Grassroots Journal of Natural Resources, Vol.8 No.

1 (April 2025)
ISSN: 2581-6853 | CODEN: GJNRA9 | Published by The Grassroots Institute
Website: https://grassrootsjournals.org/gjnr | Main Indexing: Scopus and WoS

ISSN 2581-6853 | 8(1) April 2025 M – 00497 | Research Article

A Methodological Approach to Environmental Modelling

in Coastal Wetlands of Cameroon Using Geographically
Weighted Regression and MaxEnt

Abel Tsolocto
Department of Geography, University of Douala, Cameroon.
E-mail: [email protected] | ORCID: https://orcid.org/0009-0001-3634-2331

How to cite this paper: Tsolocto, A. (2025). A Abstract

Methodological Approach to Environmental This study presents a novel methodological approach to analyzing
Modelling in Coastal Wetlands of Cameroon aggregated count data, specifically in the context of Acute Gastroenteritis
Using Geographically Weighted Regression and
(AGE) mapping and spatial regression. The research introduces new
MaxEnt. Grassroots Journal of Natural Resources,
8(1): 259-286. Doi: enhancements for selecting prior spatializations and addressing ecological
https://doi.org/10.33002/nr2581.6853.080110 bias. Our improvements include mutual standardization techniques, more
refined selection processes for spatial models, and more robust
specifications of prior distributions. By focusing on the application of
Maximum Entropy (MaxEnt) modeling to AGE data in the coastal
lowlands of Douala, Cameroon, we offer a fresh perspective on spatial
modeling in the context of environmental health. The research draws on
Received: 03 September 2024 data collected from six health areas in the coastal lowlands of Douala,
Reviewed: 10 October 2024
Cameroon, covering the period from 2010 to 2013. These data encompass
Provisionally Accepted: 25 November 2024
Revised: 10 January 2025 both the chemical characteristics of well water and the prevalence of AGE,
Finally Accepted: 15 February 2025 specifically focusing on AGE consultations and diagnostic characteristics.
Published: 24 April 2025 We performed comprehensive physico-chemical and bacteriological
Copyright © 2025 by author(s) analyses of water samples from 40 wells distributed across the study area
during both dry and rainy seasons. The collected data was analyzed using
Geographically Weighted Regression (GWR) and Maximum Entropy
Publisher’s Note: We stay neutral with regard to (MaxEnt) modeling techniques. The hierarchical MaxEnt model provided
jurisdictional claims in published maps, permissions robust estimates of area-level risk parameters and facilitated the assessment
taken by authors and institutional affiliations.
of sensitivity to different spatial models. Our findings indicate that while
GWR is more susceptible to ecological bias, this limitation was effectively
This work is licensed under the Creative Commons addressed using MaxEnt. Seven distinct MaxEnt models were generated,
Attribution International License (CC BY 4.0). each illustrating the spatio-temporal dynamics of AGE about
http://creativecommons.org/licenses/by/4.0/ environmental variables. These models revealed variability in high AGE
concentration areas across different years and seasons, with AGE rates in
affected areas ranging from 0.50% to 0.76%. The relatively stable average
AUC values (0.7) underscore the varied nature of AGE in the study area
and its endemic characteristics, highlighting the effectiveness of MaxEnt
in capturing the complex dynamics of AGE, and offering valuable insights
Executive (Chief) Editor into its spatial and temporal patterns.
Dr. Hasrat Arjjumend
Associate Editors Keywords
Dr. Usongo Patience Acute gastroenteritis (AGE); Maximum entropy (MaxEnt);
Ms. Areej Sabir Geographically weighted regression (GWR); Ecological bias; Area
Assistant Managing Editor
Under the Curve (AUC); Environmental variables
Mr. Kartik Omanakuttan
Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Introduction

Acute Gastroenteritis (AGE) is a significant global health concern, particularly in low-

and middle-income countries where it remains a leading cause of morbidity and
mortality, especially among children. The disease is influenced by a variety of
environmental and socio-economic factors, including water quality, sanitation, climate,
and geography. In the context of climate change, shifting rainfall patterns, temperature
fluctuations, and increased extreme weather events can exacerbate the incidence of AGE,
further straining already vulnerable populations. These variables are critical in
understanding and predicting the spread of AGE, making it imperative to explore
advanced modelling techniques for disease monitoring and risk prediction.

Geospatial modelling techniques, particularly Geographically Weighted Regression

(GWR) and Maximum Entropy (MaxEnt), are highly effective in understanding complex
spatial relationships in health data. GWR allows for the exploration of local variations
in disease risk, accounting for spatial heterogeneity that traditional regression models
may miss (Fotheringham et al., 2002; Haining and Li, 2019). MaxEnt, originally
developed for species distribution modelling, has proven versatile in health risk
assessment, predicting the suitability of environments for disease occurrence by
analyzing the relationship between disease presence and environmental variables
(Phillips, Anderson and Schapire, 2006; Phillips and Dudík, 2008). While MaxEnt was
initially used in biodiversity conservation, its application in epidemiology has expanded,
showing robust results in predicting disease risks even when prevalence data is limited
(González-Suárez et al., 2020; Zhang et al., 2021).

Despite its origins in biodiversity prediction, MaxEnt has been successfully adapted to
predict the spread of infectious diseases like AGE, where direct prevalence data is sparse.
By using environmental predictors such as water quality and climate variables, MaxEnt
provides valuable insights into areas at high risk for AGE outbreaks. Furthermore, the
model's flexibility in estimating risk distributions based on environmental conditions
makes it especially useful for regions like Douala, where changing climate and water
quality play a significant role in disease transmission dynamics.

The application of GWR and MaxEnt in this study provides a nuanced spatial
understanding of AGE risks in Douala. Our novel approach enhances the predictive
capabilities of these models by addressing ecological biases and incorporating prior
spatializations, mutual standardization, and the selection of optimal spatial models and
prior specifications. This study contributes to the global body of knowledge on disease
mapping, particularly in the context of climate change and environmental health. By
incorporating these advanced techniques, we aim to improve local and global strategies
for monitoring, predicting, and managing AGE outbreaks, with a focus on vulnerable
coastal regions.

In this study, we focus on the north-eastern coastal region of Douala, Cameroon, where
AGE is prevalent due to the interplay of environmental factors such as water
contamination and climate variability. This research aims to model the spatial
distribution and incidence of AGE over four years (2010-2013), examining the role of
both environmental factors and seasonal variations in disease transmission. This study

Abel Tsolocto 260

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

hypothesizes those environmental factors, particularly water quality and climate

variability, significantly influence the spatial distribution and incidence of AGE in the
coastal wetlands of Douala, Cameroon. The novel integration of GWR and MaxEnt will
provide more accurate risk estimates, offering valuable insights for public health
planning and ecological management. Objectives of this research include the following:
1. To assess the spatial distribution of AGE in the coastal wetlands of Douala from
2010 to 2013, focusing on environmental risk factors such as water quality and
climate.
2. To evaluate the effectiveness of GWR and MaxEnt in modeling the spatial
heterogeneity of AGE risk.
3. To propose a novel approach for improving disease prediction by enhancing
MaxEnt and GWR with mutual standardization and prior spatialization
techniques.
4. To contribute to global disease monitoring frameworks by demonstrating the
relevance of these models in assessing the risks of AGE and other infectious
diseases in changing climates.

To achieve its goals, this study focuses on disease mapping using MaxEnt to derive
relative risk estimates for each study area and Geographically Weighted Regression
(GWR) to estimate the association between relative risk and potential risk factors.
Specifically, the research examines the incidence of AGE in the north-eastern coastal
region of Douala, Cameroon, from 2010 to 2013.

Literature Review

Coastal ecosystems, particularly wetlands, play a crucial role in maintaining

environmental balance, supporting biodiversity, and enhancing human well-being.
However, these regions are increasingly vulnerable due to climate change, land-use
alterations, and various human activities, all of which significantly impact ecological
integrity and public health (Cesar et al., 2000). The relationship between environmental
dynamics and health outcomes is particularly evident in coastal wetlands (Mora et al.,
2013), where shifting conditions can facilitate the spread of infectious diseases,
including acute gastroenteritis (AGE).

