Methodology for Clustering Cities Affected

by Natural Disasters

Fabiana Santos Lima1, Daniel de Oliveira1, Mirian Buss Gonalves1

1
Department of Production and Systems Engineering, Federal University of Santa Catarina,
Florianopolis, Brazil

{fabiana.lima, daniel.oliveira}@posgrad.ufsc.br, mirianbuss@deps.ufsc.br

Abstract. This study aims to present a methodology to form clusters by

analyzing historical data of disasters in the state of Santa Catarina - Brazil,
using the k-means method as a tool for pattern analysis. It can therefore assist in
the strategic coordination, in the definition of priorities and in the share of
experiences between cities. Therefore, the proposed methodology aims to
identify similar regions in order to standardize and suggest a method of
prevention and, thus, improve and assist the processes of decision-making
regarding the events of Humanitarian Logistics. A computational experiment,
applying the proposed methodology, was performed and the obtained results are
presented and analyzed at the end.

Keywords: Humanitarian Logistics; Clustering; Natural Disasters

1 Introduction

Humanitarian aid agencies coordinate billions of dollars annually in the relief for
victims of natural disasters, civil conflicts and wars. Its main task is the timely
mobilization of funding and of assets of international donors and relief administration
for vulnerable beneficiaries in disaster sites around the globe. As such, logistics is
essential for their activities and strategic for their missions.
According to [1] in the year of 2010 there were more than 370 natural disasters
worldwide, with approximately 296,800 deaths and a total of 207 million people
affected, with a damage estimated at 109 billion dollars. Comparing the number of
natural disasters in 2010 such as droughts, earthquakes, high temperatures, floods,
earth movements, storms, volcanoes and fires with the average of the decade from
2000 to 2009 it was found that there was no growth on the number of natural disasters
in the world. Brazil was the 13th country most affected by natural disasters in 2008,
where nearly two million people were affected mainly by the action of rains and about
1.5 million people suffered from the disasters in the state Santa Catarina [2].
According to statistics from the Civil Defence of Santa Catarina state, only in the
months from January to March of 2011, there were recorded 152 events such as
flooding, erosions, pests, storm surges, gales and floods being the latter the
predominant event, with 143 occurrences.
 In all these cases the assistance to the disasters was difficult and complicated.
Many factors, such as the large number of actors involved in the humanitarian aid and
the lack of sufficient resources, contributed to the difficulties in the relief coordination
[3]. There are few success stories of coordination, so that coordination is still the
fundamental weakness of the humanitarian action [4]. The coordination can be
referred to the coordination of resource and information sharing, centralization of the
decisions, the realization of common projects, the regional division of tasks, or a
system based on partnerships. Studies by international humanitarian organizations
show the need to improve the effectiveness of the humanitarian response, ensuring
greater predictability, accountability and partnership. Researches related to disaster
management are being developed in order to decrease the difficulties in coordinating
the various types of disasters [5, 6, 7, 8, 9, 10, 11, 12, 13].
Given this context, this article aims to contribute to the Humanitarian Logistics
using a methodology to generating profiles based on historical events of natural
disasters by combining frequency of occurrences in municipalities in order to identify
similar regions to further assist in the strategic coordination in defining priorities and
sharing of experiences in certain regions. This will allow standardize a method for
prevention and response, helping the competent operational agencies in decision
making regarding the necessary training, investments in infrastructure, forecasts of
demand or inventory, and partnerships, in addition to enabling discover the
geographic range of disasters and relief assistance chain.
The paper is structured as follows. In the next section, we introduce some basic
concepts about Humanitarian Logistics and Disaster Management. In section 3 we
show the methodology for pattern analysis in Humanitarian Logistics and in section 4
we illustrate a implementing the proposed method. Section 5 concludes this paper and
provides an outlook on future works.

