Engineering Structures 301 (2024) 117357

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Short communication

The Climatic Action Uncertainty Chain

V. Picozzi a ,∗, F. Landi b , A.M. Avossa a , P. Croce b , P. Formichi b , F. Ricciardelli a
a
Department of Engineering, University of Campania ‘‘L. Vanvitelli’’, Italy
b
Department of Civil and Industrial Engineering, University of Pisa, Italy

ARTICLE INFO ABSTRACT

Keywords: Inspired by the Alan G. Davenport Wind Loading Chain, first a uniform format for the modelling of climatic
Structural codes actions of snow, wind, thermal actions and atmospheric icing actions is presented in the form of a Climatic
Structural reliability Action Loading Chain (CALC). The definition of the complete probabilistic loading chain is discussed providing
LRFD
the basis for a more reliable calibration of load partial safety factors. Finally, the concepts are applied to
Partial factors
an example showing that the value of the partial factor strongly depends on the assumptions made during
Eurocodes
calibration, and that as a consequence the value of the reliability index may strongly differ from the target
one.

1. Introduction calibrating load Partial Safety Factors (PSFs). There derives a Climatic
Action Uncertainty Chain (CAUC), that allows investigating the role of
Based on an idea developed over the years, in 1980’ Davenport the uncertainties embedded in each component of the action. Finally,
presented the Wind Loading Chain [1]. On July 12, 2011, during the an example is presented highlighting the role of uncertainty in the
thirteenth International Conference on Wind Engineering, the General calibration of load PSFs.
Assembly of the International Association for Wind Engineering (IAWE) The proposed framework should help clarifying the role of the
approved the name ‘‘Alan G. Davenport Wind Loading Chain’’ (AGD- different factors appearing in the climatic load equations, giving them
WLC) as the most appropriate term to describe the way in which wind a unitary formulation. On the other hand, it should help locating
loads and wind induced response are generated from the atmospheric
uncertainty sources, therefore indicating actions to be taken in order
circulation Isyumov [2]. The idea of picturing the load and response
to improve accuracy of uncertainty models.
generation process as a chain derived from the fact that the reliability
of the results depend on the strength of the weakest link. This idea has
been driving research, teaching and practice in Wind Engineering for 2. Physical modelling of climatic actions
about half a century.
In this paper, it is shown that the idea of loading chain can be 2.1. General framework
extended to other climatic actions, namely snow loads, thermal actions
and atmospheric icing actions, and that the format used by Codes For the purpose of reliability analyses, the effect 𝐸 of the climatic
to derive characteristic values of the actions already complies with action 𝐴 can be evaluated as:
such description. In particular, reference will be made to the relevant
Eurocode 1 parts, both the existing versions and those currently under 𝐸 = 𝐴 ⋅ 𝜒𝐸 ⋅ 𝑒 (1)
development within Mandate M/515 of the European Commission to
CEN: EN 1991-1-3:2003 [3] and prEN 1991-1-3:2022 [4] for snow in which 𝑒 is the considered effect deriving from a unit action, and
loads, EN 1991-1-4:2005 [5] and prEN 1991-1-4:2022 [6] for wind 𝜒𝐸 is the quantity that transforms the intensive action 𝐴 into an
actions, EN 1991-1-5:2003 [7] and prEN 1991-1-5:2022 [8] for thermal extensive action for the structural model. For example, wind pressure
actions, and ISO 12494:2017 [9] and prEN 1991-1-9:2022 [10] atmo- is transformed into a distributed load acting on a beam or into a point
spheric icing of structures. The same approach applies to other Codes, load acting in a node of a framed structure by assuming 𝜒𝐸 as a
provided the appropriate definition is given to the terms appearing in reference length or a tributary area, respectively.
the equations. The climactic action 𝐴 can be defined as:
Furthermore, it is possible to show that to each Climatic Action
Loading Chain (CALC), an uncertainty chain is associated, which allows 𝐴 = 𝐴(𝐶, 𝑎1 , 𝑎2 , …) (2)

