Rain versus Snow in the Sierra Nevada, California ... · mosphere must be related to the melting...

Rain versus Snow in the Sierra Nevada, California: Comparing Doppler ProfilingRadar 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/EarthSystem 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 areextremely sensitive to flooding from warm winter storms, primarily because rain falls at higher elevationsand over a much greater fraction of a basin’s contributing area than during a typical storm. Accuratepredictions of this rain–snow line are crucial to hydrologic forecasting. This study examines how remotelysensed atmospheric snow levels measured upstream of a mountain range (specifically, the bright bandmeasured above radar wind profilers) can be used to accurately portray the altitude of the surface transitionfrom 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 cyclesin solar radiation, and fluctuations in the free-tropospheric melting level. At 1.5°C, 50% of precipitationevents fall as rain and 50% as snow, and on average, 50% of measured precipitation contributes to increasesin snow water equivalent (SWE). Between 2.5° and 3°C, snow is equally likely to melt or accumulate, withmost cases resulting in no change to SWE. Qualitatively, brightband heights (BBHs) detected by 915-MHzprofiling radars up to 300 km away from the American River study basin agree well with surface meltingpatterns. Quantitatively, this agreement can be improved by adjusting the melting elevation based on thespatial location of the profiler relative to the basin: BBHs decrease with increasing latitude and decreasingdistance to the windward slope of the Sierra Nevada. Because of diurnal heating and cooling by radiationat the mountain surface, BBHs should also be adjusted to higher surface elevations near midday and lowerelevations near midnight.

1. Introduction

In the maritime mountain ranges of western NorthAmerica, most precipitation falls during the wintermonths, with a large percentage falling in the form ofsnow. Reservoir managers must balance the objectiveof storing water for summer use with the need to re-

lease water to provide flood protection from warm win-ter storms. Coastal basins spanning a wide range ofelevations, such as the North Fork of the AmericanRiver basin in California (Fig. 1), are extremely sensi-tive to rain falling on snow at mid- to high elevations(Kattelmann 1997; Osterhuber 1999). Many studies ofclimatic change show that these same river basins areextremely sensitive to regional warming. Over the past50 yr, an increasingly greater percentage of precipita-tion has fallen in the form of rain rather than snow, andthis trend is likely to continue (Dettinger et al. 2004;Jeton et al. 1996; Knowles et al. 2006; Lettenmaier and

Corresponding author address: Jessica D. Lundquist, Civil andEnvironmental Engineering, University of Washington, Seattle,WA 98195-2700.E-mail: [email protected]

194 J O U R N A L O F H Y D R O M E T E O R O L O G Y VOLUME 9

DOI: 10.1175/2007JHM853.1

© 2008 American Meteorological Society

JHM853

Gan 1990). However, while the importance of floodingand erosion associated with warm rain events has beenrepeatedly stressed (Brunengo 1990; Ffolliott andBrooks 1983; Hall and Hannaford 1983; Harr 1986;Kattelmann 1997; Marks et al. 1998; McCabe et al.2007; Zuzel et al. 1983), hydrologic forecasting centersstill struggle to predict the elevation at which snowturns to rain in mountain watersheds (E. Strem, Cali-fornia–Nevada River Forecast Center, 2006, personalcommunication).

In the West Coast mountains, particularly warmstorms result in floods primarily because rain falls athigher elevations and over a much larger contributingarea for runoff than during a typical storm. Figure 2

illustrates the fraction of the North Fork AmericanRiver basin contributing to runoff based on the eleva-tion below which precipitation falls as rain. As the rain–snow line rises from 800 to 2800 m (a common range forthe melting level in California winter storms), the con-tributing fraction increases from 25% to 100% of thebasin, resulting in up to 4 times as much runoff for agiven rain rate. White et al. (2002) tested the impor-tance of the melting level by comparing river forecastsimulations of peak flow as a function of melting levelin four California watersheds. In three of the four wa-tersheds, runoff tripled if the snow level rose 600 m(White et al. 2002). These results, combined with thesteep slope of the curve in Fig. 2, indicate that signifi-cant uncertainty in streamflow prediction can occur as aresult of relatively small errors in characterizing theprecise altitude at which snow accumulates in a givenstorm. For these reasons it is important to monitor,understand, and predict the details of the snow level towithin relatively small margins of error. This paper usesunique measurements, collected during the NationalOceanic and Atmospheric Administration’s (NOAA)Hydrometeorological Test Bed field campaign (Ralphet al. 2005), to improve understanding of the meltinglevel and snow accumulation transition zone. Theanalysis includes a novel radar technique to measureconditions aloft in combination with more traditional insitu observations at the earth’s surface.

Because of the difference in radar reflectivity and fallspeeds between frozen and liquid hydrometeors, Dopp-ler profiling radars are able to detect the altitude in theatmosphere where rain changes to snow (Battan 1973;

FIG. 1. Locations of radar profilers (black circles), surface mea-surement sites (plus signs), and surface snow and temperaturesites (white circled plus signs). KOAK identifies the Oaklandsounding location.

FIG. 2. Cumulative fraction of total basin area contributing torunoff as a function of altitude of rain–snow line for the NorthFork of the American River, CA.

APRIL 2008 L U N D Q U I S T E T A L . 195

Cunningham 1947; Fabry and Zawadzki 1995; Marshallet al. 1947; Mittermaier and Illingworth 2003; Stewartet al. 1984; White et al. 2002). The dielectric constantincreases as melting ice particles become coated withwater, 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 Cali-fornia coast) at temperatures usually between 1° and2°C (Stewart et al. 1984; for measurements near theSierra Nevada). Doppler wind-profiling radars alongthe coast and central valley of California have beenobserving brightband heights since 1996, and an auto-mated algorithm developed by White et al. (2002)makes this information available in real time to riverforecasters.

For hydrologic applications, the BBH in the free at-mosphere must be related to the melting level on thesurface some distance away. First, snow falling in thefree atmosphere starts melting at 0°C and changes torain over time and space as a complex function of tem-perature, humidity, winds, and the type and fall rate ofhydrometeors. Most winter storms impacting themountains of the west coast of North America possessa fairly stable, stratiform structure (Medina et al. 2005),with ice crystals forming aloft and then falling throughwarmer atmospheric layers below, where they melt andbecome rain. The vertical distance through which thesnow crystals pass before they are completely melteddepends on the size and density of the crystals, the fallrate 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), theventilation of the hydrometeors, and the balance ofheat at the particle surface (due to conduction of heatfrom the warmer surrounding air and to latent heatreleased onto the melting particles during water vaporcondensation; Szyrmer and Zawadski 1999). Aircraftobservations indicate that, in general, the ratio of liquidto total hydrometeor mass increases linearly as tem-peratures 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 correspondingmelting levels change as an air mass intersects a moun-tain range, such as the Sierra Nevada, and is influencedby boundary layer effects. Specifically, orographic lift-ing causes adiabatic cooling, and greater precipitationrates increase diabatic cooling, frequently resulting in alowering of the BBH as it intersects the windwardslopes of a mountain range (Marwitz 1983, 1987; Me-dina et al. 2005). Diabatic cooling can also result in

down-valley airflow in mountain valleys, often in theopposite direction of the general storm movement(Steiner et al. 2003).

