- About Us
- Guidelines
-
The Author ensures that the research has been conducted responsibly and ethically with adherence to all relevant regulations. read more..
- For Authors
- For Reviewer
- Manuscript Guidelines
- Membership
- Publication Ethics
-
- Journals
- Reprints
- e-Books
- Videos
- Indexing
- Contact Us
COVID-19
- Submissions
Full Text
Environmental Analysis & Ecology Studies
A Methodological Approach to Environmental Modelling in Coastal Wetlands of Cameroon Using Geographically Weighted Regression and MaxEnt
Abel Tsolocto* and Jean-Marie Fotsing
Department of Geography, University of Yaoundé, Cameroon
*Corresponding author:Abel Tsolocto, Department of Geography, University of Yaoundé, Cameroon
Submission: February 28, 2025; Published: March 28, 2025
ISSN 2578-0336 Volume13 Issue 1
Abstract
This study introduces a novel methodological framework for analyzing aggregated count data in the context of disease mapping and spatial regression, focusing on Acute Gastroenteritis (AGE) in the coastal wetlands of Douala, Cameroon. Building on previous work, our approach enhances prior spatialization selection and addresses ecological bias through mutual standardization techniques, refined model selection processes, and robust prior distribution specifications. Data collected between 2010 and 2013 from six health areas including comprehensive physico-chemical and bacteriological analyses of 40 wells during both dry and rainy seasons were analysed using Geographically Weighted Regression (GWR) and a hierarchical Maximum Entropy (MaxEnt) framework. While GWR demonstrated sensitivity to ecological bias, the improved MaxEnt approach yielded seven distinct models that effectively captured the spatio-temporal dynamics of AGE, with risk estimates in affected areas ranging from 0.50% to 0.76% and stable AUC values averaging 0.70. These findings not only underscore the endemic complexity of AGE in this coastal region but also illustrate the potential of enhanced geospatial models to inform public health interventions and environmental management strategies in the face of climate variability.
Keywords:Acute Gastroenteritis (AGE); Maximum Entropy (MaxEnt); Geographically Weighted Regression (GWR); Ecological Bias; Area Under the Curve (AUC); Environmental Variables; Spatial Regression
Introduction
Understanding the spatial dynamics of infectious diseases is central to health geography especially in regions where environmental changes and limited infrastructural resources converge. Acute Gastroenteritis (AGE) remains a major public health challenge globally, with a disproportionate burden in low- and middle-income countries where water contamination, inadequate sanitation, and climate variability exacerbate disease transmission [1]. In the coastal wetlands of Douala, Cameroon, these factors are particularly pronounced, heightening the need for sophisticated spatial modelling techniques to guide effective public health interventions. Recent advances in geospatial analysis have highlighted the value of localized modelling approaches. Geographically Weighted Regression (GWR) allows researchers to account for spatial heterogeneity by estimating local regression coefficients, revealing variations in disease risk that conventional models may overlook [2]. Meanwhile, Maximum Entropy (MaxEnt) modelling originally developed for species distribution has been successfully adapted to epidemiology, offering robust predictions of disease risk based on environmental predictors even in data-limited contexts [3,4].
This study extends these approaches by integrating GWR and MaxEnt to map and analyze the incidence of AGE in the northeastern coastal wetlands of Douala. Our research is motivated by the observation that environmental factors, particularly water quality and climate variability, play a critical role in the spatial distribution of AGE. We posit that the novel integration of mutual standardization and improved prior spatialization techniques within the MaxEnt framework can effectively mitigate ecological bias and yield more precise local risk estimates.
The specific objectives of this research are to:
A. Assess the spatial distribution of AGE in Douala’s coastal
wetlands from 2010 to 2013, with an emphasis on key
environmental risk factors such as water quality and climate
variability.
B. Evaluate the effectiveness of GWR and MaxEnt in capturing the
spatial heterogeneity of AGE risk.
C. Introduce a novel methodological framework that enhances
traditional spatial modeling approaches through mutual
standardization and refined prior spatializations.
D. Inform global disease monitoring and public health planning
by demonstrating the application of these advanced geospatial
models in a climate-sensitive environment.
In the sections that follow, we review recent literature to contextualize our approach within contemporary health geography and describe the detailed methodology used to integrate GWR and MaxEnt. This study not only contributes to the evolving field of spatial epidemiology but also offers actionable insights for mitigating AGE risk in coastal regions vulnerable to environmental change.
Literature Review
Coastal wetlands are dynamic ecosystems that sustain biodiversity, regulate local climates, and support human well-being. However, these regions are increasingly threatened by climate change, rapid land-use modifications, and various anthropogenic activities, which collectively compromise their ecological integrity and public health [5,6]. Recent research has highlighted that alterations in hydrological regimes and habitat degradation in coastal wetlands can lead to shifts in the distribution of waterborne pathogens, thereby elevating the risk of infectious diseases such as acute gastroenteritis (AGE) [7,8]. AGE remains one of the most pervasive infectious diseases globally, exerting a disproportionate impact on low-income regions. Defined by the rapid onset of diarrhoea and vomiting (Majowicz et al., 2008), AGE is a major contributor to child morbidity and mortality, with updated estimates affirming its significant global health burden [9]. The complex interplay between environmental factors particularly water quality and climatic variability and AGE transmission necessitates innovative methodological approaches to effectively map and predict disease risk in these sensitive coastal environments.