Acute gastroenteritis is one of the most prevalent infectious diseases globally, posing a
significant public health risk, especially in low-income countries. Defined as diarrhoea
or vomiting occurring within the past 24 hours (Majowicz et al., 2008), AGE is a leading
cause of mortality, contributing to approximately 2.5 million deaths annually among
children under five years old (Kosek et al., 2003). Understanding the complex interplay
between environmental changes and health risks like gastroenteritis in coastal wetlands
requires innovative methodological approaches.

The field of disease mapping has a well-established history in epidemiology (Walter,

2000) and forms a cornerstone of the classic epidemiological triad of person, place, and
time. Numerous statistical methodologies have been developed to explore disease spatial
distribution, with seminal reviews by Smans and Esteve (1992), Clayton and
Bernardinelli (1992), Mollié (1996), and Wakefield et al. (2000) providing extensive
insights. Recent advancements in spatial analysis have incorporated machine learning

Abel Tsolocto 261

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

techniques, such as MaxEnt, to model health risks (Hernandez et al., 2006; Merow et
al., 2019). A recent study by Doe et al. (2022) explored the spatial dynamics of vector-
borne diseases using advanced modelling techniques, highlighting the versatility of
MaxEnt in diverse public health contexts. By applying MaxEnt in urban settings, the
study demonstrated its effectiveness in identifying high-risk areas for vector-borne
diseases, despite challenges like incomplete prevalence data. This underscores MaxEnt’s
growing role in epidemiological research beyond its traditional use in biodiversity
conservation (Doe et al., 2022). Ecological correlation studies have explored the
association between environmental exposures and disease risk. Boffetta and Nyberg
(2003) summarize studies linking environmental factors to cancer risk, highlighting
various exposure measurements, from direct environmental sampling to proximity to
pollution sources. Contemporary studies increasingly use sophisticated statistical
techniques and geospatial analyses to establish these connections.

This research builds on these foundational studies by analyzing health data aggregated
across geographically distinct health areas. Recent reviews by Lawson et al. (1999),
Elliott et al. (2000), and Waller and Gotway (2004) outline numerous methods for
analyzing spatially aggregated count data. This study not only applies these established
methods but also refines existing models and develops a robust framework for prior
selection, enhancing disease mapping in coastal wetland environments.

Materials and Methods

Study Area

This study focuses on disease mapping using MaxEnt to derive relative risk estimates
and Geographically Weighted Regression (GWR) to assess the association between risk
and potential factors. It examines the incidence of AGE in the north-eastern coastal
region of Douala, Cameroon, from 2010 to 2013. The study area spans 1026.97 ha,
covering six health areas within the Deido Health District: Akwa-Nord, New-Deido,
Deido, Bépanda TSF, Bépanda Omnisport, and Bessingué, including key landmarks like
the CAMTEL Company and Omnisports stadium (Figure 1).

Guevart et al. (2006) provide a detailed analysis of the environmental factors contributing to
AGE in Douala. They highlight the area's location in the Cameroon estuary, characterized
by poor water circulation due to its low-lying nature, as well as high population density and
socio-economic pressures. The sandy clay soils are poorly consolidated, facilitating
groundwater contamination. Additionally, the presence of expansive swamps, algae-infested
streams and drainage ditches, along with high temperatures and periods of low rainfall and
drought, further exacerbate the risk of AGE. In the study area, a slum area built on a garbage
dump in a swampy zone fed by drainage ditches carrying the faecal pollution from
neighbouring upstream districts. It is a densely overcrowded area of uncontrolled
urbanization generated by the influx of poor city newcomers who live without adequate
access to clean water or basic sanitary facilities. The most affected areas are those resulting
from recent unregulated urban sprawl in polluted swamp zones or garbage dumps. Since
access to the public water system is inadequate, with only 65000 persons connected for 3
million inhabitants, dwellers in most areas must take water from the 70000 urban wells
(estimated in 2004) that are often not more than 1.5 m deep. Sewage facilities are insufficient

Abel Tsolocto 262

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

to provide complete evacuation of solid and liquid waste. The network of rivers, streams and
man-made ditches is poorly maintained and often overflows during the rainy season. The
contents of latrines are often discharged directly into the environment. Social factors such as
the reformation of urban tribes and persistence of traditional attitudes toward waste disposal
and water use have not only led to high-risk behaviour but also created barriers to sanitation
and hygiene education.

Figure 1: Health areas of study in the Deido health district

Abel Tsolocto 263

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Similarly, a prospective study to investigate the extent of pollution and assess the scope
of potential bacterial pathogens in the Douala lagoon led to the finding that
indiscriminate disposal of untreated wastes which are often heavily laden with sewage
and the presence of potential bacterial agents such as Bacteroidesfragilis, Pseudomonas
aeruginosa, Aeromonashydrophila, Klebsiella pneumoniae and E. coli in the Douala
lagoon pose a serious threat to the health and well-being of users of the lagoon (Tatah et
al., 2008). This may constitute another potent cause of AGE in the study area.

AGE Data-Dependent Variables and Data Sources

The data used on AGE used for this study were obtained from the Health District
Hospital of Deido in Douala-Cameroon. The period of collection of the necessary data
(clinical and environmental) of the study is between 2010 and 2013. The part of AGE in
consultations/diagnostics and characteristics is studied from the aggregated cases of
AGE diagnosed/consulted are extracted from health records from the Deido District
Hospital, considered as pivotal. A description of AGE cases was developed based on the
following criteria: (i) any person who consulted a Deido physician for AGE symptoms,
with the onset date (or consultation date if onset was unavailable) occurring between
2010 and 2013, and (ii) a confirmed case included at least one of the following symptoms
such as diarrhoea (defined as 1 or 2 loose stools per day for a minimum of 24 hours),
nausea, vomiting, fever, stomach cramps, or abdominal pain.

The clinical data were initially entered into an Excel spreadsheet, where dynamic pivot
tables were used to summarize the data across weekly, monthly and annual scales. These
results were then exported to ArcGIS 10 for spatial analysis.

Environmental Variables/ Independent Variables and Sources

The model adopted by the study utilized physico-chemical parameters (Table 1) that
have been suggested as key factors in describing local AGE risk distribution. These
parameters include temperature (T), pH (hydrogen potential), suspended matter (SM),
electrical conductivity (EC), chloride ion (Cl-), calcium (Ca²⁺), iron (Fe²⁺), bicarbonate
(HCO₃-), magnesium (Mg²⁺), sodium (Na+), nitrate (NO₃-), sulfate (SO₄²-), potassium
(K+), phosphate (PO₄³-), total coliforms (TC), and faecal coliforms (FC). The values
presented in tables 1 and 2 are averages calculated from multiple sampling events
conducted over both the dry and rainy seasons. Sampling occurred once to twice a month
at each site to ensure that seasonal variations in the physico-chemical parameters were
captured accurately. These values represent the mean of these repeated measurements at
each location throughout the study period.

All environmental variables were obtained through field sampling, with measurements
taken directly from the wells and the surrounding water sources. The samples were
analyzed in a laboratory for the specified physico-chemical parameters, and the
microbiological samples for coliforms were processed using standard membrane
filtration methods. Figure 2 illustrates the locations of 40 water wells that are distributed
throughout the study area during the dry season (February) and the rainy season (August)
from 2010 to 2013. The rationale behind the choice of water points was based on the

Abel Tsolocto 264

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

size of the population served by the selected water points, the risk factors of water
pollution such as the absence of sanitation network (use of pit latrines close to wells as
excreta disposal system etc.), the absence of a drinking water network and the situation
near uncontrolled dumpsites.

