Urban Flood Susceptibility Mapping Using GIS and Analytical Hierarchy Process: Case of City of Uvira, Democratic Republic of Congo

Abstract

The city of Uvira, located in the eastern Democratic Republic of Congo (DRC), is increasingly experiencing flood events with devastating impacts on human life, infrastructure, and livelihoods. This study evaluates flood susceptibility in Uvira using Geographic Information Systems (GISs), and an Analytical Hierarchy Process (AHP)-based Multi-Criteria Decision Making approach. It integrates eight factors contributing to flood occurrence: distance from water bodies, elevation, slope, rainfall intensity, drainage density, soil type, topographic wetness index, and land use/land cover. The results indicate that proximity to water bodies, drainage density and slope are the most influential factors driving flood susceptibility in Uvira. Approximately 87.3% of the city’s land area is classified as having high to very high flood susceptibility, with the most affected zones concentrated along major rivers and the shoreline of Lake Tanganyika. The reliability of the AHP-derived weights is validated by a consistency ratio of 0.008, which falls below the acceptable threshold of 0.1. This research provides valuable insights to support urban planning and inform flood management strategies.

1. Introduction

A flood hazard refers to the likelihood of a flood event with a specific intensity and recurrence interval occurring in a given location and timeframe [1,2]. Building on this concept, flood susceptibility represents an area’s inherent vulnerability to flooding based on its physical characteristics, indicating where floods are likely to occur without predicting their frequency [3,4].

The recent increase in the frequency and severity of extreme weather events (wildfires, floods, heatwaves, etc.) underscores the devastating impacts of climate change worldwide [5,6]. Due to climate change impacts and the increasing concentration of populations near coastal areas, lakeshores, and river basins, flooding has become one of the most frequent and devastating disasters, causing severe damage to communities, livelihoods and critical infrastructure [7,8]. Floods have accounted for 44% of all disaster events during the last 5 decades (1970–2019), affecting over 2.3 billion people worldwide between 1980 and 2009 and contributing to mass displacement, injuries and the spread of waterborne diseases [7,9].

According to [9], economic losses from weather-, climate- and water-related disasters have been estimated at USD 3.6 trillion globally between 1970 and 2019, with floods alone accounting for 31% of that total. In Africa, floods were responsible for 34% of the USD 38.5 billion disaster-related economic losses. In the Democratic Republic of Congo, floods between October 2023 and January 2024 affected at least 300,000 people, caused approximately 300 deaths and the destruction of more than 43,000 houses, and severely impacted infrastructure and agricultural land [10,11].

In the past five years, the city of Uvira in the Democratic Republic of Congo has experienced several flood events. In April 2020, catastrophic flooding affected 9 out of the city’s 14 neighborhoods, resulting in at least 38 deaths, 185 injuries, 100 separated children and the displacement of approximately 70,000 people [12,13]. Flooding in Uvira predominantly occurs during the rainy season, when rivers overflow due to extreme rainfall events. The most recent incidents were reported in 2024, when the city experienced intense flooding in May, June, and July, leading to the displacement of 1200 people to camps under deplorable conditions [14].

Uvira features several rivers, including the Kavimvira, Mulongwe, and Kalimabenge, which frequently overflow during periods of heavy rainfall. The city is also located along the shores of Lake Tanganyika, whose water level has risen by approximately 2 m since 2019 [15,16]. In total, 61.9% of Uvira’s land area has a slope of less than 2%, and 78.2% is considered susceptible to flooding when slope is taken into account [17].

Accurate flood susceptibility mapping is essential for risk mitigation. GISs and remote sensing have emerged as powerful tools for assessing flood-prone areas, particularly in data-scarce regions [18]. Machine Learning (ML) has also emerged as another powerful tool for flood susceptibility mapping (FSM), offering predictive capabilities through algorithms like Neural Networks and ensemble models [19]. Its methods, such as Neural Networks (NNs), ensemble models, and hybrid approaches, have shown promise in flood modeling but face challenges like overfitting, data inconsistency, and limited real-time predictive capacity [20], especially in data-scarce regions like Uvira. In contrast, GISs and RS techniques are essential and widely applied to assess flood risks and enhance disaster preparedness and response [18]. Multi-Criteria Decision Making (MCDM) techniques like the AHP offer a robust framework for integrating diverse flood-conditioning factors (e.g., slope, land use, rainfall, and proximity to water) into spatial assessments [21,22,23]. This approach aligns with methodologies applied in other contexts, such as railway infrastructure resilience studies, where GIS-based AHP models identified flood-prone zones to prioritize maintenance and disaster planning [24].

