Analytical Hierarchical Process for Modelling Malaria Vulnerability Index Among Local Government Areas in Bayelsa State, Nigeria
A
Okpachi
Abbah
1✉
Olalekan
John
Taiwo
2,4
Emailolalekantaiwo@gmail.com
Emailolalekan.taiwo@coronams.com
James
Olaoye
Oyeleye
2
Ganiyat
Eshikhena
2
Tamaraebi
Borme
2
German
Wisdom
2
Dupsy
Akoma
3
Fayokemi
Olususi
3
Ifeoma
Ezenyi
3
Chioma
Oluebubechukwu
Unogu
1
Chijioke
Kaduru
5
1
Department of Climate Change, Corona Management Systems
Abuja
Nigeria
2
Department of Infectious Disease, Corona Management Systems
Abuja
Nigeria
3
Department of Clinical, Epidemiological and Social Research
Corona Management Systems
Abuja
Nigeria
4
A
Faculty of the Social Sciences, Department of Geography
University of Ibadan
Oyo State
Nigeria
5
Management Unit
Corona Management Systems
Abuja
Nigeria
Okpachi Abbah*1, Olalekan John Taiwo2,4, James Olaoye Oyeleye2, Ganiyat Eshikhena2, Tamaraebi Borme2, German Wisdom2, Dupsy Akoma3, Fayokemi Olususi3, Ifeoma Ezenyi3, Chioma Oluebubechukwu Unogu1, Chijioke Kaduru5
Authors Affiliations:
1. Department of Climate Change, Corona Management Systems, Abuja, Nigeria
2. Department of Infectious Disease, Corona Management Systems, Abuja, Nigeria
3. Department of Clinical, Epidemiological and Social Research, Corona Management Systems, Abuja, Nigeria
4. Faculty of the Social Sciences, Department of Geography, University of Ibadan, Oyo State, Nigeria
5. Management Unit, Corona Management Systems, Abuja, Nigeria
*Corresponding author: Olalekan John Taiwo; olalekantaiwo@gmail.com, olalekan.taiwo@coronams.com;
Abstract
Background
Persistent malaria transmission in Africa underscores the need for spatially explicit tools that identify highly endemic areas for targeted control. Although multi-criteria decision analysis (MCDA) offers a structured approach, its application has been limited by outdated environmental inputs and inconsistent factor aggregation methods. This study developed an ecology-informed Malaria Vulnerability Index (MVI) for Bayelsa State, Nigeria, using up-to-date, open-source geospatial datasets and a transparent weighting framework.
Methods
Thirteen environmental predictors were sourced from OpenStreetMap, Google Earth Engine, WorldPop, and GRID3. Using the Analytical Hierarchy Process (AHP), a 13×13 pairwise comparison matrix was constructed and solved using the eigenvalue method to derive criterion weights. Weighted predictors were combined to generate the MVI, which was overlaid with gridded population data to quantify population exposure. Associations between population counts across low, medium, and high vulnerability zones and reported malaria cases were assessed using correlation analysis.
Results
The highest-priority criteria were distance to streams, distance to wetlands, precipitation, topographic wetness index, and land surface temperature. Medium vulnerability dominated the landscape (77.1%), followed by low (17.2%) and high (5.7%) vulnerability. High-vulnerability areas were concentrated in riverine LGAs, particularly Southern Ijaw (40.25%), Brass (19.30%), Ekeremor (17.36%), and Sagbama (15.89%). Population exposure reflected these patterns: 3.63% of residents lived in high-vulnerability zones, 74.66% in medium, and 21.70% in low zones. Population in low-vulnerability areas showed a strong correlation with reported malaria cases (r = 0.914), while total population also correlated with cases (r = 0.719).
Conclusion
Malaria vulnerability in Bayelsa State is primarily driven by hydrological and hydroclimatic conditions, especially proximity to streams and wetlands, rainfall, and microtopographic wetness. The AHP-based MCDA framework provides a rigorous and transparent approach for integrating environmental factors, supporting hydrology-focused targeting of malaria surveillance and vector control, and enabling reproducible MVI mapping using open-source geospatial data.
Keywords:
Malaria Vulnerability Index (MVI)
Analytic Hierarchy Process (AHP)
Multicriteria Decision Analysis (MCDA)
Geographic Information System (GIS)
Bayelsa State
Spatial Epidemiology
Health Geographics
A
1. Introduction
Malaria remains a leading cause of preventable morbidity and mortality in sub-Saharan Africa, with Nigeria persistently accounting for a substantial share of the continent’s Plasmodium falciparum burden (1–4). Transmission is intensely heterogeneous over short distances due to interactions among climate, hydrology, vector ecology, human settlement, and access to prevention and care (5–7). Bayelsa State, situated in the coastal Niger Delta, exemplifies a hydro-ecological template of low elevation, mangrove-swamp mosaics, and high, year-round rainfall that sustains Anopheles receptivity and complicates uniform program strategies (8, 9). Socio-environmental conditions, including housing quality and peri-domestic exposure, further modulate risk, reinforcing the need to integrate multiple determinants when assessing vulnerability (10).
