Extent of the rain-snow transition zone in the western U.S. under historic and projected climate

Authors


Abstract

This study investigates the extent of the rain-snow transition zone across the complex terrain of the western United States for both late 20th century climate and projected changes in climate by the mid-21st century. Observed and projected temperature and precipitation data at 4 km resolution were used with an empirical probabilistic precipitation phase model to estimate and map the likelihood of snow versus rain occurrence. This approach identifies areas most likely to undergo precipitation phase change over the next half century. At broad scales, these projections indicate an average 30% decrease in areal extent of winter wet-day temperatures conducive to snowfall over the western United States. At higher resolution scales, this approach identifies existing and potential experimental sites best suited for research investigating the mechanisms linking precipitation phase change to a broad array of processes, such as shifts in rain-on-snow flood risk, timing of water resource availability, and ecosystem dynamics.

1 Introduction

The western United States (U.S.) is strongly dependent on wintertime precipitation phase and snowpack accumulation to sustain a multitude of ecosystem goods and services [Magoun and Copeland, 1998; Barnett et al., 2005; Bales et al., 2006]. Assessment of the region's sensitivity of water resource availability to climate change is confounded by complex terrain and large heterogeneity in temperature and precipitation projections [Elsner et al., 2010]. A detailed spatial assessment of these projected changes is particularly important in the climatic rain-snow transition zone, defined here as the transition between wintertime precipitation regimes that are strongly (near 100% climatically) rain and snow dominated. These areas are in the process of undergoing a major hydrologic shift as the phase of wintertime precipitation changes from predominantly snow to rain [Knowles et al., 2006; Abatzoglou, 2011]. Observations support changes in hydrologic indicators dependent on precipitation phase, including: widespread decreased spring snowpack [Mote et al., 2005; Mote, 2006; Bales et al., 2006; Knowles et al., 2006; Pederson et al., 2011; Kapnick and Hall, 2012], increased rain-on-snow flood risk [McCabe et al., 2007], and earlier snowmelt-driven streamflows in mountain catchments [Cayan et al., 2001; Barnett et al., 2005; Regonda et al., 2005; Bales et al., 2006; Luce and Holden, 2009; Nayak et al., 2010; Fritze et al., 2011].

The present location of the climatic rain-snow transition zone has been identified in several locales based on relatively long-term climatic observations [e.g., Nayak et al., 2010; Hunsaker et al., 2012; Minder and Kingsmill, 2013]. Maps of “at-risk” snowpack at the 4 km scale have been created for the maritime snow region in the U.S. Pacific Northwest [Nolin and Daly, 2006], for potential late season snowpack changes in the western U.S. at 0.5 km scale [McKelvey et al., 2011], and at hemispheric to global scales at coarse and static temporal (e.g., annual mean) and spatial (e.g., 200 to 50 km) resolutions [Sturm et al., 1995; Barnett et al., 2005; Kapnick and Delworth, 2013; Krasting et al., 2013]. Assessment of the sensitivity of snow-dominated landscapes to projected future temperatures at finer spatial and temporal resolutions is needed to develop climate change adaptation strategies, particularly within the highly heterogeneous complex terrain of the western U.S.

The primary objective of this study is to estimate the location of the climatic rain-snow transition zone based specifically on temperatures from days with appreciable precipitation accumulation. We also pair observed and projected changes in wet-day temperature at high spatial resolution to better aid the research and management communities to evaluate where, when, and how hydrologic impacts of rain-snow changes are likely to occur throughout the western U.S. Finally, the study summarizes changes within different sub-regions of the western U.S. and highlights existing and potential experimental sites ideal for investigations that explore the mechanisms linking precipitation phase changes to a wide array of other coupled processes, including the timing of water resource availability, the risk of rain-on-snow flooding, and ecosystem dynamics.

