=Paper=
{{Paper
|id=Vol-2805/paper19
|storemode=property
|title=Event-Based Approach to Multi-Hazard Risk Assessment
|pdfUrl=https://ceur-ws.org/Vol-2805/paper19.pdf
|volume=Vol-2805
|authors=Maryna Zharikova,Volodymyr Sherstjuk,Oleg Boskin,Iryna Dorovska
|dblpUrl=https://dblp.org/rec/conf/citrisk/ZharikovaSBD20
}}
==Event-Based Approach to Multi-Hazard Risk Assessment==
https://ceur-ws.org/Vol-2805/paper19.pdf
Event-Based Approach to Multi-Hazard Risk Assessment
Maryna Zharikova1[0000-0001-6144-480X], Volodymyr Sherstjuk2[0000-0002-9096-2582], Oleg
Boskin3[0000-0001-7391-0986] and Irina Dorovska4[0000-0001-9280-8098]
Kherson National Technical University, Berislav Road, 24, Kherson, Ukraine
1marina.jarikova@gmail.com, 2vgsherstyuk@gmail.com,
3aandre.lenoge@gmail.com, 4irina.dora07@gmail.com
Abstract. This work presents an event-based spatially-distributed dynamic mul-
ti-hazard risk model for the objects of critical infrastructure. The multi-hazard
spatially-distributed risk model is based on the five-level spatial model, as well
as the dynamic models of the socio-economic system, vulnerability, and event-
based scenario model of multi-hazards represented on macro and micro levels.
Each hazard on micro level can be represented as a sequence of events plunged
into a certain context, where each event can initiate scenarios describing the
multi-hazard dynamics. The risk for a certain object at a certain time point is a
combination of the following components: the object state, disaster threat, the
vulnerability of the object, and the potential damage. Thus, the area of interest
at a certain time point will be characterized by integrated dynamic spatially-
distributed assessments of the multi-risk in the conditions of multi-hazards.
Keywords: Multi-hazard risk, Model, Events, Scenario, Hierarchy, Critical in-
frastructure, Socio-economic infrastructure, Socio-economic system
1 Introduction
Economy and society in the globalized world are increasingly dependent on the relia-
ble availability of essential goods and services provided by technical and socio-
economic infrastructures. Industrial facilities and critical infrastructure are vulnerable
to the impact of hazards that can generate cascading effects. Different sectors of the
infrastructures are interdependent. Being under influence of hazards or multi-hazards
such interdependencies extend affected area and increase damage. Climate change
also gives rise to the increase in the frequency, intensity, spatial extent, and duration
of extreme events [1].
Disaster prevention and mitigation require analysing risks from hazards and multi-
hazards, as well as their cascading effects to critical infrastructure elements. A subject
matter of such analysis is not only single hazard but also chains of hazards. At that,
one specific event can trigger different possible paths of hazard chains. The effects
from hazards can also be cascading. Cascading effects are associated with the level of
vulnerability and interdependency of critical infrastructure objects being at risk.
The chains of events are usually represented using event-tree structures, also
called event trees [2] where the nodes are associated with the events, and the arcs are
Copyright © 2020 for this paper by its authors. This volume and its papers are published under
the Creative Commons License Attribution 4.0 International (CC BY 4.0).
associated with the conditional probability of the next event occurring given that the
previous event occurred.
2 Related works
Much has been done in the field of single- and multi-hazard risk analysis [3]. The
classical definition of risk maintains that risk is a probability of occurrence of an un-
wanted event multiplied by the amount of loss [4]. In disaster risk case, unwanted
event is disaster that can’t be represented as a single event, and disaster risk analysis
can’t be assessed using classical approach. There are several reasons for this. At first,
unwanted event (hazard or multi-hazard) is dynamic spatially-distributed process
spreading in uncertain conditions. At second, disaster risk is analysed to protect some
objects of critical infrastructure influenced by disaster that are also spatially-
referenced and can be interrelated [5-12].
In recent years a range of approaches has been developed to disaster risk analysis.
