Temporal Variability of Rain-on-Snow Events in the Western United States
Gregory J. McCabe1, Martyn P. Clark2, and Lauren E. Hay1
ABSTRACT
VARIABILITY OF RAIN-ON-SNOW EVENTS
The temporal and spatial variability of the monthly frequency of
rain-on-snow (ROS) events for 280 sites in the western United
States (U.S.) are characterized for the period 1981-2003. The
ROS events were identified using daily snow-telemetry
(SNOTEL) data and were defined as days when precipitation
occurred at a site and snowpack decreased. Results indicate
that for most sites analyzed the largest number of ROS events
occurred during the months of March through June when
snowpacks are prevalent and temperatures become warm
enough for rain to occur rather than snow. During these months
with the greatest frequency of ROS events there is little
correlation with elevation. However, during the winter months,
the frequency of events is strongly related with elevation, such
that the largest number of events occurs for low elevation sites.
In these instances winter temperatures at only low elevation
sites are warm enough for rain to occur. Correlations between
the frequency of ROS events and spring/early-summer
temperature and precipitation indicate that the frequency of
ROS events is greatest for periods that are cool and wet. In
addition, the periods with high ROS event frequencies are
related to periods with increased zonal atmospheric flow over
the western US which increases the intrusion of moist air from
the North Pacific Ocean into the western U.S.
The frequency of ROS events for most sites is largest for the
months of March through June (Figure 2). It is during these
months that snowpacks are prevalent in the western U.S. and
temperatures become high enough for rain to occur rather
than snow. After June, although temperatures are high
enough for most precipitation to be rain, snowpack
accumulations are depleted for most sites.
(1U.S. Geological Survey, Denver, Colorado, 2University of Colorado, Boulder, Colorado)
During the 1990s not only was there an increase in the
frequency of ROS events, but ROS events also
represented a greater proportion of all precipitation events
(Pevents) in the western U.S. (Figure 5).
Figure 5
Figure 2. Mean ROS event frequency (days per month)
for the period 1981-2003.
The increased frequency of ROS events during the 1990s
was driven mostly by an increase in the frequency of
Pevents rather than by increased temperature (see Table
1). Analysis of climatic records indicates that the temporal
variability of ROS events is positively correlated with
precipitation and negatively correlated with temperature,
which indicates that for the period analyzed, high-thanaverage ROS frequencies occur during cool/wet periods
(see Table 1).
RELATIONS WITH ELEVATION
There is only a weak correlation between the annual
frequency of ROS events and elevation, however on a
monthly basis there are strong correlations between monthly
ROS frequencies and elevation (Figure 7). During cold
season months (i.e. October through April) there is a strong
negative correlation between ROS frequency and elevation. It
is during these cold months that temperatures only at low
elevation sites reach high enough temperatures for rain to
occur rather than snow on some days. Thus, correlations
between the number of ROS events and elevation across all
sites produces significant negative correlations during these
months. During summer months (e.g. June through August)
the correlations between ROS frequency and elevation are
positive and weaker than those for the winter months. The
positive correlation with elevation results because during the
summer months only the highest elevations continue to have
snowpacks on which ROS events can occur.
Figure 7. Correlations between elevation and the
mean monthly frequency of ROS events for sites
in the western U.S.
DATA
Daily SNOTEL data for the years 1981 through 2003 were used
for the analysis. SNOTEL data provide observations of daily
precipitation and snowpack (expressed as snow water
equivalent (SWE)). An ROS event was defined as a day when
precipitation occurred but the snowpack was diminished. For
these days precipitation amount and snowpack reduction were
computed.
Table 1. Correlations between ROS event
characteristics and climate statistics.
Figure 1A illustrates the location of the sites used in the analysis
and figure 1B shows the number of sites with nearly complete
data (not missing more than 30 days of data in any year) for
periods ending in 2003, but with different beginning dates. The
period 1981-2003 was chosen for analysis yielding 280 sites.
EFFECTS OF ATMOSPHERIC CIRCULATION
Figure 1
The mean monthly number of ROS events averaged for all
sites in the western U.S. (Figure 3) clearly shows the
concentration of ROS events during March through June
(MAMJ). These data also suggest that there was a period of
increased ROS event frequency during the 1990s.
In addition to examining the frequency of ROS events, the
magnitude of SWE loss for ROS events also was
examined. On average the largest SWE losses per ROS
event occurred during the summer months when
temperatures are warm (Figure 6). Results also indicate
that the magnitude of SWE loss per ROS event is strongly
positively correlated with temperature (see Table 1).
Figure 3
Figure 6. Distributions of mean SWE losses for
ROS events for all sites in the western U.S.
Because temperature measurements were limited for SNOTEL
sites before the 1990s, monthly temperature and precipitation data
for the 84 climate divisions of the western U.S. were used to
compute mean temperature and precipitation for the western U.S.
The frequency of ROS events is related to atmospheric
pressure patterns over the eastern Pacific Ocean and North
America. During years with few ROS events atmospheric
pressure anomalies were positive over the western U.S.
(Figure 8A). This anomaly pattern indicates a strengthening of
the high pressure ridge that normally exists over the western
U.S. and indicates an increase in atmospheric subsidence
which results in a drying and warming of the air. In contrast,
during years with frequent ROS events the atmospheric
pressure pattern indicates lower-than-average atmospheric
pressures that extend from the eastern Pacific Ocean across
North America (Figure 8B). This pressure pattern suggests an
increase in the frequency and/or the magnitude of storm
events.
Figure 8. Mean 700-hectoPascal
height anomalies (in meters) for
MAMJ periods with (A) low and
(B) high frequencies of MAMJ
ROS events.