M a n a g e m e n t

a n d

A n a l y s i s

o f

S n o w,

A v a l a n c h e

a n d

C l i m a t e

D a t a

web[http://snow.water.ca.gov/]. Thefirst coursesweremeas-

ured in 1910,and most contain 50 years of monthly SWE
andsnowdensitymeasurements,whichwerecollectedfrom
one to six times perseason.Many of these courses were
createdandheavilysampled in the1930's, thoughmeasure-
ments in the 1940's are sparse, sampling density increases
againin the1950's.Most stations weresampledat leastfour
times per year from the 1960's until the present.

METHODOLOGY

These data present several challenges forclimate analy-
ses because they were collected with the intended use for
waterresources management.Ideally, stations would be
measured daily throughout the season and continuously
overmany decades,oreven millennia.However,about
20% of the 44,000 SWE measurementswere unusable for
timeseriesanalysis duetoinfrequentsampling,having
less than three measurements perstation-year.
Sampling was conducted on a monthly basis,usually
clustered within a few days of the first day of the month.
Therefore,maximum monthly SWE values were interpo-
lated to the first of the month.Furthermore, missing data
points were calculatedwith linear regressions.Additional
problems include that themaximum SWEday is estimated
and hence possibly an artifact if the actual maximum day
is afterthe last sample, and finally,stations were added
and removed throughout the years.
Newerstationswere notevenlydistributedbyeleva-
tion, which accounts for about 6% of the SWE variability
[Aguado, 1990], so regressions were grouped by elevation
zonestominimize errorstemmingfromstationfluctua-
tions.Also, individual station SWE values and snowmelt
timingwasstandardizedbyconvertingestimatesinto
monthlyfractionsofthestation'sseasonmaximumand
differences from this station's average snowmelt month.
The three samples peryearscreen leaves 281qualify-
ingstations(figure12)ranging between1500and3500
meters inelevation,constituting34,500station-yearsof
data.Thesedata alsomeet thefollowingqualityassur-

2

ance:The day of maximumSWE is notsimply the last
day measured unless that month is the average month of
meltand itisanaverage-to-wetyear.Thisdiscourages
false snowmelt timing calculations due to sampling bias,
yetaccountsforprecipitationvariabilityand associated
snowmelt fluctuations.
Theyearlydatapointsandsnowcourseinformation
was imported into the statistics software package SAS for
analysisanddata screening.Measurementswere inter-
polated to the first of the month by adding a station's SWE
to the product of the slope to the adjacent month's value
and the number of days from the first for that station-year.

(1) SWE
inte r po lated^{= SWE}
x^{+ days}x ^(SWE+1 ^-SWE
x ^)/(DOYx +1 ^-
DOY_x )^x

SWE x= given monthly SWE measurement SWE x +1= following month's SWE measurement DOY= day of year on which a given month was sam- pled x DOYx +1= day of yearon which following month was sampled daysx= number of days after the first of the month sam- ple was measured

Equation (1)applies if the sample day is afterthe first of
the month.Whenthe sample day is before the first,then
a similar equation applies to the previous month to mini-
mizetheinterpolationerror.Equation(1)requiresfre-
quentsamplingwhichdoesnotexistformany station-
years.Therefore, station-decade averages and overall sta-
tion averages were used as the slope when required.In
some cases, especially January or June adjustments, a sta-
tionaveragedidnotexist,soanaverageaccumulation
slopefromthatelevationzone wassubstituted.Meas-
urements taken before January 1 used a January-February
slope.Likewise,samples taken afterJune 1 used a May-
June slope since onlyseven measurements were carried
out in Decemberand none in July.