Comments prEN 1991-1-3:2020

Snow Loads on Roofs

Part V: Influences and Snow load distribution

Cylindrical roofs

Figure 1 (Photo Domke): Arched greenhouse roofs clad with plastic film under a snow cover

Autors:
Dr.-Ing. I. Pertermann, IB Puthli, Schüttorf
Prof. Dr.-Ing. R. Puthli, KIT, Karlsruhe Date: 16.11.2020
 2

Table of content

Symbols ……………………………………………………………………………... 2
1. Introduction and basic principles of EN 1991-1-3 ……………………...……… 4

2. Development of the Eurocode: unsolved problems ……………………………. 5

3. Background and measurements for cylindrical roofs …...…………………….. 7

3.1 Case studies in Canada (Taylor, 1979, 1980) …………………………………..…….. 7
3.2 Snow load limit according to Taylor and NBCC 77 ………………………….……… 8
3.3 Wind tunnel Jules Verne, France, 2000 …………………………….…………….….. 12
3.4 Other measurements …………………………………………………………………. 14
3.5 Measurements and CFD in Norway (Thiis et.al., 2009) …………………...…….….. 14

4. Development of codes and influences ………………………………..………… 17

4.1 EN 1991-1-3 versus NBCC, ISO, CSA - roof angle influence ……………….…….. 17
4.2 Comparison ISO 4355: 1998 - ISO 4355: 2013 …………………………………...… 18
4.3 Comparison unbalanced loads ASCE-7, NBCC and ISO 4355:2013 ……………… 22

5. Arched greenhouse roofs ……………………………………………………...... 24

5.1 Shapes, sizes and cladding of arched roofs for greenhouses ……………..…...…..... 24
5.2 Classification for the snow load distribution ……………………………..…...…..... 25
5.3 Snow load distributions for single-span greenhouse arches ……………...…...…..... 26

Literature to Part V …………………………………………………………….…. 33

Symbols
Abbreviations:
Luv windward side of the structure
Lee leeward side of the structure

Symbols:
NOTE The following symbols are based on EN 1990 and EN 1991.

Latin upper-case letters:

Ce exposure coefficient
Cesl exceptional snow load coefficient
Cm surface material coefficient
Ct thermal coefficient
N number of days with vm,10 > 5 m/s during the three coldest months of the year
Rsi internal surface heat transfer resistance in m2K/W: Rsi = 1/hi
Rse external surface heat transfer resistance in m2K/W: Rse = 1/he
Sr,1 total roof snow load in kN for load case 1
Sr,2 total roof snow load in kN for load case 2
 3
U overall heat transmittance in W/(m2K)
Uo special heat transmittance for snowmelt conditions in W/(m2K): Uo = U / (1 – U · Rse)
W width of a roof trough of a multi-pitched roof in m

Latin lower-case letters:

d depth of the snow in m
fd(Ce) drift surcharge factor for pitched roofs
fd(x; Ce) drift surcharge factor for arched roofs
hd roof height in m; arched roof ISO-format
hr roof height in m
he external surface heat transfer coefficient in W/(m2K), total: he = hr,e + hc,e
hr,e external surface heat transfer coefficient for radiation in W/(m2K)
hc,e external surface heat transfer coefficient for convection in W/(m2K)
hro external surface heat transfer coefficient for Black-Body-radiation in W/(m2K)
hi internal surface heat transfer coefficient in W/(m2K), total: hi = hr,i + hc,i + hcd,i
hr,i internal surface heat transfer coefficient for radiation in W/(m2K)
hc,i internal surface heat transfer coefficient for convection in W/(m2K)
l roof width or span in m
l1 snow covered roof width in m
n return period in years
s roof span in m
s snow load in kN/m2
sb basic roof snow load in kN/m2 ISO-format
sd drift snow load in kN/m2 ISO-format
ss slide load in kN/m2 ISO-format; slide load limit in kN/m2
sgm roof snow weight at the slide limit in kg/m3
s0 characteristic ground snow load in kN/m2 ISO-format
sk,n characteristic ground snow load in kN/m2 for the reference period n
si,n,t characteristic roof snow load in kN/m2 for the location i, the shape coefficient µi, the reference
period n, the exposure coefficient Ce and the thermal coefficient Ct (Note: for Ct < 1; Ce = 1)
min s1,n,t minimum roof snow load in kN/m2 for load case 1: uniform distribution
vm mean wind speed in m/s
vm,10 mean wind speed in m/s, measured in 10 m height
w width of the snow in the trough of a multi-pitched roof after sliding in m
x horizontal distance from the ridge of an arched roof (crown: x = 0 and b = 0°)
x30° horizontal distance from the ridge to the point with the roof slope b = 30°

Greek upper-case letters:

Dqe · t accumulated temperatures above zero degrees over time in °C · h

Greek lower-case letters:

a angle of roof pitch, measured from the horizontal
as roof angle at the slide load limit
as,mean mean value of the slip angle in a standardized sliding test
b roof angle of an arched or cylindrical roof, measured from the horizontal
e emissivity of a surface
qi internal air temperature in °C
qe external air temperature in °C
qe,m coldest monthly mean temperature
r equivalent density of a snow layer in kg/m3
µb basic shape coefficient ISO-format
µd drift load coefficient ISO-format
µs slide load coefficient ISO-format; slide load limit value
µs,mean mean value of the friction coefficient in a standardized sliding test
µi shape coefficient for the location i EN-format
 4
1. Introduction and basic principles of EN 1991-1-3

For the calculation of roof snow loads EN 1991-1-3:2003 is based on a multiplicative format, as
opposed to other standards, such as the ISO 4355, which have an additive format, because drifting
and sliding snow are added to a basic snow, uniformly distributed.

EN 1991-1-3: 2003: Roof snow load: si,n,t = µi · Ce · Ct · sk,n

where µi is the shape coefficient for the location i, with µ1 = 0,8 as
a reference value for flat roofs with a roof angle a = 0°
Ce is the exposure coefficient with 0,8 ≤ Ce ≤ 1,2 with Ce = 1 for
normal “topography” as a reference
Ct is the thermal coefficient (Ct = 1 for normal applications)
sk,n is the characteristic ground snow load* for a reference period n
with n = 50 years as a reference value in CC2
Note: * Exceptional snow loads sAd = Cesl · sk,n can be treated in the same way.

The final draft versions of the future Eurocode, e.g. prEN 1991-1-3:2020, indicate a shift of the
exposure coefficient from the equation for the roof snow load into the shape coefficient µi, where it
can be adapted to the shape of the roof. Within the shape coefficient the exposure coefficient Ce
becomes more prominent and can accommodate drift losses as well as drift surcharges better than
before. Therefore, this format is also used in the greenhouse standard EN 13031-1:2019.

prEN 1991-1-3: 2020: Roof snow load: si,n,t = µi · Ct · sk,n

where µi is the shape coefficient for the location i, with µ1 = 0,8 · Ce as
a reference value for small roofs with a roof angle a = b = 0°;
µ1 = 0,8 · Ce,F for flat roofs with Ce,F = 1,25 for large flat roofs
with Lc = ³ 400 m
Ce is the exposure coefficient with 0,8 ≤ Ce ≤ 1,2 with Ce = 1 for
normal “wind exposure conditions” as a reference
Ct is the thermal coefficient (Ct = 1 for normal applications; Ct =
1,2 for freezer buildings and ice-skating arenas)
sk,n is the characteristic ground snow load* for a reference period n
with n = 50 years as a reference value in CC2
Note: * Exceptional snow loads sAd = Cesl · sk,n can be treated in the same way.

For the design roof snow loads, different snow load cases exist. These load cases may not necessarily
be realistic but have to cover the most unfavourable distributions for the design of all parts of the
structure. They are limit states.

In the Eurocode they are called “load arrangements”. To avoid misconceptions and fruitless
discussions, the terms “undrifted” and “drifted” load arrangements in EN 1991-1-3:2003 are now
in the final draft prEN 1991-1-3:2020 corrected into “balanced” and “unbalanced” load
arrangements. This is also an improvement.

It has been cleared, that balanced roof snow loads are “uniformly distributed” and “affected by the
shape of the roof and its exposure to wind”. Unbalanced roof snow loads “resulting from snow having
been moved from one location to another location on the roof or off the roof, depending on the
exposure of the roof to wind and the effects of sliding.”
 5
2. Development of the Eurocode: unsolved problems

The snow load distributions for cylindrical roofs in the first and second generation of the Eurocode
are shown in comparison in Table 1.

Load case (i) Undrifted Load Arrangement

EN 1991-1-3:2003 Uniform distribution across ls for DNG b in °
Undrifted Loads
b < 60° b ³ 60°
Main document: Shape Coefficient µ2 = 0,8 µ2 = 0
Load case (ii) Drifted Load Arrangement (along ls)
EN 1991-1-3:2003 Windward (along ls/2) Leeward (along ls/2)
Drifted Loads:
µLuv = 0,5 · µ4(b) µLee = µ4(b)
2 asymmetric drift mounds
Maximum for 30°: µ4,30° = 0,2 + 10 · h / b ≤ 2
Format Roof Snow Load: s = sk · Ce · Ct · µi

Load case (i) Balanced Load Arrangement

prEN 1991-1-3:2020 Uniform distribution (along ls) for roof angle b in °
Balanced Loads
b < 70° b ³ 70°
µ2 = 0,8 · Ce µ2 = 0
Load case (ii) Unbalanced Load Arrangement (along ls / 2)
prEN 1991-1-3:2020 Windward Leeward
Unbalanced Loads
µLuv = 0 µLee = µ4(b)
Maximum for 30°: µ4,30° = 1/Ce · (0,2 + 10 h/b) ≤ 2/Ce
Format Roof Snow Load: s = sk · Ct · µi

Where: h/b is the ratio of roof height h to roof width or span b;

ls is the snow-covered width of the roof for -60° ≤ b ≤ 60° or new: -70° ≤ b ≤ 70°
Ce is the exposure coefficient and Ct is the thermal coefficient.
Table 1: Shape coefficients and snow load distributions for cylindrical roofs in EN 1991-1-
3:2003+A1:2015 and in prEN 1991-1-3:2020 ((below))

The snow-covered width of the arches ls has been increased by covering roof angles up to 60° to up
to 70. This in line with the innovations for pitched roofs but remains questionable for certain roofs
and climatic conditions, see Comments Part II - Pitched roofs.

The most important innovation can be seen on the windward side in the unbalanced load case (ii),
which is assumed to be free of snow in the future. This will increase the asymmetry considerably.
While this is certainly justified for small snow loads and warm, slippery surfaces, for large snow
loads a limitation may be required, see Comments Part II - Pitched roofs.

Opposed to EN 1991-1-3:2003 in the new prEN 1991-1-3:2020 the exposure coefficient Ce has an
increasing effect on the shape coefficient µ4 for the drift maximum in load case (ii). Because of the
new roof snow load format si = sk · Ct · µ4 excluding Ce, this effect remains in the drift snow load
ratio, as shown in Figure 12 in comparison to the ratio according to ISO 4355:2013.
 6
In prEN 1991-1-3:2020 only one trend of two opposed trends is considered, the increase of the amount
of drifting snow deposited on the leeward side of the roof. The other trend, the decreasing total snow
load on the roof, is not considered. In the previous version EN 1991-1-3:2003 the other trend was
considered, what was also not correct, because it led to increasing drift mounds in sheltered locations.
How the model is calibrated can also be questioned. With µ4 = 2/Ce, for windy, exposed locations
with Ce = 0,8, the maximum increases to µ4 = 2,5. Values above Cs = 2 have been measured in Canada
(Taylor, 1979) in two cases only. However, these cases were special. Whether this is appropriate for
Europe needed to be investigated. Thiis & Ramberg (2008) measured maxima of up to 4,5 of the
ground snow loads. However, these years did not have much snow in comparison to the characteristic
snow loads of the locations in Norway. These measurements are not representative.

The roof angle function is still missing in load case (i) and the surface roughness is not considered in
any model in the Eurocode EN 1991-1-3. In NBCC, ASCE 7 and in ISO 4355 the roof angle function
in the balanced load case and the influence of the surface roughness can be considered in all models.