A
To guide intervention planning in such settings, decision-ready metrics that integrate multiple vulnerability drivers are essential. Multicriteria decision analysis (MCDA) provides a principled framework for integrating diverse malaria determinants into a single, interpretable index, aligning with the broader movement toward transparent, evidence-based priority setting in health(11) (12). Within Geographic Information System (GIS) based MCDA, the Analytic Hierarchy Process (AHP) formalises expert judgment into weights while enforcing internal coherence via a consistency ratio (CR), and Weighted Linear Combination (WLC) offers an intuitive overlay for combining normalised criteria into continuous vulnerability surfaces (13, 14). In infectious disease and malaria risk assessment, GIS-MCDA approaches have proved useful for stratification, site selection, and targeting of interventions by combining remotely sensed layers (e.g., land cover, elevation, Normalized Difference Vegetation Index (NDVI), rainfall) with program data (e.g., Insecticide Treated Net/ Indoor Residual Spraying (ITN/IRS) coverage, test positivity) to capture vulnerability that no single indicator can reveal (5, 14). Methodological advances, including sensitivity analysis to examine the stability of results underweight perturbations, strengthen the credibility and transferability of MCDA outputs (15). While malaria risk mapping has widely leveraged remote sensing and routine surveillance to characterise spatial heterogeneity reproducible AHP–WLC applications that transparently document weighting, produce LGA-resolved indices, estimate population exposure by risk class, and validate composite outputs against routine caseloads remain sparse in high-burden, ecologically complex settings like Bayelsa State, Nigeria (2, 5, 8).These gaps are operationally consequential in Bayelsa, where decision makers must prioritise interventions across hydrologically diverse LGAs with varying access to prevention and care. Existing subnational risk assessments often rely on single-domain proxies (e.g., environmental suitability alone or routine incidence alone), apply ad-hoc or opaque weighting, and seldom quantify uncertainty or assess convergence with epidemiological indicators, and these limitations blunt their programmatic value (14–16); (17); (18). To address these needs, this study develops a transparent, reproducible Malaria Vulnerability Index (MVI) for Bayelsa State using AHP and WLC, tightly coupled to routine data. Our objectives are to develop a comprehensive malaria vulnerability assessment for Bayelsa State through the creation of a consistent pairwise comparative matrix of vulnerability indicators, generation of a composite malaria vulnerability index map, ranking of Local Government Areas based on vulnerability scores, estimation of population proportions within each vulnerability category, and analysis of the relationship between confirmed uncomplicated malaria cases and areas under different vulnerability classifications. Together, these objectives link mechanistic vulnerability to observed burden and quantify how many people reside in each class, which is key information for prioritisation (6, 7).
2 Methodology
2.1 Data sources and processing
We constructed a spatially explicit Malaria Vulnerability Index (MVI) for Bayelsa State, Nigeria, by integrating harmonised geospatial datasets on hydroclimate, terrain, land use/land cover (LULC), hydrology, accessibility, population, administrative boundaries, and routine malaria surveillance. Details of these are presented in Table 1.
|
Dataset
|
Source
|
Resolution
|
Application
|
|---|---|---|---|
|
Digital Elevation Model (DEM)
|
NASA/USGS Shuttle Radar Topography Mission (SRTM v3)
|
1 arc-second (~ 30 m)
|
Topographic and terrain derivatives; slope and Topographic Wetness Index (TWI) computation for hydrological modelling.
|
|
Slope & Topographic Wetness Index (TWI)
|
Derived from SRTM DEM using hydrologic conditioning (D8 flow direction/accumulation framework)
|
~ 30 m
|
Indicators of surface runoff, soil moisture, and mosquito habitat suitability.
|
|
Precipitation (CHIRPS v2.0)
|
Climate Hazards Group InfraRed Precipitation with Stations (19)
|
0.05° (~ 5 km)
|
Hydroclimate driver of mosquito breeding and malaria transmission risk.
|
|
Land Surface Temperature (LST, MOD11A2 v6.1)
|
MODIS Terra (20)
|
1 km
|
Thermal suitability for malaria vectors and parasite development.
|
|
Vegetation Index (NDVI, MCD13Q1 v6.1)
|
MODIS Terra/Aqua combined (21)
|
250 m
|
Proxy for vegetation cover, mosquito resting/breeding habitats.
|
|
Land Use/Land Cover (LULC)
|
Esri/Microsoft Impact Observatory Global Land Cover (22)
|
10 m
|
Binary masks & proximity surfaces for open water, wetlands, cropland, built-up areas, trees, rangeland; linked to malaria vulnerability.
|
|
Hydrologic Networks (Rivers/Streams))
|
HydroSHEDS/HydroRIVERS (23)
|
Vector (polyline)
|
Hydrologically consistent networks for computing proximity to rivers/water bodies
|
|
Accessibility Boundaries (Road Networks)
|
OpenStreetMap (24)
|
Vector (line features)
|
Accessibility surfaces; Euclidean distance to transportation routes.
|
|
Health Facilities (Public & Private)
|
GRID3 Nigeria (25)
|
Geocoded point features
|
Euclidean distance to facilities; health service accessibility.
|
|
Administrative Boundaries (LGA)
|
GRID3 Nigeria (25)
|
Vector (Polygon)
|
Alignment with reporting units; aggregation of surveillance and demographic data
|
|
Population Counts Data
|
WorldPop 2020 Nigeria gridded counts (26, 27)
|
~ 100 m
|
Denominators for malaria incidence calculation; population vulnerability mapping.
|
|
Routine Malaria Surveillance
|
DHIS2-based Health Management Information System (2024)
|
LGA-level counts
|
Laboratory-confirmed malaria cases; incidence calculation following WHO standards.
|
2.2 Data Analysis
2.2.1 Data harmonisation
All spatial data sets were standardised to a single analysis grid to avoid resampling errors(28)(29) The study used the WGS 84/UTM Zone 32N coordinate system for metric accuracy. The Landuse/Landcover (LULC) raster, with a native resolution of 10m, served as the reference (snap raster)and defined the processing cell size. All rasters were aligned to this grid, while vector layers were reprojected before rasterisation.
We utilised malaria case data for 2024, whereas population denominators were derived from WorldPop 2020. Although this introduces a temporal mismatch, such is common in subnational studies. The use of externally validated WorldPop estimates reduces the risk of denominator bias and was accounted for in the uncertainty assessment (30, 31).
2.2.2 Criteria selection and surface derivation
Criteria were chosen to represent well-established drivers of malaria receptivity, exposure, and vulnerability in sub-Saharan Africa (1, 2, 7, 8, 32, 33). These included:
Hydrology and moisture:open water, wetlands, Topographic Wetness Index (TWI), proximity to rivers. These capture larval habitat availability and persistence (34–36).
Climate and greenness: Precipitation, land surface temperature (LST), NDVI. These approximate moisture and thermal suitability for vector and parasite development (9, 37–40).
Terrain: Elevation, slope. These modulates temperature and drainage (36).
LULC and accessibility variables:Proximity to cropland, rangeland, trees, built-up and bare ground; distance to roads; distance to health facilities. These reflect human–-=environment interactions and care access (7, 41–43).
2.2.3 Standardisation and vulnerability coding
To place heterogeneous inputs on a common scale and direction, all rasters were linearly rescaled to [0, 1], with 0 denoting the highest vulnerability and 1 denoting the lowest. Transformations were guided by epidemiological evidence:
Variables positively associated with vulnerability (higher raw values imply higher risk) were mapped with decreasing transforms so that higher raw values yield lower standardised scores. This applied to open water and wetlands (presence), TWI, precipitation, proximity to rivers (shorter distances imply higher vulnerability), cropland presence, NDVI within the local dynamic range, and LST within transmission-relevant bounds (9, 34, 35, 37, 38, 40).