They are as following: quantitative (deterministic and probabilistic), indicator-based
approaches, risk matrix approaches, event-tree approaches [13], data mining ap-
proaches.
Quantitative deterministic approach allows considering only one individual hazard
(such as landslides, floods, wildfires, etc.) or a small subset of potential hazardous
events and can’t be applied to a wide range of hazards, as well as their interactions
and cascading effects. However, in real life most of the regions are prone to multiple
hazards that can lead to cascading effects. Quantitative probabilistic approach is that
risk is assessed quantitatively taking into account a given set of hazard scenarios and
the probabilities of their occurrence, at that each hazardous scenario is treated as an
undivided whole. However, probability is associated to frequency of hazards, and the
researcher faces an issue where the event of interest is quite rare. To cope with this issue
and to increase the representativeness of posterior statistical samples, large territorial
entities and big-time intervals (10-100 years) are considered [14]. In most cases,
probabilistic approach is based not on imitation of many thousands of events using
Monte Carlo method, which is connected with high computational complexity.
Indicator-based approach allows to carry out relative holistic risk assessment di-
vided into a number of components such as hazard, exposure, vulnerability and capac-
ity. The relative risk assessments don’t provide information on actual expected losses.
The risk matrix represents semi-quantitative approach to risk analysis focusing on
categorizing risks by comparative scores. Such matrix is made of classes of frequency
of the hazardous events and the expected losses. Risk is represented as a combination
of these two dimensions.
Existing event-tree approaches allow analyzing disaster chains. The nodes corre-
spond to hazards, and the links between nodes depict the situations when one hazard
causes another. The main drawbacks of existing event-tree approaches are as follows:
each hazard is treated as undivided whole without referencing to spatial locations.
Data mining approach to risk analysis work well in conditions of incompleteness,
inaccuracy, ambiguity and uncertainty of both the initial data and the rules for their
transformation and can be an important tool in finding the correlation or uncertainty
of risk factors [15]. In spite of flexibility of data mining methods, they are character-
ized by high computational complexity and can’t be used for risk analysis in real-time
decision making.
Currently, individual hazards and risks are analyzed and treated by disaster risk
managers separately, especially, natural, social, and technical risks are not combined.
A severe gap is a fact that critical infrastructure is often recognized as important but
treated only regarding its technical components without representation of the popula-
tion and its vulnerability. When the vulnerability is captured, even in multi-risk anal-
yses, it is done under static conditions, not in real-time. This emphasizes a knowledge
gap in understanding the dynamic interaction of destructive processes with their po-
tential receptors (the elements of CI, people) [16].
3 Materials and methods
Giving foregoing literature analysis, some gaps in the state of the art of multi-hazard
risk analysis can be distinguished. Individual hazards and risks are analyzed inde-
pendently. It’s necessary to accumulate the knowledge about dynamics of multi-
hazards, their interactions, and their cascade/simultaneous effects on CI elements.
Risk assessment is also usually represented as a static value. We propose to con-
sider risk as a spatially-distributed process that provides a more comprehensive un-
derstanding of the multi-hazard risk concept. Spatially-temporal approach in multi-
hazard risk analysis provides critical information on hazard areas, impact zones, and
location of populations and vulnerable infrastructure within hazardous area.
Thus, the objective of this paper is to develop a model of spatially distributed dy-
namic multi-hazard risks for CI elements on different time and spatial scales, includ-
ing cascading risks and risk-related processes driven by environmental and socio-
economic changes, based on the models of socio-economic system (SES), multi-
hazard dynamics, and vulnerability of CI elements.
3.1 Underlying models
Taking into account both temporal and spatial distributions of multi-hazard risks, the
comprehensive approach to multi-risk assessment proposed in this paper is based on
the dynamic models of the socio-economic system, vulnerability, multi-hazards, and
spatially-distributed risks.
SES is a complex dynamic system resulting from the interaction between people,
environment, technical and socio-economic infrastructures, which represent their
interdependencies.