2 Humanitarian Logistics and Disaster Management

Humanitarian Logistics (HL) proposes the effective use of logistics concepts adapted
to the specific chain of humanitarian assistance. These concepts can be the big
difference in minimizing actions of improvisation, very common in these instances,
maximizing efficiency and response time to the emergency situation.
However, HL has to deal with special circumstances and enormous challenges
[14]. The specific feature in this context is the involvement of human life, which
depends on the efficient and effective performance of the response operations, which
means that in the humanitarian logistics, the aid must reach its destination accurately
and on time, always focusing in the relief of suffering and in the preservation of life.
Aspects linked to infrastructure, assistance centres location, resources allocation,
coordination of processes (people, information, and goods) are worth mentioning in
the logistic processes systematized in the humanitarian logistics.
To better understand the chain of humanitarian assistance is essential to understand
disaster management which can be defined as the set of activities designed to
maintain control over disaster and emergency situations and provide a framework to
help people at risk situations. This deals with situations that occur before, during and
after the disaster and the objectives are to prevent or reduce loss of human lives,
physical and economical suffered by individuals, society and by the country in
general, reduce the suffering of the people and accelerate the recovery of the affected
area [15, 11, 6, 13].

2.1 Phases and Cycles of disasters relief

The coordination of disasters involves three phases: pre-disaster (prevention,

mitigation and preparedness), focus of this research, response (warning, impact and
emergency response) and post-disaster (transition, rehabilitation and reconstruction).
In the pre-disaster phase occurs the prevention, mitigation and preparedness for
emergency situations and the response performance to the disaster is related to the
level of preparedness and to the methodologies used in this phase [11].
This work turns to the region of clustering and risk analysis proposed in the first
item of the Logistics Process Model of Relief in International Disasters (MLSCI) of
[5], which consists of three functions: (i) determine the profile of the length of each
disaster, based on the analysis of data obtained from the past disasters. The profile can
be identified through a historical analysis; (ii) identify threatened regions and main
threats, resulting in different scenarios regarding the impact of different types of
disasters, resources and projected needs; (iii) the combination of information obtained
from the performance of the first two functions allows creating profiles of resources
requirements, depending on the region as well as on the type of disaster. Thus,
different regions with similar levels of criteria will be gathered to form clusters.
When working with a clusters system, it is intended to assist in the strategic
coordination in defining priorities and in the share of experiences. A common
approach is the risk analysis, which takes advantage that each region or municipality
contributes with its knowledge and experience so that, finally, a common
understanding returned to the planning can be applied in several regions.

3 Methodology for Pattern Analysis in Humanitarian Logistics

In this section, we will be presented the proposed methodology for the clusters
formation, as well as the computational experiment for validation and verification of
the presented model, using for that real data on such natural disasters of the state of
Santa Catarina Brazil.
This was motivated due in a most specific bibliographical survey we found a
quantity with small representativeness of works using methods of cluster historical
data and k-means algorithm in humanitarian logistics operation of natural disasters.
We can emphasize papers presented by [16, 17, 18, 19, 20, 21] where, in a general
manner, utilize specific methods of grouping with for example k-means, fuzzy c-
means, hierarchical clustering, seeking to develop an analysis and categorize the data
in different events related to natural disasters.
It is worth emphasizing the work done by [22] where the main objective was to
suggest ways to predict demand for goods and services for relief operations in
humanitarian logistics. With the use of historical data related to humanitarian
assistance information was available order to understand and predict future demand
and identify the main standards in terms of needs and demand response. It is
noteworthy that the methodology proposed in this paper resembles the proposal by
[22] to search for patterns in major humanitarian logistics analysis using historical
data, however differentiates itself in the process and objective.
It should be noticed that the purpose of this work is related to the pre-disaster
phase, it may be a policy of preparation to be used by responsible organizations in
attendance to respond to disaster situations. It is considered that the same disaster
could reach different municipalities at the same time, requiring the coordination
between them for the development of actions of relief or aid. Moreover, the impact of
a disaster on a specific municipality may require the assistance of other municipality,
either in the form of rescue units, or on loan of equipments, expert assistance or
recollection of donations. The operations in these situations may have very specific
traits. They are characteristics that make the coordination of these operations very
difficult if there is not a standardized system, common for planning and implementing
of the actions.