The integration of GISs and RS with the AHP facilitates the evaluation of flood susceptibility factors within a given area. This approach has consistently proven effective for flood susceptibility mapping across various regions [21,22]. For example, ref. [23] demonstrated the application of the AHP in Kota Belud, Malaysia, where topographic and environmental factors were integrated to provide a comprehensive assessment of flood-prone zones. Furthermore, RS techniques are particularly useful in data-scarce regions, offering valuable spatial insights to support flood management efforts. The AHP also enables the prioritization of risk factors by assigning them scores based on their relative importance in a specific context [25].

Despite Uvira’s acute vulnerability to recurrent floods, existing studies have focused narrowly on post-disaster impact assessments [13] or community resilience strategies [16], leaving a critical gap in proactive, spatially explicit flood susceptibility mapping. While GISs and AHP-based approaches have been widely applied elsewhere [21,22,23], their use in data-scarce African urban contexts particularly in the Great Lakes region remains limited. No prior study has integrated Uvira’s unique drivers (lake-level rise, dense river networks, and rapid urbanization) into a comprehensive risk model, hindering evidence-based mitigation planning.

This study provides the first high-resolution flood susceptibility map of Uvira, addressing an urgent need for actionable risk intelligence. By combining open access remote sensing data with AHP-weighted multi-criteria analysis, the methodology offers a replicable framework for flood-prone cities with similar data constraints. The results directly inform land use zoning, early-warning systems, and infrastructure priorities while filling a key gap in climate adaptation planning for Central Africa’s rapidly urbanizing lakeshore communities.

The primary objective of this study is to assess and map flood susceptibility in Uvira, the Democratic Republic of Congo, using a GIS, RS and the AHP. This approach aims to identify flood-prone areas and inform urban planning and flood management strategies. The significance of this study lies in its potential to enhance flood management practices in Uvira by pinpointing high-risk zones, thereby benefiting urban planners, policymakers, and communities at risk of flooding. Given the demonstrated effectiveness of these methods in various regions, the study’s methodology can also be replicated to other flood-prone areas worldwide to improve disaster preparedness and resilience.

This study contributes to flood risk assessments by integrating the AHP with a map removal sensitivity analysis to produce and verify a flood susceptibility map for Uvira. While the AHP is a widespread method used in multi-criteria flood assessments, it is usually coupled with expert opinion in previous studies without any further empirical verification. With the incorporation of a spatially explicit sensitivity analysis, this study enables a more efficient evaluation of factor influence, leading to more transparent and reliable results. Implementing this methodology in Google Earth Engine also ensures computational efficiency and replicability, which is vital in data-constrained urban environments. Applying this integrated approach to the less-studied Uvira context addresses a crucial void and offers practical information for flood risk management in analogous contexts.

2. Materials and Methods

2.1. Study Area

Uvira is a growing city located in South Kivu Province in the eastern part of the Democratic Republic of Congo. Figure 1 shows the location and boundaries of the study area. It lies between 3°20′ S and 4°20′ S latitude, and 29°00’ E and 29°30′ E longitude, at an altitude of approximately 800 m, with a population of 760,857 inhabitants [26]. Geographically, Uvira is situated between the shores of Lake Tanganyika and the Mitumba Mountains, with an average width of only 2 km, making it a relatively narrow urban area [13,17]. The city is bordered to the north by the Ruzizi Plain (River Kawizi), to the south by the Territory of Fizi (River Gengeza), to the east by Lake Tanganyika, and to the west by the Mitumba Mountains [17].

Uvira is traversed by three major rivers, Mulongwe, Kavimvira, and Kalimabenge, that flow directly into Lake Tanganyika, as well as the Kiliba River, which drains into the Ruzizi River. Located to the east of the city, the Ruzizi River is the most significant watercourse, serving as natural border between the Democratic Republic of Congo, Rwanda, and Burundi and connecting Lake Kivu to Lake Tanganyika [17]. Surrounded and intersected by rivers, the city usually experiences flash floods [13]. Another source of flooding is Lake Tanganyika, whose water level has risen by approximately two meters since 2019.

Uvira experiences a humid tropical climate characterized by a prolonged rainy season from October to May and a shorter dry season from July to September. Average temperatures range between 24 °C and 28 °C throughout the year.

2.2. Data

The datasets used in this study, presented in

Table 1. Datasets, source and description.