Technical and socio-economic infrastructure (SEI) is a dynamic spatially-
distributed system of systems consisting of the elements important to the activity of
people and society.
Critical Infrastructure (CI) is a part of SEI containing elements which are essential
for the maintenance of vital societal functions. The damage to critical infrastructure,
its destruction or disruption by natural disasters, terrorism, criminal activity, or mali-
cious behavior, may have a significant negative impact on people's security.
3.2 Spatial model
All the above-mentioned models will be grounded on a multi-level spatial model
(Fig.1). The lower level represents a system of geographic coordinates. On the second
level, the spatial model is discretized by a grid of isometric cells, which makes it pos-
sible to model the dynamics of processes occurring within a certain territory. A cell is
a homogeneous object of minimal size that can be variable.
Fig. 1. Spatial model
On the third level, each considered territory can be divided into a finite set of disjoint
spatial objects (geotaxons) representing geo-referenced natural parts of the terrain
with homogeneous characteristics. Using this level, we can describe land-use objects
that have a spatial extent, such as fields, forests, ponds, etc.
The forth level represents an administrative structure of the considered territory
(municipalities, districts, provinces/regions, countries). Such hierarchy will be used to
help stakeholders (risk managers, decision-makers, etc.) be aware of threatened areas
and infrastructure elements being at risk from multi-hazards and make risk-informed
decisions to forecast, prevent, respond, mitigate, adapt to multiple hazards, their
chains and interactions.
On the fifth level, the area of interest is subdivided into zones representing homog-
enous areas with respect to the definite assessments of some indicators such as dan-
ger, threat, risk, etc.
All levels of the spatial model can be implemented based on a multitude of corre-
sponding layers within a Geo-Information System (GIS).
3.3 SES model
We will consider a novel SES model as a network of networks containing people,
technical and socio-economic infrastructures such as food supply, health service,
agriculture, forestry, energy supply, traffic, the economy as a whole, and others. Some
subsystems of SES can be organized hierarchically. All of its infrastructures consume
natural resources such as water, air, soil.
The proposed model of SES can reflect all kinds of possible relations between a
large variety of its elements. It will be grounded on the GIS-based spatial model of
the territory. Thus, all objects of SES can be referenced spatially using this model.
3.4 Vulnerability model
Since decision-makers are interested in the protection of some valuable elements of
SEI exposed to hazards and minimizing their risks, we need to use corresponding
models of the dynamic vulnerability of various elements of infrastructures against
hazards of different classes.
We consider the interaction of SEI elements in dynamics to take into account vari-
ous short-term (meteorological), mid-term, and long-term (climatic, migration pro-
cesses) impacts, as well as the influence of people on the vulnerability of the objects.
A property of vulnerability accumulation and the possible existence of recovery func-
tion should be also taken into account.
4 Event-Based Multi-Hazard Model
We propose to consider multi-hazard model on macro and micro levels. On the both
levels multi-hazard model is represented by even-tree [17, 18]. On the macro level,
the nodes of event tree are associated with the hazards as a whole, and the arcs repre-
sent the facts when one hazard causes another. Each node of event tree built on the
macro level can be represented as more detailed tree on the micro level. The trees on
the micro level represent the model of each hazard spreading within the grid of iso-
metric cells (the first level of spatial model) (Fig. 2).
Hazard 2 Hazard 4
Hazard 1
Hazard 3 Hazard 5
Fig. 1. Event
Fig. 2. Macro and micro levels of multi-hazard model
We propose to consider every occurrence of an observed hazardous process on the
micro level as a certain kind of dynamic case, so we can use a case-based approach to
accumulate and store the scenarios of dynamics of various hazards and multi-hazards
on the micro level, as well as their combination, and chains on the macro level. Each
case can be represented as a sequence of events plunged into a certain context, where
each event can initiate scenarios describing the multi-hazard dynamics.
One of the important properties of the models mentioned above is that any object
has its state available for the use of threat/risk assessment methods. This applies to
any building, any infrastructure component, any element of any group, or ever any
hierarchies within the models, including areas of any level of the spatial model.
