ARTICLE IN PRESS

0360-1323/$ - se

doi:10.1016/j.bu

�Correspond124, Blindern, N

E-mail addr

Building and Environment 42 (2007) 3726–3736

www.elsevier.com/locate/buildenv

Effects of wind exposure on roof snow loads

Vivian Meløysunda,b,�, Kim Robert Lisøa,c, Hans Olav Hygend,Karl V. Høisethb, Bernt Leirae

aSINTEF Building and Infrastructure, P.O. Box 124, Blindern, NO-0314 Oslo, NorwaybNorwegian University of Science and Technology (NTNU), Department of Structural Engineering, NO-7491 Trondheim, NorwaycNorwegian University of Science and Technology, Department of Civil and Transport Engineering, NO-7491 Trondheim, Norway

dNorwegian Meteorological Institute, P.O. Box 43, Blindern, NO-0313 Oslo, NorwayeNorwegian University of Science and Technology, Department of Marine Technology, NO-7491 Trondheim, Norway

Received 20 June 2006; received in revised form 22 August 2006; accepted 26 September 2006

Abstract

This paper presents results from an investigation of the suitability of the exposure coefficient as defined in ISO 4355 ‘‘Bases for design

of structures—Determination of snow loads on roofs’’, based on thorough analyses of weather data from 389 weather stations in Norway

for the reference 30-year period 1961–1990. First, the background of the exposure coefficient is examined. Historical field investigations

of snow loads on roofs are also evaluated. Next, values for the exposure coefficients in Norway are calculated according to ISO 4355.

Finally, possible approaches aiming at improving calculations of wind exposure on roof snow loads are suggested. It is shown that the

exposure coefficient as defined in ISO 4355 does not reflect the actual effects of wind exposure on roof snow loads in Norway, the main

reasons being oversimplifications in the definition of the coefficient and the extreme variations of the climate in Norway. The definition is

based on coarse simplifications of snow transport theories, and must be revised and improved to serve as an applicable tool for

calculations of design snow loads on roofs in Norway.

r 2006 Elsevier Ltd. All rights reserved.

Keywords: Buildings; Roofs; Snow; Snow loads; Structural design; Wind loads

1. Introduction

In the current Norwegian snow load standard NS 3491-3‘‘Design of structures—Design actions—Part 3: Snowloads’’ [1] snow loads on roofs are defined as

s ¼ mCeCts0, (1)

where s0 is snow loads on the ground. The parameters m, Ce

and Ct describe conditions on the roof. The exposurecoefficient Ce takes into account that wind removes snowfrom flat roofs. Using this coefficient the snow load on asheltered roof becomes twice as large as the snow load on awindswept roof. The shape coefficient m describes thedistribution of snow load on the roof due to geometry. Thethermal coefficient Ct defines the reduction of the snow

e front matter r 2006 Elsevier Ltd. All rights reserved.

ildenv.2006.09.005

ing author. SINTEF Building and Infrastructure, P.O. Box

O-0314 Oslo, Norway. Fax: +47 22 69 94 38.

ess: vivian.meloysund@sintef.no (V. Meløysund).

load on the roof as a function of the heat flux through theroof. An equivalent expression can be found in ISO 4355‘‘Bases for design of structures—Determination of snowloads on roofs’’ [2].In practice, it has turned out difficult for consultants in

structural engineering to determine the exposure coefficientCe. The main reason is the meteorological input needed.According to an informative annex in ISO 4355 and NS3491-3, the exposure coefficient is a function of the meantemperature, y, in the coldest winter month and number ofdays, N, with a wind velocity above 10m/s where N isdefined as an average for the three coldest months of theyear (see Table 1). Mean values for ‘‘many years’’ arerecommended (usually 30 years). This meteorologicalinformation is available merely at advanced weatherstations. If a building site happens to be located near sucha station, the data needed is still not easily accessible.In this paper weather data from meteorological stations

in Norway for the reference 30-year period 1961–1990 is

ARTICLE IN PRESS

Table 1

Exposure coefficient according to ISO 4355

Winter temperature Winter wind category

Category Mean temp. (1C) I (No1) II (1pNp10) III (N410)

A y42.5 1.0 1.0 0.8

B �2.5pyp2.5 1.0 0.8 0.6

C yo�2.5 0.8 0.8 0.5

Fig. 1. Balanced snow load sb and drift snow load sd on pitched roofs.

1.0

0.8

0.6

0.4

0.2

0.0

0 10 20 30

Roof slope (deg.)

40 50 60

µ b

Fig. 2. Slope reduction coefficient mb for simple pitched roofs with non-

slippery surface.