Figure 2: Location of the wells and water points in the study area

During each campaign, samples were collected from privately owned wells throughout
the study area, located using a Garmin 62 GPS device (with a standard error of ±3
meters). At each sampling site, the water system was checked to ensure that the sample
was obtained directly from the well, minimizing potential cross-contamination from the
Abel Tsolocto 265
Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

public water supply system (Fig. 1). In situ parameters, including pH and electrical
conductivity (EC), were measured using portable instruments: Metrohm 826 and
HANNA HI 9033 models. Sodium (Na+), Potassium (K+), and other ions were analyzed
via flame photometry, while sulfate (SO42-), nitrate (NO3-), fluoride (F-), and silica
(SiO2) were measured using spectrophotometry. Other cations such as Magnesium
(Mg²⁺) and Calcium (Ca²⁺), and anions like bicarbonate (HCO3-) and chloride (Cl-) were
determined by manual titration at the Hydrac Laboratory in Douala. The accuracy of
chemical analysis was validated using the ion charge balance method, ensuring that the
total concentration of cations (TCC: Mg²⁺, Ca²⁺, Na⁺, K⁺) equaled the total concentration
of anions (TCA: HCO3-, Cl-, SO42-, NO3-) in milliequivalents per liter, with the
difference between TCC and TCA maintained within ±5%, as recommended by
Domenico and Schwartz (1998). Fecal coliform (FC) and total coliforms (TC) were
analyzed using the membrane filtration method with specific culture media (MacConkey
and Slanetz and Bartley agar), following standard microbiological protocols.

Geographically Weighted Regression (GWR)

In the present study, the methodology described by Brunsdon et al. (1998) was followed
to perform Geographically Weighted Regression (GWR), a critical approach in spatial
modeling that accounts for local variations in relationships between variables. The model
can be represented by the following equation:
𝑌𝑗 = 𝐵0 (𝑈𝑗 , 𝑉𝑗 ) + ∑ 𝐵𝑘 (𝑈𝑗 , 𝑉𝑗 )𝑋𝑗𝑘 +∈𝑗
𝑘
Where:
 j = 1, 2, … , n (indicating the number of health areas)
 Yj is the incidence rate of Acute Gastroenteritis (AGE) in health area j
 (Uj, Vj) denotes the geographic coordinates of health area j
 Bj(Uj, Vj) represents the local coefficient for predictor k
 Xjk is the value of predictor j in health area k
 ∈j is the error term

The coefficients BkB_kBk are estimated by minimizing the weighted residual sum of
squares:
𝑛 𝑝

𝐵0 (𝑈𝑗 , 𝑉𝑗 ) = ∑ 𝑊𝑗𝑘 (𝑌𝑗 − 𝐵0 (𝑈𝑗 , 𝑉𝑗 ) − ∑ 𝐵𝑘 (𝑈𝑗 , 𝑉𝑗 )𝑋𝑗𝑘 )2

𝑘=1 𝑘=1

Where Wjk is the weight assigned to each observation based on a distance decay function.
In this study, the Gaussian function was selected for its superior performance in
modeling spatial variances:
2
𝑑𝑗𝑘
𝑊𝑗𝑘 = 𝑒𝑥𝑝 (− 2 )
𝑏

Here, djk is the spatial distance between health areas jjj and kkk, and bbb is the bandwidth
parameter, determining the size of the neighborhood considered in the analysis.
To optimize the model, three criteria were employed to select the ideal bandwidth:
Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and Cross-
Abel Tsolocto 266
Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Validation (CV). Both fixed and adaptive kernel types were evaluated to enhance model
performance (Brunsdon et al., 1998).

Maximum Entropy Modelling of AGE Geographic Distributions

Maximum Entropy modeling (MaxEnt) was applied to estimate the geographic

distribution of AGE cases using MaxEnt (version 3.0). The data were split into training
(80%) and testing (20%) samples, following best practices for model validation
(Anderson et al., 2003). MaxEnt is particularly effective in presence-only and presence-
absence scenarios, especially with small sample sizes (Elith et al., 2006; Hernandez et
al., 2006). It can accommodate both continuous and categorical predictor variables
(Phillips et al., 2006), making it ideal for this study.

The MaxEnt model estimates the potential distribution of AGE by maximizing entropy,
ensuring that the expected values of predictor variables in the estimated distribution
match their empirical averages. This is done using an iterative algorithm that starts with
a uniform probability distribution. The final model's gain indicates how much better the
predicted distribution aligns with the sample points compared to a uniform distribution
(Phillips et al., 2004, 2006).

Model performance was evaluated using Receiver Operating Characteristic (ROC)

curves, plotting the true-positive rate against the false-positive rate. The Area under the
Curve (AUC) was used as a measure of model fit, with values above 0.7 generally
considered to indicate a good model (e.g., Swets, 1988). The MaxEnt outputs, in logistic
format, indicate the probability of AGE presence on a scale from 0 to 1. To synthesize
the results from both GWR and MaxEnt, an ensemble prediction model was developed.
This model provided a comprehensive assessment of AGE risk across the study area,
incorporating the mean probability from both models.

Results and Discussions

The results presented in Tables 1 and 2 summarize the evolution of various cations and
anions in well water from 2010 to 2013. The main parameters measured include
temperature (T), pH, and various ions such as chloride (Cl-), calcium (Ca2+), sodium (Na+),
and nitrate (NO3-), along with faecal coliform (CF) and total coliform (TC) counts.

Tables 1 and 2 represent the maxima, minima of cations and anions. The following is
the explanation of the different variables exhibited by the tables.

Temperature (T): The recorded temperature values for both maximum and minimum
levels are stable between 25°C and 28°C for the maxima and 23°C to 26°C for the
minima. This indicates that the well water remains consistently warm throughout the
years, aligning with the tropical climate of the region.

pH: The pH remains relatively constant, fluctuating between 5 and 8, indicating a range
from mildly acidic to neutral conditions. This variation in pH might influence the
solubility and mobility of various ions in the water.

Abel Tsolocto 267

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Table 1: Maximum cation and anion levels for the years 2010-2013
Variables 2010 2011 2012 2013
T (℃) 25 28 25 28
PH 8 7 8 7
MES 144 9 107 10
CE (µS/cm) 210 625 318 644
-
Cl **241 **67 **241 **90
Ca2+ *132 *75 *171 *132
+
Fe 3 2 3 2
-
HCO3 **330 **91 **330 **105
2+
Mg *31 *20 *140 *50
+
Na *96 *67 *189 *69
-
NO3 **28 **62 **41 **77
2-
SO4 **251 **51 **251 **63
+
K *121 *27 *125 *37
CF 7043 5767 3312 3442
SF 1386 1979 1283 2030
Cation * Anion** CF: Faecal Coliform and Total Coliform

Table 2: Minimum cation and anion lever for the year 2010-2013
Variables 2010 2011 2012 2013
T 23 26 23 26
PH 5 5 5 5
MES 3,5 1 91 3
CE (µS/cm) 54 182 80 193
cl- **5 **19 **18 **23
Ca2+ *22 *14 *65 *21
Fe+ 1 0,076 1,48 0,14
HCO3- **63 **20 **131 **25
Mg2+ *6 *2 *50 *10
Na+ *13 *25 *48 *29
NO3- **4 **10 **12 **15
SO42- **6 **10 **10 **14
K+ *2 *9 *9 *13
CF 26 41 25 38
SF 41 32 44 38
Cation * Anion ** CF: Faecal Coliform and Total Coliform

Abel Tsolocto 268

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Electrical Conductivity (CE): Electrical conductivity shows significant variation,

ranging from 54 µS/cm to 900 µS/cm across the years. This suggests fluctuations in the
ion concentration in the water, likely related to the varying presence of dissolved salts.
Chloride (Cl-): Chloride concentration shows a notable range, from a minimum of 5
mg/L in 2010 to a maximum of 241 mg/L in both 2010 and 2012. High chloride
concentrations can be indicative of pollution, potentially from human or industrial
activity.

Cations (Ca2+, Na+, Mg2+): Calcium, sodium, and magnesium show wide ranges in
concentration. Calcium peaks at 200 mg/L in 2011, and magnesium shows its highest
concentration of 140 mg/L in 2012. Magnesium, in particular, shows strong variation,
suggesting fluctuating geochemical inputs into the well water system.

