn: an R package for nonstationary extreme value analysis by generalized extreme value conditional density estimation network. Comput. Geosci. 37 (9), 1532–1533. Cannon, A.J., 2012. Neural networks for probabilistic environmental prediction: conditional density estimation network creation and evaluation (CaDENCE) in R. Comput. Geosci. 41, 126–135. Cayan, D.R., Das, T., Pierce, D.W., Barnett, T.P., Tyree, M., Gershunov, A., 2010. Future dryness in the southwest US and the hydrology of the early 21 st century drought. Proc. Natl. Acad. Sci. 107 (50), 21271–21276. Cheng, L., AghaKouchak, A., Gilleland, E., Katz, R.W., 2014. Non-stationary extreme value analysis in a changing climate. Clim. Change 127 (2), 353–369. Christensen, N.S., Lettenmaier, D.P., 2007. A multimodel ensemble approach to assessment of climate change impacts on the hydrology and water resources of the Colorado River Basin. Hydrol. Earth Syst. Sci. 11 (4), 1417–1434. Clow, D.W., 2010. Changes in the timing of snowmelt and streamflow in Colorado: a response to recent warming. J. Clim. 23 (9), 2293–2306. Coles, S., 2001. An Introduction to Statistical Modeling of Extreme Values. Springer Verlag, London, UK 209 p. Dai, A., Qian, T., Trenberth, K.E., Milliman, J.D., 2009. Changes in continental freshwater discharge from 1948 to 2004. J. Clim. 22 (10), 2773–2792. Dudley, R.W., Hodgkins, G.A., McHale, M.R., Kolian, M.J., Renard, B., 2017. Trends in snowmelt-related streamflow timing in the conterminous United States. J. Hydrol. 547, 208–221. Easterling, D.R., Evans, J.L., Groisman, P.Y., Karl, T.R., Kunkel, K.E., Ambenje, P., 2000. Observed variability and trends in extreme climatic events: a brief review. Bull. Am. Meteorol. Soc. 81, 417–425. Falcone, J.A., Carlisle, D.M., Wolock, D.M., Meador, M.R., 2010. GAGES: A stream gage database for evaluating natural and altered flow conditions in the conterminous United States. Ecology 91 (2), 621. Ficklin, D.L., Stewart, I.T., Maurer, E.P., 2013. Climate change impacts on streamflow and subbasin-scale hydrology in the Upper Colorado River Basin. PLoS One 8 (8), e71297. Groisman, P.Y., Knight, R.W., Karl, T.R., 2001. Heavy precipitation and high streamflow in the contiguous United States: trends in the twentieth century. Bull. Am. Meteorol. Soc. 82 (2), 219–246. Hannaford, J., Buys, G., Stahl, K., Tallaksen, L.M., 2013. The influence of decadal-scale variability on trends in long European streamflow records. Hydrol. Earth Syst. Sci. 17, 2717–2733. Hansen, J., Sato, M., Ruedy, R., Lo, K., Lea, D.W., Medina-Elizade, M., 2006. Global temperature change. Proc. Natl. Acad. Sci. 103 (39), 14288–14293. Hirsch, R.M., Archfield, S.A., De Cicco, L.A., 2015. A bootstrap method for estimating uncertainty of water quality trends. Environ. Model. Softw. 73, 148–166. Kampf, S.K., Lefsky, M.A., 2016. Transition of dominant peak flow source fromsnowmelt to rainfall along the Colorado Front Range: historical patterns, trends, and lessons from the 2013 Colorado Front Range floods. Water Resour. Res. 52, 407–422. Karl, T.R., Knight, R.W., 1998. Secular trends of precipitation amount, frequency, and intensity in the United States. Bull. Am. Meteorol. Soc. 79 (2), 231–241. Karl, T.R., Knight, R.W., Plummer, N., 1995. Trends in high-frequency climate variability in the twentieth century. Nature 377, 217–220. Katz, R.W., 2010. Statistics of extremes in climate change. Clim. Change 100 (1), 71–76.

376

Journal of Hydrology: Regional Studies 12 (2017) 363–377

K.C. Solander et al.