4 Implementing the Proposed Method

There were used data of disasters occurred in the period between the years of 2000
and 2010 in the state of Santa Catarina. These data were collected through the
database of the CEPED - University Centre for Disaster Studies and Research at the
UFSC. This database used official documents such as AVADANS - Damage
Assessment Form; NOPRED - Preliminary Notification of Disaster; DECREE -
Municipal Decree; ordinances, damages reports and other considered as non official
such as the newspaper (CIVIL DEFENSE, 2011). These documents are being
referenced as recorded events in this work.
Initially the occurrences of natural disasters were separated according to the state's
municipalities. Then, as data of the CEPED included various types of disasters, there
were filtered from the databases only the occurrences of natural disasters, resulting in
3.247 selected events. There were considered the types of natural disasters as classify
the Secretary of National Civil Defence (CIVIL DEFENSE, 2011). Fig. 1 shows the
total of events presented in the selected period by natural disasters.
1 400
1.287
1 200
971
1 000

800

600

400 304
247 241
197
200

0
Winstorms Hails Floods Landslides and Droughts Others
Flash Floods

Figure 1. Sum of events recorded by type of disaster.

 As data from CEPED were already organized, it was not necessary pre-processing
or cleaning of the data before proceeding with the database clustering, such as
removal of incomplete data or elimination of data repetition. Thus, for the purpose of
this work, data were processed into a matrix containing the frequency of each type of
natural disaster by municipality. This matrix has become, then, in the database to be
exploited by the k-means algorithm aiming at the formation of clusters based on the
number of municipalities impacted by natural disasters in Santa Catarina.

4.1 Results and discussion

In general, clustering techniques calculate the matrix partition [ kj ]K n , where kj

represents the degree of belonging to j-th point x j , of the clustering Ck , k = 1, ..., K.
Thus, kj = 1 if the point x j belongs to Ck and kj = 0, otherwise [24]. To carry out
the processing of the information and determine the clusters, we used the K-means
algorithm. This algorithm performs a sequence of iterative procedures that form K
compact spherical clusters over input data, with the objective of minimizing
n K
J kj D 2 ( zk , x j ) (1)
j 1 k 1

where D2(zk, xj), in our work, is the Euclidean distance of xj to zk; and z1,z 2, ..., zK are
the centroids of the groups.
According to [24], the algorithm, first of all, randomly initializes the K centroids.
These are used to partition the data by assigning each point to the cluster of its closest
centre. Succeeding, for each cluster, the mean value of all the data points assigned to it
is computed, and this is considered as the new centre. This completes one iteration of
the K-means clustering. Next iteration uses the newly computed centres for reassigning
the data points to the K clusters, and the process continues until the measure of J falls
beneath a certain threshold or a maximum number of iterations have been executed.
To set the number of clusters, it was analyzed the mean of the centroid distribution
and the Davies Bouldin index [23], which is based on the measurement of dispersion
and a measure of similarity between the clusters. On this basis, we decided that the
number of 4 clusters responds to the computational test expected in this article. Soon,
running the k-means algorithm it was come to the result summarized in Table 1.

Table 1. Result of the k-means algorithm.

Occurrence of Events
Cluster Total Events Municipalities
Windstorms Hails Floods Runoffs Droughts Others
1 87 73 22 100 845 33 1160 90
2 125 101 133 316 174 36 885 54
3 35 32 33 201 145 78 524 42
4 57 41 53 354 123 50 678 108
 Graphically Fig. 2 shows the percentage of influence of each type of disaster, as
well as the number of events in each cluster, Fig. 3 shows the disasters average
compared with the overall average, and Fig. 4 shows the geographically clusters
distributions. Detailed results are given as follows.

Figure 2. Importance and quantities of each cluster by disaster

Figure 3. Averages of occurrences for each cluster relative to the total average.