Detailed microphysical models are able to representall of the processes related to the melting level heightand thickness in one dimension but become computa-tionally unwieldy in two or three dimensions, whereslight variations in precipitation microphysics can set upmesoscale circulation patterns, which in turn feed backand change hydrometeor fall rates and melting rates(Szyrmer and Zawadski 1999). Thus, most operationalweather forecast models parameterize ice microphysics(Barros and Lettenmaier 1994; Lin et al. 1983; Schultz1995), and different parameterization schemes often re-sult in quite different precipitation distributions over abasin (Wang and Georgakakos 2005; Wang and Geor-gakakos 2007, manuscript submitted to J. Geophys.Res.). Most hydrologic models discount atmosphericmicrophysics altogether, instead relying on surface tem-perature cutoffs to delineate rain from snow (Anderson1976; Storck 2000; Westrick and Mass 2001).

At the mountain surface, these near-surface air tem-peratures provide a good estimate of precipitation type,such that at temperatures below 0°C, over 90% of pre-cipitation events occur as snow; at temperatures above3°C, over 90% of precipitation events occur as rain; andrain and snow are equally likely at 1.5°C (U.S. ArmyCorps of Engineers 1956). These empirically deriveddistributions of rainfall versus snowfall as a functionof surface temperature are remarkably consistentacross 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). De-tailed observations in a mixed precipitation event in theOregon Cascades also fell within this range, whereYuter et al. (2006) used a Parsivel disdrometer to ob-serve that rain particles made up 5% of the total con-centration 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 coverin a river basin affect the final form precipitation takeson the surface—rain may freeze within a snowpack andbe stored for some time, or snow may melt on bareground and contribute to runoff. This last scenario—what happens on the surface—is most important inflood forecasting.

While the processes governing melting snowflakes inthe atmospheric melting layer and the relationship be-tween surface temperature and snow accumulation arerelatively well known, the precise relationship betweenmeasurements of the melting level aloft and snow ac-cumulation behavior at the surface has not been welldocumented, especially in complex terrain. This paper


determines if BBHs measured atmospherically up-stream of a mountain range can be used to infer meltinglevels within a hydrologic basin. Specifically, we trackthe altitude of the melting level from the free atmo-sphere, to the mountain boundary layer, to the ground(or snow covered) surface for winters 2001–2005, asillustrated in Fig. 3. Significant heights and tempera-tures are detailed in Table 1. Section 2 describes theobservations and the methods used for classifying pre-cipitation as rain or snow. Section 3 presents two case

studies of mixed rain and snow events in the AmericanRiver basin to illustrate the rapid changes in meltinglevel and the complex surface responses associated withwinter storms. Section 4 establishes temperature as anindex for where surface snowmelt is taking place andcompares near-surface and free-atmosphere measure-ments during storm events. Section 4 also presents 5-yrstatistics of how these temperatures vary with locationand time of day, and how they relate to different melt-ing 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 theOakland 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 BBis the FAML.

BBH � brightband height. Height of maximum reflectivity in the radar melting layer bright band. Below the FAML but abovethe bottom of the BB.

Zs-half � half snow-runoff melting level at surface. Height along the Sierra mountainside at which half of the precipitation fallingin 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 altitudesabove 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.

FIG. 3. Schematic diagram of subjects discussed in this paper.


North Fork of the American River. Section 5 discussesapplications for streamflow forecasting and directionsfor further research.

2. Observations and methods

Free atmosphere melting levels (FAML; 0°C iso-therm) were determined by temperature measurementsfrom radiosondes launched at Oakland, California, andcompared with BBHs at four 915-MHz radar profilers.Boundary layer melting levels were determined from atransect of surface temperatures in the American Riverbasin, and the fate of rain versus snow at the surfacewas determined from precipitation, temperature, snowwater equivalent (SWE), and snow depth measure-ments at California Department of Water Resources(CA DWR) snow pillow stations in the same basin. TheAmerican River basin is located downwind of the pro-filers along the primary winter storm track, which isfrom the west (Fig. 1).

a. Radar profilers

The 915-MHz Doppler wind profilers provide windand precipitation profile measurements in the bound-ary layer and lower free troposphere, and data fromfour profilers in California are analyzed here (Table 2;Fig. 1). Signal-to-noise ratios (SNRs) indicate regionsof high reflectivities, and Doppler vertical velocities(DVVs) approximate the fall speeds of hydrometeors.BBHs were determined using the automated algorithmdetailed in White et al. (2002), which picks out the al-titude where DVV starts decreasing with height, indi-cating the height of slower-falling snow particles, in tan-dem with SNR increasing with height, indicating thegreater reflectance of large, water-coated snow and iceparticles (Fig. 3). This level is determined with hourlytemporal resolution and 100-m vertical resolution.

b. Oakland radiosonde

Upper-air measurements of temperature, dewpointtemperature, wind speed, and direction were obtainedfrom the National Weather Service (NWS) RadiosondeNetwork site at Oakland, California, where they aremeasured at 0000 and 1200 UTC [0400 and 1600 Pacific

standard time (PST)] each day. This site is generallyupwind of the American River basin and representsfree-air conditions before they encounter the topogra-phy of the Sierra. Wet-bulb temperatures and virtualtemperatures were calculated from the measured vari-ables (Jensen et al. 1990). The altitude of the 0°C iso-therm (FAML, Table 1) and the temperature at theBBH were both determined using linear interpolationfrom available measurements. While this approach mayresult in errors during the hours surrounding a frontalpassage, it is the best available approximation given theexisting data.

c. Surface meteorology, snow, and hydrology

The CA DWR manages a network of 41 automatedmeteorological monitoring stations in or near theAmerican River basin (Fig. 1; Table 3). These stationsall measure precipitation and 23 of them also measureair temperature, typically from a sensor mounted on apole or mast, approximately 10 m above ground level.Thirteen of the stations have snow pillows that measurethe weight of snow accumulation and thereby indicatethe SWE of the snow column. Data were obtained fromCA DWR (http://cdec.water.ca.gov).

Most snow pillows are located in flat meadows, sur-rounded by forested areas that shelter the pillow fromwind scouring (Farnes 1967). Compared to nearby for-ested areas, these sites tend to accumulate more snowand are also more exposed to sunlight during the meltseason. Because snowmelt during a precipitation eventis caused primarily by turbulent energy transfer, whichgrows with increasing wind speed (Marks et al. 1998),snow pillows may record more melt than a completelyforested area and significantly less melt than an unshel-tered area. While snow pillows do not record the fullrange of snow properties due to varying slope, aspect,and vegetative cover, they provide the most compre-hensive set of high-elevation measurements available.

The CA DWR snow pillows report information athourly intervals, but because pillows can experienceseveral hours of delay in responding to changes in SWE(Beaumont 1965; Trabant and Clagett 1990), 12-h av-erages were used here. An increase in SWE may be dueto snow falling on the pillow or to liquid water fallingon snow already on the pillow and freezing into thesnowpack, thus increasing its density. Decreases inSWE are due to snowmelt or sublimation. Redistribu-tion of SWE from or to areas outside of the snow pillowarea may also be responsible for changes. However, thiseffect is slight at most California snow pillows becauseof the high density of the snowpack and their flat loca-tions with wind shelter from the surrounding trees.