2 Methods

The historic rain-snow transition zone was estimated using daily temperatures for wet days (precipitation >5 mm) using a 4 km horizontal resolution surface meteorology data set [Abatzoglou, 2013] developed using data from NASA's Land Data Assimilation Systems (Phase 2) and the Parameter-elevation Relationships on Independent Slopes Model (PRISM [Daly et al., 2008]) over the period 1979–2012. Grid elements that received fewer than 30 wet-days total within the 34 year combined record for each month (i.e., <30 out of approximately 1020 days) were excluded from analysis as they contribute little to precipitation-derived water resources. Monthly averaged wet-day temperatures were transformed using an empirically derived hyperbolic tangent function [Dai, 2008] to estimate the precipitation phase probability. The Dai [2008] equation was based on worldwide temperature-precipitation phase relationships from land-based stations. The function is bounded approximately by −2° and +4°C, where the phase probability ranges from 100% snow to 100% rain. We created precipitation phase-probability maps for individual months and for a mid-winter average of December–February (DJF) wet-day temperatures.

Projected changes in wet-day temperatures were calculated using daily output of temperature and precipitation downscaled for 20 CMIP5 models for the late 20th (late20C, 1979–2012) and mid-21st century (mid21C, 2036–2065) to the aforementioned 4 km resolution grid using the Multivariate Adaptive Constructed Analogues method [Abatzoglou and Brown, 2012]. We constrain our projections to a single emissions scenario, RCP 8.5, given that inter-model variability generally exceeds scenario uncertainty for the first half of the 21st century [Hawkins and Sutton, 2009]. We calculated the average change in monthly and DJF wet-day temperatures across the 20 CMIP5 models, as the multi-model mean better isolates the signal of forced change and is often regarded as being more credible than any individual model [Reichler and Kim, 2008]. Projected changes in the multi-model mean wet-day temperature varied both by month and spatially, but generally showed a 1.5° to 4°C warming across the western U.S. by the 2036–2065 period. A delta-change procedure [e.g., Wilby and Wigley, 1997] was applied by adding historic monthly wet-day temperature from observations to projected changes in monthly wet-day temperature.

To quantify changes in precipitation phase across the entire western U.S., we aggregated the monthly wet-day snow-likelihood probabilities across the study area to calculate the proportion of land area receiving snowfall under both the historical and projected wet-day temperature regimes. We also performed an analysis using an overlay of both U.S. EPA Level-III Ecoregions and USGS HUC-4 Watersheds to assess changes in wintertime (mean DJF) phase regime at finer ecologically and hydrologically relevant scales across the region.

3 Results

Monthly maps (Figure 1) based on observed temperatures reveal strongly snow-dominated (100% snow-phase likelihood) precipitation in DJF for the mountains of the western U.S., particularly the high Rocky Mountains, Cascades, and Sierra Nevada (Figure 1). The shoulder months of November and March exhibit similar spatial patterns, but with less spatially contiguous probability of snowfall relative to DJF. April qualitatively exhibits the highest spatial variability in phase distributions across the landscape, with a strong contrast in the likelihood of rain versus snow between lower and higher elevations as the rain-snow transition zone moves upslope over the spring.

Figure 1.

The current and future extent of the strongly rain-dominated (blue), strongly snow-dominated (white), and rain-snow mix (pink to red) areas within the western U.S. based on wet-day mean temperature. Future extents are based upon the RCP8.5 scenario using a 20-model global climate model (GCM) mean. (ΔT ranging from ~1.5 to ~4°C spatially).

The projected rain-snow zones for the mid21C period (Figure 1) show continued snow dominance for DJF across western mountains, particularly in the high Rocky Mountains and Sierra Nevada. Large areas, particularly many that were previously strongly snow dominated in March and April in late20C, will likely begin to experience increased frequency of rainfall during these months. Furthermore, results suggest that many mountainous areas will be characterized by a mixed rain-snow regime in November, in contrast to the historic strongly snow-dominated precipitation regime. Exceptions to these changes are the mid-continental, higher elevation regions including the western portions of Wyoming, the greater Yellowstone ecoregion, the Uinta and Bighorn Ranges, mountains of east-central Idaho, central Colorado Rockies, and the high Sierra Nevada which remain relatively snow-dominated in November (Figure 1). Although DJF are still strongly snow dominated in mid21C projections, there is a reduction in the total months of snow-conducive temperatures overall with the likelihood of rain increasing in the autumn (October, November) and spring (March, April, May) months.