Variables negatively associated with vulnerability were mapped with increasing transforms so that higher raw values imply higher standardised scores; this applied to elevation, slope, and LULC classes generally protective or less favourable for stable transmission in the West African urban context (trees, rangeland, bare ground, built-up) (7, 44–46).
For access variables, greater distance to roads or health facilities increases vulnerability; rescaling, therefore, yielded lower standardised scores at larger distances (2, 41, 43). Binary LULC presences received scores consistent with their direction (e.g., open water present = 0; absent = 1). Continuous variables were min–max transformed within the study area.
2.2.4 AHP Weighting
Weights for the criteria selected after expert consultation and judgement were assigned using the Analytic Hierarchy Process (AHP).Expert judgments were elicited via pairwise comparisons on Saaty’s 1–9 scale to form a 13 × 13 reciprocal judgment matrix. Normalised criteria weights were extracted from the principal right eigenvector. Internal consistency was evaluated using the Consistency Index (CI) and Consistency ratio (CR).
Where
is the Principal eigenvalue of the pairwise comparison matrix, n is the number of criteria(size of the matrix), RI is the random index, and RI13 is the Random index at the 13th criteria. Where
was considered an acceptable coherence (13, 47–49).
2.2.4 Composite MVI construction and classification
The continuous MVI was computed as a weighted linear combination:
where
s is the standardised score for criterion
at location
and
is its AHP-derived weight. Because of the coding, lower
values denote greater vulnerability. The continuous surface was reclassified into three ordinal classes (high, medium, low) using equal interval categorisation to ensure comparability (14).
2.2.5 Population at risk
We quantified the population residing in each vulnerability class within each LGA by overlaying the raster WorldPop 2020 with class-specific binary masks (low, medium and high vulnerability). Class totals and proportions were summarised using zonal statistics with LGA polygons. Islands and open water within LGA boundaries were masked before extraction to avoid over-counting uninhabitable areas.
2.2.6 Malaria incidence and association analysis
Confirmed malaria cases for 2024 were aggregated at the LGA level. Incidence per 1000 population was calculated using the WorldPop denominators.(31). Associations between incidence and MVI-derived population classes were evaluated using Pearson correlation coefficients, with 95% confidence intervals (CIs) reported. To adjust for multiple testing, the Benjamini–Hochberg correction was applied. Because of potential non-normality and the modest sample size, Spearman rank correlations were also conducted as sensitivity analyses.
Paired differences between class-specific population counts and malaria cases were tested using paired t-tests, with statistical significance set at p ≤ 0.05 (two-tailed). Effect sizes were calculated as Hedges’ g, with corresponding 95% CIs. All interpretations emphasised the scale dependence of counts and the ecological nature of the analysis.
2.2.7 Quality assurance and uncertainty
We appraised key sources of uncertainty: (i) temporal mismatch between 2020 denominators and 2024 cases, (ii) the modifiable areal unit problem (MAUP) due to aggregation at the LGA level, (iii) the subjectivity inherent in AHP, and (iv) classification and standardisation choices We mitigated these risks by aligning all raster operations to the Landuse/cover grid, analysing proportions where possible, using fixed reclassification thresholds, and checking the internal consistency of AHP weights. Spatial dependence was considered using Moran’s I for LGA-level outcomes where relevant, with permutation-based inference to guard against inflated significance(50, 51).
3 Results
3.1 AHP-derived criterion weights
Bayelsa State is a Nigerian state located within the Niger Delta region. Administratively, it has eight Local Government Areas (LGAs) and 105 local wards (Fig. 1). The LGAs are divided into two: four upland and four riverine. Dominant vegetation comprises mangroves and freshwater swamp forest. (52)
Fig. 1
Administrative Map of Bayelsa State, Nigeria, Showing Local Government Area.
We derived the criterion weights for the Bayelsa State malaria vulnerability index using Saaty’s Analytic Hierarchy Process (AHP). After redundancy screening and expert review, thirteen criteria were retained for the Analytic Hierarchy Process (AHP) weighting to balance parsimony and ecological coverage. Class-specific rasters and distance surfaces were derived as described above. Distances to rivers, roads, and health facilities were computed under planar (UTM) geometry and constrained to the state boundary before normalisation. Slope was expressed as a per cent rise. The 13×13 pairwise comparison matrix was built to reflect the ecology of malaria vectors in the Niger Delta, where dense hydrographic networks, seasonally inundated wetlands, and low relief dominate the landscape, and was solved with the eigenvalue method (Table 2). The normalised priority vector places the strongest emphasis on hydrological and hydro-meteorological determinants. Distance to streams emerged as the most influential criterion (w ≈ 0.218), followed by distance to wetland areas (w ≈ 0.199), precipitation (w ≈ 0.128), and the topographic wetness index (TWI; w ≈ 0.099) (Table 3). Land surface temperature (LST) carried a moderate weight (w ≈ 0.082), consistent with its well-documented role in modulating vector and parasite development. Secondary contributions were assigned to distance from cropland (w ≈ 0.068) and elevation (w ≈ 0.049), while slope and NDVI had smaller, roughly equivalent weights (both w ≈ 0.032). Access and infrastructure-related factors such as distance to healthcare facilities, distance to the road network, and distance from built-up areas were given modest and nearly identical weights (each w ≈ 0.026). Distance from bare ground made the smallest contribution (w ≈ 0.011). The priority ordering therefore followed: Distance to stream > Distance to wetland > Precipitation > TWI > LST > Distance from cropland > Elevation > Slope ≈ NDVI > Distance to healthcare ≈ Distance to roads ≈ Distance from built-up > Distance from bare ground. Model consistency was high: the maximum eigenvalue was approximately 13.64, yielding a consistency index (CI) of 0.0536 and a consistency ratio (CR) of 0.0343 using Saaty’s random index for n = 13 (RI = 1.56), well below the 0.10 threshold (13, 49).
