PHYSICSAccess Research Paper

ISSN Online: 2756-3898, ISSN Print: 2714-500X [Link]

Cluster Analysis on Radioactivity Levels of K-40,

Ra-226, and Th-232 at Chikun Environment of
Kaduna State, Nigeria, Using the R Package
Handan T Elisha1, Sarki S Habila2, Shirley O Yakubu2, Umar Isah3, and Davou H
Daze4
1 Department of Statistics, Nuhu Bamalli Polytechnic Zaria, Kaduna State, Nigeria
2
Department of Science and Laboratory Technology, Federal Polytechnic, Kaltungo, Gombe State,
Nigeria
3
Department of Medical Imaging Technology, Shehu Idris College of Health Sciences and Technology,
Makarfi, Kaduna State, Nigeria
4
Department of General Studies, Plateau State College of Nursing Sciences, Vom, Plateau State,
Nigeria

Corresponding E-mail: ethandan14@[Link]

Received 02-10-2025
Accepted for publication 07-11-2025
Published 10-11-2025

Abstract
Exposure to ionizing radiation from both natural and anthropogenic sources poses serious
environmental and health concerns. This study used hierarchical clustering to investigate the
spatial distribution of naturally occurring radionuclides K-40, Ra-226, and Th-232 in soil
samples collected from 14 locations within Chikun Local Government Area (LGA), Kaduna
State, Nigeria. Using the R statistical package, single and complete linkage methods were
implemented on Euclidean distance matrices to identify natural groupings among the sampling
sites. The optimal number of clusters, determined through the elbow and silhouette methods,
was found to be two (2), with further sub-groupings evident in the complete linkage
dendrogram. The first cluster comprises the KRPC1 and KRPC2 sampling sites, situated near
the Kaduna Refining and Petrochemical Company. These sites exhibited the highest mean
activity concentrations of the measured radionuclides, with values of 106.14 Bq/kg for K-40,
265.84 Bq/kg for Ra-226, and 827.52 Bq/kg for Th-232, respectively. The second cluster
comprised residential neighborhoods and nearby waste disposal sites, demonstrating
comparatively lower activity concentrations of the radionuclides. This spatial distribution
suggests that refinery operations exert a pronounced influence on localized radioactivity levels,
while the surrounding residential and dumpsite areas exhibit only residual contamination.
These results underscore the effectiveness of cluster analysis as a tool for interpreting
environmental radioactivity patterns and reinforce the need for sustained monitoring and public
health safeguards in industrialized settings.

Keywords: Cluster Analysis; Dendrogram; Linkage; Radioactivity; R package.

VOLUME 05, ISSUE 02, 2025 48 ©DOP_KASU Publishing

PHYSICSAccess Handan et al.