As expected for such a simple model as the drift load case (ii), it tends to fail for large snow loads on
small arch radii. The height of the leeward drift mound should be limited by adaptation to the available
space behind the ridge (wind flow field), see chapter 3.2.

For large snow loads the unlimited drift mounds reach considerable heights. This has driven countries
in their National Annex towards regulations such as the one in DIN EN 1991-1-3/NA:2019 with
symmetrical and slightly unsymmetrical uniform loads over the entire width of the arch (Fingerloos
& Schwind, 2019). This regulation may be suitable for tunnels of small height in locations with large
snow loads, where the ground snow cover supports any snow to slide. The design of large slender
tunnels or arched roofs on higher walls in locations with small snow loads remains questionable in
the light of old and new research results, see chapter 3.1 and 3.5.

For warm roofs or in climates with moist snowfall events, sliding will be relevant instead of drifting.
The drift load case (ii) is not suitable to cover sliding on warm and cold slippery surfaces. Therefore,
the design of greenhouses required a different approach, separating unbalanced load cases for sliding
and drifting, see chapter 5.3.

As for other roof shapes, for cylindrical roofs applies as well, that previous national snow load
standards in Europe and the Eurocode EN 1991-1-3:2003 are not based on threshold wind speeds,
transport rates, meteorological data, thermodynamic models or any other scientific approaches for the
estimation of snow load distributions on roofs. The drafting of EN 1991-1-3:2003 has been influenced
by ISO 4355:1981. Further revisions were no longer taken into account, because it was considered
too complicated etc.

As Tory (1992) concluded: ”It is admitted by most countries that their snow load shape coefficients
are largely empirical, owing more to crude approximations of a very few observations than to
scientific derivation and justification.”

For the recent review of the second generation, according to the final draft prEN 1991-1-3:2020, it
has been claimed that international codes such as ISO 4355:2013 and ASCE 7-15 have been
compared. However, this cannot be true for all types of cylindrical roofs, as it will be explained in
detail. For cylindrical roofs, the redistribution of snow on the limited space of a curved roof and the
related sliding are to be specified. This has not been done for prEN 1991-1-3:2020.
Also, the background needs to be understood, before any coefficients or values, such as a thermal
coefficient Ct, exposure coefficient Ce or shape coefficients µi are taken into another standard.
Especially the American codes are calibrated quite differently.
 7

3. Background and measurements for cylindrical roofs

3.1 Case studies in Canada (Taylor (1979, 1980))

In an extended survey (aiming at 22 years field experience) of snow loads on arena-type structures in
Canada, Taylor (1979) presented measurement data of roof snow loads and valuable information
about the snow load distributions, snow heights, snow densities, shelter, orientation to the sun,
climate, date of collapse or critical state on 32 arched roofs, including 3 Cathedral arches and 15
Quonset arches, covering spans from 9,1 m to 61 m. However, most of the roofs (19) had small spans
up to 24,4 m and heights up to 14 m with a rise to span ratio of 0,125 to 0,55, dimensions also used
for greenhouses. The roof cladding is not described often, but from the use of the buildings it can be
guessed, that metal (sandwich) or corrugated plastic sheets were used. Some of the buildings (sheds,
workshops, storages, farm buildings) were unheated, few were heated (warehouse) and some were
“intentionally kept below zero” (curling rinks, ice skating alleys).

Out of 27 registered snow load distributions in the critical state there were:
• 27% fairly uniform, however not constant, with maximum snow above the ridge,
• 19% not uniform, but symmetrical, with less snow on the ridge, more snow lower down,
• 54% unbalanced, with accumulation on one side, less or no snow on the other side. The peak
snow accumulations reached heights of Cs = 2,9 times the ground snow.
The measured roof snow loads were referenced to characteristic ground snow loads between 1,1
kN/m2 and 4 kN/m2 of these sites according to NBCC 1977 using a singular shape coefficient Cs.
Note: The characteristic ground snow loads according to NBCC 1977 had a return period of n = 30 years. In the
meantime, slightly larger ground snow loads for n = 50 years are used. Should the shape coefficients Cs not be
adapted?
For the roof snow in the critical state snow weight densities between 1,9 kN/m3 and 2,78 kN/m3 were
measured and calculated. The overall climate is cold maritime and continental with continuous snow
cover between 3 and 5 months. This should be considered, when comparisons to Europe are made.
Uniform snow load distributions (27%) were found only at sheltered sites. The snow height
remained nearly constant from the ridge (0°) down to 30° roof slope. It reduced to zero towards the
edges, at semi-circular arches at about 70°. The largest roof - ground snow conversion was Cs = 0,78
≈ 0,8 on an unheated Quonset arch sheltered by trees.
Non-uniform, but symmetrical snow load distributions (19%) could be found on roofs with a
smooth, not granulated metal surface with high rise to span ratio of 0,36 to 0,55. The snow was
sliding from the ridges down and collected within the increasing snow cover at the edges. This could
be a problem for the proper use of the building, but it was not critical for the arch itself. The arch
itself experiences smaller pressures. It was concluded, that for roof snow loads drifting of snow is
more important than sliding. Therefore, the symmetrical sliding would not be investigated further at
this point. For modern roof claddings, such as plastic film, this should be revisited.
Unbalanced asymmetric snow load distribution (54%) as a result of continental snow drifting was
of the main interest here. In all cases the windward side tended to be free of snow, in some cases the
roof was snow free far onto the leeward side. The higher arches had triangular snow distributions on
the leeward side with a peak around 30° roof slope. The snow load decreased down to zero at 70°.
Shallow arched with side walls or on top of flat roofs had the snow load maximum at the edge further
down. Very often the snow surface of the drift is described as “horizontal” and “not protruding above
the ridge level”. These leeward drift surcharges did increase the previous design maximum at Cs = 2
in 14% of all cases (2 out of 14 drift cases for 32 cases in total). The largest value was Cs = 2,9.
However, both roofs were very large (61 m span), very shallow (h/b = 0,2) with side walls and with
 8
annexes, which could have prevented further sliding and drifting down. Both arches were not
freestanding. Therefore, the maximum value Cs = 2 would remain in the Canadian NBCC 1977.
These studies confirmed both models for arched roofs used in the Canadian building regulations until
that time in 1980, triangular mounds with maximum at 30° or triangle with maximum at the edge.
However, the snow load towards the edges had to be reduced. Also, the models seemed to apply only
within certain ranges and to fail in other combinations, e.g. for small loads and large radii and for
high snow loads and small width.
The shape of the leeward drift surcharge and the position of the maximum depends not only from the
rise to span ratio h/b, the “roof shape”, but also from the ratio of roof height to snow height h/sk, that
means from the snow load itself.

With reference to the revision of the Canadian National Building Code (NBCC) having two load
cases (case I: uniform distribution, case II: asymmetric distribution with two snow mounds) similar
to EN 1991-1-3:2003, Taylor (1980) noted the following:

“The treatment of case I was not entirely satisfactory, for there should probably have been a
reduction allowed for slopes above 30°.”
“There were two mayor problems with case II:
1) The first was inconsistency in the use of a slope reduction formula for all sloped roofs but this
one. Was it reasonable to expect a load of 2 g (2 sk) to adhere to a steeply sloped surface, even
90° as would be recommended for a semi-circular arch on vertical side walls? In a sense the
load was largely independent of geometry.
2) In moderate to heavy ground load area the profile of the recommended design snow protruded
far above the crown of the arch,” ... “an apparently unreasonable situation.” “Case II should
clearly be reconsidered to accommodate the geometry of the roof.”
As a result, NBCC 1977 was revised accordingly, as described by Taylor (1980):
“Case I: The uniformly distributed load was modified to make it consistent with the slope
reduction formula, allowing a reduction in load on slopes over 30°.”
”Case II: a) The windward side of the roof was assumed to be clear of snow.
b) The upper surface of the snow drift on the leeward side was assumed to be parallel
to the ground and no higher than the crown of the arch.
c) The maximum load at any point on the surface was to be not greater than twice
the 30-year return ground load.
d) The slope reduction factor ß, as previously described, was applied to all points on
the surface after application of c).”
These observations and measurements were of great importance for the development of the
international building codes. In recognition of the snow load influences on arched roofs the climate
in Canada, the roof sizes and the state of the art for roof cladding materials should not be forgotten.

3.2 Snow load limit according to Taylor and NBCC 77

For the balanced loads as well as for the unbalanced loads on arched roofs NBCC 1977 introduced
the same roof angle function as for pitched roofs. With the slope factor Cs = 0,8 and the format si =
Cs · Cw · sk this leads to a reference level at the ridge of the roof of si / sk = 0,8 Cw with Cw being the
wind exposure factor. In the following formulae Cw = 1 for windy, but sheltered locations in Canada.
Cw = 0,75 is recommended for windy and exposed locations.
 9

Load Case I: Balanced Loads

a ≤ 30°: with b(a) = 1: Cs = 0,8
30° < a < 70°: with b(a) = 1 - (a - 30°) / 40°: Cs = 0,8 · (1 - (a - 30°) / 40°)
a ≥ 70°: with b(a) = 0: Cs = 0
with: a roof angle in °;
b(a) Roof angle function for pitched roofs, if the sliding is not being obstructed.

Note: Since 1977 NBCC has been reviewed several times. Now roof snow loads can be differentiated much
better for surface roughness, wind and exposure.

For the unbalanced load case the observations by Taylor (1980) have been applied as follows:

Load Case II: Unbalanced Loads

Windward side: Cs = 0
Leeward side: a ≤ 30°: Cs = g · (h - hx) / sk ≤ 2
30° < a < 70°: Cs = Cs,30° · (1 - (a - 30°) / 40°)
a ≥ 70°: Cs = 0
with: g equivalent density of the snow in the drift with g = 2,35 kN/m3;
h maximum height of the roof at the ridge for x = 0 in m;
hx vertical height of the point x on the roof with max hx = h at the ridge and hx = 0 at the
base; with the coordinate hx = h · cos a for a semi-circular cylinder with radius r = h
h-hx height difference between the point x on the roof and the ridge;
with (h - hx) = h · (1 - cos a) for a semi-circular cylinder

Figure 2: Load case II: Unbalanced snow loads for semi-circular arches (h/b = 0,5) of different
sizes (h = 10 m to 40 m and b = 8 m to 80 m) for sk = 1,5 kN/m2 and g = 2,35 kN/m3

As can be seen in the example in Figure 2 the drift mounds on small roofs remain small as well. For
larger roofs, the drift surface remains horizontal on ridge level within the aerodynamic shade. The
increase from the ridge to the maximum at 30° is not linear but follows the curvature of the arch with
r · (1 - cos a) for a circular cylinder with the radius r.
 10
Because of the limit at Cs = 2, for larger roofs (large radii) a plateau is reached at the level of si / sk =
2. The plateau increases in length and can lead to very large amounts of roof snow. This is an effect
of this model, where the aerodynamic shade behind the ridge is filled up with snow, without checking,
whether enough snow is available. In load case II the roof snow load does not depend on the available
ground snow load, it depends on the wind field created by the geometry of the arch. However, there
have been no measurements to support this effect for very large roofs.

Therefore, Taylor suggests a limit for larger roofs after which the design model must be changed. In
this case the two alternatives from previous Canadian standards, load case III and load case IV are
suggested, see Figure 3. The condition for a model change is that the total snow load in case II
increases half the total snow load on the arch sk · b/2.

Load Case III:

Windward side: Cs = 0
Leeward side: a ≤ 30°: Cs = 2 · x/(b/2)
30° < a < 70°: Sliding not obstructed: Cs = Cs,30° · (1 - (a - 30°) / 40°)
Sliding obstructed: Cs = 2 · x/(b/2)
a ≥ 70°: Cs = 0
Load Case IV:
Windward side: Cs = 0
Leeward side: a ≤ 30°: Cs = 2 · x/x30°
30° < a < 90°: Cs = Cs,30° · (1 - (a - 30°) / 60°)

with: b Roof span in m;

x horizontal coordinate of the point x on the roof with x = 0 at the ridge and x = b/2
at the base; with x = (b/2) · sin a for semi-circular cylinder with the radius r = h;
x30° horizontal coordinate of the point with 30° roof angle; with x30° = 0,5 · (b/2) for a
semi-circular cylinder

Figure 3: Load case III and IV: Alternative unbalanced load distributions for shallow arches if the
snow loads in load case II are larger than sk · b/2

In the years around the publication of these regulations the use of computers was not common yet.
Therefore, the calculation of the snow load distribution in load case II, the test for the limit and the
change of models seemed too cumbersome for many, especially if the background was not
understood. As a much simpler snow load distribution, the load case IV with the maximum at 30°
and an adapted roof angle function was taken into most other international standards. That load case
 11
IV was only one of two alternative models for arched roofs with very large radii and that it required
unobstructed sliding, was forgotten.