Further, we will consider a generalized concept called “object” against all above-
mentioned. Thus, the object Oi has its state wtO at the time t represented by a sub-
i
set of attributes wtO = {aij ,...aim } . Suppose W = {W0 ,...WF } is an ordered set of the ob-
i
ject state classes. Clearly, a variation of the value of any attribute ak ⊆ AD of the ob-
ject Oi at the time t can change its class. We consider this is an event denoted by y ,
so that y : wtO → wtO+1 , where wtO ∈ W j , wtO+1 ∈ Wk , W j ,Wk ∈ W , and W j ≠ Wk .
i i i i
Thus, during the lifecycle, each object can pass through a sequence of different
classes of its states.
In our model, the event is a basic concept that reflects causality and is referenced
in space and time. Accordingly, each event has its “cause” influenced by a natural
phenomenon or an anthropogenic action that induce a hazard, and “effect” resulted in
a change of the state of some object.
Consider a threat as a result of the hazard materialization that leads to the event oc-
currence. Let us describe the threat τ k as a couple
τ k = tk , lk , ck , mk , (1)
where tk is a reference of τ k in time, lk is a reference of τ k in space, ck is a class of
τ k , and mk is its magnitude.
5 Spatially-Distributed Dynamic Risk Analysis
Risk should be assessed for each SEI object that has a spatial reference. To analyze
the dynamics of risk for a certain object, it is necessary to analyze state transitions for
the cell within which the object is located. Under the influence of multi-hazard, the
cell passes from one category of state to another. Each category of states can be at-
tributed to the degree of undesirability with regard to the decision-maker. Then risk
will be a likelihood of the transition from a less undesirable state to a more undesira-
ble one.
The risk Ri ( t ) for the object oi at the time point t is a combination of the fol-
lowing components: the object state, disaster threat, the vulnerability of the object,
and the potential damage [19]. The threat to an object results from the presence of
disturbance events. The threat can be expressed as a likelihood that a system will face
a particular type of disturbance event at a specific target location in a given period.
An object´s vulnerability to particular disturbance events determines the likelihood
that damage will occur if it takes place at a specific object location. Vulnerability is
determined by the nature of the system (e.g. technical infrastructure, socio-economic
infrastructure) and the set of baseline coping capacities that are already in place to
protect the objects. Consequences refer to the nature of the damage (e.g. injures or
death of people, property damage, economic damage) that will occur if the disturb-
ance succeeds. The consequences of a disturbance event can impact either people or
the functioning of the system (as damage and/or business interruption).
Thus, the area of interest will be characterized by integrated dynamic spatially-
distributed assessments of the multi-risk in the conditions of multi-hazard at the time
t: {Ri ( t ) ∀oi ∈ O* ( t )} .
6 The Results of the Research
The proposed models have been implemented in multi-hazard risk analysis frame-
work using Python programming language, as well as the framework Django, its GIS
extension GeoDjango, DBMS PostgreSQL, and geospatial extension PostGIS to im-
plement a GIS-based risk management environment. The event-based model, event
structures, and hierarchies are based on the double indexed lists and provide enough
performance of the framework for the multi-hazard risk assessment.
The proposed model and framework can be used for risk-informed decision sup-
port within spa-tially distributed SES of any level of complexity in conditions of spa-
tially distributed destructive processes (such as flood, drought, epidemic spread, etc.)
and their cascades under uncertainty.
The framework has been approbated on the simulated area covering Okeshky
Sands, Kherson region, Southern Ukraine. Oleshky Sands is the largest expanse of
sand in Ukraine and the second in Europe. It’s situated near the Dnipro river and the
coast of Black Sea. In XX century moving sands was limited by planting artificial
coniferous forest around sandy areas. Although a relatively small sandy steppe, the
Oleshky Sands have sand and dust storms. Giving the fact, that at summer air temper-
ature rises to 40°C, in summer this forest often catches fire. There is underground
water reserve that forms an indispensable part of local environment as a source of
fresh water [20].
