In many hydrologic models, determination of precipitation type is indexed to surface air temperature, and the selection of the maximum snow and minimum rain thresholds are chosen empirically, by calibration or using published values (e.g. U.S. Army Corps

1
Partitioning of precipitation into rain and snow
in distributed hydrologic simulations in the
Western Cascades, Oregon, USA. Edwin P. Maurer1,
Jasmine Cetrone2, Socorro Medina2, and Clifford
Mass2 1. Civil Engineering Department, Santa
Clara University, Santa Clara, CA 95053-0563 2.
Department of Atmospheric Sciences, University of
Washington, Seattle, WA 98195-1640
AMS Annual Meeting 2004 Poster Session 3 Poster
P3.5
3
Meteorology in South Santiam Basin during IMPROVE
Observation Period
ABSTRACT One of the greatest challenges in
hydrologic modeling in areas with significant
orographic influences is accurate simulation of
the precipitation fields, since this drives the
streamflow response. In the northwest United
States, where most of the precipitation occurs
during the cool season, another major factor in
streamflow simulation is the determination of
whether the precipitation is falling as rain or
snow, since these strongly influence the timing
of the resulting runoff. The partitioning of
precipitation in distributed hydrologic models
into rain, snow, or a mixture of the two is often
based on surface air temperature, since this is
included in the station observation records that
provide the precipitation and other
meteorological data used to force the model. This
study examines the adequacy for hydrologic
modeling of using surface air temperature to
determine this partitioning of rain and snow in
the Santiam River basin, Oregon. The western
slopes of the Cascade mountain range in Oregon,
specifically an area including the South Fork of
the Santiam River, was the geographical focus for
the second phase of the research effort dubbed
Improvement of Microphysical PaRameterization
through Observational Verification Experiment
(IMPROVE-2). This intensive field observation
campaign was carried out from 26 November through
22 December 2001, with measurements used to
perform comprehensive verification of cloud and
precipitation microphysical processes
parameterized in mesoscale models. Included in
the suite of IMPROVE-2 observations were both
scanning and vertically pointing radar. While
scanning radar observations in areas of complex
terrain, such as the western Cascades, are
problematic due to ground clutter and beam
blocking, vertically pointing radar does not
suffer from this. We show that, by replacing the
surface air temperature-based algorithm in a
distributed hydrologic model with a freezing
level determined with S-band radar supplemented
by other observations, significant improvement in
the simulated hydrograph can be obtained.
Elevation and surface air temperature at each
pixel (interpolated from observations with a
lapse rate of -5.5 C/km) were combined with the
radar-observed freezing level to illustrate the
variability in surface temperatures associated
with rain and snow. At each time step, the radar
detected 0 level was projected across the basin.
The surface air temperatures for pixels with
elevations
Examination of P/T/SWE relationships at 2 SNOTEL
sites in basin
within 10m of this 0 level, were taken as
samples of the minimum surface air temperature at
which any rain occurs, Tmin(rain). At a distance
300 meters below the 0 level all melt is assumed
complete, and pixels with elevations close to
this are representative of the maximum surface
air temperature at which snow occurs, Tmax(snow).
A linear mixture is assumed between these levels.
Surface Air Temperatures for Snow/Rain Inferred
from Radar 0C Level
Sample reflectivity data from the NOAA/ETL S-band
vertically pointing radar. The data shown are for
2215 UTC 13 December - 0115 UTC 14 December 2001.
Note the bright band in red, the top of which is
typically associated with 0C temperatures.
(Houze and Medina, 2002)
JUMP_OFF_JOE
LITTLE_MEADOWS
1
Focus of This Study
Minimum during IMPROVE-2 period Maximum during IMPROVE-2 period Average during IMPROVE-2 period
Tmin(Rain) -9.7 -0.6 -4.9
Tmax(Snow) -6.7 1.7 -2.4