|
Criteria
|
C1
|
C2
|
C3
|
C4
|
C5
|
C6
|
C7
|
C8
|
C9
|
C10
|
C11
|
C12
|
C13
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Distance to Stream (C1)
|
1
|
1
|
3
|
3
|
3
|
5
|
5
|
7
|
7
|
7
|
7
|
7
|
9
|
|
Distance to Wetland Area (C2)
|
1
|
1
|
3
|
3
|
3
|
3
|
5
|
5
|
5
|
7
|
7
|
7
|
9
|
|
Precipitation (C3)
|
0.33
|
0.33
|
1
|
1
|
3
|
3
|
3
|
5
|
5
|
5
|
5
|
5
|
9
|
|
Topographic Wetness Index (TWI) (C4)
|
0.33
|
0.33
|
1
|
1
|
1
|
1
|
3
|
3
|
3
|
5
|
5
|
5
|
9
|
|
Land Surface Temperature (LST) (C5)
|
0.33
|
0.33
|
0.333
|
1
|
1
|
1
|
3
|
3
|
3
|
3
|
3
|
3
|
9
|
|
Distance from Cropland Area (C6)
|
0.2
|
0.33
|
0.333
|
0.333
|
1
|
1
|
1
|
3
|
3
|
3
|
3
|
3
|
9
|
|
Elevation (C7)
|
0.2
|
0.2
|
0.333
|
0.333
|
0.333
|
1
|
1
|
1
|
1
|
3
|
3
|
3
|
5
|
|
Slope (C8)
|
0.14
|
0.2
|
0.2
|
0.333
|
0.333
|
0.333
|
1
|
1
|
1
|
1
|
1
|
1
|
5
|
|
Normalised Vegetation Index (NDVI) (C9)
|
0.14
|
0.2
|
0.2
|
0.333
|
0.333
|
0.333
|
1
|
1
|
1
|
1
|
1
|
1
|
5
|
|
Distance to Healthcare Facilities (C10)
|
0.14
|
0.14
|
0.2
|
0.2
|
0.333
|
0.333
|
0.333
|
1
|
1
|
1
|
1
|
1
|
3
|
|
Distance to Road Network (C11)
|
0.14
|
0.14
|
0.2
|
0.2
|
0.333
|
0.333
|
0.333
|
1
|
1
|
1
|
1
|
1
|
3
|
|
Distance from Built-up Area (C12)
|
0.14
|
0.14
|
0.2
|
0.2
|
0.333
|
0.333
|
0.333
|
1
|
1
|
1
|
1
|
1
|
3
|
|
Distance from Bareground (C13)
|
0.11
|
0.11
|
0.111
|
0.111
|
0.111
|
0.111
|
0.2
|
0.2
|
0.2
|
0.333
|
0.333
|
0.333
|
1
|
|
Column Sum
|
4.225
|
4.472
|
10.110
|
11.043
|
14.109
|
16.776
|
24.199
|
32.200
|
32.200
|
38.333
|
38.333
|
38.333
|
79.000
|
|
Criteria
|
Priority Weight
|
Percentage
|
|---|---|---|
|
Distance to Stream
|
0.2186
|
21.90%
|
|
Distance to Wetland Area
|
0.1999
|
20.00%
|
|
Precipitation
|
0.1288
|
12.90%
|
|
Topographic Wetness Index (TWI)
|
0.0991
|
9.90%
|
|
Land Surface Temperature (LST)
|
0.082
|
8.20%
|
|
Distance from Cropland Area
|
0.0686
|
6.90%
|
|
Elevation
|
0.0492
|
4.90%
|
|
Slope
|
0.0321
|
3.20%
|
|
Normalised Vegetation Index (NDVI)
|
0.0321
|
3.20%
|
|
Distance to Healthcare Facilities
|
0.0261
|
2.60%
|
|
Distance to Road Network
|
0.0261
|
2.60%
|
|
Distance from Built-up Area
|
0.0261
|
2.60%
|
|
Distance from Bareground
|
0.0112
|
1.10%
|
3.2 Statewide distribution of vulnerability classes
Across Bayelsa State, the MVI was dominated by the medium vulnerability class, which covered 77.1% of the area, with low and high classes covering 17.2% and 5.7%, respectively (Figs. 2 and 3). Medium vulnerability was the largest component in every LGA, ranging from 57.20% in Kolokuma/Opokuma to 89.05% in Nembe. Brass (8.84%), Sagbama (8.67%), and Southern Ijaw (7.74%) exhibited comparatively larger high-vulnerability areas, whereas Ogbia (0.24%), Kolokuma/Opokuma (0.67%), and Yenagoa (1.75%) had very small high-vulnerability fractions (Table 4). Low vulnerability was most prominent in Kolokuma/Opokuma (42.12%), Ogbia (34.25%), and Yenagoa (32.77%).
Fig. 2
Malaria Vulnerability Map of Bayelsa State, Nigeria
Fig. 3
Malaria Vulnerability Categories in Bayelsa State, Nigeria
|
LGANAME
|
Highly Vulnerable
|
Moderately Vulnerable
|
Low Vulnerable
|
Total
|
|---|---|---|---|---|
|
Ekeremor
|
89.42 (4.99%)
|
1399.9 (78.05%)
|
304.29 (16.97%)
|
1793.61
|
|
Southern Ijaw
|
207.38 (7.74%)
|
2317.48 (86.47%)
|
155.38 (5.80%)
|
2680.24
|
|
Nembe
|
21.73 (2.79%)
|
692.88 (89.05%)
|
63.5 (8.16%)
|
778.11
|
|
Brass
|
99.41 (8.84%)
|
866.62 (77.10%)
|
158.06 (14.06%)
|
1124.09
|
|
Ogbia
|
1.61 (0.24%)
|
445.63 (65.51%)
|
232.96 (34.25%)
|
680.2
|
|
Yenegoa
|
11.34 (1.75%)
|
423.39 (65.48%)
|
211.89 (32.77%)
|
646.62
|
|
Kolokuma/Opokuma
|
2.41 (0.67%)
|
204.53 (57.20%)
|
150.6 (42.12%)
|
357.54
|
|
Sagbama
|
81.88 (8.67%)
|
593.55 (57.20%)
|
268.82 (42.12%)
|
944.25
|
|
Total
|
515.18
|
6943.98
|
1545.5
|
9004.66
|
LGAs contributing most to the statewide high-vulnerability area were Southern Ijaw (40.25%), Brass (19.30%), Ekeremor (17.36%), and Sagbama (15.89%) (Table 5). For medium vulnerability, Southern Ijaw (33.37%) and Ekeremor (20.16%) dominated, with Brass (12.48%) and Nembe (9.98%) also contributing appreciably. Low-vulnerability contributions were led by Ekeremor (19.69%) and Sagbama (17.39%), followed by Ogbia (15.07%) and Yenagoa (13.71%).
|
LGANAME
|
High Vulnerability
|
Medium Vulnerability
|
Low Vulnerability
|
|---|---|---|---|
|
Ekeremor
|
17.36
|
20.16
|
19.69
|
|
Southern Ijaw
|
40.25
|
33.37
|
10.05
|
|
Nembe
|
4.22
|
9.98
|
4.11
|
|
Brass
|
19.30
|
12.48
|
10.23
|
|
Ogbia
|
0.31
|
6.42
|
15.07
|
|
Yenegoa
|
2.20
|
6.10
|
13.71
|
|
Kolokuma/Opokuma
|
0.47
|
2.95
|
9.74
|
|
Sagbama
|
15.89
|
8.55
|
17.39
|
3.3. Population at risk by LGA
At the state level, 3.63% of residents lived in high-vulnerability zones, 74.66% in medium, and 21.70% in low vulnerability zones. The medium vulnerability areas represented the majority in every LGA, ranging from 56.91% in Kolokuma/Opokuma to 84.56% in Nembe (Table 6). High-vulnerability proportions were modest but heterogeneous, spanning 0.32–8.25%, with Sagbama (8.25%), Southern Ijaw (6.09%), Brass (5.40%), and Ekeremor (4.64%) at the upper end. Low vulnerability was most prominent in Kolokuma/Opokuma (42.77%), Ogbia (29.08%), Sagbama (28.46%), and Yenagoa (27.63%).
|
LGA Name
|
High Vulnerability Population
|
Medium Vulnerability Population
|
Low Vulnerability Population
|
Total
|
|---|---|---|---|---|
|
Ekeremor
|
17,900.07 (4.64%)
|
314,440.13 (81.47%)
|
53,635.30 (13.90%)
|
385,975.51
|
|
Southern Ijaw
|
30,070.70 (6.09%)
|
407,011.32 (82.42%)
|
56,726.32 (11.49%)
|
493,808.34
|
|
Nembe
|
2,401.26 (1.41%)
|
143,834.60 (84.56%)
|
23,861.34 (14.03%)
|
170,097.20
|
|
Brass
|
7,700.30 (5.40%)
|
107,423.43 (75.31%)
|
27,514.50 (19.29%)
|
142,638.23
|
|
Ogbia
|
1,357.97 (0.48%)
|
199,348.57 (70.44%)
|
82,295.23 (29.08%)
|
283,001.77
|
|
Yenegoa
|
5,381.89 (0.93%)
|
411,440.51 (71.43%)
|
159,151.00 (27.63%)
|
575,973.40
|
|
Kolokuma/Opokuma
|
345.54 (0.32%)
|
61,018.05 (56.91%)
|
45,858.84 (42.77%)
|
107,222.44
|
|
Sagbama
|
23,742.34 (8.25%)
|
182,155.90 (63.29%)
|
81,920.12 (28.46%)
|
287,818.36
|
|
Total
|
88,900.08 (3.63%)
|
1,826,672.51 (74.66%)
|
530,962.65 (21.70%)
|
2,446,535.25
|
3.4 Association between vulnerability classes and malaria cases
Across LGAs, the mean malaria cases in 2024 were 8,143.6. Paired comparisons between class-specific population counts and case counts primarily reflected differences in scale; as expected, medium- and low-vulnerability populations substantially exceeded case counts (medium: mean difference 220,190.4, 95% CI 111,877.0–328,503.9, p = 0.002; low: 58,226.7, 95% CI 27,551.5–88,901.9, p = 0.003), yielding large effect sizes (Table 7). For the high-vulnerability class, the mean difference was small and non-significant (2,968.9, 95% CI − 8,644.4 to 14,582.1, p = 0.565).
|
Paired Samples Test
|
|||||||||
|---|---|---|---|---|---|---|---|---|---|
|
Paired Differences
|
t
|
Df
|
Sig. (2-tailed)
|
||||||
|
Mean
|
Std. Deviation
|
Std. Error Mean
|
95% Confidence Interval (CI) of the Difference
|
||||||
|
Lower
|
Upper
|
||||||||
|
Pair 1
|
High Vulnerability Population - Malaria2024
|
2968.88375
|
13891.08420
|
4911.23992
|
-8644.35327
|
14582.12077
|
.605
|
7
|
.565
|
|
Pair 2
|
Medium Vulnerability Population - Malaria2024
|
220190.43875
|
129558.29020
|
45805.77278
|
111876.99758
|
328503.87992
|
4.807
|
7
|
.002
|
|
Pair 3
|
Low Vulnerability Population - Malaria2024
|
58226.70625
|
36691.88070
|
12972.53883
|
27551.52633
|
88901.88617
|
4.488
|
7
|
.003
|
|
Pair 4
|
Total - Malaria 2024
|
297673.28125
|
163889.89324
|
57943.82744
|
160657.90166
|
434688.66084
|
5.137
|
7
|
.001
|
Pearson correlations between class-specific population counts and LGA case counts indicated heterogeneity across vulnerability strata. The low-vulnerability population correlated strongly and significantly with case counts (r = 0.914, p = 0.001; Benjamini–Hochberg-adjusted p = 0.003). The medium-vulnerability population showed a moderate but non-significant correlation (r = 0.618, p = 0.102; adjusted p = 0.153). The high-vulnerability population was uncorrelated with case counts (r = − 0.069, p = 0.870). Total population size correlated positively with cases (r = 0.719, p = 0.044), underscoring the scale dependence of absolute counts.
4. Discussion
The AHP results, which prioritised proximity to streams and wetlands alongside rainfall and microtopographic wetness, are epidemiologically coherent for a low-relief, estuarine delta where the Anopheles gambiae complex and brackish-tolerant An. melas exploit shallow, sunlit, and seasonally persistent water bodies (8, 53). Land surface temperature’s moderate weight aligns with the strong temperature dependence of vector development and Plasmodium sporogony, with transmission potential peaking at intermediate temperatures typical of the lowland tropics (9, 38). The comparatively smaller weights assigned to slope and NDVI reflect Bayelsa’s exceptionally low relief and the dominance of fine-scale hydrology and inundation over terrain steepness and broad greenness in shaping larval habitat suitability. Accessibility and built environment variables made modest contributions, appropriately reflecting their roles in modulating exposure, care-seeking, and reporting rather than intrinsic entomological hazard (10, 54).
The spatial patterning of the MVI shows a predominantly medium vulnerability with localised high-vulnerability pockets, which matches the heterogeneous malaria ecology of southern Nigeria and supports risk-stratified planning advocated by the National Malaria Elimination Programme and WHO (31, 55–57). LGAs such as Southern Ijaw, Brass, and Sagbama contain sizeable high-vulnerability areas, whereas Kolokuma/Opokuma, Ogbia, and Yenagoa retain substantial low-vulnerability extents despite statewide predominance of medium vulnerability. Population overlays underline that large absolute numbers of residents live in medium-vulnerability zones, which may serve as substantial reservoirs of transmission even when the ecological hazard is not extreme.
The empirical relationships between vulnerability strata and routine case counts require careful interpretation. The lack of correlation between high-vulnerability population and reported cases could reflect limited statistical power (only eight LGAs), classification mismatch with realised transmission in 2024, or reporting artefacts in hard-to-reach, riverine areas where barriers to diagnosis and care can attenuate the link between ecological vulnerability and routine surveillance (58). It is also plausible that targeted interventions prioritised to more vulnerable localities (e.g., LLIN campaigns, community case management, IPTp) reduced symptomatic case burdens relative to population at risk within the public reporting system (1, 31).
By contrast, the strong positive correlation between the number of residents in low-vulnerability areas and case counts likely reflects two mechanisms. First, urban and peri-urban settings are often classified as lower ecological vulnerability, yet can sustain appreciable transmission via man-made larval habitats, insecticide resistance, and indoor biting despite LLIN availability (59, 60). Second, better care access and reporting completeness in such settings strengthen the correspondence between population size and notified cases (56, 58). The positive correlation between total population and cases is epidemiologically expected because absolute counts scale with population size; this underscores the value of rate-based metrics or count models with population offsets when comparing LGAs (61, 62).
Methodologically, future work should relate LGA-level incidence (or case counts with a log-population offset) to the proportions of residents in each vulnerability class, along with climate covariates and intervention coverage, ideally using Poisson or negative binomial models with random effects to handle extra-Poisson variation and unmeasured heterogeneity (2, 61). Given the compositional nature of the three class proportions, an isometric log-ratio transform or the use of a referent class can prevent collinearity. The MVI itself could be refined by incorporating explicit indicators of health care access, diagnostic use, and reporting completeness as a second dimension of vulnerability, separate from ecological receptivity, to improve its ability to anticipate observed routine burdens (58).
Limitations include the temporal mismatch between 2020 denominators and 2024 cases, potential MAUP effects from LGA aggregation, the subjectivity inherent in AHP, and small-sample uncertainty. We mitigated these by aligning all raster operations to the Landuse/cover grid, using proportions where feasible, preserving fixed thresholds for class definitions, and verifying internal consistency of AHP judgments. Nonetheless, conclusions regarding associations with routine case counts should be considered preliminary and hypothesis-generating.
5. Conclusion
In Bayelsa State, vulnerability to malaria is overwhelmingly structured by hydrology and hydroclimate, with proximity to streams and wetlands, rainfall, and microtopographic wetness driving the MVI. The landscape is dominated by medium vulnerability, with localised high-vulnerability pockets concentrated in riverine LGAs. In routine 2024 data, absolute case counts rose with total population and correlated strongly with the number of residents in areas classified as low vulnerability, but not with those in high-vulnerability zones. These findings likely reflect a combination of ecological and health-system processes and the intrinsic scaling of counts with population size. Incorporating access and reporting indicators into the vulnerability framework and moving to rate-based or offset count models will yield more policy-relevant inference on how vulnerability classes map onto malaria risk, thereby strengthening subnational stratification and targeting in Bayelsa.
List of Abbreviations
MVI
Malaria Vunerability Index
LGA
Local Government Areas
MAUP
Modifiable area unit problem
AHP
Analytic Hierarchy Process
MCDA
Multi–Criteria Decision Analysis
NDVI
Normalized Difference Vegetation Index
GIS
Geographic Information System
CR
Consistency Ratio
CI
Consistency Index
CIs
Confidence Interval
WLC
Weighted Linear Combination
ITN/IRS
Insecticide Treatment Net/ Indoor Residual Spraying
LULC
Land use/ Land Cover
NASA
National Aeronautics and Space Administration
USGS
United States Geological Survey
SRTM
Shuttle Radar Topography Mission
DEM
Digital Elevation Model
TWI
Topographic Wetness Index
WHO
World Health Organization
LST
Land Surface Temperature
LLIN
Long Lasting Insecticide Treatment
IPTP
Intermittent Preventive Treatment in Pregnancy
Declarations
Ethics approval and consent to participate
Not applicable
Consent for publication
Not applicable
A
Data Availability
The datasets generated and/or analysed during this study are not publicly available due to ethical and data-sharing restrictions but are available from the corresponding author on reasonable request.
Competing interests
The authors declare that they have no competing interests
A
Funding
The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
A
Author Contribution
**OA** prepared and wrote the initial draft, performed and computed the analysis. **OJT** and **JOO** supported the draft preparation and the analysis and the modelling computations. **TB** and **GW** contributed to data curation, including maintaining the research data for initial use and later reuse **GE** , **DB, FO, COU** and **IE** critically reviewed and edited the manuscript. **CK** provided leadership of project design, supervision of project delivery, and supervisory authorship of our manuscript. All authors read and approved the final manuscript.
Acknowledgements
References
1.
Bhatt S, Weiss DJ, Cameron E, Bisanzio D, Mappin B, Dalrymple U, et al. The effect of malaria control on Plasmodium falciparum in Africa between 2000 and 2015. Nature. 2015;526(7572):207–11.
2.
Weiss DJ, Nelson A, Vargas-Ruiz CA, Gligorić K, Bavadekar S, Gabrilovich E, et al. Global maps of travel time to healthcare facilities. Nat Med. 2020;26(12):1835–8.
3.
Snow RW. Global malaria eradication and the importance of Plasmodium falciparum epidemiology in Africa. BMC Med. 2015;13(1):23.
4.
Snow RW, Guerra CA, Noor AM, Myint HY, Hay SI. The global distribution of clinical episodes of Plasmodium falciparum malaria. Nature. 2005;434(7030):214–7.
5.
Gething PW, Smith DL, Patil AP, Tatem AJ, Snow RW, Hay SI. Climate change and the global malaria recession. Nature. 2010;465(7296):342–5.
6.
Bousema T, Griffin JT, Sauerwein RW, Smith DL, Churcher TS, Takken W, et al. Hitting Hotspots: Spatial Targeting of Malaria for Control and Elimination. PLoS Med. 2012;9(1):e1001165.
7.
Hay SI, Guerra CA, Tatem AJ, Atkinson PM, Snow RW. Urbanization, malaria transmission and disease burden in Africa. Nat Rev Microbiol. 2005;3(1):81–90.
8.
Sinka ME, Bangs MJ, Manguin S, Coetzee M, Mbogo CM, Hemingway J, et al. The dominant Anopheles vectors of human malaria in Africa, Europe and the Middle East: occurrence data, distribution maps and bionomic précis. Parasit Vectors. 2010;3(1):117.
9.
Mordecai EA, Paaijmans KP, Johnson LR, Balzer C, Ben-Horin T, de Moor E, et al. Optimal temperature for malaria transmission is dramatically lower than previously predicted. Ecol Lett. 2013;16(1):22–30.
10.
Tusting LS, Bottomley C, Gibson H, Kleinschmidt I, Tatem AJ, Lindsay SW, et al. Housing Improvements and Malaria Risk in Sub-Saharan Africa: A Multi-Country Analysis of Survey Data. PLoS Med. 2017;14(2):e1002234.
11.
Gaievskyi S, Delfrate N, Ragazzoni L, Bahattab A. Use of multi-criteria decision analysis (MCDA) to support decision-making during health emergencies: a scoping review. Frontiers in Public Health. Volume 13. Frontiers Media SA; 2025.
12.
Băcescu Ene GV, Stoia MA, Cojocaru C, Todea DA. SMART Multi-Criteria Decision Analysis (MCDA)—One of the Keys to Future Pandemic Strategies. J Clin Med. 2025;14(6).
13.
Saaty TL. How to make a decision: The analytic hierarchy process. Eur J Oper Res. 1990;48(1):9–26.
14.
Malczewski J. GIS-based multicriteria decision analysis: a survey of the literature. Int J Geogr Inf Sci. 2006;20(7):703–26.
15.
Chen Y, Yu J, Khan S. Spatial sensitivity analysis of multi-criteria weights in GIS-based land suitability evaluation. Environ Model Softw. 2010;25(12):1582–91.
16.
Thokala P, Devlin N, Marsh K, Baltussen R, Boysen M, Kalo Z, et al. Multiple Criteria Decision Analysis for Health Care Decision Making—An Introduction: Report 1 of the ISPOR MCDA Emerging Good Practices Task Force. Value Health. 2016;19(1):1–13.
17.
Onyiri N. Estimating malaria burden in Nigeria: A geostatistical modelling approach. Geospat Health. 2015;10(2):163–70.
18.
Ebenezer A, Noutcha AEM, Agi PI, Okiwelu SN, Commander T. Spatial distribution of the sibling species of Anopheles gambiae sensu lato (Diptera: Culicidae) and malaria prevalence in Bayelsa State, Nigeria. Parasit Vectors. 2014;7(1).
19.
Funk C, Peterson P, Landsfeld M, Pedreros D, Verdin J, Shukla S et al. The climate hazards infrared precipitation with stations - A new environmental record for monitoring extremes. Sci Data. 2015;2.
20.
Wan Z. New refinements and validation of the MODIS Land-Surface Temperature/Emissivity products. Remote Sens Environ [Internet]. 2008 Jan 15 [cited 2025 Sep 29];112(1):59–74. Available from: https://www.sciencedirect.com/science/article/abs/pii/S0034425707003665
21.
Tucker CJ. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens Environ [Internet]. 1979 May 1 [cited 2025 Sep 29];8(2):127–50. Available from: https://www.sciencedirect.com/science/article/abs/pii/0034425779900130
22.
Karra K, Kontgis C, Statman-Weil Z, Mazzariello JC, Mathis M, Brumby SP, GLOBAL LAND USE/LAND COVER WITH SENTINEL 2 AND DEEP LEARNING. International Geoscience and Remote Sensing Symposium (IGARSS). 2021;2021-July:4704–7.
23.
Lehner B, Verdin K, Jarvis A. New global hydrography derived from spaceborne elevation data. Eos (Washington DC). 2008;89(10):93–4.
24.
OpenStreetMap [Internet]. [cited 2025 Sep 29]. Available from: https://www.openstreetmap.org/#map=6/9.12/8.67
25.
Home. - GRID3 [Internet]. [cited 2025 Sep 29]. Available from: https://grid3.org/
26.
Stevens LA, Behn MD, McGuire JJ, Das SB, Joughin I, Herring T, et al. Greenland supraglacial lake drainages triggered by hydrologically induced basal slip. Nature. 2015;522(7554):73–6.
27.
Tatem AJ. WorldPop, open data for spatial demography. Sci Data [Internet]. 2017 Jan 31 [cited 2025 Sep 29];4(1):1–4. Available from: https://www.nature.com/articles/sdata20174
28.
Vernon CR, Mongird K, Nelson KD, Rice JS. Harmonized geospatial data to support infrastructure siting feasibility planning for energy system transitions. Sci Data. 2023;10(1).
29.
Pogson M, Smith P. Effect of spatial data resolution on uncertainty. Environmental Modelling & Software [Internet]. 2015 Jan 1 [cited 2025 Sep 28];63:87–96. Available from: https://www.sciencedirect.com/science/article/abs/pii/S1364815214002850?via%3Dihub
30.
Stevens FR, Gaughan AE, Linard C, Tatem AJ. Disaggregating Census Data for Population Mapping Using Random Forests with Remotely-Sensed and Ancillary Data. PLoS ONE. 2015;10(2):e0107042.
31.
WHO. Report of the first and second meetings of the Technical Advisory Group on Malaria Elimination and Certification. World Health Organization. 2023; 2023.
32.
Caminade C, Kovats S, Rocklov J, Tompkins AM, Morse AP, Colón-González FJ et al. Impact of climate change on global malaria distribution. Proceedings of the National Academy of Sciences. 2014;111(9):3286–91.
33.
Githeko AK, Lindsay SW, Confalonieri UE, Patz JA. Climate change and vector-borne diseases: a regional analysis. Bull World Health Organ Special Theme– Environ Health. 2000;78(9):1136–47.
34.
Keiser J, Jürg Utzinger MC, De Castro TA, Smith M, Tanner, Burton H. Singer. Urbanization in Sub-Saharan Africa and Implication for Malaria Control. The Intolerable Burden of Malaria II: What’s New, What’s Needed: Supplement to Volume 71 (2) of the American Journal of Tropical Medicine and Hygiene. 2nd ed. American Society of Tropical Medicine and Hygiene, Northbrook (IL); 2004.
35.
Kibret S, Wilson GG, Ryder D, Tekie H, Petros B. Malaria impact of large dams at different eco-epidemiological settings in Ethiopia. Trop Med Health. 2017;45(1):4.
36.
Wilson JP, Gallant JC. Terrain analysis : principles and applications. Wiley; 2000. p. 479.
37.
Ceccato P, Connor SJ, Jeanne I, Thomson MC. Application of Geographical Information Systems and Remote Sensing technologies for assessing and monitoring malaria risk. Parassitologia. 2005;47:81–96.
38.
Paaijmans KP, Read AF, Thomas MB. Understanding the link between malaria risk and climate. Proceedings of the National Academy of Sciences. 2009;106(33):13844–9.
39.
Shapiro LLM, Whitehead SA, Thomas MB. Quantifying the effects of temperature on mosquito and parasite traits that determine the transmission potential of human malaria. PLoS Biol. 2017;15(10):e2003489.
40.
Tucker CJ. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens Environ. 1979;8(2):127–50.
41.
Guagliardo MF. Spatial accessibility of primary care: concepts, methods and challenges. Int J Health Geogr. 2004;3(1):3.
42.
Jha AKumar B, Robin. Lamond Jessica. Cities and flooding : a guide to integrated urban flood risk management for the 21st century. World Bank; 2012. p. 631.
43.
Noor AM, Zurovac D, Hay SI, Ochola SA, Snow RW. Defining equity in physical access to clinical services using geographical information systems as part of malaria planning and monitoring in Kenya. Tropical Med Int Health. 2003;8(10):917–26.
44.
Munga S, Minakawa N, Zhou G, Mushinzimana E, Barrack OOJ, Githeko AK, et al. Association Between Land Cover and Habitat Productivity of Malaria Vectors in Western Kenya Highlands. Am J Trop Med Hyg. 2006;74(1):69–75.
45.
Atieli HE, Zhou G, Lee MC, Kweka EJ, Afrane Y, Mwanzo I, et al. Topography as a modifier of breeding habitats and concurrent vulnerability to malaria risk in the western Kenya highlands. Parasit Vectors. 2011;4(1):241.
46.
Krefis AC, Schwarz NG, Krüger A, Fobil J, Nkrumah B, Acquah S, et al. Modeling the Relationship between Precipitation and Malaria Incidence in Children from a Holoendemic Area in Ghana. Am Soc Trop Med Hygiene. 2011;84(2):285–91.
47.
Forman EH, Gass SI. The Analytic Hierarchy Process—An Exposition. Oper Res. 2001;49(4):469–86.
48.
Ishizaka A, Labib A. Review of the main developments in the analytic hierarchy process. Expert Syst Appl. 2011.
49.
Saaty TL. In. Fundamentals of the Analytic Hierarchy Process. 2001. pp. 15–35.
50.
Moran PAP. Notes on Continuous Stochastic Phenomena. Biometrika. 1950;37(1/2):17.
51.
Waller LanceA, Gotway CarolA. Applied Spatial Statistics for Public Health Data. Wiley; 2004.
52.
Kaduru C, Eshikhena G, Abe E, Ojielo N, Aworabhi N, Masa H et al. The impact of strengthened ward development committees on utilisation of reproductive maternal and child health services in Bayelsa State, Nigeria. BMC Public Health [Internet]. 2025 Dec 1 [cited 2025 Sep 25];25(1):1–14. Available from: https://bmcpublichealth.biomedcentral.com/articles/10.1186/s12889-025-23304-z
53.
Coetzee M, Craig M, le Sueur D. Distribution of African Malaria Mosquitoes Belonging to the Anopheles gambiae Complex. Parasitol Today. 2000;16(2):74–7.
54.
Noor A, Aftab A, Aslam M, Imanpour S. Household heating fuels impact on Acute Respiratory Infection (ARI) symptoms among children in Punjab, Pakistan. BMC Public Health. 2023;23(1):2380.
55.
FMoH NMEP. National Malaria Strategic Plan 2021–2025. Oct: Abuja; 2020.
56.
NMEP, NPC. ICF. Nigeria Malaria Indicator Survey 2021 Final Report. Abuja; 2022.
57.
WHO. Global technical strategy for malaria, 2016–2030. Global Malaria Programme. World Health Organization; 2021. p. 29.
58.
Alegana VA, Wright JA, Pentrina U, Noor AM, Snow RW, Atkinson PM. Spatial modelling of healthcare utilisation for treatment of fever in Namibia. Int J Health Geogr. 2012;11(1):6.
59.
Hancock PA, Hendriks CJM, Tangena JA, Gibson H, Hemingway J, Coleman M, et al. Mapping trends in insecticide resistance phenotypes in African malaria vectors. PLoS Biol. 2020;18(6):e3000633.
60.
Tatem AJ, Gething PW, Smith DL, Hay SI. Urbanization and the global malaria recession. Malar J. 2013;12(1):133.
61.
Hilbe JM. Negative binomial regression. Cambridge University Press; 2011. p. 553.
62.
Morgenstern H. Ecologic Studies in Epidemiology: Concepts, Principles, and Methods. Annu Rev Public Health. 1995;16(1):61–81.