∗ Corresponding author.
E-mail address: vincenzo.picozzi@unicampania.it (V. Picozzi).

https://doi.org/10.1016/j.engstruct.2023.117357
Received 19 June 2023; Received in revised form 25 November 2023; Accepted 13 December 2023
0141-0296/© 2023 Elsevier Ltd. All rights reserved.
V. Picozzi et al. Engineering Structures 301 (2024) 117357

Fig. 1. Climatic Action Loading Chain.

where the 𝑎𝑖 coefficients transform the climatic variable 𝐶 into the the method of structural analysis. In particular, when the PSF format
climatic action 𝐴, taking into account the effects of the surroundings is used, the fifth link includes partial safety factors and combination
and of the structure geometry and mechanical properties. The climatic factors, transforming the representative value of the action into its
variable is to be understood as complying with the standards set by design value.
the World Meteorological Organization [11] and generally adopted by The structure of the CALC allows identifying the three components
structural loading Codes. Thus, the 𝑎𝑖 factors also transform the climatic of risk due to climatic actions: the Climatic Variable quantifies the basic
variable 𝐶 when non-standard conditions apply. For snow loads, as well Hazard, then amplified by the Effects of Surroundings, i.e. local effects;
as for wind and icing actions, Eq. (2) can be simplified into a product Geometric and Dynamic Interaction coefficients determine Vulnerabil-
chain: ity; Design Criteria account for the consequences of failure within the
considered limit state, thus determining Exposure.
𝐴 = 𝐶 ⋅ 𝑎1 ⋅ 𝑎2 ⋅ ... (3) In the bottom part of Fig. 1, the way the chain links interact with
The factors 𝑎𝑖 appearing in Eqs. (2) and (3) can be grouped into three each other in the reliability assessment is shown. In particular, the
categories: Exposure coefficients, Geometric Interaction coefficients, Climatic Action derives from interaction of the Climatic Variable with
the surroundings and the structure. Therefore, it is affected by the un-
and Dynamic Interaction coefficients. Exposure coefficients take into
certainties associated with the definition of the surroundings properties
account the effects of the environment surrounding the structure, the
and of the structural properties (mainly geometry). The Climatic Action
height from the ground, and the exposure to sun. Geometric Interaction
produces effects, that are transformed into equivalent static loads in
coefficients allow for the effects deriving from the interaction of the
the case of static behaviour or of Mild Dynamic Interaction. Climatic
climatic phenomenon with the structure. Finally, Dynamic Interaction
Action Effects are then combined with the effects of Other Actions, and
coefficients, where applicable, allow for vibrations induced by the
by assuming a model for Resistance, a Limit State Function (LSF) is
dynamic nature of the action.
defined. On the other hand, when Strong Dynamic Interaction takes
The climatic variable is random in nature and can be expressed as:
place, as it can happened in the case of the wind action, then specific
𝐶 = 𝐶o ⋅ 𝑏1 ⋅ 𝑏2 ⋅ ... (4) models are required to define the LSF.