N°	Data	Description	Spatial Resolution	Time Scale, Date of Acquisition or Latest Update	Provider	Source
1	Water data (rivers and lakes)	Shapefile containing all DRC’s river networks and lake polygons	-	-	OSM	https://download.geofabrik.de/ (accessed on 14 November 2024)
2	SRTM	Digital elevation model	30 m	2013	NASA	https://developers.google.com/earth-engine/datasets/catalog/USGS_SRTMGL1_003 (accessed on 16 November 2024)
3	CHIRPS rainfall	Daily rainfall data	5.6 km	1 January 1994 to 31 December 2023	UCSB	https://developers.google.com/earth-engine/datasets/catalog/UCSB-CHG_CHIRPS_DAILY (accessed on 9 December 2024)
4	Soil type	Harmonized World Soil Database v2.0	1 km	2023	FAO	https://www.fao.org/soils-portal/data-hub/soil-maps-and-databases/harmonized-world-soil-database-v20/en/ (accessed on 9 December 2024)
5	Sentinel 2	Surface reflectance data	10 m	10 July 2024	ESA	https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S2_SR_HARMONIZED (accessed on 22 December 2024)

SRTM: Shuttle Radar Topography Mission, DRC: Democratic Republic of Congo, CHIRPS: Climate Hazards Center InfraRed Precipitation With Station data; OSM: OpenStreetMap; NASA: National Aeronautics and Space Administration; UCSB: University of California, Santa Barbara; FAO: Food and Agriculture Organization; ESA: European Space Agency.

, were collected from various sources and used for different purposes, detailed in Section 2.3. The analysis focuses on long-term flood susceptibility rather than a specific flood event. Rainfall data were obtained, such as 30-year averages (1994–2023), to reflects long-term precipitation patterns. Other datasets such as elevation, slope, drainage density, and soil type represent stable biophysical characteristics of the landscape. To derive a recent land use/land cover map, a Sentinel 2 image captured in 2024 was used and a single-date classification was applied to represent current LULC conditions. Finally, the topographic wetness index was derived from a static Digital Elevation Model. All eight input layers were harmonized to a common spatial resolution of 30 m.

2.3. Methodology

The first step was to identify the factors that influence the occurrence of floods in the city of Uvira. This was accomplished through several observations and an intensive literature review. Satellite images were visually observed, and topographic features were extracted from a Digital Elevation Model to provide a clear idea of the city’s physical characteristics. Eight factors were identified as contributing to the occurrence of floods in the city, including distance from water, elevation, slope, rainfall, drainage density, soil type, the topographic wetness index, and land cover. The selection of these factors was guided by their widespread use and proven relevance in flood susceptibility assessments across diverse African urban contexts [27,28]. All these factors were employed in previous studies due to their strong influence on flood occurrence through surface runoff generation, water accumulation, and terrain control. While other variables such as hydraulic conductivity, geology, the NDVI, or proximity to roads have also been used in the literature, they were excluded from this analysis either due to the unavailability of reliable spatial data (e.g., for hydraulic conductivity) or their limited relevance in the context of Uvira. The overall workflow of the methodology is presented in Figure 2.

2.3.1. Calculation of Factors

a.

Distance from water

There is an inverse relationship between distance from rivers and flood risk, with the areas closest to the rivers facing the highest risk of flooding [29]. The city of Uvira is threatened by both the rivers that cross it and Lake Tanganyika, which borders it to the east. Thus, the distance to water was calculated considering both the lake and the rivers simultaneously, using the Distance Accumulation tool of ArcGIS Pro. This tool calculates accumulated distance for each cell to sources, allowing for straight-line distance, cost distance, and true surface distance [30]. The river network and the lake polygon were used to calculate the distance from water.

b.

Elevation

According to [29], there is also an inverse relationship between elevation and flood risk, as lowland and downstream areas are more susceptible to floods compared to their surrounding highlands. A Shuttle Radar Topography Mission (SRTM) Digital Elevation Model was used to create an elevation map. It was imported and clipped to the study area using Google Earth Engine and then exported and visualized in ArcGIS Pro.

c.

Slope

Slope is directly proportional to surface runoff, thereby influencing the occurrence of floods, with flat areas being at a higher risk. Steeper slopes produce higher velocity compared to flatter or gentler slopes, allowing for faster runoff [31]. Slope (in degrees) was extracted from the imported SRTM DEM. It was calculated using the Terrain Analysis Slope tool in Google Earth Engine and then exported and visualized in ArcGIS Pro.

d.