Katz, R.W., 2013. In: In: AghaKouchak, A., Easterling, D., Hsu, K. (Eds.), Statistical Methods for Nonstationary Extremes, Chapter 2, Extremes in a Changing Climate: Detection, Analysis, and Uncertainty, vol. 65 Springer, New York 423 p. Kendall, M.G., 1975. Rank Correlation Methods. Griffin, London. Knowles, N., Dettinger, M.D., Cayan, D.R., 2006. Trends in snowfall versus rainfall in the western United States. J. Clim. 19, 4545–4559. Kormos, P.R., Luce, C.H., Wenger, S.J., Berghuijs, W.R., 2016. Trends and sensitivities of low streamflow extremes to discharge timing and magnitude in Pacific Northwest mountain streams. Water Resour. Res. 613–615. Lins, H.F., Slack, R., 1999. Streamflow trends in the United States. Geophys. Res. Lett. 26 (2), 227–230. MacDonald, G.M., 2010. Water, climate change, and sustainability in the southwest. Proc. Natl. Acad. Sci. 107 (21), 21,256–21,262. Mann, H.B., 1945. Nonparametric tests against trend. Econometrica 13, 245–259. Mannshardt-Shamseldin, E.C., Smith, R.L., Sain, S.R., Mearns, L.O., Cooley, D., 2010. Downscaling extremes: a comparison of extreme value distributions in pointsource and gridded precipitation data. Ann. Appl. Stat. 6 (1), 484–502. Mantua, N., Tohver, I., Hamlet, A., 2010. Climate change impacts on streamflow extremes and summertime stream temperature and their possible consequences for freshwater salmon habitat in Washington State. Clim. Change 102 (1–2), 187–223. McCabe, G.J., Clark, M.P., 2005. Trends and variability in snowmelt runoff in the western United States. J. Hydrometeor. 6 (4), 476–482. McCabe, G.J., Wolock, D.M., 2002. A step increase in streamflow in the conterminous United States. Geophys. Res. Lett. 29 (24), 8–11. McCabe, G.J., Wolock, D.M., 2014. Spatial and temporal patterns in conterminous United States streamflow characteristics. Geophys. Res. Lett. 41, 6889–6897. McDowell, N.G., Bennett, K.E., Michaletz, S., Solander, K.C., Xu, C., Maxwell, R.M., Allen, C., Middleton, R.S., 2017. Predicting chronic climate-driven disturbances and their mitigation. Trends in Ecol. Evol (in Review). Meinshausen, M., Smith, S.J., Calvin, K., Daniel, J.S., Kainuma, M.L.T., Lamarque, J.-F., Matsumoto, K., Montzka, S.A., Raper, S.C.B., Riahi, K., Thomson, A., Velders, G.J.M., van Vuuren, D.P.P., 2011. The RCP greenhouse gas concentrations and their extensions from 1765 to 2300. Clim. Change 109, 213–241. Miller, W.P., Piechota, T.C., 2008. Regional analysis of trend and step changes observed in hydroclimatic variables around the Colorado River Basin. J. Hydrometeor. 9 (5), 1020–1034. Milly, P.C.D., Betancourt, J., Falkenmark, M., Hirsch, R.M., Kundzewicz, Z.W., Lettenmaier, D.P., Stouffer, R.J., 2008. Climate change, Stationarity is dead: whither water management? Science 319, 573–574. Mishra, A.K., Singh, V.P., Özger, M., 2011. Seasonal streamflow extremes in Texas river basins: uncertainty, trends, and teleconnections. J. Geophys. Res. Atmos. 116 (8), 1–28. Mote, P.W., 2006. Climate-driven variability and trends in mountain snowpack in western North America. J. Clim. 19 (23), 6209–6220. Murphy, K.W., Ellis, A.W., 2014. An assessment of the stationarity of climate and stream flow in watersheds of the Colorado River Basin. J. Hydrol. 