Global warming leads to decrease in groundwater levels, forests are being affected
by invasions of insects, they also become more prone to forest fires. All these factors
can cause rapid destruction of forests in large areas, desertification of the territory,
and the revival of sand movement.
A five-level spatial model of the territory was built. The third level of spatial
model is shown in the Figure 3.
Fig. 3. Third level of Oleshky Sands spatial model
Figure 4 shows a fragment of the multi-hazard model on the macro level, connected to
the region level of spatial model, and a fragment of SES model.
The experiment has been conducted to evaluate query response time-varying the
cell size within the spatial model from 5 m to 50 m. The results of the experiment are
shown in Fig. 5.
The proposed model and framework make it possible to simulate emergencies that
begin with forest fires, and then the chain of consequences may include tornadoes,
sand and dust storms, floods, etc.
The results of the simulation experiment confirmed the adequacy of the proposed
event-based model and the efficiency of the framework; the developed framework
provides acceptable performance for the GIS-based multi-risk assessments.
Wood- Water supply Fishing
processing industry
industry
SES model
Multi-hazard model: macro level
Damage of Dust
forest by insect storms
invasion Forest
mortality
Forest
Global
fires Sand
warming
Decrease in storms
groundwater
level
Disappearance
of flood water Shallowing
the rivers
and lakes
Territorial model: region level
Lake
Sand
Forest
Fig. 4. Multi-hazard model
7 Conclusion
The proposed approach to multi-hazard risk analysis addresses the need of society to
minimize the negative effects of disasters and climate changes through adequate adap-
tation actions. Our approach is primarily targeting the value of CI elements but it also
takes into account the citizens’ health and wellbeing by making the connection be-
tween the infrastructure monetary value and its perceived value for the citizens.
The proposed approach will contribute towards a more resilient and more sustain-
able society based on the concepts of increased awareness, better preparedness, in-
formation enhancement, appropriate behavior during disasters. This will result in
fewer expenses related to fixing the hazards as well as in the improved quality of life
and health of the citizens.
Fig. 5. The simulation results
8 References
1. Climate change 2014: mitigation of climate change. In: Contribution of Working Group III
to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. IPCC,
Cambridge University Press (2014).
2. Zuccaro, G., De Gregorio, D., Leone, M.F.: Theoretical model for cascading effects anal-
yses. International journal of disaster risk reduction 30, 199-205 (2018). DOI:
10.1016/j.ijdrr.2018.04.019.
3. Ciurean, R., Gill, J., Reeves, H.J., O’Grady, S., Aldridge, T.: Review of Multi-hazards Re-
search and Risk Assessment. In: Engineering Geology & Infrastructure Programme, Open
Report OR/18/057 (2018).
4. Simmons, D.C., Corbane, C., Menoni, S., Schneiderbauer, S., Zschau, L.: Understanding
disaster risk: risk assessment methodologies and examples. In: Poljanšek, K., Marín Fer-
rer, M., De Groeve, T., Clark, I. (eds.) Science for disaster risk management: knowing
better and losing less Luxembourg. Publications Office of the European Union (2017).
5. Tachtsidou, K., Kirkos, E.: Risk analysis and decision support systems with data mining
techniques in custom administrators. In: Proc. of 3rd International Conference on quantita-
tive, social, biomedical & economic issues (ICQSBE2019). Athens, Greece (2019).
6. Kappes, M.S., Keiler, M., Glade, T.: From Single- to Multi-hazard Risk Analyses: a Con-
cept Addressing Emerging Challenges. In: Malet, J.P., Glade, T., Casagli, N. (eds.) Moun-
tain Risks: Bringing Science to Society, pp. 351–356. Strasbourg, France (2010).
7. Kunz, M., Hurni, L.: Hazard maps in Switzerland: state-of-the-art and potential improve-
ments. In: Proceedings of the 6th ICA Mountain Cartography Workshop. Lenk, Switzer-
land (2008).
