RESULTS

4.1 Snow simulations

The hydrological model performance in simulating SWE is adequate for all winter seasons at the snow pillow site in Reynolds Mountain (Figure 4), which provides confidence that the model can represent snow melt and other fluxes that cannot be easily measured. The Nash-Sutcliffe efficiency (NSE ), normalized mean bias (NBIAS ), and root mean squared error (RMSE ) are 0.95, -0.01, and 4.8 cm, respectively. The cumulative snowfall shown in Figure 4 illustrates the accumulation or loss of snow at the snow pillow site. There is a pronounced difference between cumulative snowfall and peak SWE (Figure 4), indicating the snow depletion by midwinter melting, winter snow redistribution, and sublimation during most of the simulation period. Further diagnosis, using model outputs, suggests that melting processes during midwinter and early spring are responsible for large snow depletion due to a series of warm spells having a maximum air temperature of ~10 ºC. For example, during 1983-84 winter, five warm spells (one week duration each) occurred with a mean air temperature of 5 ºC that resulted in melting 176 mm of SWE before it peaked on April 15, 1984.
The simulated spatial snow depth agrees well with the observed snow depth from airborne lidar (Figure 5; Table 2). The spatial NSEbetween observations and simulations is 0.7 for all HRUs. The mean of observed and simulated spatial snow depths are 0.99 m and 1.03 m, respectively. The model captures the areas of snow sinks and sources with reasonable error ranges (Figure 5). Both simulated and observed snow depth maps provide the first sign of underlying spatial controls on the distributed snow processes. The snowpack usually recedes markedly in almost 50% of the basin by mid-March. Snowpack in HRUs with north facing aspects and forest cover (Figure 2c, Figure 3), however, does not reach the peak value (Figure 5). Table 2 shows that sagebrush and aspen HRUs having north facing steep slopes have high snow depths as these areas have slow snow depletion rates with low solar radiation and receive transport of snow from other HRUs. For example, a total sunshine period during 2008-2009 winter for north facing steep slopes in sagebrush is 45 days, which is lower than that of the east facing sagebrush regions (66 days). Aspen and riparian forest on flat-lying areas in the valley bottom also experience high snow accumulations as snow is redistributed from sagebrush and grassland areas to wooded areas in both basins.
Modeled net blowing snow transport into the riparian forest is up to 62 mm, which is 10% of the 2008-09 total precipitation. In contrast, the blowing snow transport from the sagebrush HRU is 32 mm. Thus, vegetation type or height, aspect, slope, and topographic depressions play an essential role in the spatial variability of snow accumulation. Previous studies (Reba et al., 2011a; Kumar et al., 2013; Rasouli et al., 2015) emphasized the impacts of vegetation induced blowing snow transport as a major process responsible for the spatial variation of SWE. Vegetation, aspect, and slopes are found as partial factors affecting the spatial variation of snow accumulation at a regional scale in western North America (Tennant et al., 2017).

4.2 Relation between climatic teleconnection patterans and local hydrology

The Pearson correlation coefficients were calculated between basin-scale hydrological variables, including observed mean annual and winter air temperatures, modeled annual rainfall, the ratio of modeled rainfall to observed precipitation, modeled snowfall, observed annual runoff, the ratio of annual runoff to total annual precipitation, observed peak SWE and its timing in Reynolds Mountain, and teleconnection patterns (Figure 6). The partitioning of precipitation into rainfall and snowfall was used to calculate the ratio of rainfall to total precipitation, which showed the strongest relation with AAO in the same year with a correlation coefficient of -0.7 (Figure 6a). The ratio of rainfall to precipitation also has an intermediate positive correlation with SST (Figure 6b) and a strong negative correlation with PNA in the preceding year (Figure 6e). The climate teleconnection of NAO showed an intermediate positive relation with winter air temperature and a negative relation with the annual ROS runoff (Figure 6d).
Six major hydroclimatic phases with distinct climatic conditions were identified based on the decomposed multiple year frequency time series of daily precipitation (Figure 7). Each phase was classified into a wet or dry (above or below average, respectively) span, lasting for three to eight years. Characteristics of the identified hydroclimatic phases (Figure 7) in relating the hydrological variables and climatic teleconnection patterns are demonstrated in Figure 8. Phase (1) is a wet and cold period under negative AO and SST from 1984 to 1986; phase (2) is a dry and warm period from 1987 to 1994 under positive AO and NAO; phase (3) is a wet and cold period from 1995 to 1999 under positive AAO; phase (4) is a warm period from 2000 to 2003 under the SST transitioning from negative to positive and positive PNA; phase (5) is in a transition from warm to cold conditions under positive PNA and negative NAO from 2004 to 2011; and finally, phase (6) is a low flow period from 2012 to 2014 under negative PNA and positive NAO.