V. Meløysund et al. / Building and Environment 42 (2007) 3726–3736 3727

used to determine the exposure coefficient Ce according tothe definition in ISO 4355. First, historical field investiga-tions studying snow loads on roofs are evaluated giving thebackground of the exposure coefficient. Next, values forthe exposure coefficients are calculated for 389 meteor-ological stations, and the suitability of the definition inorder to describe the effects of wind exposure is discussed.Finally, possible approaches aiming at improving calcula-tions of wind exposure on roof snow loads are suggested.

2. Background

2.1. Snow load on roofs according to ISO 4355

In ISO 4355 ‘‘Bases for design of structures—Determi-nation of snow loads on roofs’’ [2] the snow load on theroof is defined as the sum of a balanced load sb, a drift loadpart sd and a slide load part ss (see Fig. 1):

s ¼ sb þ sd þ ss. (2)

The balanced load sb is uniformly distributed on the roof(except for curved roofs) and a function of characteristicsnow load on the ground s0, exposure coefficient Ce,thermal coefficient Ct and slope reduction coefficient mb:

sb ¼ s0CeCtmb. (3)

The slope reduction coefficient, mb, defines the reductionof the snow on the roof due to roof slope and surfacematerial. High slopes and smooth surface materials makethe snow slide from the roof. In Fig. 2 slope reductioncoefficient mb is shown for a single pitched roof with non-slippery surface.

The thermal coefficient Ct defines the reduction of thesnow load on the roof as a function of the heat fluxthrough the roof, causing snow melting.

The exposure coefficient Ce defines the balanced load ona flat horizontal roof of a cold building, as a fraction of thecharacteristic snow load on the ground. The coefficientincludes the effect of snow being removed from flat roofsby wind. According to an informative annex in ISO 4355 itis a function of the mean temperature, y, in the coldestwinter month and number of days, N, with a wind velocityabove 10m/s (N is an average for the three coldest monthsof the year, see Table 1).

In addition to the balanced load sb, a drift load part sdhas to be included in order to take into account snow

accumulation on leeward side of the roof due to drifting.The drift load sd is a function of characteristic snow loadon the ground s0, exposure coefficient Ce, thermalcoefficient Ct, slope reduction coefficient mb and drift loadcoefficient md:

sd ¼ s0CeCtmbmd. (4)

The drift load coefficient md multiplies with mb anddefines the amount and distribution of additional load on aleeward side or part of a roof. The coefficient depends onwind exposure and geometry of the roof. In Fig. 3 themultiplication mb � md for various exposure coefficients Ce isshown for a single pitched roof with non-slippery surface.The slide load ss takes into account slide from an upper

roof onto a lower roof, or a lower part of a roof.

2.2. The historical background of the exposure coefficient

In the period 1956–1967, the National ResearchCouncil of Canada established more than 50 observationstations across Canada, where snow depth and density

ARTICLE IN PRESS

0.5

0.4

0.3

0.2

0.1

0.0

0 10 20 30

Roof Slope (deg.)

40 50 60

µ b*µ

dCe=1.0Ce=0.8Ce=0.6Ce=0.5

Fig. 3. Drifted shape coefficients mb � md for simple pitched roofs with non-

slippery surface.

V. Meløysund et al. / Building and Environment 42 (2007) 3726–37363728

measurements were registered once a week and immedi-ately following major snowfalls on the ground and for avariety of roofs and wind exposures [3]. The roofs wereboth flat and sloped and varied in size. The surveyconcluded that the roof-to-ground ratios depend primarilyon the degree to which a roof is sheltered from wind. Well-sheltered roofs had ratios up to approximately 0.9, whereasnearly all unobstructed roofs had ratios of less than 0.6.Well-exposed, unobstructed roofs in generally open areashad ratios of less than 0.3. Lutes [3] did not discuss whetherheat loss was included in these results.

Otstavnov and Rosenberg [4] presented a method whereaverage wind velocities for the whole winter season wereused in order to describe snow drifting from flat roofsduring snowfall. Experiments were performed on flat roofsand an expression for the roof-to-ground conversion factorwas developed (effect of heating was excluded). Inaddition, an expression was developed for the total amountof snow drifted from the roof between snowfalls. Thisamount was found to be a function of the average windvelocity between snowfalls and the summarized time thiskind of drifting occurs during the winter season. Otstavnovand Rosenberg [4] reported a close correlation betweenthe number of days with snowfall and the total groundsnow load.