area of Kaduna Metropolis using gamma-ray spectrometry

I. INTRODUCTION and reported elevated radioactivity levels at several locations.

I onizing radiation is a natural environmental hazard that can However, given that many of these sites are characterized by
cause significant physiological (somatic) and genetic waste deposits from the Kaduna Refining and Petrochemical
damage. With the discovery of atomic and nuclear energy and Company (KRPC), alongside other industrial and domestic
advancements in technology, the use of man-made radiation, activities, the study did not examine whether shared
especially in power production, along with the increased environmental or anthropogenic factors might explain the
likelihood of human exposure to harmful radiation, has spatial variations observed.
continued to rise [1]. The present study, therefore, seeks to determine the
Several studies have been conducted to assess the level of presence of natural groupings among the sampling locations
radioactivity of radioactive elements, both within and outside in Chikun Local Government Area (LGA), Kaduna State,
Nigeria, to understand the sources, effects, and risks Nigeria, based on the radionuclide activity concentrations
associated with ionizing radiation. These studies have reported in [1]. To achieve this, Complete Linkage and Single
emphasized the importance of having accurate information Linkage hierarchical clustering algorithms were applied using
about the health effects of exposure to high levels of the R statistical software.
background radiation [2].
Reference [3] studied the impact of environmental radiation II. MATERIALS AND METHODS
on the incidence of cancer and birth defects in areas with high A. Materials
natural radioactivity and found that natural radioactive 1) R Package
elements released from crystalline rocks can enter the food The primary analytical tool employed in this study was the R
chain, including soil, plants, animals, and humans, leading to statistical software (version 4.5.1), an open-source
increased radiation exposure and potential health effects. programming environment recognized for its robustness,
Natural occurring radioactivity in the Earth's crust may be
versatility, and broad accessibility [9].
categorized into virgin and modified natural sources. Virgin
sources are of cosmogenic or primordial (terrestrial) origin 2) Data Source
and have been present since the Earth’s early geological The dataset utilized in this study was derived from the work
history. In contrast, modified natural sources result from reported in [1], which determined the activity concentrations
anthropogenic activities such as mining, fossil fuel of K-40, Ra-226, and Th-232 in soil samples collected from
combustion, fertilizer production, and the use of naturally the Chikun area of Kaduna Metropolis using gamma-ray
derived materials in construction. Gamma emissions from spectrometry (see Table I).
radionuclides such as K-40 and members of the Th-232 decay
series, along with their progeny, constitute the principal Table I. Result of K-40, Ra-226 and Th-232 Obtained
external source of irradiation to the human body [4]. from Soil Samples in Chikun L.G.A [1]
The application of multivariate analytical techniques has Sample ID K-40 Ra-226 Th-232
proven valuable for researchers, as it facilitates the KP1 228.3 28.2 28.8
identification of underlying patterns and relationships among KP2 517.6 19.6 50.7
variables, as well as associations between variables and the KP3 122.2 89.5 147.5
corresponding sampling sites [5]. Comparable observations SBR1 29.40 48.3 106.2
were reported in [6], where the authors employed the R SBR2 152.6 33.0 105.1
statistical software to investigate the relationships between SBR3 378.1 36.2 37.9
metal concentrations and their spatial distribution, as well as GN1 118.5 73.1 22.1
the associations between the metals and radiological activity GN2 175.9 60.1 24.2
levels. GN3 144.9 17.6 84.7
Other multivariate approaches have also been effectively RIDO1 63.00 40.4 107.8
applied within the Nigerian context. For example, [7] RIDO2 97.20 16.0 25.1
examined soils in the vicinity of a coal-fired cement plant and RIDO3 211.0 35.2 133.8
attributed the elevated activity concentrations of natural KRPC1 889.73 108.6 243.1
radionuclides to industrial emissions and combustion residues. KRPC2 765.3 103.7 288.6
Similarly, [8] reported increased levels of uranium- and
B. Methods
thorium-series radionuclides in peri-urban Emure-Ekiti,
southwestern Nigeria, linking the observed spatial variations 1) Data Importation
in radioactivity to both anthropogenic activities and The acquired dataset was exported from Microsoft Excel as a
underlying geological formations. comma-separated values (CSV) file before being imported
Reference [1] assessed the activity concentrations of K-40, into R for subsequent analysis.
Ra-226, and Th-232 in soil samples collected from the Chikun

VOLUME 05, ISSUE 02, 2025 49 ©DOP_KASU Publishing

PHYSICSAccess Handan et al.