4.3 Time variation of rain on snow (ROS) events

The hydrological importance of ROS events in generating high flows and its potential relation with NAO and AAO, warrants studying these events in more detail for different snow regimes in Reynolds Mountain. Snowmelt generates substantial runoff during ROS events. The ROS contribution to total runoff, however, depends on (i) snow cover of the HRUs and (ii) rainfall occurrence. Heterogeneity of snow cover due to topography and redistribution of snow by wind controls the runoff during ROS events. Snow transport to sinks and topographic depressions with drifted snow can intensify snowmelt in spring, and early summer when the likelihood of rainfall is high.
HRUs with drifted snow have deep snowpacks (Figure 5) and generate high ROS runoff depths and contribute more than other HRUs to basin streamflow. Figure 9 shows the annual runoff generated in four snow regimes. HRUs were grouped into four blowing snow regimes (Rasouli et al., 2015), including sink and source, and intercepted and sheltered snow. These categories are based on topographic exposure and vegetation height (Pomeroy et al., 1997). Blowing snow sink HRUs include drift HRUs, riparian, and tall sage HRUs. The HRUs covered with short vegetation were grouped as source HRUs. The forested landscapes were divided into those that are subject to interception (coniferous fir) and those that are cleared or have negligible winter interception capacity (Deciduous aspen, Pomeroy et al., 2002). Runoff generated during ROS events varies with blowing snow regimes in different hydroclimatic phases (Figure 9). Forest landscapes with intercepted snow on canopies generate larger ROS runoff than other snow regimes in all six hydroclimatic phases (Figure 9a). The sheltered forest landscapes with minimal blowing wind generated the highest ROS runoff during phase three among all snow regimes, with 60% above normal (Figure 9a). This is likely due to the above-normal precipitation (206 mm, 21%) and below-normal winter air temperature (0.5 °C) during phase three, which prolonged the snow cover period by 16 days above normal (Table 3) and increased the frequency of the ROS events. In contrast to the sheltered HRUs, the blowing snow source HRUs showed the lowest (40% below normal) ROS runoff generation during phase two (Figure 9b). In phase two, annual precipitation was 163 mm (17%) below normal and the winter air temperature was 0.3°C warmer than the normal values (Table 3). Despite the high rainfall ratio in this phase, the modeled ROS runoff averaged for the entire basin was the lowest among the phases with 56 mm (31%) below normal (Table 3 and Figure 9). This is because of a short period of snow cover (21 days shorter than normal, Table 3) and the effect of a strong positive phase of NAO in hydroclimatic phase two (Figure 8). A mechanistic diagnosis on ROS runoff is critical as the mid-latitude basins are expected to experience warmer conditions and precipitation phase change from snow to rain in the future.
The observed runoff ratio, defined as the ratio of total annual runoff to total annual precipitation, varies between 13% above normal in phases one and three and 13% below normal in phase six (Table 3). Time series of the runoff ratio (Figure 8) are consistent with the values reported by Sridhar and Nayak (2010).