in which 𝐶o is the fundamental value (usually associated with a return 2.2. Eurocode climatic action models
period of 50 years) of the climatic variable meeting standard condi-
tions, and the coefficients 𝑏𝑖 account for possible deviations from the The variables defining the components of the climatic action models
statistical definition of 𝐶o , including seasonality, directionality, and of Eurocode 1 are given in Table 1, following the classification of the
return period. CALC. In particular, the climatic variables are defined as:
Inspired by the AGD-WLC, Eq. (2) can be seen as a Climatic Action
Loading Chain, as sketched in Fig. 1. The last link of the chain is • the basic ground snow load 𝑠𝑏 = 𝑠𝑏,o ⋅ 𝑐𝑝𝑟𝑜𝑏
not explicitly included in the CALC; indeed, it represents the design • the basic velocity pressure 𝑞𝑏 = 0.5𝜌 ⋅ 𝑣2𝑏 deriving from the basic
criteria, i.e. the criteria adopted to assess the reliability of the structure, wind speed 𝑣𝑏 = 𝑣𝑏,o ⋅ 𝑐𝑝𝑟𝑜𝑏 ⋅ 𝑐𝑑𝑖𝑟 ⋅ 𝑐𝑠𝑒𝑎𝑠𝑜𝑛
including both the effects and the resistance models, the method used • the shade air temperature 𝑇𝑏
for the reliability assessment and the required safety level, as well as • the basic ice load 𝑖𝑏 = 𝑖𝑏,o ⋅ 𝑐𝑝𝑟𝑜𝑏 ⋅ 𝑐𝑑𝑖𝑟 ⋅ 𝑐𝑠𝑒𝑎𝑠𝑜𝑛

2
V. Picozzi et al. Engineering Structures 301 (2024) 117357

Table 1
Classification of the terms composing the climatic actions according to the Climatic Action Loading Chain.
Climatic action Climatic Exposure Geometric Dynamic Reference
variable interaction interaction
Snow 𝑠𝑏 𝑐𝑒 𝑐𝑡 , 𝜇𝑖 1 [3,4]
Wind 𝑞𝑏 𝑐𝑒 , 𝑐𝑠 𝑐𝑝 𝑐𝑑 [5,6]
Thermal 𝑇𝑏 𝑇0 , 𝑇𝑜𝑢𝑡 𝑇𝑖𝑛 , 𝑇𝑜𝑢𝑡 1 [7,8]
Atmospheric Icing 𝑖𝑏 𝑐ℎ 𝑐𝑜𝑏𝑗𝑒𝑐𝑡 , 𝑐𝑜𝑟𝑖𝑒𝑛𝑡 1 [9,10]

