<p> 1 Supplemental Materials </p><p>2</p><p>3 Methods </p><p>4</p><p>5 Comparison of Simulated to Observed SWE </p><p>6 The VIC model has been widely used to represent hydrological conditions and to </p><p>7 study future hydrological change (e.g. Nijssen et al 2001). Mote et al (2005) examined </p><p>8 long-term trends in April 1 SWE and found that the fraction of negative trends was nearly</p><p>9 identical for observations and VIC output (75% for snow course observations, 73% for </p><p>10 VIC). Mao et al (2015) compared elevation profiles of VIC-simulated SWE and </p><p>11 observations over California and found good agreement. </p><p>12 To evaluate the performance of the VIC snow model under historic conditions, we</p><p>13 performed a limited set of point simulations for comparison with SNOTEL observations. </p><p>14 VIC simulations at 1/16º spatial resolution using the Livneh et al (2013) forcing data </p><p>15 were run for the period 1987-2005 to provide overlap with the SNOTEL observations. </p><p>16 The simulations were run for grid cells that contained SNOTEL sites as well as for the </p><p>17 eight surrounding grid cells. Sub-grid cell elevation heterogeneity was accounted for </p><p>18 through the use of up to five elevation bands within each grid cell. Using our simulations,</p><p>19 we binned simulated SWE and observed SWE by 300 m elevation increments </p><p>20 corresponding to the elevation of the SNOTEL site, ranging from 500 m to 3500 m, and </p><p>21 the average over each was computed. </p><p>22 23 Past comparisons of VIC-modeled SWE have been over limited parts of our </p><p>24 domain, and the Mote et al. (2005) comparisons are now somewhat out of date (through </p><p>25 1997). Our hydrologic simulations, using historical gridded data (described in Section, </p><p>26 2.3) addressed this by comparing VIC-simulated SWE to observations of SWE from </p><p>27 SNOTEL stations, shown in Online Resource 1, temporally averaged over 1987-2005. </p><p>28 Although the range of values differs between observed and simulated SWE, with </p><p>29 simulated SWE often exhibiting a larger range particularly around 2600 m, the mean, 10th</p><p>30 and 90th percentile values are in relatively good agreement. This is particularly the case </p><p>31 for middle elevations (2000 m to 3000 m). The White Mountains are the only range that </p><p>32 shows little agreement between observed and simulated SWE, particularly in the 10th and </p><p>33 90th percentile values. This is likely due in part to the small number of available </p><p>34 observations for this region (the area of the White Mountains is substantially smaller than</p><p>35 that of the other ranges, corresponding to a much smaller number of SNOTEL stations, </p><p>36 e.g. less than ten versus hundreds). Additionally, some of the lack of agreement in the </p><p>37 White Mountains may be due to the large interannual variability in simulated SWE </p><p>38 relative to observed. </p><p>39 Comparisons to SNOTEL SWE can be problematic, with SNOTEL SWE being </p><p>40 unrepresentative of SWE distribution in the surrounding area due to the majority of </p><p>41 SNOTEL stations being located at mid to lower elevations (Nolin 2012). However, </p><p>42 Meromy et al (2013) studied the representativeness of SNOTEL-observed SWE to </p><p>43 modeled SWE at three resolutions – 1, 4 and 16 km2 – in California, Idaho, Oregon, </p><p>44 Wyoming and Colorado and found that snow depth during the accumulation period was </p><p>45 generally less biased, with the ablation period exhibiting greater subgrid variability. At a 46 larger scale over the Colorado Rockies, Chen et al (2015) compared observed SWE from </p><p>47 SNOTEL networks to modeled SWE (using the VIC model) and found relatively good </p><p>48 agreement. We are primarily concerned with capturing the interannual variability of April</p><p>49 1 SWE thus catchment-scale issues with the representativeness of SNOTEL SWE are