Snow depth measurements, which are only available

TABLE 2. Radar profilers.

Name ID Elev (m) Lat (°N) Lon (°W)

Bodega Bay BBY 12 38.32 123.07Chowchilla CHL 76 37.11 120.24Chico CCO 41 39.69 121.91Grass Valley GVY 689 39.17 121.11


at a small subset of the stations (Blue Canyon, MeadowLake, and Caple’s Lake; Fig. 1b), acoustically detectchanges in snow depth at an hourly time scale. An in-crease in snow depth can only occur with snowfall orwith deposits of drifting snow. However, the latter israre during a rain event in the Sierra Nevada. A de-crease in snow depth could be due to melt or compac-tion. These changes are accurate on an hourly timescale but also include noise, particularly during a snowstorm, when the acoustic signal can reflect off of fallingsnowflakes. To minimize the noise, we smoothed the

depth record with a 20-h Blackman-window filter andthen established a cutoff where we declared snowfall tooccur in any hour when the depth increased by at least0.35 cm.

Most precipitation sensors in the region consist of areservoir with antifreeze. These report accumulatedprecipitation and are also subject to instrumental noise.Incremental precipitation was determined by smooth-ing the record with a 20-h Blackman-window filter andtaking the difference between hourly reports of thesmoothed accumulated precipitation. A cutoff of

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 8 38.555 121.416 1 1 0 0Arden Way 11 38.596 121.413 1 0 0 0Hurley 11 38.587 121.407 1 0 0 0Rio Linda W.C. 14 38.7 121.448 1 0 0 0Arcade Creek 21 38.645 121.347 1 0 0 0Cresta Park 21 38.593 121.368 1 0 0 0Rancho Cordova 22 38.604 121.311 1 0 0 0Royer Park/Dry Creek 43 38.749 121.28 1 0 0 0Roseville Fire Station 46 38.76 121.315 1 0 0 0Lincoln 61 38.882 121.272 1 1 0 0Chicago 64 38.652 121.254 1 0 0 0Orangevale W.C. 71 38.686 121.219 1 0 0 0Prairie City 95 38.592 121.161 1 0 0 0Folsom Dam 107 38.7 121.167 1 1 0 0Roseville Water Tr. 116 38.724 121.226 1 1 0 0Folsom Water Tr. 129 38.687 121.149 1 1 0 0Caperton Reservoir 189 38.863 121.218 1 0 0 0Loomis Observatory 223 38.815 121.113 1 0 0 0New Castle 271 38.874 121.135 1 0 0 0Auburn Dam Ridge 366 38.882 121.045 1 1 0 0Pilot Hill (CDF) 366 38.832 121.012 1 0 0 0Georgetown (USBR) 991 38.925 120.789 1 1 0 0Pacific House 1048 38.76 120.5 1 1 0 0Sugar Pine 1171 39.128 120.75 1 1 0 0Owens Camp 1372 38.733 120.245 1 1 0 0Hell Hole (USFS) 1396 39.078 120.437 1 1 0 0Robb’s Powerhouse* 1570 38.903 120.375 1 1 1 0Blue Canyon* 1609 39.276 120.708 1 1 1 1Sugarloaf near Kyburz 1696 38.7839 120.31 1 1 0 0Greek Store* 1707 39.075 120.558 1 1 1 0Robb’s Saddle* 1798 38.912 120.378 1 1 1 0Huysink* 2012 39.282 120.527 1 1 1 0Van Vleck* 2042 38.945 120.305 1 1 1 0Duncan 2164 39.144 120.509 0 1 0 0Silver Lake 2164 38.678 120.118 1 1 1 0Alpha (SMUD)* 2316 38.805 120.215 1 1 1 0Forni Ridge* 2316 38.805 120.213 1 1 1 1Beta 2316 38.8 120.2 1 0 1 0Caple’s Lake (DWR)* 2438 38.71 120.042 1 1 1 1Carson Pass 2546 38.6924 120.0021 1 1 1 1Schneider’s* 2667 38.747 120.068 1 1 1 0Meadow Lake 2194 39.417 120.508 1 1 1 0

* Stations used in snowmelt vs temperature contour plots.


0.04 cm h�1 was used to indicate that precipitation ac-tually occurred. Blue Canyon was equipped with aheated tipping bucket in September 2004 to measureincremental precipitation. Because this instrument ismore sensitive than the antifreeze reservoirs, a cutoff of0.02 cm h�1 was used to signify the occurrence of pre-cipitation at this station. This is equivalent to 1 tip ofthe bucket in the rain gauge (0.01 in. or 0.254 mm).

Mean daily discharge measurements were examinedfor the North Fork (NF) of the American River (Fig. 1).It does not have dams or diversions and is included inthe U.S. Geological Survey (USGS) hydroclimaticdataset (Slack and Landwehr 1992). The record beganin 1941, the gauge is at an elevation of 218 m, basin areais 886 km2, and mean basin elevation is 1190 m.

d. Temperature–SWE PDF

The fate of precipitation after it reached the groundwas determined by creating a probability density func-tion (PDF) of change in SWE given a precipitationevent at a given temperature. Data at all snow pillowswere quality controlled, and 10 stations (indicated inTable 3) were selected for having reliable measure-ments of both SWE and collocated temperature.Hourly SWE, surface air temperature, and precipita-tion observations were averaged every 12 h. Next, thechange in SWE and the incremental precipitation overeach 12-h interval was calculated. Because of the lack ofquality precipitation data in the snow zone, precipita-tion data at all 40 stations (Table 3) were considered. If10 or more of these stations reported increased precipi-tation, then the precipitation rate was determined as

the average of all stations. If less than 10 stations re-ported precipitation, then the precipitation rate was setat zero, and that 12-h time step was not considered inthe subsequent analysis. Data were binned by tempera-tures (at 0.5°C increments) and normalized by the num-ber of measurements at each temperature to produceprobability density functions.

3. Case studies—Radar brightband levels andsurface rain, temperature, and snow

Two case studies from winter 2005 were chosen toillustrate how radar measurements of BBH correspondwith measurements of temperature, rainfall, and snowgain or loss at various surface elevations. BBHs variedfrom 600 to 3000 m during the winter 2005 period, andincreased NF American River streamflow generallycorresponded to warm storms, indicated by high-alti-tude BBHs (Fig. 4). Warm storms not only resulted inliquid precipitation falling over large fractions of thebasin, but they also carried the most moisture and,hence, the highest precipitation rates. The two casestudy periods (marked, Fig. 4) both encompassedstorms with a wide range of melting levels, but the twohad very different streamflow responses.

a. January 2005

The 25 to 29 January case provided a simple exampleof a cold front passing with dropping melting levels inearly winter. Rain falling at Blue Canyon (1600 m) wasincorporated into the snowpack, increasing SWE, withminimal contributions to streamflow. Radar returns at

FIG. 4. (a) Radar brightband heights and (b) NF American River discharge (normalized bybasin area on the left axis and actual quantity on the right axis) for winter 2005 (PST). Casestudy periods are marked by vertical dashed lines.


the Grass Valley wind profiler showed decreasingBBHs during this period (Fig. 5a). Surface tempera-tures at Blue Canyon (1600 m) fell from 10° to �1°C,and temperatures at Caple’s Lake (2400 m) fell from10° to �11°C (Fig. 6a). On 25 January, precipitationincreased (Fig. 6b), and snow depth decreased slightlyat both Caple’s Lake and Blue Canyon (Fig. 6c), indi-cating that rain compacted the snowpack. On 26 Janu-ary, snow depth mostly increased at Caple’s Lake (witha slight decrease near noon) and decreased at BlueCanyon, indicating that while mostly snow fell at Ca-ple’s Lake, rain fell at Blue Canyon (Fig. 6c). SWEincreased at both locations (Fig. 6d), indicating that atleast some of the rain at Blue Canyon was incorporatedinto the snowpack, increasing its density and water con-tent, and not directly contributing to streamflow (Fig.4). By the 28 January precipitation event, temperatureswere at or below freezing at both locations, and snowfell at both locations, as indicated by increases in snowdepths and SWE.

