194
JOURNAL OF HYDROMETEOROLOGY
VOLUME 9
Rain versus Snow in the Sierra Nevada, California: Comparing Doppler Profiling
Radar and Surface Observations of Melting Level
JESSICA D. LUNDQUIST
Civil and Environmental Engineering, University of Washington, Seattle, Washington
PAUL J. NEIMAN
Physical Sciences Division, NOAA/Earth System Research Laboratory, Boulder, Colorado
BROOKS MARTNER, ALLEN B. WHITE,
AND
DANIEL J. GOTTAS
Cooperative Institute for Research in Environmental Science, University of Colorado, and Physical Sciences Division, NOAA/Earth
System Research Laboratory, Boulder, Colorado
F. MARTIN RALPH
Physical Sciences Division, NOAA/Earth System Research Laboratory, Boulder, Colorado
(Manuscript received 22 November 2006, in final form 16 July 2007)
ABSTRACT
The maritime mountain ranges of western North America span a wide range of elevations and are
extremely sensitive to flooding from warm winter storms, primarily because rain falls at higher elevations
and over a much greater fraction of a basin’s contributing area than during a typical storm. Accurate
predictions of this rain–snow line are crucial to hydrologic forecasting. This study examines how remotely
sensed atmospheric snow levels measured upstream of a mountain range (specifically, the bright band
measured above radar wind profilers) can be used to accurately portray the altitude of the surface transition
from snow to rain along the mountain’s windward slopes, focusing on measurements in the Sierra Nevada,
California, from 2001 to 2005. Snow accumulation varies with respect to surface temperature, diurnal cycles
in solar radiation, and fluctuations in the free-tropospheric melting level. At 1.5°C, 50% of precipitation
events fall as rain and 50% as snow, and on average, 50% of measured precipitation contributes to increases
in snow water equivalent (SWE). Between 2.5° and 3°C, snow is equally likely to melt or accumulate, with
most cases resulting in no change to SWE. Qualitatively, brightband heights (BBHs) detected by 915-MHz
profiling radars up to 300 km away from the American River study basin agree well with surface melting
patterns. Quantitatively, this agreement can be improved by adjusting the melting elevation based on the
spatial location of the profiler relative to the basin: BBHs decrease with increasing latitude and decreasing
distance to the windward slope of the Sierra Nevada. Because of diurnal heating and cooling by radiation
at the mountain surface, BBHs should also be adjusted to higher surface elevations near midday and lower
elevations near midnight.
1. Introduction
In the maritime mountain ranges of western North
America, most precipitation falls during the winter
months, with a large percentage falling in the form of
snow. Reservoir managers must balance the objective
of storing water for summer use with the need to re-
Corresponding author address: Jessica D. Lundquist, Civil and
Environmental Engineering, University of Washington, Seattle,
WA 98195-2700.
E-mail: [email protected]
DOI: 10.1175/2007JHM853.1
© 2008 American Meteorological Society
lease water to provide flood protection from warm winter storms. Coastal basins spanning a wide range of
elevations, such as the North Fork of the American
River basin in California (Fig. 1), are extremely sensitive to rain falling on snow at mid- to high elevations
(Kattelmann 1997; Osterhuber 1999). Many studies of
climatic change show that these same river basins are
extremely sensitive to regional warming. Over the past
50 yr, an increasingly greater percentage of precipitation has fallen in the form of rain rather than snow, and
this trend is likely to continue (Dettinger et al. 2004;
Jeton et al. 1996; Knowles et al. 2006; Lettenmaier and
APRIL 2008
LUNDQUIST ET AL.
195
FIG. 2. Cumulative fraction of total basin area contributing to
runoff as a function of altitude of rain–snow line for the North
Fork of the American River, CA.
FIG. 1. Locations of radar profilers (black circles), surface measurement sites (plus signs), and surface snow and temperature
sites (white circled plus signs). KOAK identifies the Oakland
sounding location.
Gan 1990). However, while the importance of flooding
and erosion associated with warm rain events has been
repeatedly stressed (Brunengo 1990; Ffolliott and
Brooks 1983; Hall and Hannaford 1983; Harr 1986;
Kattelmann 1997; Marks et al. 1998; McCabe et al.
2007; Zuzel et al. 1983), hydrologic forecasting centers
still struggle to predict the elevation at which snow
turns to rain in mountain watersheds (E. Strem, California–Nevada River Forecast Center, 2006, personal
communication).
In the West Coast mountains, particularly warm
storms result in floods primarily because rain falls at
higher elevations and over a much larger contributing
area for runoff than during a typical storm. Figure 2
illustrates the fraction of the North Fork American
River basin contributing to runoff based on the elevation below which precipitation falls as rain. As the rain–
snow line rises from 800 to 2800 m (a common range for
the melting level in California winter storms), the contributing fraction increases from 25% to 100% of the
basin, resulting in up to 4 times as much runoff for a
given rain rate. White et al. (2002) tested the importance of the melting level by comparing river forecast
simulations of peak flow as a function of melting level
in four California watersheds. In three of the four watersheds, runoff tripled if the snow level rose 600 m
(White et al. 2002). These results, combined with the
steep slope of the curve in Fig. 2, indicate that significant uncertainty in streamflow prediction can occur as a
result of relatively small errors in characterizing the
precise altitude at which snow accumulates in a given
storm. For these reasons it is important to monitor,
understand, and predict the details of the snow level to
within relatively small margins of error. This paper uses
unique measurements, collected during the National
Oceanic and Atmospheric Administration’s (NOAA)
Hydrometeorological Test Bed field campaign (Ralph
et al. 2005), to improve understanding of the melting
level and snow accumulation transition zone. The
analysis includes a novel radar technique to measure
conditions aloft in combination with more traditional in
situ observations at the earth’s surface.
Because of the difference in radar reflectivity and fall
speeds between frozen and liquid hydrometeors, Doppler profiling radars are able to detect the altitude in the
atmosphere where rain changes to snow (Battan 1973;
196
JOURNAL OF HYDROMETEOROLOGY
Cunningham 1947; Fabry and Zawadzki 1995; Marshall
et al. 1947; Mittermaier and Illingworth 2003; Stewart
et al. 1984; White et al. 2002). The dielectric constant
increases as melting ice particles become coated with
water, and a layer of enhanced radar reflectivity, or
“bright band” appears. The brightband height (BBH),
defined as the altitude of maximum radar reflectivity,
occurs on average about 200 m below the 0°C isotherm
(White et al. 2002; for measurements taken on the California coast) at temperatures usually between 1° and
2°C (Stewart et al. 1984; for measurements near the
Sierra Nevada). Doppler wind-profiling radars along
the coast and central valley of California have been
observing brightband heights since 1996, and an automated algorithm developed by White et al. (2002)
makes this information available in real time to river
forecasters.
For hydrologic applications, the BBH in the free atmosphere must be related to the melting level on the
surface some distance away. First, snow falling in the
free atmosphere starts melting at 0°C and changes to
rain over time and space as a complex function of temperature, humidity, winds, and the type and fall rate of
hydrometeors. Most winter storms impacting the
mountains of the west coast of North America possess
a fairly stable, stratiform structure (Medina et al. 2005),
with ice crystals forming aloft and then falling through
warmer atmospheric layers below, where they melt and
become rain. The vertical distance through which the
snow crystals pass before they are completely melted
depends on the size and density of the crystals, the fall
rate of the crystals (1 to 1.5 m s⫺1 for most snowflakes,
increasing to 4 to 9 m s⫺1 for mostly melted snowflakes;
Szyrmer and Zawadski 1999 and Mitra et al. 1990), the
ventilation of the hydrometeors, and the balance of
heat at the particle surface (due to conduction of heat
from the warmer surrounding air and to latent heat
released onto the melting particles during water vapor
condensation; Szyrmer and Zawadski 1999). Aircraft
observations indicate that, in general, the ratio of liquid
to total hydrometeor mass increases linearly as temperatures increase from 0.3° to 2.2°C (Stewart et al.
1984), over an atmospheric thickness of 200 to 500 m
(Fabry and Zawadzki 1995).