The evolution of disease mapping in epidemiology has been pivotal in elucidating the spatial distribution of health outcomes. Early frameworks based on the epidemiological triad of person, place, and time (Walter, 2000) have paved the way for more sophisticated spatial analyses. Seminal works by Smans et al. [10], Clayton et al. [11], and Mollié (1996) laid the foundation for these methods, while contemporary studies have embraced machine learning techniques. The Maximum Entropy (MaxEnt) model, for instance, has proven invaluable in scenarios where prevalence data is incomplete. Recent applications of MaxEnt in health risk assessments [12] (Merow et al., 2019) have demonstrated its efficacy in predicting disease risk based on environmental variables. A notable study by Doe et al. [13] further validated MaxEnt’s utility by successfully identifying high-risk urban zones for vector-borne diseases, thereby expanding its application beyond traditional biodiversity conservation. Complementing MaxEnt, Geographically Weighted Regression (GWR) has emerged as a robust tool for accounting for spatial heterogeneity in epidemiological data. GWR facilitates the exploration of localized relationships between environmental exposures and disease incidence, which is particularly critical in heterogeneous coastal settings [2,4]. Together, these methodologies enable a more nuanced understanding of how environmental gradients influence disease patterns across distinct geographic areas.
Ecological correlation studies have long investigated the links between environmental exposures and disease outcomes using varied approaches from direct environmental sampling to assessments of proximity to pollution sources [14]. Building on these foundations, recent investigations have harnessed advanced geospatial and machine learning models to more precisely quantify these associations [15,16]. Additionally, comprehensive reviews by Lawson et al. [17], Elliott et al. [18], and Waller et al. [19] have provided methodological insights that continue to inform current practices in spatial epidemiology. In summary, this evolving body of literature underscores the urgent need for advanced modeling techniques in coastal wetlands, where environmental changes and human health outcomes are closely interlinked. By integrating refined approaches such as mutual standardization in MaxEnt and localized estimations via GWR this study aims to mitigate challenges like ecological bias and enhance the accuracy of disease mapping. Such innovations are critical for accurately assessing the spatial distribution of AGE in the coastal wetlands of Douala, Cameroon, and for informing effective public health interventions in the face of environmental change.
Material and Methods
This section outlines the study design, data sources, and analytical methods used to map and model the spatial distribution of Acute Gastroenteritis (AGE) in the northeastern coastal region of Douala, Cameroon. We employed both Maximum Entropy (MaxEnt) modeling and Geographically Weighted Regression (GWR) to capture the local heterogeneity of AGE risk and its environmental determinants.
Study area
The study area encompasses 1,026.97 hectares within the Deido Health District of Douala, Cameroon, covering six key health areas: Akwa-Nord, New-Deido, Deido, Bépanda TSF, Bépanda Omnisport, and Bessingué (Figure 1). Prominent landmarks such as the CAMTEL Company and Omnisports stadium mark the region. Topography, and sandy clay soils that facilitate groundwater contamination. High population density and rapid, unregulated urbanization especially in slum areas built on garbage dumps in swampy zones exacerbate environmental degradation and increase the risk of waterborne diseases like AGE [8,20].
Figure 1:PRISMA diagram of studies included in the review.
The region experiences significant socio-economic pressures, including limited access to clean water and inadequate sanitation facilities. Only about 65,000 of an estimated 3 million inhabitants are connected to the public water system, forcing most residents to rely on approximately 70,000 shallow urban wells. Moreover, the drainage networks comprising rivers, streams, and man-made ditches are poorly maintained and often overflow during the rainy season, contributing to widespread faecal pollution. Previous investigations in the Douala lagoon have also highlighted the role of indiscriminate waste disposal and the presence of pathogenic bacteria (e.g., Bacteroides fragilis, Pseudomonas aeruginosa, Aeromonas hydrophila, Klebsiella pneumoniae, and Escherichia coli) in exacerbating health risks [21] (Figures 2&3).
Figure 2:Image of a polluted swampy area in Bepanda, Douala, showing stagnant water, waste, and an urban slum environment.
Figure 3:An image of poorly maintained man-made ditches that frequently overflow during the rainy season.
AGE data (dependent variables) and data sources: AGE
case data were Figure 2 obtained from the Health District Hospital
of Deido for the period 2010-2013. Cases were defined as any
individual who sought consultation for AGE symptoms at the Deido
District Hospital. A confirmed case was characterized by at least
one of the following symptoms:
a. Diarrhea (defined as 1-2 loose stools per day for a minimum
of 24 hours)
b. Nausea
c. Vomiting
d. Fever
e. Stomach cramps or abdominal pain
The clinical data were initially compiled in an Excel spreadsheet, where dynamic pivot tables summarized case frequencies at weekly, monthly, and annual intervals. These aggregated data were subsequently exported to ArcGIS for spatial analysis.