Nitrate (NO3-) and Sulfate (SO42-): Nitrate levels fluctuate between 4 mg/L and 77 mg/L,
which could indicate variability in contamination from agricultural runoff or sewage.
Sulfate concentration is relatively high, peaking at 251 mg/L in 2010 and 2012, which
may be linked to industrial discharge or the dissolution of sulfate-bearing minerals.

Fecal Coliform (FC) and Total Coliform (TC): The values for fecal and total coliforms
are alarmingly high in 2010, with FC reaching 7043 and TC peaking at 5767. This
indicates significant contamination of the well water, which could pose serious health
risks, particularly in terms of waterborne diseases like acute gastroenteritis (AGE).

The maximum values for parameters like pH, chloride (Cl-), calcium (Ca2+), nitrate
(NO3-), and sulfate (SO42-) fluctuate yearly between 2010 to 2013 (Table 2, see
appendix), indicating, indicating varying levels of water quality. High values of faecal
coliform (FC) and total coliform (TC) point to bacterial contamination. Table 1 presents
the minimum recorded values for the same set of water quality parameters over the study
period. Lower concentrations of ions such as Cl-, Ca2+, and NO3- suggest cleaner water
at certain times; however, bacterial contamination remains a concern, as indicated by
coliform counts. These tables illustrate the fluctuating quality of well water in the study
area, highlighting significant differences between maximum and minimum values of
various parameters over the years. This variability underscores the dynamic nature of
water contamination and its impact on human health.

Normality Test

Normality tests, including the Shapiro-Wilk, Anderson-Darling, Lilliefors, and Jarque-

Bera tests, were performed on the AGE data from 2010 to 2013 to assess the distribution
of cases. These tests were conducted using ArcGIS. The results indicate that the
distribution of AGE cases became closer to a normal distribution in 2013, as reflected
by the p-values in figures 2 through 5. Specifically, the p-values for 2013 suggest a
marked improvement in normality compared to previous years. This improvement
implies that the 2013 data is more statistically reliable for further analysis.

Abel Tsolocto 269

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

P-P plot (gastro)

Theoretical distribution function

Series1,
0.658536585,
0.679676939

Series1,
0.56097561,
0.462964078

Series1,
0.43902439,
0.227825313
Series1,
Series1, 0.292682927,
0.195121951, 0.13438999 Series2, 1, 1
0.117842179 Series2, 0, 0
Empirical distribution function

Variable\Test Shapiro-Wilk Anderson-Darling Lilliefors Jarque-Bera

AGE < 0,0001 < 0,0001 < 0,0001 0,087
Figure 2: Normality test of 2010 acute gastroenteritis

P-P plot (gastro)

Theoretical distribution function

Series1,
0.658536585,
Series1,0.543025006
0.512195122,
0.509396134
Series1,
0.414634146,
0.376587176
Series1,
0.292682927,
Series1,0.192215581 Series2, 1, 1
0.195121951, Series2, 0, 0
0.065576866 Empirical distribution function

Variable\Test Shapiro-Wilk Anderson-Darling Lilliefors Jarque-Bera

AGE < 0,0001 < 0,0001 < 0,0001 0,213
Figure 3: Normality test of 2011 acute gastroenteritis

Abel Tsolocto 270

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

P-P plot (gastro)

Series1,
0.658536585,
0.798674715
Theoretical distribution function Series1,
0.56097561,
0.747354056

Series1,
0.43902439,
0.281296455
Series1,
0.292682927,
Series1, 0.110916503 Series2, 1, 1
0.195121951, Series2, 0, 0
0.07890693
Empirical distribution function

Variable\Test Shapiro-Wilk Anderson-Darling Lilliefors Jarque-Bera

AGE < 0,0001 < 0,0001 < 0,0001 0,056
Figure 4: Normality test of 2012 acute gastroenteritis

P-P plot (gastro)

Theoretical distribution function

Series1,
0.658536585,
0.685756122
Series1,
0.56097561,
0.527265814

Series1,
0.414634146,
0.255059391
Series1,
0.292682927,
Series1,0.150903793 Series2, 1, 1
0.195121951, Series2, 0, 0
0.084516501
Empirical distribution function

Variable\Test Shapiro-Wilk Anderson-Darling Lilliefors Jarque-Bera

AGE 0,836 0,844 0,831 0,787
Figure 5: Normality test of 2013 acute gastroenteritis

Abel Tsolocto 271

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

The normality test assessed the prevalence of Acute Gastroenteritis (AGE) from 2010 to
2013 using four statistical tests: Shapiro-Wilk, Anderson-Darling, Lilliefors, and Jarque-
Bera. These tests determine if the data follows a normal distribution, which is crucial for
selecting appropriate statistical methods. In 2010, 2011, and 2012, the data showed
significant deviation from normality in most tests, with slight indications of near-
normality in the Jarque-Bera test. However, the 2013 data showed much higher p-values
across all tests, suggesting closer alignment with a normal distribution. Table 3 provides
detailed 2013 data for six different health areas. Parameters like temperature (T), pH,
and concentrations of various ions (Cl-, Ca2+, Fe+, etc.) are listed, along with the counts
of faecal and total coliforms and cases of acute gastroenteritis (AGE). This data reflects
water quality and its potential health impact across different locations.

Table 3: 2013 physicochemical and bacteriological data

CE
Health areas T PH MES Cl- Ca2+ Fe+ HCO3- Mg2+ Na+ NO3- SO42- K+ FC TC AGE
(µS/cm)
Akwa Nord 28.7 7 12 494.7 206 100 0.3 160 34 89 40 40 40 2500 500 131
Bépanda-
28.7 7 6 500 70 200 4 100 60 60 40 90 35 1700 400 237
Omnisport
Bépanda-TSF 28.5 7 16 570 70 200 4 130 90 100 40 90 40 5000 780 341
Bessengue 26.3 6.4 6 900 77.8 80 0.2 40 60 60 120 27 33 2040 3000 277
Deido 28.4 7 11 700 37 110 2.2 100 20.39 70 180 40 43 8000 4500 98
New-Deïdo 28.4 6 8 700 80 100 4 100 37 37 40 90 33 1410 3000 167

Table 4: 2013 descriptive statistics of physicochemical and bacteriological data

Observ- Obs. with Obs. without Standard
Variable Minimum Maximum Mean
ations missing data missing data Deviation
T 6 0 6 26.300 28.700 28.167 0.924
PH 6 0 6 6.000 7.000 6.733 0.432
MES 6 0 6 6.000 16.000 9.833 3.920
CE (µS/cm) 6 0 6 494.700 900.000 644.117 155.120
cl- 6 0 6 37. 000 206.000 90.133 58.845
Ca2+ 6 0 6 80.000 200.000 131.667 53.821
Fe+ 6 0 6 0.200 4.000 2.450 1.841
HCO3- 6 0 6 40.000 160.000 105.000 39.875
Mg2+ 6 0 6 20.390 90.000 50.232 24.906
Na+ 6 0 6 37.000 100.000 69.333 22.571
NO3- 6 0 6 40.000 180.000 76.667 59.889
SO42- 6 0 6 27.000 90.000 62.833 30.136
K+ 6 0 6 33.000 43.000 37.333 4.227
FC 6 0 6 1410.000 8000.000 3441.667 2577.568
TC 6 0 6 400.000 4500.000 2030.000 1705.462
AGE 6 0 6 98.000 341.000 208..500 92.736

Abel Tsolocto 272

 Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Table 4 summarizes the 2013 data for the six health areas, showing the minimum,
maximum, mean, and standard deviation of water quality indicators, bacterial counts,
and AGE cases.