The cluster 1 has as main feature events drought, and its average was far above the
overall average for this disaster and showed values below the overall average for the
other attributes. Notice that the geographic distribution of this cluster was quite
concentrated in the western of the state and covered a portion of the mountain region
(Fig. 4). The droughts events due to its higher recurrence rate attracted to itself a
particular cluster.
Cluster 2 differs by having the municipalities with the highest recurrence disaster,
containing the highest averages in the event of windstorms, floods and runoffs.
Although it has only 54 cities, is present in all regions of the state. In terms of flooding
it is the most critical cluster having 37.05% of all occurrences.
 Figure 4. Geographical distribution for resulting clusters.

In the case of cluster 3, its characteristic has owning the smallest amount of
affected municipalities, but having the second highest average for runoff events. That
is, has a high frequency of these events in just a few areas of the state. Many
municipalities that compose this cluster are situated on the coast of the state.
Finally cluster 4 has the highest number of municipalities affected by runoffs,
however presents the smallest averages in the other events. That is, although the large
number of cities, the recurrence of the events is low. This cluster covers the largest
number of municipalities that have a low frequency of disasters. Their lower range is in
the region west of the state.
We can notice also that both types of disasters most significant in the state have
peculiar characteristics and distinct from each other. The occurrences of drought are
strongly concentrated in western Santa Catarina economy where most municipalities
revolves around agricultural activities, so this event causes elevated economic
damages to the region. Already, most of the municipalities affected by the flash flood
are mainly in the Vale do Itajai, Great Florianpolis and north of the state. Most of
these municipalities have the particularity of irregular relief favoring the landslides.
Thus, the impact generated reaches more urbanized areas. The urbanization process
which results in several implications and hence the disordered occupation aggravate
the impacts caused by flash floods.

5 Conclusions

In this paper there were presented some concepts involved in Humanitarian Logistics
and Disaster Management in order to locate the focus of the research. There were
discussed concepts and features of a clusters system having the intention to verify the
application of the clusters algorithm to profile municipalities affected by natural
disasters.
 In this work, due to the unavailability of data, we do not take into account the
intensity or geographic area affected by natural disasters, as well as costs and number
of people affected to data analysis. This demonstrates the weakness in Brazilian
public policies that address planning for humanitarian crises. Thus, another question
which this paper aims to contribute is to demonstrate the importance of having
methodologies keepers of historical data against natural disasters.
Thus the responsible organizations will be able to develop standards for a different
form of prevention method ,that is, it is verified the profile of each cluster and the
decision making regarding which method can be adapted to the groups of
municipalities considering the typical behaviour of this group. For each of the clusters
of municipalities, the experiences and knowledge can be shared, which could serve as
a basis for integrated planning, aiming the preparation and the logistics processes on
the imminence of the disasters.
It can be verified that the proposed methodology corresponds to the expectations
and brings satisfactory results. These results could be either used as a previous step or
as an input data for a big process, which aims to identify potential donors or form
partnerships with suppliers during the pre-disaster phase, since it could be useful to
assist in evaluating the kind of products required for each type of disaster, prediction
of demand and generating inventories. As future work, we intend to further
investigate the clusters validation indices and to evaluate the results found in this
paper using the k-means algorithm with other clustering algorithms as well.

Acknowledgements

The authors acknowledge the body of experts of the CEPED/UFSC and the financial
support of the CAPES and CNPQ.

