HYDROLOGICAL PROCESSES Hydrol. Process. (2014) Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/hyp.10203

Elevation-dependent responses of streamflow to climate warming Christopher J. Tennant,* Benjamin T. Crosby and Sarah E. Godsey Department of Geosciences, Idaho State University, 921 S. 8th Ave. Stop 8072, Pocatello, ID, 83209-8072, USA

Abstract: Warming will affect snowline elevation, potentially altering the timing and magnitude of streamflow from mountain landscapes. Presently, the assessment of potential elevation-dependent responses is difficult because many gauged watersheds integrate drainage areas that are both snow and rain dominated. To predict the impact of snowline rise on streamflow, we mapped the current snowline (1980 m) for the Salmon River watershed (Idaho, USA) and projected its elevation after 3 °C warming (2440 m). This increase results in a 40% reduction in snow-covered area during winter months. We expand this analysis by collecting streamflow records from a new, elevation-stratified gauging network of watersheds contained within high (2250–3800 m), mid (1500–2250 m) and low (300–1500 m) elevations that isolate snow, mixed and rain-dominated precipitation regimes. Results indicate that lags between percentiles of precipitation and streamflow are much shorter in low elevations than in mid- and high-elevation watersheds. Low elevation annual percentiles (Q25 and Q75) of streamflow occur 30–50 days earlier than in higher elevation watersheds. Extreme events in low elevations are dominated by low- and no-flow events whereas mid- and high-elevation extreme events are primarily large magnitude floods. Only mid- and high-elevation watersheds are strongly cross correlated with catchment-wide flow of the Salmon River, suggesting that changes in contributions from low-elevation catchments may be poorly represented using mainstem gauges. As snowline rises, mid-elevation watersheds will likely exhibit behaviours currently observed only at lower elevations. Streamflow monitoring networks designed for operational decision making or change detection may require modification to capture elevation-dependent responses of streamflow to warming. Copyright © 2014 John Wiley & Sons, Ltd. KEY WORDS

snowline; streamflow; climate change; rain–snow transition; hypsometry

Received 25 July 2013; Accepted 19 March 2014

INTRODUCTION Mountainous terrain gives rise to some of the most pronounced climatic gradients on Earth. It is not uncommon for high elevations to receive as much as five times more precipitation than nearby low-elevation topography (Sinclair, 1994; Katzfey, 1995a,1995b; Roe, 2005). Large temperature gradients and orography generate strong contrasts in the amount and phase of precipitation and the duration of snow cover at different elevations (Rolland, 2003; Lundquist and Cayan, 2007; Minder et al., 2010). Heterogeneous patterns of precipitation and energy in mountain systems (Niu and Yang, 2004; Beniston, 2006) are also influenced by high spatial variability in topographic characteristics such as slope and aspect (Poulos et al., 2012), as well as biotic characteristics such as vegetation type and spatial extent (Niu and Yang, 2004). The meteorological and topographic characteristics typical of mountainous terrain produce strong contrasts in the timing, magnitude and frequency of streamflow events within different elevation zones (Hunsaker et al., 2012), however, few studies consider *Correspondence to: Christopher J. Tennant, Department of Geosciences, 921 S. 8th Ave. Stop 8072, Idaho State University, Pocatello, ID 83209-8072, USA. E-mail: [email protected]

Copyright © 2014 John Wiley & Sons, Ltd.

how this elevation-driven heterogeneity complicates our understanding of how rising snowlines could modify elevation-dependent streamflow patterns. Recent warming (Intergovernmental Panel on Climate Change (IPCC), 2008, 2009) has caused an increase in rain during winter months (Knowles et al., 2006), reduced spring snowpack (Mote et al., 2005; Bedford and Douglass, 2008; Stewart, 2009) and changed the timing and magnitude of peak and mean annual flows (Stewart et al., 2005; Luce and Holden, 2009; Stewart, 2009). These trends could be exacerbated by anticipated global warming in the range of 2 to 5 °C (Intergovernmental Panel on Climate Change (IPCC), 2008, 2009). In cold regions, this warming results in earlier and reduced runoff as more precipitation falls as rain instead of snow (Christensen and Lettenmaier, 2007; Elsner et al., 2010; Jin and Sridhar, 2012; Tang and Lettenmaier, 2012; Tang et al., 2012; Vano et al., 2012). These changes have the potential to increase drought and wildfire activity (Westerling et al., 2003), impact water management practices (Mote et al., 2003) and alter the character of disturbances in snowmelt-dependent ecosystems (Hauer et al., 1997; Rieman et al., 2007; Davis et al., 2013; Isaak and Rieman, 2013).

C. J. TENNANT ET AL.

Most studies of climate-driven changes in streamflow typically employ discharge records from large, mainstem channels in order to detect trends or to calibrate hydrologic models (e.g. Muttiah and Wurbs, 2002; Barnett et al., 2004; Stewart et al., 2005; Christensen and Lettenmaier, 2007; Hamlet and Lettenmaier, 2007; Rauscher et al., 2008; Luce and Holden, 2009; Stewart, 2009; Fritze et al., 2011; Tang and Lettenmaier, 2012). These mainstem mountain channels typically integrate waters sourced from a large range of elevations where there is high spatial variability in the amount, dominant phase and timing of precipitation as well as its delivery to the channel. While discharge records from large drainage area channels will remain useful for understanding how integrated land areas respond to changes, there is little research evaluating whether mainstem signals are representative of processes occurring throughout the basin or only in discrete regions. We expect that streamflow records from mainstem channels limit our ability to (1) isolate and map the diversity of precipitation and streamflow conditions that exist within mountain watersheds, (2) clearly understand how precipitation phase influences local streamflow characteristics; and (3) make accurate predictions of how changes in the snowline elevation will affect streamflow patterns and water resource availability in different elevation zones in mountainous topography. To address these limitations, we designed and implemented a streamflow monitoring network in elevation-stratified tributaries of the Salmon River watershed (SRW) located in central Idaho, USA (Figure 1). The network was specifically designed to measure streamflow from high-, mid- and low-elevation catchments that isolate regions where precipitation is either snow dominated, a mixture of rain and snow, or rain dominated, respectively. Collectively, our study catchments span the full relief of the

SRW, yet unlike mainstem gauging networks, the discharge records from our network are sourced from distinct elevation bands/precipitation regimes (Figure 1a). We explore elevation-dependent streamflow responses in three ways. We first evaluate the sensitivity of the SRW to loss of snow-covered area by identifying the contemporary snowline elevation and its potential shift after 3 °C of warming. Second, we compare this sensitivity analysis of snow-covered area with streamflow patterns observed in our elevation-stratified network. We discuss differences in precipitation and streamflow under current climate conditions and the anticipated changes under a warmer climate. Finally, we demonstrate that the not all elevation zones within the SRW are well represented by the existing regional streamflow network. In response to these findings, we discuss the suitability of mainstem gauge networks for making predictions regarding climate warming impacts on streamflow patterns.

NETWORK DESIGN AND SITE CHARACTERISTICS Network design

High-relief watersheds are characterized by large temperature gradients (Smith, 1979; Roe, 2005) that result in varying proportions of rain and snow at different elevations. To limit this variability, we selected tributary watersheds contained within limited elevation ranges that isolate different temperature regimes (Figure 1a). These 12 watersheds are evenly divided between high-, midand low-elevation zones and differentiate basins that are either snow dominated, receive a mixture of rain and snow or are rain dominated, respectively. Though all 12 selected watersheds cross the defined boundaries of individual precipitation regimes (Figure 1a), the mean,

Figure 1. (a) Spindle plots of elevation distributions for our 12 study watersheds and the entire Salmon River watershed. Within each spindle, the mean (white square) and median elevations (grey circle) are displayed. The current snowline (~1980 m, dashed line) and its potential rise (to ~2440 ± 80 m with 3 ± 0.5 °C warming, dotted line with light shading showing uncertainty) are shown. (b) Color-shaded map showing aerial extent of precipitation domains, the locations of study watersheds, the Salmon River and its major tributaries (black lines), the USGS Salmon River at Whitebird, ID mainstem gauge (circle with shaded cross in upper northwest corner; gauge elevation is 430 m) and our snow-dominated (SD, stars), mixed rain and snow (MRS, squares) and rain-dominated (RD, triangles) study sites. The location of the Salmon River watershed in central Idaho is shown beneath the color shaded map

Copyright © 2014 John Wiley & Sons, Ltd.

Hydrol. Process. (2014) DOI: 10.1002/hyp

ELEVATION-DEPENDENT RESPONSES OF STREAMFLOW

median and interquartile range of each basin is within the specified range. Verification of the meteorological divisions of watersheds as snow dominated, mixed rain and snow and rain dominated is provided below. Regional setting

The SRW (Figure 1) and its tributaries have no major dams, diversions or withdrawals, and there is little direct human alteration of the landscape. The mainstem Salmon is one of the longest free-flowing rivers in the contiguous US, second only to the Yellowstone River. Streamflow within the basin provides a natural laboratory to investigate its sensitivity to changes caused by increases in air temperature (Kunkel and Pierce, 2010; Tang et al., 2012; Davis et al., 2013; Sridhar et al., 2013). Much of the SRW is underlain by Idaho batholith granites and a mixture of metamorphic, sedimentary and extrusive volcanic rocks comprise the remainder of the basin (Lund, 2004; Lewis et al., 2012). The USGS currently operates 23 streamflow gauges within the SRW, 14 of which are located on mainstem rivers draining high-relief topography that mixes inputs from multiple phases of precipitation. The remaining 9 gauges primarily measure streamflow from high-elevation, snow-dominated watersheds. There are currently 22 SNOTEL stations in operation within the SRW; 14 of these stations are located at elevations greater than 2000 m. The lowest elevation station is at 1600 m, and the mean elevation of all 22 sites is 2600 m. Because most of these sites are located well within snow-dominated elevations, they provide little information about conditions at elevations near the transient snowline, the region most likely to be sensitive to future warming (Trujillo et al., 2012). The Pacific Northwest and Intermountain West is predicted to warm as much as 3 to 5 °C by 2090 (Muttiah and Wurbs, 2002; Christensen and Lettenmaier, 2007; Sridhar et al., 2013), which motivates our exploration of elevationdependent streamflow patterns and their sensitivity to warming. We explore how these potential changes could alter streamflow in high- (~35% of SRW drainage area), mid- (~50% of SRW drainage area) and low-elevation (~15% of SRW drainage area) domains. Snow-dominated watersheds (SD)

The snow-dominated (SD) watersheds selected for this study are contained within the upper elevations of the SRW (~2250–3800 m; Figure 1a and b) and receive an average annual precipitation of 1120 mm. Approximately 70% (± 6%) of this precipitation falls as snow, with a mean annual peak snow water equivalent (SWE) of 670 mm (Figure 2; Table I). Peak SWE is lower than the total annual snowfall because some snow melts or sublimates during the winter. The percentages of precipitation falling as rain or snow (~ 30% and 70%, respectively) are Copyright © 2014 John Wiley & Sons, Ltd.

Figure 2. (a) Mean daily snow water equivalent and (b, inset) the percentages of precipitation that fall as rain or snow (indicated by solid colors and snowflakes, respectively) for the SD, MRS and RD watersheds between 2004 and 2011. Note that the axis for snow water equivalent in the RD watersheds is reduced by a factor of 20 compared to those for the MRS and SD watersheds. Also note that the dotted standard deviation around the mean daily snow water equivalent in RD watersheds always crosses the zero line, indicating that even when these catchments experience snow cover, they are not completely snow covered; the lowest elevations remain snow free

calculated based on the cumulative totals during a year (Table I). The uplands of these watersheds are either bare rock or well forested, with lodgepole pine dominating (Whitlock et al., 2011); the lower valley floors are vegetated by grasses and sage, with riparian zones composed of various species of alder and willow. The SD watersheds generally have the highest mean slope and the greatest area with slopes greater than 50% (Table I). While high slopes are present in the watershed uplands, our gages are located in wide, low-slope, open valley bottoms that have experienced several episodes of glaciation between the Pliocene and early Holocene (Meyer et al., 2004; Thackray et al., 2004), and are thus filled with thick deposits of coarse-grained alluvium. We recognize that because our gages are located in regions of highly permeable alluvium some portion may be lost to shallow groundwater, thus leading to flow underestimation. The mean annual temperature in the SD catchments is ~2 °C. More importantly, the average temperature during the dominant precipitation season (December–May) is 4 °C; as a consequence, warming in the range of 3 to 5 °C is likely to increase the proportion of rain during winter and spring months. Mixed-rain-and-snow watersheds (MRS)

The mixed-rain-and-snow (MRS) watersheds are contained within mid-elevation zones (~1500 m–2250 m; Figure 1a and b) and receive a mean annual precipitation of 987 mm, ~50% (± 5%) of which falls as snow resulting in a mean annual peak SWE of 304 mm (Figure 2; Table I). Higher elevation portions of these watersheds are consistently snow covered through winter months, while lower elevations experience transient snow cover. The MRS watersheds have the greatest relief (average of 1286 m) (Figure 1b) and the greatest percent of forest cover of the 3 regimes (Table I). Densely forested areas are predominantly at higher elevations and are composed Hydrol. Process. (2014) DOI: 10.1002/hyp

Copyright © 2014 John Wiley & Sons, Ltd.

47

40

36

Salmon headwaters

Smiley

Snow-dominated mean

34

43

55

N.F. Slate

Slate

Mixed regime mean

1092

1067

1216

1106

978

1717

1862

1591

1695

1719

2540

2557

2563

2527

2514

1700

1871

1786

1734

1408

2327

2542

1978

2323

2466

3094

3109

3170

2987

3108

(m)

Max. elev.

438

421

402

472

457

1041

1070

872

1289

933

2216

2216

2240

2234

2173

(m)

Min. elev.

a

35

18

34

51

36

33

38

37

24

31

40

40

38

43

37

(%)

Mean slope

23

14

30

23

24

85

87

87

85

80

63

60

67

68

57

(%)

Forest area

480 ± 107

487 ± 128

427 ± 100

430 ± 92

576 ± 117

987 ± 319

1097 ± 380

953 ± 291

1005 ± 351

894 ± 280

1120 ± 485

1134 ± 490

1068 ± 469

1098 ± 480

1181 ± 117

(mm)

Total precip.

Rain

373 ± 82

379 ± 99

336 ± 78

345 ± 74

433 ± 83

471 ± 151

503 ± 170

522 ± 131

425 ± 172

434 ± 146

351 ± 243

356 ± 244

340 ± 237

345 ± 242

365 ± 251

(mm)

106 ± 59

107 ± 61

91 ± 59

85 ± 53

142 ± 65

516 ± 204

778 ± 299

430 ± 188

580 ± 220

459 ± 165

769 ± 293

778 ± 299

728 ± 278

754 ± 287

816 ± 311

(mm)

Snow

29 ± 9

39 ± 22

23 ± 18

21 ± 14

34 ± 17

304 ± 110

306 ± 186

176 ± 112

439 ± 112

424 ± 121

670 ± 190

693 ± 209

579 ± 170

668 ± 176

733 ± 222

(mm)

Max SWE

characteristicsb

Meteorological

Date of max SWE

Jan 06 ± 12

Jan 13 ± 30

Dec 27 ± 43

Dec 21 ± 46

Jan 12 ± 32

Mar 15 ± 23

Mar 17 ± 43

Feb 19 ± 45

Mar 24 ± 12

Mar 29 ± 15

Apr 15 ± 15

Apr 15 ± 17

Apr 11 ± 15

Apr 15 ± 17

Apr 18 ± 11

(± number of days)

Total

165 ± 65

122 ± 45

241 ± 142

192 ± 100

94 ± 26

556 ± 149

785 ± 78

486 ± 277

431 ± 185

523 ± 60

493 ± 26

562 ± 58

342 ± 54

460 ± 21

609 ± 56

(mm)

Min.

0.01 ± 0.01

0.02 ± 0.02

0.03 ± 0.02

0.002 ± 0.004

0.0002 ± 0.0003

0.24 ± 0.01

0.28 ± 0.10

0.14 ± 0.02

0.16 ± 0.07

0.37 ± 0.14

0.11 ± 0.02

0.11 ± 0.03

0.16 ± 0.02

0.09 ± 0.04

0.06 ± 0.03

(mm/day)

b

Elevation data (10 m) from USGS National Elevation Dataset, percent forest area from USGS StreamStats calculated using the National Land Cover Dataset 2006 Meteorological parameters calculated using SNODAS products, 2004-2011 (see Meteorological Methods for details) c Flow characteristics are for the 2009-2011 water years

a

79

234

Rock

Rain regime mean

48

8

Gregory

Rice

25

Baker

Rain dominated

42

Little Goose

Boulder

102

18

Mixed rain and snow

39

Frenchman

(m)

Beaver

Snow dominated

Mean elev.

Area

(km2)

characteristics

Physical

0.22 ± 0.16

0.18 ± 0.12

0.33 ± 0.24

0.26 ± 0.20

0.13 ± 0.06

0.48 ± 0.10

0.60 ± 0.08

0.44 ± 0.23

0.25 ± 0.06

0.61 ± 0.13

0.40 ± 0.02

0.40 ± 0.06

0.40 ± 0.07

0.42 ± 0.02

0.39 ± 0.08

(mm/day)

Median

c

4.1 ± 1.1

3.4 ± 1.6

5.4 ± 1.8

5.3 ± 0.73

2.2 ± 0.73

13.1 ± 5.8

17.0 ± 1.8

14.3 ± 10.4

11.8 ± 5.2

9.3 ± 2.3

15.1 ± 2.5

18.5 ± 3.6

11.9 ± 3.0

11.7 ± 1.5

18.3 ± 2.2

(mm/day)

Peak

characteristics

streamflow

Mean annual

May 04 ± 34

Apr 29 ± 47

Apr 26 ± 33

Apr 25 ± 42

May 25 ± 18

May 28 ± 10

Jun 06 ± 2

Jun 01 ± 11

May 14 ± 21

May 31 ± 11

Jun 12 ± 11

Jun 11 ± 11

Jun 11 ± 12

Jun 12 ± 11

Jun 12 ± 65

(± number of days)

Date of peak

Table I. Physical, meteorological and runoff characteristics for the snow-dominated (SD), mixed rain and snow (MRS) and rain-dominated (RD) watersheds

C. J. TENNANT ET AL.

Hydrol. Process. (2014) DOI: 10.1002/hyp

ELEVATION-DEPENDENT RESPONSES OF STREAMFLOW

of a mixture of ponderosa pine, lodgepole pine and Douglas fir (Doerner and Carrara, 1999). The mean annual temperature of these watersheds is ~6 °C, with average December–May temperatures around 0 °C. Expected warming of 3 to 5 °C will likely change the dominant form of precipitation from a nearly equal mixture of rain and snow to one dominated by rain. Rain-dominated watersheds (RD)

The rain-dominated (RD) watersheds are located in lowelevation regions of the SRW (~300 m–1500 m; Figure 1a and b) and receive a mean annual precipitation of 480 mm with ~20% (± 5%) in the form of snow; the mean annual peak SWE is 29 mm (Figure 2; Table I). Snowfall in these watersheds primarily occurs in the upper elevations and persists only for short periods of time. Vegetation is primarily composed of grasslands and deciduous tree species, with some conifers in the uplands. The raindominated watersheds have the lowest amount of forestcovered area (Table I). The mean annual temperature of the RD watersheds is ~10 °C, with temperatures between December and May averaging ~5 °C. Predicted warming will likely eliminate the scarce amount of snow that falls within the highest elevations of these watersheds.

METHODS Meteorological methods

Because this study demanded finer spatial scale precipitation data than is available through instrumental records, we used the National Weather Service’s Snow Data Assimilation System (SNODAS) (Carroll et al., 2001; Barrett, 2003; Carroll et al., 2003). This freely available, spatially distributed climate product was used to estimate the meteorological inputs to our study watersheds at 1 km2 resolution and to identify the snowline elevation for the SRW. The model utilizes an energy and mass-balance approach that synthesizes observations from ground stations, satellites and airborne platforms to provide daily estimates of precipitation and snowpack. Data from SNODAS grids were clipped to the extents of our study watersheds to derive daily values for SWE, liquid, and solid precipitation. To confirm the accuracy of these data, we compared the modeled precipitation and SWE values from SNODAS against the closest available instrumental records and found good agreement in the amount and timing of meteorological events; annual totals were within ±10% and the timing of peaks occurred on the same days between 2004 and 2011. Throughout this paper, we report SNODAS values for each precipitation domain that are calculated in a two-step process. We calculated the daily mean value of all cells contained within each study watershed and then calculated the mean of the 4 watersheds Copyright © 2014 John Wiley & Sons, Ltd.

in each precipitation zone. This was done to maximize the representativeness of our precipitation characterization and to minimize the effect of any spurious model cells. The average annual snowline for the SRW (water years 2004–2011) was calculated by identifying the elevation where there is a 50% probability of encountering an annual average SWE value greater than zero. Hantel and Maurer (2011) defined the snowline in a similar way to address the transience of snow cover associated with individual storm events. Evaluating the snowline on an annual basis provides an unbiased and relevant metric for how climate affects streamflow patterns. To calculate the snowline, daily SNODAS grids of SWE were paired with corresponding elevation grids at the same resolution. The probability of snow cover was calculated by identifying the number of days where SWE was greater than zero for each year. We then calculated the average probability as the mean number of snow-covered days per year for all points within each elevation range. Our calculations of snowline elevation and precipitation metrics (mean annual totals, mean annual SWE and the proportions of rain and snow) were calculated from SNODAS data from water years 2004 through 2011. Average annual and winter temperatures were calculated using the closest Cooperative Observer Program stations, run by the National Oceanic and Atmospheric Administration (records used are from 1950 to 2006). Hydrologic methods

To monitor surface flow, we installed vented pressure loggers (In-Situ Level Troll 500) in 2-in diameter perforated stilling wells that recorded stage level at 10-min intervals. Pressure loggers were installed in the spring of 2009 and were maintained throughout the 2011 water year. We used an YSI-Flowtracker Acoustic Doppler Velocimeter to measure discharge within the channel at a variety of flow conditions following the methods outlined by Blanchard (2004). Erroneous stage data were removed, and any offsets in stage data caused by changes in channel geometry (i.e. scour or deposition) were shifted to match the starting datum. Power law regressions were used to calibrate the relationship between stage and discharge as described by Rantz (1982). To assure robust stage-discharge relations >20 discharge measurements, spanning high- and low-flow conditions were made over 3 years at each site. To facilitate comparisons against USGS streamflow and SNODAS data, we integrated the 10-min data into daily discharge measurements, as reported in this paper. RESULTS AND DISCUSSION In this section, we discuss snowline sensitivity to changing climate, highlight contemporary differences in Hydrol. Process. (2014) DOI: 10.1002/hyp

C. J. TENNANT ET AL.

streamflow from our elevation-stratified monitoring network (Figure 1a) and then couple a space-for-time substitution (Budyko, 1974; Dooge, 1992; Dooge et al., 1999; Harman et al., 2011; Sivapalan et al., 2011) with our snowline analysis to explore how warming could alter elevation-dependent streamflow patterns. We close by demonstrating that our elevation-stratified streamflow gauging network offers insights into elevation-dependent streamflow patterns (Figure 4) that are not well represented in existing regional gauging networks. Snowline elevation and the effect of a warmer climate

Changes in the elevation of the snowline and its variability through time will influence streamflow patterns and their local response to a warmer climate. To better estimate this sensitivity, we mapped the snowline elevation of the SRW and how it may shift in response to predicted warming. The average snowline elevation from 2004 to 2011 was ~1980 m (Figure 3). On an annual basis, any elevation above this value has more than a 50% probability of having snow cover. Using a conservative estimate of 3 °C warming (Muttiah and Wurbs, 2002; Christensen and Lettenmaier, 2007; Sridhar et al., 2013) and typical lapse rates of 0.65 °C per 100 m (Barry and Chorley, 2003; Blandford et al., 2008), the snowline is likely to rise from its current location of ~1980 m to an elevation of ~2440 m (± 80 m with ± 0.5 °C uncertainty; Figure 1a). Under the current snowline elevation ~60% of the SRW has a >50% probability of being snow covered on an annual basis. In other words, on average, 60% of the watershed area is snow covered for at least 6 months out of the year. If warming of 3 °C drives the snowline to an elevation of 2441 m, only 18% of the SRW area will have a 50% or greater probability of being snow-covered for

Figure 3. Probability that average annual snow water equivalent is greater than zero at a particular elevation in the Salmon River watershed based on the method proposed by (Hantel and Maurer, 2011). Probabilities were determined by calculating the average fraction of days per year with snow cover at each unique grid elevation for each water year from 2004 to 2011. The elevation where there is a 50% probability of snow water equivalent greater than zero is defined as the snowline elevation (~1980 m). The regression line is a 2nd-degree polynomial and has an adjusted R squared of 0.97

Copyright © 2014 John Wiley & Sons, Ltd.

>6 months of the year (Figure 1a). The hypsometry (elevation–area relationship) of individual watersheds (Figure 1a) will likely exert a strong influence on watershed-specific responses to warming and snow cover loss. The diverse hypsometries of our study catchments (Figure 1a) illustrate that measured changes in streamflow patterns for the whole Salmon River may under represent the dramatic changes occurring in some sub-watersheds that function as important water sources for anthropogenic and ecological systems. It is important to point out that there is uncertainty in our snow line shift estimate related to seasonal and synoptic variation in lapse rate (Lundquist and Cayan, 2007; Blandford et al., 2008; Lundquist and Lott, 2008; Minder et al., 2010), climatic variations that influence orographic precipitation enhancement (Luce et al., 2013) and how these variables could change under warmer conditions (Intergovernmental Panel on Climate Change (IPCC), 2008, 2009). The potential for these factors to vary make a precise prediction of snowline increase difficult, nonetheless, we expect that the lapse rate provides a credible, first-order metric from which to predict changes in the snowline. Comparison of streamflow metrics across elevation bands Number and magnitude of hydrograph peaks. Highelevation, SD catchments generally have one dominant snowmelt hydrograph peak whereas hydrographs for the mid-elevation MRS, and especially low-elevation RD regimes, have multiple peaks (Figure 4). Streamflow patterns in the SD watersheds are characterized by long and distinct rising and falling limbs, which tend to remain unpunctuated by ‘flashy’ rain events (Figure 4a). The large pulse of spring snowmelt is preceded and followed by baseflow conditions and typically results in a dominant singular peak in the annual hydrograph. In contrast, hydrographs for MRS watersheds tend to have two or more significant peaks in discharge. Observations of streamflow and precipitation indicate that earlier and more frequent rain, as well as a higher number of snowmelt events, drives the multi-peaked nature of spring melt in mid-elevation MRS catchments (Figure 4a and b). Peak streamflow in the SD watersheds tends to be slightly higher (Table I), which is likely caused by the release of a large volume of snowpack over a short time frame (Figure 4a and b). Streamflow patterns in the RD watersheds exhibit significant contrasts from the MRS and SD catchments. Streamflow events are numerous and of short duration; streamflow values start low, quickly rise, peak and return to baseflow values (Figure 4c). While a seasonal rise and fall in baseflow conditions are observable in RD hydrographs, it is much less pronounced than the seasonal baseflow increases in MRS and SD catchments (Figure 4). Hydrol. Process. (2014) DOI: 10.1002/hyp

ELEVATION-DEPENDENT RESPONSES OF STREAMFLOW

Figure 4. Streamflow, snow water equivalent (SWE) and rainfall for three years in the (a) SD, (b) MRS and (c) RD watersheds. Shading around mean SWE shows the standard deviation. Note that the scale for streamflow for the RD watersheds is reduced by a factor of 3 compared to the SD and MRS watersheds. Differences in precipitation phase and amount strongly affect the shape of the hydrographs

Copyright © 2014 John Wiley & Sons, Ltd.

Hydrol. Process. (2014) DOI: 10.1002/hyp

C. J. TENNANT ET AL.

As warming increases, higher probabilities of rapid melt or rain-on-snow events will lead to flashier flows, potentially altering the frequency of sediment transport events and magnitude of bed scour (Hassan et al., 2006; Goode et al., 2013) in mid to high elevations. Lags between precipitation and streamflow percentiles. Lags between percentiles of precipitation and streamflow provide a metric to evaluate storage (surface or subsurface) of precipitation and are important in gaining insights into variability of precipitation routing to stream channels in different elevation zones. We interpret the lags reported below as a function of the dominant precipitation phase. The lag between the 25th percentile of precipitation and the 25th percentile of streamflow provides an indication of how rapidly early season precipitation is transformed to streamflow. These lags are 157, 138 and 86 days for the SD, MRS and RD regimes, respectively (comparing Figure 5a to b). These times are generally much longer than lags between the 75th percentiles which are 79, 56 and 3 days, respectively. Because the lags include the storage and melt times for snow, greater snowpack leads to longer melt times and thus longer lags between precipitation and streamflow. The lag between 75th percentiles for the RD regime is much shorter because the bulk of flows are fed by rain, with little snowpack storage to delay the response time. If lag times shorten considerably as more of the landscape transitions to higher proportions of rain, it will be valuable to explore whether water residence times will also shift, potentially leading to changes in chemical weathering rates and water quality.

requires 46 days and in the RD catchments, it requires 72 days. RD gages receive the 25th percentile of flow in April, followed by the MRS catchments in early May and the SD catchments in late May. These differences suggest that under a warmer scenario MRS and SD watersheds could experience an earlier rise in streamflow by as much as 30–50 days (Figure 5b). Changes in the timing and rates of water movement through the landscape and stream channels could have strong impacts on the resilience and dynamics of snow-melt dominated ecosystems by altering the timing of phenological events and increasing streambed scour during winter months (Hauer et al., 1997; Rieman et al., 2007; Davis et al., 2013; Goode et al., 2013; Isaak and Rieman, 2013).

Timing of streamflow. Streamflow passes through the SD catchments more quickly and later in the water year than in the MRS and RD catchments (Figure 4 and 5b). On average, the inter-quartile range (IQR) of annual streamflow in the SD catchments passes in 36 days whereas the IQR of streamflow in the MRS catchments

Extreme flow events. Extreme flow events are typically classified based on their frequency, intensity or severity (Beniston et al., 2007). Understanding the type(s) of extremes that occur within different elevation zones is important for adapting water management plans to meet the needs of human and ecological systems. We use a statistical approach to define extreme events; streamflow outliers are defined as values below Q25 [1.5 × (Q75 Q25)] or above Q75 + [1.5 × (Q75 Q25)], where Qn is the nth percentile of flow (Tukey, 1977). Both SD and MRS watersheds exhibit a large number of high-flow outliers during the spring and early summer (Figure 3a and b). As a consequence, most of the annual streamflow occurs during a relatively brief period. Rain-dominated catchments, on the other hand, only exhibit low-flow outliers (Figure 6). We attribute the increase in low-flow outliers in low-elevation zones to lower amounts of precipitation rather than longer recession times. (Table I). We found the duration of hydrograph recession (time of peak flow – time of minimum flow) to be uncorrelated with the number of low-flow outliers. The SD watersheds also exhibit a small group of outliers at low-flow related to winter freezing events where only a small portion of the

Figure 5. Cumulative distribution plots for the average timing of delivery of (a) precipitation and (b) streamflow. Values were calculated by taking the mean for (a) precipitation or (b) streamflow for each day of the year for water years 2009, 2010 and 2011. Thicker line sections show the inter-quartile range for precipitation and streamflow. Note that this range is transmitted over a much shorter time period in the MRS and SD catchments

Figure 6. Boxplots of streamflow for the mainstem Salmon River at Whitebird, ID, and the SD, MRS and RD watersheds. Large black circles display the median value, the edges of the boxes show the 25 and 75 percentiles (left and right edges respectively), the whiskers extend to 1.5 times the inter-quartile range and any values that fall outside of this range are considered outliers. Note that the x-axis is log scale

Copyright © 2014 John Wiley & Sons, Ltd.

Hydrol. Process. (2014) DOI: 10.1002/hyp

ELEVATION-DEPENDENT RESPONSES OF STREAMFLOW

channel sustains flow (Figure 6). Low baseflow at all sites in early fall is common because of minimal precipitation from June to Oct. (Figure 4a). Like the SD watersheds, the MRS watersheds exhibit a right-skewed distribution, but with far fewer high-flow outliers. Higher proportions of rain in the MRS catchments and warmer winter and spring temperatures increase the probability of water delivery to stream channels throughout the wet winter-spring season. Because flows are more evenly distributed throughout the year, the flush of spring snowmelt has a smaller impact on streamflow than in SD watersheds and also drives mid-elevation MRS catchments to have a higher median streamflow (Figure 6). Extreme events in RD watersheds are dominated by low-flow or no-flow events, reflecting channel drying (Figure 6). Greater variability in streamflow and a high probability of channel drying suggests that aquatic ecosystems and riparian plant communities in the RD elevations of mountainous watersheds experience higher frequency of drought-driven disturbance. The characteristics of the low-elevation RD watersheds match what Poff (1996) described as intermittent-flashy streams: channels have greater than 10 days of zero flow, high daily coefficients of variation and low predictability concerning the timing of flood events (Poff, 1996). In contrast, the MRS and SD watersheds have higher magnitude flood events and a greater seasonality of streamflow (Figure 4a and b). The larger proportion of high flows in MRS and SD watersheds than in RD ones suggests that these channels have a greater capacity to adjust channel form (Dunne and Leopold, 1978). As warming proceeds and higher proportions of precipitation fall as rain, one might expect fewer flooding extremes as streamflow seasonality is reduced. One exception could be an increase of rain-on-snow events (McCabe et al., 2007). Comparison of elevation-stratified and regional gauging approaches

Regional-scale hydrologic models are used to predict climate driven changes in streamflow throughout the world (e.g. Muttiah and Wurbs, 2002; Barnett et al., 2004; Stewart et al., 2005; Christensen and Lettenmaier, 2007; Hamlet and Lettenmaier, 2007; Rauscher et al., 2008; Luce and Holden, 2009; Stewart, 2009; Fritze et al., 2011; Tang and Lettenmaier, 2012) and in the SRW by Tang et al. (2012) and Sridhar et al. (2013). Our field-based predictions of change are generally consistent with these modeling results, but we provide new insight into the fundamental differences in the timing, magnitude and patterns of streamflow within different elevation bands (see above). Based on our hypsometric analysis (see above) we argue that the magnitude of climate warming impacts on snow-dominated mountain watersheds will be elevation dependent. Copyright © 2014 John Wiley & Sons, Ltd.

Figure 7. Cross correlation of daily mean streamflow between the rain dominated (RD, dashed line), mixed rain and snow (MRS, dotted line) and snow dominated (SD, solid line) watersheds and streamflow at the mainstem Salmon River at Whitebird, ID (SR@WB, see Figure 1a for gauge locations). The offset of the peak for cross correlation between the rain-dominated and regional-scale (SR@WB) streamflow indicates that the regional gage is a poor predictor of changes in the rain-dominated portions of the Salmon River Watershed

To test how well regional streamflow gauging networks (typically focused on mainstem channels) reflect elevationdependent streamflow patterns, we tested the correlation and cross correlation of flow records between our elevation-stratified network and a regional scale record from the mainstem Salmon River at Whitebird, ID (SR@WB; Figure 1b). We find a strong correlation between streamflow at mid and high elevations and the mainstem, but a weaker correlation with the low-elevation, rain-dominated watersheds. For example, the average Pearson correlation coefficients between the mainstem and the mid and high MRS and SD watersheds are 0.95 and 0.93, respectively, whereas the mainstem and the lowelevation RD watersheds have an average correlation coefficient of 0.45. The maximum cross-correlation between streamflow at the SR@WB and the MRS and SD watersheds occurred with lags of 2 days and 0 days, with correlation coefficients of 0.97 and 0.94, respectively (Figure 7). In contrast, the RD watersheds have a much lower maximum cross-correlation coefficient (0.74), which occurs at a lag of 42 days. The timing as well as the streamflow patterns in mid and high elevations are much more strongly correlated with regional flow than low elevations where streamflow patterns not only show lower correlation with regional trends but also occur 42 days earlier in the season (Figure 7). Thus, our unique elevationstratified dataset shows regional-scale flow networks that focus on mainstem observations poorly capture the behaviour of lower elevation catchments.

CONCLUSIONS If temperatures warm by 3 °C, as climate models predict, the current snowline elevation in the SRW may rise from ~1980 to ~2440 m. This warming would lead to a 42% reduction in the snow-dominated area of the basin, Hydrol. Process. (2014) DOI: 10.1002/hyp

C. J. TENNANT ET AL.

shifting precipitation phase from snow to rain across that area. To understand how streams respond to rain versus snow, we used an elevation-stratified streamflow network to examine how such warming might affect streamflow patterns. We found that streamflow patterns, the timing of streamflow delivery, lags between precipitation and streamflow, and the type and frequency of extreme events all differed across elevation zones. We demonstrate that regional, mainstem gauges do not reflect streamflow patterns across all elevation zones. Thus, models that are calibrated or validated using mainstem flow records may not be capable of predicting lower elevation streamflow responses to warming. Increased gaging on lowerelevation, rain-dominated and mixed rain–snow streams may help to improve our ability to detect and adapt to climate change impacts in these mountainous catchments.

ACKNOWLEDGEMENTS

This work was supported by NSF award numbers EPS0814387 and EPS-1006968 from the NSF Idaho EPSCoR Program. This study’s experimental design and larger implications benefited from discussions with Dr. Colden Baxter. Drs. John Welhan and DeWayne Derryberry provided support regarding analysis and interpretation of results. Kara Gergely of the National Snow and Ice Data Center was very helpful with technical issues associated with SNODAS data. We also would like to thank Patrick Calhoun, Matt Schenk, Aaron Trevino and Casey McCarty for field assistance. Discharge data from our 12 gaging stations are available via the CUAHSI-HIS system.

REFERENCES Barnett T, Malone R, Pennell W, Stammer D, Semtner B, Washington W. 2004. The effects of climate change on water resources in the west: Introduction and overview. Climatic Change 62(1–3): 1–11. Barrett AP. 2003. National operational hydrologic remote sensing center snow data assimilation system (SNODAS) products at NSIDC. National Snow and Ice Data Center, Cooperative Institute for Research in Environmental Sciences. Barry RG, Chorley RJ. 2003. Atmosphere, weather and climate [Electronic book]. Psychology Press: New York, NY. Bedford D, Douglass A. 2008. Changing properties of snowpack in the Great Salt Lake basin, western United States, from a 26-year SNOTEL record. Professional Geographer 60(3): 374–386. Beniston M. 2006. Mountain weather and climate: a general overview and a focus on climatic change in the Alps. Hydrobiologia 562(1): 3–16. Beniston M, Stephenson DB, Christensen OB, Ferro CA, Frei C, Goyette S, Halsnaes K, Holt T, Jylhä K, Koffi B. 2007. Future extreme events in European climate: an exploration of regional climate model projections. Climatic Change 81(1): 71–95. Blanchard SF. 2004. Policy on the use of the Flow Tracker for discharge measurements, United States Geological Survey Office of Surface Water Technical Memorandum 2004.04. Blandford TR, Humes KS, Harshburger BJ, Moore BC, Walden VP, Ye HC. 2008. Seasonal and synoptic variations in near-surface air temperature lapse rates in a mountainous basin. Journal of Applied Meteorology and Climatology 47(1): 249–261.

Copyright © 2014 John Wiley & Sons, Ltd.

Budyko MI. 1974. Climate and life, 508. Elsevier: New York. Carroll T, Cline D, Fall G, Nilsson A, Li L, Rost A. 2001. NOHRSC operations and the simulation of snow cover properties for the coterminous US, 69th Annual Meeting of the Western Snow Conference, Sun Valley, Idaho; 16–19. Carroll T, Cline D, Fall G, Li L, Nilsson A. 2003. Seasonal snow cover monitoring and modeling for the coterminous United States, EGSAGU-EUG Joint Assembly, 1396. Christensen NS, Lettenmaier DP. 2007. A multimodel ensemble approach to assessment of climate change impacts on the hydrology and water resources of the Colorado River Basin. Hydrology and Earth System Sciences 11(4): 1417–1434. Davis JM, Baxter CV, Minshall GW, Olson NF, Tang C, Crosby BT. 2013. Climate-induced shift in hydrological regime alters basal resource dynamics in a wilderness river ecosystem. Freshwater Biology 58(2): 306–319. Doerner JP, Carrara PE. 1999. Deglaciation and postglacial vegetation history of the West Mountains, west-central Idaho, USA. Arctic Antarctic and Alpine Research 31(3): 303–311. Dooge JC. 1992. Sensitivity of runoff to climate change: A Hortonian approach. Bulletin-American Meteorological Society 73: 2013–2013. Dooge J, Bruen M, Parmentier B. 1999. A simple model for estimating the sensitivity of runoff to long-term changes in precipitation without a change in vegetation. Advances in Water Resources 23(2): 153–163. Dunne T, Leopold LB. 1978. Water in environmental planning. WH Freeman: New York, NY. Elsner MM, Cuo L, Voisin N, Deems JS, Hamlet AF, Vano JA, Mickelson KEB, Lee SY, Lettenmaier DP. 2010. Implications of 21st century climate change for the hydrology of Washington State. Climatic Change 102(1–2): 225–260. Fritze H, Stewart IT, Pebesma E. 2011. Shifts in western North American snowmelt runoff regimes for the recent warm decades. Journal of Hydrometeorology 12(5): 989–1006. Goode JR, Buffington JM, Tonina D, Isaak DJ, Thurow RF, Wenger S, Nagel D, Luce C, Tetzlaff D, Soulsby C. 2013. Potential effects of climate change on streambed scour and risks to salmonid survival in snow-dominated mountain basins. Hydrological Processes 27(5): 750–765. Hamlet AF, Lettenmaier DP. 2007. Effects of 20th century warming and climate variability on flood risk in the western U.S. Water Resources Research 43(6): 17. Hantel M, Maurer C. 2011. The median winter snowline in the Alps. Meteorologische Zeitschrift 20(3): 267–276. Harman CJ, Troch PA, Sivapalan M. 2011. Functional model of water balance variability at the catchment scale: 2. Elasticity of fast and slow runoff components to precipitation change in the continental United States. Water Resources Research 47: W02522. Hassan MA, Egozi R, Parker G. 2006. Experiments on the effect of hydrograph characteristics on vertical grain sorting in gravel bed rivers. Water Resources Research 42(9): W09408. Hauer FR, Baron JS, Campbell DH, Fausch KD, Hostetler SW, Leavesley GH, Leavitt PR, McKnight DM, Stanford JA. 1997. Assessment of climate change and freshwater ecosystems of the Rocky Mountains, USA and Canada. Hydrological Processes 11(8): 903–924. Hunsaker CT, Whitaker TW, Bales RC. 2012. Snowmelt Runoff and Water Yield Along Elevation and Temperature Gradients in California’s Southern Sierra Nevada. Journal of the American Water Resources Association 48(4): 667–678. Intergovernmental Panel on Climate Change (IPCC). 2008, 2009. Climate change and water, Intergovernmental Panel on Climate Change Technical Paper VI, IPCC Secretariat, Geneva. Isaak DJ, Rieman BE. 2013. Stream isotherm shifts from climate change and implications for distributions of ectothermic organisms. Global Change Biology 19(3): 742–751. Jin X, Sridhar V. 2012. Impacts of Climate Change on Hydrology and Water Resources in the Boise and Spokane River Basins. Journal of the American Water Resources Association 48(2): 197–220. Katzfey JJ. 1995a. Simulation OF Extreme New-Zealand precipitation events .1. Sensitivity to orography and resolution. Monthly Weather Review 123(3): 737–754. Katzfey JJ. 1995b. Simulation of extreme New-Zealand precipitation events .2. Mechanisms of precipitation development. Monthly Weather Review 123(3): 755–775.

Hydrol. Process. (2014) DOI: 10.1002/hyp

ELEVATION-DEPENDENT RESPONSES OF STREAMFLOW Knowles N, Dettinger MD, Cayan DR. 2006. Trends in snowfall versus rainfall in the Western United States. Journal of Climate 19(18): 4545–4559. Kunkel ML, Pierce JL. 2010. Reconstructing snowmelt in Idaho’s watershed using historic streamflow records. Climatic Change 98 (1–2): 155–176. Lewis RS, Link PK, Stanford LR, Long SL. 2012. Geologic map of Idaho: Idaho Geological Survey. University of Idaho: Moscow Idaho. Luce CH, Holden ZA. 2009. Declining annual streamflow distributions in the Pacific Northwest United States, 1948–2006. Geophysical Research Letters 36: 6. Luce CH, Abatzoglou JT, Holden ZA. 2013. The Missing Mountain Water: Slower Westerlies Decrease Orographic Enhancement in the Pacific Northwest USA. Science 342(6164): 1360–1364. Lund K. 2004. Geology of the Payette National Forest and vicinity, westcentral Idaho. United States Geological Survey. Lundquist JD, Cayan DR. 2007. Surface temperature patterns in complex terrain: Daily variations and long-term change in the central Sierra Nevada, California. Journal of Geophysical Research-Atmospheres 112(D11). Lundquist JD, Lott F. 2008. Using inexpensive temperature sensors to monitor the duration and heterogeneity of snow-covered areas. Water Resources Research 44: 6. McCabe GJ, Clark MP, Hay LE. 2007. Rain-on-snow events in the western United States. Bulletin of the American Meteorological Society 88(3): 319–328. Meyer GA, Fawcett PJ, Locke WW. 2004. Late-Pleistocene equilibriumline altitudes, atmospheric circulation, and timing of mountain glacier advances in the interior northwestern United States. Geological field trips in southern Idaho, eastern Oregon, and northern Nevada: US Geological Survey Open-File Report, 1222: 61–68. Minder JR, Mote PW, Lundquist JD. 2010. Surface temperature lapse rates over complex terrain: Lessons from the Cascade Mountains. Journal of Geophysical Research-Atmospheres 115: D14122. Mote PW, Parson E, Hamlet AF, Keeton WS, Lettenmaier D, Mantua N, Miles EL, Peterson D, Peterson DL, Slaughter R, Snover AK. 2003. Preparing for climatic change: The water, salmon, and forests of the Pacific Northwest. Climatic Change 61(1–2): 45–88. Mote PW, Hamlet AF, Clark MP, Lettenmaier DP. 2005. Declining mountain snowpack in western north America. Bulletin of the American Meteorological Society 86(1): 39 49. Muttiah RS, Wurbs RA. 2002. Modeling the impacts of climate change on water supply reliabilities. Water International 27(3): 407–419. Niu GY, Yang ZL. 2004. Effects of vegetation canopy processes on snow surface energy and mass balances. Journal of Geophysical ResearchAtmospheres 109(D23): 15. Poff NL. 1996. A hydrogeography of unregulated streams in the United States and an examination of scale-dependence in some hydrological descriptors. Freshwater Biology 36(1): 71–91. Poulos MJ, Pierce JL, Flores AN, Benner SG. 2012. Hillslope asymmetry maps reveal widespread, multi-scale organization. Geophysical Research Letters 39: L06406. Rantz SE. 1982. Measurement and computation of streamflow: Volume 1. Measurement of stage and discharge. Available from Supt. of

Copyright © 2014 John Wiley & Sons, Ltd.

Documents, GPO, Washington DC 20402. USGS Water Supply Paper 2175, 284 pages. Rauscher SA, Pal JS, Diffenbaugh NS, Benedetti MM. 2008. Future changes in snowmelt-driven runoff timing over the western US. Geophysical Research Letters 35(16): L16703. Rieman BE, Isaak D, Adams S, Horan D, Nagel D, Luce C, Myers D. 2007. Anticipated climate warming effects on bull trout habitats and populations across the interior Columbia River basin. Transactions of the American Fisheries Society 136(6): 1552–1565. Roe GH. 2005. Orographic precipitation, Annual Review of Earth and Planetary Sciences. Annual Review of Earth and Planetary Sciences. Annual Reviews, Palo Alto; 645–671. Rolland C. 2003. Spatial and seasonal variations of air temperature lapse rates in Alpine regions. Journal of Climate 16(7): 1032–1046. Sinclair MR. 1994. A diagnostic model for estimating orographic precipitation. Journal of Applied Meteorology 33(10): 1163–1175. Sivapalan M, Yaeger MA, Harman CJ, Xu XY, Troch PA. 2011. Functional model of water balance variability at the catchment scale: 1. Evidence of hydrologic similarity and space-time symmetry. Water Resources Research 47: W02522. Smith RB. 1979. The influence of mountains on the atmosphere. Advances in Geophysics 21: 87–230. Sridhar V, Jin X, Jaksa WTA. 2013. Explaining the hydroclimatic variability and change in the Salmon River basin. Climate Dynamics 40(7–8): 1921–1937. Stewart IT. 2009. Changes in snowpack and snowmelt runoff for key mountain regions. Hydrological Processes 23(1): 78–94. Stewart IT, Cayan DR, Dettinger MD. 2005. Changes toward earlier streamflow timing across western North America. Journal of Climate 18(8): 1136–1155. Tang QH, Lettenmaier DP. 2012. 21st century runoff sensitivities of major global river basins. Geophysical Research Letters 39: L06403. Tang C, Crosby BT, Wheaton JM, Piechota TC. 2012. Assessing streamflow sensitivity to temperature increases in the Salmon River Basin, Idaho. Global and Planetary Change 88–89(0): 32–44. Thackray GD, Lundeen KA, Borgert JA. 2004. Latest Pleistocene alpine glacier advances in the Sawtooth Mountains, Idaho, USA: Reflections of midlatitude moisture transport at the close of the last glaciation. Geology 32(3): 225–228. Trujillo E, Molotch NP, Goulden ML, Kelly AE, Bales RC. 2012. Elevation-dependent influence of snow accumulation on forest greening. Nature Geoscience 5(10): 705–709. Tukey JW. 1977. Exploratory data analysis. Reading, MA, 231. Vano JA, Das T, Lettenmaier DP. 2012. Hydrologic Sensitivities of Colorado River Runoff to Changes in Precipitation and Temperature. Journal of Hydrometeorology 13(3): 932–949. Westerling AL, Gershunov A, Brown TJ, Cayan DR, Dettinger MD. 2003. Climate and wildfire in the western United States. Bulletin of the American Meteorological Society 84(5): 595–604. Whitlock C, Briles CE, Fernandez MC, Gage J. 2011. Holocene vegetation, fire and climate history of the Sawtooth Range, central Idaho, USA. Quaternary Research 75(1): 114–124.

Hydrol. Process. (2014) DOI: 10.1002/hyp