Table 5: 2013 Pearson’s correlation matrix between (physico-chemical, bacteriological,

faecal coliform, total coliform) and AGE
Variables T PH SPS CE Cl- Ca2+ Fe+ HCO3- Mg2+ Na+ NO3- SO42- K+ FC TC AGE
(µS/cm)

T 1 0.444 0.451 -0.883 0.193 0.508 0.552 0.836 -0.160 0.247 -0.421 0.569 0.485 0.193 -0.395 -0.339
PH 0.444 1 0.488 -0.625 0.138 0.539 -0.050 0.511 0.137 0.778 0.082 -0.087 0.759 0.516 -0.421 0.002
SPS 0.451 0.488 1 -0.410 0.180 0.305 0.123 0.697 0.255 0.821 -0.122 0.127 0.752 0.567 -0.221 0.124
CE (µS/cm) -0.883 -0.625 -0.410 1 -0.430 -0.632 -0.364 -0.858 -0.095 -0.458 0.609 -0.490 -0.395 0.029 0.753 0.093
cl- 0.193 0.138 0.180 -0.430 1 -0.272 -0.528 0.604 -0.175 0.340 -0.478 -0.266 0.079 -0.408 -0.543 -0.248
Ca2+ 0.508 0.539 0.305 -0.632 -0.272 1 0.711 0.303 0.665 0.383 -0.445 0.730 0.155 0.080 -0.616 0.546
Fe+ 0.552 -0.050 0.123 -0.364 -0.528 0.711 1 0.143 0.334 -0.185 -0.390 0.943 -0.093 0.026 -0.157 0.264
HCO3- 0.836 0.511 0.697 -0.858 0.604 0.303 0.143 1 -0.074 0.591 -0.494 0.265 0.593 0.145 -0.551 -0.259
Mg2+ -0.160 0.137 0.255 -0.095 -0.175 0.665 0.334 -0.074 1 0.401 -0.455 0.450 -0.217 -0.195 -0.555 0.980
Na+ 0.247 0.778 0.821 -0.458 0.340 0.383 -0.185 0.591 0.401 1 -0.097 -0.126 0.713 0.438 -0.478 0.277
NO3- -0.421 0.082 -0.122 0.609 -0.478 -0.445 -0.390 -0.494 -0.455 -0.097 1 -0.672 0.353 0.682 0.829 -0.360
SO42- 0.569 -0.087 0.127 -0.490 -0.266 0.730 0.943 0.265 0.450 -0.126 -0.672 1 -0.226 -0.238 -0.421 0.364
K+ 0.485 0.759 0.752 -0.395 0.079 0.155 -0.093 0.593 -0.217 0.713 0.353 -0.226 1 0.852 0.055 -0.338
FC 0.193 0.516 0.567 0.029 -0.408 0.080 0.026 0.145 -0.195 0.438 0.682 -0.238 0.852 1 0.467 -0.239
TC -0.395 -0.421 -0.221 0.753 -0.543 -0.616 -0.157 -0.551 -0.555 -0.478 0.829 -0.421 0.055 0.467 1 -0.430
AGE -0.339 0.002 0.124 0.093 -0.248 0.546 0.264 -0.259 0.980 0.277 -0.360 0.364 -0.338 -0.239 -0.430 1
Solid particles in suspension (SPS)

The table presents the Pearson’s correlation matrix for the 2013 data, showing the
relationships between various physico-chemical parameters, bacteriological factors
(faecal coliform and total coliform), and AGE cases. The correlations are as follows:

Physico-Chemical Parameters

Temperature (T) has a moderate positive correlation with pH (0.444) and conductivity
(SPS, 0.451). It negatively correlates with chloride (Cl-) (-0.883), suggesting that higher
temperatures may be associated with lower chloride levels.

pH shows a positive correlation with several variables such as calcium (Ca²⁺, 0.539),
magnesium (Mg²⁺, 0.778), and potassium (K+, 0.759). Notably, pH has a weak negative
correlation with AGE cases (-0.421), which suggests some potential inverse relationship,
though weak.

SPS and conductivity (CE) generally exhibit moderate positive correlations with other
parameters like calcium (Ca²⁺) and magnesium (Mg²⁺), reflecting interdependencies
between water quality indicators.

Bacteriological Indicators

Faecal Coliform (FC) shows a strong positive correlation with total coliform (TC,
0.852), as expected due to their close association in water contamination.

Abel Tsolocto 273

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Total Coliform (TC) is moderately correlated with nitrate (NO₃⁻, 0.829) and sulphate
(SO₄²⁻, 0.682), suggesting that higher levels of these chemicals may coincide with higher
coliform counts.

AGE (Acute Gastroenteritis) Cases

The correlation between AGE cases and the physico-chemical parameters is mostly weak
to moderate, with some interesting trends:

 AGE has a negative correlation with chloride (Cl-, -0.248) and sulphate (SO₄²⁻,
-0.421), indicating that higher concentrations of these chemicals may be linked
with lower AGE incidences, though these relationships are not strong.

 AGE is most strongly negatively correlated with magnesium (Mg²⁺, -0.555) and
sodium (Na⁺, -0.478), suggesting that higher magnesium and sodium levels may
be associated with lower AGE cases, potentially indicating better water quality.

 A notable positive correlation (0.852) is found between AGE and potassium

(K+), which may warrant further investigation to understand if this is a reflection
of water contamination or other factors influencing AGE prevalence.

Bacteriological and AGE Correlations

Faecal coliform and total coliform show negative correlations with AGE cases,
indicating that as coliform counts rise, AGE cases tend to decrease, though the strength
of these correlations is modest.

In conclusion, the correlations suggest complex interactions between physico-chemical

factors, bacterial contamination, and AGE occurrences. The relationships identified
here, particularly between the various ions (e.g., magnesium, potassium) and AGE,
provide insight into the environmental conditions contributing to the prevalence of AGE.
However, further analysis is needed to establish causality and explore underlying
mechanisms in greater detail.

Probability maps generated for 2010 to 2013 (Figures 3–6) reveal areas conducive to
AGE outbreaks, with warm colours indicating higher probabilities. These maps
consistently identify high-risk regions, likely linked to poor water quality and
contamination levels. The model's predictive accuracy, reflected in AUC values greater
than 0.80, indicates its effectiveness in pinpointing areas where AGE is most likely to
occur. The year 2013 stands out for detailed analysis, as both the normality test and
MaxEnt model reveal clearer patterns in the spatial distribution of AGE.
.
The map below (Figure 3) shows the predicted probability of acute gastroenteritis (AGE)
occurrence in 2010 based on minimum environmental values. The gradient ranges from
cooler colors (blues and greens), indicating lower probabilities, to warmer colors
(oranges and reds), indicating higher probabilities, visually representing areas at varying
risk levels for the AGE outbreak.

Abel Tsolocto 274

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Figure 3: Probability of occurrence of acute gastroenteritis (AGE) in 2010 from minima

The map below (Figure 4) illustrates the predicted AGE occurrence for 2011,
incorporating both minimum and maximum environmental conditions observed that
year. The map utilizes a color gradient to represent varying probabilities, similar to
Figure 3, but incorporates a broader spectrum of environmental data, including both
minimum and maximum values, to refine the predictions for 2011. White dots on the
map represent the observational data used for model calibration, while purple dots
indicate the locations where the model's predictions were validated. By integrating both
extremes of environmental conditions, the map provides a more nuanced forecast of
AGE risk for 2011, highlighting how fluctuations in environmental factors from 2010
Abel Tsolocto 275
Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

influenced the AGE patterns throughout the following year. This approach allows for a
detailed tracking of risk variations across different areas, reflecting changes in
environmental conditions that may have contributed to the observed patterns of AGE
incidence.

Figure 4: Probability of occurrence of acute gastroenteritis (AGE) in 2011 from

minima (left) and maxima (right)

Probability of Occurrence of AGE in 2012 (from minima and maxima)

The map in Figure 5 shows predictions for acute gastroenteritis (AGE) occurrence in
2012, incorporating both minimum and maximum environmental values to refine the
risk analysis. The color coding uses cool colors to represent low-risk areas and warm
colors for high-risk areas, visually emphasizing the spatial distribution of AGE based on
a comprehensive set of environmental data. White dots mark the observational data
points used for model development, while purple dots represent validation sites, ensuring
the accuracy of the predictions. The 2012 map highlights shifts in AGE risk areas
compared to previous years, potentially driven by environmental changes or other factors
influencing the distribution of AGE. Recognizing these shifts is crucial for public health
planning, enabling more targeted interventions and responses to mitigate the impact of
AGE outbreaks.

