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
|
|
The Northwest Avalanche Center (NWAC) began building
a network of automated weather stations shortly after they
beganforecastingin1976.Sincethenthenetworkhas
grown to be one of the most valuable sources of mountain
weatherdatainthe northwesternUnitedStates.Instru-
mentationateach sitehasbeentestedand modifiedto
withstand theharsh winter environment. Thedataareused
to help determine snow layering and avalanchepotential,
the depth and extent of freezing rain that impairs driving
conditions,and overall mountain weatherconditionsin
the Olympic and Cascade mountains of Washington and
northern Oregon.The location ofsensors has proved in-
valuable forobserving and defining unique phenomenon
associated with easterly pass flows, arctic inversions, and
topographically forced convergence. Until now, however,
the NWAC data have been available only for daily, opera-
tional use by forecasters.We have undertaken the task of
reformatting the NWAC data and adding quality control
flags to each variable to make them more readilyavailable
forabroadrangeofweathermodelverificationand
climatological purposes. Because this is the first network
of automated, hourly, mountain weather data to attempt
a qualitycontrol analysis,we have solicited assistance
from the Western Regional Climate Center and the USDA
Climate Data AccessFacility.Itisanticipated thatthe
experienceobtainedfromexaminationofNWACdata
can be applied in developing guidelines for quality con-
trolofothernetworks,suchastheUSDA-USDI/BLM
RAWS.During theretrospective analysisof dataqual-
ity,problems andissueswere identified thatcouldbe
traced both to the original establishment of the network
and its sensorsand tothe waythenetworkwas main-
tained.Inthiswaytheuncertaintyofeachvaluewas
assessedandappropriatequalityassurancemeasures
were assigned using a system of three flags, 1) data qual-
ity,2) data problem(if any),and 3) adjustmentmethod
(ifany).Although the originaldata values always were
preserved, adjusted values were suggested as substitutes
whenever a value could be identified as "suspect" with
reasonableconfidence.TheNWACdataare organized
in a way that the file can be read by typical Fortran or C
programs or imported intoa spreadsheet.
|
|
Key Words: avalanche, storm cycle, classification trees, ava-
lancheforecast
ABSTRACT
Weather and avalanche observationsfromAlta,Utah for
the period 1983-1994provided the basis of thisanalysis,
which used classification and regression trees to rank and
scoresimpleweatherfactorscontributingtoavalanche
activity.Weatherfactorsincluded dailyobservationsof
wind speed, airtemperature and snow fall.The product
of new snow times average wind speedwas also included.
The analysis tested these variables,along with 2-day and
3-daytrends,maximum and minimum airtemperatures
and total snow depth in terms of theirimportance to ex-
plaining dailyavalanche activity.We constructed deci-
siontreestorelatethe stormvariablestotheavalanche
responsevariables:avalanchesornot(avalancheday),
number of avalanche releases, maximum avalanche size,
sum of avalanche sizes and lumped size class.Backwards
timeshifting ofavalanche records, to more closely corre-
spondtoslab-formingconditionsduringstormcycles,
improved classificationand regression accuracy. Thehigh-
est classification accuracieswere found for the avalanche
day (ornot) and maximum avalanche size.
INTRODUCTION
Mostareasinthe USsubject toavalanche controlprac-
tices record weather variables with the aim to use these in
daily forecasts of avalanche activity.Some of these areas
simply record the weather observations,and rely prima-
rilyonapurelyhumanforecast,devoidofquantitative
analysis.With a forecaster who has much experience, the
forecaststendtohavehighaccuracyandgreatutility
(LaChapelle,1980).Nevertheless,the records of weather
and avalanche occurrence provide the basis forquantita-
tive analysis,forexample to explore relationships in the
|
|
![[CONVERTED BY MYRMIDON]](Imgs/convBy.GIF)