However, if only the model in load case IV is used, the height of the drift mound cannot be limited
any more. The model fails for large snow loads in comparison to the roof height. Also, the snow load
distribution cannot be adapted to the limited space on the leeward side of smaller roofs. The model
fails for small arches as well.

For greenhouses with arched roofs, which have rather small radii, it is important to remember the
background of load case II, where the drift mounds are adapted to the available space behind the
ridge. In the wind field above the arch the drifting snow is deposited in the aerodynamic shade behind
the ridge and below, where the wind flow is separated from the surface, causing turbulences with
reversed flow, see Figure 4.

Figure 4: Wind field above a shallow arched roof according to Thiis & Ramberg (2008), Figure 6a

The drifts grow downwind until the wind flow above the surface can reattach itself to the newly
created drift surface (equilibrium state, see Tabler (2003)). Drifts on arched roofs cannot grow
upwards higher than the crown of the arch, where the wind flow from the windward side arrives.

This condition can be adapted as a limit for any model with leeward drift surcharges, when the
windward side is free of snow. For the maximum leeward drift at the roof slope 30°, a simple test is
carried out, whether the local snow load protrudes the ridge of the roof. If the maximum drift
surcharge is larger than the limit, it can be reduced.

max si (a = 30°) ≤ g · (h - hx,30°) = lim si

with: h, b height and width of the arched roof in m;

r radius of the arched roof; with r = h/2 · (1 + ¼ (b/h)2) for circular cyliners;
hx,30 vertical height of the point x30° at 30° roof angle; with hx30° = h - r · (1 - cos 30°);
with the available height above the arch (h - hx,30°) = r · (1 - cos 30°) = 0,134 · r;
g equivalent weight density of the drifted snow in kN/m3

In Figure 2 can be recognised, that for small roof heights (and radii) in comparison to the snow height
for ratios r/(sk/g) < 14,9, the limit is relevant, and the maximum drift snow load of 2 sk can be reduced
due to lack of space in the shade of the arch. In this example with h/b = 0,5 and g = 2,35 kN/m3 for
the snow load level at sk = 1,5 kN/m2 this applies to roof heights below 9,5 m and would affect most
of the greenhouse tunnels.
 12
3.3 Wind tunnel Jules Verne, France 2000

Because there were no provisions for small roofs with very smooth cladding, such as plastic film and
plastic sheets, snow load distributions on arched greenhouse roofs have been investigated in the
Climatic Wind Tunnel Jules Verne in Nantes, France (Scarascia-Mugnozza et.al. (2000)). Artificial
snow was applied on unheated arched greenhouse models in full and in slightly reduced scale. The
models were made with plywood, but realistically covered with plastic film and semi-rigid plastic
corrugated sheets. The models with sidewall and without walls (tunnels) were between 0,66 m and 2
m high. A full-scale tunnel with plastic film and a height of 2,31 m was also tested.

Unlike natural snow with diverse precipitation particles (Fierz et.al., (2009)), the artificial snow from
the snow cannons had compacter particles (mean particle diameter 0,25 mm) with a density of 352
kg/m3 (dry snow produced at -10°C with less than 2% liquid water) and 397 kg/m3 (wet snow
produced at -5°C with 5% to 8% liquid water in the snow volume). On the roof models the snow
densities increased only little up to 408 kg/m3. Natural fresh snow has densities between 50 kg/m3
and 120 kg/m3 depending on the air temperature and moisture saturation.

The wind speeds for the transportation of the artificial snow were also very low with 2,5 m/s in the
beginning and 1,5 m/s at the end. This is below any threshold wind speed for drifting snow, even if a
certain scale reduction is taken into account. The Climatic Wind Tunnel Jules Verne can simulate
wind up to 40 m/s. However, the small wind speeds were chosen on purpose, to obtain the “most
uniform snow distribution”.

This was justified with dramatic failure events for polytunnels under masses of snow. Here, the return
period of those failure events may be questioned. Is it economical to build lightweight plastic tunnels
to resist the snowstorm or blizzard of the century or better to rebuild it eventually?
In Japan, such questions are explored with a trial and error method (Moriyama et.al. (2015)),
compare Bras (2014), with a photo-report about the results of a dramatic 2 m-snowfall event.

A better argument for investigating uniform distributions is the sensitivity of slender arches to large
loads on the ridge (as well as to large asymmetries). Symmetrical snow accumulations (vertical loads)
on the lower parts of the tunnel have not much influence on the structure of the arches, but snow loads
on the top have (especially if they are asymmetrical). The snow at the bottom can have an influence
on the cladding and cause serviceability problems. In the tests, the so-called ponding was investigated.
Due to deformation of the plastic film more snow can accumulate in the pockets. Ratios of 1,1 to 1,22
were measured, 2,78 for a Gothic arch under wet snow. However, when the plastic film was
supported, there was no ponding and the snow could slide. This is why plastic film on greenhouses is
always pre-stressed. In snow rich countries additional snow supports are used, where the plastic film
can rest comfortably. Ponding should be prevented instead of designing for unknown load increases.

Independent from the wind speed and the moisture of the artificial snow, no leeward drift surcharges
formed. The snow cover was distributed nearly symmetrical (asymmetry ratios max. 0,76 to 0,81)
with a minimal cover at the top and increasing snow heights lower down due to immediate sliding.
More snow accumulated on the windward side. Full sliding was not possible on the small-scale
models. The accumulation on the bottom of the arches restricted further sliding.

However, if there was no restraint on the bottom for the full-scale arch with steep sides and for the
two large models, the snow cover on the ridge broke apart and slid down in large pieces. The sliding
could be symmetrical or asymmetrical at one side only. On the larger full-size model, the snow did
slide earlier and much more than on the smaller scale models. This can be explained due to the larger
size and weight of sliding snow pieces.
 13
The conclusion is, that experiments with smaller roof models under-predict sliding. Very small
models in combination with larger snow heights do not allow sliding at all, because the snow is
supported at the base. This effect may have caused the increasing load on the reduced scale models.

Multi-span roofs and tunnels

Multi-span greenhouses and tunnels were also tested in the Climatic Wind Tunnel Jules Verne in
France, see Scarascia-Mugnozza et.al. (2000). All of the multi-span models were too small for snow
sliding from the roofs. Also, the roof arches of the greenhouse models were rather shallow. With a
full-size model the snow would have filled the troughs, if enough snow had been available. The snow
on the side acting vertically does not endanger the arch construction. It is rather supportive. The
plastic film between the arches, however, can be deformed (film displacement effect) and cause
serviceability problems.

Under good sliding conditions the snow load on multi-span tunnels becomes only critical after the
troughs are filled and the snow rises above the ridge. Considerable amounts of snow are required to
achieve this. The return period of such events could be questioned, also in the view of future climate
trends.

Conclusions from all tests

The proposed shape factors for the cases A1, A2 and A3 on the basis of the tests by Scarascia-
Mugnozza et.al. (2003) cannot be used generally. They represent half-finished symmetrical or
slightly non-symmetrical sliding of artificial snow on very small arches with restricted sliding, but
without much influence of wind, see Figure 5. Accumulations on the windward side (left side in
Figure 5) are larger than on the leeward side.

Figure 5: Proposed shape factors for not heated arched greenhouse roofs based on tests in the
climatic Wind Tunnel Jules Verne by Scarascia-Mugnozza et.al. (2003).

In the same publication the outcome of a full-scale test on a greenhouse tunnel in France at about
1000 m altitude was shown. During snowfalls in the winter 2000/2001 a quite rare full sliding of
snow from one side of the roof with restricted sliding on the other side was observed (maximum
asymmetry). The separation was at the ridge over the whole length of the arch. The snow load level
 14
at the sliding event was high, but well below the characteristic snow load. This shows that the proposal
according to Figure 5 is not sufficient for the design of greenhouse tunnels and arched roofs.

For design relevant shape coefficients more natural conditions and greenhouse arches with up to 15
m height have to be investigated. However, these preliminary tests have shown, how slippery the
cladding of greenhouses (plastic film and corrugated plastic sheets) can be and that snow load
distributions as for normal roofs do not apply. In contrast the conclusions by Taylor (1979), for
modern roof claddings on cylindrical roofs such as metal, glass, plastic film or textile
membranes, sliding will be more important than drifting.

3.4 Other measurements

Otsuka & Homma (1992) reported about the efficient removal of snow from large curved membrane
roofs. There mean values for slide load limits depending on the roof angle are given. It is not possible
to use it for design yet, because the variations are large. Further tests and a statistical approach are
necessary to estimate design values for the slide load limit. As long as there are no limits, the snow
load on arches has to be based on the ground snow load as a reference.

In the last decade case studies, measurements, numerical analysis and advanced CFD- and FAE-
simulations have been carried out to expand the research on newly developed membranes and ETFE-
plastic film used for lightweight arched structures, for example in Japan (Tomabechi et.al. (1997)).

Paek et.al. (2013) conducted wind tunnel investigations on small arched greenhouse models. Sodium
bicarbonate (baking soda) was used to model snow particles. The models were made from balsa, a
tropical wood, and sandpaper, to “catch the snow particle well on the roof”. As a result, the model
showed maximum “snow” on the rooftop, however no sliding. There was also hardly any
accumulation in the troughs of the multi-span model or between the three single-span models. These
conditions do not resemble natural conditions with design relevant snow loads on greenhouse roofs.

Side pressure

Sometimes it is speculated that snow collecting in a confined space such as a roof trough would
develop side pressure against the trough. For water or for pure ice, frozen onto the surface and filling
the trough up to the ridge, that could be the case. Looking at the small density of snow in comparison
to water and ice, it becomes clear, that snow contains a lot of empty spaces filled with air. Ice crystals
such as dendrites, stellars, needles and plates are breakable and give way. The snow in the trough
would have an average density between 200 kg/m3 for dry drifted snow and 400 kg/m3 with melting
at the bottom, too small to offer any resistance.

The problem of roof snow sliding between two vertical walls has been investigated in Japan finding
no pressure in the first meter below the snow surface. For deeper snow accumulations the side
pressure increases. The variations are large. A formula for the design side pressure for confined snow
accumulations much larger than 1 m is given in AIJ (2006).

3.5 Measurements and CFD in Norway (Thiis et.al. (2009))

In Norway according to Potac (2013), Thiis & Ramberg (2008), Thiis et.al. (2009) and O´Rourke,
Potac & Thiis (2016) new case studies on modern arched roofs were conducted. Based on these
measurements in connection with the meteorological data, CFD-models were developed, and
parameter studies made. It was realized, that the snow load distributions on arched roofs need to be
updated. However, this process is not finished yet. This is shown by the different snow load
 15
distributions international standards still have. None of the models can convincingly be used for every
type of roof, every roof cladding in every snow climate. Therefore, discussions are ongoing.

Potac (2013) collected measurements of snow load distributions on 3 arched roofs over a period of 4
years in total. Two roofs with radii of 60 m and of 32 m were quarter cylinders with 45° roof angle
at the eaves and height to span ratios of h/b = 0,2. The third roof was a multi-span roof with three
smaller spans of 18 m. As sport or shopping halls they were closed, maybe heated buildings well
insulated but with a very slippery roof cladding (“plastic membrane”).

Thiis & Ramberg (2008) started to analyse the measurements on the quarter cylinder with 60,5 m
radius, 85,56 m wide and 17,72 m high at the base (h/b = 0,207) in Oslo with a characteristic snow
load of 3,5 kN/m2. The arched roof with a roof angle of 45° at the base was sitting on a flat roof annex
covering a ground area of 123 m x 106 m. The roof geometry can be seen in Figure 4.