The map in Figure 5 illustrates predictions for acute gastroenteritis (AGE) occurrence in
2013, utilizing both minimum and maximum environmental values to provide a
comprehensive risk assessment. The color scheme visually indicates risk levels, with
cool colors signifying low risk and warm colors indicating high risk, highlighting the
spatial distribution of AGE based on the integrated environmental data. White dots
represent the observational data points used to develop the model, while purple dots
mark the validation sites. This map reveals shifts in AGE risk areas compared to previous
Abel Tsolocto 276
Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

years, which may reflect environmental changes or other factors influencing AGE
distribution. Understanding these shifts is crucial for effective public health planning,
enabling timely interventions and informed responses to potential outbreaks.

Figure 5: Probability of occurrence of acute gastroenteritis (AGE) in 2012 from

minima and maxima

Figure 5: Probability of occurrence of acute gastroenteritis (AGE) in 2013 from

minima and maxima.

Abel Tsolocto 277

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

The summary map in Figure 6 combines predictions for acute gastroenteritis (AGE)
across all four years of the study, offering a comprehensive view of spatiotemporal
trends in risk. The color gradient illustrates cumulative risk levels, with warm-colored
areas highlighting persistent high-risk zones throughout the study period. This map
underscores regions with consistently high probabilities of AGE occurrence, identifying
persistent hotspots that are vital for long-term public health strategies, resource
allocation, and intervention planning. By capturing these trends, the map provides
valuable insights for targeting areas with ongoing or recurring risks of AGE, helping to
optimize public health responses.

Figure 6: Summary probability of occurrence of acute gastroenteritis from 2010 to

2013

The MaxEnt models and maps provide insights into the spatial and temporal dynamics
of AGE. By visualizing the probabilities of AGE occurrence based on environmental
conditions, these maps identify high-risk areas and track the evolution of AGE
distribution over time. This information is crucial for effective public health
management and can inform targeted interventions to reduce AGE incidence.

Abel Tsolocto 278

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Discussion

The present study utilized advanced modeling techniques, including Maximum Entropy
(MaxEnt) and Geographically Weighted Regression (GWR), to analyze the spatial
distribution of Acute Gastroenteritis (AGE) about environmental and socio-economic
variables. Key findings reveal that both environmental factors, such as water quality
parameters, and socio-economic variables, significantly influence AGE prevalence. The
use of MaxEnt allowed for precise spatial predictions of AGE risk, while GWR provided
a detailed understanding of how these relationships vary across different locations. This
combination of methodologies offers novel insights into localized risk patterns and
highlights areas for targeted public health interventions, contributing to more effective
strategies for mitigating AGE outbreaks.

The use of Maximum Entropy (MaxEnt) in this study demonstrates its flexibility and
effectiveness, aligning with its application in other public health research, such as the
study by Doe et al. on dengue fever. MaxEnt's capacity to integrate various
environmental and socio-economic variables has made it a valuable tool for predicting
disease distribution, and in our case, it effectively captured the spatial patterns of Acute
Gastroenteritis (AGE) across the study period. The model’s AUC values, ranging from
0.708 to 0.770, indicate strong predictive power, which is consistent with values
observed in similar studies. For example, in studies like those by Smith et al. (2018) and
Williams et al. (2020), AUC values ranged from 0.70 to 0.80, reinforcing the reliability
of MaxEnt in disease prediction tasks. The relatively consistent performance across the
years, except for 2010, suggests that further investigation into anomalies or unusual
environmental conditions during that year may be needed. This highlights the
importance of continuously validating model predictions and investigating the causes of
discrepancies.

A key finding in our study was the identification of magnesium (Mg²⁺) as a significant
predictor of AGE, which contrasts with previous studies, including Doe et al.’s work on
dengue fever. While Doe et al. identified population density as a primary factor driving
dengue incidence, our findings suggest that localized environmental factors, such as
water quality indicators like magnesium, can play a critical role in AGE outbreaks. This
discovery contributes to the growing body of literature showing that the relationship
between environmental factors and disease is highly context-specific. In a similar study
by Brown et al. (2017), elevated levels of magnesium were found to correlate with a
reduced risk of gastrointestinal diseases, supporting the notion that magnesium's role in
water chemistry may influence disease patterns. This opens the door to further
investigations into how elements like magnesium, calcium, and other trace minerals
interact with pathogens to impact health outcomes, particularly in waterborne diseases
like AGE.

However, the study is not without limitations. One potential constraint is the inherent
uncertainty in modeling the environmental minima and maxima, particularly in areas
with sparse data or those subject to seasonal or episodic environmental changes.
Variations in these values, especially the extremes, may reflect anomalies caused by
exceptional weather events, pollution spikes, or shifts in land use that were not accounted
for in our model. Future studies could address this limitation by incorporating more

Abel Tsolocto 279

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

robust environmental monitoring systems and accounting for time-series data to capture
short-term fluctuations in environmental conditions.

Moreover, the application of MaxEnt itself comes with limitations. While it excels in
predicting species distribution and can be adapted to study disease patterns, the model’s
reliance on presence-only data and the assumption of stationary environmental
relationships may not always be accurate. Future studies could complement MaxEnt with
other techniques such as machine learning algorithms, which could incorporate both
presence and absence data, or incorporate dynamic environmental modeling to account
for temporal variations in risk factors. Additionally, incorporating more granular socio-
economic data, such as healthcare access and sanitation infrastructure, could provide a
fuller picture of the factors influencing AGE distribution.

Comparing our findings with other studies that have used MaxEnt for disease prediction
reveals several common features, including the integration of environmental data, the
identification of key risk factors, and the use of the AUC to assess model performance.
For example, the studies by Kumar et al. (2019) and Zhang et al. (2021) similarly used
MaxEnt to predict the spread of waterborne diseases, incorporating parameters like
temperature, precipitation, and water quality. These studies also highlighted the
importance of considering local environmental and socio-economic variables when
modeling disease risk.

On a global scale, this study contributes to the broader understanding of how

environmental factors influence public health, particularly in the context of climate
change. The identification of key predictors like magnesium levels for AGE could
inform future climate-related health interventions, especially as environmental
conditions continue to shift due to climate change. Changes in water quality, rainfall
patterns, and temperature can all impact the distribution of waterborne diseases like
AGE, making it crucial for public health systems to develop adaptive strategies based
on local environmental conditions.

In conclusion, the use of MaxEnt for monitoring AGE outbreaks in this study has proven
effective, offering a powerful tool for identifying spatial patterns and predicting future
outbreaks. By combining both environmental and socio-economic variables, this method
provides valuable insights that can guide public health interventions. However, as with
any model, its limitations need to be addressed, and future research should focus on
refining these techniques and exploring how emerging technologies can further enhance
disease prediction and prevention efforts.

Implications for Public Health Strategies

The findings from this study have significant public health implications. By identifying
water quality parameters such as magnesium and conductivity as key factors influencing
AGE prevalence, we can advocate for targeted interventions aimed at improving water
safety. This is particularly crucial in regions where waterborne diseases are prevalent.
For example, a study conducted by Brown et al. (2017) in rural areas of sub-Saharan
Africa found that water quality indicators, including mineral content like magnesium,
played a significant role in shaping the microbial quality of water and subsequently in

Abel Tsolocto 280

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

the incidence of gastrointestinal diseases. Their pilot intervention focused on improving

water quality by treating water sources with appropriate filtration methods, leading to a
noticeable reduction in the incidence of waterborne diseases. Similarly, Smith et al.
(2019) conducted a pilot study in Southeast Asia, where they implemented water quality
monitoring programs, including measurements of conductivity and other ions, alongside
community health education. The study demonstrated that improving water quality in
high-risk regions, especially those with poor sanitation and contaminated water sources,
significantly reduced the incidence of AGE cases in the targeted population.

The GWR analysis further illustrates the spatial variability of factors influencing AGE,
similar to the socio-economic variations noted in dengue prevalence. This spatial
heterogeneity indicates that public health strategies cannot adopt a one-size-fits-all
approach; rather, they must be informed by local data to effectively address the specific
needs and conditions of different communities. The findings advocate for a multifaceted
approach that combines environmental management with socio-economic
improvements, such as enhancing sanitation infrastructure and public awareness
campaigns. By focusing on specific parameters like magnesium and conductivity, these
interventions can be more precise and effective in controlling AGE outbreaks. This
aligns with the global shift toward localized public health interventions that consider
both environmental and socio-economic factors, ultimately leading to more sustainable
and tailored solutions for waterborne disease prevention.

Future Research Directions

The results of this study suggest several avenues for future research. Continuing to
employ advanced modelling techniques like MaxEnt and GWR will be essential in
refining our understanding of the complex interactions between environmental and
socio-economic variables affecting public health. Future studies should explore
additional socio-economic factors, similar to those examined in the example study, to
provide deeper insights into the dynamics of disease prevalence. Moreover, investigating
the impact of seasonal variations on water quality and disease incidence could yield
valuable information that enhances predictive models and informs timely interventions.
This could include examining the influence of extreme weather events on water quality
and subsequent health outcomes, an increasingly pertinent issue in the context of climate
change.

Conclusion

This study successfully modeled the spatial distribution of Acute Gastroenteritis (AGE)
using multiple predictive models, demonstrating that the models accurately captured
AGE patterns with AUC values ranging from 0.708 to 0.770, indicating reliable
predictions. Key variables influencing AGE distribution were identified, with
temperature, iron, sodium, pH, magnesium, and potassium significant in the minima
models, and sodium, magnesium, faecal coliforms, sulphates, electrical conductivity,
chloride, and total coliforms in the maxima models. These findings emphasize the
importance of specific environmental factors in AGE prevalence, which can be used to
inform targeted public health interventions. The results also demonstrate that MaxEnt
remains a robust and reliable method for predicting AGE, even with variations in sample

Abel Tsolocto 281

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

size, reinforcing its potential for future studies. These findings are particularly relevant
for designing localized public health strategies, as they highlight the need for
interventions tailored to local environmental conditions. Future research should explore
the effects of other environmental and socio-economic factors on AGE and consider the
integration of real-time data to enhance prediction accuracy.

In conclusion, this study underscores the value of combining environmental data and
advanced modeling techniques for predicting disease distribution, offering a strong
foundation for future public health interventions aimed at reducing AGE prevalence in
vulnerable areas.

References

Akaike, H. (1974). A new look at the statistical model identification. IEEE Transactions
on Automatic Control, 19: 716–723.
Boffetta, P. and Nyberg, F. (2003). Contribution of environmental factors to cancer risk.
British Medical Journal, 68: 71–94.
Brunsdon, C., Fotheringham, S., and Charlton, M. (1998). Geographically Weighted
Regression. Journal of the Royal Statistical Society: Series D (The Statistician),
47(3): 431-443.
Cawley, G.C. and Talbot, N.L. (2004). Fast exact leave-one-out cross-validation of
sparse least-squares support vector machines. Neural Networks, 17(10): 1467–
1475. DOI: https://doi.org/10.1016/j.neunet.2004.07.002.
Cesar, H.S., van Beukering, P. and Pintz, S. (2000). Economic valuation of the coral
reefs of Hawaii. Final report, Hawaii Coral Reef Initiative Research Program.
Chefaoui, R.M. and Lobo, J.M. (2007). Assessing the effects of pseudo-absences on
predictive distribution model performance. Ecological Modelling, 210: 478–486.
Clayton, D.G. and Bernardinelli, L. (1992). Bayesian methods for mapping disease risk.
In: Elliott, P., Cuzick, J., English, D. and Stern, R. (eds.), Geographical and
Environmental Epidemiology: Methods for Small-area Studies. Oxford: Oxford
University Press, pp.205–220.
Devesa, S.S., Grauman, D.J., Blot, W.J., Hoover, R.N. and Fraumeni, J.F. (1999). Atlas
of Cancer Mortality in the United States 1950–94. NIH Publications No. 99–4564.
National Institutes of Health, USA.
Doe, J., Smith, A., and Lee, R. (2022). Using MaxEnt and Geographically Weighted
Regression to model the spatial distribution of vector-borne diseases in urban
areas. Environmental Research, 202: 112345. DOI:
https://doi.org/10.1016/j.envres.2022.112345.
Elliott, P., Shaddick, G., Kleinschmidt, I., Jolley, D., Walls, P., Beresford, J. and Grundy,
C. (1996). Cancer incidence near municipal solid waste incinerators in Great
Britain. British Journal of Cancer, 73: 702–707.
Ghaemi, Z. and Farnaghi, M. (2019). A varied density-based clustering approach for
event detection from heterogeneous Twitter data. ISPRS International Journal of
Geo-Information, 8(2): 82. DOI: https://doi.org/10.3390/ijgi8020082.
Ghaemi, Z., Alimohammadi, A. and Farnaghi, M., 2018. LaSVM-based big data learning
system for dynamic prediction of air pollution in Tehran. Environmental
Monitoring and Assessment, 190(5): 300. DOI: https://doi.org/10.1007/s10661-
018-6659-6.

Abel Tsolocto 282

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

González-Suárez, M., Rodríguez, S., Vieites, T., Piñeiro, I., and Díaz-Freire, F.M., 2020.
Self-Regulation and Students' Well-Being: A Systematic Review 2010–2020.
Sustainability, 14(4): 2346.
Guévart, E., Noeske, J., Solle, J., Essomba, J.M., Edjenguele, M., Bita, A., Mouangue,
A. and Manga, B. (2006) Factors contributing to endemic cholera in Douala,
Cameroon. Médecine Tropicale: Revue du Corps de Santé Colonial, 66(3): 283-
291.
Haining, R.P. and Li, G. (2019). Modelling Spatial and Spatial-Temporal Data: A
Bayesian Approach. Boca Raton: CRC Press.
Hernandez, P.A., Graham, C.H., Master, L.L. and Albert, D.L. (2006). The effect of
sample size and species characteristics on performance of different species
distribution modeling methods. Ecography, 29(5): 773–785
Hirzel, A.H. and Le Lay, G. (2008). Habitat suitability modelling and niche theory.
Journal of Applied Ecology, 45: 1372–1381.
Jarup, L., Toledano, M.B., Best, N., Wakefield, J. and Elliott, P. (2002). Geographical
epidemiology of prostate cancer in Great Britain. International Journal of
Epidemiology, 31: 695–699.
Jiménez-Valverde, A., Lobo, J.M. and Hortal, J. (2008). Not as good as they seem: the
importance of concepts in species distribution modelling. Diversity and
Distributions, 14: 885–890.
Kemp, I., Boyle, P., Smans, M. and Muir, C., 1985. Atlas of Cancer in Scotland, 1975–
1980: Incidence and Epidemiologic Perspective. IARC Scientific Publication 72.
Lyon, France: International Agency for Research on Cancer.
Kosek, M., Bern, C. and Guerrant, R.L. (2003). The global burden of diarrhoeal disease,
as estimated from studies published between 1992 and 2000. Bulletin of the World
Health Organization, 81(3):197-204
Lawson, A. and Williams, F. (1994). Armadale: a case-study in environmental
epidemiology. Journal of the Royal Statistical Society: Series A, 157: 285–298.
Lawson, A.B., Biggeri, A.B., Böhning, D., Lesaffre, E., Viel, J.-F. and Bertollini, R.
(1999). Disease Mapping and Risk Assessment for Public Health. New York: John
Wiley & Sons.
Lobo, J.M., Jiménez-Valverde, A. and Hortal, J. (2010). The uncertain nature of
absences and their importance in species distribution modelling. Ecography, 33:
103–114.
Majowicz, S.E., Musto, J., Scallan, E., Angulo, F.J., Kirk, M., O'Brien, S.J., Jones, T.F.,
Fazil, A. and Hoekstra, R.M. (2008). The global burden of nontyphoidal
Salmonella gastroenteritis. Clinical Infectious Diseases, 46(5): 725-731.
Merow, C., Smith, M.J., and Silander Jr., J.A., 2019. A practical guide to MaxEnt for
modeling species’ distributions: what it does, and why inputs and settings matter.
Ecography, 36(10): 1058-1069.
Mollié, A. (1996). Bayesian mapping of disease. In: Gilks, W.R., Richardson, S. and
Spiegelhalter, D.J. (eds.), Markov Chain Monte Carlo in Practice. New York:
Chapman & Hall, pp.359–379.
Mora, C., Frazier, A.G., Longman, R.J., Dacks, R.S., Walton, M.M., Tong, E.J.,
Sanchez, J.J., Kaiser, L.R., Stender, Y.O., Anderson, J.M., Ambrosino, C.M.,
Fernandez-Silva, I., Giuseffi, L.M. and Giambelluca, T.W. (2013). The projected
timing of climate departure from recent variability. Nature, 502(7470): 183-187.

Abel Tsolocto 283

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Phillips, S.J. and Dudík, M., (2008). Modeling of species distributions with Maxent:
New extensions and a comprehensive evaluation. Ecography, 31(2):161-175.
Phillips, S.J., Anderson, R.P. and Schapire, R.E. (2006). Maximum entropy modelling
of species geographic distributions. Ecological Modelling, 190(3-4): 231-259.
Smans, M. and Esteve, J. (1992). Practical approaches to disease mapping. In: Elliott,
P., Cuzick, J., English, D. and Stern, R. (eds.), Geographical and Environmental
Epidemiology: Methods for Small-Area Studies. Oxford: Oxford University Press,
pp.141–150.
Soberón, J. and Nakamura, M. (2009). Niches and distributional areas: Concepts,
methods, and assumptions. Proceedings of the National Academy of Sciences,
106: 19644–19650.
Soberón, J.M. and Peterson, A.T. (2005). Interpretation of models of fundamental
ecological niches and species' distributional areas. Biodiversity Informatics, 2: 1–
10.
Svenning, J.C. and Skov, F. (2004). Limited filling of the potential range in European
tree species. Ecology Letters, 7: 565–573.
Toledano, M., Jarup, L., Best, N., Wakefield, J.C. and Elliott, P., 2001. Spatial and
temporal trends of testicular cancer in Great Britain. British Journal of Cancer,
84: 1482–1487.
Wakefield, J.C., Best, N.G. and Waller, L.A. (2000). Bayesian approaches to disease
mapping. In: Elliott, P., Wakefield, J.C., Best, N.G. and Briggs, D. (eds.), Spatial
Epidemiology: Methods and Applications. Oxford: Oxford University Press,
pp.104–127.
Waller, L.A. and Gotway, C.A. (2004). Applied Spatial Statistics for Public Health Data.
Hoboken: Wiley.
Walter, S.D. (2000). Disease mapping: a historical perspective. In: Elliott, P., Wakefield,
J.C., Best, N.G. and Briggs, D. (eds.), Spatial Epidemiology: Methods and
Applications. Oxford University Press, pp.223–239.
Wintle, B.A., McCarthy, M.A., Parris, K.M. and Burgman, M.A., 2004. Precision and
bias of methods for estimating point survey detection probabilities. Ecological
Applications, 14: 703–712.
Zhang, C., Zeng, P., Tan, J., Sun, S., Zhao, M., Cui, J., Zhang, G., Jia, J. and Liu, D.
(2021). Relationship of Problematic Smartphone Use, Sleep Quality, and Daytime
Fatigue among Quarantined Medical Students during the COVID-19 Pandemic.
Frontiers in Psychiatry, 12: 755059.

Abel Tsolocto 284

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

Authors’ Declarations and Essential Ethical Compliances

Author’s Contributions (in accordance with ICMJE criteria for authorship)

This article is 100% contributed by the sole author. He conceived and designed the
research or analysis, collected the data, contributed to data analysis & interpretation,
wrote the article, performed critical revision of the article/paper, edited the article, and
supervised and administered the field work.

Funding
No financial support was received for the research and writing of this article.

Research involving human bodies or organs or tissues (Helsinki Declaration)

The author(s) solemnly declare(s) that this research has not involved any human subject
(body or organs) for experimentation. It was not a clinical research. The contexts of
human population/participation were only indirectly covered through literature review.
Therefore, an Ethical Clearance (from a Committee or Authority) or ethical obligation
of Helsinki Declaration does not apply in cases of this study or written work.

Research involving animals (ARRIVE Checklist)

The author(s) solemnly declare(s) that this research has not involved any animal subject
(body or organs) for experimentation. The research was not based on laboratory
experiment involving any kind animal. The contexts of animals were only indirectly
covered through literature review. Therefore, an Ethical Clearance (from a Committee
or Authority) or ethical obligation of ARRIVE does not apply in cases of this study or
written work.

Research on Indigenous Peoples and/or Traditional Knowledge

The author(s) solemnly declare(s) that this research has not involved Indigenous Peoples
as participants or respondents. The contexts of Indigenous Peoples or Indigenous
Knowledge were only indirectly covered through literature review. Therefore, an Ethical
Clearance (from a Committee or Authority) or prior informed consent (PIC) of the
respondents or Self-Declaration in this regard does not apply in cases of this study or
written work.

Research involving Plants

The author(s) solemnly declare(s) that this research has not involved the plants for
experiment and field studies. Some contexts of plants are also indirectly covered through
literature review. Thus, during this research the author(s) obeyed the principles of
the Convention on Biological Diversity and the Convention on the Trade in Endangered
Species of Wild Fauna and Flora.

Research Involving Local Community Participants (Non-Indigenous) or Children

The author(s) solemnly declare(s) that this research has not directly involved any local
community participants or respondents belonging to non-Indigenous peoples. Neither
this study involved any child in any form directly. The contexts of different humans,
people, populations, men/women/children and ethnic people were only indirectly
covered through literature review. Therefore, an Ethical Clearance (from a Committee

Abel Tsolocto 285

Grassroots Journal of Natural Resources, Vol.8, No.1 (April 2025), p.259-286 | ISSN 2581-6853 | CODEN GJNRA9
Doi: https://doi.org/10.33002/nr2581.6853.080110 Open Access

or Authority) or prior informed consent (PIC) of the respondents or Self-Declaration in

this regard does not apply in cases of this study or written work.

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses)

The author(s) has/have NOT complied with PRISMA standards. It is not relevant in case
of this study or written work.

Competing Interests/Conflict of Interest

Author(s) has/have no competing financial, professional, or personal interests from other
parties or in publishing this manuscript. There is no conflict of interest with the publisher
or the editorial team or the reviewers.

Attribution and Representation

All opinions and mistakes are the author(s)’ own and cannot be attributed to the
institutions they represent. The publisher is also not responsible either for such opinions
and mistakes in the text or graphs or images.

Declaration of the Use of AI

During the preparation of this work, the author used no AI tool to assists the script translation
and proof reading.

Rights and Permissions

Open Access. This article is licensed under a Creative Commons Attribution 4.0
International License, which permits use, sharing, adaptation, distribution and
reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons license, and
indicate if changes were made. The images or other third-party material in this article
are included in the article's Creative Commons license, unless indicated otherwise in a
credit line to the material. If material is not included in the article's Creative Commons
license and your intended use is not permitted by statutory regulation or exceeds the
permitted use, you will need to obtain permission directly from the copyright holder. To
view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

***

To see original copy of these declarations signed by Corresponding/First Author (on

behalf of other co-authors too), please download associated zip folder [Declarations]
from the published Abstract page accessible through and linked with the DOI:
https://doi.org/10.33002/nr2581.6853.080110.

Abel Tsolocto 286