2) Cluster Analysis Step 3. Combine objects r and s into a new cluster (rs)
Cluster analysis, a data exploration (mining) tool for dividing employing some criterion, and reduce the number of clusters
a multivariate dataset into “natural” clusters (groups) which by 1 by deleting the row and column for objects r and s.
may exist in a dataset, was used to sort the different objects Calculate the dissimilarities between the cluster (rs) and all
(location) into groups such that the degree of association remaining clusters, using the criterion, and add the row and
between two objects is maximal if they belong to the same column to the new dissimilarity matrix.
group and minimal otherwise [10]. To achieve this, distance Step 4. Repeat steps 2 and 3, (𝑛 − 1) times until all objects
measures were computed to construct a proximity matrix from form a single cluster. At each step, identify the merged clusters
which a specific iterative clustering algorithm is applied. In and the value of the dissimilarity at which the clusters are
this case, the single and complete linkage algorithms were merged.
The agglomerative hierarchical procedures were adopted, viz,
employed to comparatively assess areas of similarity by both
the single and complete linkage algorithms.
algorithms.
i) Single Link (Nearest-Neighbour) Method
a) Distance Measures To implement the single-link method, we combine objects in
To initiate a cluster analysis, a proximity matrix, which clusters using the minimum dissimilarity between clusters.
represents the strength of the relationship in terms of Then new distances between R and S are calculated using (6).
similarities/distance (dissimilarities) between pairs of rows in 𝑑( )( ) = 𝑚𝑖𝑛{𝑑 |r ∈ R and s ∈ S} (6)
𝑦 × is constructed. As in the case of continuous (interval,
ratio scale) variables, the Euclidean distance between two ii) Complete Link (Farthest-Neighbour) Method
objects, the most common dissimilarity measure, was used For the complete link procedure, we also combine objects in
[11]. clusters using the minimum dissimilarity between clusters.
Given an (n × p) matrix Y with (1 × p) row vectors 𝑦 , the Then, new distances between clusters R and S are calculated
Minkowski measure between two rows 𝑦 𝑎𝑛𝑑 𝑦 is defined using the rule: -
as; 𝑑( )( ) = 𝑚𝑎𝑥{𝑑 |r ∈ R and s ∈ S} (7)
𝑑 = ∑ 𝑦 −𝑦 (1)
c) Cluster Optimality
With this, the study adopts the Euclidean distance measure, Normally, for a hierarchical clustering algorithm, a visual
which is the most popular variant of the Minkowski measure. inspection or a priori information might suffice in determining
When 𝑟 = 2, this gives (2). the number of clusters constituted. However, to quantitatively
assess the adequacy of the clustering structure obtained
𝑑 = ∑ 𝑦 −𝑦 through the hierarchical methods, two validation measures
were computed.
𝑑 = [(𝑌 − 𝑌 ) (𝑌 − 𝑌 )] (2)
i) Elbow Plot
The (n × n) data matrix 𝑫 = [𝑑 ] is called the Euclidean This is a two-dimensional plot of the Within-Cluster Sum of
distance matrix, and given two objects 𝑌 𝑎𝑛𝑑 𝑌 in a p- Squares (WCSS) against increasing k-values and looking for
dimensional space, a dissimilarity measure satisfies the a point (the elbow) where the improvement slows down [12].
following conditions. ii) Silhouette Index Plot
𝑑 ≥ 0 for all objects 𝑦 𝑎𝑛𝑑 𝑦 (3) The silhouette index (𝑠(𝑖) ) measures the degree of separation
between clusters and compactness within clusters and is
𝑑 = 0 if and only if 𝑦 = 𝑦 (4) defined as:
() ()
𝑑 =𝑑 (5) 𝑠(𝑖) = (8)
[ ( ), ( )]

b) Clustering Algorithms Where 𝑎(𝑖) represents the average dissimilarity of sample i

Given the proximity matrix 𝐷 × = [𝑑 ], the steps for the with all other samples in the same cluster, and 𝑏(𝑖) is the
agglomerative hierarchical clustering algorithm as executed in minimum average dissimilarity of i with samples in other
this research are as follows [10,11]: clusters.
Step 1. Begin with n clusters, each containing only a single Hence, our interest was to plot the 𝑠(𝑖) values against
object. increasing k-values and spot the highest point to mark the
Step 2. Search the dissimilarity matrix D for the most similar optimal number of clusters.
pair. Let the pair chosen be associated with the element 𝑑 ,
so that objects r and s are selected. In the case of single III. RESULTS AND DISCUSSION
linkage, the minimum value that connects a pair is considered,
whereas the largest value is considered in the case of complete This section presents the result of the cluster analysis
linkage. performed on the data obtained from [1]. More importantly,
analyzing the result of dendrograms produced by the single