In the period 1966–1986 professor Høibø at theAgricultural University of Norway measured snow depthsand densities on cold pitched roofs on approximately 200agricultural buildings in Norway [5]. The measurementswere performed at the assumed maximum seasonal roofload. A total of 1300 measurements were done. Measure-ments were also performed in ‘‘undisturbed’’ areas close tothe buildings. The effect of wind was not evaluated. Basedon his observations, Høibø proposed roof-to-groundconversion factors for both leeward and windward sideof the roof depending on roof angle and ground snow load.The magnitude of the ground snow load was found to

affect the conversion factor strongly. A ground snow loadequal to 1.0 kN/m2 gave a conversion factor of, respec-tively, 0.75 and 0.82 for windward and leeward side of theroof. A ground snow load equal 3.5 kN/m2 produced aconversion factor of, respectively, 0.48 and 0.62 forwindward and leeward side. These formulas were restrictedto buildings with ground snow load equal 3.5 kN/m2 or lessand buildings not fully sheltered.After heavy snowfalls in the winter of 1975–1976, snow

depths and densities were measured at 55 pitched roofs(roof angles 18–251) in Trondheim, Norway [6]. The snowwas measured at 6–12 points at the most heavily loadedroof side. In addition, snow load was measured on theground at 94 locations in the area of Trondheim. The roofswere anticipated to be cold roofs. The authors of this paperhave calculated an average roof-to-ground ratio of 0.27 forwindswept areas (based on the data from Løberg [6]). Forsheltered areas the average roof-to-ground ratio is calcu-lated to 0.55.Taylor [7] performed a survey of snow loads on roofs of

arena-type buildings in Canada. Data were collected for 32curved roofs and 16 gable roofs through a 4-year pilotstudy of snow on buildings, case histories and newspaperreports (snow events in the period 1956–1977). It wasconcluded that the maximum of the uniformly distributedloads for both gable and curved roofs, sheltered from wind,was approximately 80% of the specified 30-year returnground load. Five of these buildings were reported to beunheated. The effect of heat loss was not consideredseparately.In case studies performed by O’Rourke et al. [8], roof

and ground snow loads were measured for 199 buildings inNortheast, Midwest and Northwest USA, during threewinter seasons in the period 1975–1978. A total of 253roofs were measured. Conversion factors defined as theratio between the maximum roof load and the maximumground load were calculated. Areas with infrequent snow-falls and small accumulations were reported to have higherground-to-roof conversion factors than colder areas withsubstantial ground snow accumulation. Average conver-sion factors were calculated to be 0.78 for the shelteredroofs, 0.59 for the semi-sheltered roofs and 0.53 for thewindswept roofs when the effects of roof slope and heatingwere included. Based on this study, the average ground-to-roof conversion factors for unheated flat roofs arerecalculated by the authors of this paper to be 0.76 forthe sheltered roofs, 0.57 for the semi-sheltered roofs and0.55 for the windswept roofs.In the European Snow Load Program 1997–1999, roof

snow loads were measured for 55 pitched roofs and 26 flatroofs in Switzerland, Italy, Great Britain and Germany inthe winter season 1998–1999 [9]. The roof-to-ground ratiofor flat roofs was calculated to 0.90 for sheltered roofs, 0.74for semi windswept roofs and 0.58 for windswept roofs.When selecting the buildings for this project, unheated orvery high thermal insulated roofs were required (whetherthis requirement was met was not considered).

ARTICLE IN PRESS

Table 2

Exposure coefficients for flat roofs

References Exposure coefficient

Sheltered Semi-sheltered Windswept

Otstavnov and Rosenberg [4]a 0.98 0.72 0.46

Lutes [3] 0.90 0.60 0.30

Taylor [7] 0.80b — —

O’Rourke et al. [8]c 0.76 0.57 0.55

Høibø [5]

s0 ¼ 1.0 kN/m2 0.82d — —

s0 ¼ 3.5 kN/m2 0.62d — —

Løberg [6] — 0.55 0.27

Com. Eur. Comm [9] 0.90 0.74 0.58

aAssumed snow cover for 3.5 months. Average winter wind velocity in

sheltered, semi-sheltered and windswept area are assumed to be

respectively 2, 4 and 6m/s.bSnow ground load with 30-year return period was used when

calculating roof-to-ground ratio.cValues are recalculated by the authors of this paper in order to apply

unheated roofs.dDegree of wind exposure was not registered. s0—ground snow load.

V. Meløysund et al. / Building and Environment 42 (2007) 3726–3736 3729

The findings of the investigations, which indicate theeffect of wind blowing snow from roofs, are summarized inTable 2.

In investigations performed by Irwin et al. [10], the effectof roof size on snow loads was studied. It was concludedthat there was a trend towards increased uniform snowloads on flat roofs with increasing size. It was recom-mended to account for roof size when considering roofswith characteristic lengths above, respectively, 75 and200m for sheltered and open wind exposure (characteristiclength equals width * (2–width/length)).