Figure 2 depicts the historic rain-snow transition zone derived from the mean wet-day temperatures across the winter months (DJF), during which greater than 45% of the land area is strongly snow dominated in the western U.S. (Figure 3). The actual extent of areas dominated by rain and snow are variable over time and change throughout the winter; however, this aggregated DJF calculation provides a means to estimate the areas that are generally characterized by winter precipitation in the form of rain versus snow during the historic late20C period. The difference in average wintertime strongly snow-dominated extent between the late20C and mid21C time periods is shaded light gray to highlight the areas that this analysis indicates are likely to shift from strongly snow dominated (100% snow likelihood) to a rain-snow mix (<100% snow likelihood).

Figure 2.

Current extent of strongly snow-dominated (white and light gray), strongly rain-dominated (blue), and mixed phase (pink to red) winter precipitation regimes based on the mean wet-day winter temperature (1979–2012 Climate Period, December–February (DJF)mean) and the encroachment (light gray) of the mixed-phase rain-snow transition zone into previously 100% snow-dominated areas. Inset of Yosemite National Park to display spatial resolution. Locations of selected experimental sites include: Boulder Creek Critical Zone Observatory (BCCZO), Beaver Creek Experimental Watershed (BCEW), Dry Creek E. W. (DCEW), Fraser Experimental Forest (FEF), H. J. Andrews E. F. (HJA), Jemez River Basin C. Z. O. (JRBCZO), Little Bear River WATERS testbed (LBR), Mica Creek E. W. (MCEW), Priest River E. F. (PREF), Reynolds Creek E. W., C. Z. O., and WATERS testbed (RCEW), Santa Catalina C. Z. O. (SCCZO), Sevilleta Research Site (SEV), Sheep Range Meteorological Transect (ShRMT), Snake Range M. T. (SnRMT), Southern Sierra C. Z. O. (SSCZO), and Tenderfoot Creek E. F. (TCEF).

Figure 3.

A quantitative summary of historical and projected areal extent of snowfall-conducive temperatures across the entire western U.S. (west of −100° longitude). Future extents are based upon the RCP8.5 scenario using a 20-model GCM mean. (ΔT ranging from ~1.5 to ~4°C spatially).

Figure 3 quantifies the extent of change in snow dominance across the western U.S. When comparing the historic to the mid21C October through April periods, all show reductions in the percent of land area within a snowfall-conducive temperature regime across the western U.S. December through February show the largest reductions, ranging from a 26 to 32% decrease in total land area containing a snowfall-conducive temperature regime. For a conceptual comparison, by mid-century it is predicted that the strongly snow-dominated portions of the western U.S. will decrease by ~2 months in length, with the mid21C spatial patterns for December, January, and February qualitatively similar to the late20C patterns of November and March (Figures 1 and 3).

Table 1 showcases the changes in areal extent between late20C and mid21C for mean DJF strongly snow-dominated and strongly rain-dominated (100% rain-phase likelihood) areas across all the EPA Level-III ecoregions and HUC-4 watersheds that comprise the western U.S. Table 1 ranks these quantified changes by the proportion of loss (by mid 21C) in the percent of late20C strongly snow-dominated area, with areas projected to see 100%-loss of their strongly snow-dominated area at the top of the list. Within the ecoregions, mountain ranges of the northwestern U.S. (e.g., Northern Rockies, North Cascades, Blue Mountains) display 100%-loss of strongly snow-dominated area by late20C, and the highest percent of total internal area lost (56%, 48%, and 27%, respectively) within this 100%-loss grouping. The findings indicate that portions of many western ecoregions are likely to undergo a fundamental change in mid-winter hydrologic regime with rain events projected to be more common in areas where historically they were relatively rare. Some ecoregions display only a small proportional area moving out of the strongly snow-dominated phase regime, but do display large increases in the proportional area of the strongly rain-dominated phase regime, demonstrating that much of the area within these regions is projected to move out of the climatic DJF rain-snow transition zone by mid21C (e.g., Colorado Plateaus, Central Basin and Range, Columbia Plateau).

