Professional Documents
Culture Documents
net/publication/295245529
Risk assessment of cultural heritage sites clusters using satellite imagery and
GIS: the case study of Paphos District, Cyprus
CITATIONS READS
20 545
4 authors, including:
Diofantos G. Hadjimitsis
Cyprus University of Technology
406 PUBLICATIONS 2,967 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Cyprus Continuously Operating Natural Hazards Monitoring and Prevention System (CyCLOPS) View project
(JPI-CH) CLIMA - Cultural Landscape risk Identification Management and Assessment View project
All content following this page was uploaded by Kyriacos Themistocleous on 31 October 2017.
A. Agapiou, V. Lysandrou,
K. Themistocleous & D. G. Hadjimitsis
Natural Hazards
Journal of the International Society
for the Prevention and Mitigation of
Natural Hazards
ISSN 0921-030X
Nat Hazards
DOI 10.1007/s11069-016-2211-6
1 23
Your article is protected by copyright and all
rights are held exclusively by Springer Science
+Business Media Dordrecht. This e-offprint
is for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.
1 23
Author's personal copy
Nat Hazards
DOI 10.1007/s11069-016-2211-6
ORIGINAL PAPER
D. G. Hadjimitsis1
Abstract This paper aims to assess the risk of natural and anthropogenic hazards for
cultural heritage in Cyprus by integrating multi-temporal GIS and earth observation
analysis, in the area of Paphos District. The work presented here attends to re-evaluate
previous results from earth observations and GIS analysis and go a step forward targeting
more reliable outcomes for cultural heritage management. The scope of the paper was to
develop a more accurate methodology for risk assessment against natural and anthro-
pogenic hazards (e.g., soil erosion; urban expansion), based on homogeneous clustering of
the monuments under consideration. The accomplished assessment approach, being lop-
sided and generic, cannot be applied across the board and undistractedly for cultural
heritage management of all types of monuments of the district. Instead, the proposed
clustering of monuments based on a variety of parameters is taking into consideration
characteristics of their immediate environment, resulting rational local-based outcomes
more useful for monuments and sites safeguarding and for prevention measurements. For
each one of the five clusters of monuments located in the Paphos District, an analytical
hierarchy process (AHP) method was followed in order to address the individual and
unique characteristics of the monuments and sites within the same cluster area. Subse-
quently, the weight factors from these clusters were interpolated to the whole district, prior
to the application of the overall AHP risk assessment. Ultimately, the results were com-
pared with the overall AHP method applied for the entire Paphos District, indicating that
& A. Agapiou
athos.agapiou@cut.ac.cy
V. Lysandrou
vasiliki.lysandrou@cut.ac.cy
K. Themistocleous
k.themistcleous@cut.ac.cy
D. G. Hadjimitsis
d.hadjimitsis@cut.ac.cy
1
Department of Civil Engineering and Geomatics, Remote, Sensing and Geo-Environment Research
Lab, Cyprus University of Technology, Limassol, Cyprus
123
Author's personal copy
Nat Hazards
the proposed methodology can be more accurate and realistic for the different groups of the
monuments.
1 Introduction
123
Author's personal copy
Nat Hazards
large extent of some properties, especially in the case of the world natural heritage sites or
the level of detail needed for the assessment of the cultural sites, satellite-based earth
observation (EO) enables the responsible authorities to protect, monitor, document, present
and share the World Heritage Sites (Negula et al. 2015) as well as of course other
important and less important archaeological sites and monuments.
AHP is an intuitively easy approach widely applied to help decision-makers who face
several conflicting criteria and alternatives simultaneously. AHP has been proposed by
Saaty in 1970s (Saaty 1977). It is a multi-criteria decision-making method based on
comparing concepts (alternatives) in pairs. Since then, it has been applied successfully in
several different fields, such as education, social, political, manufacturing and engineering,
for different purposes: planning, selection of most important factors and best alternatives,
allocation of resources, optimization and conflict resolution (Sara et al. 2015). However, as
Fan et al. (2015) and Lin and Kou (2015) argue, there are still some deficiencies and
limitations when applying this methodology. First, the comparative judgments are sub-
jective because they rely heavily on expert opinion, which may sometimes cause incon-
sistency. Furthermore, AHP lacks the ability to adequately cope with any inherent
uncertainty and imprecision in the data. Finally, in a real situation, an expert may have
limited knowledge of and experience with alternatives; the preferred information may
contain fuzziness and incompleteness, and AHP is unable to handle this incomplete
information.
