GIS-Driven Multi-Criteria Assessment of Rural Settlement Patterns and Attributes in Rwanda’s Western Highlands (Central Africa)

Abstract

This study investigates rural settlement patterns and land suitability in Rwanda’s Western Highlands, a mountainous region highly vulnerable to geohazards like landslides and flooding. Its primary aim is to inform sustainable, climate-resilient development planning in this fragile landscape. We employed high-resolution satellite imagery, a Digital Elevation Model (DEM), and comprehensive geospatial datasets to analyze settlement distribution, using Thiessen polygons for influence zones and Kernel Density Estimation (KDE) for spatial clustering. The Analytic Hierarchy Process (AHP) was integrated with the GeoDetector model to objectively weight criteria and analyze settlement pattern drivers, using population density as a proxy for human pressure. The analysis revealed significant spatial heterogeneity in settlement distribution, with both clustered and dispersed forms exhibiting distinct exposure levels to environmental hazards. Natural factors, particularly slope gradient and proximity to rivers, emerged as dominant determinants. Furthermore, significant synergistic interactions were observed between environmental attributes and infrastructure accessibility (roads and urban centers), collectively shaping settlement resilience. This integrative geospatial approach enhances understanding of complex rural settlement dynamics in ecologically sensitive mountainous regions. The empirically grounded insights offer a robust decision-support framework for climate adaptation and disaster risk reduction, contributing to more resilient rural planning strategies in Rwanda and similar Central African highland regions.

1. Introduction

The strategic organization of rural settlements and sustainable land use planning are increasingly critical for economic development, particularly in regions where agriculture forms the backbone of livelihoods [1,2,3]. In Rwanda, a land-scarce and rapidly developing nation, rural transformation is essential for achieving long-term socio-economic objectives—particularly those outlined in Rwanda’s Vision 2050 [4] and the United Nations Sustainable Development Goals (SDGs), notably SDG 11 (Sustainable Cities and Communities) and SDG 13 (Climate Action) [5,6]. This study aims to provide spatially informed insights for guiding rural development in alignment with these national and international frameworks.

Rwanda’s Western Highlands, located in Central Africa, represent a geologically and socio-economically sensitive region [7]. Characterized by steep slopes, high rainfall, and fragile volcanic and metamorphic substrates, this mountainous zone faces persistent risks of soil erosion, landslides, and flash flooding during its biannual wet seasons [8,9]. Human activities such as terracing, road construction, and land clearing increasingly disrupt the natural hydrological balance, exacerbating vulnerability to both geohazards and anthropogenic surface pollution [10,11]. Despite these constraints, the region has experienced rapid growth in rural settlements and infrastructure, largely driven by national policies promoting economic diversification and global integration [12].

Research on rural settlement patterns in mountainous African regions emphasizes terrain and proximity to infrastructure—such as roads and rivers—as key determinants of settlement location [13,14,15]. While such proximity enhances access and agricultural productivity, it also elevates exposure to environmental hazards [16,17,18,19,20,21]. Previous suitability assessments have often relied on simple overlay techniques or expert judgment, with limited consideration of spatial heterogeneity or the interaction between environmental and infrastructural factors [22,23,24]. Consequently, many past interventions in rural development have failed to meet resilience and sustainability goals [25,26]. To address these challenges, this study analyzes rural settlement suitability in Rwanda’s Western Highlands by examining the complex interactions between environmental constraints, infrastructural access, and climate-related risks, thereby supporting policy implementation under frameworks such as Rwanda’s National Land Use and Development Master Plan (RNLUDMP, 2020) and Vision 2050 [26,27,28,29,30,31,32].

To guide the investigation, this study addresses the following research questions:

• What are the spatial distribution characteristics and typologies of rural settlements in Rwanda’s Western Highlands (Central Africa)?
• How can suitability indicators be systematically selected, graded, and weighted to reflect their roles in promoting climate-resilient development?
• What relationships exist between settlement distribution patterns and suitability evaluations?
• How can integrated spatial and statistical techniques facilitate adaptive rural development planning in mountainous regions?

2. Materials and Methods

2.1. Study Area

This study focuses on the Western Highlands of Rwanda, encompassing an area of approximately 5882 km² within the Western Province. Geographically, the region is situated between latitudes 2°22′18″ S and longitudes 29°12′21″ E, covering seven administrative districts: Karongi, Ngororero, Nyabihu, Nyamasheke, Rubavu, Rusizi, and Rutsiro. The landscape is characterized by steep slopes and rugged terrain, with elevations spanning from 900 m to over 4000 m. This varied topography, coupled with a tropical highland climate that receives annual precipitation exceeding 1200 mm, contributes to high susceptibility to environmental hazards, including landslides, flash floods, and severe soil erosion. The region’s hydrographic network, composed of numerous small rivers and streams, contributes to rapid runoff during heavy rainfall events. The underlying geological formations, predominantly volcanic and metamorphic, further contribute to terrain instability. The natural landscape’s response to rainfall events, amplified by human interventions such as extensive terracing, road construction on unstable slopes, and historical deforestation, often results in altered runoff patterns and increased vulnerability to these hazards, including surface and potential deep anthropogenic pollution.

The study area is predominantly rural, supporting an estimated population of 2.9 million people, which translates to a high population density of approximately 490 persons per km². This intense human pressure on a topographically complex and environmentally fragile landscape creates a socio-environmental context highly vulnerable to climatic and geomorphological risks. These characteristics necessitate the development of a comprehensive, spatially explicit framework for assessing settlement suitability that promotes climate resilience and sustainable rural development within this delicate mountainous environment (Figure 1).