less</p><p>50 concerning for our application. </p><p>51</p><p>52 Calculation of Dead Fuel Moisture (DFM)</p><p>53 Daily minimum and maximum temperature, relative humidity and precipitation were </p><p>54 used to derive Equilibrium Moisture Content (EMC) and calculate 100-hr and 1000-hr </p><p>55 DFM as detailed in Section 2.4. We derived minimum and maximum daily relative </p><p>56 humidity (RH) from specific humidity (calculated internally by VIC) by assuming that </p><p>57 specific humidity is approximately equal to the mixing ratio (w). We then computed </p><p>58 minimum and maximum daily RH as , where , is the saturation vapor pressure calculated</p><p>59 with the minimum and maximum daily temperatures and p is the atmospheric pressure. </p><p>60 We accounted for precipitation duration using an empirical transform that translates daily</p><p>61 precipitation amount to precipitation duration (Holden and Jolly 2011) and constrained </p><p>62 precipitation duration not to exceed 8 hours. 63</p><p>64 Results</p><p>65 Projected Changes in Snowpack</p><p>66 Most of the differential effects of climate change on SWE can be explained in </p><p>67 terms of elevation and thus temperature. Online Resource 7 characterizes changes in the </p><p>68 ensemble mean of simulated April 1 SWE between the historical period and the future </p><p>69 period for RCP8.5. For this analysis, we applied locally weighted scatterplot smoothing </p><p>70 (LOWESS), a method of nonparametric regression (Cleveland 1981). The curves show </p><p>71 the distribution of April 1 SWE for the historical period (1970-1999) and the future </p><p>72 climatological periods (2020s, 2050s and 2080s) for RCP8.5. Underlying the SWE </p><p>73 curves, we show the distribution of grid cell elevations. In the Sierra Nevada, the largest </p><p>74 decrease in SWE occurs at 1500-2000 m, while in the Cascades and Northern Rockies, it </p><p>75 occurs around 1500 m. Maximum SWE in the Cascades occurs around 1400 m during the</p><p>76 historical period, but it nearly disappears at that elevation by the end of the twenty-first </p><p>77 century. Since much of the Cascades is below 1400 m, projected changes in climate will </p><p>78 have large impacts on volumetric SWE storage. We also classified grid cells in the five </p><p>79 mountain ranges as rain dominant (RD), transient (TR1 or TR2) or snow dominant (SD) </p><p>80 (e.g. Elsner et al 2010) by modifying the classification regime of Hamlet and Lettenmaier</p><p>81 (2007) to include two transient classifications, corresponding to average winter </p><p>82 temperature of between -6C and 0C and 0C and +5C, respectively (Online Resource </p><p>83 8). By the end of the 21st century for RCP 8.5 projections, the majority of TR1 grid cells </p><p>84 in the Cascades and the western part of the Northern and Southern Rockies have </p><p>85 transitioned to TR2. The grid cells on the windward slopes of the Cascades and Sierra 86 Nevada that were TR1 in the historical period have almost entirely transitioned to RD. </p><p>87 Thus, windward-facing areas and mid- to low-elevation areas such as the White </p><p>88 Mountains will be most affected by warming temperatures in western US mountain </p><p>89 ranges.</p><p>90</p><p>91</p><p>92</p><p>93</p><p>94</p><p>95</p><p>96</p><p>97</p><p>98</p><p>99</p><p>100</p><p>101</p><p>102</p><p>103</p><p>104</p><p>105</p><p>106</p><p>107</p><p>108 109 FIGURES </p><p>110 Online Resource 1: Comparison of simulated and observed April 1 SWE (averaged over </p><p>111 1987 – 2005). Observations for April 1 SWE were taken from SNOTEL stations </p><p>112 in the five mountain ranges and were compared with VIC-simulated SWE </p><p>113 averaged over nine grid cells surrounding the SNOTEL station. </p><p>114 Online Resource 2: Selected Global Climate Models (GCMS) from CMIP5 used in this </p><p>115 study. </p><p>116 Online Resource 3: Average winter (November through March) historical temperature </p><p>117 and projections for increases in temperature in western US mountain ranges </p><p>118 averaged over the ten GCMs.