Second, the atmospheric structure and corresponding
melting levels change as an air mass intersects a mountain range, such as the Sierra Nevada, and is influenced
by boundary layer effects. Specifically, orographic lifting causes adiabatic cooling, and greater precipitation
rates increase diabatic cooling, frequently resulting in a
lowering of the BBH as it intersects the windward
slopes of a mountain range (Marwitz 1983, 1987; Medina et al. 2005). Diabatic cooling can also result in
VOLUME 9
down-valley airflow in mountain valleys, often in the
opposite direction of the general storm movement
(Steiner et al. 2003).
Detailed microphysical models are able to represent
all of the processes related to the melting level height
and thickness in one dimension but become computationally unwieldy in two or three dimensions, where
slight variations in precipitation microphysics can set up
mesoscale circulation patterns, which in turn feed back
and change hydrometeor fall rates and melting rates
(Szyrmer and Zawadski 1999). Thus, most operational
weather forecast models parameterize ice microphysics
(Barros and Lettenmaier 1994; Lin et al. 1983; Schultz
1995), and different parameterization schemes often result in quite different precipitation distributions over a
basin (Wang and Georgakakos 2005; Wang and Georgakakos 2007, manuscript submitted to J. Geophys.
Res.). Most hydrologic models discount atmospheric
microphysics altogether, instead relying on surface temperature cutoffs to delineate rain from snow (Anderson
1976; Storck 2000; Westrick and Mass 2001).
At the mountain surface, these near-surface air temperatures provide a good estimate of precipitation type,
such that at temperatures below 0°C, over 90% of precipitation events occur as snow; at temperatures above
3°C, over 90% of precipitation events occur as rain; and
rain and snow are equally likely at 1.5°C (U.S. Army
Corps of Engineers 1956). These empirically derived
distributions of rainfall versus snowfall as a function
of surface temperature are remarkably consistent
across the globe, with similar results found in the Alps
(Rohrer 1989), the Bolivian Andes (L’Hôte et al. 2005),
and Sweden (Feiccabrino and Lundberg 2008). Detailed observations in a mixed precipitation event in the
Oregon Cascades also fell within this range, where
Yuter et al. (2006) used a Parsivel disdrometer to observe that rain particles made up 5% of the total concentration of hydrometeors at 0°C, 23% at 0°–0.5°C,
and 93% at 0.5°–1.5°C.
Finally, the antecedent snow cover and ground cover
in a river basin affect the final form precipitation takes
on the surface—rain may freeze within a snowpack and
be stored for some time, or snow may melt on bare
ground and contribute to runoff. This last scenario—
what happens on the surface—is most important in
flood forecasting.
While the processes governing melting snowflakes in
the atmospheric melting layer and the relationship between surface temperature and snow accumulation are
relatively well known, the precise relationship between
measurements of the melting level aloft and snow accumulation behavior at the surface has not been well
documented, especially in complex terrain. This paper
APRIL 2008
LUNDQUIST ET AL.
197
FIG. 3. Schematic diagram of subjects discussed in this paper.
determines if BBHs measured atmospherically upstream of a mountain range can be used to infer melting
levels within a hydrologic basin. Specifically, we track
the altitude of the melting level from the free atmosphere, to the mountain boundary layer, to the ground
(or snow covered) surface for winters 2001–2005, as
illustrated in Fig. 3. Significant heights and temperatures are detailed in Table 1. Section 2 describes the
observations and the methods used for classifying precipitation as rain or snow. Section 3 presents two case
studies of mixed rain and snow events in the American
River basin to illustrate the rapid changes in melting
level and the complex surface responses associated with
winter storms. Section 4 establishes temperature as an
index for where surface snowmelt is taking place and
compares near-surface and free-atmosphere measurements during storm events. Section 4 also presents 5-yr
statistics of how these temperatures vary with location
and time of day, and how they relate to different melting altitudes and different contributing areas in the
TABLE 1. Summary of significant heights and temperatures.
FAML ⫽ free atmosphere melting level. 0°C isotherm. Height of the top of the melting layer in the free air, as measured by the
Oakland radiosonde near the coast.
BB ⫽ bright band. Layer of the atmosphere within which snow changes to rain (also called the melting layer). The top of the BB
is the FAML.
BBH ⫽ brightband height. Height of maximum reflectivity in the radar melting layer bright band. Below the FAML but above
the bottom of the BB.
Zs-half ⫽ half snow-runoff melting level at surface. Height along the Sierra mountainside at which half of the precipitation falling
in the basin contributes to snow accumulation and half to runoff. Also height with 50% chance rainfall and 50% snowfall.
Zs-50 ⫽ 50% probability melting level at surface. Height along the Sierra mountainside at which precipitation has a 50%
probability of at least partially accumulating as snow on the ground. Snow accumulation is increasingly more likely at altitudes
above this height, and snowmelt is increasingly more likely at altitudes below.
TFA ⫽ temperature in the free air, as measured well upwind of the Sierra Nevada by the Oakland radiosonde.
TS ⫽ temperature at the surface, along the Sierra mountainside.
BBH ⫽ Z km MSL, TFA ⬇ ⫹1.8°C.
FAML ⬇ Z ⫹ 0.2 km, TFA ⫽ 0°C.
Zs-half ⫽ BBH ⫺ f1(profiler site, time of day), TS ⫽ ⫹1.5°C.
Zs-50 ⫽ BBH ⫺ f2(profiler site, time of day), TS ⫽ ⫹3°C.
198
JOURNAL OF HYDROMETEOROLOGY
TABLE 2. Radar profilers.
Name
ID
Elev (m)
Lat (°N)
Lon (°W)
Bodega Bay
Chowchilla
Chico
Grass Valley
BBY
CHL
CCO
GVY
12
76
41
689
38.32
37.11
39.69
39.17
123.07
120.24
121.91
121.11
North Fork of the American River. Section 5 discusses
applications for streamflow forecasting and directions
for further research.
VOLUME 9
standard time (PST)] each day. This site is generally
upwind of the American River basin and represents
free-air conditions before they encounter the topography of the Sierra. Wet-bulb temperatures and virtual
temperatures were calculated from the measured variables (Jensen et al. 1990). The altitude of the 0°C isotherm (FAML, Table 1) and the temperature at the
BBH were both determined using linear interpolation
from available measurements. While this approach may
result in errors during the hours surrounding a frontal
passage, it is the best available approximation given the
existing data.
2. Observations and methods
c. Surface meteorology, snow, and hydrology
Free atmosphere melting levels (FAML; 0°C isotherm) were determined by temperature measurements
from radiosondes launched at Oakland, California, and
compared with BBHs at four 915-MHz radar profilers.
Boundary layer melting levels were determined from a
transect of surface temperatures in the American River
basin, and the fate of rain versus snow at the surface
was determined from precipitation, temperature, snow
water equivalent (SWE), and snow depth measurements at California Department of Water Resources
(CA DWR) snow pillow stations in the same basin. The
American River basin is located downwind of the profilers along the primary winter storm track, which is
from the west (Fig. 1).
The CA DWR manages a network of 41 automated
meteorological monitoring stations in or near the
American River basin (Fig. 1; Table 3). These stations
all measure precipitation and 23 of them also measure
air temperature, typically from a sensor mounted on a
pole or mast, approximately 10 m above ground level.
Thirteen of the stations have snow pillows that measure
the weight of snow accumulation and thereby indicate
the SWE of the snow column. Data were obtained from
CA DWR (http://cdec.water.ca.gov).
Most snow pillows are located in flat meadows, surrounded by forested areas that shelter the pillow from
wind scouring (Farnes 1967). Compared to nearby forested areas, these sites tend to accumulate more snow
and are also more exposed to sunlight during the melt
season. Because snowmelt during a precipitation event
is caused primarily by turbulent energy transfer, which
grows with increasing wind speed (Marks et al. 1998),
snow pillows may record more melt than a completely
forested area and significantly less melt than an unsheltered area. While snow pillows do not record the full
range of snow properties due to varying slope, aspect,
and vegetative cover, they provide the most comprehensive set of high-elevation measurements available.