References

1. Sapir, G. D.: Disasters in Numbers 2010. CRED, Catholic University of Louvain, Brussels
(Belgium) January 24, Geneva (2011)
2. CREA, Revista do Conselho Regional de Engenharia, Arquitetura e Agronomia de Santa
Catarina. Ano 6- n 9 maio, Brasil (2011)
3. Fenton, G.: Coordination in the Great Lakes, Forced Migration. Review, pp. 2324,
September (2003)
4. Rey, F.: The complex nature of actors in humanitarian action and the challenge of
coordination. In: Humanitarian Studies Unit (Ed.), Reections on Humanitarian Action:
Principles, Ethics and Contradictions. TNI/Pluto Press with Humanitarian Studies Unit and
ECHO (European Commission Humanitarian Ofce), London (2001)
5. Tufinkgi, P.: Logistik im kontext internationaler katastrophenhilfe: Entwicklung eines
logistischen referenzmodells fr katastrophenflle. Ed. Haupt Verlag. Bern.Stuttgart Wien
(2006)
6. Schultz, S.F.: Disaster Relief Logistics:Benefits of and Impediments to Horizontal
Cooperation between Humanitarian Organizations. Tese. Technischen Universitt Berlin
(2008)
 7. Balcik, B.; Beamon, B.M.; Krejci, C; Muramatsu, K.M., Ramirez, M.: Coordination in
humanitarian relief chains: Practices, challenges and opportunities. International Journal
Production Economics. Science Direct, vol.126 issue 1, pp.22-34 (2010)
8. Chandes, J., Pach, G.:Investigating humanitarian logistics issues: from operations
management to strategic action. Journal of Manufacturing Technology Management Vol.
21 No. 3 , pp.320-340 (2011)
9. Kovacs, G , Spens, K.M.:Humanitarian logistics in disaster relief operations. International
Journal of Physical Distribution & Logistics Management, Vol. 37 Iss: 2, 99 114(2007)
10. Kovacs, G ,Spens, K.M.: Identifying challenges in humanitarian logistics, International
Journal of Physical Distribution & Logistics Management, Vol. 39 Iss: 6, 506 528 (2009)
11. Van Wassenhove, L. V.: Humanitarian Aid Logistics: Supply Chain Managent in High
Gear, Journal of the Operational Research Society, vol.57, n5, p.475-489 (2006)
12. Tomasini, R , Van Wassenhove, L.V.: Humanitarian logistics. Insead Business Press (2009)
13. Blecken, A.: A Reference Task Model for Supply Chain Processes of Humanitarian
Organizations. Doctorate Thesis. Institute of the University of Paderborn (2010)
14. Meirim, H.: Logstica humanitria e logstica Empresarial, mmrbrasil. Information,
www.mmrbrasil.com.br
15. Thomas, A. , Kopczak, L.: From logistics to supply chain management. The path forward in
the humanitarian sector, Fritz Institute, Information, www.fritzinstitute.org
16. Chu H.J. , Liau C.J., Lin C.H., Su B.S.:Integration of fuzzy cluster analysis and kernel
density estimation for tracking typhoon trajectories in the Taiwan region. Original Research
Article Expert Systems with Applications, Volume 39, Issue 10, August, Pages 9451-9457
(2012)
17. Chang L.C., Shen H.Y., Wang Y.F., Huang J.Y.& Lin Y.T.: Clustering-based hybrid
inundation model for forecasting flood inundation depths. Original Research Article
Journal of Hydrology, Volume 385, Issues 14, 7 May, Pages 257-268 ( 2010)
18. Wan S. : Entropy-based particle swarm optimization with clustering analysis on landslide
susceptibility mapping . Springer-Verlag (2012)
19. Acosta, M. Goncalves, M. , Vidal, M.E.: CAREY: ClimAtological ContRol of EmergencY
Regions. OTM Workshops, LNCS 7046, pp. 494503, c Springer-Verlag Berlin
Heidelberg (2011)
20. Jahre,M., Jensen, M.: Coordination in humanitarian logistics through clusters. International
Journal of Physical Distribution & LogisticsManagement Vol. 40 No. 8/9, pp. 657-674
(2010)
21. Dalal J., Mohapatra P.K.J., Mitra G.C.: Locating cyclone shelters: a case. Emerald Article:
Disaster Prevention and Management Vol. 16 No. 2, pp. 235-244 .Emerald Group
Publishing Limited 0965-3562 (2007)
22. Jahre, M., Navangul, A.K.:Predisting the unpredictable Demand Forecasting in
International Humanitarian Response. In: Proceedings of the 23rd Annual NOFOMA
Conference. Harstad, Norway. P. 265-281(2011)
23. Halkidi, M; Batistakis , Y. , Vazirgiannis, M..: On Clustering Validation Techniques.
Journal of Intelligent Information Systems, 17:2/3, 107145 (2001)
24. Bandyopadhyay, S. : Genetic algorithms for clustering and fuzzy clustering.WIREs Data
Mining and Knowledge Discovery1, 524531 (2011)

All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.