Table 1. Changes in Wintertime Precipitation Phase by Regiona
RegionStrongly Snow-dominated Extent in Late20c (%)Change in Strongly Snow-dominated Extent by Mid21c (%)Strongly Rain-dominated Extent in Late20c (%)Change in Strongly Rain-dominated Extent by Mid21c (%)
  1. aBased on mean strongly snow- and rain-dominated areal extent for the winter months (DJF) between historic (late20C) and projected (mid21C) climate; ranked by greatest proportional loss by mid21C in the displayed percent of late20C strongly snow-dominated areal extent.
U.S. EPA Level-III Ecoregions
15 NORTHERN ROCKIES56−560+3
77 NORTH CASCADES48−481+18
11 BLUE MOUNTAINS27−270+29
80 NORTHERN BASIN AND RANGE18−180+23
20 COLORADO PLATEAUS18−186+50
04 CASCADES6−617+42
13 CENTRAL BASIN AND RANGE6−62+66
09 EASTERN CASCADES SLOPES AND FOOTHILLS5−50+27
10 COLUMBIA PLATEAU3−30+65
23 ARIZONA/NEW MEXICO MOUNTAINS1−136+54
12 SNAKE RIVER PLAIN44−420+29
16 IDAHO BATHOLITH88−790+1
19 WASATCH AND UINTA MOUNTAINS68−600+8
18 WYOMING BASIN44−3700
05 SIERRA NEVADA19−1426+26
41 CANADIAN ROCKIES100−6700
22 ARIZONA/NEW MEXICO PLATEAU3−220+71
17 MIDDLE ROCKIES82−4600
21 SOUTHERN ROCKIES69−340+3
08 SOUTHERN CALIFORNIA MOUNTAINS0077+16
14 MOJAVE BASIN AND RANGE0095+4
01 COAST RANGE0088+11
02 PUGET LOWLAND0095+5
03 WILLAMETTE VALLEY0099+1
06 SOUTHERN AND CENTRAL CALIFORNIA CHAPARRAL AND OAK WOODLANDS0099+1
07 CENTRAL CALIFORNIA VALLEY001000
78 KLAMATH MOUNTAINS0065+30
79 MADREAN ARCHIPELAGO0099+1
81 SONORAN BASIN AND RANGE001000
USGS HUC-4 Watersheds
MIDDLE SNAKE29−290+23
GREAT SALT LAKE16−160+56
YAKIMA16−160+25
ESCALANTE DESERT SEVIER LAKE14−140+70
BLACK ROCK DESERT HUMBOLDT14−140+49
OREGON CLOSED BASINS13−130+26
UPPER COLORADO DIRTY DEVIL12−1210+64
PUGET SOUND7−740+18
MIDDLE COLUMBIA6−60+54
CENTRAL LAHONTAN5−51+75
UPPER CANADIAN4−430+18
CENTRAL NEVADA DESERT BASINS4−411+53
SALT1−159+31
SACRAMENTO1−154+26
LOWER COLORADO LAKE MEAD1−159+35
NORTH LAHONTAN1−10+50
KLAMATH NORTHERN CALIFORNIA COASTAL1−146+27
WILLAMETTE1−164+25
UPPER PECOS1−171+23
BEAR79−760+3
RIO GRANDE ELEPHANT BUTTE12−1122+51
UPPER COLORADO DOLORES17−150+54
LOWER SNAKE50−440+25
LOWER GREEN55−480+15
WHITE YAMPA81−6900
POWDER TONGUE36−2900
SAN JOAQUIN10−874+8
MISSOURI MUSSELSHELL35−2700
MISSOURI MARIAS44−3300
UPPER SNAKE74−550+8
LOWER YELLOWSTONE24−1700
SAN JUAN10−71+77
LOWER COLUMBIA3−245+34
NORTHERN MOJAVE MONO LAKE3−279+11
MISSOURI HEADWATERS69−4400
TULARE BUENA VISTA LAKES10−676+7
GREAT DIVIDE UPPER GREEN52−3100
BIG HORN40−2300
UPPER YELLOWSTONE62−3200
RIO GRANDE HEADWATERS65−2900
COLORADO HEADWATERS73−300+11
GUNNISON68−260+13
OREGON WASHINGTON COASTAL0079+15
LITTLE COLORADO0013+83
SOUTHERN CALIFORNIA COASTAL0085+3
SOUTHERN MOJAVE SALTON SEA0095+1
CENTRAL CALIFORNIA COASTAL0097+1
LOWER COLORADO0082+3
LOWER GILA0099+1
MIDDLE GILA0096+1
SAN FRANCISCO BAY00980
UPPER GILA0072+27