The CA DWR snow pillows report information at
hourly intervals, but because pillows can experience
several hours of delay in responding to changes in SWE
(Beaumont 1965; Trabant and Clagett 1990), 12-h averages were used here. An increase in SWE may be due
to snow falling on the pillow or to liquid water falling
on snow already on the pillow and freezing into the
snowpack, thus increasing its density. Decreases in
SWE are due to snowmelt or sublimation. Redistribution of SWE from or to areas outside of the snow pillow
area may also be responsible for changes. However, this
effect is slight at most California snow pillows because
of the high density of the snowpack and their flat locations with wind shelter from the surrounding trees.
Snow depth measurements, which are only available
a. Radar profilers
The 915-MHz Doppler wind profilers provide wind
and precipitation profile measurements in the boundary layer and lower free troposphere, and data from
four profilers in California are analyzed here (Table 2;
Fig. 1). Signal-to-noise ratios (SNRs) indicate regions
of high reflectivities, and Doppler vertical velocities
(DVVs) approximate the fall speeds of hydrometeors.
BBHs were determined using the automated algorithm
detailed in White et al. (2002), which picks out the altitude where DVV starts decreasing with height, indicating the height of slower-falling snow particles, in tandem with SNR increasing with height, indicating the
greater reflectance of large, water-coated snow and ice
particles (Fig. 3). This level is determined with hourly
temporal resolution and 100-m vertical resolution.
b. Oakland radiosonde
Upper-air measurements of temperature, dewpoint
temperature, wind speed, and direction were obtained
from the National Weather Service (NWS) Radiosonde
Network site at Oakland, California, where they are
measured at 0000 and 1200 UTC [0400 and 1600 Pacific
APRIL 2008
199
LUNDQUIST ET AL.
TABLE 3. American River surface sites and types of measurements available. Note that Meadow Lake is adjacent to, but not within,
the American River basin.
Station name
Elev (m)
Lat (°N)
Lon (°W)
Precipitation
Temperature
SWE
Depth
Cal State Sacramento
Arden Way
Hurley
Rio Linda W.C.
Arcade Creek
Cresta Park
Rancho Cordova
Royer Park/Dry Creek
Roseville Fire Station
Lincoln
Chicago
Orangevale W.C.
Prairie City
Folsom Dam
Roseville Water Tr.
Folsom Water Tr.
Caperton Reservoir
Loomis Observatory
New Castle
Auburn Dam Ridge
Pilot Hill (CDF)
Georgetown (USBR)
Pacific House
Sugar Pine
Owens Camp
Hell Hole (USFS)
Robb’s Powerhouse*
Blue Canyon*
Sugarloaf near Kyburz
Greek Store*
Robb’s Saddle*
Huysink*
Van Vleck*
Duncan
Silver Lake
Alpha (SMUD)*
Forni Ridge*
Beta
Caple’s Lake (DWR)*
Carson Pass
Schneider’s*
Meadow Lake
8
11
11
14
21
21
22
43
46
61
64
71
95
107
116
129
189
223
271
366
366
991
1048
1171
1372
1396
1570
1609
1696
1707
1798
2012
2042
2164
2164
2316
2316
2316
2438
2546
2667
2194
38.555
38.596
38.587
38.7
38.645
38.593
38.604
38.749
38.76
38.882
38.652
38.686
38.592
38.7
38.724
38.687
38.863
38.815
38.874
38.882
38.832
38.925
38.76
39.128
38.733
39.078
38.903
39.276
38.7839
39.075
38.912
39.282
38.945
39.144
38.678
38.805
38.805
38.8
38.71
38.6924
38.747
39.417
121.416
121.413
121.407
121.448
121.347
121.368
121.311
121.28
121.315
121.272
121.254
121.219
121.161
121.167
121.226
121.149
121.218
121.113
121.135
121.045
121.012
120.789
120.5
120.75
120.245
120.437
120.375
120.708
120.31
120.558
120.378
120.527
120.305
120.509
120.118
120.215
120.213
120.2
120.042
120.0021
120.068
120.508
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
1
0
0
0
1
1
1
0
0
0
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
1
1
1
1
0
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
0
1
1
0
0
* Stations used in snowmelt vs temperature contour plots.
at a small subset of the stations (Blue Canyon, Meadow
Lake, and Caple’s Lake; Fig. 1b), acoustically detect
changes in snow depth at an hourly time scale. An increase in snow depth can only occur with snowfall or
with deposits of drifting snow. However, the latter is
rare during a rain event in the Sierra Nevada. A decrease in snow depth could be due to melt or compaction. These changes are accurate on an hourly time
scale but also include noise, particularly during a snow
storm, when the acoustic signal can reflect off of falling
snowflakes. To minimize the noise, we smoothed the
depth record with a 20-h Blackman-window filter and
then established a cutoff where we declared snowfall to
occur in any hour when the depth increased by at least
0.35 cm.
Most precipitation sensors in the region consist of a
reservoir with antifreeze. These report accumulated
precipitation and are also subject to instrumental noise.
Incremental precipitation was determined by smoothing the record with a 20-h Blackman-window filter and
taking the difference between hourly reports of the
smoothed accumulated precipitation. A cutoff of
200
JOURNAL OF HYDROMETEOROLOGY
VOLUME 9
FIG. 4. (a) Radar brightband heights and (b) NF American River discharge (normalized by
basin area on the left axis and actual quantity on the right axis) for winter 2005 (PST). Case
study periods are marked by vertical dashed lines.
0.04 cm h⫺1 was used to indicate that precipitation actually occurred. Blue Canyon was equipped with a
heated tipping bucket in September 2004 to measure
incremental precipitation. Because this instrument is
more sensitive than the antifreeze reservoirs, a cutoff of
0.02 cm h⫺1 was used to signify the occurrence of precipitation at this station. This is equivalent to 1 tip of
the bucket in the rain gauge (0.01 in. or 0.254 mm).
Mean daily discharge measurements were examined
for the North Fork (NF) of the American River (Fig. 1).
It does not have dams or diversions and is included in
the U.S. Geological Survey (USGS) hydroclimatic
dataset (Slack and Landwehr 1992). The record began
in 1941, the gauge is at an elevation of 218 m, basin area
is 886 km2, and mean basin elevation is 1190 m.
d. Temperature–SWE PDF
The fate of precipitation after it reached the ground
was determined by creating a probability density function (PDF) of change in SWE given a precipitation
event at a given temperature. Data at all snow pillows
were quality controlled, and 10 stations (indicated in
Table 3) were selected for having reliable measurements of both SWE and collocated temperature.
Hourly SWE, surface air temperature, and precipitation observations were averaged every 12 h. Next, the
change in SWE and the incremental precipitation over
each 12-h interval was calculated. Because of the lack of
quality precipitation data in the snow zone, precipitation data at all 40 stations (Table 3) were considered. If
10 or more of these stations reported increased precipitation, then the precipitation rate was determined as
the average of all stations. If less than 10 stations reported precipitation, then the precipitation rate was set
at zero, and that 12-h time step was not considered in
the subsequent analysis. Data were binned by temperatures (at 0.5°C increments) and normalized by the number of measurements at each temperature to produce
probability density functions.
3. Case studies—Radar brightband levels and
surface rain, temperature, and snow
Two case studies from winter 2005 were chosen to
illustrate how radar measurements of BBH correspond
with measurements of temperature, rainfall, and snow
gain or loss at various surface elevations. BBHs varied
from 600 to 3000 m during the winter 2005 period, and
increased NF American River streamflow generally
corresponded to warm storms, indicated by high-altitude BBHs (Fig. 4). Warm storms not only resulted in
liquid precipitation falling over large fractions of the
basin, but they also carried the most moisture and,
hence, the highest precipitation rates. The two case
study periods (marked, Fig. 4) both encompassed
storms with a wide range of melting levels, but the two
had very different streamflow responses.
a. January 2005
The 25 to 29 January case provided a simple example
of a cold front passing with dropping melting levels in
early winter. Rain falling at Blue Canyon (1600 m) was
incorporated into the snowpack, increasing SWE, with
minimal contributions to streamflow. Radar returns at
APRIL 2008
LUNDQUIST ET AL.