Changes in temperature (Fig. 7a) and precipitationtype (Fig. 7b) with elevation corresponded with drop-ping radar BBHs at both Bodega Bay (the profiler onthe coast, west of the American River basin) and atGrass Valley (the profiler in the Sierra foothills, justslightly north of the American River basin). On 25January the entire basin from 1500- to 2500-m elevationwas warmer than 2°C, and rain fell at both Caple’s Lake

and Blue Canyon. Bodega Bay melting levels on 25January were at about 2000 m, but this cooler stormsector did not arrive at the American River until mid-night that night, when Grass Valley reported meltinglevels at about 2000 m, and temperatures above thataltitude dropped below 0°C. At this time, snow fell atCaple’s Lake and Meadow Lake, while rain continuedat Blue Canyon. On 26 January daytime rainfall andtemperatures 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 being107 km southeast of Grass Valley (Fig. 1), thus remain-ing in a warmer sector of the storm. When precipitationresumed near midnight on 27 January, temperatureswere much colder, and the BBH at Grass Valley wasbelow 1500 m. Snow fell at all monitored locations.Some profiler measurements at Bodega Bay and GrassValley indicated BBHs slightly above the altitude ofBlue Canyon. While surface measurements indicatedthat Blue Canyon had a net gain of snow, a mixture ofrain and snow may have been falling from the sky. TheBBH may also have dropped in the distance betweenGrass Valley and Blue Canyon (about 44 km).

b. March 2005

The 19–22 March case illustrated a more complicatedsequence of storms, alternating between warm andcold conditions (Fig. 5b). During the second warm

FIG. 5. Wind profiles and BBHs for the Grass Valley profiler (PST) for (top) case study 1 and (bottom) the last 4 days of casestudy 2.


period, snow melted while rain fell at high rates at BlueCanyon, and American River discharge rose steeply. Atthe start of the storm on 19 March, snow depth de-creased slightly at Blue Canyon but then began increas-ing at midday as temperatures dropped and rainchanged to snow (Fig. 8). Snowfall continued throughmidday on 21 March, and then temperatures rose dur-ing a break in the storm. The second storm began sev-eral hours later and was considerably warmer, with adecrease in snow depth and SWE (indicating melting

snow in addition to falling rain) at Blue Canyon (Fig.8). At midday on 22 March, Caple’s Lake reported anincrease in temperature, a sharp increase in precipita-tion, and slight drops in snow depth and SWE. Afterthis time, temperatures dropped at both stations, rainswitched to snow, and snow depths and SWE increased.

These patterns of alternating warm and cold stormsectors were also observed by the profilers (Fig. 9).Temperatures at all altitudes above 1500 m droppedbelow 0°C on 19 and 20 March, following the drop ofthe BBH at Grass Valley. BBH changes at Bodega Baypreceded those at Grass Valley by about 4 h. The switchfrom rain to snow at Blue Canyon on 22 March coin-cided with drops in BBHs at both Bodega Bay andGrass Valley. However, data at Caple’s Lake indicatedan earlier switch to snow, and Meadow Lake reportedincreases in snow depth, that is, continuous snow ratherthan rain, throughout the warm sector of the storm.During the warm sector (early 22 March), rain fell at arate greater than 4 mm h�1 over more than 90% of thebasin. This, combined with already-saturated snow-packs and soils, resulted in a large streamflow responseto this storm (Fig. 4).

c. Case study summary

The case studies above illustrate that qualitatively,radar BBHs represent the patterns and trends of melt-ing levels on the mountain surface at an hourly timescale. However, the case studies also illustrate the dif-ficulty of quantitatively determining the form and fateof precipitation hitting the ground, particularly in re-gions without human observers. From the case studies,several important points emerge. First, while cloud mi-crophysics are difficult to model and represent spatiallyacross a basin, preexisting surface snow characteristicsare often more important than hydrometeor distribu-tions in determining where precipitation will contributeto runoff. For example, rain at Blue Canyon was incor-porated into the snowpack during case 1 but contribut-ed to snowmelt and runoff during case 2, likely becausecase 1 occurred earlier in the year when the snowpackwas colder and not yet saturated. Thus, improvementsin model representations of snowpack temperature andliquid water content would help improve hydrologicforecasts.

Second, distributions of rain versus snow vary spa-tially and are not solely functions of time and elevation.For example, Caple’s Lake to the southeast appearedto be in a warmer sector of the storm during case study1, receiving rain when the altitude of the BBH mea-sured at the Bodega Bay and Grass Valley profilersindicated snow should fall at that elevation. Duringcase 1, the Bodega Bay and Grass Valley profilers

FIG. 6. Surface measurements at Blue Canyon (1600 m) andCaple’s Lake (2400 m) between 24 and 30 Jan 2005 (PST) of (a)air temperature, (b) precipitation rates, (c) changes in snowdepth, and (d) changes in SWE. To aid in comparison of snowchanges, both snow depth and SWE are offset to equal 0 at thestart of the analysis period.


agreed on the BBH when they reported measurementsat the same time, but during case 2, Grass Valley con-sistently reported a lower-altitude BBH than BodegaBay. In both cases, the Grass Valley measurement cor-responded better with surface temperature measure-ments. This highlights the danger of comparing oneprofiler with one surface station, and the importance ofexamining the profiler measurements in relation notonly to each other but to as many surface measure-ments as possible to better understand statistical varia-tions across the region of interest.

4. Temperatures and melting statistics from 2001to 2005

Few measurements of precipitation type exist, andthese are noisy. However, temperature measurementsare reliable, continuous, and widely available. By in-dexing SWE gain or loss during a precipitation event totemperature and then comparing elevational transectsof surface temperatures to radar-detected BBHs, wecan determine statistically whether and by how muchthe BBH should be adjusted up or down to representthe surface melting level as a function of location andtime.

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 ofOakland). Free-air temperatures (T) above Oakland atthe Bodega Bay BBH averaged 1.8°C, and wet-bulbtemperatures (Tw) averaged 0.3°C for the five winters.The standard deviation for both T and Tw over the 5-yrperiod was approximately 1.5°C, probably representingchanging thicknesses of the melting region and differ-ent storm structures, orientations, and motions relativeto these two sites. These values did not vary with thehour of the day and are consistent with prior observa-tions.