2.2. Data Sources and Preprocessing

This study employed a robust spatial analysis framework, integrating multi-source geospatial and socio-economic datasets from reputable public repositories and institutional sources. All spatial layers were standardized to the WGS 1984 UTM Zone 35S coordinate system (EPSG: 32735) to ensure geospatial consistency and analytical compatibility. Table 1 provides a concise summary of the key datasets, detailing their format, spatial resolution, sources, and applications within the research.

Remote sensing data, specifically Landsat 8 Operational Land Imager (OLI) imagery (Path 171, Row 65) acquired on 5 December 2020, at a 30 m spatial resolution, was obtained from the USGS EarthExplorer archive. This imagery was crucial for the supervised classification of land use and land cover [33]. Furthermore, a 30 m resolution Digital Elevation Model (DEM), also sourced from USGS [34], yielded essential topographic variables, including elevation, slope, aspect, and terrain ruggedness. These terrain indicators were vital for delineating geomorphic constraints on rural settlement distribution, consistent with prior research [35,36].

Hydrological and transportation networks, encompassing rivers and roads, were acquired as vector line features from Bigemap. These datasets facilitated proximity and buffer analyses, which were instrumental in assessing flood risk and accessibility [37,38,39,40]. All data underwent rigorous preprocessing and cleaning using ArcGIS 10.7 software, which was subsequently utilized for comprehensive spatial transformation, classification, and modeling operations throughout the analysis [41].

Rural settlement locations, collected as point features from Bigemap and rigorously validated through targeted field surveys in 2021–2022 [42], formed the foundation for Thiessen polygon (Voronoi) generation and Kernel Density Estimation (KDE), thereby enabling spatial characterization of settlement dispersion and clustering. Administrative centers and urban nodes, specifically towns and sector headquarters, were derived from Bigemap vector datasets (polygon format) to delineate zones of urban influence [43].

The 2020 national census population density data, provided by the Rwanda National Institute of Statistics (NISR) and structured as polygon-based administrative units, were incorporated as a continuous proxy for demographic intensity and service demand. Despite the absence of more granular income, employment, and land tenure data, population density remains a widely recognized and policy-relevant indicator in spatial planning and vulnerability studies [44].

Table 1. Summary of datasets used in the study.

Data Type	Dataset Description	Format	Source	Spatial Resolution	Application	Software
Remote Sensing	Landsat 8 OLI (5 December 2020, Path 171, Row 65)	Raster	USGS EarthExplorer [33]	30 m	Land use/land cover classification	ArcGIS 10.7
Topography	DEM	Raster	USGS EarthExplorer [34]	30 m	Elevation, Slope, Aspect, Ruggedness	ArcGIS 10.7
Hydrology	River network	Line	Bigemap [37,38]	Vector (1:50,000 scale)	River Proximity analysis (flood risk)	ArcGIS 10.7
Transportation	Road network	Line	Bigemap [39,40]	Vector (1:50,000 scale)	Accessibility analysis	ArcGIS 10.7
Settlement Locations	Rural settlements (field-verified)	Point	Bigemap + Field Survey [42]	Point	Spatial analysis (Thiessen polygons, KDE)	ArcGIS 10.7
Urban Centers	Towns and sector headquarters	Polygon	Bigemap [45]	Polygon (Administrative level)	Urban influence zones	ArcGIS 10.7
Population density	2020 national census by administrative sector	Polygon	Rwanda National Institute of Statistics	Polygon (Sector level)	Proxy for socio-economic pressure	ArcGIS 10.7

Table 1. Summary of datasets used in the study.

Data Type	Dataset Description	Format	Source	Spatial Resolution	Application	Software
Remote Sensing	Landsat 8 OLI (5 December 2020, Path 171, Row 65)	Raster	USGS EarthExplorer [33]	30 m	Land use/land cover classification	ArcGIS 10.7
Topography	DEM	Raster	USGS EarthExplorer [34]	30 m	Elevation, Slope, Aspect, Ruggedness	ArcGIS 10.7
Hydrology	River network	Line	Bigemap [37,38]	Vector (1:50,000 scale)	River Proximity analysis (flood risk)	ArcGIS 10.7
Transportation	Road network	Line	Bigemap [39,40]	Vector (1:50,000 scale)	Accessibility analysis	ArcGIS 10.7
Settlement Locations	Rural settlements (field-verified)	Point	Bigemap + Field Survey [42]	Point	Spatial analysis (Thiessen polygons, KDE)	ArcGIS 10.7
Urban Centers	Towns and sector headquarters	Polygon	Bigemap [45]	Polygon (Administrative level)	Urban influence zones	ArcGIS 10.7
Population density	2020 national census by administrative sector	Polygon	Rwanda National Institute of Statistics	Polygon (Sector level)	Proxy for socio-economic pressure	ArcGIS 10.7

2.2.1. Remote Sensing and Topographic Data

Satellite imagery from the Landsat 8 Operational Land Imager (OLI), specifically Path 171, Row 65, acquired on 5 December 2020, was chosen due to its optimal spatial resolution (30 m) and minimal cloud cover (less than 5%), rendering it highly suitable for accurate land cover and rural settlement classification. This imagery, retrieved from the United States Geological Survey (USGS) EarthExplorer platform, underwent standard preprocessing procedures, including atmospheric correction and radiometric calibration, to ensure data quality and consistency [46].

Complementary topographic data were derived from a 30 m DEM, also sourced from the USGS EarthExplorer. From this DEM, critical terrain attributes—including elevation, slope, aspect, and terrain ruggedness—were extracted [47,48]. These derived terrain indicators provided essential inputs for assessing terrain stability and informed slope-sensitive rural settlement planning within the study area.

2.2.2. Demographic and Settlement Distribution Data

Vector datasets delineating river networks, road infrastructure, and rural settlement locations were acquired from authoritative geospatial repositories to ensure data accuracy and consistency [49,50]. Population density, aggregated at the administrative sector level, was utilized as a proxy for development intensity and service demand [51,52]. This metric provided a spatially continuous and policy-relevant indicator of human presence and associated land use pressure, thereby integrating socioeconomic influence into the spatial analysis [53]. It is important to note that more granular economic variables, such as income levels, employment categories, and land tenure patterns, were not uniformly available at the requisite spatial resolution for consistent integration across the entire study area.

Future research endeavors should prioritize the acquisition and integration of more detailed, spatially disaggregated socioeconomic datasets. Such an enhancement would significantly improve the robustness and policy relevance of spatial suitability models, particularly within data-scarce rural regions like Rwanda’s Western Highlands. Despite the availability of high-resolution satellite imagery, settlement boundaries were represented as point data due to challenges posed by temporal inconsistencies, variability in image resolution, and computational limitations across the expansive and topographically complex study area.

The inherently dispersed and irregular settlement patterns characteristic of the Western Highlands further complicated the automated extraction of precise building footprints or residential polygons. Consequently, a point-based representation was adopted as a standardized and computationally efficient method for effectively capturing settlement distribution patterns across a large geographic extent.

2.2.3. Data Preprocessing and Validation

To ensure data quality and consistency, all raw spatial datasets were standardized to the WGS 1984 UTM Zone 35S coordinate reference system (EPSG: 32735). Landsat 8 imagery underwent comprehensive preprocessing, including radiometric calibration to correct sensor-induced distortions, geometric correction, and multi-spectral band registration using ArcGIS 10.7 [46]. Subsequent land use and land cover classification was conducted through manual interpretation tailored to capture Rwanda-specific landscape features, and validated through field surveys.

Classification accuracy was assessed using overall accuracy and the Kappa coefficient, based on comparisons with ground truth data collected through field surveys. This evaluation was used to ensure agreement between the classified imagery and actual land cover patterns, justifying its use in the suitability analysis [54].

Topographic variables, such as slope, elevation, aspect, and terrain ruggedness, were derived from the DEM employing ArcGIS 10.7 Spatial Analyst tools. Key parameters including slope and elevation were reclassified into discrete suitability classes based on established ecological and geomorphological thresholds pertinent to landslide susceptibility and agricultural viability [9,55]. This classification framework enabled precise delineation of site suitability, ensuring that critical environmental constraints were appropriately incorporated into the spatial modeling process.

Overall, the preprocessing and validation steps established a robust spatial dataset foundation critical to the integrity of the analysis. The combination of rigorous image calibration, targeted land cover classification, and ecologically informed topographic reclassification underpinned the generation of reliable and contextually relevant layers necessary for accurate site suitability assessment in the study area.

2.3. Analytical Framework

This study adopts a GIS-based multi-criteria evaluation framework designed to produce robust, interpretable, and policy-relevant outputs. The framework combines expert-driven indicator weighting using the Analytic Hierarchy Process (AHP) with empirical spatial statistical analysis through the GeoDetector model. By integrating the subjective insights of domain experts with data-driven evaluation of spatial heterogeneity and factor interactions, this hybrid approach enhances both the reliability and validity of rural settlement suitability assessments [43]. Figure 2 illustrates the overall methodological workflow, highlighting the systematic integration of these complementary methods to deliver a comprehensive evaluation of settlement suitability within the study area.

Selected suitability criteria operationalize climate resilience by incorporating slope and elevation parameters to mitigate landslide risks, buffered proximities to rivers and roads to capture flood exposure and accessibility constraints, and distances to towns proxying adaptive capacities associated with service availability. This multidimensional approach enables nuanced assessments of rural settlement resilience under combined climatic and demographic pressures [56].

2.3.1. Environmental Factors and Suitability Grading Scheme

Environmental suitability was systematically evaluated by reclassifying key physical variables (elevation, slope) and performing buffer analyses relative to rivers, roads, and urban centers. These parameters were selected following the integrated multi-criteria evaluation framework due to their combined influence on rural settlement distribution and climate resilience, as supported by prior studies [57].

Table 2. Environmental factors and grading scheme used for suitability assessment.

Environmental Factor	Analysis Method	Reclassification and Buffer Level
Elevation (E)	Reclassification	<1000 m, 1000–1500 m, 1500–2000 m, >2000 m
Slope (S)	Reclassification	<10°, 10–20°, 20–30°, >30°
Major Road (MR)	Buffer Analysis	<2 km, 2–5 km, 5–10 km, >10 km
River (R)	Buffer Analysis	<2 km, 2–4 km, 4–6 km, 6–8 km, >8 m
Towns/Urban centers (T)	Buffer Analysis	0–1000 m, 1000–2000 m, 2000–3000 m, 3000–4000 m, >4000 m

summarizes the classification schemes and buffer zones applied to assess the spatial suitability gradients across Rwanda’s Western Highlands.

2.3.2. Thiessen Polygon Analysis

Thiessen polygons (also known as Voronoi diagrams) were generated around rural settlement points to partition the study area into discrete spatial units, each representing the zone of influence of a single settlement. This method has been widely applied in spatial analysis to delineate areas of influence for discrete points [43]. Unlike conventional administrative boundaries, Thiessen polygons are defined solely by spatial proximity, providing a data-driven framework for analyzing settlement clustering and spatial distribution across the heterogeneous terrain of the Western Highlands.

To quantify the spatial heterogeneity of settlement distribution, the coefficient of variation (CV) of polygon areas was computed using the following formula:

$$CV={\displaystyle \frac{\sigma }{\mu }}$$

where $\sigma$ is the standard deviation and $\mu$ is the mean area of the Thiessen polygons. This normalized metric enables comparative assessment of settlement compactness and dispersion across different subregions, serving as a foundational input for evaluating variability in settlement patterns. By integrating Thiessen polygon delineation with CV analysis, the study captures essential spatial dynamics that inform how settlements influence land use and vulnerability to environmental hazards, underpinning spatially explicit assessments foundational to climate-resilient rural development planning in the region.

2.3.3. Kernel Density Estimation

KDE was applied to produce a continuous spatial surface representing the intensity of rural settlements across the study area. This non-parametric smoothing technique, commonly used for visualizing point patterns, employed a bandwidth of 1.5 km, determined based on the observed spatial scale of local settlement patterns. The resulting density surface effectively identifies clusters of settlements with heightened concentration, which are of particular interest for assessing exposure to climate-related hazards such as flooding and landslides. KDE computations were performed using ArcGIS 10.7 [16,43].

The derived density surfaces were validated against independent population density data to ensure spatial accuracy and robustness, thereby bolstering confidence in using KDE outputs to delineate meaningful settlement hotspots [58]. The integration of KDE with complementary spatial analyses provides a thorough characterization of settlement distribution dynamics, which is essential for informing targeted vulnerability assessments and advancing climate-resilient rural development strategies in the Western Highlands

2.3.4. GeoDetector for Spatial Factor Analysis

The GeoDetector method was employed to quantitatively assess the explanatory power of key environmental and socio-economic factors on the spatial variability of rural settlement distribution. This approach is particularly well-suited for complex, non-linear landscapes as it evaluates the strength of association between the dependent variable—in this case, settlement density—and each independent factor, without presupposing linear relationships. Additionally, GeoDetector facilitates the analysis of interaction effects among variables, offering deeper insights into the combined influence of multiple factors on settlement patterns. Its application in similar environmental studies has proven effective in understanding spatial drivers [59].

Central to GeoDetector is the q-statistic, which measures the proportion of spatial variance in settlement density explained by a given factor [60]. The q-statistic is mathematically defined as follows:

$$q=1-{\displaystyle \frac{{\displaystyle {\sum }_{h=1}^L{N}_h}{{\partial }^2}_h}{N{\partial }^2}}$$

where $L$ is the number of strata or categories within the factor, $N_h$ and ${{\partial }^2}_h$ represent the sample size and variance of settlement density within stratum h, respectively, and $N$ and ${\partial }^2$ denote the total sample size and variance across the entire study area. The q values range from 0 to 1, with higher values indicating stronger explanatory power of the factor in accounting for spatial variations in settlement distribution [61].

For this analysis, all explanatory variables, including elevation, slope, proximity to roads and rivers, and population density, were rasterized at a 30 m spatial resolution to enable granular, spatially explicit assessment. The GeoDetector results elucidated not only the relative importance of individual factors but also revealed significant interactions between variables, thereby providing a comprehensive understanding of the determinants shaping rural settlement patterns within Rwanda’s Western Highlands.

2.3.5. Integrated AHP and GeoDetector Weighting Scheme

To derive robust and objective weights for the suitability factors, this study employed an integrated weighting scheme combining expert-based AHP assessments with empirical spatial influence measures from GeoDetector analyses. AHP is a well-established multi-criteria decision-making method used for structuring complex decisions and determining criteria weights through pairwise comparisons. Experts conducted pairwise comparisons of the criteria within the AHP framework, with consistency ratios (CR) rigorously evaluated to ensure reliability (CR < 0.1), aligning with common practice in AHP applications [43]. To mitigate the inherent subjectivity of AHP and reinforce data-driven rigor, the GeoDetector-derived q-values were normalized and quantitatively integrated with the AHP weights through a weighted average approach:

$$W_i=\alpha .W_{AHPi}+(1-\alpha ).W_{Geoi}$$

where W_i represents the final combined weight for factor i, W_AHPi and W_Geoi, denote the respective normalized weights from AHP and GeoDetector, and $\alpha$ is an empirically determined coefficient (set to 0.5) balancing expert judgment and empirical evidence. This hybrid-weighting scheme was applied at the spatial polygon level, thereby enhancing the precision of spatial decision-making and ensuring the policy relevance of rural settlement suitability assessments within Rwanda’s Western Highlands. The integration of complementary weighting sources strengthens the validity and applicability of the resulting suitability models [62].

Consistent with the GeoDetector findings (Section 3.3), natural factors were assigned predominant influence, collectively accounting for 87.5% of the total weight, underscoring their critical role in shaping settlement patterns and resilience. Among these, slope (X₂) was the most significant criterion with a weight of 66.16%, reflecting its overriding impact on site stability and accessibility. Distance from rivers (X₃) and elevation (X₁) contributed 13.87% and 7.47%, respectively. Locational factors including distance from towns (X₅, 9.38%) and roads (X₄, 3.13%) comprised the remaining 12.5%. This weighting scheme leverages empirical evidence and expert judgment to enhance the robustness of the suitability model (

Table 3. Weights assigned to evaluation criteria and sub-factors in AHP-based suitability assessment.

Criterion Layer	Weights	Sub-Factor	Weight
Natural Factors	0.875	Elevation (X1)	0.0747
		Slope (X2)	0.6616
		Distance from Rivers (X3)	0.1387
Location Factors	0.125	Distance from Roads (X4)	0.0313
		Distance from Towns (X5)	0.0938

). It is important to note that the weights used in the suitability analysis were derived exclusively through the AHP method. The GeoDetector results served as a validation tool, providing empirical support for the AHP weighting scheme but were not averaged or combined with the AHP weights.

2.3.6. Suitability Evaluation and Classification

Suitability scores for each Thiessen polygon, representing discrete spatial units of rural settlement influence, were computed by aggregating weighted and normalized indicator values within ArcGIS 10.7 [63]. The composite suitability score S was calculated as follows:

$$S={\displaystyle \sum _{i=1}^n{w}_i}.X_i$$

where $w_i$ denotes the integrated weight of the ith factor derived from the combined AHP–GeoDetector approach, X_i represents the normalized and reclassified score of the corresponding factor, and n is the total number of factors considered. This weighted summation approach enables a quantitative synthesis of multiple criteria to robustly delineate spatial variation in settlement suitability [62,64].

The resulting suitability scores were stratified into five discrete classes: Highly Suitable, Suitable, Relatively Suitable, Basically Suitable, and Unsuitable for Settlement. These classification thresholds were informed by recognized international and region-specific standards on land use and climate resilience to ensure contextual relevance and applicability. The suitability grading framework is detailed in

Table 4. Suitability classification framework based on cumulative AHP scores.

Grade	Description	Score
Highly Suitable	Priority for settlement	4
Suitable	Suitable for settlement	3
Relatively Suitable	Moderate development effort	2
Basically Suitable	Considerable development effort	1
Unsuitable	Not recommended for settlement	0

.

To validate the classification outcomes, the spatial distribution of suitability classes was cross-verified using field observations, as well as existing land use and settlement pattern datasets [63]. This validation underscored the reliability and practical significance of the assessment in guiding climate-resilient rural settlement planning within Rwanda’s Western Highlands.

2.3.7. Sensitivity Analysis of Suitability Weights

To evaluate the robustness of the suitability model to variations in factor weights, a sensitivity analysis was conducted. The weights obtained from the integrated AHP–GeoDetector approach were systematically perturbed within a ± 10% range for each criterion, with the constraint that the total sum of weights remained fixed at 1. A one-at-a-time (OAT) approach was employed, whereby the weight of each factor was individually increased or decreased, and the resultant effects on suitability scores and classification distributions were quantified. Spatial analysis of the variations in settlement suitability classifications allowed identification of areas exhibiting high sensitivity to changes in weight values. Factors inducing significant shifts in suitability outcomes from minor weight adjustments were characterized as highly influential within the model framework [24,65]. This sensitivity assessment enhances the credibility of the model by demonstrating its stability and pinpointing factors where weight calibration may improve predictive accuracy.

2.4. Statistical Validation with Normal Probability Plots

To confirm the spatial distribution trends of rural settlements in relation to environmental factors, normal probability plots were utilized. This statistical technique is employed to visually assess whether a dataset follows a normal distribution. In this study, normal probability plots were generated for settlement proximity to rivers, elevation zones and slope zones, roads, and urban centers [36].

The plots illustrate the cumulative distribution of settlement counts against the expected cumulative distribution of a normal dataset. A linear pattern on the plot suggests that the observed settlement distribution generally conforms to a normal distribution in relation to the factor. Deviations from linearity indicate non-normal patterns, which can highlight specific environmental influences. This method provides a clear visual confirmation of the observed spatial patterns, reinforcing the quantitative analyses presented in the results [66].

3. Results

This section presents the findings from the spatial distribution analyses, the GeoDetector model, and the AHP-based suitability evaluation, detailing the characteristics and influencing factors of rural settlements in Rwanda’s Western Highlands.

3.1. Spatial Distribution Characteristics of Rural Residential Areas

3.1.1. Point-Based Spatial Analysis Using Thiessen Polygons

The spatial configuration of rural settlements in Rwanda’s Western Highlands was systematically analyzed through Thiessen polygons, which delineate the zone of influence around each settlement point, complemented by the calculation of the coefficient of variation (CV) derived from Voronoi diagrams (Figure 3a). The overall mean CV value of 23.0% across the study area reflects moderate variability in settlement spacing, indicating a heterogeneous pattern of rural habitation influenced by complex environmental and socio-economic factors.

At the district level, the analysis revealed significant disparities in settlement distribution. Rubavu District exhibited a notably high CV of 0.7491, characteristic of concentrated settlement clusters, whereas Ngororero District showed a markedly lower CV of 0.2629, consistent with a more uniform and dispersed settlement pattern. Settlements with CV values below 0.33, such as Bwira and Nyabirasi, were classified as randomly distributed, while those surpassing 0.64, including Gisenyi and Rugerero, were identified as clustered (Figure 3b). Collectively, 14.29% of settlements demonstrated clustering, with 7.14% exhibiting strong aggregation, highlighting the spatial diversity within the highlands.

These spatial patterns are closely linked to the region’s fragmented terrain and limited infrastructure networks, conditions that promote random settlement dispersion and elevate vulnerability to environmental hazards such as landslides and flooding. The distinct distribution typologies observed underscore the critical need for spatially nuanced, hazard-informed rural planning strategies aimed at enhancing community resilience. These findings align with prior studies emphasizing the interplay between topography, infrastructure, and vulnerability in mountainous rural contexts [67,68,69,70].

Although terrain ruggedness served as the primary proxy for assessing geohazard susceptibility in this study, we recognize that a more comprehensive risk evaluation requires incorporation of additional factors such as lithology, vegetation cover, and land use patterns. The exclusion of these variables was necessitated by limitations in the spatial availability and resolution of relevant datasets across the entire study area. Future research should integrate these environmental parameters to refine hazard modeling and enhance the robustness of rural settlement planning under geohazard risk.

3.1.2. Kernel Density Analysis of Rural Settlement Areas

KDE, conducted using a 4 km search radius, revealed significant spatial heterogeneity in the concentration of rural settlements across Rwanda’s Western Highlands (Figure 4). High-density clusters were predominantly observed in the northeastern and central subregions, notably within Rubavu District surrounding Gisenyi Town, where population densities approached approximately 2.5 persons per km². In contrast, low-density settlement zones, such as Ruharambuga in Nyamasheke District, exhibited densities near 0.6 persons per km², largely attributable to restrictive topographic conditions and limited infrastructural connectivity.

This pronounced spatial variability holds important implications for resilience-oriented rural planning. Areas of concentrated settlements may benefit from enhanced efficiency in service provision and collective resource management, whereas dispersed, low-density zones face increased challenges related to accessibility, service delivery, and heightened vulnerability to environmental hazards. These patterns align with established rural resilience frameworks, emphasizing the critical interplay between settlement density, infrastructure, and hazard exposure [71,72].

The KDE results, integrated with population density mapping, provide a granular spatial foundation critical for designing targeted interventions that optimize climate-resilient development. By delineating hotspots of rural settlement concentration alongside structurally vulnerable low-density areas, this analysis informs policymakers and planners in prioritizing resource allocation and adaptive infrastructure investments to sustainably support the diverse settlement typologies of the Western Highlands.

3.2. Influence of Environmental Factors on Settlement Distribution

3.2.1. Influence of Proximity to Rivers

Buffer zone analysis indicates that approximately 85% of rural settlements lie within 6 km of rivers, with nearly 45% located within 2 km (Figure 5a). Rivers provide critical water resources essential for domestic use and agriculture; however, their proximity in steep and rugged terrain significantly increases flood risk, posing a dual challenge for rural settlements. This trade-off between resource accessibility and hazard exposure underscores the necessity for integrating floodplain zoning and protective infrastructure into planning frameworks to enhance climate resilience [67,69]. The settlement proximity trends to rivers were confirmed through normal probability plotting (Figure 5b), reinforcing the observed spatial patterns.

3.2.2. Influence of Elevation and Slope

Elevation and slope significantly influence settlement patterns due to their effects on agricultural potential, accessibility, and hazard exposure. Approximately 80% of settlements are located below 1500 m elevation, primarily between 1000 and 1500 m, where climatic and soil conditions support farming and habitation. Settlements above 2000 m are sparse, reflecting constraints posed by rugged terrain and a harsher climate (Figure 6a,b).

The slope analysis shows that over 70% of settlements occupy slopes less than 20°, with nearly half on slopes under 10°, correlating with reduced landslide and erosion susceptibility and supporting community resilience (Figure 6c,d). Although slope serves as a practical initial indicator of landslide risk, it does not encompass the full range of controls influencing hazard occurrence. Factors including soil characteristics, lithology, land use, vegetation cover, and hydrological conditions are critical determinants of landslide dynamics [73,74]. The absence of consistent, high-resolution spatial data for these variables across the study region limited their incorporation in this analysis. Future work should integrate these environmental parameters to improve landslide risk assessment accuracy in the Western Highlands.

3.2.3. Influence of Proximity to Roads and Urban Centers

Accessibility is a critical determinant of settlement viability, impacting service delivery, market access, and emergency response capacity. Buffer analyses show that over 80% of rural settlements are within 5 km of major roads, with 40% located less than 2 km away (Figure 7a,b). Similarly, 70% of settlements lie within 5 km of urban centers, which function as hubs for economic activities and essential services (Figure 7c,d). The proximity to transport infrastructure and towns enhances adaptive capacity, facilitating climate change mitigation and socio-economic development [75]. These spatial relationships underscore the importance of targeted infrastructure investment to strengthen rural resilience networks.

3.2.4. Aggregate Spatial Distribution Insights

Integrating physical and infrastructural factors reveals that the majority of rural settlements in the Western Highlands are concentrated in areas characterized by elevations below 1500 m, slopes less than 20°, and proximal access to rivers and road networks. Specifically, over 60% of settlements reside within 4 km of rivers and 2–5 km of roads, reflecting a preference for landscapes that balance resource access with relative safety from hazards. These spatial–environmental conditions delineate zones of higher suitability and resilience potential, providing a robust empirical basis for guiding sustainable rural development and climate-adaptive planning in this vulnerable mountainous region [57].

3.3. Influencing Factors of Spatial Distribution Based on GeoDetector Model

Building upon the spatial distribution analyses, the GeoDetector model was applied to quantitatively assess the explanatory power of key environmental and infrastructural factors on rural settlement patterns within Rwanda’s Western Highlands. Using the kernel density estimation values as the dependent variable (Y), five principal explanatory factors were evaluated: elevation (X₁), slope (X₂), distance from rivers (X₃), distance from roads (X₄), and distance from towns (X₅) (Figure 8).

3.3.1. Single-Factor Detection Results

Slope (X₂) emerged as the dominant factor influencing settlement distribution, exhibiting the highest q-statistic value of 0.87. This indicates that gentler terrains are strongly preferred for settlement, consistent with topographic constraints related to landslide susceptibility and accessibility. Distance from rivers (X₃) followed with a moderate explanatory power (q = 0.24), underscoring the dual role of proximity to water resources against flood risk. Elevation (X₁) and distance from towns (X₅) presented moderate influences with q-values of 0.12 and 0.015, respectively. Conversely, distance from roads (X₄) displayed minimal impact (q = 0.003), reflecting recent enhancements in the regional road network that have likely reduced infrastructural limitations on settlement location (Figure 8a). These findings corroborate the primacy of terrain and hydrological factors identified in earlier spatial analyses and align with established theories on rural settlement suitability and climate resilience [43,76].

3.3.2. Two-Factor Interaction Detection Results

The investigation of interactions between factors revealed synergistic effects that significantly enhanced explanatory power for settlement spatial variability beyond individual factors. The interaction between slope (X₂) and distance from rivers (X₃) yielded the highest q-value of 0.9348, confirming that moderate slopes in close proximity to water bodies represent the most favorable settlement environments in the Western Highlands. Other notable interactions included slope–town distance (X₂ × X₅) with q = 0.80, slope–road distance (X₂ × X₄) with q = 0.78, and elevation–slope (X₁ × X₂) exhibiting q = 0.71. The weakest interaction observed was between elevation and road distance (X₁ × X₄) at q = 0.2233, emphasizing the relatively subordinate role of infrastructure proximity when compared to topographic and hydrological determinants (Figure 8b). Collectively, these interaction results highlight the integrated influence of terrain and water accessibility in shaping rural settlement patterns and substantiate the use of multifactorial evaluation frameworks for climate-resilient land use planning [77].

3.4. Suitability Evaluation of Rural Settlement Areas Using AHP

Building on the preceding spatial and statistical analyses, AHP was employed to integrate environmental and locational factors into a comprehensive suitability assessment for rural settlements in Rwanda’s Western Highlands. This approach quantitatively weighted key criteria identified in earlier sections, generating an overall suitability score that informs prioritized zones for resilient development.

Spatial Suitability Patterns and Implications for Rural Development

The weighted overlay analysis, executed using ArcGIS 10.7, generated a comprehensive spatial suitability map delineating distinct gradients across the Western Highlands. A significant congruence was observed between the spatial distribution of existing rural settlements and areas classified as “Highly Suitable” or “Relatively Suitable,” thereby confirming the ecological validity of the predictive model. This methodological approach, which integrates spatially explicit environmental criteria with multi-criteria weighting, aligns with studies in analogous mountainous regions, such as the Loess Plateau in China, where similar GeoDetector and Analytic Hierarchy Process (AHP) methods have proven effective in assessing rural settlement suitability and livability [78]. Districts including Rubavu, Karongi, and Rusizi exhibited the highest proportions of suitable land, which correlates with their moderate topographic gradients, favorable elevations, and enhanced access to critical resources and infrastructure (Figure 9;

Table 5. Suitability grade distribution and summary statistics by district.

District	Highly Suitable (%)	Suitable (%)	Relatively Suitable (%)	Basically Suitable (%)	Unsuitable (%)	Mean Score (M)	Standard Deviation (SD)
Rubavu	49.3	31.8	11.5	6.2	1.2	0.8045	26.3
Karongi	37.5	27.2	17.0	9.1	9.2	0.703	24.08
Rusizi	28.6	35.4	20.0	9.0	7.0	0.674	22.80
Nyabihu	5.6	42.1	31.5	15.6	8.2	0.56825	21.80
Rutsiro	4.6	33.2	40.1	19.4	2.7	0.544	21.50
Nyamasheke	15.2	34.1	22.3	18.0	10.4	0.557	21.20
Ngororero	2.5	28.0	30.0	22.0	17.5	0.51925	17.45

). Conversely, districts such as Nyabihu, Ngororero, Rutsiro, and Nyamasheke displayed substantial areas of reduced suitability. These findings underscore the imperative for targeted interventions in these regions, specifically focusing on infrastructure enhancements, the implementation of sustainable land management practices, and the development of adaptive planning strategies to fortify climate resilience and foster sustainable livelihoods.

This spatially explicit suitability evaluation provides a scientifically grounded framework to guide policymakers and practitioners in prioritizing investment and adaptation strategies, thereby contributing to resilience-oriented rural development in complex mountainous landscapes.

The results demonstrate that rural settlement distribution in Rwanda’s Western Highlands is strongly influenced by slope, elevation, and proximity to water sources, with topography emerging as the primary constraint on spatial suitability.

The AHP-based composite suitability analysis confirms the ecological logic of existing settlement patterns while identifying spatial disparities in development potential across districts. These insights offer a scientifically grounded foundation for planning climate-resilient rural settlement strategies in mountainous environments prone to geohazards.

4. Discussion

This study presents a comprehensive spatial and multi-criteria evaluation of rural settlement patterns and land suitability within Rwanda’s Western Highlands, employing an integrated framework that combines GIS, GeoDetector modeling, and an AHP analysis. The results illuminate the intricate interplay between environmental and infrastructural factors shaping settlement distributions and offer empirically grounded insights for climate-resilient rural development. This discussion synthesizes key findings within their environmental and socio-economic contexts, highlights methodological innovations, explores policy implications, and identifies limitations to guide future research directions.

4.1. Spatial Settlement Patterns and Environmental Context

Rural settlements in the Western Highlands exhibit heterogeneous spatial configurations—clustered, dispersed, and random—that reflect the complex influence of topography, hydrology, and infrastructure. Clustered settlements, notably around Gisenyi and Kamembe, benefit from proximity to transport and service networks, facilitating efficient access to economic opportunities and public services. However, this concentration also amplifies localized vulnerability to climate hazards, including flooding. Conversely, dispersed settlements in districts such as Rutsiro, Ngororero, and Nyamasheke occupy steeper slopes and higher elevations, facing challenges related to accessibility but potentially reduced exposure to density-dependent hazards. These patterns corroborate prior studies emphasizing terrain-driven settlement heterogeneity and underscore the necessity for differentiated spatial planning strategies that promote densification and infrastructure development in clustered zones while enhancing connectivity and resilience in more isolated areas [79,80].

4.2. Dominant Spatial Drivers and Factor Interactions

GeoDetector analysis identified slope and proximity to rivers as primary determinants of settlement distribution, underscoring the critical balance between terrain stability and water accessibility essential for sustainable rural habitation. The significant interaction between slope and river proximity reveals a settlement preference for moderate slopes adjacent to water sources, mitigating both landslide and flood risks. Although variables such as distance to roads and urban centers exhibited lower individual explanatory power, their interactions with natural factors significantly increased the overall explanatory capacity, which suggests value of integrated, multivariate spatial models over simplistic univariate approaches [43,81]. These findings advance the understanding of rural settlement dynamics within complex mountainous topographies by elucidating synergistic environmental and infrastructural influences.

4.3. Methodological Advancement: Integrating AHP with GeoDetector

A principal contribution of this research lies in the integration of AHP with GeoDetector-derived q-statistics to empirically inform and validate factor weighting within the multi-criteria evaluation framework.

This hybrid approach addresses the subjectivity often inherent in expert-based AHP by incorporating quantitative spatial influence metrics, while preserving critical expert knowledge. The resulting weighting schema, emphasizing natural factors such as slope and river proximity alongside socio-infrastructural variables, enhances the robustness, transparency, and replicability of suitability assessments. This methodological fusion represents a substantive advancement in spatial decision support systems aimed at climate-resilient rural land use planning [82].

4.4. Policy Implications for Climate-Resilient Rural Development

Spatial suitability mapping highlights key regions, particularly within Rubavu, Nyamasheke, and Karongi, where moderate slopes, favorable elevations, and infrastructural accessibility converge to support sustainable settlement expansion. Targeted interventions including investments in green infrastructure, sustainable agricultural practices, and eco-sensitive housing can leverage these advantages to accommodate population growth while minimizing environmental degradation. Conversely, districts such as Nyabihu, Ngororero, and Rutsiro, characterized by extensive low-suitability zones, require prioritized measures such as slope stabilization, reforestation, flood mitigation, and decentralization of critical services.

Beyond prevalent hazards like landslides and flooding, the Western Highlands face multifaceted climate-related risks, including drought, temperature variability, and ecosystem disruptions that threaten agricultural productivity and water security. These challenges necessitate integrated, risk-informed planning approaches transcending simple proximity metrics to embrace comprehensive landscape resilience. Participatory mapping and spatial decision support systems are essential for incorporating diverse hazard profiles and stakeholder perspectives, aligning with Rwanda’s Vision 2050 and broader climate action agendas. This study advocates for holistic rural development strategies, proactively addressing climate-induced vulnerabilities to bolster long-term community resilience.

4.5. Limitations and Future Research Directions

While this study advances rural settlement suitability assessment in mountainous contexts, several limitations warrant attention. The analyses predominantly rely on static spatial datasets that do not capture temporal dynamics such as seasonal climatic variability, land use change trajectories, or demographic shifts. Future research integrating multi-temporal remote sensing, migration data, and dynamic hazard modeling would enable more adaptive and forward-looking evaluations. Furthermore, expanding the participatory dimension of AHP weighting through community-based GIS (PGIS) would strengthen the social legitimacy and contextual relevance of weighting criteria [83]. Incorporating additional environmental variables such as soil quality, lithology, biodiversity indices, and microclimatic factors would deepen the multidimensional characterization of settlement suitability and vulnerability. Finally, transitioning from point-based to polygonal delineations of settlements may enhance spatial precision in exposure and land use analyses, enabling finer-scale planning and risk assessment.

Collectively, these directions underscore the imperative for continued methodological refinement and data enrichment, thereby enhancing the applicability and impact of spatially explicit rural development frameworks within Rwanda’s Western Highlands and comparable mountainous regions.

5. Conclusions

This study developed a robust and replicable framework that integrates GIS-based spatial analysis, AHP, and GeoDetector statistical modeling to evaluate rural settlement patterns and land suitability in Rwanda’s Western Highlands (Central Africa). By combining environmental and infrastructural variables with a novel hybrid-weighting approach—empirically validating expert-derived AHP weights through GeoDetector’s spatial explanatory power—this methodology substantially enhances objectivity, spatial precision, and contextual relevance. This approach effectively addresses inherent biases and limitations common to conventional, less empirically grounded suitability assessments.

The findings confirm that natural factors, particularly slope and proximity to rivers, serve as the primary determinants of rural settlement distribution in complex mountainous terrains. The study also highlights significant synergistic interactions among environmental attributes and accessibility variables, such as distances to roads and urban centers, underscoring the necessity of integrated multi-criteria frameworks over simplistic, uni-dimensional models. These insights deepen the understanding of settlement viability under climate hazard exposure and provide nuanced perspectives directly applicable to mountainous rural contexts globally.

From an applied perspective, the spatially explicit suitability maps identify high-potential zones in Rubavu, Rusizi, and Karongi districts, where favorable terrain and infrastructure converge, offering strong opportunities for sustainable rural development. These priority areas can guide targeted investments in green infrastructure, hazard mitigation, and enhanced service provision to accommodate population growth sustainably. Conversely, regions with lower suitability, including Nyabihu, Rutsiro, Nyamasheke, and Ngororero, require tailored interventions emphasizing slope stabilization, flood risk management, and improved connectivity to reduce vulnerability and support resilient livelihoods.

Methodologically, this research advances spatial decision support by integrating qualitative expert judgment with quantitative spatial statistics, yielding a more balanced and empirically grounded weighting scheme. This hybrid approach significantly improves the transparency and defensibility of land suitability models, representing a critical advancement for policy-relevant applications. Moreover, the framework’s scalability and transferability enable adaptation to other mountainous, hazard-prone regions facing similar development and climate resilience challenges.

Recognizing the current limitations, such as reliance on static spatial data and a focus on expert-driven weighting, future research should incorporate temporal dynamics through multi-temporal remote sensing and dynamic hazard modeling. Expanding stakeholder participation via community-based participatory GIS (PGIS) approaches will further increase social legitimacy and adaptive capacity. Additionally, incorporating more granular environmental variables including soil properties, lithology, biodiversity, and microclimatic factors will refine the multidimensional assessment of settlement suitability and vulnerability. Finally, transitioning from point-based settlement representations to polygonal delineations will improve spatial accuracy and exposure estimation, providing a more detailed understanding of settlement extent and impact.

In summary, this comprehensive and methodologically innovative study provides a scientifically robust and operationally relevant foundation for climate-resilient rural planning in Rwanda’s Western Highlands.

It offers a valuable blueprint for comparable regions worldwide, where the integration of advanced spatial analytics, multi-criteria evaluation, and participatory insights can foster more adaptive, inclusive, and sustainable rural development in the face of ongoing socio-environmental change.