201
FIG. 5. Wind profiles and BBHs for the Grass Valley profiler (PST) for (top) case study 1 and (bottom) the last 4 days of case
study 2.
the Grass Valley wind profiler showed decreasing
BBHs during this period (Fig. 5a). Surface temperatures at Blue Canyon (1600 m) fell from 10° to ⫺1°C,
and temperatures at Caple’s Lake (2400 m) fell from
10° to ⫺11°C (Fig. 6a). On 25 January, precipitation
increased (Fig. 6b), and snow depth decreased slightly
at both Caple’s Lake and Blue Canyon (Fig. 6c), indicating that rain compacted the snowpack. On 26 January, snow depth mostly increased at Caple’s Lake (with
a slight decrease near noon) and decreased at Blue
Canyon, indicating that while mostly snow fell at Caple’s Lake, rain fell at Blue Canyon (Fig. 6c). SWE
increased at both locations (Fig. 6d), indicating that at
least some of the rain at Blue Canyon was incorporated
into the snowpack, increasing its density and water content, and not directly contributing to streamflow (Fig.
4). By the 28 January precipitation event, temperatures
were at or below freezing at both locations, and snow
fell at both locations, as indicated by increases in snow
depths and SWE.
Changes in temperature (Fig. 7a) and precipitation
type (Fig. 7b) with elevation corresponded with dropping radar BBHs at both Bodega Bay (the profiler on
the coast, west of the American River basin) and at
Grass Valley (the profiler in the Sierra foothills, just
slightly north of the American River basin). On 25
January the entire basin from 1500- to 2500-m elevation
was warmer than 2°C, and rain fell at both Caple’s Lake
and Blue Canyon. Bodega Bay melting levels on 25
January were at about 2000 m, but this cooler storm
sector did not arrive at the American River until midnight that night, when Grass Valley reported melting
levels at about 2000 m, and temperatures above that
altitude dropped below 0°C. At this time, snow fell at
Caple’s Lake and Meadow Lake, while rain continued
at Blue Canyon. On 26 January daytime rainfall and
temperatures above 2°C occurred at Caple’s Lake,
when the Grass Valley BBH was about 500 m lower.
This may be explained by Caple’s Lake’s location being
107 km southeast of Grass Valley (Fig. 1), thus remaining in a warmer sector of the storm. When precipitation
resumed near midnight on 27 January, temperatures
were much colder, and the BBH at Grass Valley was
below 1500 m. Snow fell at all monitored locations.
Some profiler measurements at Bodega Bay and Grass
Valley indicated BBHs slightly above the altitude of
Blue Canyon. While surface measurements indicated
that Blue Canyon had a net gain of snow, a mixture of
rain and snow may have been falling from the sky. The
BBH may also have dropped in the distance between
Grass Valley and Blue Canyon (about 44 km).
b. March 2005
The 19–22 March case illustrated a more complicated
sequence of storms, alternating between warm and
cold conditions (Fig. 5b). During the second warm
202
JOURNAL OF HYDROMETEOROLOGY
VOLUME 9
snow in addition to falling rain) at Blue Canyon (Fig.
8). At midday on 22 March, Caple’s Lake reported an
increase in temperature, a sharp increase in precipitation, and slight drops in snow depth and SWE. After
this time, temperatures dropped at both stations, rain
switched to snow, and snow depths and SWE increased.
These patterns of alternating warm and cold storm
sectors were also observed by the profilers (Fig. 9).
Temperatures at all altitudes above 1500 m dropped
below 0°C on 19 and 20 March, following the drop of
the BBH at Grass Valley. BBH changes at Bodega Bay
preceded those at Grass Valley by about 4 h. The switch
from rain to snow at Blue Canyon on 22 March coincided with drops in BBHs at both Bodega Bay and
Grass Valley. However, data at Caple’s Lake indicated
an earlier switch to snow, and Meadow Lake reported
increases in snow depth, that is, continuous snow rather
than rain, throughout the warm sector of the storm.
During the warm sector (early 22 March), rain fell at a
rate greater than 4 mm h⫺1 over more than 90% of the
basin. This, combined with already-saturated snowpacks and soils, resulted in a large streamflow response
to this storm (Fig. 4).
c. Case study summary
FIG. 6. Surface measurements at Blue Canyon (1600 m) and
Caple’s Lake (2400 m) between 24 and 30 Jan 2005 (PST) of (a)
air temperature, (b) precipitation rates, (c) changes in snow
depth, and (d) changes in SWE. To aid in comparison of snow
changes, both snow depth and SWE are offset to equal 0 at the
start of the analysis period.
period, snow melted while rain fell at high rates at Blue
Canyon, and American River discharge rose steeply. At
the start of the storm on 19 March, snow depth decreased slightly at Blue Canyon but then began increasing at midday as temperatures dropped and rain
changed to snow (Fig. 8). Snowfall continued through
midday on 21 March, and then temperatures rose during a break in the storm. The second storm began several hours later and was considerably warmer, with a
decrease in snow depth and SWE (indicating melting
The case studies above illustrate that qualitatively,
radar BBHs represent the patterns and trends of melting levels on the mountain surface at an hourly time
scale. However, the case studies also illustrate the difficulty of quantitatively determining the form and fate
of precipitation hitting the ground, particularly in regions without human observers. From the case studies,
several important points emerge. First, while cloud microphysics are difficult to model and represent spatially
across a basin, preexisting surface snow characteristics
are often more important than hydrometeor distributions in determining where precipitation will contribute
to runoff. For example, rain at Blue Canyon was incorporated into the snowpack during case 1 but contributed to snowmelt and runoff during case 2, likely because
case 1 occurred earlier in the year when the snowpack
was colder and not yet saturated. Thus, improvements
in model representations of snowpack temperature and
liquid water content would help improve hydrologic
forecasts.
Second, distributions of rain versus snow vary spatially and are not solely functions of time and elevation.
For example, Caple’s Lake to the southeast appeared
to be in a warmer sector of the storm during case study
1, receiving rain when the altitude of the BBH measured at the Bodega Bay and Grass Valley profilers
indicated snow should fall at that elevation. During
case 1, the Bodega Bay and Grass Valley profilers
APRIL 2008
LUNDQUIST ET AL.
203
FIG. 7. (a) Surface temperatures at seven snow pillow locations, plotted at each location’s
elevation between 24 and 30 Jan 2005 (PST). Times with no mark indicate the freeatmosphere transition zone of a mixture of rain and snow with 0° ⬍ T ⬍ 2°C. Black squares
are brightband melting levels detected at Bodega Bay (BBY). Gray circles are brightband
levels detected at Grass Valley (GVY). (b) Increases in precipitation and snow depth at three
snow pillow locations (Caple’s Lake, Meadow Lake, and Blue Canyon). Precipitation at
Meadow Lake is inferred from records at Alpha, the station nearest in elevation. Black
triangles (snow) indicate increases in snow depth, and gray upside-down triangles (rain)
indicate increases in precipitation. Brightband heights are the same as in (a).
agreed on the BBH when they reported measurements
at the same time, but during case 2, Grass Valley consistently reported a lower-altitude BBH than Bodega
Bay. In both cases, the Grass Valley measurement corresponded better with surface temperature measurements. This highlights the danger of comparing one
profiler with one surface station, and the importance of
examining the profiler measurements in relation not
only to each other but to as many surface measurements as possible to better understand statistical variations across the region of interest.
4. Temperatures and melting statistics from 2001
to 2005
Few measurements of precipitation type exist, and
these are noisy. However, temperature measurements
are reliable, continuous, and widely available. By indexing SWE gain or loss during a precipitation event to
temperature and then comparing elevational transects
of surface temperatures to radar-detected BBHs, we
can determine statistically whether and by how much
the BBH should be adjusted up or down to represent
the surface melting level as a function of location and
time.
a. Free-air temperatures at BBHs
Oakland sounding temperatures, sampled every 12 h,
were interpolated to hours when BBHs were detected
at Bodega Bay (located about 100 km northwest of
Oakland). Free-air temperatures (T ) above Oakland at
the Bodega Bay BBH averaged 1.8°C, and wet-bulb
temperatures (Tw) averaged 0.3°C for the five winters.