ISO 4355 [2] defines wind categories and temperatureclasses in connection with determination of the exposurecoefficients Ce (see Table 1). The justification of therecommendations is somewhat vague. According to Ot-stavnov and Rosenberg [4] drifting occurs at average windvelocities above 4m/s during snowfall and above 6.5m/swith no snowfall.

Other studies have focused on a more instant thresholdwind velocity and not a wind velocity averaged over alonger period. According to Mellor [11] threshold windvelocities of 3–8m/s at a height of 10m are needed in orderto transport loose and unbounded snow. If the surfacesnow is densely packed and firmly bounded threshold windvelocity above 30m/s may occur.

According to Kind [12] the threshold wind velocity isapproximately 5m/s at a height of 10m for fresh dry snow,11m/s for slightly aged or hardened snow and 23m/s forsnow hardened by very strong winds.

Li and Pomeroy [13] evaluated hourly observations fromthe period 1970–1976 at 16 meteorological stations in theCanadian prairies. Based on this studies threshold windvelocities were recorded and presented as a function oftemperature. It was concluded that threshold wind

increased nonlinearly with ambient air temperature above�25 1C. An average threshold wind velocity of 9.9 and7.7m/s was observed for, respectively, wet and dry snowtransport. An average threshold wind velocity of 7.5 and8.0m/s was observed for, respectively, fresh and agedsnow.Results from field investigations show a reduction in

roof snow load with increasing wind exposure (Table 2).The values of the exposure coefficients vary, possibly as aresult of differences in the definitions of the categories.Although it can be concluded that wind exposure is of largeimportance for the resulting snow loads on roofs.

3. The exposure coefficients for Norway according to ISO

4355

Data from 389 meteorological stations in Norway is usedin order to derive temperature zones and wind categories asdefined in ISO 4355 [2]. Within the normal period(1961–1990), stations with at least 15 years of data areused. Temperature zones are based on reference grids forthe normal period developed by the Norwegian Meteor-ological Institute. As seen in Fig. 4, almost none of thestations have mean temperatures above 2.5 1C in thecoldest winter month. Temperature category A (as definedin ISO 4355) is only represented at small offshore islands inthe southern part of Norway, and it is therefore not visibleon the map in Fig. 4. Mean temperatures between �2.5 and2.5 1C are mainly found in the coastal areas in south andwest. For a majority of the stations, mean temperaturesbelow �2.5 1C are registered for the coldest winter month.Almost none of the meteorological stations situated in

the zone with mean temperatures between �2.5 and 2.5 1Chave less than one day with wind velocity above 10m/s. Insouth and west, stations situated at places highly exposedto wind (e.g. at lighthouses on islands/peninsulas) havemore than 10 days with wind velocities above 10m/s. Insettled areas, the number of days with wind velocitiesabove 10m/s is mainly between 1 and 10. In the northernpart of Norway, the number of days with wind velocitiesabove 10m/s exceeding 10 days is found for thistemperature zone also in settled areas.Considering the areas with mean temperatures below�2.5 1C in the coldest winter month, some characteristicscan be observed. In the inland of southern Norway, thestations situated in the mountainous areas have mainlybetween 1 and 10 days with wind velocity above 10m/s inthe three coldest winter months. For lower regions, wherethe building density is highest, the number of days withwind velocity above 10m/s is mainly below 1. Furthernorth the number of days with wind velocity above 10m/sis mainly between 1 and 10 days. At some of the stationshighly exposed to wind, the number of days with windvelocity above 10m/s exceeds 10 days. These stations aremainly situated in areas close to the sea where the buildingdensity is low.

ARTICLE IN PRESS

72°0'0"N

67°0'0"N

62°0'0"N

57°0'0"N

72°0'0"N

67°0'0"N

62°0'0"N

57°0'0"N

10°0'0"E

0°0'0" 10°0'0"E 20°0'0"E 30°0'0"E

20°0'0"E

Fig. 4. Winter temperature categories and winter wind categories as defined in ISO 4355 for 389 meteorological stations (weather data from the reference

30-year period 1961–1990). See also Table 1 for explanation.

V. Meløysund et al. / Building and Environment 42 (2007) 3726–37363730

In Fig. 5 the exposure coefficients (according to ISO4355) for 389 meteorological stations are presented. For85% of the meteorological stations the exposure coefficientis 0.8, for 8% the value is 0.6, for 6% the value is 1.0 whilefor 2% the value is 0.5.

Six meteorological stations (2%) achieve values of theexposure coefficient equal to 0.5. Two of these stations aresituated in the mountains of southern Norway 1828 and2062m above sea level, respectively. The other fourstations are situated in the northernmost parts ofNorwegian mainland, 701 North at the gateway to theNorth-east Passage and to the Barents Sea. Three of thesestations are situated at lighthouses close to the sea andexposed to the weather. Only one station is situated in asettled area: ‘‘Vardø radio’’ in Vardø.