8. Collins, T., Grinseki, S., Romo Aguilar, M.: Vulnerability to environmental hazards in the
Ciudad Juárez (Mexico) - El Paso (USA) metropolis: a model for spatial risk assessment in
transnational context. Applied Geography 29, 448 (2009). DOI:
10.1016/j.apgeog.2008.10.005
9. Fleming, K., Parolai, S., Garcia-Aristizabal, A., Tyagunov, S., Vorogushyn, S., Kreibich,
H., Mahlke, H.: Harmonising and comparing single-type natural hazard risk estimations.
Annals of Geophysics 59(2), So216 (2016). DOI: 10.4401/ag-6987
10. Carpignano, A., Golia, E., Di Mauro, C., Bouchon, S., Nordvik, J.-P.: A methodological
approach for the definition of multi-risk maps at regional level: first application. Journal of
Risk Research 12(3-4), 513 (2009). DOI: 10.1080/13669870903050269
11. Choine, M.N., O’Connor, A., Gehl, P., D’Ayala, D., Garcia-Fernández, M., Jiménez, M.,
Gavin, K., van Gelder, P., Salceda, T., Power, R.: A multi-hazard risk assessment method-
ology accounting for cascading hazard events. In: Proc. of 12th International Conference
on Applications of Statistics and Probability in Civil Engineering (ICASP12). Vancouver,
Canada (2015). DOI: 10.14288/1.0076192
12. Pescaroli, G., Alexander, D.: Critical infrastructure, panarchies and the vulnerability paths
of cascading disasters. Natural Hazards 82(1), 175–192 (2016). DOI: 10.1007/s11069-016-
2186-3
13. Van Westen, C.J., Quan Luna, B., Vargas Franco, R.D.: Development of training materials
on the use of geo - information for multi - hazard risk assessment in a mountainous envi-
ronment. Mountain risks : bringing science to society : proceedings of the Mountain Risks
International Conference, Firenze, Italy, 24-26 November 2010 / ed. by J.-P. Malet, T.
Glade and N. Casagli. - Strasbourg : CERG, ISBN 2-95183317-1-5. pp. 469-475 (2010).
14. Risk assessment and mapping guidelines for disaster management. Staff Working Paper,
SEC 1626 final of 21.12.2010. European Commission (2010). DOI: 10.2788/688605
15. Hyde, K., Dickinson, M.B., Bohrer, G., Calkin, D., Evers, L., Gilbertson-Day, J., Nikolet,
T., Ryan, K., Tague, C.: Research and Development Supporting Risk-based Wildfire Ef-
fects Prediction for Fuels and Fire Management: Status and Needs. International Journal of
Wildland Fire 22, 37–50 (2013). DOI: 10.1071/WF11143.
16. Sherstjuk, V., Zharikova, M.: Situation Diagnosis Based on the Spatially-Distributed Dy-
namic Disaster Risk Assessment. Shakhovska N., Medykovskyy M. (eds) Advances in In-
telligent Systems and Computing 1080, 453-472 (2020). DOI: 10.1007/978-3-030-33695-
0_31.
17. Sherstjuk, V., Zharikova, M.: Heterogeneous Hybrid Neural Network for Modeling Spa-
tially Distributed Destructive Processes. In: Proc. of the 15th Int. Conf. on ICT in Educa-
tion, Research, and Industrial Applications. Integration, Harmonization, and Knowledge
Transfer. Vol. I: Main Conference, pp. 81–97. Kherson, Ukraine (2019).
18. Sherstjuk, V., Zharikova, M.: Plausible Event-Tree Networks for Knowledge Representa-
tion in Real-time GIS-Based Decision Support Systems. CEUR Workshop Proceedings
2105 (2018).
19. Miller, C., Ager, A.A.: A Review of Recent Advances in Risk Analysis for Wildfire Man-
agement. International Journal of Wildland Fire 22(1), 1–14 (2013). DOI:
10.1071/WF11114.
20. Bogdanets, V.: Land cover dynamics of Oleshky Sands: time-series analysis 1987-2017.
Monitoring and protection of land 4, 71-75 (2017).