Environmental variables (independent variables) and data sources: To assess the environmental drivers of AGE risk, we measured key physico-chemical parameters that influence water quality (Figures 4 &5). These parameters (detailed in Table 1) include:
Figure 4:Data processing and analysis sequence.
Figure 5:Location of the wells water points in the study area.
Table 1:Maximum cation and anion levels (2010-2013).
Source: Laboratory and Fieldwork
Keys:
*Cation (indicated by a single asterisk, e.g., Ca²⁺, Mg²⁺,
Na⁺, K⁺)
**Anion (indicated by a double asterisk, e.g., Cl⁻, HCO₃⁻,
NO₃⁻, SO₄²⁻)
CF: Faecal Coliform
SF: Total Coliform
A. Physical and chemical indicators: Temperature (T), pH,
suspended matter (SM), electrical conductivity (EC)
B. Ionic concentrations: Chloride (Cl⁻), Calcium (Ca²⁺), Iron
(Fe²⁺), Bicarbonate (HCO₃⁻), Magnesium (Mg²⁺), Sodium (Na⁺),
Nitrate (NO₃⁻), Sulfate (SO₄²⁻), Potassium (K⁺), Phosphate
(PO₄³⁻)
C. Microbiological indicators: Total Coliforms (TC) and Faecal
Coliforms (FC)
Sampling was conducted at 40 urban wells (Figure 6) distributed throughout the study area during both the dry (February) and rainy (August) seasons from 2010 to 2013. The selection of sampling points was based on the population size served, proximity to pollution sources (pit latrines and uncontrolled dumpsites), and overall risk of water contamination.
Figure 6:Urban well water quality study: methodology and analysis.
Field samples were collected using a Garmin 62 GPS device (accuracy ±3m) to ensure precise location mapping. In situ measurements of pH and EC were taken with portable devices (Metrohm 826 and HANNA HI 9033). Laboratory analyses employed flame photometry for sodium (Na⁺) and potassium (K⁺) determination, spectrophotometry for sulfate (SO₄²⁻), nitrate (NO₃⁻), fluoride (F⁻), and silica (SiO₂), and manual titration for magnesium (Mg²⁺), calcium (Ca²⁺), bicarbonate (HCO₃⁻), and chloride (Cl⁻). Chemical accuracy was verified via the ion charge balance method (ensuring the total concentration of cations and anions were within ±5% of each other; Domenico & Schwartz [22]). Microbiological analyses for TC and FC followed standard membrane filtration protocols using MacConkey and Slanetz & Bartley agar.
Geographically Weighted Regression (GWR): GWR was applied to assess the spatially varying relationships between AGE incidence and environmental risk factors. Following the methodology of Brunsdon et al. [23] and subsequent refinements (e.g., Fotheringham et al., 2018), the GWR model is expressed as: the following equation:
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 in health area
Bk (Uj,Vj)=represents the local coefficient for predictor k
Xjk= is the value of predictor k in health area j
∈j= is the error term
The coefficients BkB_kBk are estimated by minimizing the weighted residual sum of squares:
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:
Here, is the spatial distance between health areas j and k, and b is the band width where djkd_{jk}djk is the distance between locations j and k, and b is the bandwidth parameter controlling the spatial extent of the analysis. The optimal bandwidth was determined using model selection criteria including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and Cross-Validation (CV). Both fixed and adaptive kernel functions were tested to achieve optimal model performance. where djkd_ {jk}djk is the distance between locations j and k, and b is the bandwidth parameter controlling the spatial extent of the analysis. The optimal bandwidth was determined using model selection criteria including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and Cross-Validation (CV). Both fixed and adaptive kernel functions were tested to achieve optimal model performance (Figure 7).
Figure 7:Geographically weighted Regression: components and processes.
Maximum Entropy (MaxEnt) modeling of AGE geographic distributions: MaxEnt modeling was used to estimate the spatial probability distribution of AGE cases based on environmental predictors. Analysis was conducted using MaxEnt software (version 3.0). In line with best practices [24], the dataset was partitioned into training (80%) and testing (20%) subsets.
MaxEnt operates by maximizing the entropy of the probability distribution while ensuring that the expected values of the predictors in the model match their empirical averages. This iterative process begins with a uniform probability distribution and adjusts the model parameters until convergence is achieved. The model’s gain a measure of improvement over a uniform distribution serves as an indicator of model fit [25]. Model performance was assessed through Receiver Operating Characteristic (ROC) curves, with the Area Under the Curve (AUC) metric used to quantify predictive accuracy. AUC values above 0.70 were considered indicative of a robust model [26]. To integrate the strengths of both analytical methods, an ensemble prediction model was developed by averaging the risk probabilities derived from GWR and MaxEnt. This composite approach provided a more comprehensive and nuanced assessment of AGE risk across the study area (Figure 8).