VOLUME 05, ISSUE 02, 2025 50 ©DOP_KASU Publishing

 PHYSICSAccess Handan et al.

and complete linkage algorithms, given the computed 𝑘 = 2 and 𝑘 = 3, with the steeper bend at 𝑘 = 3, implying
Euclidean distance measures as shown in the proximity matrix that the optimal number of clusters could be two or three.
in Table II. The Elbow plot and silhouette index plot for Hence, given the unclear nature of this check, further
cluster optimality are depicted in Figs. 1 and 2, respectively. investigation using the silhouette index plot in Fig. 2 clearly
The elbow plot shows the presence of two elbow-like bends at reveals an optimal cluster number of 𝑘 = 2.
Table II. Proximity Matrix for the Locations in Chikun LGA Computed from their Radioactive Level [1].
Location GN1 GN2 GN3 KRPC1 KRPC2 KP1 KP2 KP3 SBR1 SBR2 SBR3 RIDO1 RIDO2
GN2 58.89
GN3 87.73 80.17
KRPC1 803.05 748.21 766.91
KRPC2 700.22 647.46 658.70 132.58
KP1 118.81 61.52 100.96 699.91 601.30
KP2 403.68 345.11 374.25 428.28 353.59 290.26
KP3 126.52 137.66 98.13 773.70 658.55 170.60 413.03
SBR1 125.01 168.30 121.43 873.24 760.19 214.37 492.18 109.61
SBR2 98.28 88.44 26.69 753.74 643.48 107.59 369.27 76.90 124.15
SBR3 262.69 204.07 238.58 555.98 466.19 150.29 141.07 283.44 355.53 235.32
RIDO1 107.21 141.86 88.10 840.50 727.96 183.61 458.64 86.55 34.55 89.95 322.79
RIDO2 61.02 90.22 76.35 827.17 723.52 131.72 421.19 144.94 110.53 98.78 281.92 92.76
RIDO3 149.90 117.75 84.20 691.38 579.57 106.65 318.04 104.98 184.15 65.11 192.67 150.36 158.54

1000000
Fig. 3(a) and (b) present the dendrograms that were
generated using the single and complete linkage algorithms,
respectively. As recommended by the silhouette plot, the two
750000
Total Within Sumof Square

clusters consist of KRPC1 and KRPC2 as cluster 1, with mean

activities of K-40, Ra-226, and Th-232 at 106.14 Bq/kg,
265.84 Bq/kg and 827.52 Bq/kg, respectively, while every
500000
other location belongs to cluster 2.
Analysis of Fig. 4, which provides a visual representation
250000
of elemental levels, reveals that Cluster 1 exhibits the highest
concentrations across all elements. This finding is
unsurprising, as the locations within Cluster 1 are situated
0 around the Kaduna Refinery, which likely contributes to the
1 2 3 4 5 6 7 8 9 10 elevated radiation levels. This observation agrees with [7],
Number of clusters k
who attributed high natural radionuclide concentrations
Fig. 1. Elbow plot for the number of clusters. around a coal-fired cement plant to increased industrial
emissions.
To further elucidate the similarity patterns within Cluster 2,
comparative visual examination of Fig. 3(a) and Fig. 3(b)
0.6
indicates that, at linkage heights of 150 and 200, respectively,
Cluster 2 further resolves into three distinct sub-clusters:
Average silhouette width