The exposure of the site was characterised as “normal” with Ce = 1. Governing wind speeds during
snowfall were 4 to 6 m/s. According to prEN 1991-1-3:2020, Table 7.1 this roof would qualify already
as “windswept, exposed” with Ce = 0,8.

It was believed that drifts were formed mainly during snowfalls. Snow was not picked up from the
windward side and transported afterwards to the leeward side of the roof but settled there at the first
place due to the characteristically smooth wind flow over the roof finding aerodynamic shade far
behind and below the ridge.

This large and shallow arched roof showed asymmetrical snowdrifts of a nearly triangular shape on
one leeward side only, the surface of the drifted snow remained far below the top of the arch.
Despite the small roof angles at the base (45°), the windward side was nearly free of snow. The
maximum snow depth 2007 (F07) was h = 0,73 m (1,5 kN/m2; g = 2,055 kN/m3) measured at x = 29
m (a = 28,6°) from the middle of the roof. This was 4,5 times the measured ground snow load of 0,33
kN/m2. The snow drift was the result of a month of subsequent minor snowfalls with some melting
in between. In 2008 (J08) a single snowfall with a SWE (snow-water-equivalent) of 23 mm did lead
to h = 0,35 m (0,23 kN/m2) at x = 22 m (a = 21,5°) from the middle of the roof, same shape, but much
lower, see Figure 6. Both drifts remained well below the top of the arch.

Figure 6: Measured snow depths 2007 and 2008 on a shallow arched roof in Oslo according to
Thiis & Ramberg (2008), Figure 2
 16
These measurements show that unbalanced drift mounds can be formed also on very slippery arch
roof surfaces of heated buildings, if there is enough space available. However, to conclude, such
snowdrifts could develop everywhere on all arched roofs “for instance” in central Europe with sk =
0,3 kN/m2 is not correct.
As Hochstenbach, Irwin & Gamble concluded in connection with the research about the shape of
drift mounds in Canada: “However, at lower altitudes, the warmer climate would likely reduce the
unbalanced loading due to less drifting.”
And about drift mounds in general: “Also, although not within the scope of this study, the use of
slippery roofing materials (e.g. standing seem metal, PVC membrane) can lead to sliding, even on
low slope roofs, which would result in potentially large unbalanced loads.”

The wind field, the snow accumulation and the redistribution were simulated numerically using
previous experiences, achieving good agreements with the measurements, if properly calibrated, but
only then. Not all influences are really understood or can be simulated at the same time. However,
such simulations do not have to reproduce measurements in detail but can be used to investigate the
influences of different parameters.
In the numerical simulation it could be shown where and how the drifts develop. Very light snow was
simulated with a density of 50 kg/m3 and a particle radius of 0,2 mm. The drifts started to grow where
the separation point in the wind flow begins, see Figure 4. The newly developed snow surface above
the drift parallel to the ridge moved the separation point further downwind. Later the snowdrift growth
rate decreased with increasing snow transport until an equilibrium shape was reached. The maximum
was finally located at about 30°. The simulated drift mounds looked very much like in the model Case
II according to Taylor (1980) in Figure 2. The concave shape of the drift indicates, that the drifted
snow remaines parallel to the roof. However, even the largest drift of about d = 2,7 m height at x =
30,25 m (a = 30°) as shown in Figure 7 remained much smaller than the top of the arch with h - hx,30°
= 8,11 m >> 2,7 m = d.

3
Simulated snow depth [m]

Time 1
2 Time 4
Time 6
Time 12

1 Time 27
Time 31

0
10 14 18 22 26 30 34 38
Distance from top [m]
Figure 7: Numerically simulated drift growth on the shallow arched roof (h/b = 0,2) in Oslo
according to Thiis & Ramberg (2008), Figure 5

These measurements and simulations indicate, that the ground snow load might not be sufficient to
calculate roof snow loads using a single shape coefficient. The snow load distribution for the
unbalanced load arrangement case (ii) “Drift” depends on the size and shape of the roof in comparison
to the snow height and the wind, to create the characteristic wind field above the arch. Furthermore,
for cases where drifting snow is not relevant, more than one unbalanced load arrangements may be
required as for greenhouses according to EN 13031-1:2019, see chapter 5.3.
 17

4. Development of codes and consideration of influences

4.1 EN 1991-1-3 versus NBCC, ISO, CSA - roof angle influence

The models for balanced snow load distributions on cylindrical roofs in EN 1991-1-3:2003, prEN
1991-1-3/NA:2020 and as a worst case in DIN EN 1991-1-3/NA:2019 are inconsistent.

On pitched roofs, roof angle related drift losses with some sliding can be taken into account for roof
angles above 30° (at x = 0,5 · b/2 on a semi-circular arched roof with the radius b/2). For arched roofs
the snow is not reduced, and the full roof snow height remains unchanged up to roof angles of 60° (x
= 0,866 · b/2), 70° (x = 0,94 · b/2) or even 90° (x = b/2), over the whole width b of the arched roof.

Other international standards allow the reduction of roof snow loads consistent with their regulations
for pitched roofs, see Figure 8. The consistent treatment of arched roofs is recognised since the
measurement campaign in Canada (Taylor (1979, 1980)), as commented in chapter 3.1.

Figure 8: Comparison of balanced snow load distributions on arched roofs in international

standards for reference cases, e.g. Ct = 1, Cm = 1 und Ce = 1 for ISO 4355:2013

The assumption according to EN 1991-1-3 may be based on observations of sudden sliding events,
typical for small cylindrical roofs with smooth or even slippery surface. The snow depths of the
remaining snow after such sliding events are often not very high. This shows, that sliding occurs well
below the characteristic ground snow level. However, as a limit state for balanced roof snow load
distributions (load case (i) or I or 1) with reference to the characteristic ground snow load, it is not
relevant.

The most likely reason for this limited approach was, that symmetrical slide load limits are just not
known, see Taylor (1979, 1980). Also, sliding can cause large unbalanced snow load distributions,
which would be relevant for the design, if they could not be covered with the unbalanced load case
(ii) or II or 2 (leeward drift mounds).

Sliding, State of the Art:

For the design of modern arched roofs, the sliding of snow should be revisited. Modern arched roofs
are more likely to have smooth or slippery roof surfaces than pitched roofs, e.g. metal, plastic sheets,
 18
plastic film or textile membranes. The sliding of snow on Teflon-coated glass fibre membranes is
“remarkably superior” to other materials, e.g. float glass (Ito et.al., 1997). Float glass performs in
standard sliding tests (Jelle, 2012) much better than metal, metal somewhat better than traditional
roof claddings such as ceramic or concrete tiles, asphalt roofing or wood shingles, used in most of
the measurements (Høibø, 1988).

The surface roughness as a parameter for the probability of sliding depends on several other properties
of the roof surface, the heat flux from below, the thermal properties of the snow cover and the climate
above the snow (temperature and moisture). The roof surface alone contributes with three properties,
the geometric roughness (e.g. as mean depth of irregularities per m2), the roundness of the
irregularities and the water repellence (e.g. as cosine of the contact angle of a water drop). Float glass
has a small geometric roughness but is not water repellent. Teflon coatings are very water repellent
(Lotus-effect), see (Ito et.al., 1997).

According to these findings only Canada has published in CSA S367-09 alternative roof snow load
distributions for membrane roofs opposed to normal building roofs. Design relevant are two load
cases, a balanced load case with roof angle influences and an unbalanced load case with
asymmetric sliding of the balanced load from one side. Drift mounds are not considered.

Note: EN 13031-1:2019 does not follow this route entirely because the roof snow load measurements in
Norway showed the development of leeward drift surcharges on shallow arches covered with plastic
membrane, if there is enough space on the roof (large shallow arch) and the sliding is obstructed at the base.

Of great importance for the sliding are the thermal properties of the roof on a building or structure.
It helps to differentiate between two main causes for the sliding.

1. Active Sliding on warm surfaces is initiated by controlled heating of the surface itself or
of the air below the surface. In the last case it depends on the heat permeability of the roof,
the snowfall rate and the air temperature and moisture outside (Otsuka & Homma (1992),
Tomabechi et.al. (1997)). For heated greenhouses this is practised for more than a
century.

2. Passive Sliding on cold surfaces is initiated by climatic influences. It requires the

accumulation of temperatures above 0°C and moisture in the snow cover. The effect can
only be captured statistically by evaluation of the meteorological data during snowfall
events of a certain reference time scale, e.g. 50 years, see Chiba et.al. (2012) and Zhou
et.al. (2015). Lower and upper limits for slide loads are to be estimated depending on the
purpose.

4.2 Comparison ISO 4355: 1998 - ISO 4355: 2013

4.2.1 Balanced load case

Some of the conclusions after the measurement campaign in Canada (Taylor, 1979, 1980) are
included in ISO 4355:1998, see Table 2 and Table 3. The influence of the roof angle is consistent.
In prEN 1991-1-3:2020 this is not the case yet.

The basic shape coefficient according to ISO 4355 is not only a function of the roof angle, it allows
also the differentiation according to surface roughness by using the Surface Material Coefficient Cm.
The difference between ISO 4355:1998 and ISO 4355:2013 is the roof angle function, see Figure 9
and Table 2. The resulting total snow load in load case (i) changes by less than 5%.
 19

Load Case 1 Basic Load Case

Balanced Loads Roof angle b in °
Surface Material Coefficient b ≤ 30° / Cm 30° / Cm < b < 60° / Cm b ≤ 60° / Cm
Cm = 1 b ≤ 30° 30° < b < 60° b ³ 60°
Cm = 1,2 b ≤ 25° 25° < b < 50° b ³ 50°
Cm = 1,333 b ≤ 22,5° 22,5° < b < 45° b ³ 45°
Basic Shape Coefficient 1998 µb = (cos (1,5 Cm · b) )0,5 µb = 0
Basic Shape Coefficient 2013 µb = 1 µb = (60° - Cm · b) / 30° µb = 0
Roof Snow Load 1998 sb = s0 · Ce· Ct · µb for 0,5 ≤ Ce ≤ 1
Roof Snow Load 2013 sb = 0,8 s0 · Ce · Ct · µb for 0,8 ≤ Ce ≤ 1,2
Table 2: Balanced Loads: Shape coefficients and snow load distribution for arched roofs
according to ISO 4355:1998, 5.4.5.5 and ISO 4355:2013, B.4

The continuous basic shape coefficient using the square root of a trigonometric term of the roof angle
function µb = (cos (1,5 Cm · b))0,5 based on measurements by Høibø (1988) has been replaced by the
recent bi-linear function in the new ISO 4355:2013. This leads to larger loads around 30° roof angle,
where the maximum drift is located too. Therefore, this simplification is not economical. The only
argument for a bi-linear function is simplicity, an argument of lower importance in times, when
computers are used. The new bi-linear function is an unfortunate step backwards. It would have
been better to improve the trigonometric function (replace by e-function for angles from 0° to 90°:
easier to program and to calculate with).

Figure 9: Comparison of roof angle functions in ISO 4355:1998 and 2013

 20

Note: The snow load distribution for warm roofs with Ct < 1 and Cm = 1,333 is not shown in Figure 9 to avoid
misconceptions about the possibility to vary the exposure coefficient Ce in the case of short-term snow loads.
This remains one of the mistakes in ISO 4355.

4.2.2 Unbalanced load case: Drift loads

For the unbalanced loads there is progress, see Table 3. In ISO 4355:1998 50% of the basic snow
load remains on the windward side of the roof (as so-called movable load), see also Figure 10.