• In many hydrologic models, determination of
  precipitation type is indexed to surface air
  temperature, and the selection of the maximum
  snow and minimum rain thresholds are chosen
  empirically, by calibration or using published
  values (e.g. U.S. Army Corps of Engineers, 1956),
  or are selected arbitrarily (e.g., Bowling et
  al., 2003). Some models use one fixed temperature
  as a division between rain and snow, rather than
  using a range with mixed (frozen and solid)
  precipitation (e.g. Bicknell et al., 2002). The
  NWSRFS implementation of the Sacramento model
  (Office of Hydrologic Development, 2002) is rare
  in allowing the incorporation of freezing level
  data.
• The following questions take advantage of the
  availability of radar-based freezing level
  observations in the study region to look for
  opportunities for improving streamflow
  simulations in regions of complex topography and
  strong orographic influence
• How well do surface temperature-based methods
  work for determining whether precipitation is
  falling as rain, snow, or a mixture?
• Does the radar-detected 0C level differ
  substantially from the air-temperature-based
  method?
• Can the observed radar-based 0C level be used to
  improve streamflow simulations during the events
  studied during IMPROVE-2?

Observed 0 Level Based on Bright Band
Identification
This example, from LITTLE_MEADOWS, highlights 2
periods where the air temperature indexing and
radar 0C levels can give different results. Here
two periods are highlighted where the freezing
level is well above the station elevation
(indicating rain), while the surface air
temperature is below zero (indicating at least
partial snow).

• Observations show
• Surface air temperature is not a good indicator
  of whether precipitation is falling as rain or
  snow.
• This is especially evident for lower elevation
  JUMP_OFF_JOE site, closer to valley bottom, where
  there is essentially no correlation between air
  temperature during a precipitation even and
  whether snow is accumulating or melting.
• Even at LITTLE_MEADOWS, closer to the ridge, at
  air temperatures between 2 and 4C during this
  period, air temperature is a poor indicator of
  precipitation type.
• There is a wide discrepancy between the 2
  locations in the air surface temperatures
  associated with both rain and snow, indicating
  the use of one index for the basin could be
  problematic.

The average 700 - 925 mb wind speed for the
entire IMPROVE-2 period is 13.5 ms-1, thus
applying the 0 level at the S-Band location to
the entire basin introduces at most a 1 hour
timing error for any point in the basin on
average.
2
IMPROVE-2 Overview and River Basin for Study
IMPROVE is aimed at comprehensively checking and
improving the parameterization schemes currently
implemented in the Penn State/NCAR Mesoscale
Model (MM5), a mesoscale model that has been
extensively used for both research and
operational forecasting. The primary goal of
IMPROVE is to utilize quantitative measurements
of cloud microphysical parameters in a variety of
mesoscale features to improve the representation
of cloud and precipitation processes in mesoscale
models. The IMPROVE-2 field study, focused on
orographic clouds and precipitation in the Oregon
Cascade Mountains, was conducted 26 November
through 22 December 2001.  For more details,
see http//improve.atmos.washington.edu/
5
4
Summary
Effect of Precipitation Type Determination on
Hydrologic Simulations

• Based on surface observations of air temperature
  and snow accumulation, the surface air
  temperature is not a strong indicator of
  accumulation or melt of snowpack (or whether
  precipitation type is rain or snow).
• The first method of partitioning precipitation
  into rain and snow used radar detected 0
  elevation as that below which snow begins to
  melt 300 m below this elevation marked that
  below which all precipitation is rain.
• A second method combined radar data with the
  elevations and surface air temperatures at each
  pixel and time step in the basin. Populations of
  Tmin(rain) and Tmax(snow) (see Box 3 for
  definitions) were derived, the average of which
  provided temperature indices for partitioning
  precipitation into rain, snow, or a mixture.
• The use of the second method in a hydrologic
  model (see Trial 2, Box 4) produced measurable
  improvements in simulated peak flows over using
  values of Tmin(rain) and Tmax(snow) from
  literature.
• Using the first method (see Trial 3, Box 4)
  produced further improvements in the simulation
  of one flood peak, but the incremental
  improvement over using the second method was
  small.
• At two locations where snow was observed, the
  simulations by the hydrologic model were better
  with the first method than the second, though
  local effects complicate the accurate simulation
  of snow at one site.