where 𝑐𝑝𝑟𝑜𝑏 accounts for return periods different from 50 years, 𝑐𝑑𝑖𝑟 and (iii) uncertainties in the parameters of the physical and statistical
accounts for directionality, and 𝑐𝑠𝑒𝑎𝑠𝑜𝑛 accounts for seasonality. Notice models. Since they cannot be predicted, aleatory uncertainties cannot
that for thermal actions, the shade air temperature refers to both max- be avoided. Instead, model and statistical uncertainties are epistemic,
imum 𝑇𝑚𝑎𝑥,o and minimum 𝑇𝑚𝑖𝑛,o fundamental values, thus including therefore can be reduced [14]. In particular, they consist of a systematic
seasonality. component that can be corrected via a bias factor, and of a random
The effects of surroundings are accounted by exposure coefficients component requiring statistical treatment [15]. The randomness of
defined as: the climatic action derives from the aleatory variability, 𝐶(𝑡), of the
climatic phenomenon describing the climatic variable, 𝐶; limited to the
• for snow load, 𝑐𝑒 accounts for wind exposure, i.e. for possible
wind action, also the aleatory variability of the aerodynamic coefficient
removal or accumulation of snow on the construction due to wind
has to be considered, accounting for the high-frequency signature tur-
and its interaction with the construction and other obstacles
bulence [16]. On the other hand, the physical modelling of the climatic
• for wind action, 𝑐𝑒 accounts for the site exposure, i.e. for orog-
action and of each of its components, and the fitting of the statistical
raphy, surface roughness, and atmospheric turbulence, while 𝑐𝑠
models introduce epistemic uncertainties in the CALC.
accounts for the reduction of the action due to the lack of spatial
The recognition and the quantification of the two types of uncer-
correlation of wind speeds
tainty is essential in the calibration process of structural Codes, i.e. in
• for thermal action the difference 𝑇𝑜𝑢𝑡 − 𝑇o accounts for exposure the calibration of load factors [17–19]. Indeed, in the evaluation of
effects, 𝑇o being the temperature of the structure element at the structural reliability, results are strongly dependent on the stochastic
time of completion, and 𝑇𝑜𝑢𝑡 accounts for the orientation of the models assumed for the basic variables. However, the availability of ac-
surfaces of the sun-exposed elements curate probabilistic description is not enough for successful calibration
• for icing load, 𝑐ℎ accounts for the height above the ground of partial factors, as it is demonstrated that the choice of the strategy
Finally, geometric interaction coefficients are defined as: for calibration of partial factors strongly affects the final results.
In the case of climatic actions, three different levels can be con-
• for snow load, 𝑐𝑡 accounts for the effects on snow of the heat flux sidered in order to highlight the influence of the different sources of
through the roof, and 𝜇𝑖 accounts for the geometry of the snow uncertainty in the reliability problem:
cover, depending on roof geometry
• for wind action, 𝑐𝑝 accounts for the static interaction of the wind • Level 1, in which only the aleatory uncertainty in the Climatic
flow with the structure geometry, i.e. aerodynamics Variable is considered
• for thermal action the difference 𝑇𝑜𝑢𝑡 − 𝑇𝑖𝑛 accounts for the geo- • Level 2, in which also the epistemic uncertainties in the Climatic
metric interaction effects, 𝑇𝑖𝑛 being the temperature of the inner Action are considered
environment, in turn depending on thermal and heat transfer • Level 3, in which the Climatic Action Effects are evaluated ac-
properties, thus on building materials and their layering, thermal counting also for the model effect uncertainty
insulation, etc.; In details, Level 1 allows evaluation of the impact on the PSF of the
• for icing action, 𝑐𝑜𝑏𝑗 accounts for the uniformity of the ice accre- time-variant nature of the climatic variable, 𝐶. Level 2 allows evalua-
tion, and 𝑐𝑜𝑟𝑖𝑒𝑛𝑡 accounts for the structural member orientation, tion of the influence on the PSF of the complete climatic action model,
e.g. vertical or horizontal. 𝐴, including both time-variant and time-invariant load components.
The dynamic interaction coefficient 𝑐𝑑 applies only to wind action, Level 3 allows evaluation of the influence on the PSF of the climatic
due to its high frequency fluctuations interacting with the dynamics of action and action effect models, and lends itself to the definition of
the structure. Indeed, some dynamic effect could also arise from snow appropriate Limit State Functions, in which also the structure resistance
sliding from an upper roof on a lower roof. This has the nature of an is accounted for.
impact load and is not directly dealt with by the Eurocodes. Availability Calibration at levels 1 to 3 allows the use of factorized PSFs:
of an appropriate model to quantify such action, would allow defining 𝛾𝐴 = 𝛾𝐶 ⋅ 𝛾𝑎 ⋅ 𝛾𝐸 (5)
a dynamic interaction also in the case of snow.
in which:
3. Statistical modelling of climatic actions • 𝛾𝐶 quantifies the uncertainty in the climatic variable, thus it is
related to the climatology of the site of construction;
In Section 2 the CALC is introduced for the evaluation of the design • 𝛾𝑎 accounts for the uncertainty in the 𝑎𝑖 coefficients in Eq. (2),
climatic actions and effects on structures. Both the climatic variable and applied to the climatic variable;
the coefficients of Eq. (2) are either specified fractiles of their statistical
• 𝛾𝐸 accounts for the error in the evaluation of the effects of the
distribution, or nominal values. To the aim of calibration of partial
climatic action.
factors to be used within PSF format, the uncertainties associated with
each link of the chain have to be assessed. Denoting with 𝛾𝐴(𝑖) the factor calibrated at the 𝑖th Level, there follows
In reliability analysis, the Bayesian approach is used incorporating that 𝛾𝐶 = 𝛾𝐴(1) , 𝛾𝑎 = 𝛾𝐴(2) ∕𝛾𝐶 and 𝛾𝐸 = 𝛾𝐴(3) ∕(𝛾𝐶 ⋅ 𝛾𝑎 ).
both aleatory and epistemic uncertainties [12]. According to Der Ki-
ureghian and Ditlevsen [13], seven sources of uncertainty can be 4. Illustrative example
identified. In the case of climatic actions, they can be grouped into:
(i) aleatory uncertainties that can be directly measured, (ii) model The concepts discussed in the previous sections are applied to a low-
uncertainties, related to both the physical and probabilistic models, rise building, which resistant system is composed of the steel portal