For the analysis of regional watersheds (HUC-4), 100%-loss of strongly snow-dominated area occurs mainly in basins with moderate relief and elevation, (e.g., Middle Snake, Great Salt Lake, Oregon Closed Basins). Basins with greater relief, however, have higher late20C strongly snow-dominated areal extent, but are not projected to have a 100%-loss of strongly snow-dominated area because some limited areas at higher elevations are projected to stay within a strongly snow-dominated wintertime regime (e.g., Bear, White-Yampa, Upper Snake). Basins that comprise the highest elevation headwater regions are still projected to have large proportional losses in the strongly snow-dominated area (all >38% loss) but also are projected to retain some of the largest strongly snow-dominated areal extents by mid21C (e.g., Upper Yellowstone, Rio Grande Headwaters, Colorado Headwaters, Gunnison).

4 Discussion

Recent studies across the western U. S. have highlighted research sites within the context of either being rain-dominated, snow-dominated, or transitional rain-snow climate. Qualitatively our findings are supported by analyses of empirical data from the Southern Sierra Critical Zone Observatory [Hunsaker et al., 2012], Reynolds Creek Critical Zone Observatory [Nayak et al., 2010], H.J. Andrews Experimental Forest [Jones and Perkins, 2010], and Mica Creek Experimental Watershed [Hubbart et al., 2007], all of which are located within or near the rain-snow transition zone (Figure 2). Quantitatively, the accuracy of these findings is controlled directly by the phase/temperature relationships derived from global land-based observational data sets [Dai, 2008], which may differ slightly from regional phase/temperature relationships unique to specific locales in the western U.S. Additionally, the empirical Dai [2008] precipitation phase relationships are based on seasonal means calculated for each annual quarter and are derived from 3-hourly precipitation phase observations. This is somewhat different than the daily-scale temperature analysis (with aggregation to monthly and seasonal time-steps) used for wet-day phase calculations within this study. The accuracy of the findings could be enhanced through additional work with improved empirical relationships to estimate the phase likelihood; for example, by including variables such as vapor pressure, as has been done at more localized scales when sufficient data are available [Marks et al., 2013]. Furthermore, additional improvements within high-resolution spatially continuous data sets of surface meteorological variables would provide increased accuracy for the driving interpolated climate data, particularly in regions of the western U.S. where data in complex terrain are especially sparse and limit the PRISM methodology, leading to increased localized error [Daly et al., 2009]. Despite these nuances, this analysis should effectively capture the general spatiotemporal trends of transitioning precipitation phase regime across the western U.S.

4.1 Implications for Timing of Peak Snow-water Equivalent (SWE)

Mountainous regions in the western U.S. have historically been strongly snow dominated from November through March. The sensitivity analysis revealed that by mid21C the length of snowfall-conducive temperatures over many western mountain ranges will be reduced from approximately five (November–March) to approximately three (DJF) months of the year (Figures 1 and 3). Considering these temporal changes, it will be critical for the water resources research and management communities to look beyond the April 1 standard for measurement of approximate peak SWE [Pederson et al., 2011]. The idea that a standard protocol, such as a fixed date, can be continually valid has been challenged previously when considering water resources and climate change [Milly et al., 2008]. Other studies have stated a similar need to change the time of peak SWE evaluation in an effort to better predict resulting springtime high flows and aid water managers [Hunsaker et al., 2012; Meromy et al., 2013].