Thirty meteorological stations (8%) achieve values of theexposure coefficient equal to 0.6. Nineteen of these stationsare situated at lighthouses. Seven stations are situated atsmall island communities on the edge of the coastlineheavily exposed to the weather. Only four stations (1%) aresituated in settled areas: Ørland III, Bodø, Andøya andLoppa. None of the meteorological stations in thiscategory is situated in the eastern part of Norway wherethe building density is highest.Twenty-two meteorological stations (6%) achieve

values of the exposure coefficient equal to 1.0. Thesestations are placed shielded from the wind typically atthe farther end of the long fjords of western Norway.None of the stations in northern Norway is in thiscategory.

ARTICLE IN PRESS

72°0'0"N

67°0'0"N

62°0'0"N

57°0'0"N

72°0'0"N

67°0'0"N

62°0'0"N

57°0'0"N

10°0'0"E 20°0'0"E

0°0'0"

1.5

2 and 2.5

3 and 3.54 and 4.5

5 and 5.5

6 and 6.5

7 and 7.58

0.5

0.6

0.8

1.0

10°0'0"E 20°0'0"E 30°0'0"E

Characteristic snowload on the ground

Exposure Coefficient (Ce)

Fig. 5. Exposure coefficients according to ISO 4355 for 389 meteorological stations (weather data from the reference 30-year period 1961–1990) and

characteristic snow load on the ground (kN/m2, 50-year return period) for municipality centres.

V. Meløysund et al. / Building and Environment 42 (2007) 3726–3736 3731

Three hundred and thirty-one meteorological stations(85%) achieve values of the exposure coefficient equal to0.8. Almost all of the stations in settled areas are found inthis category. Exceptions are one station with an exposurecoefficient of 0.5, four stations with exposure coefficients of0.6 and 22 stations with exposure coefficients of 1.0 asmentioned above.

4. Evaluation of the exposure coefficients for Norway

4.1. Historical field investigations

Results from field investigations show a reductionin roof snow load with increasing wind exposure (seeTable 2). The calculated values of the exposure coefficient

according to ISO 4355 [2] for building sites with meantemperature between �2.5 and 2.5 1C are in fairly goodagreement with these results, and to the conservative side.But there is no available research supporting ISO’sdescription of wind categories.In regions with a mean temperature above 2.5 1C, ISO

4355 allows a reduction of the snow loads on the roof onlyat building sites with more than 10 days of wind velocityabove 10m/s. This seems not to be justified by fieldinvestigations. It is nevertheless reasonable considering thefact that high temperatures reduce the ability of windactions to transport snow. Whether this temperature limitshould be 2.5 1C is uncertain.In regions with a mean temperature below �2.5 1C, ISO

4355 recommends a reduction of snow loads also when the

ARTICLE IN PRESSV. Meløysund et al. / Building and Environment 42 (2007) 3726–37363732

building is completely shielded. An exposure coefficientequal to 0.8 for this situation agrees with some of theresearch results (see Table 2). Some questions remain: werethe buildings in the historical investigations completelyshielded? Is it possible to obtain a completely shieldedbuilding? For a completely shielded building the snow loadon a flat roof is expected to be equal to the snow load onthe ground.

4.2. Snow transport theories

According to snow transport theories drifting occurs evenfor light winds (0.3–1.5m/s). At higher wind velocities(1.6–3.3m/s) the snow particles move more horizontallythan vertically. Drifting affects the deposition of snow;

72°0'0"N

67°0'0"N

62°0'0"N

57°0'0"N

10°0'0"E

0°0'0" 10°0'0"E

Mean winter temperature2.5 - 5

0 - 2.5

-2.5 - 0

-5 - -2.5

-7.5 - -5

-10 - -7.5

-12.5 - -10

-15 - -12.5

< -15

Fig. 6. Mean winter temperature (1C, December–February, w

particles are transferred through areas with high windvelocities and accumulate in areas with low wind velocities.At wind velocities between 3.4 and 5.4m/s the snow movesconsiderably faster horizontally than vertically, and signifi-cant redistribution may occur. Higher winds often blow thesnow off the roofs leaving them almost bare [7,14].The limit of 10m/s chosen for the wind categories seems

unreasonable considering the fact that drifting occurs atwind velocities as low as 0.3–1.5m/s. A larger number ofmeteorological stations are expected to achieve a value ofthe exposure coefficient below 0.8. In Canada heavysnowfalls often coincidence with high wind velocitiesaccording to [15]. This is not the pattern in Norway. InNorway heavy snowfalls may occur at low wind velocitiesas well as at higher wind velocities.