The standard deviation for both T and Tw over the 5-yr
period was approximately 1.5°C, probably representing
changing thicknesses of the melting region and different storm structures, orientations, and motions relative
to these two sites. These values did not vary with the
hour of the day and are consistent with prior observations.
Because the other stations were considerably further
from Oakland than Bodega Bay, they sampled air
masses with different vertical temperature structures.
Chico (182 km northeast of Bodega Bay) sampled
colder air masses, so that at times when both stations
had measurements during the 5-yr period, BBHs were
on average 174 m lower at Chico than at Bodega Bay.
Chowchilla (283 km southeast of Bodega Bay) sampled
warmer air masses, so BBHs at Chowchilla were on
average 49 m higher than at Bodega Bay. Grass Valley
is only slightly north of Bodega Bay, but 195 km inland,
and it averaged melting levels 106 m lower. Grass Valley is on the windward slope of the Sierra, so this result
is consistent with observations from studies in the Sierra (Marwitz 1983, 1987) and the Alps and Cascades
(Medina et al. 2005) that show that BBHs dip to lower
elevations on the windward slopes of mountains.
204
JOURNAL OF HYDROMETEOROLOGY
VOLUME 9
heights at the three nonfoothill stations yielded a decrease in average BBH of 41.4 m per increase in degree
latitude. The average BBH at Grass Valley is 73 m
lower than would be expected based on its latitude and
this line. This is smaller than the approximate 400 m
decrease in BBH observed where the bright band intersected the mountain in prior studies (Marwitz 1983,
1987; Medina et al. 2005), perhaps because Grass Valley is generally lower in elevation than the freezing
level and is only detecting the beginning of the dip in
BBH as storms intersect the mountain slopes.
b. Temperature at which snow melts on the surface
FIG. 8. Surface data for Blue Canyon and Caple’s Lake, as in
Fig. 6, but for 18 to 24 Mar 2005 (PST). Note changed limits for
each axis from Fig. 6 (smaller temperature range and increased
precipitation and depth ranges).
Mean BBHs for each station (averaged over the entire 5-yr period, without regard to having simultaneous
measurement of the melting level with a neighboring
station) were 1796 m at Bodega Bay, 1744 m at Chico,
1691 m at Grass Valley, and 1851 m at Chowchilla.
Thus, on average, BBHs in California storms decrease
with increasing latitude and with proximity to the windward slope of the Sierra Nevada. These spatial slopes in
BBHs should be considered when applying melting
level information to a river basin north, south, or inland
of a given profiler. For example, a line fit to the average
To compare BBHs with the snow response at the
surface, we created a probability density function of
change in SWE during precipitation events as a function of temperature. Because of the scarcity of humidity
data measured at the surface, these are determined
based on actual, and not wet-bulb, temperatures. Based
on the similar structure and variance of actual and webbulb temperatures measured in the free atmosphere
during these events (section 4a), we assume little information is lost by neglecting surface humidity information. Figure 10a shows the likelihood of SWE gain or
loss over a 12-h period, given a local surface temperature, at the 10 snow pillow stations indicated in Table 2
for 2001–05, as a ratio of change in SWE to the average
basin precipitation. Although there is much scatter in
the data, several features emerge. Below 0°C there is
about 90% chance that snow will accumulate, and
about 60% chance that snow will accumulate at a mean
rate equal to or larger than the measured rate of average basin precipitation. The likelihood of the latter is
due both to orographic precipitation enhancement (the
snow pillows are all in the upper elevations of the basin
while most precipitation sensors are at lower elevations) and to the precipitation sensors’ likelihood of
undercatch during mixed rain and snow events. Above
3°C the surface snowpack is ⬎60% likely to melt (Fig.
10b). In between 0° and 3°C, precipitation may fall as
rain, snow, or a mixture of the two, with probabilities
transitioning from likely snow to likely rain as the temperature rises within this range [Fig. 10c, determined by
the U.S. Army Corps of Engineers (1956) from direct
observations near the Sierra crest during winters 1946–
51]. At 1.5°C, 50% of precipitation events fall as rain
and 50% as snow, and on average, 50% of measured
precipitation contributes to increases in SWE. Between
2.5° and 3°C, snow is equally likely to melt or accumulate, with most cases resulting in no change to SWE
(Figs. 10a and 10b). This is slightly warmer than the
free-atmosphere temperature where snow converts to
APRIL 2008
LUNDQUIST ET AL.
205
FIG. 9. (a) Surface temperatures and (b) forms of surface precipitation compared to
radar-measured melting levels, as in Fig. 7, but for 18–24 Mar 2005 (PST).
rain because this value measures the fate of the precipitation phase at the surface. Cold rain and rain–snow
mixtures falling on an existing snowpack are often frozen and incorporated into the snowpack, thus resulting
in a measured increase in SWE. In other cases, rain
passes through the snowpack with no measured change
in SWE.
These statistics were checked for differences between
night and day, for differences with averaging period (6
and 24 h), and for differences in temperature bin size
(0.25° or 1°C). The resulting minor differences were not
statistically significant.
c. Surface temperatures compared with brightband
heights
Section 4a demonstrated that, on average, the BBH
corresponds to a free-air temperature of 1.8°C. Section
4b demonstrated that roughly half of the precipitation
passes through the snowpack as rain at a surface temperature of 1.5°C (Ts-half), and existing snow cover becomes likely to melt during a precipitation event with
surface temperatures above 3°C (Ts-50). These indices
provide a means to translate the radar BBHs into surface hydrologic events. Figure 11 shows the mean surface temperature in the American River basin at the
altitude of each radar’s BBH for each hour of the day.
At all stations, there was a clear diurnal cycle, where
the surface was warmer during the day and cooler at
night. Chowchilla, to the south, measured higher BBHs,
corresponding to colder surface temperatures in the
American River basin. Chico, to the north, measured
lower BBHs, corresponding to warmer surface tem-
peratures in the American River basin. Bodega Bay, on
the coast, varied the most diurnally from surface temperatures, and Grass Valley, in the Sierra foothills, varied the least diurnally. This could be because it was
often not yet precipitating in the American River basin
when it was raining at Bodega Bay, or it could be because air masses above Bodega Bay arrived straight off
the ocean and were least affected by diurnal heating
and cooling from the continent. Temperatures corresponding to radar melting levels were quite variable in
all cases, as represented by standard deviations roughly
between 1° and 2°C at all hours of the day (Fig. 11b).
Standard deviations were most variable at Bodega Bay
and least variable at Grass Valley because Grass Valley
is geographically closest to the American River basin.
The diurnal variation in surface temperatures at
BBHs can best be explained by revisiting our March
2005 case study (Fig. 12). BBHs at Bodega Bay and
Grass Valley generally fell within the 0° to 3°C temperature contours determined from Oakland soundings
(Fig. 12a). However, the correspondence between BBH
and temperature became less well defined when compared with contours of surface temperatures (from the
23 stations in Table 3) versus elevation (Fig. 12b).
These deviations were due primarily to diurnal changes
in solar radiation, which heats the surface and not the
free air even during rain events. The differences between surface and free-air temperatures were particularly noticeable during the onset and end of each storm
(Fig. 12c), when more sunlight reached the basin.
While radiation is a relatively small component of the
total energy balance during storm events (18%–30% of
the energy available for snowmelt during rain-on-snow
206
JOURNAL OF HYDROMETEOROLOGY
VOLUME 9
FIG. 11. (a) Mean and (b) standard deviation of American River
basin surface temperatures at the elevation closest to the measured BBH as a function of hour of day (PST).
