I n s t r u m e n t s

a n d

M e t h o d s

sensoris the result of many improvements overthe years
andnowgivesanaccuracyof±1cm,whateverthe
characteristicofthe snowsurface.The reliabilityofthis
kindofsensorhasbeenproven,even duringsnowfalls
(figure 1) and allows us to precisely mark accumulations,
packingdown and meltingperiods[Lecorpsand Sudul,
1989].
The idea of designinga set of several acoustic sensors to
follow the evolution of the snow surface, led to the device
as shown in the figure 2 and photo 1.
Some hardware and software adaptations to the initial
device were needed. A multiplex systemhas been devel-
oped to achieve the monitoring of all sensors, one after the
other(figure 3),a sampling rate of 15 min.used, and the
data arerecordedon a computer placedin a shelter on the
mast.It is possible to use the telephone network in order
to consult the system and to collect the most recent data.

Where can the sensor device be installed?

The "Dôme des Petites Rousses",located close to the ski
resort of Alpe d'Huez (south-east of Grenoble-France), was
chosen as the best study area because of:

*therelatively completerecord ofmeteorologicaldata
(windvelocityanddirection,airtemperature,water
equivalentofprecipitation).Thesedata are available
onan hourlybasisfromthestartofNovembertothe
end of April for5 years.

• the easy access by the Vaujany cable car whateverthe
  weather conditions.
• the complete observation of avalanche activity (natural
  orartificial) made by the ski patrollers of Alpe d'Huez
  ski resort.
  Moreover,the slope is perpendicular to prevailing wind,
  subject to high altitude weather conditions (as high wind
  speed, hoar frost formation, strong precipitation,...) and is
  relatively steep (gradient of about 35º).

Processing and Graphs

Unfortunately with data that are collectedin such an envi-
ronment,input is not alwaysconsistent.The main prob-
lems to be resolved can be summarized as follow:

*strong average winds (> 25 m/s)can shake cables and
deviate ultrasonic pulses.Then,the pulses which are
sent back bythe snow surface are unable to reach the
receiver.

*heavysnowfallscaninduceinterferences leadingto
wrong measurements.

*the frost (photo2)can stiffen and overload the whole
framework of the system. It canobstruct thesensor open-
ing and can be at the origin of mechanical vibrations.
Wrong values can be eliminated through a statistical post-
processing. This last one is able to restore the missing data
in certain conditions.The snowpack thickness isrepre-

• 
  
  Figure 1 : graph of snow thickness at the "Col du Lac Blanc" (2700m ) usinga snow depth sensor.

130