</p><p>119 Online Resource 4: Total winter (Nov-March) precipitation projections for mountain </p><p>120 ranges and lowland regions from CMIP5. </p><p>121 Online Resource 5: Total spring (March-May) precipitation projections for mountain </p><p>122 ranges and lowland regions from CMIP5 for RCP 4.5 and 8.5. </p><p>123 Online Resource 6: Projected losses in April 1 SWE storage (in km3) for five mountain </p><p>124 ranges, averaged across the ten GCMs. The maximum and minimum losses </p><p>125 denote the largest and smallest SWE storage losses projected by the ten selected </p><p>126 GCMs. </p><p>127 Online Resource 7: Change in mean simulated April 1 SWE between historical and future</p><p>128 periods for all GCMs. Changes in grey are not statistically significant. </p><p>129 Online Resource 8: Change in ensemble mean simulated April 1 SWE between the </p><p>130 historical (1970-1999) period and RCP 8.5. Curves show the distribution of April </p><p>131 1 SWE overlaid on the distribution of grid cell elevations. 132 Online Resource 9: Shifts in hydrologic model grid cell classifications based on winter </p><p>133 temperature. Classifications are based on Hamlet and Lettenmaier (2007) but </p><p>134 modified to include two transient classifications. </p><p>135 Online Resource 10: Heatmap comparing projected % change in soil moisture between </p><p>136 GCMs. Of particular interest are areas where the soil moisture signal differs </p><p>137 between models. </p><p>138 Online Resource 11: Ensemble-mean summer (JJAS) 1000-hr dead fuel moisture (DFM) </p><p>139 shown over a) the five mountain ranges, and b) the six lowland regions, for the </p><p>140 control period (1970-1999) and RCP 8.5. 2010-2039, 2040-2069, and 2070-2099. </p><p>141 For the control period, % DFM is shown, and for the future periods, the % </p><p>142 difference in DFM. DFM was calculated using the NFDRS algorithm for fuel </p><p>143 moisture. </p><p>144 Online Resource 12: Number of models projecting positive changes in 100-hour DFM </p><p>145 minus number of models projecting negative changes for the mountain ranges and</p><p>146 lowland regions. Consistent agreement between models can be seen in areas in </p><p>147 dark blue (positive changes) and dark red (negative changes), whereas lighter </p><p>148 colors indicate areas where models do not agree on the sign of change in 100-hr </p><p>149 DFM.</p><p>150 Online Resource 13: Number of models projecting positive changes in 1000-hour DFM </p><p>151 minus number of models projecting negative changes for the mountain ranges and</p><p>152 lowland regions. Consistent agreement between models can be seen in areas in </p><p>153 dark blue (positive changes) and dark red (negative changes), whereas lighter 154 colors indicate areas where models do not agree on the sign on change in 1000-hr </p><p>155 DFM. </p><p>156 Online Resource 14: Areas of mountain ranges and lowland regions in study domain. </p><p>157</p><p>158</p><p>159</p><p>160</p><p>161</p><p>162</p><p>163</p><p>164</p><p>165</p><p>166</p><p>167</p><p>168</p><p>169</p><p>170</p><p>171</p><p>172</p><p>173</p><p>174</p><p>175</p><p>176 177 References </p><p>178 Chen F, Barlage M, Tewari M, Rasmussen R, Jin J, Lettenmaier D, Livneh B, Lin C, </p><p>179 Miguez-Macho G, Niu G-Y, Wen L, Yang Z-L (2015) Modeling Seasonal </p><p>180 Snowpack Evolution in the Complex Terrain and Forested Colorado Headwaters </p><p>181 Region: A Model Intercomparison Study. J Geophys. Atmos. 