3
V. Picozzi et al. Engineering Structures 301 (2024) 117357

Table 3
Results of calibration for the illustrative example.
( ) ( )
Scenario 𝛾𝐶 𝛾𝑎 𝛾𝐴,(2) 𝛾𝐸 𝛾𝐴,(3)
W1 1.70 1.48 (2.51) 1.04 (2.62)
W2 2.22 1.34 (2.98) 1.04 (3.10)
S1 1.70 1.51(2.57) 1.01(2.60)
S2 2.22 1.37(3.04) 1.01(3.07)

accounted for; on the other hand, no dynamic effects arise on the low
Fig. 2. Main resistant system of the example low-rise building. rise structure, therefore the dynamic coefficient is equal to one and no
uncertainty is associated with this value.
Two scenarios are considered also for snow load. In detail, Scenario
frame shown in Fig. 2. To highlight the role of uncertainties in the 1 (S1) refers to a small value of the coefficient of variation (𝑉 = 0.15)
calibration process of the PSFs, wind and snow actions are considered.
of the extremes of the ground snow load, e.g. the case of Norway.
The Design Value Method is applied by assuming a target reliability
Instead, Scenario 2 (S2) refers to a large value of the coefficient of
index 𝛽𝑡 = 3.8 for a reference period of 50 years and a sensitivity factor
variation (𝑉 = 0.25) e.g. the case of Germany. In addition, a coefficient
for action effects 𝛼𝐸 = 0.7, as suggested by Eurocode 0 [20].
When a climatic action is applied to the example structure, then its of variation 𝑉 = 0.40 is considered for the roof conversion factor [25].
members are required to resist its effects. For example, snow acts as a This value derives from measurements of snow load on flat roofs
gravity load 𝑠 applied to the roof, and its main effect are axial forces 𝑁𝑠 collected in Switzerland, Italy, United Kingdom, and Germany within
in the columns and in the members of the truss beam. Consequently, the European Snow Load Research Project [26]. A similar value is
Eq. (1) can be written as: observed in the US [27].
For model effects uncertainty 𝜃𝐸 , a coefficient of variation 𝑉 = 0.10
𝑁𝑠,𝑖 = 𝑠 ⋅ 𝑎𝑠 ⋅ 𝑛𝑠,𝑖 (6)
for moments and 𝑉 = 0.05 for axial forces, were considered, respec-
where 𝑎𝑠 is the snow-loaded area, and 𝑛𝑠,𝑖 is the snow load influence tively [28].
factor for the considered structural member. To apply the DVM at the three levels of calibration, the empirical
Wind acts as a pressure 𝑤𝑗 on the vertical walls and roof, which distribution function of the load effect is obtained by Monte Carlo
intensity depends on the location of the 𝑖th surface. One of its effects simulations. The Cumulative Distribution Functions (CDFs) are shown
is the base bending moment 𝑀𝑤 of the columns, for which Eq. (1) can
in Fig. 3 where the reliability index 𝛽 is used instead of the probability
be written as:
of failure. As expected, more uncertainty is included in the model,
𝑛𝑠𝑢𝑟𝑓
∑ larger PSFs are obtained. As a consequence, accounting only for the
𝑀𝑤,𝑖 = 𝑤𝑗 ⋅ 𝑎𝑤,𝑗 ⋅ 𝑚𝑤,𝑖𝑗 (7)
𝑗=1 time-variant uncertainty in the calibration, i.e. aleatory uncertainty in
velocity pressure and ground snow load, results in unsafe PSFs.
where 𝑚𝑤,𝑖𝑗 is the wind action influence factor for the 𝑖th column due
to the 𝑗th loaded area, 𝑎𝑤,𝑗 . The results of calibration are reported in Table 3 showing that for
The statistical properties of the effects depend on the variability the considered scenarios the larger contribution to the PSFs derives
