143

WEATHER AND SNOW OBSERVATIONS FOR AVALANCHE FORCASTING:

AN EVALUATION OF ERRORS IN MEASUREMENT AND INTERPRETATION

R.T. Marriottl and M.B. Moorel

Abstract.--Measurements of weather and snow parameters forsnow stability forecasting may frequently contain false ormisleading information. Such error~ can be attributedprimarily to poor selection of the measuring sites and toinconsistent response of the sensors to changing weatherconditions. These problems are examined in detail and someremedies are suggested.

INTRODUCTION

A basic premise of snow stability analysis foravalanche forecasting is that point measurements ofsnow and weather parameters can be used to infer thesnow and weather conditions over a large area. Dueto the complexity of this process in the mountainenvironment, this "extrapolation" of data haslargely been accomplished subjectively by anindividual experienced with the area in question.This experience was usually gained by visiting theareas of concern, during many differing types ofconditions, allowing a qualitative correlationbetween the measured point data and variations inthe snow and weather conditions over the area.

In many instances today, the forecast area hasexpanded, largely due to increased putlic use ofavalanche-prone terrain (e.g. increased backcountryskiing in developed areas, large use areas forhelicopter skiing operations, or a regionalavalanche forecasting center). The ultimate effectof this expanded area of concern is less directcontact with conditions by forecasters. This hasresulted in greater reliance on both data gatheredby instruments and on the extrapolation of thesedata based on physical principles rather than directsubjective experience.

In this paper, several basic problemsassociated with this increased dependence oninstrument measurement and its interpretation areexamined. Specifically, errors in the measurement ofprecipitation, wind, and air temperature introducedby sensor site selection are considered, as well as,limitations on the sensors' responses to theenvironment. Errors introduced by poor equipmentmaintenance, line noise, and calibration problems,although frequently serious, will not be considered.

lAvalanche-Meteorologist, NorthwestAValanche Center, 7600 Sandpoint Way NE, BoxC-15700, Seattle, Wa. 98115.

SOURCES OF ERROR

Errors which arise in instrumented snow andweather measurements can be broken into two, ifsomewhat overlapping, parts: those associated withthe representativeness of the site where themeasurements are to be taken, and those associatedwith the response of the instrument to itsenvironment.

The first source of error is associated withthe site chosen for measurements. The topography ofmountains results in dramatic variations inconditions over short distances and often timesthese variations are not easily predictable. Forexample, temperature, which may often beextrapolated to other elevations using approximatelapse rates, may on some occasions be complicated byinversions generated by mesoscale or synoptic scaleweather conditions, undetectable from a valley site.Thus measurements must be taken at a site or sitesthat provide information that is unambiguousregardless of the weather conditions or they must betaken at enough sites that sufficient information isavailable to sort out any ambiguities that mightexist.

The second source of errors is caused by thewide variation in sensors available to measure eachparameter. Each type of sensor has a different typeof response to the same environmental conditionswhich can result in markedly differing readings atthe same location. Often times instruments arechosen without consideration of their differingtraits, resulting in frustration and/or confusion ininterpreting the data.

Finally, all of the above is furthercomplicated by the fact that each of t~e majorweather parameters (precipitation, wind, andtemperature) must be combined to provide meaningfulinformation on snowpack stability. As the best sitefor one type of measurement may not be the best foranother, this results in the merging of data fromseveral different areas and environments. Thus

144

40

1 Stampede Pass(J

'-..... - - - - - - - - Snoqualmie Pass~ 30Cll....C;;>

A...~ I \0-

Greatestkl 20 I \(.. J \Cll \.....c~

10

Noy Dec Jan Feb Mar Apr

Figure l.--Comparison of weekly precipitation totals inwater equivalent between Stampede and Snoqualmie Passes,Washington, 1931-1965. Data are from ClimatologicalHandbook, Columbia Basin States, Precipitation, Vol. 2,septerrber 1969.

errors introduced by either poor or unrepresentativesite selection or instrument peculiarities can beadditive, further confusing snow stability analysis.This further emphasizes the importance ofknowledgable selection of both measuring sites andinstruments.

PRECIPITATION

Site Selection

The primary information desired fromprecipitation data are the amount and rate of loadingof the snowpack and the density of new snow. It iswell accepted that the areal variation of thesequantities is affected by the interaction of windwith the topography. On the small scale (e.g.meters to kilometers) this is by wind scouring anddeposition of snow and associated crystal breakage,while on the large scale (kilometers to thousands ofkilometers) it is caused primarily by topographi­cally forced lifting and altitudinal effects ontemperature.

Concerns with these effects depend on the sizeof the forecasting area. For an area the size ofmost developed ski operations «lOkm2 ) thisonly requires consideration of the immediate terrainaround the sampling site. On this scale, theassumption can be made that an approximately equalamount of precipitation falls over the area, but issubsequently redistributed by wind interacting withthe terrain. Determination of snow loading for a

specific avalanche starting zone requiresestablishing a proportionality between the amountsreceived at a sensor site and that at the site inquestion. This proportion will be affected by windsat the starting zones, which may bear littleresemblance to the winds at the measuring site (seebelow). Thus in order to be accurate under allconditions, measurements should be made at a sitewhich is sufficiently protected to receive snowindependent of wind speed or direction. The idealsite is usually protected by a combination oftopographic features and local vegetation Marriott(1984) Although it is possible to use data from lesssuitable sites, this requires estimating themagnitude of the effects of the wind at themeasuring site and adds more uncertainty to thedata.

If the area of concern is greater thanl02-3km2, variations due to orographicallyinduced lifting must be considered in selectingmeasuring sites. Many general variations inprecipitation can be estimated from climatologicalinformation (fig. 1) and/or simple orographicprecipitation models. However, often, mesoscaleeffects of topography on the synoptic scale air flowmay produce mesoscale effects which become verysensitive to small changes in the synoptic scalewind patterns undetectable by current measurements.

An example of this is shown in figure 2, whichshows the differences in precipitation betweenParadise at 2599m (on the south side of Mt.Rainier) and Crystal Mountain located at 2079m aboutBkm to the northwest. Synoptic scale winds

145

2

1 1'-..E \1~ 0,00

fIQ

t -t..~

-2

3

-88 12 16 20 24 28

DATE~-Jla.rch

Figure 2.--Comparison of daily (12OT and our) 850-rob free airwind direction and speed from Quillayute, Washingtonversus daily precipitation differences between Paradise(Mt Rainier) and Crystal Mountain, Washington. Windsare plotted 0-360 degrees and rounded to the nearest 5m/sec, and water equivalents (D) indicate Paradiseminus Crystal Mt. data.

interacting with Mt. Rainier and the Cascade Creststrongly affect the mesoscale effects ofrainshadbwing and convergence in the area around Mt.Rainier. As can be seen from figure 2, there islittle correlation between the measured synopticscale winds (taken at the radiosonde station nearQuillayute, vlashington) and precipitationdifferences between the two stations. This showsthat measurements of synoptic scale winds are tooinfrequent and too sparse to infer the location andmagnitude of this type of effects: Detection of thistype of mesoscale effect which is sensitive tosynoptic scale winds can only be found by using a"dense" grid of stations or potentially through theuse of realistic orographic precipitation models(Speers-Hayes 1984).

Sensor Errors

In snow stability analysis, precipitation datais largely used to give an indication of the amountand rate of loading of avalanche starting zones.Historically, this has been accomplished by using asnowboard: measuring the depth of new snow, taking asnow core from the board, and subsequently weighingor melting the sample to obtain the waterequivalent. Increasingly, snowboard measurementshave been supplemented or replaced by recordingprecipitation instruments, almost exclusivelymeasuring water equivalent. A general review of thetypes of sensors in current use and their operationis given in Marriott and Moore (1984).

146

All of the current methods of water equival~nt

medsurement are subject to errors under certainconditions. In some instances these errors can bedetected and allowed for, however, this is oftennot the case, unless information from more than onetype of sensor is available fr~n the site.

Gauge Sensors

The most widely used sensors for waterequivalent measurements are gauge type devices.However, these sensors suffer from a number ofi~herent problems including missed catch (blowover),capping (for unheated gauges) and evaporation (forheated gauges).

Missed Catch.--Wind effects on precipitationmeasuring sites can introduce serious errors, andthis is particuarly true if the measurements arebeing taken using a gauge type device. Work byLarson and Peck (1974) and Goodison (1978) haveshown that wind effects can introduce substantialerrors in gauge measurements. The magnitude of theerrors is related to three factors: wind speed, windspeed vertical profile, and the obstruction to theair flow presented by the gauge. Figure 3 preparedby Larson and Peck (1974) displays the catchdeficiency compared to snowboard measurements as afunction of wind speed. This shows that evenmoderate winds at a measuring site can causesubstantial errors in gauge measurements. However,figure 3 and additional work by Goodison (1977) haveshown that good site selection and proper shieldingcan reduce or eliminate errors in the measurementdue to missed catch. It is possible to develop acorrection factor for the lost catch (Larson andPeck 1974), however, this requires information onwind conditions during precipitation, complicating·iata extraction whil~ still producing questionabledata.

Wind speed (m/s)

2 3 4 6PO

0.1

0.2>-uc 03Q)

:~~

0.4"0.£Uiii 0.5uQ)Cl:> 0.6til

C)

0.7Unshielded gauge

08

0.9

Fi~ure 3.--Mean gauge catch deficiency of shieldedand unshielded United States gauges as afunction of wind speed (Larson and Peck 1974).

Gauge Capping.-- Often during sustained moderateto heavy snowfalls, unheated gauges will accumulatesnow along the rim of the collection cylinder,especially for orifices of 30cm or less. As thisaccuIDJlation grows, the effective orifice sizedecreases, reducing the measured precipitation,sometimes to zero when complete capping occurs.Additionally, following a capping episode, warmingtemperatures usually result in the eventual meltingof the accuillJlated snow into the gauge, oftenresulting in an overestimate of the currentprecipitation amounts or indicating the occurrenceof precipitation when none is occurring.

Figure 4 shows the typical results ofmeasurements taken at the same location from both aheated and an unheated precipitation gauge duringJanuary 1984. The unheated gauge SUCCessfullymeasured light snowfall on the 20th (confirmed bysnowboard measurements). However, increasingsnowfall at cold temperatures on the 21st thru the23rd capped the orifice of the unheated gaugestopping almost all indication of precipitationduring that time, although 7 to 9 cm of waterequivalent were indicated by the heated gauge andthe snowboard. On the 24th, warming temperaturesaccompanied by rain caused the snowcap to meltcausing an overestimate of the precipitation on the24th. Obviously, any snow stability analysisprepared using the unheated gauge would be inserious error. This type of error can make unheatedgauges virtually useless for measurements ofmoderate to heavy snowfalls at temperatures belowfreezing.

Evaporation.--As can be seen in figure 4 onthe 20th and 26th and 27th, the unheated gaugeindicated more precipitation than the heated gauge.This effect can be attributed to evaporation withinthe heated gauge. The magnitude of this effect hasnot yet been quantified, however, the effects appearto be largest for dry snow at low snowfall rates,especially with propane heated precipitation gauges.

Almost all heated precipitation gauges usetipping bucket mechanisms, which require that thesnow must first melt and then drop into a movablebucket inside the heated gauge. This can result inevaporation at the external melting surface and, toa greater extent, within the gauge itself.

At low snowfall rates, thG accumulating liquid,3 exposed to the internal environment of the gaugefor a relatively long period of time. As the gaugeis heated above the ambient air temperature,resulting humidities inside the gauge are low. Forexample, an ambient air temperature of -lOoCand 100% relative humidity will result in aninternal relative humidity of 23% if the internaltemperature is kept at +100 C. Thus theenvironment inside of the gauge is quite dry and canlead to substantial evaporation. In the case of drycontinental type snowfalls, the outside relativehumidities can be well below 100% resulting in evendrier conditions and more evaporation.

This effect is seen in figure 5 which shows thenifference between an electrically heatedprecipitation gauge and snowboard water equivalents

147

12

11 1\! \

10 I \

f9 /~\8 ~ \\

'- I \\i .,l Ii'; ! "

j 6iJl ~

QI 5 ,; \...t 4... I \~

3I ;I

2 II

'~1 J0

18 20 22 24- 26 28

Da.te--JANUARY

Figure 4.--Comparison of daily (24-hr) precipitation totalsfor Stevens Pass, Washington, during February, 1984.Water equivalent data are derived from electricallyheated tipping bucket precipitation gauge (D) andunheated precipi tation storage gauge (+) measurements.

at Stevens Pass, Washington. Although Stevens Passhas a strongly maritime climate, it is obvious thatat low precipitation rates, the heated precipitationgauge reads routinely lower than the snowboard.Although these differences may not always besignificant, at unmanned locations, where theSeerrors can be cumulative, this can lead to seriousmistakes in judging snowpack stability.

The evaporation problem is particuarly acute forgauges heated by propane as opposed to electricity.In electrically heated gauges a thermostat isusually provided allowing a certain internaltemperature to be maintained independent of thecooling by the external enviroment. Thus a lowtemperature can be selected (5-100 C) minimizingthe evaporation. In the case of most propane gaugeson the market today, a set rise in temperature overthe ambient temperaturel must be selected as the

lRecently a new propane heated precipitationgauge was introduced which uses a simple thermostatto maintain the gauge funnel at or above 150 C,however, no information on the evaporative effectsof this gauge are available.

amount of heat delivered to the gauge is constantregardless of external changes. This requires thata large enough rise over ambient be selected to keepthe gauge from dropping below freezing at lowtemperatures or capping over at high precipitationrates, while still not being so warm as to causelarge amounts of evaporation at low precipi tationrates. TJ:lis necessary compromise normally resultsin a loss of accuracy at both extremes.

Snow Pillows

The only non-gauge type water equivalent sensorin use on a wide scale is the snow pillow. Thisdevice essentially measures the weight of theoverlying snowpack which is then converted directlyto the water content of the snowpack. Taking thedifference between successive readings can then giveinformation on the water equivalent of precipitationfalling on the surface of the snowpack. Althoughthese devices have proven successful for hydrologi­cal purposes (Bartee 1978), the snow pillow suffersfrom several inherent problems which may yieldinaccurate or misleading measurements (Beaumont1965) and limits their resolution especially forshort term changes (i.e. one day or less). Figure 6

148

292521

//\\ I~/ i'

f \V'17

,

\

13

t

95

2.8

2.6

2.4

2.2

1 2CJ'- 1.8~

~ 1.6Cll....tS~ 1.4...~0- 1.2~

lo- 1Cll~

tS 0.8i:

0.6

0.4

0.2

0

1

Date--FEBRUARY

Figure 5.-~omparison of daily water equivalent amountsbetween 24-hr snow board (0) and electrically heatedtipping bucket precipi tation gauge (+) measurements atStevens Pass, Washington, during February, 1984.

illustrates the significant short term differencesthat can exist between snowboard measurements and aSnotel (Barton and Burke 1977) snow pillow locatedat the same site at Stevens Pass, Washington. It isobvious that the snow pillow can read signficantlyabove and below the measurements taken from asnowboard. These differences are associated withthe layered nature of the snowpack and the complexway it is supported by the underlying ground.

The interpretation of the snow pillowdata would be easier if it measured consistentlygreater or less than the snow board, however, due tothe differing sources of error it can vary eitherway. In deep snowpacks, the response time of snowpillows to heavy snowfalls can be slow, ranging from5 hours to as much as 10 days (Tarble 1968).Warming or cooling of the snowpack may result inerroneous indications of increasing or decreasingwater =ntent, respectively. Formation of crustsand ice lenses within the snowpack may result inbridging of the pillow, giving erroneously lowvalues of water equivalent. In addition, in shallowsnowpacks diurnal variations in the temperature ofthe snow pillow itself may give erroneous values.

The net effect of these combined errors is tolimit the accuracy of short term changes measured by

a snow pillow. Thus, although they may be used toestimate the magnitude of precipitation events, eventhis use is difficult unless information on localtemperature changes and/or other independentindications of precipitation are available.Despite their lack of short term reliability, thesnow pillow's widespread distribution throughout themountains for hydrological purposes makes them auseful source of information on precipitation, butinformation that can only be used to substantiateinformation from other types of nearby precipitationsensors.

Summary

Thus selection of representative Bites forprecipitation measurement depends partly on the sizeof the area of concern and partly on the complexityof that area. In all events, the site should beprotected from wind effects either by the localtopography, vegetation, and/or artificial shieldingso that the precipitation measured at the site isindependent of local wind speed or direction. Inaddi tion, for larger forecast areas, a sufficentnurrber of sites must be measured to detect mesoscalevariations in orographically induced precipitationcaused by effects such as channelling orconvergence.

149

21.91.81.71.6

11.61.4-

'- 1.3.. 1.2E'0 1.1

-i 1, 0.9.. 0.8t 0.7..~ 0.6

0.60.4-0.30.20.1

018 20 22 24- 26 28

Da.te--FEBRUARY

Figure 6.--Comparison of daily water equivalent amountsbetween 24-hr snow board ([J) and 24-hr snow pillow (+)data at Stevens Pass, Washington, during February, 1984.

Additionally, all of the available. methods ofmeasuring water equivalent produce significanterrors under certain conditions. Realistically,most measurement sites suffer from some type ofdeficiency which may cause these errors to becomemanifest.' The practical solution to this appears tobe the use of one or more types of sensors at asite. This may not be simply another precipitationsensor, but perhaps information on temperature orwind. With several pieces of information, it maybecome possible to recognize and correct errors whent~y occur.

WIND

In snow stability forecasting, wind informationis primarily important for estimating the degree andnature of snow depostion over the topography ofconcern. Synoptic scale free air winds interactingwith the mountain topography create a large varietyof local winds depending on the orientation of thefree air wind to the terrain. However, this problemis often further complicated by winds driven bymesoscale pressure differences across a mountainrange, small scale channeling effects, and/ordrainage winds.

Site Selection

Selection of sites for wind measurementrequires evaluating the potential effects of all oft~ above winds on the avalanche starting zones ofinterest. The main consideration in choosing alocation(s) for measuring winds is finding a sitewhich gives wind information from which startingzone winds may be inferred.

For small scale forecasting concerns«lOkmf) this often means locating the wind systemclose to the main starting zones minimizing theamount of inference necessary. This is particuarlyeffective for areas whose starting zones are atapproximately the same elevation with the sameaspect. A wind system which gives accurate resultsfor directicos which would tend to load these slopesis usually sufficient. However, for areas with avariety of starting zone elevations and aspects, itis necessary to choose a site which is likely togive a reasonable estimate of the localfree air wind speed and direction. Adding localtopography to this information then allowsestimating the winds for any aspect, but notnecessarily any elevation (see below). A site whichcould satisfy this criteria would be an isloated,symmetrical peak. More often wind sites are located

150

on ridgelines or on peaks located along ridgelineswhich then provide a varying response depending onwind direction. This type of variation can beroughly accounted for, except in the extreme casewhere the ridgeline or a local obstacle blocks windfrom certain directions. In some cases this mayrequire more than one measuring location.

Small scale wind channeling, drainage winds, orwinds wi thin inversions may produce ·dramaticvariations in winds at different elevations notderivable from measurements at any singleelevations. Figure 7 illustrates the types ofvariation that can occur in a complex wind-terrainsituation. In this example, winds at Denny Mountain(1696m), Stampede Pass (1204m), and Snoqualmie Pass(llS8m) are shown. Snoqualmie Pass and Stampede Passare loca ted about 1. Skm and 8km southeast of DennyMountain, respectively. The dramatic differencesbetween wims measured at the three sites isobvious, both in directions and speeds.Importantly, notice that there is little consistencyin the differences between the stations. Thussignificant variations in winds which may be loadingavalanche stating zones are not. easily inferablefrOm the measurements taken at anyone of thesesites. Although this is an extreme case, itemphasizes the potential for major differences inthe wims which are loading starting zones in arelatively small area.

For snow stability analysis for larger scaleareas (>lOl-3km2 ) of concern, it is necessaryto measure wims which are representative of thela~est area possible. Since local effects have sucha strong influence on measured winds, site selectionbecomes critical. This usually requires placing thewind site at a location which measures the mesoscalefree air winds. Although this often requiressacrificing important information on local effects,it is required by the need to apply the informationto areas beyom the immediate vicinity of the windsite. In this case, it is usually necessary to infersignificant local effects from experientialknowledge of local behavior (when available) with acom"mensurate increase in possible errors.

Sensor Errors

Although a variety of wind sensors arecommerically available (Marriott and Moore 1984),the most commonly used are variations on the cupanemometer and wind vane or the single unit "bird"which uses a plane-like body with a F~opellor and atail. The only serious problem with wind sensorsthemselves occurs during periods of rime formation.In light riming situations heat lamp deriming hasshown some effectiveness, however, moderate to heavyriming areas require the use of a conduction heatedanemometer (Taylor 1984). The only alternative tothis is to locate the anemometer downwind of aterrain feature which may remove most of thesupercooled water droplets minimizing the riming ofthe instrument, but at the cost of lessrepresentative measurements. Some wind vane designsalso suffer from riming problems requiring heating,however, it appears that wind vanes with astreamlined profile and a mean length to width ratioof about 4:1 or higher, maintain enough of theirinitial shape to remain oriented into the wind under

~ll but the most extended heavy riming conditions.

The most difficult interpretive problemassociated with rime formation occurs when ananemometer is only partially rimed. In total r1m1ngsituations, it is frequently obvious that zero windspeed is not physically reasonable and that theanemometer is rimed up. However, in partial rimingsituations, the anemometer may continue to turn, butat a reduced speed, giving low wino Ej)eed readingsthat may mislead the forecaster, leading to a largeunderestimate of wind transport of snow. Similarpr:dJlems of interpretation occur when the wind vanerimes, especially if the anemometer melts out first.

Summary

The selection of wind sensor sites must bedetermined based on the size and complexi ty of thearea of concern. For forecasting situationsinvolving small areas «lOkm2 ) it is frequentlypossible to select a representative measuring sitewhich can be used to infer winds affecting allstarting zones of interest. However, in situationswhere wim-terrain interaction is complex, severalsites may be necessary to achieve the requiredresolution. Sites which are used to estimatecomitions over a large area (> lOl-3km2 ) mustbe placed in positions which are as free as possiblefrom the small scale local effects, accepting aconsequent loss in resolution of starting zonewinds. Errors associated with the sensormeasurements themselves are primarily limited torime deposition on the sensors, inhibiting theirmotions, and producing false readings. Rimingproblems can be overcome by heating the sensors or,in some cases, by locating the sensors at arelatively rime free site.

AIR TEMPERATURE

Air temperature measurements supply significantinformation to snow stability forecasters, allowingjudgements to be made about such things as: thelikely types and rates of metamor:phism in thesnowpack, crystal types and densities of snowfalls,surface melt and crust formation, and occurrence ofrainfall, etc., all of which relate intimately tothe strength of the snowpack.

Site Selection

Although temperatures in the mountains arela~ely controlled by a combination of free airfreezing levels and diurnal variations, thesefactors are sometimes overpowered by more localizedeffects such as terrain induced inversions.Temperature variations caused by topographic effectsmay act on the scale of a few meters and be of onlyminor significance or, in some cases, may operate onthe scale of hundreds of kilometers.

A single air temperature sensor does notprovide sufficient temperature data for either localor regional snow stabili ty analysis, particuarly inthe the case of inversions. Inversions of lOoe to200e between the top and bottom of a ski area arenot unusual, am in some instances, may bemaintained for days. A single measurement taken at

S60

SOO

~

! 260

~e 200

~ 160q

~~ 100

---- - yariab1e - - ---

50

0

151

40

so

20

10

2 6 10

TillE (PST)

14 18

Figure 7.--Comparison of hourly wind speed and direction datafrom three different elevations in the WashingtonCascades. Data are derived from wind sensors atsnoqualmie Pass (1158m, D), Stampede Pass (1204m, +) andDenny Mountain (1696m, <», Decerrber 30, 1983.

152

either the top or the bottom of the area could bemisleading, however, measurements at both can showthe existence and the magnitude of the inversion.Measurements at intermediate levels may be ofinterest when the top of the inversion lies withinthe elevation range of the forecast area, as thiswill help to locate the top of the inversion, whichis often a relatively, sharp boundary and can leadto rapid localized temperature variations.

Inversions are especially important fortemperatures near freezing as variations intemperature in this range can have strong andimmediate effects on snowpack stability. However,even at temperatures well below freezing inversionsmay have significant effects on snowpackmetamorphism and the formation of surface hoar.

For snow stability forecastin~ for areas on thescale of less than about 10km , temperaturesshould be measured near the elevation of thestarting zones of concern as a minimum, or in thecase of starting zones at multiple elevations,temperatures should be taken both at the highest andlowest elevations. Temperature measurements atadditional elevations will improve the verticalresolution of temperature but are probably onlyessential in areas where inversions at intermediate

levels may exist for extended periods.

Long lasting inversions (hours to days) canhave a significant effect on the timing of warminginduced avalanches. Figure 8 illustrates thecomplexity that can result within n eingle ski area.The time variation of temperature sensors located at914m, 1341m, and 1646m in the Alpental ski area areshown. Initially, a relatively normal lapse rateexists with temperatures cooling with increasingelevation. At about 1300 PST, warming at higherelevations produces an inversion somewhere betweenthe stations at l341m and 1646m, although, a normallapse rate continues to exist between the two lowersensors. Around 2100 PST the sensor at 1341m beginsto warm as the top of the inversion graduallylowers. Finally around 0100-0200 PST, the inversiondrops completely below the middle sensor, while anormal lapse rate develops between the two uppersensors. In this example it is obvious that one oreven two temperature sensors in this situation couldresult in an erroneous impression of thetemperatures in the area and consequently could leadto a poor estimate of avalanche stability, iftemperature sensors were not located near theelevation of the starting zones. Thus for stabilityanalysis , the vertical temperature resolutionrequired will depend on the distribution of

96

\\ f\" I '\.

'¥" ~~<S!~~'r-;&-~~

1

f!f

17 21

TillE (PST)

139

o

6

1

0

-1

-2

~-3

CIt -4-

! -6

= -6

~ -7

a -8

~ -9

~ -10

-11

-12

-13

-14-1

Figure 8.--Hourly air temperature at three elevations-­914m (0), 1204m (+), and 1646m (¢)--at Alpental skiarea near Snoqualmie Pass, Washington, December 29-30,1983.

153

avalanche starting zones with elevation and to someextent on the longevity of inversions in the area,which may be controlled by local or regionaltopography.

On the scale of about lOl-3km2, smallerscale inversion effects such as valleys or basin areusually ignored wi th temperatures from the tops ofpeak or ridges and from balloon soundings being morerepresentative of the temperature near most majorstarting zones. However, additional measurementsmay be necessary in situations where geographicalconditions produce inversions over wide areas. Inthis case additional measurements on carefullyselected ridges and in valleys can give importantinformation on the depth and trends of this cold airwhich may affect snow stability.

Sensor Errors

Numerous types of temperature sensorS are incurrent use (Marriott and Moore 1984), although themost popular recording temperature sensor continuesto be the thermistor. The single biggestcontributor to errors in air temperaturemeasurements, aside from poor line maintenance, islack of shielding from heat and radiation sources.Locating sensors close to heated buildings or indirect sunlight without proper radiation shieldingproduces erors in measurement which change duringthe day as the temperature inside the building orthe sun exposure changes, making observationsdifficult to interpret and correct. Positioning ofsensors on trees or towers a substantial distancefrom any heat sources and the use of simple doublewalled radiation shields for the probe can eliminatemoo t measurement errors.

In some poorly placed or exposed sites, it isp?ssible that the temperature sensors may becomenmed or encrusted in ice or snw. This ordinarilyresu~ts in a slwer response to temperature changes,and ~n the case of temperatures fluctuating aroundfreezing, the sensors will continue to read OoCuntil all of the ice or snw melts from the sensor.Locating the sensor at a site reasonably protectedfrom direct wind should minimize or eliminate theproolem.

Summary

Selection of temperature sensor location de­pends both on the scale of the area under consid­eration and on the complexity of the associatedterrain. Snow stability analysis for a small area(~lOkm2) may be accomplished with a singletemperature sensor if inversions tend to be weak orshort lived. However, if inversions are strong andpersistent, it may be necessary to measure thetemperat~re near the elevation of all major groupsof start~ng zones to obtain sufficient verticaltemperature resolution.

For larger scale areas of concern (lOl-3km2),sensors should be located on ridges or at sitesunaffected by local temperature effects such asdrainage winds or trapped cold air, allowingextrapolation of the information over a wide area.~hen geographical conditions can produce regional~nversions, however, it is desirable to obtain

measurements from several elevations to determinethe depth and trend of the inversion.

CONCLUSIONS

Assuming proper sensor maintenance andcalibration and a reliable telemetry a;,d datarecording system, errors in measurement andinterpretation of precipitation, wind, andtemperature quantitites arise from two primarysources: lJimproper or inadequate site selection(s)for the measurements and 2) inconsistent response ofsensors to varying weather conditions. Althoughthese deficiencies have always existed, increasingreliance an instrumentation, especially at remote,unmanned sites has made recognition and, wherepossible, correction of the problems moreimperative. Although the optimal situation for snowstability analysis remains a combination of sensorsand direct human observation, where this is notpossible or impractical, sensors can provide usefulinformation, provided a suitable combination ofproper sensors and sites and appropriate sitedensity are achieved for the intended forecast area.

In those instances where sensor data alone mustbe relied upon, the best solution appears to be anurrber of measuring sites with several types ofsensors at each site. This type of i~put oftenhelps to resolve potential arrbiguities produced byindividual sites or sensors. The use of multiplesensors, gathering information on the same parameteror on several different parameters (e.g.precipitation and temperature) at each site canelucidate effects which are artifacts of thesensors. Similarly an increase in the number ofmeasuring sites can overcome deficiencies atindividual sites and provide information onmesoscale weather features.

Practically, the optimal data system for snowstability analysis and forecasting varies markedlyfrom area to area, depending on the scale of theoperation and upon the peculiari tes, bothmeteorological and topographical, of the specificarea. Although basic considerations outlined inthis paper can aid in initial site and sensorselections, experience with an area and a datasystem are necessary to produce an optimal system.

ACKNOWLEDGEMENTS

The authors wish to extend their appreciationto the many observers who contributed data which wasused to develop this paper, only a small fraction ofwhich is shown. Special thanks to the WashingtonDepartment of Transportation Avalanche Control Crewat Stevens Pass, the Stevens Pass Pro Patrol, theSoil Conservation Service, the Crystal Mountain ProPatrol, the Snow Rangers at Stevens Pass and atParadise on Mt. Rainier, and the National WeatherService Forecast Office in Seattle. for their extraassistance.

154

LITERATURE CITED

Bartee, D.L. 1978. Snow sensor evaluation report.USDA Soil Conservation Service Special Report.West Technical Service Center, Portland,Oregon.

Barton, M. and M. Burke. 1977. SNOTEL: Anoperational data acquisiton system using meteorburst technology. Proceedings 45th AnnualMeeting Western Snow Conference, pp. 82-87.

Beaumont, R. T. 1965. Mt Hood pressure pillow snowgage and application to forecasting avalanchehazard. lASH Publication No. 69, pp. 341-349.International Symposium on Aspects of Snow andIce Avalanches, Davos, Switzerland

Goodison, B.E. 1977. Snowfall and snowcover insouthern Ontario: principles and techniques ofassessment. Ph.D. thesis, University ofToronto, Toronto, Ontario.

Goodison, B.E. 1978. Accuracy of Canadian snowgage measurements. Journal of AppliedMeteorology, Vol. 27, pp. 1542-1548.

Larson, L.W. and E.L. Peck. 1974. A~curacy ofprecipitation measurements for hydrologicalmodeling. Water Resources Res., Vol. 10, pp.857-863.

Marriott, R.T. 1984. Snow study sites. AvalancheReview, Vol. 2, No.4, pp.4.

Marriott, R.T. and M. B. Moore. 1984. Recentadvances in the collection and transmission ofmountain weather and snowpack data foravalanche forecasting. Northwest AvalancheCenter Technical Memorandum No.5, Seattle,Washington.

Speers Hayes, Pamela. 1984. Diagnosis ofprecipitation in mountainous terrain: part I,rain shadow effects. Proceedings ofInternational Snow Science Workshop, Aspen,Colorado.

Tarble, R.D. 1968. California federal-state snowsensor investigations, problems, and rewards.Proceedings 36th Annual Meeting Western SnowConference, pp. 106-109.

Taylor, P.L. 1984. Measurement problems inmountain weather. American W<'\te::- ResourcesAssociation Symposium, Seattle, Washington.