Because the other stations were considerably furtherfrom Oakland than Bodega Bay, they sampled airmasses with different vertical temperature structures.Chico (182 km northeast of Bodega Bay) sampledcolder air masses, so that at times when both stationshad measurements during the 5-yr period, BBHs wereon average 174 m lower at Chico than at Bodega Bay.Chowchilla (283 km southeast of Bodega Bay) sampledwarmer air masses, so BBHs at Chowchilla were onaverage 49 m higher than at Bodega Bay. Grass Valleyis only slightly north of Bodega Bay, but 195 km inland,and it averaged melting levels 106 m lower. Grass Val-ley is on the windward slope of the Sierra, so this resultis consistent with observations from studies in the Si-erra (Marwitz 1983, 1987) and the Alps and Cascades(Medina et al. 2005) that show that BBHs dip to lowerelevations on the windward slopes of mountains.

FIG. 7. (a) Surface temperatures at seven snow pillow locations, plotted at each location’selevation between 24 and 30 Jan 2005 (PST). Times with no mark indicate the free-atmosphere transition zone of a mixture of rain and snow with 0° � T � 2°C. Black squaresare brightband melting levels detected at Bodega Bay (BBY). Gray circles are brightbandlevels detected at Grass Valley (GVY). (b) Increases in precipitation and snow depth at threesnow pillow locations (Caple’s Lake, Meadow Lake, and Blue Canyon). Precipitation atMeadow Lake is inferred from records at Alpha, the station nearest in elevation. Blacktriangles (snow) indicate increases in snow depth, and gray upside-down triangles (rain)indicate increases in precipitation. Brightband heights are the same as in (a).


Mean BBHs for each station (averaged over the en-tire 5-yr period, without regard to having simultaneousmeasurement of the melting level with a neighboringstation) 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 decreasewith increasing latitude and with proximity to the wind-ward slope of the Sierra Nevada. These spatial slopes inBBHs should be considered when applying meltinglevel information to a river basin north, south, or inlandof a given profiler. For example, a line fit to the average

heights at the three nonfoothill stations yielded a de-crease in average BBH of 41.4 m per increase in degreelatitude. The average BBH at Grass Valley is 73 mlower than would be expected based on its latitude andthis line. This is smaller than the approximate 400 mdecrease in BBH observed where the bright band in-tersected the mountain in prior studies (Marwitz 1983,1987; Medina et al. 2005), perhaps because Grass Val-ley is generally lower in elevation than the freezinglevel and is only detecting the beginning of the dip inBBH as storms intersect the mountain slopes.

b. Temperature at which snow melts on the surface

To compare BBHs with the snow response at thesurface, we created a probability density function ofchange in SWE during precipitation events as a func-tion of temperature. Because of the scarcity of humiditydata measured at the surface, these are determinedbased on actual, and not wet-bulb, temperatures. Basedon the similar structure and variance of actual and web-bulb temperatures measured in the free atmosphereduring these events (section 4a), we assume little infor-mation is lost by neglecting surface humidity informa-tion. Figure 10a shows the likelihood of SWE gain orloss over a 12-h period, given a local surface tempera-ture, at the 10 snow pillow stations indicated in Table 2for 2001–05, as a ratio of change in SWE to the averagebasin precipitation. Although there is much scatter inthe data, several features emerge. Below 0°C there isabout 90% chance that snow will accumulate, andabout 60% chance that snow will accumulate at a meanrate equal to or larger than the measured rate of aver-age basin precipitation. The likelihood of the latter isdue both to orographic precipitation enhancement (thesnow pillows are all in the upper elevations of the basinwhile most precipitation sensors are at lower eleva-tions) and to the precipitation sensors’ likelihood ofundercatch during mixed rain and snow events. Above3°C the surface snowpack is �60% likely to melt (Fig.10b). In between 0° and 3°C, precipitation may fall asrain, snow, or a mixture of the two, with probabilitiestransitioning from likely snow to likely rain as the tem-perature rises within this range [Fig. 10c, determined bythe U.S. Army Corps of Engineers (1956) from directobservations near the Sierra crest during winters 1946–51]. At 1.5°C, 50% of precipitation events fall as rainand 50% as snow, and on average, 50% of measuredprecipitation contributes to increases in SWE. Between2.5° and 3°C, snow is equally likely to melt or accumu-late, with most cases resulting in no change to SWE(Figs. 10a and 10b). This is slightly warmer than thefree-atmosphere temperature where snow converts to

FIG. 8. Surface data for Blue Canyon and Caple’s Lake, as inFig. 6, but for 18 to 24 Mar 2005 (PST). Note changed limits foreach axis from Fig. 6 (smaller temperature range and increasedprecipitation and depth ranges).


rain because this value measures the fate of the pre-cipitation phase at the surface. Cold rain and rain–snowmixtures falling on an existing snowpack are often fro-zen and incorporated into the snowpack, thus resultingin a measured increase in SWE. In other cases, rainpasses through the snowpack with no measured changein SWE.

These statistics were checked for differences betweennight and day, for differences with averaging period (6and 24 h), and for differences in temperature bin size(0.25° or 1°C). The resulting minor differences were notstatistically significant.

c. Surface temperatures compared with brightbandheights

Section 4a demonstrated that, on average, the BBHcorresponds to a free-air temperature of 1.8°C. Section4b demonstrated that roughly half of the precipitationpasses through the snowpack as rain at a surface tem-perature of 1.5°C (Ts-half), and existing snow cover be-comes likely to melt during a precipitation event withsurface temperatures above 3°C (Ts-50). These indicesprovide a means to translate the radar BBHs into sur-face hydrologic events. Figure 11 shows the mean sur-face temperature in the American River basin at thealtitude of each radar’s BBH for each hour of the day.At all stations, there was a clear diurnal cycle, wherethe surface was warmer during the day and cooler atnight. Chowchilla, to the south, measured higher BBHs,corresponding to colder surface temperatures in theAmerican River basin. Chico, to the north, measuredlower BBHs, corresponding to warmer surface tem-

peratures in the American River basin. Bodega Bay, onthe coast, varied the most diurnally from surface tem-peratures, and Grass Valley, in the Sierra foothills, var-ied the least diurnally. This could be because it wasoften not yet precipitating in the American River basinwhen it was raining at Bodega Bay, or it could be be-cause air masses above Bodega Bay arrived straight offthe ocean and were least affected by diurnal heatingand cooling from the continent. Temperatures corre-sponding to radar melting levels were quite variable inall cases, as represented by standard deviations roughlybetween 1° and 2°C at all hours of the day (Fig. 11b).Standard deviations were most variable at Bodega Bayand least variable at Grass Valley because Grass Valleyis geographically closest to the American River basin.