Load Case 2 Drift Load Case

Unbalanced Loads Windward Side Leeward Side
Roof Snow Load 1998 s = ½ sb s = sb + sd with: sd = s0 · Ce · Ct · (µb · µd)
Roof Snow Load 2013 s=0 s = sd = s 0 · µ b · µ d
Drift Surcharge 1998 µd = (2,2 Ce - 2,1 Ce2) · sin (3 · b) for b < 60° und 0,5 ≤ Ce ≤ 1
Drift Surcharge 2013 High and normal Arch Shallow Arch
h / b ≥ 0,12 0,05 < h / b < 0,12
µd = 2 · x/x30° ≤ 2 µd = 16,7 h/b · x/x30° ≤ 16,7 h/b
Where: s0 is the characteristic snow load on the ground in kN/m2;
Ce is the exposure coefficient: 1998: 0,5 ≤ Ce ≤ 1 and 2013: 0,8 ≤ Ce ≤ 1,2;
Cm is the surface material coefficient with: 1 ≤ Cm ≤ 1,333;
Ct is the thermal coefficient: Ct ≤ 1,2;
h/b is the ratio of the roof height h to the roof span b;
x is the horizontal distance from the ridge (x = 0) with x = b/2 · sin b for semi-circular cylinders;
x/x30° is the ratio of x to x30° = x(b = 30°) ≤ b/2 for the roof angle b = 30°
at the snow load maximum. For distances x ≥ x30° is the value x/x30° = 1 applied.
Table 3: Unbalanced Loads: Shape coefficients and snow load distribution for arched roofs
according to ISO 4355:1998, 5.4.5.5 and ISO 4355:2013, B.4

Figure 10: Unbalanced Loads: Snow load distribution according to ISO 4355:1998, 5.4.5.5
 21
On the leeward side the drift surcharge is added to the basic snow load in the same way as for duo-
pitched roofs, see Figure 10. However, such distributions have not been met in the case studies in
Canada (Taylor (1979)) or could have been recreated in research.

Later the measurements and CFD-simulations in Norway (Thiis et.al. (2009)) lead to the replacement
of this model by a model with a leeward drift mound and the windward side free of snow, the same
principle as in ASCE 7-10 and NBCC 2010. Taylor (1980) suggested this type of model already as
one of two limit load cases (Case IV) for arched roofs with very large radii, see Figure 11.

Figure 11: Unbalanced Loads: Leeward drift mound according to ISO 4355:2013, B.4

The geometry of the drift surcharge depends on the surface material coefficient Cm, which is
introduced via the basic shape coefficient µb. Although the drift coefficient µd increases up to the roof
angle of 30°, the basic shape coefficient µb starts to decrease depending on the surface roughness
already at 25° (for Cm = 1,2) or even 22,5° (Cm = 1,333). It is not clear, whether the sliding of snow
would have such an effect on the geometry of the drift. If this is not anticipated, the location for the
drift maximum should be adapted by replacing 30° by 30°/Cm.

For shallow roofs with large radii in ISO 4355 there is a linear transition to the snow loads on flat
roofs. The transition for shallow arches starts at height (or rise) to span ratios of h/b = 0,12 with µd =
2 and ends at h/b = 0,05 with µd = 16,7 · 0,12 = 0,835. The smaller drift mound for h/b = 0,05 can be
recognised in Figure 11. Figure 12 shows the influence of the ratio h/b.

Opposed to the previous version ISO 4355:1998 in the new ISO 4355:2013 the exposure coefficient
Ce has no increasing effect on the drift surcharge coefficient µd. The new basic shape coefficient µb
according to ISO 4355:2013 does also not depend on Ce, see Table 3. Because of the roof snow load
format s = sd = s0 · µb · µd, the resulting drift mound is equal for all wind and snow climates and
exposures.

In the new ISO 4355 regulation it seems that two opposite trends are supposed to compensate each
other, the potential increase of the amount of drifting snow deposited on the leeward side of the roof
and the decrease of the total snow load on the complete roof.
 22
For strong winds in very cold climate with higher snow loads, the decrease of snow on the roof in
total can even dominate, compare the suggestion by Taylor (1980) for the use of Cw = 0,75. For
higher snow loads the relative ratio of the drifted snow load part (drift rate) decreases.

Figure 12: Influence of the ratio of roof height to span on the snow load maximum

4.3 Comparison unbalanced loads ASCE 7, NBCC and ISO 4355: 2013

O´Rourke, Potac & Thiis (2016) compared the measurement data (see Figure 6 and 7) with the
standardized provisions for arched roofs in ASCE 7-10, NBCC-2010 and ISO 4355: 2013. All three
codes have models with leeward drift mounds with the windward side left snow-free. The shape and
size of the drifted snow is different. They concluded that the model in ISO 4355: 2013 is mainly
conservative, except for very small snow loads on large arch radii, the models in NBCC-2010 and
ASCE 7-10 however are non-conservative or even “poor”.

However, the comparison is based on different assumptions and the standards have not been used
correctly. For ISO 4355 it was assumed, that that the roof near Oslo, Norway is a cold roof with a
non-slippery surface Cm = 1. For ASCE 7-1 it was assumed to be a warm roof (Ct = 1) with slippery
surface. For NBCC it was assumed to be cold and slippery - three different assumptions.

For the actual roof it has not even been checked whether the conditions for a classification as a warm
roof according to ASCE are satisfied or would be satisfied according to ISO. It was also not mentioned
in the publication, that the flat roof annexes at the base can be seen as an obstruction to sliding snow.
In this case the surface cannot be classified as slippery.
 23
NBCC does advise against the use of thermal coefficients (Ct = 1, no differentiation) but allows
different roof angle functions for slippery and non-slippery surfaces with the start of sliding at 15° or
at 30°. This creates a large difference, see blue lines in Figure 13. Such a difference is not caused
by any surface roughness alone but requires thermal influences too. Otherwise NBCC is calibrated
rather similar to ISO (red lines). The basic snow load factor is Cb = 0,8.

For warm unobstructed slippery surfaces (Ct = 1) according to ASCE (green lines), the thermal
resistance has to be less than 5,3 / 3,5 m2K/W for unventilated / ventilated roofs (assuming room
temperature inside). This is not very restrictive, but if these limits are not checked, the solid line for
non-slippery surfaces has to be used with the start of sliding at 30° not at 5°, resulting in much larger
loads, see Figure 13. Alternatively, the roof could be classified as a cold roof with Ct = 1,1 and the
start of sliding at 10° for a slippery surface and at 37,5° for a non-slippery surface. The performance
of ASCE does not have to be “poor”, if it is applied responsibly.

Figure 13: Differentiated comparison of drift models for arched roofs in ASCE 7-10, ISO 4355-
2013 and NBCC-2010 for warm / cold arched roofs with slippery / non-slippery surfaces

Note: These models show only design relevant unbalanced loads for the upper part of the arched roof. The
sliding snow will collect at the base on the flat roofs. Only ISO 4355 has a model for this local load case.

If only cold roofs with non-slippery surface would have been compared, all three codes would agree
well. The drift mound in ASCE has a somewhat lower peak, but more substance, see Figure 13. How
to deal with slippery surfaces and with warm and cold roofs, this is different in the three standards.

Also, the thermal and exposure coefficients Ct and Ce of ISO 4355 and ASCE cannot be compared
directly without adaptation, because the roof snow load format in total is calibrated differently. In
ASCE Ct = 1 represents a warm roof, in other codes usually a cold roof. In ASCE the reference value
is at 0,7 of the ground snow load. In ISO and NBCC the reference value is at 0,8 of the ground snow
load. A comparison has to be based on the total roof snow load. With 0,7 · (Ce · Ct)ASCE = 0,8 · (Ce ·
Ct)ISO for the comparison applies (Ce · Ct)ISO = 0,875 (Ce · Ct)ASCE.
If for the arch near Oslo, Norway in both standards Ce = 1 is assumed, for the thermal coefficient
applies Ct,ISO = 0,875 Ct,ASCE.
 24
This means a warm roof with Ct = 1 according ASCE can only be compared with a warm roof
according to ISO 4355 with Ct < 0,9 and a surface material coefficient Cm = 1,333 for a very slippery
surface. However, NBCC has no option for this case.

In such a case in reality the thermal coefficient Ct < 0,9 should have been proven by calculation, see
ISO 4355, Annex D and Annex F. Otherwise Ct = 1 and Cm = 1,2 would have been the safer choice,
at least for a design according to ISO. This refers to ASCE with Ct = 1,1.

The same problem for a fair comparison is the real exposure coefficient Ce of the location, where Ce
= 0,8 would have been an option, see prEN 1991-1-3:2020. This case would refer to Cw = 0,75
according to NBCC and complicate a direct comparison. However, the climate in Norway should
correlate better with the climate in Canada than in the USA, which on the other hand correlates better
with the European climate.

The question remains, when the tests were made in Norway, why not also compare it to the Eurocode
EN 1991-1-3:2003? Maybe, because these regulations performed so “poor”, that they could not be
compared? The comparison would have been very simple, because the Eurocode does not
differentiate between warm and cold roofs or slippery and non-slippery surfaces.

5. Arched greenhouse roofs

5.1 Shapes, sizes and cladding of arched roofs for greenhouses

Arched greenhouses roofs include shapes with curved and pointed arches with height up to 15 m and
hr / s ≥ 0,05. There are greenhouses with semi-circular or Quonset type roofs and sidewalls or without
sidewalls or greenhouse tunnels. For a semi-circular arch a height of 15 m refers to a span of 30 m as
a current limit for single-span arched roofs of greenhouses.

Most of the arched greenhouse roofs are covered with prestressed plastic film, as a single film or
layered and filled with air. They are referred to as greenhouses Type B, where the cladding is tolerant
to frame displacement. Rigid plastic corrugated sheets are also used for smaller size roofs.

Pointed arch roofs, Cathedral or Gothic types, covered with plastic film, are common. Pointed arches,
in contrast to curved arches, have a tangent or slope angle at the ridge (x = 0) of b (x = 0) > 0°. For
the choice of the snow load distribution, b (x = 0) = 15° is a limit, see ISO 4355. Pointed arches with
smaller angles are to be treated as curved roofs, with larger angles as pitched roofs. The drift
surcharges in the unbalanced load case (ii) are different because of different wind flow fields. The
uniform, continuous distribution in the balanced load case (i) is the same.

Single span versions with controlled heating allow an energy saving early sliding of roof snow,
before larger accumulations including drifts can be created. The sliding in action on a roof with curved
plastic film can be seen in many case studies in the nature in Figure 14. For the controlled heating
operation background information is available (Background Snow EN 13031 Part I - Melting).

Because arched greenhouse roofs are mostly covered in prestressed plastic film, snow sliding is very
common without being reflected in the design for snow loads according to most building codes.

Multi-span tunnels and arched roofs are not commented here, because their snow load distribution is
not covered by EN 1991-1-3. The distributions in EN 13031-1:2019 follow the principles for multi-
span pitched roofs explained in detail in the series Background Snow EN 13031.
 25
5.2 Classification for the snow load distribution

To be economical and consistent, the snow load distribution for curved and pointed arch greenhouse
roofs should be adapted to the same principles as for pitched roofs (consistency). The melting of
snow and the subsequent sliding are important, also asymmetrical. For tunnels the ratio of roof
height and snow height decides whether the sliding of snow is restricted at the bottom or not. As in
the climatic wind tunnel Jules Verne, France in tests shown, this is only the case for very small tunnels
in reduced scale, not for the arches of 2 m height onwards.

Figure 14 (Photo VDH): Repeated snow sliding off air filled plastic film during a snowfall event
with 65 cm of fresh snow with air temperatures between -13°C und -3°C; heating with internal air
temperature up to 19°C (controlled heating operation)

On larger roofs of commercial production greenhouses with very slippery surface a sliding
mechanism revealed itself in the climatic wind tunnel, subsequent sliding after reaching a limit load
on the ridge. The sliding can be asymmetrical. That is why in addition to the uniform / continuous
load case 1 (balanced loads) a one-sided slide load case 2 (unbalanced loads) should be considered
as a basis. For completely unheated greenhouses in colder climate snow drifting could lead to
asymmetric drift surcharges, as measured on larger membrane roofs in Norway. In this case a second
asymmetric load case 2 for drifting should be considered.

To enable a qualified choice for the appropriate snow load distribution including surface material
coefficient Cm and thermal coefficient Ct, greenhouses are classified according to their cladding,
heating and climate. EN 13031-1:2019, Annex C, C.1 (2) introduces the principles shown in Table
4. The values in grey fields (boxes) are open to further national regulations (NDP).
 26
According to EN 13031-1:2019, 3.1 the application is restricted to greenhouses by defining a
greenhouse as a “building structure that optimizes solar radiation transmission used for plants
requiring regulated climatic conditions”. This is a clear limitation against misuse.