72°0'0"N

67°0'0"N

62°0'0"N

57°0'0"N

20°0'0"E

20°0'0"E 30°0'0"E

eather data from the reference 30-year period 1961–1990).

ARTICLE IN PRESSV. Meløysund et al. / Building and Environment 42 (2007) 3726–3736 3733

4.3. The Norwegian climate

Norway is a country with large variations in meantemperatures and wind velocities. In Fig. 6 mean wintertemperatures are given (December–February). There is apattern of low winter temperatures in the mountainousregions of southern Norway and the inland regions in thefar north. Coastal areas in southwest have temperaturesbetween 0 and �2.5 1C, while the inland in the far northhas winter temperatures less than �15 1C. In Fig. 7 numberof days with wind stronger than 5m/s for months withnormal temperature less than 1 1C is presented. Tempera-ture 1 1C is chosen as a limit to consider all months withhigh probability of snow and snowdrift. The map also

72°0'0"N

67°0'0"N

62°0'0"N

57°0'0"N

10°0'0"E

0°0'0" 10°0'0"E

Days with wind > 5 m/s

Months withTemp <= 1 C

0 - 25

26 - 5051 - 75

76 - 100101 - 150

151 - 250

0 - 2

3 - 45 - 6

7 - 8> 8

Fig. 7. Days with wind velocity above 5m/s in months with mean temperatu

reference 30-year period 1961–1990).

shows number of months with normal average temperatureless than 1 1C. Approximately 90% of the Norwegianmainland has six or more months with average temperatureless than 1 1C. That is, six or more months with possiblesnow and snowdrift. In Fig. 7, there is a clear pattern ofhigher probability of high winds in the costal areas. Itshould be noted that the costal areas do not always havethe highest number of occurrences (days with windstronger than 5m/s for months with normal temperatureless than 1 1C), but these regions also have fewermonths with temperatures below 1 1C. In the middle ofNorway (approximately 62–671 North) this pattern is mostpronounced. In this region the coastal areas have at leastone month with mean temperatures below 1 1C less than

72°0'0"N

67°0'0"N

62°0'0"N

57°0'0"N

20°0'0"E

20°0'0"E 30°0'0"E

re 1 1C or below for 389 meteorological stations (weather data from the

ARTICLE IN PRESSV. Meløysund et al. / Building and Environment 42 (2007) 3726–37363734

the inland, but a higher number of days with wind above5m/s.

Norway also has large variations in snow loads (seeFig. 5). Characteristic snow loads on the ground variesfrom 1.5 kN/m2 in coastal areas in the south up to9.0 kN/m2 in some inland areas (50-year return period).

As documented, Norway has areas with high snow loads(Fig. 5) and high frequency of wind (Fig. 7). In these areasthe wind affects roof snow loads, and the exposurecoefficient is expected to achieve its lowest value. As seenin Fig. 4 the definition of the exposure coefficient as givenin ISO 4355 does not point out these areas as areas wherewind blows snow away from roofs.

Many of the meteorological stations with an exposurecoefficient of 0.8 are situated in areas exposed to wind andsnow (e.g. the mountainous areas of southern Norway andstations along the coastline). Exposure coefficients lowerthan 0.8 are therefore expected for these stations.

4.4. Main findings

Historical field investigations show that the effects ofwind on roof snow loads are of significance for a large partof buildings. According to snow transport theories snowdrifting occurs even for low wind velocities. Norway has aclimate with low winter temperatures, large snow amountsand high frequency of wind. The definition of the exposurecoefficient found in ISO 4355 is too conservative and doesnot manage to differentiate the buildings in settled areas.The stations found to be most windswept are situated inareas where no or a very few buildings are located. Otherstations situated in areas known as windswept and withhigh snow loads are found in the same category as moreshielded stations. According to the definition in ISO 4355the buildings in the field investigations of Høibø [5] andLøberg [6] have all exposure coefficient equal to 0.8although the documentation shows exposure coefficientsconsiderably lower for the investigated buildings.

Norway has large variations in snow loads (Fig. 5). Thelowest snow loads are due to heavy snowfalls over a shortperiod while the higher snow loads are a result ofaccumulation over a long winter season. It seems reason-able that areas with low snow loads have higher exposurecoefficients than areas with high snow loads. This is takeninto account in ISO 4355 when differentiating the exposurecoefficient according to mean temperatures, but the limitschosen are not substantiated thoroughly through researchresults.