Given the above differences in air
temperature-based versus radar-based
discrimination of rain and snow, we investigate
the sensitivity of streamflow simulations in
IMPROVE to incorporation of freezing level data,
using the DHSVM model (Wigmosta et al., 1994),
modified to ingest freezing level data.
RMSE for peak events (observed flows gt 60 m3/s)
Gauge 14185000 Gauge 14185900
Trial 1 43 57
Trial 2 40 40
Trial 3 39 39
Trial 1 Using air temperature-based indexing
of 2C Tmax(snow) and 0C Tmin(rain) (following
Bowling, et al., 2003 Office of Hydrologic
Development, 2002)
Using the radar detected 0C level, either
directly (Trial 3) or combining it with the
basin DEM and surface air temperatures to
estimate a Tmax(snow) and Tmin(rain) (Trial 2)
produces simulated improved hydrographs compared
to literature-based Tmax(snow) and Tmin(rain)
values (Trial 1), as reflected in the RMSE
values above. Trial 2 achieves most of the
decrease in RMSE, implying that calibrating
Tmax(snow) and Tmin(rain) using radar data may be
possible and beneficial to hydrologic
simulations. The simulation of snow at the 2
SNOTEL sites in the basin is more problematic.
The use of the variable radar-detected freezing
level (trial 3) improves the snow water
equivalent simulations at these points compared
to trial 1, while trial 2 is not a consistent
improvement. None of the trials could reproduce
the complete removal of snow at the Jump_off_Joe
site, indicating that other local factors are
important.
Trial 2 Using air temperature-based indexing,
as in Trial 1, but with average vales inferred
from radar-detected level (see box 3 above)
-2.4C Tmax(snow) and -4.9C Tmin(rain)
The IMPROVE-2 domain overlaps largely with the
South Santiam River basin, shown here, which has
a total basin area of 1,440 km2. This catchment
provides a spatial integrator for the observed
and modeled precipitation, and a valuable
validation tool for assessing precipitation
fields simulated by forecast models.
The South Santiam basin, during the IMPROVE-2
period, included many observational assets. The
subset of observations used in this study
included a vertically pointing S-Band radar,
daily and hourly cooperative observer stations,
SNOTEL stations, precipitation gauges installed
for this study (labeled IMPROVE), and USGS
streamflow gauges, shown on this map.
Trial 1
Trial 3 Using elevation indexing Observed 0
Level Based on Bright Band Identification (see
plot in Box 3 above), varies with time
Trial 2
REFERENCES Bicknell, B.R., J.C. Imhoff, J.L.
Kittle Jr., A.S. Donigian, Jr. and R.C. Johanson.
1997. Hydrological Simulation Program -- FORTRAN,
User's Manual for Version 11. EPA/600/R-97/080.
U.S. EPA, National Exposure Research Laboratory,
Athens, GA. Bowling, L.C., D.P. Lettenmaier, B.
Nijssen, L.P. Graham, et al. 2003, Simulation of
high latitude hydrological processes in the
Torne-Kalix basin PILPS Phase 2(e) 1 Experiment
description and summary intercomparisons, Journal
of Global and Planetary Change, 38(1-2), 1-30.
Houze, R.A. and S. Medina, 2002 comparison of
orographic precipitation in MAP and IMPROVE II,
In Preprints 10th Conference on mountain
meteorology and MAP meeting, Park City, UT, 17-21
June, 2002. Office of Hydrologic Development,
2002 National Weather Service Forecast System
Model User Manual, Section 3.3-RSNWELEV, National
Weather Service. U.S. Army Corps of Engineers,
1956 Summary Report of the snow investigations
Snow Hydrology, North Pacific Division, Portland,
OR, June 1956. Wigmosta, M.S., L.W. Vail, and
D.P. Lettenmaier, 1994, A distributed
hydrology-soil-vegetation model for complex
terrain, Water Resour. Res. 30, 1665-1679.
Trial 3