FIG. 10. The probability of (a) fractional SWE gain or loss
during 12-h periods with precipitation for the 10 American River
basin snow pillow stations marked with an asterisk in Table 3 for
the five winters from 2001 to 2005. Contours show probability of
distributions at each temperature, and the heavy dashed line
shows the mean. (b) Percent of occurrences with 100% of precipitation contributing to SWE gain (⌬SWE/⌬precip ⱖ 1, dashed
line, left axis), only part of precipitation contributing to SWE gain
(0 ⱖ ⌬SWE/⌬precip ⬍ 1, width of region between dashed and
solid lines), and SWE loss (⌬SWE/⌬precip ⬍ 0, solid line, right
axis). (c) Percent of occurrences of snow (dashed line, left axis),
mixed rain and snow (width of region between dashed and solid
lines), and rain (solid line, right axis) observed falling from the sky
at Donner Summit. [Note: (c) is based on data collected by the
U.S. Army Corps of Engineers (1956).]
events in Oregon in February 1996; Marks et al. 1998),
it is by far the most predictable component and has
consistent diurnal patterns. Marks et al. (1998) observed daily fluctuations in net radiation during rainon-snow events ranging from ⫺30 W m⫺2 at night to a
peak of 50 to 100 W m⫺2 at solar noon. While this is
about half (or less) of the diurnal variation observed
during a subsequent sunny period, it is not negligible.
This range in radiation alone would result in peak daytime melt rates about 1.3 mm h⫺1 higher than nighttime
minimum melt rates. Assuming that the specific heat of
vegetation (2928 J kg⫺1 K⫺1; Thom 1975) controls local
surface temperatures, this radiation increase would
raise peak daytime surface temperatures about 1.5°C
higher than nighttime minimum temperatures. This
value is similar to the observed differences in daily
variations between the radar and surface temperatures.
These results suggest that melting levels at the surface should be adjusted up (to higher elevations) during
daylight hours and down (to lower elevations) during
nighttime hours, particularly at the onset of a storm.
APRIL 2008
LUNDQUIST ET AL.
207
FIG. 13. (a) Mean and (b) std dev of the height difference in
meters between the altitude of the 1.5°C contour on the surface of
the American River basin and the radar-measured brightband
height.
FIG. 12. (a) Profile of temperatures observed at Oakland sounding compared with observations of melting level at Bodega Bay
(black squares) and Grass Valley (white circles) for 18 to 25 Mar
2005, in local time. Heavy black contour represents 0°C, and each
contour line represents a change of 1°C, with lighter shades indicating warmer and darker shades indicating colder temperatures.
(b) As in (a), but for surface temperatures in the American River
basin. (c) Difference between surface and free-air temperatures
shown in (b) and (a). Positive values indicate warmer surface
temperatures.
Figure 13 illustrates what these adjustments might be,
on average, for corresponding each radar’s measured
BBH with the 1.5°C contour, that is, the level where
precipitation is equally likely to fall as rain or snow, in
the American River basin. Bodega Bay BBHs needed
to be adjusted up 270 m during daylight hours (0600 to
1800 PST) and adjusted down 8 m during nighttime
hours (1800 to 0600 PST). Grass Valley (the profiler
closest to the ARB) needed to be adjusted down 78 m
during the day and down 205 m at night. When compared with the surface 3°C contour, where surface snow
cover is likely to melt, Bodega Bay heights were an
average of 223 m too high during the day and 488 m too
high during the night. Grass Valley heights were an
average of 318 m too high during the day and 433 m too
high at night. Thus, surface snowmelt is always lower
than the elevation where falling snowflakes melt to
rain.
208
JOURNAL OF HYDROMETEOROLOGY
VOLUME 9
TABLE 4. RMSE (m) resulting from assuming each station’s brightband height corresponds to a given surface temperature contour;
1.5°C represents the point where snow switches to rain, and 3°C represents the point where an existing snowpack melts.
Ts ⫽ 1.5°C
Ts ⫽ 1.5°C
Ts ⫽ 1.5°C
Ts ⫽ 3°C
Ts ⫽ 3°C
Ts ⫽ 3°C
RMSE (m)
No correction
Constant offset
Diurnal correction
No correction
Constant offset
Diurnal correction
Bodega
Grass Valley
Chico
Chowchilla
443
326
328
457
425
296
323
320
371
273
296
302
551
478
444
644
431
301
338
317
377
283
312
306
d. Uncertainties, corrections, and application to
hydrologic nowcasting
The spatial and diurnal patterns of how BBHs relate
to surface temperatures in the Sierra Nevada provide
guidance for how these radar measurements can best be
employed in hydrologic forecasting. There is a great
deal of uncertainty in all of the measurements, but the
data presented here can help quantify that uncertainty
and, to some degree, reduce it. BBH can be determined
to within 100-m vertical resolution (White et al. 2002).
However, the error associated with directly interpreting
that height as the elevation where snowfall changes to
rain near the surface (here inferred to be 1.5°C) ranged
from 326 to 457 m (Table 4). Assigning constant elevation offsets based on how each location related geographically to the American River basin reduced those
RMSEs to 296 to 425 m (Table 4). Applying a further
correction, with different offsets based on time of day
(i.e., as illustrated in Fig. 13), reduced the RMSEs to
273 to 371 m (Table 4). The average improvement in
RMSE through this process was 78 m. If we want to
define the surface melting level as the height where the
snowpack is likely to melt (here inferred to be 3°C),
offset corrections become even more important in adjusting the radar melting levels to lower elevations.
Table 4 details the RMSEs for the original and corrected heights at this level, which resulted in an average
improvement of roughly 200 m.
Over most of the North Fork of the American River
basin, the increase in contributing area with elevation is
linear (Fig. 2), such that a change in melting level of 100
m corresponds to a change in runoff contributing area
of about 48 km2, or 5% of the total basin area. Thus, an
RMSE of 500 m corresponds to an error of 25% of the
basin’s contributing area. A decrease of RMSE of 78 m
represents a hydrologic forecasting improvement of 4%
contributing area, and a decrease in RMSE of 200 m
represents an improvement of 10% contributing area.
For example, consider the March 2005 case study peak
of 1.3 mm h⫺1 (11 300 ft3 s⫺1). Assuming a constant rain
rate over the basin, 4% corresponds to 0.05 mm h⫺1
(441 ft3 s⫺1), and 10% corresponds to 0.13 mm h⫺1
(1130 ft3 s⫺1). However, even after making the corrections detailed above, the remaining RMSE is about 300
m, corresponding to 15% of the basin contributing area,
leaving room for further improvements.
5. Conclusions
a. Summary
The elevation where precipitation transitions from
contributing to snow water storage versus contributing
to runoff is critical for hydrologic monitoring and forecasting all along the western coast of North America.
Vertically pointing profiling radars are able to detect
the elevation where snow changes to rain in the free
atmosphere. Thus, a critical step toward improving hydrologic forecasts in these mountain basins is to quantitatively relate the radar-observed snow level to what
is happening on the ground. Specifically, forecasters
need to know which altitude offset would best distinguish between surface snow accumulation and runoff
and need to characterize the uncertainty in this vertical
offset.
To accomplish these objectives, this study tracked
the altitude of the melting level from the free atmosphere, to the mountain boundary layer, to the ground
(or snow covered) surface for winters 2001–05 (Fig. 3)
to address three major issues. First, how well do BBHs
measured at profiler sites located some distance from
each other and from a river basin represent melting
levels on the mountain surface? Because of the largescale nature of most of California’s winter storm events,
radars quite distant (300 km) from a river basin can be
used to interpret surface melting levels. On average,
RMSEs of 326 to 457 m occur when BBHs are directly
assumed to represent surface melting levels. Additional
local errors may arise when different basin locations are
in warm versus cold sectors of a frontal passage or when
drainage flows resulting from evaporative cooling develop in canyons (Steiner et al. 2003).
Second, what offsets, as a function of profiler and
basin locations and of time of day, should be used to
adjust BBH measurements to better match surface
APRIL 2008
LUNDQUIST ET AL.
melting levels? Forecasters should incorporate spatial
gradients of decreasing BBHs with increasing latitude
(on average, ⫺41.4 m °⫺1), due to regional temperature
gradients, and with decreasing distance to where the
bright band intersects the windward mountain slope,
due to adiabatic and diabatic cooling (Medina et al.
2005). In addition to these spatial offsets, the fate of
precipitation at the mountain surface will also be modified by diurnal cycles in solar radiation, which result in
warmer temperatures and greater melt rates during
daylight hours, even during cloudy precipitation events.