• Sub-cluster A (SBR1 and RIDO1) corresponds to the

0.4
dumpsite locations, exhibiting mean activity concentrations of
46.2 Bq/kg (K-40), 44.35 Bq/kg (Ra-226), and 107 Bq/kg (Th-
232). The elevated levels observed here are likely attributable
0.2
to the accumulation of heterogeneous waste materials and
enhanced organic matter decomposition processes.
• Sub-cluster B (KP2 and SBR3) comprises residential areas
0.0
situated in proximity to the refinery. These sites display
1 2 3 4 5 6
Number of clusters k
7 8 9 10
intermediate radionuclide activities, with mean concentrations
Fig. 2. Silhouette index plot. of 447.85 Bq/kg (K-40), 27.9 Bq/kg (Ra-226), and 44.3 Bq/kg

VOLUME 05, ISSUE 02, 2025 51 ©DOP_KASU Publishing

PHYSICSAccess Handan et al.

(Th-232), suggesting secondary dispersion and localized activity concentrations of 156.33 Bq/kg (K-40), 44.09 Bq/kg
environmental redistribution associated with the refinery’s (Ra-226), and 71.41 Bq/kg (Th-232). These values indicate
operations. that radioactivity levels here are predominantly influenced by
• Sub-cluster C (KP1, RIDO2, GN1–GN3, SBR2, and RIDO3) routine domestic and commercial activities rather than
represents residential zones located farther from direct industrial emissions.
industrial influence, characterized by comparatively lower

Fig. 3. Dendrograms from the (a) Single and (b) Complete Linkage Algorithms for Chikun LGA.

a discernible influence on radionuclide levels in its immediate

Th-232
800

Ra-226 surroundings, with diminishing yet notable effects extending

K-40 to more distant locations. The observed increase in human
settlement and activities around these affected zones is
Radiation level(Bq/Kg)

600

therefore a cause for concern and underscores the need for

targeted environmental management and public health
intervention strategies.
400

APPENDIX
R Source Code
200

# DATA IMPORTATION
# -----------------------------------------
[Link] <- [Link]([Link](), [Link] = 1)
#Note: all object assignments in the R code block are
0

KRPC1 KRPC2 SBR1 RIDO1 KP2 SBR3 GN1 GN2 GN3 KP1 KP3 SBR2 RIDO2 RIDO3
done specifically to this research.

Sample ID # PROXIMITY MATRIX CALCULATION

Fig. 4. Barchart showing levels of K-40, Ra-226 and Th-232 # -----------------------------------------
# Compute Euclidean distance matrix and Display
in Chikun LGA [1].
proximity matrix
[Link] <- dist([Link], method =
IV. CONCLUSION "euclidean")
This study examined the natural grouping of sampling print([Link])
locations within Chikun Local Government Area (LGA),
# HIERARCHICAL CLUSTERING - SINGLE LINKAGE
Kaduna State, Nigeria, based on measured activity
# ------------------------------------------
concentrations of K-40, Ra-226, and Th-232. Radioactivity # Execute single linkage clustering
data were obtained from fourteen locations across the Chikun [Link] <- hclust([Link], method = "single")
environs. Cluster analysis was subsequently applied, and the # Plot single linkage dendrogram
outcomes of both single-linkage and complete-linkage plot([Link],
hang = -1,
methods indicate that operations of the Kaduna Refinery exert
cex = 0.6,

VOLUME 05, ISSUE 02, 2025 52 ©DOP_KASU Publishing

PHYSICSAccess Handan et al.