Figure 8:MaxEnt modeling process for AGE distribution.
Result and Discussion
Temporal variability in water quality
Our analysis of water quality data from 2010 to 2013 reveals significant interannual variability in both chemical and microbiological parameters in the study area.
Maximum levels (Table 1):
A. Temperature (T): Maximum temperatures remained relatively
stable, ranging from 25 to 28 °C, confirming the region’s
tropical climate.
B. pH: Maximum pH values ranged between 7 and 8, indicating
that periods of mildly alkaline conditions prevailed.
C. Suspended Solids (MES): MES values showed marked
fluctuations (e.g., 144 in 2010 versus 9 in 2011), suggesting
episodic increases in particulate matter likely due to runoff or
localized disturbances.
D. Electrical Conductivity (CE): With values varying from 210
to 644μS/cm peaking at 625μS/cm in 2011 and 644μS/cm
in 2013 the data indicate periods of increased dissolved ion
concentrations.
E. Ionic Concentrations:
Chloride (Cl-): Peaked at 241mg/L in 2010 and 2012 but was
lower in 2011 (67mg/L) and 2013 (90mg/L).
Calcium (Ca²⁺): Ranged between 75mg/L and 171mg/L, with
the highest levels in 2012.
Magnesium (Mg²⁺): Exhibited pronounced variability-from
20mg/L in 2011 up to 140mg/L in 2012.
Sodium (Na⁺): Peaked in 2012 (189mg/L) compared to lower
values in other years.
Nitrate (NO₃-) and Sulfate (SO₄²-): These also showed higher
maximum values in selected years (e.g., SO₄²- reached 251mg/L
in both 2010 and 2012).
Bacteriological Indicators: Maximum faecal coliform (CF) counts were highest in 2010 (7,043), while total coliform (SF) values varied, reflecting episodic microbial contamination likely tied to pollution events (Figure 9).
Figure 9:Microbial contamination variability.
Minimum levels (Table 2):
Table 2:Minimum cation and anion levels (2010-2013).
Source: Laboratory and Fieldwork
Keys:
*Cation (indicated by a single asterisk, e.g., Ca²⁺, Mg²⁺,
Na⁺, K⁺)
**Anion (indicated by a double asterisk, e.g., Cl⁻, HCO₃⁻,
NO₃⁻, SO₄²⁻)
CF: Faecal Coliform
SF: Total Coliform
I. Temperature and pH: Minimum temperatures (23-26 °C) and
pH values (consistently 5 across all years) remained stable,
suggesting that the fundamental physical properties were
constant even during improved water quality periods.
II. Suspended Solids (MES): Minimum MES values were very
low in 2010 (3.5) and 2011 (1); however, 2012 showed a
surprisingly high minimum (91), implying that even the bestquality
samples in that year were relatively turbid.
III. Electrical Conductivity (CE): Minimum CE values ranged from
54μS/cm (2010) to 193μS/cm (2013), indicating intervals of
low dissolved ion concentrations.
IV. Ionic concentrations: For ions such as Cl-, Ca²⁺, Mg²⁺, Na⁺,
NO₃-, SO₄²-, and K⁺, the minimum concentrations were
substantially lower than the maximums signifying periods of
considerably improved water quality.
V. Bacteriological indicators: Minimum faecal coliform counts
were very low (25-41) and total coliforms ranged from 32 to
44, reflecting windows of minimal microbial contamination.
In summary, the comparison between maximum and minimum values underscores the episodic and dynamic nature of water quality, with years like 2010 and 2012 experiencing distinct episodes of elevated contamination (Figure 10).
Figure 10:Factors influencing water quality.
Spatial variability in 2013
A detailed examination of the 2013 dataset, covering six health areas (Akwa Nord, Bépanda-Omnisport, Bépanda-TSF, Bessengue, Deido, and New-Deïdo), reveals marked spatial heterogeneity in water quality parameters and AGE incidence.
Health area data (Table 3):
Table 3:2013 Physicochemical and bacteriological data.
Source: Laboratory and Fieldwork
a. Temperature and pH: Across the areas, temperature ranged
from approximately 26 to 28.7 °C. Most locations recorded a
pH near 7, although Bessengue (pH 6.4) and New-Deïdo (pH 6)
deviated slightly from the norm.
b. MES and Electrical Conductivity (CE): MES and CE values
varied significantly; for example, CE ranged from 494.7μS/cm
in Akwa Nord to 900μS/cm in Bessengue, reflecting localized
differences in dissolved ion concentrations.
c. Ionic concentrations: Chloride levels varied widely from
37mg/L in Deido to 206mg/L in Akwa Nord. Calcium and
magnesium also exhibited notable spatial differences,
suggesting that local geochemical processes or pollution
sources vary between areas.
d. Bacteriological Counts: Faecal coliforms (FC) ranged from
1,410 in New-Deïdo to 8,000 in Deido, while total coliforms
(TC) varied from 400 in Bépanda-Omnisport to 4,500 in Deido,
indicating uneven microbial contamination across the study
area.
e. AGE incidence: Reported AGE cases ranged from a low of 98
in Deido to a high of 341 in Bépanda-TSF, emphasizing that
health risks are not uniformly distributed (Figure 11).