Although this April 1 standard is used widely in current practice for evaluating snowpack, spatially explicit evaluation dates, which are unique to the changing precipitation regimes of different western U.S. mountain ranges, may be more appropriate. A spatially explicit evaluation may be particularly important in areas where the majority of snowfall historically occurred within the spring window of March, April, and May (common for many regions closer to the continental interior). Regions historically dependent on large snowfall amounts in these spring months may be more sensitive to the impacts of changing climate on snowpack because although DJF remain cold enough to be strongly snow dominated, large changes may still occur when spring temperatures approach the rain-snow threshold. Thus, as a response to this spatial complexity, and to advance the public and scientific understanding of regional- to local-scale effects of changing climate, this study provides high-resolution, spatial data products that allow for this spatially explicit assessment of snowfall-conducive temperature regimes.

4.2 Regional Considerations

In many areas, the rain-snow transition zone is broad because controlling horizontal temperature gradients are subtle, due primarily to a combination of location (elevation and latitude) and low topographic relief. In contrast, the transition zone is narrower in regions with steep elevational, and hence steep temperature gradients. These relationships suggest that many relatively large areas that contain lower relief, mid-elevation mountain ranges will likely shift relatively quickly into new precipitation phase regimes (e.g., the Northern Rockies, North Cascades, and Blue Mountains ecoregions, Table 1). Alternatively, areas with steeper elevational gradients will likely have a smaller portion of land area shift into, or exit the transition zone in the near future (e.g., the Sierra Nevada, Central Rockies, and Southern Rockies ecoregions, Table 1). The interior northwestern U.S. shows a greater sensitivity of its strongly snow-dominated areas to warming because much of the region is characterized by relatively warm winter temperatures and by mainly mid-elevation mountain ranges. This is in accordance with empirical work that suggests that the timing of peak flows have advanced at a faster rate in the northwestern U.S., compared to other regions in the western U.S. [Regonda et al., 2005]. Our findings highlight the spatial and temporal complexity of changes and indicate how certain experimental sites within the western U.S. are better positioned to help assess the impact of these changes on precipitation phase and associated ecohydrological processes (Figure 2).

4.3 Resources for the Greater Scientific Research Community

The locations of selected long-term research sites are included (Figure 2) to help the research community identify where these sites occur within the rain-snow transition zones, and where potential future sites are warranted to fill gaps within the existing monitoring and hydrologic research network. An inset of Yosemite National Park provides a detailed view that highlights the relatively fine resolution of the results across a steep elevational gradient (Figure 2). This high spatial resolution allows individual small watersheds, down to tens of km2, to be characterized within the rain-snow transition zone, showing where a gradient of rain- or snow- dominated temperature regimes existed historically and where new wintertime rain-snow transitional temperature regimes are projected to exist by the mid21C. Similarly, tabulated results of the spatial analysis (Table 1) allow the research and planning community to compare and estimate projected changes relevant to more localized questions focused within specific watersheds and/or ecoregions.

5 Conclusions

As demonstrated through the mapped temperature-precipitation phase relationship across the western U.S., the climatic rain-snow transition zone will move up in altitude and latitude. The western U.S. is projected to see an average monthly reduction from ~53% to ~24% in the extent of the land area within a wintertime snowfall regime (Figure 3). The climatic mean annual duration of the snowfall regime will also be reduced across the western U.S. with the annual duration of 100% snow-dominated precipitation decreasing by mid21C from approximately 5 to approximately 3 (DJF) months of the year on average for many of the western mountain ranges (Figures 1 and 3). Findings demonstrate that changes in the climatic extent will be complex and that many established research sites are, or will be, better poised than others to conduct research advancing the understanding of how these shifts in precipitation phase at a climatic scale may impact integrated hydrologic, ecologic, and social systems. As a resource for the research and planning communities, full-resolution maps and data sets of historic and projected rain-, transitional-, and snow-dominated extent at monthly and integrated DJF time intervals are available for public download through the corresponding author's website.

Acknowledgments

The authors provide thanks to the Oregon State Climate Group for creating and providing free online access to the PRISM climate data, and to Dr. Aiguo Dai for his rapid response in addressing questions regarding his work. Financial support was provided by the National Science Foundation's IGERT Program (Award 0903479) and by the National Science Foundation's CBET Program (Award 0854553).

The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.

Ancillary