Type of greenhouse / greenhouse compartment / roof / climate and exposure

Slippery roof cladding / closed building Others
T: Transparent cladding a, e.g. glass, plastic film or NT: Non- Non-slippery
membrane, plastic sheets with smooth surface, BIPV f transparent e surface:
H: Heated NH: not heated e.g. tiles,
e.g. granulated
with minimum air temperature under the roof: Not
surfaces
min qi in °C heated metal plates,
sandwich, Open,
S: Heat transmittance sufficient: NS: Heat textile exposed or
Uo ≥ 1 W/(m2K) transmittance membrane, ventilated
not sufficient: BIPV e roof:
C: controlled F: frost-free c e.g. porch,
Uo < 1W/(m2K)
heating b canopy,
e.g. heat
protective mounted PV
cladding
min qi >> 5°C min qi ≥ 5°C min qi < 5°C
Thermal coefficient Ct and Surface material coefficient Cm according to ISO 4355

Cm = 1,333 Cm = 1,2+ d
Cm = 1,2 Cm = 1
Sliding without drift surcharges - Drift surcharges with sliding
Ct < 1 Ct = 1
a
T (transparent cladding): solar radiation transmission possible, required for the growth of plants. Side effect: Snow
accumulation visible, so that immediate measures can be taken to remove snow and restore full radiation transmission.
b
C (controlled heating): Heating device is intended to melt a required amount of snow, to keep the roof snow load limited
and remove all snow from the roof as fast as possible.
c
F (frost-free): Heating device is not intended to melt snow, but keeps the temperature just above the required limit for
optimum crop production or survival, at least qi = 5°C.
d
NH (not heated): No heating available. In warm and moderate climate with snowfall temperatures around 0°C where
sliding of snow is more likely than snow drifting: Cm = 1,2+. In cold climates where drift surcharges are possible, the
surface material coefficient Cm = 1,2 should be used.
e
NT (non-transparent cladding, not heated): The area with non-transparent cladding is an integrated and minor part of
the greenhouse area, meant for specific functions, directly related to the professional production of plants and crops, such
as heating, storage, water management, etc.
f
BIPV (Building Integrated Photovoltaic) roof cladding, in contrast to roof-mounted PV
Table 4: Greenhouse categories due to cladding, heating, climate and exposure according EN
13031-1:2019; Annex C, Table C.1

5.3 Snow load distribution for mono-span arches

5.3.1 Motivation for the review
The models in EN 1991-1-3:2003 and in the previous EN 13031-1:2001 for a single span arched roof
are compared in Table 5. Load case (i) for balanced loads refers to load case 1. It shows a uniform
distribution for roof angles smaller than 60°. The asymmetric load case (ii) for unbalanced loads or
Load case 2+ for greenhouses shows snow mounds with a height depending on the ground snow load
and the shape of the arch, the ratio of height hr to span l = s.
 27
The only difference between EN 1991-1-3:2003 and EN 13031-1:2001 is the snow height limit ≤ 1
instead of ≤ 2 for greenhouses in load case 2+. No background could be provided for this reduction.
Only temperature and surface roughness could have been used as arguments for such a reduction.
However, the model was also used for cold roofs disregarding the type of cladding. Therefore, it met
criticism in the past and was in need for a review.

Load Case (i) or 1 Uniform Distribution (along l1)

EN 1991-1-3:2003 Roof pitch angle b in °
EN 13031-1:2001
b < 60° b ³ 60°
Shape coefficient µ1 = 0,8 µ1 = 0
Load Case (ii) Asymmetric Distribution (along l1)
2 snow mounds Windward Side Leeward Side
EN 1991-1-3:2003
µLuv = 0,5 · µ2 µLee = µ2
For the maximum snow height µ2 = 0,2 + 10 · hr/l ≤ 2
Load Case 2+ Asymmetric Distribution (along l1)
2 snow mounds Windward Side Leeward Side
EN 13031-1:2001
µLuv = 0,5 · µ2 µLee = µ2
For the maximum snow height: µ2 = 0,2 + 10 · hr/l ≤ 1
Where: hr/l is the ratio of arch roof height hr to the roof span l.
l1 is the width of the roof covered with snow (-60° ≤ b ≤ 60°)
The uniform distribution and the snow mounds are distributed across the width l1.
Table 5: Shape Coefficients and Snow Load Distribution for single-span arched roofs according to
EN 1991-1-3:2003 and according to EN 13031-1:2001

For a semi-circular arch with hr/l = 0,5 and the snow covered width l1 = sin (60°) · l = 0,866 · l, the
total roof snow loads Sr as a function of the ground snow load sk,n and the width of the arch l can be
calculated as follows:
Load case (i): Sr,1 = µ1 · sk,n · l1 = 0,8 · sk,n · l1 = 0,6928 · sk,n. · l (69% of the ground snow)
Load case (ii): Sr,2 = ((1 + 0,5) / 2) µ2 · sk,n · l1/2 = 0,375 µ2 · sk,n · l1 = 0,325 µ2 · sk,n · l
EN 1991-1-3:2003: µ2 = 2: Sr,2 = 0,65 · sk,n · l < Sr,1 (94%)
Load case 2 : EN 13031-1:2001: µ2 = 1: Sr,2 = 0,325 · sk,n · l << Sr,1 (47%)
+

5.3.2 Load case 1: Balanced Loads

In EN 13031-1: 2019 the shape coefficient µ1 from ISO 4355:1998 is used, together with the surface
material coefficients Cm, which govern the sliding. According to ISO 4355:2013 arched roofs with
cold, rough claddings (Cm = 1) are snow-free under design loads at a roof slope of 60°, as in EN 1991-
1-3:2003.

In prEN 1991-1-3:2020 the limit will increase to 70°. However, according to ISO 4355, cold, smooth
cladding (Cm = 1,2) is snow-free at roof slopes of 50°, warm, slippery cladding (Cm = 1,333) at 45°,
if unrestricted sliding is possible.

The snow load distributions in EN 13031-1:2019 for mono-span arched roofs of greenhouses are
summarized in Table 6.
 28

Load Case 1 Uniform / continuous Distribution

Surface Material Coefficient Roof slope angle b in °
Cm = 1 b < 60° b ³ 60°
Cm = 1,2+ and Cm = 1,2 b < 50° b ³ 50°
Cm = 1,333 b < 45° b ³ 45°
Shape coefficient µ1 = 0,8 Ce · (cos (1,5 Cm · b) )0,5 µ1 = 0
Table 6: Shape coefficient for arched roofs: EN 13031-1:2019, Annex C, Table C.4

The different snow load distributions for the continuous distribution (balanced load case 1) are
compared in Figure 15 for the example of a semi-circular arch in normal exposure (Ce = 1).

Figure 15: Load case 1 (balanced loads): Snow load distribution

Table 7 shows a comparison of the total snow loads of EN 13031-1:2019 to other regulations.

Comparison of Total Loads: Ratio of Roof to Ground Snow Load

Load Case 1: Balanced Loads
Surface Sliding Snow Integral (b) Area Total Snow
Roughness Cm limit covered (Numerical Load
ISO 4355 lim b l1 / l solution) sin (b) Sr,1/(sk,n · l)
Cold, rough 1 60° 0,8660 0,79876 0,71649 0,5732
Slippery 1,2 and 1,2+ 50° 0,7660 0,66563 0,61755 0,4939
Warm, slippery 1,333 45° 0,7071 0,59907 0,56387 0,4511
Control 90° 1 1,198 ✓ 0,93
Comparison
EN 1991-1-3: 2003 60° 0,8660 1 0,8660 0,6928
prEN 1991-1-3: 2020 70° 0,9397 1 0,9397 0,75175
CSA S367-09 70° 0,9397 0,71985 0,6764 0,54115
Ground snow 1 1 1 1
Note: The integral of µ1 = 0,8 (cos (1,5 · Cm · b)) over the snow-covered width l1 = sin (90° / (1,5 Cm))
0,5

· l has no closed elementary solution. It can be reduced to an elliptical integral and calculated numerically.
Table 7: Ratio of roof to ground snow for a semi-circular arch in load case 1: uniform /
continuous distribution, normal exposure (Ce = 1)
 29

The snow load according to ISO 4355:1998 is smaller than the snow load according to EN 1991-1-3
because the sliding with increasing roof slope angle is taken into account in the same way as for
pitched roofs. Further reductions are possible for cold smooth (Cm = 1,2), warm smooth (Cm = 1,2+)
and for warm slippery surfaces (Cm = 1,333).

5.3.3 Load Case 2b: Unbalance loads, Sliding

For warm roofs with controlled heating the thermal coefficient of Ct < 1 according to ISO 4355,
Annex D (Sandvik formula) is used. It accounts for the full precipitation despite using µ1 = 0,8. The
exposure coefficient Ce = 1 has to be used, independent from the real exposure. For other warm roofs
with Ct = 1 and Cm = 1,2+ or for cold roofs with Cm = 1,2 or Cm = 1 the exposure coefficient can vary.
This is a national choice. Figure 16 shows the remaining snow load distribution on one half of the
roof after sliding. For warm roofs it is the only unbalanced load case, see Table 8.

Figure 16: Asymmetric sliding for arched roofs with smooth and slippery cladding, Load Case 2 in
EN 13031-1:2019 and in CSA S367-09

5.3.4 Load Case 2a: Unbalanced loads, Drift

Because of very small wind speeds, drifting has not been investigated sufficiently in the Climatic
Wind Tunnel by Scarascia-Mugnozza et.al. (2000, 2003). That is why it cannot be excluded for cold
roofs with non-slippery cladding, open or exposed roofs (Cm = 1) and for cold roofs with non-
transparent, slippery cladding (Cm = 1,2) in cold climate.

For such cases the so-called drifted load case according to ISO 4355:2013, chapter B.4, Figure B.5
should be considered. These show additional snowdrifts on the leeward slopes of single span roofs.
For use together with the EN-format, a drift surcharge factor fd (x, Ce) has been derived.
 30
The regulation in ISO 4355:2013, Annex B, clause B.4 for arched roofs is based on research results
by Thiis & Ramberg (2008) and Thiis et.al. (2009) using field measurements, tests as well as
extensive and advanced CFD simulations.

Load Case 2 Asymmetric Distribution

Sliding and / or Drift Windward Side Leeward Side
Warm Cm = 1,33 and Cm = 1,2+ µLuv = 0 µLee = µ1
Cold Cm = 1 and Cm = 1,2 µLuv = 0 µLee = µ1 · fd(x; Ce)
Regulation has been removed and replaced by the following note:
Shape coefficients for not heated single span arched roofs and tunnels with non-transparent cladding can be
chosen nationally depending on cladding, climate and experience. For very smooth plastic film and textile
membranes Cm = 1,2+ should be used. A consistent approach as in Table C.5 may also be used, if more
appropriate. Drift surcharge factors are given in Table C.6.
Drift Surcharge Factor High and normal arches Shallow arches
hr / s ≥ 0,12 0,05 < hr / s < 0,12
fd (x; Ce) = 2,5 / Ce2 · x/x30° fd (x; Ce) = hr/s · 20,875/Ce2 · x/x30°
Where: Ce - is the exposure coefficient with 0,8 ≤ Ce ≤ 1,2
hr/s - is the ratio of the roof height hr to the roof span s
x - is the horizontal distance from the ridge (x = 0) with x = s/2 · sin b for semi-circular arches
x/x30° - is the ratio of x to the distance x30° = x(b = 30°) ≤ s/2
for the roof angle b = 30° as a maximum, with the ratio x/x30° ≤ 1
Table 8: Shape Coefficients and Unbalanced Snow Load Distribution for mono-span arched
roofs according to EN 13031-1:2019, Annex C, Table C.6

Figure 17: Unbalanced Snow load distributions (Load Case 2) on arched greenhouse roofs
according to EN 13031-1:2019
 31
5.3.5 Derivation of the drift surcharge factor:

Arched roofs: ISO 4355:2013, clause B.4

Drifted load case: Windward side: s = 0 with: µ = 0
Leeward side: s = sd = so · µb · µd Attention, no factor 0,8; Ce or Ct!
When using s0 = sk,n, roof angle b = a, roof height hd = hr and roof span b = s:
Basic load shape factor: µb = (cos (1,5 · Cm · b))0,5 according to ISO 4355:1998
High and normal arches: hd/b ≥ 0,12: x/x30° < 1: µd = 2 · x/x30°
x/x30° ≥ 1: µd = 2
Shallow arches: 0,05 ≤ hd/b ≤ 0,12: x/x30° < 1: µd = hd/b · 16,7 · x/x30°
x/x30° ≥ 1: µd = hd/b · 16,7
where: x is the horizontal distance from the ridge (with x = 0)
x30° is the horizontal distance to the point with a slope of b = 30°
x/x30° is the ratio of x to the distance x30°
Note: For a semi-circular arch applies: hr / s = 0,5 with x = s/2 · sin a and x30° = 0,5 · s/2 = s/4.