The aim of this paper is to propose a modification of the AHP methodology applied for
cultural heritage sites and monuments by grouping together sites with similar character-
istics facing similar threats. Clustering of monuments will lead to homogenous subgroups
of sites, followed by AHP method for estimating hazards. The example given in this paper
is the case study of Paphos District, Cyprus, where the risks have been identified solely
from remote sensing data (see Agapiou et al. 2015a, b).
The current study concerns Paphos District in western Cyprus, an area abound in antiq-
uities. More than 150 declared Ancient Monuments of First (Ancient Monuments on State
Land) and Second Schedule (Ancient Monuments on Private Land) protected by the
Antiquities Law have been mapped with high accuracy (Fig. 1). Previous studies in this
area have been conducted by the authors in order to create a common geo-database of all
monuments, as well as to estimate hazards and produce risk maps from remote sensing data
(Agapiou et al. 2013).
Both natural and anthropogenic hazards have been mapped using solely remote sensing
data. Based on a variety of remote sensing data including low-, medium- and high-reso-
lution images (e.g., MODIS; Landsat; QuickBird) along with ready satellite products (e.g.,
ASTER Global Digital Elevation Model, ASTER GDEM) each hazard examined in this
paper has been analyzed, while the overall risk was estimated based on the AHP
methodology. A series of risk maps was created, relative to anthropogenic (urban sprawl,
modern road network, fires) and environmental (erosion, salinity, neotectonic activity)
hazards affecting cultural heritage (CH) sites in Paphos District. The resulting risk maps
for each hazard are illustrated in Fig. 2 (for further information, see Agapiou et al. 2015a, b).
In addition, another variable used for clustering monuments and sites into groups was their
elevation.
123
Author's personal copy
Nat Hazards
Prior to the AHP methodology, clustering of monuments in the Paphos District was
accomplished. In total, five classes (i.e., groups of monuments) have been identified as the
optimal number of groups for this area (see Sect. 3). For each one of these classes, the AHP
method has been applied. Therefore, for each class and for each factor (risk) a different
weight has been given. Then based on the overall results from all classes, six maps
indicating the distribution of the weight for each factor (risk) have been produced in a GIS
environment. These maps were then multiplied by the risk factor, in a similar procedure as
this is applied in AHP methodology. The overall methodology of the paper is given in
Fig. 3.
From a statistical analysis carried out based on Paphos’ Ancient Monuments average
nearest neighbor distance, it was found that they do not follow a random pattern, but are
formed into clusters. Given the z-score of -6.36, there is less than 1 % probability that this
clustered pattern could be the result of random chance (Fig. 4). This is not atypical for
archaeology since normally several sites are found clustered based on occasion’s variables,
such as the topography of a site. Therefore, it is evident that the protected monuments
under consideration are to be found in clusters within a small distance between each other.
Hence, a clustering of monuments and sites was performed using spatial constraints in
order to group together nearby monuments with similar characteristics. This observation
allows us to apply clustering methodology conducive to group monuments facing similar
threats.
123
Author's personal copy
Nat Hazards
Fig. 2 Map indicating the different anthropogenic and natural hazards over Paphos District. a Archaeo-
logical sites of the Paphos District, b erosion map: areas where the soil loss is greater than the mean soil loss
value of the whole district are indicated with red color, c salinity map: areas close to the sea are indicated
with red, d tectonic activity: high and very high hazard areas are indicated with red, while very low hazard
is indicated with blue color, e urban expansion is indicated with red color, f road network proximity (250 m)
is indicated with red color which is linked with potential future urban expansion, and g fires map observed
during the period 2010–2013 is indicated with red color (revised map from Agapiou et al. 2015a, b)
123
Author's personal copy
Nat Hazards
Fig. 4 Average nearest neighbor summary of the protected monuments of Paphos District
3 Results
123
Author's personal copy
Nat Hazards
Fig. 5 Clustering of monuments based on their elevation characteristics. Five groups are formed: 0–55.53,
55.53–169.97, 169.97–349.74, 349.74–539.98 and 539.98–879.98 m
factors discussed in this paper. As it is indicated, each monument has different values of
threat or elevation. Therefore, clustering analysis seeks, in the finest way (R2), to group
together into five main classes all the declared monuments of Paphos District taking into
consideration all relative information afore described.
Figure 7 indicates the final results after the overall clustering methodology was applied
for each group separately. Five main classes have been identified for all monuments and
sites of Paphos District: classes 1–5. As it is demonstrated in Fig. 7, the classes have no
similar total number of monuments, while two of them (classes 1 and 2) have only one
monument registered under their class. On the contrary, other classes, such as classes 4 and
5, seem to group together many monuments (39 and 97 monuments, respectively). It
should be mentioned that the number of the class reflects only the sequence numbering.
This number of group for the monuments in the Paphos District has been selected after an
evaluation of different numbers of classes. Grouping effectiveness is measured using the
Calinski–Harabasz pseudo-F-statistic, which is a ratio reflecting within-group similarity
and between-group difference:
2
R
nc 1
1R2
nnc
where
123
Author's personal copy
Nat Hazards
SST SSE
R2 ¼
SST
nc X
X nv
ni X 2
SST ¼ vkij vk
i¼1 j¼1 k¼1
nc X
X nv
ni X 2
SSE ¼ vkij vki
i¼1 j¼1 k¼1
A description of each class is given in Table 1 and in Fig. 8. Figure 8 presents the
parallel box for each class (class 1 with blue dots; class 2 with red dots; class 3 with green
dots; class 4 with yellow dots; and class 5 with purple dots). The parallel box plot graph
summarizes both the groups and the variables (i.e., risks) within them.
123
Author's personal copy
Nat Hazards
Fig. 7 Clustering of monuments based on their overall attributed characteristics (vicinity to urban areas,
vicinity to rivers, elevation, soil erosion, fires, etc.) and specific spatial constrains
Table 1 Description of each class after the clustering in the GIS environment
Class Description
1 Monuments/sites located in low elevations, far from the coastline, within active tectonic regions;
close to main road network and urban areas. Limited threats from soil erosion, but within areas
with high possibility of fires
2 Monuments/sites located on hilly areas, far from the coastline, within active tectonic regions; away
from main road network and urban areas. Limited threats from soil erosion, but within areas with
high possibility of fires
3 Monuments/sites located in very high elevations, far from the coastline, away from active tectonic
regions; close to main road network and urban areas. Threats from soil erosion and limited threats
from fires
4 Monuments/sites located nearby the coastline, in low altitude and in the vicinity of urban areas and
main road network. Limited threats from soil erosion and limited threats from fires. Within active
tectonic regions
5 Monuments/sites located in medium elevations, far from the coastline, within active tectonic
regions; close to main road network and urban areas. Threats from soil erosion and limited threats
from fires
123
Author's personal copy
Nat Hazards
Fig. 8 Parallel box indicating the characteristics of each class: class 1 with blue dots, class 2 with red dots,
class 3 with green dots, class 4 with yellow dots and class 5 with purple dots
matrix, and the appropriate eigenvector solution method is then employed to calculate the
factor weightings (Table 2). As demonstrated in Table 2, different weight factors have
been estimated for each class based on their characteristics (see Table 1). The highest
weight for each class is highlighted in the table. It is obvious that the weights might vary a
lot for each hazard since this might be or not of high importance for each of the five
classes. For instance, the highest weight for class 1 is linked to urban areas, while for the
rest of the classes the highest weight values are tectonic (classes 2 and 5), fires (class 3) and
again urban areas (class 4). This difference in weights that is recorded for the same hazards
in the different classes is normal, since each group of monuments (class) faces in dissimilar
proportion the different hazards depending on the location of the site. For instance, the
distance from the sea—salinity—for class 1, is not so important factor as it is for class 4,
where all monuments are located nearby the coastline.
Based on the overall weights of Table 2, the following different equations for overall
risk hazard for each class can be defined:
123
Author's personal copy
Nat Hazards
123
Author's personal copy
Nat Hazards
where F1–F6 are the different factors studied in this paper (tectonic; salinity; road net-
work; urban areas; soil erosion; and fires, respectively). Based on the results of Table 2, the
normalized weights for each risk have been added in the attribute table of the monuments
in the GIS environment. Then interpolation was carried out in a GIS environment based on
the inverse distance weight (IDW) algorithm. The results from the interpolation of the
weight factors are presented in Fig. 9. It should be noticed that in comparison with tra-
ditional AHP methodology, a single value would be allocated for all monuments in the
Paphos District.
Then the overall risk hazard map was produced by multiplying the weight factor
(Fig. 9) and the hazard (Fig. 2), as demonstrated below:
Overall Risk Hazard ¼ Weight 1 F1 þ Weight 2 F2 þ Weight 3 F3 þ Weight 4
F4 þ Weight 5 F5 þ Weight 6 F6
The final overall risk hazard map is demonstrated in Fig. 10, while Fig. 11 presents the
results from the traditional AHP methodology (from Agapiou et al. 2015a, b). The overall
risk hazard was classified into five main categories using the natural breaks values. These
categories are as follows: very low hazard; low hazard; medium hazard; high hazard; and
very high hazard. As shown in Fig. 10, the very high hazard areas are found in the
surroundings of Paphos town. Monuments and sites of class 4 seem to be the most
threatened monuments in the area of Paphos District, while other monuments of class 5 are
also endangered.
Although similar results might be seen between the traditional and the proposed AHP
methodology (Figs. 10, 11), some differences are also recorded, especially in the rural
areas of Paphos District. The proposed AHP methodology based on clustering of the
monuments was able to provide more realistic results, since the weight factors are not
uniformly distributed within Paphos District as a single value (i.e., traditional AHP), but
rather they changed depending on the location of the monument (i.e., proposed AHP).
4 Discussion
Risk assessment of cultural heritage sites are of great importance. Prevention of threatened
sites and monuments is sometimes the only way to protect tangible heritage against both
natural and anthropogenic hazards. In this paper, more than 150 protected ancient mon-
uments and sites found in Paphos District (Cyprus) have been analyzed. Six different
123
Author's personal copy
Nat Hazards
Fig. 9 Interpolation of the different normalization weights for each hazard (weight 1–6: tectonic; salinity;
road network; urban areas; soil erosion; and fires, respectively)
hazards have been studied in this paper: tectonic; salinity; road network (linked with
potential future expansion); urban areas; soil erosion; and fires threats.
The sites have been grouped into five main categories based on their characteristics. For
each cluster, weight factors have been allocated. Interpolation was then carried out pro-
viding six maps with the distributed weight. Then the overall AHP methodology was
applied. The results indicated that the majority of the sites grouped into class 4 (sites
located nearby the coastline, with low altitude and in the vicinity of urban areas and main
road network; limited threats from soil erosion and limited threats from fires; within active
tectonic regions) seem to be the most endangered sites from the whole Paphos District.
The proposed methodology and the overall results were confirmed from systematic
in situ visits in the area and observation of specific monuments, especially in the urban
center of Paphos. This observation is further in agreement with experts’ opinion (personal
communication) operating/working in the area for many years, which indicates that
123
Author's personal copy
Nat Hazards
Fig. 10 Overall risk hazard map for the Paphos District, based on the clustering of the sites
Fig. 11 Overall risk hazard map for the Paphos District after applying traditional AHP methodology (from
Agapiou et al. 2015a)
123
Author's personal copy
Nat Hazards
Acknowledgments Thanks are due to the Remote Sensing and Geo-environment Research Laboratory
(http://www.cyprusremotesensing.com/) of the Department of Civil Engineering and Geomatics of the
Cyprus University of Technology for the support (http://www.cut.ac.cy).
References
Agapiou A, Nisantzi A, Lysandrou V, Mammouri R, Alexakis DD, Themistocleous K, Sarris A, Hadjimitsis
DG (2013) Mapping air pollution using earth observation techniques for cultural heritage. In: Pro-
ceedings of SPIE 8795. First international conference on remote sensing and geoinformation of the
environment (RSCy2013), p 87950K. doi: 10.1117/12.2028234
Agapiou A, Lysandrou V, Alexakis DD, Themistocleous K, Cuca B, Sarris A, Argyrou N, Hadjimitsis DG
(2015a) Cultural heritage management and monitoring using remote sensing data and GIS: the case
study of Paphos area, Cyprus. Comput Environ Urban Syst 54:230–239. doi:10.1016/j.compenvurbsys.
2015.09.003
Agapiou A, Alexakis DD, Lysandrou V, Sarris A, Cuca B, Themistocleous K, Hadjimitsis DG (2015b)
Impact of urban sprawl to archaeological research: the case study of Paphos area in Cyprus. J Cult
Herit 16(5):671–680. doi:10.1016/j.culher.2014.12.006
Drdácký M (2007) Protecting the cultural heritage from natural disasters. Policy Department Structural and
Cohesion Policies, Culture and Education, European Parliament, IP/B/CULT/IC/2006_163
Fan G, Zhong D, Yan F, Yue P (2015) A hybrid fuzzy evaluation method for curtain grouting efficiency
assessment based on an AHP method extended by D numbers. Expert Syst Appl. Available online 11
September 2015, ISSN 0957-4174. doi:10.1016/j.eswa.2015.09.006
Hadjimitsis D, Agapiou A, Alexakis D, Sarris A (2011) Exploring natural and anthropogenic risk for cultural
heritage in Cyprus using remote sensing and GIS. Int J Digit Earth 6(2):115–142
Karaca F (2015) An AHP-based indoor Air Pollution Risk Index Method for cultural heritage collections.
J Cult Herit 16(3):352–360
Lin G, Kou G (2015) Bayesian revision of the individual pair-wise comparison matrices under consensus in
AHP–GDM. Appl Soft Comput 35:802–811. ISSN 1568-4946
Negula ID, Sofronie R, Virsta A, Badea A (2015) Earth observation for the world cultural and natural
heritage. Agricult Agricult Sci Procedia 6:438–445
Oppio A, Bottero M, Ferretti V, Fratesi U, Ponzini D, Pracchi V (2015) Giving space to multicriteria
analysis for complex cultural heritage systems: the case of the castles in Valle D’Aosta Region, Italy.
J Cult Herit. Available online 7 April 2015, ISSN 1296-2074. doi:10.1016/j.culher.2015.03.003
Saaty TL (1977) A scaling method for priorities in hierarchical structures. J Math Psychol 15(3):234–281
123
Author's personal copy
Nat Hazards
Sara J, Stikkelman MR, Herder MP (2015) Assessing relative importance and mutual influence of barriers
for CCS deployment of the ROAD project using AHP and DEMATEL methods. Int J Greenh Gas
Control 41:336–357
Taboroff T (2000) Cultural heritage and natural disasters: incentives for risk management and mitigation. In:
Kreimer A, Arnold M (eds) Managing disaster risk in emerging economies, Disaster Risk Management
Series No, 2, pp 71–79
Tarchi D, Rudolf H, Pieraccini M, Atzeni C (2000) Remote monitoring of buildings using a ground-based
SAR: application to cultural heritage survey. Int J Remote Sens 21(18):3545–3551
Wang J-J (2015) Flood risk maps to cultural heritage: measures and process. J Cult Herit 16(2):210–220
123
View publication stats