The diurnal variation in surface temperatures atBBHs can best be explained by revisiting our March2005 case study (Fig. 12). BBHs at Bodega Bay andGrass Valley generally fell within the 0° to 3°C tem-perature contours determined from Oakland soundings(Fig. 12a). However, the correspondence between BBHand temperature became less well defined when com-pared with contours of surface temperatures (from the23 stations in Table 3) versus elevation (Fig. 12b).These deviations were due primarily to diurnal changesin solar radiation, which heats the surface and not thefree air even during rain events. The differences be-tween surface and free-air temperatures were particu-larly 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 thetotal energy balance during storm events (18%–30% ofthe energy available for snowmelt during rain-on-snow

FIG. 9. (a) Surface temperatures and (b) forms of surface precipitation compared toradar-measured melting levels, as in Fig. 7, but for 18–24 Mar 2005 (PST).


events in Oregon in February 1996; Marks et al. 1998),it is by far the most predictable component and hasconsistent diurnal patterns. Marks et al. (1998) ob-served daily fluctuations in net radiation during rain-on-snow events ranging from �30 W m�2 at night to apeak of 50 to 100 W m�2 at solar noon. While this isabout half (or less) of the diurnal variation observedduring a subsequent sunny period, it is not negligible.This range in radiation alone would result in peak day-time melt rates about 1.3 mm h�1 higher than nighttimeminimum melt rates. Assuming that the specific heat ofvegetation (2928 J kg�1 K�1; Thom 1975) controls localsurface temperatures, this radiation increase wouldraise peak daytime surface temperatures about 1.5°Chigher than nighttime minimum temperatures. Thisvalue is similar to the observed differences in dailyvariations between the radar and surface temperatures.

These results suggest that melting levels at the sur-face should be adjusted up (to higher elevations) duringdaylight hours and down (to lower elevations) duringnighttime hours, particularly at the onset of a storm.

FIG. 10. The probability of (a) fractional SWE gain or lossduring 12-h periods with precipitation for the 10 American Riverbasin snow pillow stations marked with an asterisk in Table 3 forthe five winters from 2001 to 2005. Contours show probability ofdistributions at each temperature, and the heavy dashed lineshows the mean. (b) Percent of occurrences with 100% of pre-cipitation contributing to SWE gain (�SWE/�precip � 1, dashedline, left axis), only part of precipitation contributing to SWE gain(0 � �SWE/�precip � 1, width of region between dashed andsolid lines), and SWE loss (�SWE/�precip � 0, solid line, rightaxis). (c) Percent of occurrences of snow (dashed line, left axis),mixed rain and snow (width of region between dashed and solidlines), and rain (solid line, right axis) observed falling from the skyat Donner Summit. [Note: (c) is based on data collected by theU.S. Army Corps of Engineers (1956).]

FIG. 11. (a) Mean and (b) standard deviation of American Riverbasin surface temperatures at the elevation closest to the mea-sured BBH as a function of hour of day (PST).


Figure 13 illustrates what these adjustments might be,on average, for corresponding each radar’s measuredBBH with the 1.5°C contour, that is, the level whereprecipitation is equally likely to fall as rain or snow, inthe American River basin. Bodega Bay BBHs neededto be adjusted up 270 m during daylight hours (0600 to1800 PST) and adjusted down 8 m during nighttimehours (1800 to 0600 PST). Grass Valley (the profilerclosest to the ARB) needed to be adjusted down 78 mduring the day and down 205 m at night. When com-pared with the surface 3°C contour, where surface snowcover is likely to melt, Bodega Bay heights were anaverage of 223 m too high during the day and 488 m toohigh during the night. Grass Valley heights were anaverage of 318 m too high during the day and 433 m toohigh at night. Thus, surface snowmelt is always lowerthan the elevation where falling snowflakes melt torain.

FIG. 12. (a) Profile of temperatures observed at Oakland sound-ing compared with observations of melting level at Bodega Bay(black squares) and Grass Valley (white circles) for 18 to 25 Mar2005, in local time. Heavy black contour represents 0°C, and eachcontour line represents a change of 1°C, with lighter shades indi-cating warmer and darker shades indicating colder temperatures.(b) As in (a), but for surface temperatures in the American Riverbasin. (c) Difference between surface and free-air temperaturesshown in (b) and (a). Positive values indicate warmer surfacetemperatures.

FIG. 13. (a) Mean and (b) std dev of the height difference inmeters between the altitude of the 1.5°C contour on the surface ofthe American River basin and the radar-measured brightbandheight.


d. Uncertainties, corrections, and application tohydrologic nowcasting

The spatial and diurnal patterns of how BBHs relateto surface temperatures in the Sierra Nevada provideguidance for how these radar measurements can best beemployed in hydrologic forecasting. There is a greatdeal of uncertainty in all of the measurements, but thedata presented here can help quantify that uncertaintyand, to some degree, reduce it. BBH can be determinedto within 100-m vertical resolution (White et al. 2002).However, the error associated with directly interpretingthat height as the elevation where snowfall changes torain near the surface (here inferred to be 1.5°C) rangedfrom 326 to 457 m (Table 4). Assigning constant eleva-tion offsets based on how each location related geo-graphically to the American River basin reduced thoseRMSEs to 296 to 425 m (Table 4). Applying a furthercorrection, with different offsets based on time of day(i.e., as illustrated in Fig. 13), reduced the RMSEs to273 to 371 m (Table 4). The average improvement inRMSE through this process was 78 m. If we want todefine the surface melting level as the height where thesnowpack is likely to melt (here inferred to be 3°C),offset corrections become even more important in ad-justing the radar melting levels to lower elevations.Table 4 details the RMSEs for the original and cor-rected heights at this level, which resulted in an averageimprovement of roughly 200 m.

Over most of the North Fork of the American Riverbasin, the increase in contributing area with elevation islinear (Fig. 2), such that a change in melting level of 100m corresponds to a change in runoff contributing areaof about 48 km2, or 5% of the total basin area. Thus, anRMSE of 500 m corresponds to an error of 25% of thebasin’s contributing area. A decrease of RMSE of 78 mrepresents a hydrologic forecasting improvement of 4%contributing area, and a decrease in RMSE of 200 mrepresents an improvement of 10% contributing area.For example, consider the March 2005 case study peakof 1.3 mm h�1 (11 300 ft3 s�1). Assuming a constant rainrate 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 correc-tions detailed above, the remaining RMSE is about 300m, corresponding to 15% of the basin contributing area,leaving room for further improvements.

5. Conclusions

a. Summary

The elevation where precipitation transitions fromcontributing to snow water storage versus contributingto runoff is critical for hydrologic monitoring and fore-casting all along the western coast of North America.Vertically pointing profiling radars are able to detectthe elevation where snow changes to rain in the freeatmosphere. Thus, a critical step toward improving hy-drologic forecasts in these mountain basins is to quan-titatively relate the radar-observed snow level to whatis happening on the ground. Specifically, forecastersneed to know which altitude offset would best distin-guish between surface snow accumulation and runoffand need to characterize the uncertainty in this verticaloffset.

To accomplish these objectives, this study trackedthe altitude of the melting level from the free atmo-sphere, 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 BBHsmeasured at profiler sites located some distance fromeach other and from a river basin represent meltinglevels on the mountain surface? Because of the large-scale nature of most of California’s winter storm events,radars quite distant (300 km) from a river basin can beused to interpret surface melting levels. On average,RMSEs of 326 to 457 m occur when BBHs are directlyassumed to represent surface melting levels. Additionallocal errors may arise when different basin locations arein warm versus cold sectors of a frontal passage or whendrainage flows resulting from evaporative cooling de-velop in canyons (Steiner et al. 2003).

Second, what offsets, as a function of profiler andbasin locations and of time of day, should be used toadjust BBH measurements to better match surface

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.