xlab = "Sample ID", Report to the General Assembly, with Scientific

main = "Single Linkage Dendrogram") Annexes, vol. I, United Nations, New York, NY,
# Demarcate clusters (k=2)
USA, 2000.
[Link]([Link], k = 2, border = "blue")
[5] L. L. Harlow, The Essence of Multivariate Thinking:
# HIERARCHICAL CLUSTERING - COMPLETE LINKAGE Basic Themes and Methods, 2nd ed. New York, NY,
# --------------------------------------- USA: Routledge, 2014.
[6] F. Caridi, A. F. Mottese, G. Paladini, L. Pistorino, F.
# Execute complete linkage clustering Gregorio, S. Lanza, et al., “Multivariate statistics,
[Link] <- hclust([Link], method = radioactivity and radiological hazard evaluation in
"complete") marine sediments of selected areas from Sicily,
# Plot complete linkage dendrogram
Southern Italy,” J. Mar. Sci. Eng., vol. 13, no. 4, p.
plot([Link], 769, 2025.
hang = -1, [7] M. T. Kolo, M. U. Khandaker, and H. K. Shuaibu,
cex = 0.6, “Naturally radioactivity in soils around mega coal-
xlab = "Sample ID", fired cement factory in Nigeria and its implication on
main = "Complete Linkage Dendrogram") human health and environment," Arab. J. Geosci.,
# Demarcate clusters (k=2) vol. 12, no. 15, 2019.
[Link]([Link], k = 2, border = "red")
[8] O. J. Popoola, O. E. Olubi, O. S. Adewalure, and A.
# CLUSTER OPTIMALITY - ELBOW METHOD
# --------------------------------------- E. Raphael, "Elevated natural radionuclides in soils
library(factoextra) and stream sediments: pollution, spatial distribution,
fviz_nbclust([Link], hcut, method = "wss") + radiological hazards, and cancer risks in peri-urban
labs(subtitle = "Elbow Method") Emure-Ekiti, southwest Nigeria." Discover Soil, vol.
# CLUSTER OPTIMALITY - SILHOUETTE METHOD 2, no. 1, p. 70, 2025.
# --------------------------------------- [9] A. Kassambara, Practical Guide to Cluster Analysis
fviz_nbclust([Link], hcut, method =
in R: Unsupervised Machine Learning, 1st ed., vol.
"silhouette") +
labs(subtitle = "Silhouette Method")
1. STHDA, 2017.
[10] N. H. Timm, Applied Multivariate Analysis. New
# BARCHART York, NY, USA: Springer-Verlag, 2002.
# --------------------------------------- [11] A. K. Jain and R. C. Dubes, Algorithms for
barplot(t([Link]), Clustering Data. Englewood Cliffs, NJ, USA:
beside=TRUE,col=c("lightblue", "red", "lightcyan"), Prentice-Hall, 1988.
[Link]=c("Th-232","Ra-226"," K-40"),
[12] P. N. Tan, M. Steinbach, and V. Kumar, Introduction
[Link]=0.4,
xlab="Sample ID",
to Data Mining. New Delhi, India: Pearson
ylab="Radiation level(Bq/Kg)", Education, 2016.
[Link] = list(x = "topright", cex = 0.6))

References
[1] P. M. Gyuk, S. S. Habila, M. D. Dogara, N. Kure, H.
I. Daniel, and T. E. Handan, “Determination of
radioactivity levels in soil samples at Chikun
environment of Kaduna metropolis using gamma ray
spectrometry,” Sci. World J., vol. 12, no. 2, pp. 52–
55, 2017.
[2] S. H. Sarki, S. Abew, M. A. Musa, and A. M.
Ibrahim, “Determination of activity concentration
level of ²²⁶Ra, ⁴⁰K and ²³²Th in soil within Igabi Local
Government Area of Kaduna State, Nigeria,” Sci.
World J., vol. 15, no. 1, pp. 113–118, 2020.
[3] A. Zlobina, I. Farkhutdinov, F. P. Carvalho, N.
Wang, T. Korotchenko, N. Baranovskaya, and A.
Farkhutdinov, “Impact of environmental radiation on
the incidence of cancer and birth defects in regions
with high natural radioactivity,” Int. J. Environ. Res.
Public Health, vol. 19, no. 14, p. 8643, 2022.
[4] United Nations Scientific Committee on the Effects
of Atomic Radiation (UNSCEAR), Sources and
Effects of Ionizing Radiation: UNSCEAR 2000

VOLUME 05, ISSUE 02, 2025 53 ©DOP_KASU Publishing