Transformation into the EN-format:

With: roof span s = b and roof height hr = hd
Shape coefficient: µ1 = 0,8 Ce · µb = 0,8 Ce · (cos (1,5 · Cm · b))0,5 acc. EN 13031-1:2019
Load case 2: Windward side: si = 0 with: µLuv = 0
Leeward side: si = sk,n · Ct · µLee with: Ct = 1 (no drift for Ct < 1)
µLee = µ1 (Ce) · fd (x; Ce)

For cold roofs with Ct = 1 and Cm = 1 or 1,2 can be written for the leeward roof face:
ISO 2013: s = sd = s0 · µb · µd
EN 2020: si = sk,n · µ1 (Ce) · fd (x; Ce)
With s0 = sk,n applies: µb · µd = µ1 (Ce) · fd (x; Ce)
µd = 0,8 Ce · fd (x; Ce)
Drift surcharge factor: fd (x; Ce) = µd / (0,8 · Ce)
Introducing the factor 1/Ce acc. prEN 1991-1-3:2020: fd (x; Ce) = µd / (0,8 · Ce2)

High and normal arches: hr/s ≥ 0,12: x/x30° ≤ 1: fd (x; Ce) = 2,5 / Ce2 · x/x30°
x/x30° > 1: fd (x; Ce) = 2,5 / Ce2
Shallow arches: 0,05 ≤ hr/s ≤ 0,12: x/x30° ≤ 1: fd (x; Ce) = hr/s · 20,875 / Ce2 · x/x30°
x/x30° > 1: fd (x; Ce) = hr/s · 20,875 / Ce2

For the semi-circular arch with hr/l = 0,5 of the examples above, the snow-covered width l1 depends
on the surface material coefficient Cm and the angle lim b, for which the snow slides off. The snow
covered roof width is l1 = sin (lim b) · l = sin (90°/(1,5 · Cm)) · l. The total roof snow loads Sr as a
function of the ground snow load sk,n and the width of the arch l is given in Table 9.

As can be seen in Table 9, the largest difference to EN 1991-1-3: 2003 is the complete sliding on one
side. This increases the asymmetry. However, prEN 1991-1-3:2020 indicates the same asymmetry
increase.
The peak values at 30° and the total snow loads on one side of the arch remain lower because of the
continuous roof angle function and the surface material coefficient Cm according to ISO 4355:1998,
which are based on the measurement data on pitched roofs by Høibø (1988). This allows a more
economical design of slender arches of greenhouses despite complete sliding, see Figure 17.
 32
Comparison of Total Loads: Ratio of Roof to Ground Snow Load
Load Case 2: Unbalanced Loads
Roof Distance Drift Surcharge Factors fd(x; Ce)
Ratio
slope angle from the ridge Exposure Coefficient
b in ° x/(s/2) x/x30° Ce = 1,2 Ce = 1 Ce = 0,8
0 0 0 0 0 0
10 0,174 0,347 0,6029 0,8682 1,3566
20 0,342 0,684 1,1876 1,7101 2,672
30 - 60 0,5 1 1,7361 2,5 3,90625
Snow ratio roof – ground 1): s2,n / sk,n = Ct · µ1 (Ce) · fd(x; Ce)
Roof
(Example hr/s > 0,12 for Ct = 1 and Ce = 1)
slope angle
Surface material coefficient EN 1991-1-3:2003 2)
b in ° Cm = 1,2 Cm = 1 Leeward side Windward side
0 0 0 0 0
5 0,346 0,347 0,333 0,167
10 0,677 0,683 0,667 0,333
15 0,977 0,995 1 0,5
20 1,230 1,273 1,333 0,667
25 1,421 1,506 1,667 0,833
30 1,533 1,682 2 1
35 1,347 1,561 1,667 0,833
40 1,112 1,414 1,333 0,667
45 0,791 1,237 1 0,5
50 0 1,017 0,667 0,333
55 0,723 0,333 0,167
60 - 90 0 0 0
Total 0,587 · sk,n · l/2 0,728 · sk,n · l/2 0,866 · sk,n · l/2 0,433 · sk,n · l/2
snow load 0,294 · sk,n · l 0,364 · sk,n · l 0,65 · sk,n · l
1)
Ct = 1, because for Ct < 1 there is no drift surcharge.
2)
EN 1991-1-3: 2003: µ2 and ½ µ2 are used, Ct = 1 and Ce = 1. Attention: Using Ce = 1,2 (sheltered)
would increase and using Ce = 0,8 (windy) would decrease the drifted snow. This is not correct!
Table 9: Drift surcharge factor according to ISO 4355:2013 in EN-format and ratio of roof to
ground snow load for cold roofs (load case 2) for a semi-circular arch

According to prEN 1991-1-3:2020 the snow-covered width l1 = ls will be increased by shifting the
limit of 60° to 70°. Also, the exposure coefficient Ce will be introduced as described. The total snow
load on one side of the arch will increase to Sr,Lee,2 = 0,9397/Ce · Ct · sk · l/2. For Ce = 1 this is 8,5%.
Because of the snow-free windward side, in load case (ii) the total snow load as a stabilising element
will decrease to Sr,2 = 0,46985/Ce · Ct · sk · l with l – total width of the arch.

5.3.6 Proposal for drift limitations in the National Annex to EN 13031-1:2019

Measurements and CFD analysis in Norway (Potac (2013), Thiis et.al. (2008, 2009); O´Rourke,
Potac & Thiis (2016)) showed the shape and size of asymmetrical snowdrifts on one side of the roof,
the surface of the drifted snow remaining below the top of the arch within the aerodynamic shade
behind the ridge. This is in line with the initial case studies on traditional arched roofs in Canada
(Taylor (1979, 1980)), who found the same influences of the roof shape (rise to span ratio hr/s) and
the ratio of roof height to snow height.

From these initial regulations according to NBCC 77 a limit for the leeward drift mound can be
derived, see chapter 3.2. It can be introduced into the National Annex to EN 13031-1: 2019 by a
simple additional check as shown in the blue box.
 33
For unheated single-span arched greenhouse roofs with Ct = 1 and Cm = 1,2 or Cm = 1 two asymmetric
loads cases are to be considered in design:

Windward side: µLuv = 0

Leeward side:
Load case 2a (drifting): µLee = µ1 ∙ fd(x; Ce) ≤ gs,equ · (hr - hx,30°) / sk
Load case 2b (sliding): µLee = µ1
with: µLuv = 0 for sk ≤ 1,8/Cm2 in kN/m2
µLuv = µ1 ∙ (1 - 1,8/(Cm2 · sk)) for sk > 1,8/Cm2 in kN/m2
Where:
gs,equ is the equivalent weight density of the snow accumulation, with: gs,equ = 2 kN/m3
x30° is the horizontal distance of the point with 30° roof angle from the ridge (x = 0); with
x30° = r · sin 30° = 0,5 r for a circular cylindrical roof with the radius r = hr/2 · (1 + ¼ · (s/hr)2)
hx,30° is the vertical height of the point x30° at 30° roof angle; with hx30° = hr - r · (1 - cos 30°);
with the available height (hr - hx,30°) = r · (1 - cos 30°) = 0,134 · r
with: sk, µ1, Cm, s, hr, x/x30° ≤ 1 and fd(x; Ce) according to EN 13031-1:2019, Table C.6

If in load case 2a the shape coefficient µLee is larger than the limit value lim µLee = rs,equ (hr - hx,30°)/sk,
it can be corrected. With the limit value the drift mound is as high as the ridge and it remains there.
Also, in load case 2a the snow load distribution for the length between 0° and 30° can be corrected.
It is not linear, but concave, because it follows the roof curvature, to remain horizontal, see Figure 2,
Figure 6 and Figure 7.

Special regulations for small tunnels or for very shallow arches are not required, because the snow
accumulations due to drifting remain small by the adaptation to the available space behind the ridge.
For small tunnels under very large snow loads, the drift mound and the asymmetry could become
very small. The reason is, that in the model, the snow on the windward side is blown off completely,
a not very realistic assumption for the maximum asymmetry with very large snow loads. Large snow
loads are the result of subsequent snowfall events. Further snowfall under less ideal drifting
conditions would enlarge the snow loads on both sides of the roof until the slide load limit is reached.
This would happen on the leeward side first, because of the larger load. The smaller load remains.
This is not the worst case for one-sided sliding.
Therefore, at the drift equilibrium state when the drift mound reaches the limit in height of the arch
crown, with the largest asymmetry and the smallest vertical load, the model should be changed.

In this case, the asymmetry due to one-sided sliding becomes relevant over drifting (Load case 2b).
Therefore, for unheated arched roofs two unbalanced load cases should be checked.

For the one-sided slide load in Load case 2b, a limit for the height difference exists (1,8 kN/m2) based
on the measurement campaign in Canada (Taylor, 1979, 1980), see Comment prEN 1991-1-3 Part
II - Snow Load Distribution - Pitched Roofs.
For arched roofs the largest roof snow load measured on a Quonset arch in Canada after a complete
sliding event was 1,4 kN/m2. With an assumed roof to ground conversion of Cb = 0,8 and Cw = 1, this
leads to 1,4 / 0,8 = 1,75 ~ 1,8 kN/m2.

Literature
Standards and References:
- CSA S367-09 Air-, cable-, and frame-supported membrane structures
- EN 1991-1-3: 2003: Actions on structures – Part 1-3: General actions – Snow loads
- prEN 1991-1-3: 2020: Actions on structures – Part 1-3: General actions – Snow loads
 34
- Final Report of the Commission of the European Communities DG III-D3 (1998-3), Scientific
Support Activity in the Field of Structural Stability of Civil Engineering Works: Snow Loads.
- Final Report of the Commission of the European Communities DG III-D3 (1999-09), Scientific
Support Activity in the Field of Structural Stability of Civil Engineering Works: Snow Loads.
- Formichi, P. (2017): Report from CEN/TC250/SC1, Project Team SC1.T2 – EN 1991-1-3 Snow
Loads (M515), 40th Meeting June 2017 in Brussels.
- Formichi, P. (2019): Report from CEN/TC250/SC1, Project Team SC1.T2 – Second Intermediate
Draft, 43th Meeting February 2019 in Paris.
- ISO 4355: 1981: Bases for design of structures – Determination of snow loads on roofs
- NS 3491-3: Prosjektering av konstruksjoner, Dimensjonerende laster, Del 3: Snolaster
- ISO 4355: 1998: Bases for design of structures – Determination of snow loads on roofs
- ISO 4355: 2013: Bases for design of structures – Determination of snow loads on roofs
- AIJ Japan Building Code, Commentary on Recommendations for Loads on Buildings. Chapter 5
Snow Loads, pp. C5-1-C5-28.
- JIS Japanese Industrial Standard C 8955: 2011 (E): Design guide on Structures for Photovoltaic
Array
- NGMA 2004, National Greenhouse Manufacturers Association, Structural Design Manual:
Structural Loads: Chapter 6 and Annex C.6: Snow loads.
- ASCE 7-10, 2010: Chapter 7: Snow Loads.
- NBCC, In: Ontario Regulation 88/19: Building Code filed May 2, 2019
- World Meteorological Organisation: Solid precipitation measurement intercomparison. Final
Report. WMO No. 67 (1998)
Background information of general interest will be accessible for Preview or Download under
www.greenhousecodes.com. The following other documents are currently available:
- Background Snow EN 13031 Part I.pdf: Melting
- Background Snow EN 13031 Part II.pdf: Sliding and Drift
- Background Snow EN 13031 Part II-1.pdf: Snow Load Distribution – Duo-pitched Roof
- Background Snow EN 13031 Part II-2.pdf: Snow Load Distribution – Multi-pitched Roof
- Background Snow EN 13031 Part II-3.pdf: Snow Load Distribution – Arched Roof
- Comments prEN 1991-1-3:2020 Part I: General Influences – Flat Roofs.pdf
- Comments prEN 1991-1-3:2020 Part II: Snow Load Distribution – Pitched Roofs.pdf
- Comments prEN 1991-1-3:2020 Part III: Snow Load Distribution – Multi-pitched Roofs.pdf
- Comments prEN 1991-1-3:2020 Part IV: Local Effects – Obstructions – Parapet – Retention
Devices – Solar Panels on Flat Roofs.pdf
- Comments to ISO-4355-2013 Part I.pdf: The use of the thermal coefficient for large roofs
- Comments to ISO-4355-2013 Part II.pdf: Critical review of the limits of the thermal coefficient
- Comments to ISO-4355-2013 Part III.pdf: Influence of the roof angle on the thermal coefficient
- Comments to ISO-4355-2013 Part IV.pdf: Drift / slide model multi-pitched roof - Corrigenda
- Comments to ISO-4355-2013 Part V.pdf: Drift / slide model for duo-pitched roofs - Comment
- Extreme snowdrift Multi-span Roof Ottawa 1971.pdf
- Comparison of Formulae Heat Conductivity and Density Snow.pdf
- Roof Angle Function for Thermal Coefficient.pdf
Citations from these documents are not marked as such, because of the same authorship.
Literature: References and sources used generally and for Part V
- Apeland, K. (2000): Global and European Standardization, ISO and CEN. In: Snow Engineering,
Recent Advances, Proceedings of the 4rd International Conference on Snow Engineering,
Trondheim, Norway, pp. 157-160.
 35
- Bras, J. (2014): Japan: Greenhouses buckle under the weight of meters of snow. Available online:
http://www.hortidaily.com. Downloaded 2014-07-03.
- Chiba,T., Tomabechi, T, Takahashi, T. (2012): Study on evaluation of snow load considering roof
snow-slide on gable roofs. In: 7th International Conference on Snow Engineering, Fukui, Japan,
pp. 231-241.
- DaGaetano, A.T., O`Rourke, M.J. (2003): A Climatological Measure of Extreme Snowdrift
Loading on Building Roofs. In: Journal of Applied Meteorology, Volume 43, pp. 134-144.
- Delpech, P., Guilhot, J. (2008): Quantitative assessment of snow load on complex and extended
roof shapes. In: 6th International Conference on Snow Engineering, Whistler, Canada, Session 9.
- Dufresne de Virel, M., Delpech, P, Sacre, C (2000): Wind tunnel investigations of snow loads on
buildings. In: Snow Engineering, Recent Advances, Proceedings of the 4th International
Conference on Snow Engineering, Trodheim, Norway, pp. 171-178.
- Fierz, C., Armstrong, R.L., Durand, Y., Etchevers, P., Greene, E., McClung, D.M., Nishimura,
K., Satyawali, P., Sokratov, S.A. (2009): The International Classification for Seasonal Snow on
the Ground. IHP-VII Technical Documents in Hydrology No. 83, IACS Contribution No. 1, Paris,
France.
- Fingerloos, F. & Schwind W. (2019): Zur Neuausgabe es Nationalen Anhangs DIN EN 1991-1-
3/NA “Schneelasten” in 2019-04. In: Bautechnik 96, 2019-4, Berlin, Germany, pp.352-359.
- Gandemer, J., Palier, P., Boisson-Kouznetzoff, S. (1997): Snow simulation within the closed
space of the Jules Verne Climatic Wind Tunnel. In: Snow Engineering, Recent Advances,
Proceedings of the 3rd International Conference on Snow Engineering, Sendai, Japan, pp. 347-
352.
- Grammou, N. (2015): Ermittlung von geometrischen Formbeiwerten für Flachdächer mit
aufgeständerten Photovoltaikanlagen anhand eines physikalischen Analogiemodells. Doctoral
Thesis, TU Darmstadt, Germany.
- Hochstenbach, F.M., Irwin, P.A., Gamble, S.L. (): Parametric studies of unbalanced snow loads
on arched roofs. In: Rowan William Davies & Irwin, Inc. 650 Woodlawn Road West, Guelph,
Ontario, Canada; [email protected].
- Høibø, H. (1988): Snow Load on Gable roofs - Results from Snow Load Measurements on Farm
Building in Norway. 1st International Conference on Snow Engineering, Santa Barbara,
California, USA, CRREL Special Report No. 89-6, pp. 95-104.
- Irwin, P.A., Gamble, S.L., Taylor, A.D. (1995): Effects of roof size, heat transfer and climate on
snow loads. Studies for the 1995 NBC, Canadian Journal of Civil Engineering, 22, 9, pp. 770-
784.
- Irwin, P.A. (2016): Prediction of snow loads: past, present and future. In: 8th International
Conference on Snow Engineering, Nantes, France, pp. 206-213.
- Isyumov, N., Mikitiuk, M. (2008): Sliding Snow and Ice from Sloped Building Surfaces – Its
Prediction, Potential Hazards and Mitigation. In: 4th International Conference on Snow
Engineering, Whistler, Canada 2008.
- Ito, T., Tomabechi, T., Mihashi, H. (1997): Characteristics of roofing materials as related to
adfreezing of snow. In: Snow Engineering, Recent Advances, Proceedings of the 3rd International
Conference on Snow Engineering, Sendai, Japan, pp. 209-214.
 36
- Jelle, B.P. (2012): The Challenge of Removing Snow Downfall on Photovoltaic Solar Cell Roofs
in order to Maximize Solar Energy Efficiency – Research Opportunities for the Future.
Department of Materials and Structures, SINTEF Building and Infrastructures, Trondheim,
Norway.
- Li, L, Pomeroy, J.W. (1997): Estimates of Threshold Wind Speeds for Snow Transport Using
Meteorological Data. In: Journal of Applied Meteorology, Volume 36, March 1997, pp. 205-213.
- Lutes, D.A. (1970): Snow loads for the design of roofs in Canada. Proceedings of the Western
Snow Conference, Victoria, British Columbia, Canada, pp. 61-67.
- Meløysund, V., Lisø, K.R., Hygen, H.O., Høiseth, K.V, Leira, B. (2006): Effect of wind exposure
on roof snow loads. In: ScienceDirect, Building and Environment 42 (2007) pp. 3726-3736.
- Meløysund, V. (2010): Prediction of local snow loads on roofs. Doctoral Thesis, Norwegian
University of Science and Technology, Trondheim, Norway.
- Mihashi, H., Takahashi, T. (1996): Japanese recommendations 1993 for snow loads on buildings.
In: 3th International Conference on Snow Engineering, Sendhai, Japan 1996.
- Moriyama, H., Sase, S., Okushima, L., Ishi, M. (2015): REVIEW: Which Design Constraints
Apply to a Pipe-Framed Greenhouse? From Perspectives of Structural Engineering,
Meteorological Conditions and Wind Engineering. JARQ 49 (1), pp.1-10.
- O`Rourke, M., Koch, P., Redfield, R: (1983): Analysis of roof snow load case studies: Uniform
loads. Cold Regions Research & Engineering Laboratory, CRREL Report 83-1.
- O´Rourke, M, Auren, M. (1997): Unbalanced snow loads on gable roofs. In: Snow Engineering,
Recent Advances, Proceedings of the 3rd International Conference on Snow Engineering, Sendai,
Japan, pp.201-208.
- O`Rourke, M., Cocca, J. (2016): Improved design relations for roof snow drifts. In: 8th
International Conference on Snow Engineering, Nantes, France, pp. 146-150.
- O`Rourke, M., Potac, J., Thiis, T.K. (2016): Analysis of snow drifts on arch roofs. In: 8th
International Conference on Snow Engineering, Nantes, France, pp. 221-227.
- Otsuka, K., Homma, Y. (1992): Removal of snow from membrane structures. In: Proceedings of
the 2nd International Conference on Snow Engineering Santa Barbara, USA, pp. 263-274.
- Potac, J. (2013): Field measurements and numerical simulations of snow transport and deposition
around structures and into ventilations. PhD Thesis Norwegian University of Live Sciences,
Norway.
- Paeck, S.-Y., You, Y.-Y., You, K.-P., Kim, Y.-M. (2013): A study on the snow load of arched
houses using wind tunnel simulation. The 8th Asia-Pacific Conference on Wind Engineering,
Chennai, India, pp. 273-280.
- Pomeroy, J.W., Brun, E. (2012): Physical Properties of Snow. Available online:
http://www.inscc.utah.edu. Downloaded 2013-03-24.
- Sakurai, S., Abe, O., Joh, O. (2012): Practical method for predicting snow accumulations on roofs
based on wind pressure characteristics. In: 7th International Conference on Snow Engineering,
Fukui, Japan, pp. 463-475.
- Sandvik, R., Apeland, K. (1996): A comparison of the new ISO 4355 with CEN ENV 1991-2-3.
IABSE report 74, 1996, pp. 101-108.
 37
- Scarascia-Mugnozza, G., Castellano, S., Roux, P., Gratraud, P., Palier, P., Dufresne de Virel, M.,
Robertson, A. (2000): Snow distributions on greenhouses. In: Snow Engineering: Recent
Advances & Developments, 4th International Conference on Snow Engineering Trondheim,
Norway, pp. 265-274.
- Scarascia-Mugnozza, G.S., Castellano, S., Roux, P., Gratraud, P., Palier, P., Robertson, A., Palier,
P. Dufresne de Virel, (2003): Snow load shape coefficients for vaulted roof greenhouse structures.
In: Riv. Di Ing. Agr. 2, pp. 3-12.
- Tabler, R.D. (2003): Controlling Blowing and Drifting Snow with Snow Fences and Road Design.
Final Report, National Cooperative Highway Research Program Transportation Research Board
of the National Academies.
- Taylor, D.A. (1979): A survey of snow loads on roofs of arena-type buildings in Canada. In:
Canadian Journal of Civil Engineering, Vol. 6, No. 1, March 1979, pp. 85-96.
- Taylor, D.A. (1980): Roof snow loads in Canada. In: Canadian Journal of Civil Engineering, Vol.
7, No. 1, pp. 1-18.
- Taylor, D.A. (1983): Sliding snow on sloping roofs. Canadian Building Digest, National Research
Council of Canada.
- Thiis, T.K., O`Rourke, M. (2012): A model for the distribution of snow load on gable roofs. In:
7th International Conference on Snow Engineering, Fukii, Japan, pp. 260-271.
- Thiis, T.K. Ramberg, J.F. (2008): Measurements and numerical simulations of development of
snowdrifts on curved roofs. In: 4th International Conference on Snow Engineering, Whistler,
Canada 2008.
- Thiis, T.K., Potac, J, Ramberg, J.F. (2009): 3D numerical simulations and full-scale
measurements of snow depositions on a curved roof. In: 5th European & African Conference on
Wind Engineering, Florence, Italy.
- Tominaga, Y. (2017): Computational fluid dynamics simulation of snowdrift around buildings:
Past achievements and future perspectives. Cold Regions Science and Technology,
http://dx.doi.org/10.1016/j.coldregions.2017.05.004, pp. 1-13.
- Tomabechi, T., Ito, T., Takakura, M. (1997): Snow Sliding on a Membrane Roof. In: Snow
Engineering, Recent Advances, Proceedings of the 3rd International Conference on Snow
Engineering, Sendai, Japan, pp. 341-344.
- Tory, J.R. (1992): The Codification of European Snow Loading. In: 2nd International Conference
on Snow Engineering, Santa Barbara, California, pp. 453-464.
- Yan, K., Zhou, X., Kasperski, M. (2010): Wind drifted snow load simulation on gable roof. 5th
International Symposium on Computational Wind Engineering (CWE 2010), Chapel Hill, USA.
- Zhou, X. (2013): Simulation method of sliding snow load on roofs and its application in some
representative regions of China. In: Natural Hazards 2013, 67, pp. 295-320.
- Zhou, X., Li, J., Huang, P., Gu, M., Sun, L. (2016): A new method of predicting slide snow load
for sloped roofs. In: 8th International Conference on Snow Engineering, Nantes, France, pp. 197-
183.