RMSE (m)

Ts � 1.5°C Ts � 1.5°C Ts � 1.5°C Ts � 3°C Ts � 3°C Ts � 3°C

No correction Constant offset Diurnal correction No correction Constant offset Diurnal correction

Bodega 443 425 371 551 431 377Grass Valley 326 296 273 478 301 283Chico 328 323 296 444 338 312Chowchilla 457 320 302 644 317 306


melting levels? Forecasters should incorporate spatialgradients of decreasing BBHs with increasing latitude(on average, �41.4 m °�1), due to regional temperaturegradients, and with decreasing distance to where thebright band intersects the windward mountain slope,due to adiabatic and diabatic cooling (Medina et al.2005). In addition to these spatial offsets, the fate ofprecipitation at the mountain surface will also be modi-fied by diurnal cycles in solar radiation, which result inwarmer temperatures and greater melt rates duringdaylight hours, even during cloudy precipitation events.Depending on a profiler’s location, BBHs would needto be shifted down 127 to 278 m further during the nightthan during the day to correspond with surface obser-vations. Incorporating both space and time of day toBBH offsets reduced RMSEs in predicted surface melt-ing levels to between 273 and 371 m, an average of 73m less error than when these corrections were notimplemented.

Third, given an existing snowpack, what are the prob-abilities of snow accumulating versus melting at a giventemperature during a precipitation event? Surface mea-surements indicate that precipitation has an equal like-lihood of falling as rain or snow at a surface tempera-ture of 1.5°C, but a preexisting snowpack is 50% likelyto melt at a surface temperature of 3°C. Case studiesillustrate that rain may be incorporated into the snow-pack, pass through the snowpack, or be accompaniedby melting snow, depending on antecedent snow tem-perature and liquid water content.

b. Applications and implications for modeldevelopment

Accurate forecasts or nowcasts of the altitude wheresnow changes to rain are crucial for hydrologic fore-casting not only in mountain basins in California butalso all along the western coast. The current “rule-of-thumb” used by the California–Nevada River ForecastCenter (CNRFC) for forecasting the surface meltinglevel is to take the forecasted freezing level (0°C iso-therm) and subtract 300 to 460 m (1000 to 1500 ft; E.Strem, CNRFC, 2006, personal communication). Thismelting level, combined with basin areas above andbelow, is then used to determine how mean areal pre-cipitation for a basin is partitioned into rain and snow.This paper provides statistical relationships for howBBHs could be used to check and improve these fore-cast adjustments.

The observations described here illustrate the physi-cal processes that must be included in a model to accu-rately represent elevation zones of rain versus snow todetermine hydrologic runoff in any area with oro-graphic precipitation. First, the model must accurately

represent how the atmospheric melting layer elevationand thickness change with space and time, by reproduc-ing large-scale latitudinal gradients and topographicallyinduced changes, such as slope-induced cooling (Me-dina et al. 2005). Second, the model must include spa-tial representation of the surface energy balance to ac-count for the effects of radiation on near-surface airtemperatures and melting. This could be accomplishedmost precisely by modeling the thickness and opticalproperties of the cloud cover (radiation transmission)and the radiative properties (albedo) of the land sur-face cover. Radiative heating is neglected in most at-mospheric models of orographically induced precipita-tion (Barros and Lettenmaier 1994), but could haveimportant feedback effects on not only the type andfate of hydrometeors but also on the type and locationof secondary mesoscale circulation patterns (Szyrmerand Zawadski 1999). Finally, the model needs a spatialaccounting of snowpack characteristics, such as density,temperature, and liquid water content (as in the UtahEnergy Balance Model; Tarboton and Luce 1996), todetermine when and where rain will be incorporatedinto the snowpack, as opposed to directly contributingto runoff. Collectively, these modeling requirementsconsist of model components that have already beendeveloped in various separate forms and should bemuch simpler to implement than those required for ac-curate quantitative precipitation estimation (QPE;Wang and Georgakakos 2007, manuscript submitted toJ. Geophys. Res.) Accurate implementation of theabove processes should improve model representationsof where precipitation contributes to runoff, and thuswould provide immediate benefits to hydrologic fore-casting in coastal mountain regions.

In addition to providing ad hoc information for op-erational forecasts, vertically pointing radars also pro-vide a resource for model testing and development. TheCalifornia and Oregon coasts are unique in having over10 yr (since 1996) of vertically pointing Doppler radarmeasurements, which include BBHs. These could beused to check model representations of atmosphericmelting levels as a function of space and time. Mosttests of microphysics parameterization schemes areconstrained to the short duration of major field cam-paigns [Sierra Cooperative Pilot Project (SCPP; Meyersand Cotton 1992; Wang and Georgakakos 2007, manu-script submitted to J. Geophys Res.) or the Improve-ment of Microphysical Parameterization through Ob-servational Verification Experiment (IMPROVE;Colle et al. 2005)] because these are the only timesvalidation data are available. A multiyear investigationcomparing model melting levels with BBHs could pro-vide statistics of how well various schemes represent


atmospheric melt processes under different meteoro-logical conditions.

Most operational and many scientific hydrologymodels rely on temperature to identify whether precipi-tation falls as rain or snow, such as Snow-17 (Anderson1976), Variable Infiltration Capacity (VIC; Cherkaueret al. 2003), and Distributed Hydrology Soil VegetationModel (DHSVM; Wigmosta et al. 1994, 2002). Theprobability distribution functions presented here oflikelihood of snow gain or loss as a function of tem-perature may help to improve such model parameters.

As temperatures continue to warm across the west-ern United States, basins at progressively higher eleva-tions will experience mixed rain and snow during winterprecipitation events, impacting both storm runoff andsnowpack storage for the following summer. Changesin the frequency of high melting levels represent achanging baseline for hydrologic runoff and will makestatistical forecasts based on historic datasets less reli-able. The spatial patterns of precipitation type and in-tensity will influence landslides and ecosystem pro-cesses. These applications provide impetus to improveour physical knowledge of processes in this regime, andlessons learned from the relatively lower-elevationAmerican River basin may become crucial for under-standing and modeling streamflow in mountain basinsacross the west.

Acknowledgments. We thank the dedicated engineer-ing staff in the Physical Sciences Division of NOAA/ESRL for deploying and maintaining the NOAA windprofilers used in this study. Funding for this study wasprovided by a CIRES Postdoctoral Fellowship throughNOAA and the University of Colorado, Boulder.Thank you to Socorro Medina for discussions of thepaper’s content and to two anonymous reviewers forcomments.

REFERENCES

Anderson, E. A., 1976: A point energy and mass balance model ofa snow cover. NOAA Tech. Rep. NWS 19, 150 pp.

Barros, A. P., and D. P. Lettenmaier, 1994: Dynamic modeling oforographically induced precipitation. Rev. Geophys., 32, 265–284.

Battan, L. J., 1973: Radar Observations of the Atmosphere. Uni-versity 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 char-acteristics of daily snow amount for stochastic simulation ofrain-on-snow events. Proc. 88th Western Snow Conf., Sacra-mento, CA, Western Snow Conference, 110–121.

Cherkauer, K. A., L. C. Bowling, and D. P. Lettenmaier, 2003:Variable infiltration capacity cold land process model up-dates. Global Planet. Change, 38, 151–159.

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 sensitivitystudies. J. Atmos. Sci., 62, 3535–3558.