Another way of taking into consideration the accumula-tion length is to include the length of the winter seasonwhen deciding the exposure coefficient for a specificbuilding site. In the procedure presented by Otstavnovand Rosenberg [4] both the length of the winter season andnumber of days with snowfalls are included. Number ofdays with average wind velocities above 10m/s in the threecoldest months defines the wind category according to ISO4355. When selecting three months, it is indirectly assumed

that this is the length of the winter season. Other periodsshould be considered when evaluating areas with lower orhigher accumulation period.Mean temperatures for the coldest winter month are

needed when determining the value of the exposurecoefficient. At first thought this temperature could alsobe considered as a value taking into account the possibilityof snow to be transported by wind actions. In this situationthe actual length of the winter season should be chosen. Asmentioned above it can also be a measure of the size of thesnow ground load. This correlation should then bescientifically documented.

5. Discussion and further work

Meteorological stations are located to enable a goodrepresentation of regional climate. Typical locations are inagricultural and settled areas, airports and lighthouses.That is, these areas have a better representation thanmountainous regions.Maximum snow loads on the roof often do not appear

simultaneously with maximum snow loads on the ground.In the measurements reported by O’Rourke et al. [8]maximum snow loads on roofs were measured independentof maximum snow loads on the ground. In the measure-ments performed by Høibø [5] the roof and the groundwere measured simultaneously.It also seems reasonable that the exposure coefficients

decrease as the return period of the characteristic snowloads on the ground increases. When measuring snow loadsthe exposure coefficients therefore are expected to behigher in a normal year compared with a year whenextreme loads occur. In measurements reported by Taylor[7] snow loads on the ground with return period of 30 yearswere used when calculating the roof-to-ground ratio.Similar evaluation should be performed also for othermeasurement data, for instance the data of Høibø [5].In Norway the meteorological data needed is not easily

accessible for structural engineers. The data basis can bebought at the Norwegian Meteorological Institute, but thewind category (according to ISO 4355 [2]) has to becalculated either by meteorologists or by consultants instructural engineering. After deciding the mean tempera-ture and wind category for the nearest meteorologicalstation delivering such data, the structural engineer has toevaluate if these values are reasonable for the temperatureand wind climate at the specific building site. Localtopography including altitude, surrounding buildings andtrees has to be evaluated in order to decide the windcategory. High-resolution maps are required. This evalua-tion is time-consuming and demands high qualifications.The investigation presented will be used as an important

basis for ongoing studies within the ongoing NorwegianResearch and Development Programme ‘‘Climate 2000’’[16], e.g. the relationship between snow loads on roofs andwind exposure will be further investigated [17]. In thisarticle the suitability of the exposure coefficient as defined

ARTICLE IN PRESSV. Meløysund et al. / Building and Environment 42 (2007) 3726–3736 3735

in ISO 4355 is analysed. Further work should focus ondeveloping a definition reflecting the physical processesmore correctly. There is, for instance, a need for taking intoaccount the length of the snow accumulation. Thedefinition of wind categories should also be looked into,and a more detailed method specifying wind exposureshould be developed. The authors are now addressingthese issues.

When using the exposure coefficient, the snow load on asheltered roof becomes twice as large as the snow load on awindswept roof. The significance of this coefficientaccording to total building costs will be studied. Howmuch could the society profit with a more extensively use ofthis exposure coefficient? The advantage of built-in securityaccounting for future change in wind exposure or climaticimpact could be desirable.

In recent years, methods to detail the design processaccording to snow loads have been developed. However,advanced tools and data processing are required. Thesetools are often not available for structural engineers, andhigh qualifications are required. The risk of engineers usingthese methods erroneously is also present. Further workshould focus on developing tools enabling differentiationof snow loads, and thus including local topography andclimate.

The ‘‘robustness’’ of the Norwegian building stock willalso be addressed as part of the ‘‘Climate 2000’’programme, e.g. through analysis of statistical data fromthe Ground Property, Address and Building Register alongwith knowledge on process induced building defects [18].The lifetime of the built environment depends closely onthe severity of local climate conditions, and a sensible wayof ensuring high-performance building enclosures in acountry with extreme variations could be to develop moresophisticated climate classifications or exposure indexes fordifferent building materials and building enclosures. Thiswork is now concentrated on issues related to buildingtechnology or building physics, and include development ofmethods for classifying different climate parameters andtheir impact on building enclosure performance [18,19].

6. Conclusions

It is shown that the exposure coefficient as defined in aninformative annex of ISO 4355 does not reflect the actualeffects of wind exposure on roof snow loads in Norway, themain reasons being oversimplifications in the definition ofthe coefficient and the extreme variations of the climate inNorway. The definition is based on coarse simplificationsof snow transport theories. It must be revised andimproved to serve as an applicable tool for calculatingdesign snow loads on roofs, using the best available datafrom meteorological stations in Norway.

As documented in this paper, Norway has areas withhigh snow loads and high frequency of wind. In these areasthe wind affects roof snow loads, and the exposurecoefficient is expected to achieve its lowest value. The

definition of the exposure coefficient as given in ISO 4355does not point out these areas as areas where wind blowssnow away from roofs.

Acknowledgements

This paper has been written within the ongoing SINTEFResearch and Development Programme ‘‘Climate 2000—Building Constructions in a More Severe Climate’’(2000–07), strategic institute project ‘‘Impact of ClimateChange on the Built Environment’’. The authors gratefullyacknowledge all construction industry partners and theResearch Council of Norway. A special thanks toDr. Kristoffer Apeland for valuable comments on the text.

References

[1] Standards Norway. NS 3491-3 Design of structures—design ac-

tions—Part 3: snow loads. Oslo, Norway: Standards Norway; 2001

(in Norwegian).

[2] International Organization for Standardization. ISO 4355 Bases for

design on structures—determination of snow loads on roofs. 2nd ed.

Geneve, Switzerland: International Organization for Standardization;

1998.

[3] Lutes DA. Snow loads for the design of roofs in Canada. In:

Proceedings of the Western snow conference, Victoria, BC, Canada,

1970. p. 61–7.

[4] Otstavnov VA, Rosenberg LS. Consideration of wind effect in

standardization of snow load. In: A multidisciplinary approach to

snow engineering: first international engineering foundation. US

Army Corps of Engineers Cold Regions 8387.500 SR 89-6, vol. 89-6;

1989. p. 256–63.

[5] Høibø H. Snow load on gable roofs—results from snow load

measurements on farm buildings in Norway. In: Proceedings of the

first international conference on snow engineering, Santa Barbara,

CA, USA: CRREL Special Report, 89-6, 1988, p. 95–104.

[6] Løberg PA. Snølaster pa tak. Arbeidsrapport nr. 5. Oslo, Norway:

Norwegian Building Research Institute; 1976 (in Norwegian).

[7] Taylor DA. A survey of snow loads on the roofs of arena-type

buildings in Canada. Canadian Journal of Civil Engineering 1979;6:

85–96.

[8] O’Rourke M, Koch P, Redfield R. Analysis of roof snow load case

studies—uniform loads. CRREL Report 83-1 1983. Cold Regions

Research & Engineering Laboratory; 1983.

[9] Commission of the European Communities DGIII-D3. Scientific

support activity in the field of structural stability of civil engineering

works—snow loads. Final report. Pisa, Italy: University of Pisa,

Department of Structural Engineering, Prof. Luca Sanpaolesi; 1999.

[10] Irwin PA, Gamble SL, Taylor DA. Effects of roof size, heat transfer,

and climate on snow loads: studies for the 1995 NBC. Canadian

Journal of Civil Engineering 1995;22:770–84.

[11] Mellor M. Cold Regions Science and Engineering, Part III, Section A3c,

Blowing snow. Hanover, New Hampshire, USA: US Army Material

Command, Cold Regions Research & Engineering Laboratory; 1965.

[12] Kind RJ. Snow drifting. In: Gray DM, Male DH, editors. Handbook

of snow: principles, processes, management and use. Oxford:

Pergamon Press; 1981. p. 338–59.

[13] Li L, Pomeroy JW. Estimates of threshold wind speeds for snow

transport using meteorological data. Journal of Applied Meteorology

1997;36:205–13.

[14] Nordli Ø. Fjellet i snø, vind, sol og take—Dannevigs Fjellbok. Oslo,

Norway: Det Norske Samlaget; 2000 (in Norwegian).

[15] Isyumov N, Mikitiuk M. Climatology of snowfalls and related

meteorological variables with application to roof snow load

ARTICLE IN PRESSV. Meløysund et al. / Building and Environment 42 (2007) 3726–37363736

specifications. Canadian Journal of Civil Engineering 1977;4:

240–56.

[16] Lisø KR, Kvande T, Thue JV. Climate 2000—building enclosure

performance in a more severe climate. In: Proceedings of the seventh

symposium on building physics in the nordic countries. Reykjavik,

Iceland: The Icelandic Building Research Institute; 2005. p. 1195–202.

[17] Meløysund V, Lisø KR, Siem J, Apeland K. Increased snow loads

and wind actions on existing buildings: reliability of the Norwegian

building stock. Journal of Structural Engineering 2006;132(11):

1813–20.

[18] Lisø KR, Kvande T, Thue JV, Harstveit K. A frost decay exposure

index for porous, mineral building materials. Building and Environ-

ment 2006, accepted.

[19] Rydock JP, Lisø KR, Førland EJ, Nore K, Thue JV. A driving rain

exposure index for Norway. Building and Environment 2005;40(11):

1450–8.