Depending on a profiler’s location, BBHs would need
to be shifted down 127 to 278 m further during the night
than during the day to correspond with surface observations. Incorporating both space and time of day to
BBH offsets reduced RMSEs in predicted surface melting levels to between 273 and 371 m, an average of 73
m less error than when these corrections were not
implemented.
Third, given an existing snowpack, what are the probabilities of snow accumulating versus melting at a given
temperature during a precipitation event? Surface measurements indicate that precipitation has an equal likelihood of falling as rain or snow at a surface temperature of 1.5°C, but a preexisting snowpack is 50% likely
to melt at a surface temperature of 3°C. Case studies
illustrate that rain may be incorporated into the snowpack, pass through the snowpack, or be accompanied
by melting snow, depending on antecedent snow temperature and liquid water content.
b. Applications and implications for model
development
Accurate forecasts or nowcasts of the altitude where
snow changes to rain are crucial for hydrologic forecasting not only in mountain basins in California but
also all along the western coast. The current “rule-ofthumb” used by the California–Nevada River Forecast
Center (CNRFC) for forecasting the surface melting
level is to take the forecasted freezing level (0°C isotherm) and subtract 300 to 460 m (1000 to 1500 ft; E.
Strem, CNRFC, 2006, personal communication). This
melting level, combined with basin areas above and
below, is then used to determine how mean areal precipitation for a basin is partitioned into rain and snow.
This paper provides statistical relationships for how
BBHs could be used to check and improve these forecast adjustments.
The observations described here illustrate the physical processes that must be included in a model to accurately represent elevation zones of rain versus snow to
determine hydrologic runoff in any area with orographic precipitation. First, the model must accurately
209
represent how the atmospheric melting layer elevation
and thickness change with space and time, by reproducing large-scale latitudinal gradients and topographically
induced changes, such as slope-induced cooling (Medina et al. 2005). Second, the model must include spatial representation of the surface energy balance to account for the effects of radiation on near-surface air
temperatures and melting. This could be accomplished
most precisely by modeling the thickness and optical
properties of the cloud cover (radiation transmission)
and the radiative properties (albedo) of the land surface cover. Radiative heating is neglected in most atmospheric models of orographically induced precipitation (Barros and Lettenmaier 1994), but could have
important feedback effects on not only the type and
fate of hydrometeors but also on the type and location
of secondary mesoscale circulation patterns (Szyrmer
and Zawadski 1999). Finally, the model needs a spatial
accounting of snowpack characteristics, such as density,
temperature, and liquid water content (as in the Utah
Energy Balance Model; Tarboton and Luce 1996), to
determine when and where rain will be incorporated
into the snowpack, as opposed to directly contributing
to runoff. Collectively, these modeling requirements
consist of model components that have already been
developed in various separate forms and should be
much simpler to implement than those required for accurate quantitative precipitation estimation (QPE;
Wang and Georgakakos 2007, manuscript submitted to
J. Geophys. Res.) Accurate implementation of the
above processes should improve model representations
of where precipitation contributes to runoff, and thus
would provide immediate benefits to hydrologic forecasting in coastal mountain regions.
In addition to providing ad hoc information for operational forecasts, vertically pointing radars also provide a resource for model testing and development. The
California and Oregon coasts are unique in having over
10 yr (since 1996) of vertically pointing Doppler radar
measurements, which include BBHs. These could be
used to check model representations of atmospheric
melting levels as a function of space and time. Most
tests of microphysics parameterization schemes are
constrained to the short duration of major field campaigns [Sierra Cooperative Pilot Project (SCPP; Meyers
and Cotton 1992; Wang and Georgakakos 2007, manuscript submitted to J. Geophys Res.) or the Improvement of Microphysical Parameterization through Observational Verification Experiment (IMPROVE;
Colle et al. 2005)] because these are the only times
validation data are available. A multiyear investigation
comparing model melting levels with BBHs could provide statistics of how well various schemes represent
210
JOURNAL OF HYDROMETEOROLOGY
atmospheric melt processes under different meteorological conditions.
Most operational and many scientific hydrology
models rely on temperature to identify whether precipitation falls as rain or snow, such as Snow-17 (Anderson
1976), Variable Infiltration Capacity (VIC; Cherkauer
et al. 2003), and Distributed Hydrology Soil Vegetation
Model (DHSVM; Wigmosta et al. 1994, 2002). The
probability distribution functions presented here of
likelihood of snow gain or loss as a function of temperature may help to improve such model parameters.
As temperatures continue to warm across the western United States, basins at progressively higher elevations will experience mixed rain and snow during winter
precipitation events, impacting both storm runoff and
snowpack storage for the following summer. Changes
in the frequency of high melting levels represent a
changing baseline for hydrologic runoff and will make
statistical forecasts based on historic datasets less reliable. The spatial patterns of precipitation type and intensity will influence landslides and ecosystem processes. These applications provide impetus to improve
our physical knowledge of processes in this regime, and
lessons learned from the relatively lower-elevation
American River basin may become crucial for understanding and modeling streamflow in mountain basins
across the west.
Acknowledgments. We thank the dedicated engineering staff in the Physical Sciences Division of NOAA/
ESRL for deploying and maintaining the NOAA wind
profilers used in this study. Funding for this study was
provided by a CIRES Postdoctoral Fellowship through
NOAA and the University of Colorado, Boulder.
Thank you to Socorro Medina for discussions of the
paper’s content and to two anonymous reviewers for
comments.
REFERENCES
Anderson, E. A., 1976: A point energy and mass balance model of
a snow cover. NOAA Tech. Rep. NWS 19, 150 pp.
Barros, A. P., and D. P. Lettenmaier, 1994: Dynamic modeling of
orographically induced precipitation. Rev. Geophys., 32, 265–
284.
Battan, L. J., 1973: Radar Observations of the Atmosphere. University of Chicago Press, 279 pp.
Beaumont, R. T., 1965: Mt. Hood pressure pillow snow gage. J.
Appl. Meteor., 4, 626–631.
Brunengo, M. J., 1990: A method of modeling the frequency characteristics of daily snow amount for stochastic simulation of
rain-on-snow events. Proc. 88th Western Snow Conf., Sacramento, CA, Western Snow Conference, 110–121.
Cherkauer, K. A., L. C. Bowling, and D. P. Lettenmaier, 2003:
Variable infiltration capacity cold land process model updates. Global Planet. Change, 38, 151–159.
VOLUME 9
Colle, B. A., M. F. Garvert, J. B. Wolfe, C. F. Mass, and C. P.
Woods, 2005: The 13–14 December 2001 IMPROVE-2 event.
Part III: Simulated microphysical budgets and sensitivity
studies. J. Atmos. Sci., 62, 3535–3558.
Cunningham, R. M., 1947: A different explanation of the “bright
line.” J. Meteor., 4, 163.
Dettinger, M. D., D. R. Cayan, M. K. Meyer, and A. E. Jeton,
2004: Simulated hydrologic responses to climate variations
and change in the Merced, Carson, and American river basins, Sierra Nevada, California, 1900–2099. Climatic Change,
62, 283–317.
Fabry, F., and I. Zawadzki, 1995: Long-term radar observations of
the melting layer of precipitation and their interpretation. J.
Atmos. Sci., 52, 838–851.
Farnes, P. E., 1967: Criteria for determining mountain snow pillow sites. Proc. 35th Western Snow Conf., Boise, ID, Western
Snow Conference, 59–62.
Feiccabrino, J., and A. Lundberg, 2008: Precipitation phase discrimination by dew point and air temperature. Proc. 75th
Western Snow Conf., Kailua-Kona, HI, Western Snow Conference, 141–146.
Ffolliott, P. F., and K. N. Brooks, 1983: Rain-on-snow flood event
in central Arizona. Proc. 51st Western Snow Conf., Vancouver, WA, Western Snow Conference, 130–138.
Hall, R. L., and J. F. Hannaford, 1983: Analysis of three rain-onsnow floods in the Sierra Nevada, California. Proc. 51st Western Snow Conf., Vancouver, WA, Western Snow Conference,
46–54.
Harr, R. D., 1986: Effects of clearcutting on rain-on-snow runoff
in western Oregon: A new look at old studies. Water Resour.
Res., 22, 1095–1100.