509, 454–473. Painter, T.H., Deems, J.S., Belnap, J., Hamlet, A.F., Landry, C.C., Udall, B., 2010. Response of Colorado River runoff to dust radiative forcing in snow. Proc. Natl. Acad. Sci. 107 (40), 17125–17130. Piechota, T., Timilsena, J., Tootle, G., Hidalgo, H., 2004. The western US drought: how bad is it? EOS 85 (32), 301–304. Prein, A.F., Rasmussen, R.M., Ikeda, K., Liu, C., Clark, M.P., Holland, G.J., 2016. The future intensification of hourly precipitation extremes. Nat. Clim. Change 7, 48–52. Rajagopalan, B., Nowak, K., Prairie, J., Hoerling, M., Harding, B., Barsugli, J., Ray, A., Udall, B., 2009. Water supply risk on the Colorado River: can management mitigate? Water Resour. Res. 45, 1–7. Regonda, S.K., Rajagopalan, B., 2004. Seasonal cycle shifts in hydroclimatology over the western United States. J. Clim. 18, 372–384. Richer, E.E., Kampf, S.K., Fassnacht, S.R., Moore, C.C., 2013. Spatiotemporal index for analyzing controls on snow climatology: application in the Colorado Front Range. Phys. Geogr. 34, 85–107. Richter, B.D., Baumgartner, J.V., Powell, J., Braun, D.P., 1996. A method for assessing hydrologic alteration within ecosystems. Conser. Biol. 10 (4), 1163–1174. Richter, B.D., Mathews, R., Harrison, D.L., Wigington, R., 2003. Ecologically sustainable water management: managing river flows for ecological integrity. Ecol. Appl. 13, 206–224. Sagarika, S., Kalra, A., Ahmad, S., 2014. Evaluating the effect of persistence on long-term trends and analyzing step changes in streamflows of the continental United States. J. Hydrol. 517, 36–53. Salas, J.D., Obeysekera, J., 2014. Revisiting the concepts of return period and risk for nonstationary hydrologic extreme events. J. Hydrol. 19, 554–568. Sen, P.K., 1968. Estimates of the regression coefficient based on Kendall’s tau. J. Am. Stat. Assoc. 63, 1370–1389. Stewart, I.T., Cayan, D.R., Dettinger, M.D., 2005. Changes toward earlier streamflow timing across western North America. J. Clim. 1136–1155. Stewart, I.T., Ficklin, D.L., Carrillo, C.A., McIntosh, R., 2015. 21 st century increases in the likelihood of extreme hydrologic conditions for the mountainous basins of the southwestern United States. J. Hydrol. 529 (P1), 340–353. Szilagyi, J., Katul, G.G., Parlange, M.B., 2001. Evapotranspiration intensifies over the conterminous United States. J. Water Resour. Plan. Manag. 127 (6), 354–362. Taylor, R.G., Scanlon, B., Döll, P., Rodell, M., van Beek, R., Wada, Y., Longuevergne, L., Leblanc, M., Famiglietti, J.S., Edmunds, M., Konikow, L., Green, T.R., Chen, J., Taniguchi, M., Bierkens, M.F.P., MacDonald, A., Fan, Y., Maxwell, R.M., Yechieli, Y., Gurdak, J.J., Allen, D.M., Shamsudduha, M., Hiscock, K., Yeh, P.J.-F., Holman, I., Treidel, H., 2013. Ground water and climate change. Nat. Clim. Change 3 (4), 322–329. Villarini, G., Serinaldi, F., Smith, J.A., Krajewski, W.F., 2009. On the stationarity of annual flood peaks in the continental United States during the 20th century. Water Resour. Res. 45 (8), 1–17. Woodhouse, C.A., Meko, D.M., MacDonald, G.M., Stahle, D.W., Cook, E.R., 2010. A 1,200-year perspective of 21st century drought in southwestern North America. Proc. Natl. Acad. Sci. 107 (50), 21283–21288. Woodhouse, C.A., Pederson, G.T., Morino, K., Mcafee, S.A., McCabe, G.J., 2016. Increasing influence of air temperature on upper Colorado River streamflow. Geophys. Res. Lett. 1–8. Zhang, X., Zweirs, F.W., 2013. In: In: AghaKouchak, A., Easterling, D., Hsu, K. (Eds.), Statistical Indices for Diagnosing and Detecting Changes in Extremes, Chapter 1, Extremes in a Changing Climate: Detection, Analysis, and Uncertainty, vol. 65 Springer, New York 423 p. Zhang, X., Harvey, K.D., Hogg, W.D., Yuzyk, T.R., 2001. Trends in Canadian streamflow. Water Resour. Res. 37 (4), 987–998. Zwiers, F.W., Zhang, X., Feng, Y., 2011. Anthropogenic influence on long return period daily temperature extremes at regional scales. J. Clim. 24 (3), 881–892.

377