Cunningham, R. M., 1947: A different explanation of the “brightline.” J. Meteor., 4, 163.

Dettinger, M. D., D. R. Cayan, M. K. Meyer, and A. E. Jeton,2004: Simulated hydrologic responses to climate variationsand change in the Merced, Carson, and American river ba-sins, Sierra Nevada, California, 1900–2099. Climatic Change,62, 283–317.

Fabry, F., and I. Zawadzki, 1995: Long-term radar observations ofthe melting layer of precipitation and their interpretation. J.Atmos. Sci., 52, 838–851.

Farnes, P. E., 1967: Criteria for determining mountain snow pil-low sites. Proc. 35th Western Snow Conf., Boise, ID, WesternSnow Conference, 59–62.

Feiccabrino, J., and A. Lundberg, 2008: Precipitation phase dis-crimination by dew point and air temperature. Proc. 75thWestern Snow Conf., Kailua-Kona, HI, Western Snow Con-ference, 141–146.

Ffolliott, P. F., and K. N. Brooks, 1983: Rain-on-snow flood eventin central Arizona. Proc. 51st Western Snow Conf., Vancou-ver, WA, Western Snow Conference, 130–138.

Hall, R. L., and J. F. Hannaford, 1983: Analysis of three rain-on-snow floods in the Sierra Nevada, California. Proc. 51st West-ern Snow Conf., Vancouver, WA, Western Snow Conference,46–54.

Harr, R. D., 1986: Effects of clearcutting on rain-on-snow runoffin 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: Evapo-transpiration and Irrigation Water Requirements. AmericanSociety of Civil Engineers, 332 pp.

Jeton, A. E., M. D. Dettinger, and J. L. Smith, 1996: Potentialeffects of climate change on streamflow, eastern and westernslopes of the Sierra Nevada, California and Nevada. USGSWater Resources Investigations Rep. 95-4260, 44 pp.

Kattelmann, R., 1997: Flooding from rain-on-snow events in theSierra Nevada. Destructive Water: Water-Caused Natural Di-sasters, 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 insnowfall versus rainfall in the western United States. J. Cli-mate, 19, 4545–4559.

Lettenmaier, D. P., and T. Y. Gan, 1990: Hydrologic sensitivitiesof the Sacramento-San Joaquin River basin, California, toglobal warming. Water Resour. Res., 26, 69–86.

L’Hôte, Y., P. Chevallier, A. Coudrain, Y. Lejeune, and P.Etchevers, 2005: Relationship between precipitation phaseand air temperature: Comparison between the BolivianAndes and the Swiss Alps. Hydrol. Sci. J., 50, 989–997.

Lin, Y.-L., R. D. Farley, and H. D. Orville, 1983: Bulk parameter-ization 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 sensi-tivity of snowmelt processes to climate conditions and forestcover during rain-on-snow: A case study of the 1996 PacificNorthwest flood. Hydrol. Processes, 12, 1569–1587.

Marshall, J. S., R. C. Langille, and W. McK. Palmer, 1947: Mea-surements of rainfall by radar. J. Meteor., 4, 186–192.

Marwitz, J. D., 1983: The kinematics of orographic airflow duringSierra storms. J. Atmos. Sci., 40, 1218–1227.


——, 1987: Deep orographic storms over the Sierra Nevada. PartI: Thermodynamic and kinematic structure. J. Atmos. Sci., 44,159–173.

McCabe, G. J., M. P. Clark, and L. E. Hay, 2007: Rain-on-snowevents 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 poten-tial for wintertime quantitative precipitation forecasting overmountainous 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: Awind tunnel and theoretical study of the melting behavior ofatmospheric ice particles. IV: Experiment and theory forsnow flakes. J. Atmos. Sci., 47, 584–591.

Mittermaier, M. P., and A. J. Illingworth, 2003: Comparison ofmodel-derived and radar-observed freezing-level heights: Im-plications 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: Recom-mendations from a USWRP workshop. Bull. Amer. Meteor.Soc., 86, 1619–1632.

Rohrer, M., 1989: Determination of the transition air temperaturefrom snow to rain and intensity of precipitation. Instrumentsand Observing Methods, WMO Tech. Rep. 328, Rep. 48,World Meteorological Organization, 475–482.

Schultz, P., 1995: An explicit cloud physics parameterization foroperational numerical weather prediction. Mon. Wea. Rev.,123, 3331–3343.

Slack, J. R., and J. M. Landwehr, 1992: Hydro-climatic data net-work (HCDN): A U.S. Geological Survey streamflow dataset 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 valleysunder 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 stratiformclouds. J. Atmos. Sci., 41, 3227–3237.

Storck, P., 2000: Trees, snow and flooding: An investigation offorest canopy effects on snow accumulation and melt at the

plot and watershed scales in the Pacific Northwest. WaterResources Series Tech. Rep. 61, Dept. of Civil Engineering,University of Washington, 176 pp.

Szyrmer, W., and I. Zawadski, 1999: Modeling of the meltinglayer. Part I: Dynamics and microphysics. J. Atmos. Sci., 56,3573–3592.

Tarboton, D. G., and C. H. Luce, 1996: Utah Energy Balancesnow accumulation and melt model (UEB): Computer modeltechnical description and users guide. Utah Water ResearchLaboratory and USDA Forest Service Intermountain Re-search Station, 41 pp. [Available online at http://www.engineering.usu.edu/dtarb/.]

Thom, A. S., 1975: Momentum, mass and heat exchange of plantcommunities. Vegetation and the Atmosphere, J. L. Monteith,Ed., Vol. 1, Academic Press, 57–109.

Trabant, D. C., and G. P. Clagett, 1990: Measurement and evalu-ation of snowpacks. Cold Regions Hydrology and Hydraulics,W. L. Ryan and R. D. Crissman, Eds., American Society ofCivil Engineers, 39–93.

U.S. Army Corps of Engineers, 1956: Snow hydrology: Summaryreport of the snow investigations. North Pacific Division, U.S.Army Corps of Engineers, 437 pp.

Wang, J., and K. P. Georgakakos, 2005: Validation and sensitivi-ties of dynamic precipitation simulation for winter eventsover the Folsom Lake watershed: 1964–99. Mon. Wea. Rev.,133, 3–19.

Westrick, K. J., and C. F. Mass, 2001: An evaluation of a high-resolution hydrometeorological modeling system for predic-tion of a cool-season flood event in a coastal mountainouswatershed. 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 detectionalgorithm 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 dis-tributed hydrology-vegetation model for complex terrain.Water Resour. Res., 30, 1665–1679.

——, B. Nijssen, P. Storck, and D. P. Lettenmaier, 2002: The dis-tributed hydrology soil vegetation model. Mathematical Mod-els 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 char-acteristics within coexisting rain and wet snow. J. Appl. Me-teor. Climatol., 45, 1450–1464.

Zuzel, J. F., R. N. Greenwalt, and R. R. Allmaras, 1983: Rain onsnow: Shallow, transient snowpacks with frozen soils. Proc.51st Western Snow Conf., Vancouver, WA, Western SnowConference, 67–75.


Rain versus Snow in the Sierra Nevada, California ... · mosphere must be related to the melting...

Documents

Transcript of Rain versus Snow in the Sierra Nevada, California ... · mosphere must be related to the melting...