of the two climatic actions, i.e. on their statistical model, and on the from the climatic variable. In fact, the difference between the two
uncertainty related to the evaluation of the effects. The models used in scenarios is large: for both wind action and snow load, increasing
the example are summarized in the Table 2. the coefficient of variation from 0.15 to 0.25 makes the partial safety
For the wind action, two scenarios are considered for aleatory factor 𝛾𝐶 grow by 30%. Such difference reduces to approximately 18%
uncertainty. Scenario 1 (W1) refers to a small value of the coefficient of when considering uncertainty in time-invariant component of the loads.
variation of the extremes of the mean velocity pressure; this is indeed Finally, it is shown that the uncertainty in modelling climatic action
the case of Germany, for which a coefficient of variation 𝑉 = 0.15 effects has less influence on the calibration of the partial safety factors,
is calculated for the 50-year return velocity pressure [21]. Scenario with an increase of 4% when considering the bending moment in the
2 (W2) refers to a larger value of the coefficient of variation of the
columns of the example building due to wind action, and of 1% when
extremes of the mean velocity pressure; it is the case of Italy, for
considering the axial force due to snow load.
which a coefficient of variation 𝑉 = 0.25 is calculated for the 50-years
return velocity pressure [18,22]. The uncertainty in the aerodynamic Compared with the PSF 𝛾𝐸 = 1.5 adopted by the Eurocodes for
coefficient 𝑐𝑝 is summarized in a coefficient of variation 𝑉 = 0.25 of actions and actions effects, the results obtained show that a much
the 10-min maxima of the pressure coefficient, as suggested by [23,24]. larger value would be needed in order to reach the target reliability
For the sake of simplicity, no uncertainty in the exposure coefficient is level. Clearly, the PSFs in Table 3 are related to the assumption of
the sensitivity factor 𝛼𝐸 = 0.7 within the DVM, whereas it is known
that such value can change depending on design situation and on
Table 2 the uncertainty model chosen for resistance. In order to obtain a PSF
Statistical models. 𝛾𝐸 = 1.5, it is observed from Fig. 3 that a sensitivity factor of about 0.25
Variable Distribution Scenario 1 Scenario 2 shall be used for Scenario 1 and of about 0.20 for Scenario 2. In Fig. 4,
Wind Action, 𝑤 the sensitivity factor 𝛼𝐸 is shown as a function of the ratio between
Velocity Pressure EVI 𝑉 = 0.15 𝑉 = 0.25
the standard deviations of resistance, 𝜎𝑅 , and of the action effect, 𝜎𝐸 ,
Pressure Coefficient EVI 𝑉 = 0.25 𝑉 = 0.25
when both 𝑅 and 𝑆 are assumed as Normal distributed or, at least,
Snow Load, 𝑠
Ground Snow Load EVI 𝑉 = 0.15 𝑉 = 0.25 when they can be approximated as such at the design point. The figure
Roof Conversion Factor LN 𝑉 = 0.40 𝑉 = 0.40 shows that the condition 𝛼𝐸 = 0.2 is obtained when the amount of
Model Effect Uncertainty, 𝜃𝐸 random uncertainty in the resistance variable is roughly 5 times larger
Moments in frame LN 𝑉 = 0.10 𝑉 = 0.10 than that of the action effect. This condition is never met in practice,
Axial forces in frame LN 𝑉 = 0.05 𝑉 = 0.05
where the ratio 𝜎𝑅 ∕𝜎𝐸 is commonly less than 1 or close to it.