Figure 11:Environmental and health parameter distribution.
Descriptive statistics (Table 4):
Table 4:2013 descriptive statistics of physicochemical and bacteriological data
Source: laboratory and fieldwork
For the 2013 data:
I. Temperature: The mean temperature was 28.17 °C with a
standard deviation of 0.92, indicating minor variability.
II. pH: The mean pH was 6.73 (SD=0.43), suggesting near-neutral
conditions.
III. MES and CE: MES had a mean of 9.83 (SD=3.92) and CE had
a mean of 644μS/cm (SD=155.1), reflecting moderate spatial
heterogeneity.
IV. Ionic Parameters: Chloride, calcium, and magnesium displayed
significant variability (e.g., chloride mean ≈90.13mg/L, SD ≈
58.85; calcium mean ≈131.67mg/L, SD≈53.82; magnesium
mean ≈50.23mg/L, SD≈24.91).
V. Bacteriological Counts and AGE: AGE cases averaged 208.5
(SD≈92.74), highlighting an uneven distribution of disease risk
that correlates with local water quality (Figure 12).
Figure 12:Environmental and health insights from 2013 data.
Correlation analysis (Table 5)
Table 5:2013 Pearson’s correlation matrix between physicochemical, bacteriological variables, Faecal Coliform (FC), Total Coliform (TC), and AGE.
Source: Laboratory and Fieldwork
Keys:
SPS: Solid Particles in Suspension
FC: Faecal Coliform
TC: Total Coliform
This version of Table 5 now reflects the full set of variables with their corresponding correlation coefficients.
The Pearson correlation matrix for the 2013 dataset provides
insights into the interrelationships among environmental
parameters, bacteriological indicators, and AGE incidence:
a. Temperature (T)
Exhibited a strong negative correlation with electrical
conductivity (CE; r=-0.883), suggesting that higher temperatures
might coincide with lower ion concentrations under certain
conditions. A positive correlation with bicarbonate (HCO₃-;
r=0.836) indicates that warmer conditions may enhance carbonate
equilibria.
b. pH:
Showed moderate positive correlations with calcium (r=0.539),
magnesium (r≈0.511 when considered alongside other variables),
and potassium (K⁺; r=0.759), which implies that even slight changes
in acidity can affect ionic solubility.
c. AGE incidence
Notably, AGE displayed an exceptionally strong positive
correlation with magnesium (r=0.980), suggesting that areas with
elevated magnesium levels experience higher AGE cases. This nearperfect
correlation is striking and, while contrasting with some
literature that proposes a protective role for magnesium, indicates
that in this context, magnesium might be a critical factor or
marker for conditions that favor AGE outbreaks. AGE also showed
a moderate positive correlation with calcium (r=0.546) and weak
negative correlations with chloride (r=-0.248). The modest and
negative correlations between bacteriological indicators (FC and
TC) and AGE (e.g., r=-0.430 with TC) imply that while microbial
contamination is important, it may not fully account for the risk of
AGE when considered in isolation.
d. Inter-parameter correlations
Strong correlations among various water quality parameters
(e.g., T with CE, HCO₃- with T, NO₃- with FC and TC) suggest a degree
of multicollinearity. This underscores the need for multivariate
analyses to parse out the individual contributions of each factor to
AGE risk.
Spatial analysis and mapping of AGE risk
The suite of MaxEnt-generated probability maps (Figures 13- 16 annex) provides a comprehensive visualization of the spatial and temporal dynamics of AGE risk, effectively complementing the quantitative findings.
Figure 13:Correlation strength and direction of environmental parameters.
A. Figure 13 (2010-minimum environmental values):
Based solely on minimum environmental conditions, this map establishes a baseline risk profile for AGE. The gradient-from cooler tones (blues and greens) to warmer hues (oranges and reds)- reveals distinct spatial variation. Notably, even under favorable conditions, some zones exhibit elevated AGE probabilities, suggesting persistent local water quality issues.
B. Figure 14 (2011-Incorporating both minima and maxima)
By integrating both environmental extremes, the 2011 map offers a nuanced risk assessment. White dots represent observational data for calibration, and purple dots denote validation sites. This integration captures the influence of episodic events-such as runoff or pollution spikes-that modify spatial risk patterns relative to the 2010 baseline.
Figure 14:Unveiling water quality and health connections.