Rainfall

Floods are caused by rainfall events; they have a strong relationship with the total rainfall that occurs over a spell of time [32,33]. CHIRPS rainfall data covering a 30-year period (1994–2023) were imported into the Google Earth Engine platform. The maximum daily rainfall values over this period were calculated and exported as a raster layer to ArcGIS Pro for visualization

e.

Drainage density

Drainage density also highly influences the occurrence of floods. According to [18], high drainage density can contribute to accelerating water movement during heavy rainfall, thus increasing the risk of flooding. Drainage density was calculated using the Line Density tool in ArcGIS Pro [28]. Drainage density refers to the total length of streams in a watershed; high drainage density results in reduced evapotranspiration and infiltration, increased runoff, and consequently, higher flood susceptibility [34].

f.

Soil type

Soil types vary significantly in their capacity to retain water. Low infiltration capacity leads to higher surface runoff, thus increasing flood susceptibility [35]. Ref. [36] classified various soil types based on their permeability levels, categorizing Solonchaks as having low permeability, Fluvisols as moderate, and Ferralsols as high. Soils with lower permeability typically exhibit reduced infiltration rates, resulting in increased runoff and, consequently, higher flood susceptibility.

g.

Topographic Wetness Index

The topographic wetness index (TWI) represents the terrain-influenced balance of catchment water supply and local drainage [37]. Higher values of the TWI correlate with greater humidity levels, whereas lower values indicate drier conditions. Regions characterized by low TWI values are generally associated with better drainage capacity and lower flood susceptibility [38].

The topographic wetness index was calculated using the following formula [39]:

$$TWI=\ln {\displaystyle \frac{A_s}{\tan \beta }}$$

(1)

where A_s is the upslope contributing area per unit contour length in m²/m and β is the local slope angle in radians.

h.

Land Use/Land Cover

Changes in land use and land cover (LULC) influence hydrological conditions, such as runoff, infiltration, and land surface coefficient, that simultaneously may intensify the effects of hydrometeorological events, including flooding [40,41]. Land use and land cover were analyzed using the Random Forest algorithm, which is widely applied for classification and regression tasks due to its strong predictive performance and ability to minimize overfitting [42]. The random forest model was built and executed on the Google Earth Engine platform using Sentinel 2 images, and the LULC results were exported to ArcGIS 10.7 for visualization. Sentinel-2 imagery was selected for this study due to its ability to provide high-resolution multispectral data (10–60 m), making it a valuable source for accurately deriving the LULC map [43].

2.3.2. Normalization

In Multi-Criteria Decision Making (MCDM), normalization is the step of transforming data expressed in different units into a common scale with comparable units [44]. In this study, normalization was performed through factor reclassification. Each factor was reclassified into different classes, based on the literature in the case of LULC and soil type and using the equal interval classification method in ArcGIS for continuous variables. This method was selected for its simplicity and ease of interpretation, making the results accessible and useful to both technical and non-technical audiences [45]. The reclassified values range from 1 to 5, where 1 indicates the lowest influence on flood occurrence and 5 indicates the highest. All reclassified values are presented in

Table 2. Factors and class weights.

Factor	Value	Reclassified Values	Factor Weight	Class Weight
Distance from water (m)	0–486 m	5	0.319	1.595
	486–972 m	4		1.276
	972–1458 m	3		0.957
	1458–1944	2		0.638
	1944–2430	1		0.319
Elevation (m)	746–798	5	0.083	0.415
	798–851	4		0.332
	851–903	3		0.249
	903–956	2		0.166
	956–1008	1		0.083
Slope (degrees)	0–8	5	0.125	0.625
	8–15	4		0.5
	15–23	3		0.375
	23–30	2		0.25
	30–38	1		0.125
Rainfall (mm)	60–68	1	0.083	0.083
	68–75	2		0.166
	75–83	3		0.249
	83–90	4		0.332
	90–98	5		0.415
Drainage density (km/km2)	0–0.5	1	0.184	0.184
	0.5–1.1	2		0.368
	1.1–1.6	3		0.552
	1.6–2.1	4		0.736
	2.1–2.7	5		0.92
Soil type	Humic ferralsols	3	0.118	0.354
	Mollic fluvisols	4		0.472
	Gleyic solonchaks	5		0.59
Topographic wetness index	17.4–18.2	1	0.032	0.032
	18.2–19.0	2		0.064
	19.0–19.8	3		0.096
	19.8–20.5	4		0.128
	20.5–21.3	5		0.16
Land use and land cover	Built-up area	5	0.056	0.28
	Agriculture and barren areas	3		0.168
	Green areas	2		0.112
	Forest	1		0.056

Each of these values corresponds to a specific flood susceptibility class. 1: very low, 2: low, 3: moderate, 4: high, 5: very high.