4
V. Picozzi et al. Engineering Structures 301 (2024) 117357

5. Conclusions

Inspired by the Alan G. Davenport Wind Loading Chain, a consistent

format is presented for climatic actions, which reflects the structure of
the equations given by most of the National and International loading
codes, and in particular by Eurocode 1. Associated with the loading
chain is an uncertainty chain, which helps classifying the sources of
randomness that contribute to the evaluation of load Partial Safety
Factors. A simple example is also presented, aimed at showing how
calibration of the load PSF is affected by the uncertainties incorporated
in the loading model. For wind actions and snow loads a PSF only
slighter larger than the value adopted by the Eurocodes is found, when
only a moderate variability of the climatic variable is incorporated in
the calculations; such value increases by 30% when a larger variability
is given to the climatic variable. A further increase in the range of 34%
to 51% is found, when also the uncertainty in the exposure, geometric
and dynamic interaction coefficients is considered; this brings the load
PSFs to much larger values than those usually found in codes. Finally,
it is found that the uncertainty in the evaluation of the effects of the
climatic action only little affects the calibration of the load PSFs, with
differences in the order of 1% to 4%.
The illustrative example is limited to the reliability assessment and
calibration of PSFs including one design situation for each investigated
climatic action. However, a portfolio of design situations should be used
in assessing PSFs for Code calibration purposes. It is believed that the
proposed framework could be of use as a reference when modelling
climatic actions uncertainty and when calibrating PSFs for climatic
actions. In doing this, the DVM has the advantage of the ease of use; on
the other hand, full probabilistic analyses are also required to validate
the results obtained.

CRediT authorship contribution statement

V. Picozzi: Writing – original draft, Visualization, Software,

Methodology. F. Landi: Writing – original draft, Validation,
Investigation. A.M. Avossa: Writing – review & editing, Supervision,
Conceptualization. P. Croce: Writing – review & editing, Supervi-
Fig. 3. Results of the Partial Safety Factors calibration for the example building: (a) sion, Conceptualization. P. Formichi: Writing – review & editing,
wind action and (b) snow load.
Supervision, Conceptualization. F. Ricciardelli: Writing – review &
editing, Supervision, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial

interests or personal relationships that could have appeared to
influence the work reported in this paper.

Data availability

Data will be made available on request.

Acknowledgements

The authors of this paper are active within CEN TC250, CEN
TC250/SC10 and CEN TC250/SC1. Fruitful and continuous exchange
of opinions with Colleagues has contributed to the development of
Fig. 4. Sensitivity factors 𝛼𝐸 for action effects and 𝛼𝑅 for resistance in the case of the ideas that are presented in this paper.
Normal distributed variables.