Jensen, M. E., R. D. Burman, and R. G. Allen, Eds., 1990: Evapotranspiration and Irrigation Water Requirements. American
Society of Civil Engineers, 332 pp.
Jeton, A. E., M. D. Dettinger, and J. L. Smith, 1996: Potential
effects of climate change on streamflow, eastern and western
slopes of the Sierra Nevada, California and Nevada. USGS
Water Resources Investigations Rep. 95-4260, 44 pp.
Kattelmann, R., 1997: Flooding from rain-on-snow events in the
Sierra Nevada. Destructive Water: Water-Caused Natural Disasters, Their Abatement and Control, G. H. Leavesley et al.,
Eds., IAHS Publication 239, 59–65.
Knowles, N., M. D. Dettinger, and D. R. Cayan, 2006: Trends in
snowfall versus rainfall in the western United States. J. Climate, 19, 4545–4559.
Lettenmaier, D. P., and T. Y. Gan, 1990: Hydrologic sensitivities
of the Sacramento-San Joaquin River basin, California, to
global warming. Water Resour. Res., 26, 69–86.
L’Hôte, Y., P. Chevallier, A. Coudrain, Y. Lejeune, and P.
Etchevers, 2005: Relationship between precipitation phase
and air temperature: Comparison between the Bolivian
Andes and the Swiss Alps. Hydrol. Sci. J., 50, 989–997.
Lin, Y.-L., R. D. Farley, and H. D. Orville, 1983: Bulk parameterization of the snow field in a cloud model. J. Climate Appl.
Meteor., 22, 1065–1092.
Marks, D., J. Kimball, D. Tingey, and T. Link, 1998: The sensitivity of snowmelt processes to climate conditions and forest
cover during rain-on-snow: A case study of the 1996 Pacific
Northwest flood. Hydrol. Processes, 12, 1569–1587.
Marshall, J. S., R. C. Langille, and W. McK. Palmer, 1947: Measurements of rainfall by radar. J. Meteor., 4, 186–192.
Marwitz, J. D., 1983: The kinematics of orographic airflow during
Sierra storms. J. Atmos. Sci., 40, 1218–1227.
APRIL 2008
LUNDQUIST ET AL.
——, 1987: Deep orographic storms over the Sierra Nevada. Part
I: Thermodynamic and kinematic structure. J. Atmos. Sci., 44,
159–173.
McCabe, G. J., M. P. Clark, and L. E. Hay, 2007: Rain-on-snow
events in the western United States. Bull. Amer. Meteor. Soc.,
88, 319–328.
Medina, S., B. F. Smull, R. A. Houze Jr., and M. Steiner, 2005:
Cross-barrier flow during orographic precipitation events:
Results from MAP and IMPROVE. J. Atmos. Sci., 62, 3580–
3598.
Meyers, M. P., and W. R. Cotton, 1992: Evaluation of the potential for wintertime quantitative precipitation forecasting over
mountainous terrain with an explicit cloud model. Part I:
Two-dimensional sensitivity experiments. J. Appl. Meteor.,
31, 26–50.
Mitra, S. K., O. Vohl, M. Ahr, and H. R. Pruppacher, 1990: A
wind tunnel and theoretical study of the melting behavior of
atmospheric ice particles. IV: Experiment and theory for
snow flakes. J. Atmos. Sci., 47, 584–591.
Mittermaier, M. P., and A. J. Illingworth, 2003: Comparison of
model-derived and radar-observed freezing-level heights: Implications for vertical reflectivity profile-correction schemes.
Quart. J. Roy. Meteor. Soc., 129, 83–95.
Osterhuber, R., 1999: Precipitation intensity during rain-on-snow.
Proc. 67th Western Snow Conf., South Lake Tahoe, CA,
Western Snow Conference, 153–155.
Ralph, F. M., and Coauthors, 2005: Improving short-term (0–48 h)
cool-season quantitative precipitation forecasting: Recommendations from a USWRP workshop. Bull. Amer. Meteor.
Soc., 86, 1619–1632.
Rohrer, M., 1989: Determination of the transition air temperature
from snow to rain and intensity of precipitation. Instruments
and Observing Methods, WMO Tech. Rep. 328, Rep. 48,
World Meteorological Organization, 475–482.
Schultz, P., 1995: An explicit cloud physics parameterization for
operational numerical weather prediction. Mon. Wea. Rev.,
123, 3331–3343.
Slack, J. R., and J. M. Landwehr, 1992: Hydro-climatic data network (HCDN): A U.S. Geological Survey streamflow data
set for the United States for the study of climate variations,
1874–1988. USGS Open-File Rep. 92-129, 193 pp.
Steiner, M., O. Bousquet, R. A. Houze Jr., B. F. Smull, and M.
Mancini, 2003: Air flow within major Alpine river valleys
under heavy rainfall. Quart. J. Roy. Meteor. Soc., 129, 411–
431.
Stewart, R. E., J. D. Marwitz, J. C. Pace, and R. E. Carbone, 1984:
Characteristics through the melting layer of stratiform
clouds. J. Atmos. Sci., 41, 3227–3237.
Storck, P., 2000: Trees, snow and flooding: An investigation of
forest canopy effects on snow accumulation and melt at the
211
plot and watershed scales in the Pacific Northwest. Water
Resources Series Tech. Rep. 61, Dept. of Civil Engineering,
University of Washington, 176 pp.
Szyrmer, W., and I. Zawadski, 1999: Modeling of the melting
layer. Part I: Dynamics and microphysics. J. Atmos. Sci., 56,
3573–3592.
Tarboton, D. G., and C. H. Luce, 1996: Utah Energy Balance
snow accumulation and melt model (UEB): Computer model
technical description and users guide. Utah Water Research
Laboratory and USDA Forest Service Intermountain Research Station, 41 pp. [Available online at http://www.
engineering.usu.edu/dtarb/.]
Thom, A. S., 1975: Momentum, mass and heat exchange of plant
communities. Vegetation and the Atmosphere, J. L. Monteith,
Ed., Vol. 1, Academic Press, 57–109.
Trabant, D. C., and G. P. Clagett, 1990: Measurement and evaluation of snowpacks. Cold Regions Hydrology and Hydraulics,
W. L. Ryan and R. D. Crissman, Eds., American Society of
Civil Engineers, 39–93.
U.S. Army Corps of Engineers, 1956: Snow hydrology: Summary
report of the snow investigations. North Pacific Division, U.S.
Army Corps of Engineers, 437 pp.
Wang, J., and K. P. Georgakakos, 2005: Validation and sensitivities of dynamic precipitation simulation for winter events
over the Folsom Lake watershed: 1964–99. Mon. Wea. Rev.,
133, 3–19.
Westrick, K. J., and C. F. Mass, 2001: An evaluation of a highresolution hydrometeorological modeling system for prediction of a cool-season flood event in a coastal mountainous
watershed. J. Hydrometeor., 2, 161–180.
White, A. B., D. J. Gottas, E. T. Strem, F. M. Ralph, and P. J.
Neiman, 2002: An automated brightband height detection
algorithm for use with Doppler radar spectral moments. J.
Atmos. Oceanic Technol., 19, 687–697.
Wigmosta, M. S., L. W. Vail, and D. P. Lettenmaier, 1994: A distributed hydrology-vegetation model for complex terrain.
Water Resour. Res., 30, 1665–1679.
——, B. Nijssen, P. Storck, and D. P. Lettenmaier, 2002: The distributed hydrology soil vegetation model. Mathematical Models of Small Watershed Hydrology and Applications, V. P.
Singh and D. K. Frevert, Eds., Water Resource Publications,
7–42.
Yuter, S. E., D. E. Kingsmill, L. B. Nance, and M. Löffler-Mang,
2006: Observations of precipitation size and fall speed characteristics within coexisting rain and wet snow. J. Appl. Meteor. Climatol., 45, 1450–1464.
Zuzel, J. F., R. N. Greenwalt, and R. R. Allmaras, 1983: Rain on
snow: Shallow, transient snowpacks with frozen soils. Proc.
51st Western Snow Conf., Vancouver, WA, Western Snow
Conference, 67–75.