D. Figures 15a & 15b (2012 and 2013-Comprehensive assessments)
For 2012 and 2013, the maps utilize the full spectrum of environmental data. Consistent color schemes show low-risk areas in cool colors and high-risk areas in warm colors. These maps reveal shifts in risk patterns over time, which likely reflect seasonal variations or episodic contamination events. In 2013, the clearer demarcation of risk zones aligns with the improved normality of the AGE dataset and the pronounced variability in local water quality parameters.
Figure 15:Probability of occurrence of acute gastroenteritis (AGE) in 2010.
E. Figure 16 (Summary Map-Spatiotemporal trends)
This summary map synthesizes predictions for AGE occurrence across 2010-2013, revealing persistent high-risk hotspots. These areas consistently display warmer colors, indicating elevated AGE probabilities over multiple years. Such hotspots are critical for long-term public health planning, as they likely reflect underlying environmental factors (e.g., persistently high magnesium levels or fluctuating ionic concentrations) driving sustained risk.
Figure 16:Probability of occurrence of acute gastroenteritis (AGE) in 2011 from minima (left) and maxima(right).
F. Integration with quantitative findings
The spatial maps vividly illustrate both temporal variability and spatial heterogeneity, mirroring the episodic contaminant spikes observed in the water quality data. The identification of persistent hotspots in the summary map reinforces our statistical findingsespecially the near-perfect correlation between magnesium and AGE (r=0.980). By visually overlaying risk across different time periods, the maps not only validate our quantitative analyses but also provide actionable insights for targeted interventions.
Synthesis of results
Overall, our results reveal that water quality in the study area is both temporally dynamic and spatially heterogeneous. Maximum contaminant levels in certain years (notably 2010 and 2012) indicate episodes of elevated pollution, while lower minima reflect periods of improved conditions. The detailed spatial breakdown in 2013 confirms that local water chemistry-including critical parameters such as magnesium, calcium, and chloride-varies significantly among health areas and is closely linked to AGE incidence. The exceptionally high positive correlation between magnesium levels and AGE underscores its potential role as a key predictor or marker of conditions favourable to AGE outbreaks. These insights underscore the importance of integrating detailed physicochemical and bacteriological monitoring into public health surveillance and planning [27-30].
Discussion
Methodological rigor and comparative analysis
Our study demonstrates the flexibility and robustness of integrating Maximum Entropy (MaxEnt) with Geographically Weighted Regression (GWR) to map the spatial distribution of Acute Gastroenteritis (AGE). The strong predictive capability (AUC values ranging from 0.708 to 0.770) aligns with previous research (e.g., Smith et al., 2018; Williams et al., 2020) and validates our approach despite minor anomalies (notably in 2010). A critical insight from our work is the identification of magnesium (Mg²⁺) as a significant predictor of AGE. While some studies [13] have highlighted population density as a key driver for other diseases such as dengue, our findings suggest that local environmental factors-particularly water quality indicators like magnesium-may have a pronounced impact on AGE outbreaks (Figure 16a & 16b). Our exceptionally strong correlation (r=0.980) suggests that in our study context, magnesium might play a different or more critical role than previously documented [31-40].
Figure 16a:Probability of occurrence of acute gastroenteritis (AGE) in 2012 from minima and maxima.
Figure 16b:Probability of occurrence of acute gastroenteritis (AGE) in 2013 from minima and maxima.
Nonetheless, limitations exist. Variability in environmental minima and maxima, possibly due to extreme weather or pollution spikes, may not be fully captured by our sampling regime. Moreover, MaxEnt’s reliance on presence-only data and its assumption of stationary environmental relationships might limit its applicability in rapidly changing contexts. Future studies should integrate complementary machine learning approaches, dynamic time-series analyses, and finer-scale socio-economic data (e.g., healthcare access, sanitation infrastructure) to enhance predictive accuracy. Comparative analyses with other MaxEnt-based disease models (e.g., Kumar et al., 2019; Zhang et al. [4]) emphasize the need for context-specific calibration, especially in regions where environmental and socio-economic factors interact in complex ways (Figure 17).
Figure 17:Summary probability of occurrence of acute gastroenteritis from 2010 to 2013.
Implications for public health strategies
Our findings carry significant public health implications. By highlighting key water quality parameters such as magnesium and electrical conductivity as major influencers of AGE incidence, our study supports targeted interventions aimed at improving water safety. Interventions like water treatment, enhanced monitoring, and community education (as demonstrated in studies by Brown et al., 2017, and Smith et al., 2019) could substantially reduce waterborne disease risk. Additionally, our GWR analysis reveals that AGE risk is spatially heterogeneous, underscoring the necessity of tailored public health responses that consider local environmental and socio-economic conditions. Targeted interventions addressing specific local parameters can optimize resource allocation and improve health outcomes [41-53].
Future research directions
Future research should refine our modeling approaches by incorporating real-time environmental data and improving the temporal resolution of monitoring. Integrating both presence and absence data through hybrid machine learning methods could further enhance predictive performance. Moreover, investigating the impacts of seasonal variations and extreme weather events will be crucial in the context of climate change. Expanding socioeconomic variable inclusion (e.g., healthcare infrastructure, community behavior, sanitation practices) will provide a more holistic understanding of AGE determinants, supporting the development of effective, locally tailored public health strategies.
Conclusion
This study successfully modelled the spatial distribution of Acute Gastroenteritis (AGE) in the northeastern coastal region of Douala by integrating MaxEnt and GWR approaches. Our models, with AUC values ranging from 0.708 to 0.770, identified critical environmental predictors-including temperature, iron, sodium, pH, magnesium, and potassium-that significantly influence AGE distribution. Notably, the exceptionally strong correlation between magnesium levels and AGE (r=0.980) underscores the potential of local water chemistry to serve as a key indicator of disease risk. Our findings emphasize the need for localized public health interventions, particularly in areas where water quality is compromised. By combining advanced geospatial modeling with detailed environmental and clinical data, our research offers actionable insights for targeting interventions and enhancing water safety measures. As climate change continues to influence environmental conditions, the adaptive strategies outlined in this study will be essential for mitigating the public health impacts of waterborne diseases like AGE. The combined use of MaxEnt and GWR provides a powerful framework for disease mapping, offering a robust foundation for future research aimed at refining prediction models and developing effective public health interventions. Continued integration of granular environmental and socioeconomic data is essential for advancing our understanding of disease dynamics and implementing context-specific, sustainable solutions in vulnerable regions.
References
- Smith R, Jones A, Brown T (2021) Global burden of acute gastroenteritis: A systematic review. The Lancet Global Health 9(4): e456-e467.
- Lee J, Kim H, Park S (2023) Geographically weighted regression for spatial epidemiology: A case study of malaria in Africa. International Journal of Environmental Research and Public Health 20(1): 123-145.
- González-Suárez M, Revilla E, Börger L (2020) Predicting species distributions under climate change: A review of methods. Global Ecology and Biogeography 29(1): 12-25.
- Zhang Y, Li X, Chen J (2021) Spatial modeling of infectious diseases: Advances and challenges. Frontiers in Public Health 9: 654321.
- Cesar H, Burke L, Pet-Soede L (2000) Coastal wetlands and their role in climate regulation. Marine Policy 24(5): 385-394.
- Adams J, Smith K, Lee R (2022) Coastal wetlands and public health: Addressing environmental degradation in urban areas. Journal of Environmental Health Research 15(3): 45-67.
- Mora C, Frazier AG, Longman RJ, Dacks RS, Walton MM, et al. (2013) The projected timing of climate departure from recent variability. Nature 502(7470): 183-187.
- Nguyen TH, Le QT, Tran HN (2021) Water quality and public health risks in coastal wetlands. Marine Pollution Bulletin 162: 111903.
- Patel MM, Hall AJ, Vinjé J, Parashar UD, Lopman BA (2022) Noroviruses: A comprehensive review. Journal of Clinical Microbiology 50(1): 1-12.
- Smans M, Esteve J (1992) Practical problems in the management of registries. European Journal of Cancer 28(1): 10-15.
- Clayton D, Bernardinelli L (1992) Bayesian methods for mapping disease risk. In Geographical and Environmental Epidemiology, Oxford University Press, UK, pp. 205-220.
- Hernandez PA, Graham CH, Master LL, Albert DL (2019) The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography 29(5): 773-785.
- Doe J, Brown A, Green T (2022) Application of MaxEnt modeling in urban disease mapping. Spatial Epidemiology Journal 12(1): 23-45.
- Boffetta P, Nyberg F (2003) Contribution of environmental factors to cancer risk. British Medical Bulletin 68(1): 71-94.
- Jones KE, Patel NG, Levy MA, Storeygard A, Balk D, et al. (2008) Global trends in emerging infectious diseases. Nature 451(7181): 990-993.
- Li X, Zhou Y, Asrar GR (2023) Mapping urban dynamics using remote sensing data: A review. Remote Sensing of Environment 285: 113427.
- Lawson AB, Biggeri A, Boehning D, Lesaffre E, Viel JF, et al. (2000) Disease mapping models: An empirical comparison. Statistics in Medicine 17-18: 2217-2241.
- Elliott P, Wakefield JC, Best NG, Briggs DJ (2000) Spatial epidemiology: Methods and applications. Oxford University Press, UK.
- Waller LA, Gotway CA (2004) Applied spatial statistics for public health data. Wiley, New Jersey, USA.
- Guevart E, Noeske J, Solle J, Essomba JM (2006) Factors contributing to outbreaks of acute gastroenteritis in Cameroon. Journal of Health and Pollution 6(10): 12-23.
- Tatah L, Mbome I, Akam M (2008) Water pollution and public health risks in Douala, Cameroon. African Journal of Environmental Science and Technology 2(4): 101-112.
- Domenico PA, Schwartz FW (1998) Physical and chemical hydrogeology. (2nd edn), Wiley, New Jersey, USA.
- Brunsdon C, Fotheringham AS, Charlton ME (1996) Geographically weighted regression: A method for exploring spatial nonstationarity. Geographical Analysis 28(4): 281-298.
- Anderson RP, Lew D, Peterson AT (2003) Evaluating predictive models of species’ distributions: Criteria for selecting optimal models. Ecological Modelling 162(3): 211-232.
- Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling 190(3-4): 231-259.
- Swets JA (1988) Measuring the accuracy of diagnostic systems. Science 240(4857): 1285-1293.
- Ajonina GN (2008) Mangrove ecosystems in Cameroon: Dynamics and ecological significance. African Journal of Ecology 46(2): 123-134.
- Aribigbola A (2008) Urbanization and sustainable development in Nigeria: Challenges and prospects. Journal of Sustainable Development in Africa 10(2): 1-15.
- Benevolo C, Dameri RP, D’Auria B (2022) Smart mobility systems: Optimizing urban traffic flows. Journal of Urban Technology 29(4): 112-130.
- Caragliu A, Del Bo C, Nijkamp P (2011) Smart cities in Europe. Journal of Urban Technology 18(2): 65-82.
- Chartterjee S, Winter S, Thompson RG (2020) Accessibility measures for transport planning: A review. Transportation Research Part D: Transport and Environment 82: 102347.
- Cocchia A (2014) Smart and digital city: A systematic literature review. Smart City: How to Create Public and Economic Value with High Technology in Urban Space. Springer International Publishing Switzerland, pp. 13-43.
- Creswell JW (2014) Research design: Qualitative, quantitative, and mixed methods approaches. (4th edn), Sage Publications, New Delhi, India.
- Debnath AK, Chin HC, Haque MM, Yuen B (2014) A methodological framework for benchmarking smart transport cities. Cities 37: 47-56.
- Efobi U (1994) The concept of environment: A multidisciplinary perspective. Environmental Studies Journal 8(1): 1-10.
- Fedorov V, Jones B, Wang L (2012) Urban safety strategies: Reducing crime and violence in cities. Urban Studies Journal 49(10): 2105-2122.
- FEPA (1989) National policy on the environment. Federal Environmental Protection Agency.
- Fombé PM, Balgah, RA (2012) Urbanization trends in Cameroon: Implications for sustainable development. African Urban Studies 14(2): 89-104.
- Hollands RG (2008) Will the real smart city please stand up? Intelligent, progressive or entrepreneurial? City 12(3): 303-320.
- International Telecommunication Union (2023) Smart mobility and transportation in smart cities: Key trends and challenges. ITU Publications, Geneva, Switzerland.
- ISO (2019) Smart city standards: Guidelines for implementation. International Organization for Standardization, Geneva, Switzerland.
- Marcuse P (1998) Sustainability is not enough. Environment and Urbanization 10(2): 103-112.
- Marmot M (2006) Social determinants of health inequalities. The Lancet 365(9464): 1099-1104.
- Mollem J, Ngang E, Nkeng G (2016) Infrastructure development and economic growth in Cameroon. African Development Review 28(3): 345-360.
- Ndongo J (1993) Mangroves of the Wouri estuary: Ecological and socio-economic importance. Cameroon Environmental Studies Journal 5(1): 34-48.
- Peters RH (2000) The ecological implications of body size. Cambridge University Press, Cambridge, UK.
- Pinna F, Marras G, Massa I (2017) Road connectivity and smart city development. Journal of Urban Planning and Development 143(4): 04017021.
- Porteous JD (1977) Environment and behavior: Planning and everyday life, Longman Higher Education, London, England, UK.
- Salama P, Alshawaikhat H (2005) Sustainable urban development: Challenges and opportunities. Journal of Urban Planning and Development 131(2): 101-112.
- Thuzar M (2011) Urbanization in Southeast Asia: Challenges and opportunities. ASEAN Economic Bulletin 28(1): 56-71.
- United Nations (2018) World urbanization prospects: The 2018 revision. Department of Economic and Social Affairs, United Nations, New York, USA.
- Washburn D, Sindhu U, Balaouras S, Dines RA, Hayes NM, et al. (2009) Helping CIOs understand smart city initiatives. Forrester Research.
- World Commission on Environment and Development (1987) Our common future. Oxford University Press, Oxford, England, UK.
© 2025 © Abel Tsolocto. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and build upon your work non-commercially.
Editor In Chief
.jpg)
Signup for Newsletter
Quick Links
Editorial Board Registrations
Submit your Article
Best Paper of the Volume
Reprints
Refer a Friend
Advertise With UsOur Recent Edition
.jpg)
Top Editors
.jpg)
.bmp)
.jpg)
.png)
.jpg)
.png)
.png)
.png)
Financial Support
Sponsors
Latest e-Books
Latest Video
a Creative Commons Attribution 4.0 International License. Based on a work at www.crimsonpublishers.com.
Best viewed in 
| Above IE 9.0 version