4.4 Synthesis of the hydrological linkage to climate teleconnection patterns in hydroclimatic phases

The time-averaged observed precipitation (snowfall and rainfall ratio), mean annual and winter air temperatures, observed streamflow, modeled ROS, observed peak SWE and observed snow cover duration in Reynolds Mountain are reported in Table 3. The linkage of these variables in a small basin to regional climatic teleconnection patterns was synthesized for the six hydroclimatic phases as the following:
Phase one (1984-1986, cold and wet, high flow, negative phases of AO and SST). In this phase, the observed peak SWE and annual runoff were respectively 315 mm (63%) and 314 mm (57%) above normal, and the observed runoff ratio was up by 13% (Table 3). The observed mean annual air temperature was 0.6 ºC below normal, and annual precipitation was 301 mm above normal, which makes this phase the coldest and wettest among the phases. The highest annual runoff and peak SWE were linked to strong negative phases of SST and AO (Figure 8). The high SWE accumulations and subsequent runoff generations were spatially restricted to areas with drifted snow and north facing HRUs (Figure 10). Runoff generated during ROS events were, however, quite low during phase one across Reynolds Mountain.
Phase two (1987-1994, dry, positive phases of NAO and AO) . The observed annual runoff and modeled ROS runoff were respectively 163 mm (30%) and 56 mm (31%) below normal (Table 3). A strong positive phase of NAO (Figure 8) locked the polar cold air in the Arctic region, leading to a slightly warmer and drier than normal winter in the study area in phase two, which restricted the generation of ROS runoff. As a result, this decreased the annual runoff. The spatial variations of modeled peak SWE and runoff were smaller than other wet phases, such as phase one (Figure 10).
Phase three (1995-1999, cold, high ROS runoff, negative AO, positive AAO). Mean annual air temperature was 0.6 ºC below normal, and the annual ROS runoff was 57 mm (45%) above normal, the highest among the six phases (Table 3). Similar to phase one, high SWE accumulations and subsequent ROS runoff generations were spatially restricted to drift and north facing HRUs (Figure 10). The runoff depth was quite higher than peak SWE (Figure 10), indicating a substantial contribution from rainfall induced events to total runoff. Despite the similarity in snow, rain, and air temperature between this phase and phase one, a large difference is observed in ROS runoff. The only difference between the two phases is the type and phase of teleconnection patterns. Positive AAO is likely the main reason for high ROS runoff in this phase.
Phase four (2000-2003, warm, positive PNA). The annual and winter air temperatures were 0.7 °C and 0.5 °C above normal, respectively, causing this phase to be the warmest among the six phases with near-freezing winter temperatures (Table 3). Such hydrological responses suggested that Reynolds Mountain is very sensitive to changes in winter air temperature and warming of 0.5 ºC can shift winter temperatures from below-freezing to above-freezing conditions, resulting in reduction of the ROS runoff by 9% below the normal. A relatively low flow condition in this phase is associated with the positive phase of PNA.
Phase five (2004-2011, normal, negative NAO, positive PNA) . Similar to phase three, the runoff generated from ROS events is high in this phase (36 mm (20%) above normal, Table 3), which is the second largest contribution to annual runoff among the six phases. The runoff amounts are higher than peak SWEs across the basin, indicating the substantial contribution of rainfall to total annual runoff (Figure 10). Positive PNA partly moderated the effect of negative NAO on air temperature and precipitation. It, however, was not enough to offset the effect of the negative NAO on runoff, and as a result, a slight increase in annual runoff and a relatively large increase in the ROS runoff were observed in this phase.
Phase six (2012-2014, dry, low flow, low runoff ratio, positive NAO, negative PNA) . This phase represents an extreme hydroclimatic condition. The observed annual precipitation was 194 mm (20%) below normal and the observed mean annual air temperature was 0.5 ºC above normal. Warm and dry conditions caused observed peak SWE to drop 275 mm (55%) below normal, snow cover season to shorten 26 days below normal, and observed annual runoff to drop 234 mm (43%). The NAO and PNA phases are opposite of those in phase five, and as a result, there is an opposite response of the basin to phase change. The radical changes relative to phase five is associated with positive NAO, which led to dry, low snow, and low flow conditions. This clearly explains the important role of the phase of the climate teleconnection patterns in altering hydrological conditions in small basins.