.

The reclassification of LULC was conducted based on the findings of [46] who identified settlements as being at the highest risk of flooding due to human activities that alter soil structure and reduce infiltration capacity, primarily through vegetation removal, urbanization, and cultivation. In contrast, vegetated areas such as forests, grasslands, shrubs, and even agricultural lands with food crops contribute to flood mitigation by enhancing the soil’s ability to absorb water and reduce runoff. The soil type raster layer was reclassified based on the work of [36], who categorized different soil types according to their permeability levels, as outlined in f of Section 2.3.1. This classification directly aligns with the data used in this study, which considered soil types, while several other studies consider only soil texture.

2.3.3. Analytical Hierarchy Process

The method used in this study, the AHP, is a reliable, rigorous, and robust approach for capturing and measuring subjective judgments in MCDM [47]. The AHP is used to assess the consistency of criterion weightings by developing a pairwise comparison matrix [48].

The first step in the AHP involves conducting pairwise comparisons among the criteria and, where applicable, their sub-criteria [49]. To evaluate the impact of each factor on flood susceptibility, the factors were correlated and compared two by two. To construct the pairwise comparison matrix, each criterion was evaluated using the Saaty scale (

Table 3. The Saaty scale.

Scales	Degree of Importance of One Factor over Another
1	Equal importance
3	Moderate importance
5	Strong importance
7	Very strong importance
9	Extreme importance
2, 4, 6, 8	Intermediate values, for a compromise between the above values

Source: [50].

), with values ranging from 1 to 9. A score of 1 indicates that the two compared factors are equally important, whereas a score of 9 signifies that the factor in the row is significantly more important than the one in the column.

Following the construction of the matrix, the process proceeds with the normalization of the eigenvector values and concludes with the validation of the judgments’ consistency through the application of specific properties [28]. Validation was assessed using the consistency ratio (CR), according to the following formula:

$$CR={\displaystyle \frac{CI}{RI}}$$

(2)

where CI is the Consistency Index and RI is the Random Index. The matrix is sufficiently consistent when the CR ≤ 0.1.

The CI was calculated using the following formula:

$${\displaystyle \frac{{\lambda }_{max}-n}{n-1}}$$

(3)

where λ_max is the largest correct number of the matrix and n is the number of criteria.

And the RI is a table value based on the number of criteria, as found in [51].

The pairwise comparison matrix was built, and the CR was calculated using MS Excel 2019, while the AHP model was built and run using the Google Earth Engine platform.

The final flood susceptibility map was generated using a weighted linear combination method, which uses all reclassified factors to produce the final flood susceptibility zone values, according to the following equation [52]:

$$FSZ={\sum }_{\left\{i=1\right\}}^{\left\{n\right\}}{W}_i^s\ast R_i^s$$

(4)

where W_i is to the normalized weight of the factor, and R_i is the rating of the factor (reclassified values).

A weight was successfully assigned to each factor according to its own contribution to flood occurrence in Uvira, with a consistency ratio of 0.008 (<0.1), confirming the reliability of the weights. To classify the susceptibility levels, the continuous flood susceptibility values were reclassified into five classes using a threshold-based approach in Google Earth Engine. The reclassification was defined as follows:

• F ≤ 1: very low susceptibility;
• 1 < F ≤ 2: low susceptibility;
• 2 < F ≤ 3: moderate susceptibility;
• 3 < F ≤ 4: high susceptibility;
• F > 4: very high susceptibility.

Where F is the continuous flood susceptibility raster.

2.3.4. Sensitivity Analysis

To assess the relative importance of each flood susceptibility factor and to validate the results derived from the AHP, a map removal sensitivity analysis was conducted. This approach evaluates the influence of individual factors by quantifying the impact of their exclusion on the overall flood susceptibility map. It was selected for its objectivity, as it is a bias-free analytical technique that facilitates the derivation of empirically grounded and robust weighting values. The following steps were followed, described by [53].

First, a baseline flood susceptibility map was generated by assigning equal weights to all considered factors. This map served as a reference scenario to which all subsequent maps were compared. Next, a series of scenarios were developed by systematically removing one factor at a time from the analysis. In each of these scenarios, the weight of the excluded factor was reset to zero, and the weights of the remaining factors were adjusted equally to ensure the total sum remained one. For every adjusted scenario, a flood susceptibility map was generated.

The influence of each factor was assessed by comparing the flood susceptibility map with the factor removed to the baseline map. The mean absolute difference (MAD) between the two maps was calculated over the study area. The magnitude of this difference reflects the sensitivity of the flood susceptibility zoning to the removal of that particular factor.

The influence of each specific factor was quantified by comparing the newly created susceptibility map based on that particular scenario to the original baseline map. The MAD was calculated across the entire study area by performing a pixel-based subtraction, which was applied between each scenario map and the baseline map. The absolute values of the differences were used to capture both positive and negative changes. For each factor, the MAD was computed using the ee.Reducer.mean() function of Google Earth Engine, which calculates the average value of all pixels over a specified region. The total MAD was then obtained by summing each individual factor’s MADs. Subsequently, the new factors’ weights were derived using the following formula:

$$W_i={\displaystyle \frac{{MAD}_i}{Total MAD}}$$

(5)

where W_i is the new weight assigned to factor i, MAD_i is the Mean Absolute Difference for that factor, and Total MAD is the sum of the all MAD values across the eight factors.

This difference effectively measured how sensitive flood susceptibility predictions were to the removal of a given factor, indicating its relative significance. Using these sensitivity analysis-derived weights, a final flood susceptibility map was produced. This updated map served then to validate the susceptibility maps derived through the AHP.

3. Results

This study successfully assessed flood susceptibility in the city of Uvira, using a GIS and RS through the AHP that integrated eight flood-inducing factors. The results are structured in the following sections:

3.1. Flood-Inducing Factors

3.1.1. Distance from Water

In the city of Uvira, proximity to water is the most significant contributing factor to flood occurrence (factor weight = 0.319), despite the notable influence of other variables (Table 2). During extreme rainfall events, rivers overflow, resulting in flash floods. Additionally, rising water levels in Lake Tanganyika have led to gradual flooding along its shoreline. The distance to water, ranging from 0 to 2430 m, was classified into five categories. As shown in

Table 4. Distance from water.

Distance from Water (m)	Description	Area (km2)	%
0–486	Very close	18.3	63.1
486–972	Close	6.8	23.4
972–1458	Moderate	2.3	7.9
1458–1944	Far	1.1	3.8
1944–2430	Very far	0.5	1.7
Total		29	100

, up to 63.1% of the city’s total area lies in immediate proximity to water (0–486 m), with an additional 23.4% classified as close, further increasing flood susceptibility across much of the city. The map (Figure 3a) highlights areas in closest proximity to water, which are widespread, particularly along the Lake Tanganyika shoreline, with the exception of a small area in the northwest.

3.1.2. Elevation

Table 5. Elevation in Uvira city.

Elevation (m)	Description	Area (km2)	%
746–798	Lowest	18.2	62.8
798–851	Low	8.0	27.6
851–903	Moderate	2.1	7.2
903–956	High	0.6	2.1
956–1008	Highest	0.1	0.3
Total		29	100

indicates that approximately 62.8% of Uvira’s total land area lies at the lowest elevation range (746–798 m), primarily in the eastern part of the city, which borders Lake Tanganyika (Figure 3b). Another 27.6% falls within a lower elevation range (798–851 m), further increasing the city’s susceptibility to flooding, as low-lying areas are generally more vulnerable. Only a small portion of Uvira, located in the western area and covering 2.4% of the city, is situated at a higher elevation (903–1008 m) (Figure 3b). These areas represent the highlands that are likely less susceptible to flooding due to gravitational runoff and high drainage efficiency. However, their low spatial coverage reduces their significance in offering urban flood mitigation benefits.

3.1.3. Slope

The findings of this study (Table 2) indicate that, after proximity to water and drainage density, slope is the third most influential factor contributing to flood susceptibility in Uvira (Factor weight = 0.125). As shown in

Table 6. The distribution of slopes in the city of Uvira.

Slope (°)	Description	Area (km2)	%
0–4	Flat to very gentle	14.6	50.3
4–8	Gentle	7.3	25.2
8–14	Moderate	3.9	13.5
14–21	Steep	2.2	7.6
21–38	Very steep	1.0	3.4.0
Total		29	100

, 75.5% of the city’s land area lies within flat to gently sloping terrain, extending from the Lake Tanganyika shoreline in the east towards the city’s center (Figure 3c), thereby increasing flood susceptibility. Additionally, 13.5% of the area has moderate slopes, corresponding to moderate flood susceptibility based on topographic gradient. