Studying Patterns of Rainfall and Topographical Clustering for
Kingdom of Bahrain: An Application of Big Data
UNEB GAZDER
Department of Civil Engineering
University of Bahrain
Sakhir, 32038
BAHRAIN
Abstract: - Rainfall is an important aspect for urban planners, especially for professionals related to highway
design. This study was aimed at studying the patterns of rainfall in Kingdom of Bahrain using cluster analysis
with topographical data. The rainfall data was collected from the website meteorological directorate of ministry
of transportation and telecommunication of Bahrain, which consisted of approximately 14000 days, thus
referred to as big data. K-means clustering was used to identify the patterns. Moreover, Geographic Information
System (GIS) was used for topographical clustering of elevation map of Bahrain. This was done to identify the
areas which should receive more attention with regards to drainage due to the topography of Bahrain. The
results of this study showed that the time periods which receive comparatively heavier rainfall are around
January while those with significantly lower rainfall are during May and October. The topographical analysis
showed that the northern parts of the country are supposed to be paid more attention to avoid flooding and,
consequently, damage to infrastructure. Moreover, results of the study could be used by the urban and
transportation planners in Bahrain for design of efficient drainage resulting in economic benefits to the society.
Key-Words: - rainfall, cluster analysis, big data analysis, GIS, topography
Received: August 11, 2023. Revised: December 14, 2023. Accepted: February 22, 2024. Published: April 3, 2024.
1 Introduction
Rainfall patterns and their changes have been
commonly associated with the climatic changes
occurring on a global level [1]. Studying rainfall
patterns is a major concern for sustainable urban
development [2]. Rainfall and urban development
are considered interlinked to each other as one can
affect the other [3,4]. The issue is more pressing for
the highways as accumulation of water on the
highways can result in structural and functional
damage [5]. Such damages result in higher
construction and maintenance costs, economic
losses to travelers and, in some cases, loss of life [6-
9].
There are two important aspects of studying
rainfall patterns, namely, spatial and temporal.
Spatial patterns refer to the study of areas which
may be exposed to higher risk of flooding
accumulation due to higher rainfall and/or lower
elevations [10]. Temporal patterns refer to the
identification of the season/time of the year during
which heavier rainfalls are expected [11]. These
factors would eventually impact the drainage design
and emergency management activities.
Bahrain is considered one of the Arid countries
with low rainfall intensity and quantities [12]. Due
to this reason, effects of rainfall are often
overlooked by urban and transport planners.
However, consequent damages have been caused to
the infrastructure, even with low rainfall, should not
be ignored [13]. Despite this, there are few studies
available on the analysis of rainfall patterns,
especially those combining the temporal and
topographical patterns.
The analysis technique used in this study is
clustering analysis. Clustering aims to find the best
partitioning of the data based upon some criteria of
optimality. The criteria can be in the form of an
objective function or the number of clusters to be
achieved [14,15]. It has been commonly applied in
the government and private sector for the
identification of potential risks and opportunities
[16]. The applications of clustering algorithms have
been reported in the fields of statistics, computer
science and machine learning for solving classical
as well as new problems, such as traveling salesman
and bioinformatics [17].
Engineering World
DOI:10.37394/232025.2024.6.5
Uneb Gazder
E-ISSN: 2692-5079
29
Volume 6, 2024
The clustering algorithms work on distance or
similarity metric for development of clusters [18].
Some of the common clustering algorithms include
k-means, hierarchical, self-organizing maps,
expectation maximization, Density-based spatial
clustering of applications with noise (DBSCAN),
non-spherical clusters by using minimum spanning
tree over KD-tree-based micro-clusters (MCMST),
Fuzzy C-mean, and VIASCKDE Index. The
selection of clustering algorithm depends on several
factors such as popularity/familiarity, flexibility,
applicability to the available dataset, and handling
of multidimensional problems [19]. K-means and
hierarchical clustering have been used more often
than other available algorithms [20].
Due to its importance and multi-disciplinary
impacts, this study is aimed at studying the spatial
and temporal patterns of rainfall in Bahrain. Cluster
analysis and Geographic Information System (GIS)
was used for this purpose. The objectives of this
study include classifying different periods of year
according to the rainfall data and determining areas
in Bahrain which are exposed to higher risk of
flooding and accumulation. The results of this study
are expected to enable better understanding of the
rainfall patterns, possibly resulting in changes in
infrastructure planning and management practices.
Moreover, the readiness of emergency management
services can also be improved with the identification
of critical times and locations, subjected to higher
rainfall and flooding.
2. Study Area and Dataset
Bahrain is an Archipelago of 33 islands, located at
16 Km from Saudi Arabia and 21 Km from Qatar.
Its land area is estimated to be approximately 700
Km2 [21]. Similar to other neighboring countries,
Bahrain has also traditionally relied on petroleum
for its trade. However, with the depletion of its
reservoirs, it is currently facing scarcity of resources
which makes the application of sustainability
concepts crucial for long-term development of the
nation [22].
Weather of Bahrain is usually hot and humid
with temperatures going close to 500C during
summer. The average temperatures are
approximately close to 300C throughout the year
[23]. Rainfall in Bahrain is less frequent and heavy
rainfalls are seldom expected [24]. Due to this fact,
the aspect of highway drainage is overlooked at the
time of design and construction which then results
in higher investments at later stages [25]. Hence, it
is important that existing rainfall patterns are
studied, and future planning is done accordingly to
save economic losses to the country, especially in
the current situation of resource scarcity.
Big data is referred to as a massive dataset with
hidden, potentially complex trends and patterns
[26]. It was crucial for this study to utilize a large
amount of data for identifying rain patterns over the
course of several years. The dataset for this study
was taken from the website of ministry of
transportation and telecommunications of Bahrain.
The ministry has a meteorological directorate which
is responsible for monitoring weather conditions in
Bahrain. The data recorded by the sensors installed
at the geographic location of Bahrain International
Airport which is in Muharraq. It covered the period
from January 1944 to December 2014. An initial
analysis of the data showed that rainfall has not
increased/decreased substantially throughout the
above-mentioned period. Hence, it can be expected
that the analysis done on the available data is valid
for the present year as well.
The average rainfall was observed to be 0.2mm
per day and 6.5mm per month. Maximum rainfall
was 67.9mm which was observed in March 1995.
The data had a standard deviation of 2mm. Table 2
shows more detailed month-wise statistics of the
available data. It can be clearly observed that the
variation in the monthly data seems quite significant
for some of the months. This is further highlighted
by the box plots shown in Fig. 1. The vast difference
in rainfall data of different months, in terms of their
magnitude and variation, justifies the need of the
current study.
Table 1. Statistics of Available Data
Mean
Minimum
Maximum
Std.Dev.
January
18.84
0.00
135.90
24.15
February
14.05
0.00
106.80
22.60
March
14.25
0.00
139.20
22.74
April
6.95
0.00
69.90
12.79
May
1.03
0.00
11.90
2.37
June
0.00
0.00
0.00
0.00
July
0.00
0.00
0.00
0.00
August
0.01
0.00
1.00
0.12
September
0.00
0.00
0.00
0.00
October
0.32
0.00
8.90
1.32
November
7.91
0.00
101.60
20.11
December
14.81
0.00
119.60
24.35
Engineering World
DOI:10.37394/232025.2024.6.5
Uneb Gazder
E-ISSN: 2692-5079
30
Volume 6, 2024
Fig. 1. Comparison of month-wise data
3. Analysis Methods
The clusters can be used to conveniently identify the
patterns in the available dataset which has been the
reason for their utilization in different fields
including identification of faults. This was the prime
reason for selection of this technique in the present
study [27]. As an initial step, Analysis of Variance
(ANOVA) was performed to check if there is a
significant difference in the rainfall patterns of
different months. The results of this test are shown
in Table 2, which confirms that there is a significant
difference in the rainfall patterns of different months
at 5% level of confidence. This paves the way for
further investigation done with clustering analysis in
this study.
Table 2. ANOVA
Source of
Variation
SS
df
MS
F
P-
value
F
crit
Between
Groups
38700
11
3518
15
0.00
1.8
Within
Groups
183449
792
231
Total
222149
803
Two types of cluster analysis were applied, to
identify the season (months) with higher rainfall,
namely, hierarchical tree and K-means clustering.
Hierarchical clustering works by finding out the
Euclidean distance (Equation (1)) between the
samples of data and clustering the samples which
have the least distance between them [28].
Ed(p, q) = ∑(𝑝𝑖− 𝑞𝑖)2 (1)
Where Ed is the Euclidean distance between
datasets p and q.
On the other hand, K-means clustering is done
by specifying the number of clusters to be made
from the data and clustering algorithm will develop
those clusters on the K-means distance between the
samples. The clusters are developed based upon the
maximization of distances between means
(centroids) of clusters [29]. The former can be
considered under unsupervised while the latter is
part of supervised clustering approach.
As stated earlier, Bahrain is a small island,
hence, the amount of rainfall does not change
between parts of country. Therefore, it seemed
logical to identify the areas which have higher
potential of flooding based on topography.
Geographic Information System (GIS) was used to
find the parts of the country which have
significantly lower elevation and, consequently,
more likely to face water accumulation. GIS is a
platform which provides integration between the
spatial data of an area and its non-spatial
characteristics. GIS is a popular platform to analyze
spatial data based on non-spatial attributes [30]. Its
advantages include integration of data from different
sources and the ability to generate quarries and
filters for the feature of interest. Due to these
applications, GIS has been used for topographical
analysis [31], which is the same task for this it is
used in this study. An open-source software, namely
QGIS, was used for this purpose. The maps and
elevations for Bahrain were taken from Open Street
Maps and United Stated Geological Survey (USGS)
website.
4. Results
Hierarchical Tree clustering was first applied to find
out the possible clusters based on rainfall data. The
results are shown in Fig. 2. This figure shows that
the season from May to October has a similar
rainfall pattern while other months are significantly
different than these months as well as each other.
Based on this finding, it was decided to proceed
with seven clusters for K-means clustering. The
results of K-means clustering are shown in Table 3.
These results also conform with the Hierarchical
Tree clusters. Therefore, it can be said that May-
October is a season of lower rainfall with an average
of 0.23mm while other months are part of the higher
rainfall season. Among which, January had the
highest rainfall with an average of 18.84mm. Table
3 shows more detailed statistics for each cluster.
For topographical clustering, street map of
Bahrain was acquired from opensteetmap.org. Then,
a layer was created using the elevations of roads and
streets for Bahrain. The source of elevation data is
already mentioned in the previous section.
Clustering analysis involved classification of points
on the highways and streets according to their
Box Plot (rain data] 12v*67c)
Median; Box: 25%-75%; Whisker: Non-Outlier Range
Median
25%-75%
Non-Outlier Range
Outliers
Extremes
January
February
March
April
May
June
July
August
September
October
November
December
-20
0
20
40
60
80
100
120
140
160
Engineering World
DOI:10.37394/232025.2024.6.5
Uneb Gazder
E-ISSN: 2692-5079
31
Volume 6, 2024
elevations. QGIS was used for performing
topographical analysis.
Fig. 3 shows the topographical clustering of
points in Bahrain which was done by using GIS.
The colors are graduated from red (for lowest
elevations) to white (for highest elevations). It can
be observed that lower elevations are clustered
along the coastal areas. Hence, flooding could be
easily avoided in these areas with efficient drainage
systems at lower cost since the outlet (sea) is closer
to these areas. There is a higher clustering of low
elevation points in the northern part as compared to
the southern part, hence, this area requires more
attention.
Fig. 2. Hierarchical Tree Clustering
Table 3. K-Means Clustering
Jan
Feb
Mar
Apr
Dec
May-
Oct
Nov
Jan
0.00
1158
1175
874
1320
915
1112
Feb
34
0.00
862.
591
1176
697
1000
Ma
r
34
29
0.00
619
905
708
868
Apr
29
24
24
0.00
776
205
582
Dec
36
34
30
27
0.00
798
1035
Ma
y-
Oct
30
26
26
14
28
0.00
453
Nov
33
31
29
24
32
21
0.00
Table 4. Statistics of Clustered Data
Cluster
Mean
Minimum
Maximum
Std.Dev.
Jan
18.84
0.00
135.90
24.15
Feb
14.05
0.00
106.80
22.60
Mar
14.25
0.00
139.20
22.74
Apr
6.95
0.00
69.90
12.79
Dec
14.81
0.00
119.60
24.35
May-
Oct
0.23
0.00
11.90
1.16
Nov
7.91
0.00
101.60
20.11
Fig. 3. Spatial Clustering
It should be noted that Hasanean and Almazroui
[32], in a previous study, have recorded higher
rainfall quantities for Saudi Arabia, especially in the
adjoining areas of Bahrain. However, they have also
emphasized the fact that the rainfall patterns are
changing in the region. The identification of
precipitation patterns has a global impact, since
these arid regions (including Bahrain and Saudi
Arabia) comprise ¼ of the land area and are
characterized by strong intermittent hydrologic
regime [33].
Another distinct feature of this study is the
resemblance of the rainfall patterns in Bahrain with
European countries with higher precipitation during
winter season. This pattern has been linked to the
climate change [34] which seems to be evident in
arid countries such as Bahrain.
5. Conclusions
This study aimed at identifying the seasons
(months) and areas of Bahrain which have higher
potential of urban flooding. Meteorological data for
a 60-year period was used for the analysis which is
referred to as the big data. Clustering techniques,
namely, Hierarchical Tree and K-means, were
applied, on the big data of rainfall time series,
identify the season of higher rainfall while GIS was
used to identify areas which are at a higher risk of
flooding due to topography of the country.
Clustering analysis showed that the following
clusters can be made based on rainfall data; January,
February, March, April, May-October, November
Engineering World
DOI:10.37394/232025.2024.6.5
Uneb Gazder
E-ISSN: 2692-5079
32
Volume 6, 2024
and December. May to October is a period with
significantly lower rainfall. GIS analysis showed
that coastal areas are at a higher risk of flooding,
especially in the northern part of the country. The
findings of this study are expected to be beneficial
for planning authorities in Bahrain. They can be
used for better preparedness against the risks of
infrastructure damage due to urban flooding. One of
the possible applications is the necessary cleaning of
drainage systems in the area in the time and zones
identified by the clusters. Another possible
application of the results of this study is the
assignment of resources (such as pumps) in the
vicinity of the identified zones. The results of this
study can also be linked with the climatic change
patterns in the region and used to study the
hydrologic regimes.
It is recommended for future studies to combine
the spatial clustering data with temporal clusters for
a more detailed analysis. GIS platform can be
further used to study run-off patterns and designing
of efficient drainage systems. Other possible
directions of research could be to employ other
machine learning and computational intelligence
techniques for prediction of rainfall data.
References
[1] J. L. Martel, F. P. Brissette, P. Lucas-Picher, M.
Troin, and R. Arsenault, Climate Change and
rainfall intensity–duration–frequency curves:
Overview of Science and Guidelines for
Adaptation, Journal of Hydrologic
Engineering, 26(10), 2021 03121001. M. H.
Dore, Climate Change and Changes in Global
Precipitation Patterns: What do We Know?,
Environment International, 31(8), 2005, pp.
1167-1181.
[2] I. Abatcha, A. Mustapha, and A. Barkindo,
Comprehensive Analysis of Rainfall Variability
in Urban Maiduguri, Nigeria: Implications for
Climate Resilience and Sustainable
Development, International Journal of
Environment and Climate Change, 14(3), 2024,
149-159.
[3] D. Mu, P. Luo, J. Lyu, M. Zhou, A. Huo, W.
Duan, ... and X. Zhao, Impact of Temporal
Rainfall Patterns on Flash Floods in Hue City,
Vietnam, Journal of Flood Risk Management,
14(1), 2021, e12668.
[4] M. Hemmati, B. R. Ellingwood, and H. N.
Mahmoud, The Role of Urban Growth in
Resilience of Communities under Flood Risk,
Earth's Future, 8(3), 2020, e2019EF001382.
[5] D. Lu, S. L. Tighe, and W. C. Xie, Impact of
Flood Hazards on Pavement Performance,
International Journal of Pavement
Engineering, 21(6), 2020, 746-752.
[6] H. C. Z. Qing, Impact of Flooding on Highway
[J], Meteorological Monthly, 9, 2000.
[7] R. K. Dahal, S. Hasegawa, T. Masuda, and M.
Yamanaka, Roadside Slope Failures in Nepal
during Torrential Rainfall and Their Mitigation,
Disaster Mitigation of Debris Flows, Slope
Failures and Landslides, 2006, pp. 503-514.
[8] P. A. Pisano, L. C. Goodwin, and M. A.
Rossetti, US Highway Crashes in Adverse
Road Weather Conditions, In 24th Conference
on International Interactive Information and
Processing Systems for Meteorology,
Oceanography and Hydrology, New Orleans,
LA, January, 2008.
[9] A. M. Youssef, B. Pradhan, and N. H. Maerz,
Debris Flow Impact Assessment caused by 14
April 2012 Rainfall along the Al-Hada
Highway, Kingdom of Saudi Arabia using
High-Resolution Satellite Imagery, Arabian
Journal of Geosciences, 7(7), 2014, pp. 2591-
2601.
[10] S. Ghosh, D., Das, S. C. Kao, and A. R.
Ganguly, Lack of Uniform Trends but
Increasing Spatial Variability in Observed
Indian Rainfall Extremes, Nature Climate
Change, 2(2), 2012, 86.
[11] J. Panthi, P. Dahal, M. Shrestha, S. Aryal, N.
Krakauer, S. Pradhanang, ... and R. Karki,
Spatial and Temporal Variability of Rainfall in
The Gandaki River Basin of Nepal Himalaya,
Climate, 3(1), 2015, pp. 210-226.
[12] N. A. Elagib, and A. S. A. Abdu, Development
of Temperatures in The Kingdom of Bahrain
from 1947 to 2005, Theoretical and Applied
Climatology, 101, 2010, pp. 269-279.
[13] P. K. Naik, M. Mojica, F. Ahmed, and S. Al-
Mannai, Storm Water Injection in Bahrain:
Pilot Studies, Arabian Journal of Geosciences,
10(20), 2017, 452.
[14] S. Landau, and I. C. Ster, Cluster Analysis:
Overview, Á Á, 11(x12), 2010, x1p.
[15] B. S. Duran, and P. L. Odell, Cluster Analysis:
A Survey (Vol. 100), Springer Science &
Business Media, 2013.
[16] T. J. Roelandt, and P. den Hertog, Cluster
Analysis and Cluster-based Policy Making:
The State of The Art, Boosting Innovation: The
Cluster Approach, 1999, pp. 413-427.
[17] R. Xu, and D. Wunsch, Survey of Clustering
Algorithms, IEEE Transactions on Neural
Networks, 16(3), 2005, pp. 645-678.
Engineering World
DOI:10.37394/232025.2024.6.5
Uneb Gazder
E-ISSN: 2692-5079
33
Volume 6, 2024
[18] A. Nagpal, A. Jatain, and D. Gaur, Review
Based on Data Clustering Algorithms, In 2013
IEEE Conference on Information &
Communication Technologies, April 2013, pp.
298-303, IEEE.
[19] O. A. Abbas, Comparisons between Data
Clustering Algorithms, International Arab
Journal of Information Technology (IAJIT),
5(3), 2008.
[20] P. Govender, and V. Sivakumar, Application of
K-Means and Hierarchical Clustering
Techniques for Analysis of Air Pollution: A
Review (1980–2019), Atmospheric Pollution
Research, 11(1), 2020, pp. 40-56.
[21] S. Mabon, The Battle for Bahrain: Iranian-
Saudi Rivalry, Middle East Policy, 19(2), 2012,
84.
[22] F. H. Lawson, Bahrain: The Modernization of
Autocracy, Routledge, 2019.
[23] R. Hasan, S. M. Suliman, and Y. A. Malki, An
Investigation into The Delays in Road Projects
in Bahrain, International Journal of Research
in Engineering and Science, 2(2), 2014, pp. 38-
47.
[24] H. Radhi, F. Fikry, and S. Sharples, Impacts of
Urbanisation on The Thermal Behaviour of
New Built Up Environments: A Scoping Study
of The Urban Heat Island in Bahrain,
Landscape and Urban Planning, 113, 2013, pp.
47-61.
[25] J. Saundalkar, First Phase of Bahrain’s Sheikh
Zayed Highway to go Ahead; Western
Bainoona Group involved, ME Construction
News, available at
https://meconstructionnews.com/35443/first-
phase-of-bahrains-sheikh-zayed-highway-to-
go-ahead-western-bainoona-group-involved, 24
June 2019, accessed on 29th August 2019.
[26] Y. V. Burlachenko, and B. A. Snopok,
Methods of Cluster Analysis in Sensor
Engineering: Advantages and Faults,
Semiconductor Physics Quantum Electronics &
Optoelectronics, 2010.
[27] S. Sagiroglu, and D. Sinanc, Big Data: A
Review, In 2013 International Conference on
Collaboration Technologies and Systems (CTS)
(pp. 42-47), May, 2013. IEEE.
[28] R. S. Madhulatha, An Overview on Clustering
Methods, IOSR Journal of Engineering, Vol.
2(4), April 2012, pp: 719-725.
[29] A. K. Jain, Data Clustering: 50 Years beyond
K-Means, Pattern Recognition Letters, 31(8),
2010, pp. 651-666.
[30] E. Esa, and M. Assen, A GIS based Land
Suitability Analysis for Sustainable
Agricultural Planning in Gelda Catchment,
Northwest Highlands of Ethiopia, Journal of
Geography and Regional Planning, 10(5),
2017, pp. 77-91.
[31] S. Fotheringham, and P. Rogerson, Spatial
analysis and GIS, CRC Press, 2014.
[32] H. Hasanean, and M. Almazroui, Rainfall:
Features and Variations over Saudi Arabia, A
Review, Climate, 3(3), 2015, pp. 578-626.
[33] R. B. Ouarda, C., Charron, K. N. Kumar, P. R.
Marpu, H. Ghedira, A. Molini, and I. Khayal,
Evolution of The Rainfall Regime in The
United Arab Emirates, Journal of Hydrology,
514, 2014, pp. 258-270.
[34] O. Planchon, H. Quénol, N. Dupont, and S.
Corgne, Application of The Hess-Brezowsky
Classification to The Identification of Weather
Patterns causing Heavy Winter Rainfall in
Brittany (France), Natural Hazards and Earth
System Sciences, 9(4), 2009, pp. 1161-1173.
Contribution of Individual Authors to the
Creation of a Scientific Article (Ghostwriting
Policy)
Uneb Gazder is the sole author of this paper who
carried out all tasks related to the study shown in
this paper.
Sources of Funding for Research Presented in a
Scientific Article or Scientific Article Itself
No funding was received for conducting this study.
Conflict of Interest
The author has no conflicts of interest to declare that
are relevant to the content of this article.
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the
Creative Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en
_US
Engineering World
DOI:10.37394/232025.2024.6.5
Uneb Gazder
E-ISSN: 2692-5079
34
Volume 6, 2024