5
V. Picozzi et al. Engineering Structures 301 (2024) 117357

References [15] Ang AH-S, Tang WH. Probability concepts in engineering planning and design -
volume II: decision, risk, and reliability. New York: John Wiley & Sons; 1984.
[1] Davenport AG. The reliability and synthesis of aerodynamic and meteorological [16] Picozzi V, Malasomma A, Avossa A, Ricciardelli F. The relationship between wind
data for wind loading. In: Thoft-Christensen P, editor. Reliability theory and pressure and pressure coefficients for the definition of wind loads on buildings.
its application in structural and soil mechanics. Dordrecht: Springer; 1983, p. Buildings 2022;12(2).
315–35. [17] Picozzi V. Uncertainties in wind action: Modelling and code calibration [Ph.D.
[2] Isyumov N. Announcement of the Alan G. Davenport wind loading chain. thesis], University of Campania ‘‘Luigi Vanvitelli’’; 2023.
2013, http://www.iawe.org/about/Wind_Loading_Chain.pdf. [Online; accessed [18] Picozzi V, Ricciardelli F, Sørensen J. Calibration of partial factors for wind
June 2023]. action: an application to the Italian wind climate. In: Proceedings of the
[3] CEN. EN1991-1-3, eurocode 1: actions on structures - part 1-3: general actions 14th international conference on statistics and probability in civil engineering,
- snow loads. Brussels: European Committee for Standardization (CEN); 2003. ICASP14. 2023, URL http://hdl.handle.net/2262/103608.
[4] CEN. PrEN1991-1-3, eurocode 1: actions on structures - part 1-3: general actions [19] Picozzi V, Ricciardelli F, Sørensen J. Calibration of wind action combination
- snow loads. Brussels: European Committee for Standardization (CEN); 2020. factors from experimental data. In: Proceedings of the XVII conference of the
[5] CEN. EN1991-1-4, eurocode 1: actions on structures - part 1-4: general actions italian association for wind engineering, INVENTO2022. 2022.
- wind actions. Brussels: European Committee for Standardization (CEN); 2005. [20] CEN. EN1990, eurocode: basis of structural design. Brussels: European Committee
[6] CEN. PrEN1991-1-4, eurocode 1: actions on structures - part 1-3: general actions for Standardization (CEN); 2002.
- wind actions. Brussels: European Committee for Standardization (CEN); 2020. [21] Teichgräber M, Köhler J, Straub D. Hidden safety in structural design codes. Eng
[7] CEN. EN1991-1-5, eurocode 1: actions on structures - part 1-5: general actions - Struct 2022;257:114017.
thermal actions. Brussels: European Committee for Standardization (CEN); 2003. [22] Ballio G, Lagomarsino S, Piccardo G, Solari G. A first step towards the map
[8] CEN. PrEN1991-1-5, eurocode 1: actions on structures - part 1-3: general actions - of Italian extreme winds. Part 2: Results, ripercussion on standards, design
thermal actions. Brussels: European Committee for Standardization (CEN); 2020. implications. Costruzioni Met 1991;4:209–42.
[9] ISO. ISO 12494. Atmospheric icing of structures. Geneva: International [23] Davenport AG. The relationship of reliability to wind loading. J Wind Eng Ind
Organization for Standardization (ISO); 2017. Aerodyn 1983;13:3–27.
[10] CEN. PrEN1991-1-9, eurocode 1: actions on structures - part 1-9: general actions [24] Cook NJ, Mayne JR. A novel working approach to the assessment of wind loads
- atmospheric icing. Brussels: European Committee for Standardization (CEN); for equivalent static design. J Ind Aerodyn 1979;4:2984.
2020. [25] Croce P, Formichi P, Landi F. Probabilistic assessment of roof snow load and the
[11] WMO. Guide to instruments and methods of observation (WMO-No. 8). Geneva, calibration of shape coefficients in the eurocodes. Appl Sci 2021;11(7):149–64.
Switzerland: World Meteorological Organization (WMO) ; 2018. [26] Sanpaolesi L. Scientific support activity in the field of structural stability of
[12] Paté-Cornell ME. Uncertainties in risk analysis: Six levels of treatment. Reliab civil engineering works: Snow loads - final report: Phase 2. Technical Report,
Eng Syst Saf 1996;54:95–111. Commission of the European Communities, DGIII - D3; 1999, Department of
[13] Der Kiureghian A, Ditlevsen O. Aleatory or epistemic? Does it matter? Struct Saf Structural Engineering, University of Pisa.
2009;31:105–12. [27] O’Rourke MJ, Stiefel U. Roof snow loads for structural design. J Struct Eng
[14] Picozzi V, Akbaba A, Avossa A, Ricciardelli F. Correction of historical records to 1983;109:1527–37.
improve the reliability of design wind speeds. Eng Struct 2022. [28] JCSS. Probabilistic model code. Joint Committee on Structural Safety (JCSS);
2001.