119: 13795-13819. </p><p>182 doi: 10.1002/2014JD022167</p><p>183 Cleveland, WS (1981) LOWESS: A program for smoothing scatterplots by robust locally</p><p>184 weighted regression. The American Statistician 35 (1): 54. doi: 10.2307/2683591</p><p>185 Elsner MM, Cuo L, Voisin N, Deems JS, Hamlet AF, Vano JA, Mickelson, KEB, Lee S-</p><p>186 Y, Lettenmaier DP (2010) Implications of 21st century climate change for the </p><p>187 hydrology of Washington State. Clim. Change 102: 225-260. doi: </p><p>188 10.1007/s10584-010-9855-0</p><p>189 Hamlet AF, Mote PW, Clark MP, Lettenmaier DP (2007) Twentieth-century trends in </p><p>190 runoff, evapotranspiration, and soil moisture in the western United States. J. </p><p>191 Climate 20: 1468-1486. doi: 10.1175/JCLI4051.1</p><p>192 Holden ZA and Jolly WM (2011) Modeling topographic influences on fuel moisture and </p><p>193 fire danger in complex terrain to improve wildland fire management decision </p><p>194 support. Forest Ecology and Management 262, 2133-2141. doi: </p><p>195 10.1016/j.foreco.2011.08.002. </p><p>196 Mao Y, Nijssen B, Lettenmaier DP (2015) Is climate change implicated in the 2013-2014</p><p>197 California drought? A hydrologic perspective. Geophys. Res. Let. 42(8): 2805-</p><p>198 2813. doi: 10.1002/2015GL063456 199 Meromy L, Molotch NP, Link TE, Fassnacht SR, Rice R (2013) Subgrid variability of </p><p>200 snow water equivalent at operational snow stations in the western USA. Hydrol. </p><p>201 Process. 27: 2382-2400. doi: 10.1002/hyp.9355</p><p>202 Natural Resources Conservation Service (1997) SNOTEL Data Collection System. </p><p>203 Available at: http://www.wcc.nrcs.usda.gov/snow/snotel-wedata.html</p><p>204 Nijssen, B, O’Donnell GM, Lettenmaier DP, Lohmann D, Wood, EF (2001) Predicting </p><p>205 the Discharge of Global Rivers. J Climate 14: 3307-3323. doi: 10.1175/1520-</p><p>206 0442(2001)014<3307:PTDOGR>2.0.CO;2</p><p>207 Nolin AW (2012) Perspectives on Climate Change, Mountain Hydrology and Water </p><p>208 Resources in the Oregon Cascades, USA. Mountain Res. and Dev. 32(S1): S35-</p><p>209 S46. doi: 10.1659/MRD-JOURNAL-D-11-00038.S1</p><p>210</p><p>211</p><p>212</p><p>213 214 Tables </p><p>215 Table 1: Selected Global Climate Models from CMIP5</p><p>Global Climate Model Model Source BCC-CSM1-1-M Beijing Climate Center- Meteorological </p><p>Administration, China CanESM2 Canadian Centre for Climate Modeling and</p><p>Analysis CCSM4 National Center of Atmospheric Research, </p><p>US CNRM-CM5 National Centre of Meteorological </p><p>Research, France CSIRO-Mk3-6-0 Commonwealth Scientific and Industrial </p><p>Research Organization/Queensland Climate</p><p>Change Center of Excellence, Australia HadGEM2-CC Met office Hadley Center, United Kingdom HadGEM2-ES Met Office Hadley Center, United </p><p>Kingdom IPSL-CM5A Institut Pierre Simon Laplace, France MIROC5 Atmosphere and Ocean Research Institute, </p><p>University of Tokyo; Japan Agency for </p><p>Marine-Earth Science and Technology NorESM1-M Norwegian Climate Center, Norway 216 217 Table 2: Average winter (November through March) historical temperature and </p><p>218 projections for increases in temperature in western US mountain ranges </p><p>Mountain Historical RCP 4.5 RCP 4.5 RCP 4.5 RCP 8.5 RCP 8.5 RCP 8.5 Range (1970- 2010- 2040- 2070- 2010- 2040- 2070- 1999) 2039 2069 2099 2039 2069 2099 C Increase Increase Increase Increase Increase Increase (C) (C) (C) (C) (C) (C)</p><p>Sierra -1.06 +1.05 +2.14 +2.79 +1.33 +2.72 +4.55 Nevada Cascades -1.76 +1.06 +2.13 +2.81 +1.32 +2.68 +4.48 Northern -6.15 +1.23 +2.50 +3.24 +1.52 +3.16 +5.31 Rockies Southern -6.20 +1.34 +2.60 +3.30 +1.58 +3.34 +5.61 Rockies White -0.78 +1.22 +2.29 +2.90 +1.41 +2.97 +4.96 Mountains 219</p><p>220 221 Table 3: Projected losses in April 1 SWE storage (in km3) for five mountain ranges, </p><p>222 averaged across the ten GCMs. The maximum and minimum losses denote the largest </p><p>223 and smallest SWE storage losses projected by the ten selected GCMs. </p><p>224</p><p>225</p><p>226 227 Table 4: Areas of mountain ranges and lowland regions in domain </p><p>Region Area (x 1,000 km2) Mountain Ranges Sierra Nevada 53.7 Cascades 101 Northern Rockies 304 Southern Rockies 175 White Mountains 4.82 Lowland Regions Coastal North 83.7 Coastal South 154 Northwest Interior 289 Lower Colorado 308 Great Basin 287 Missouri 562 228 229</p>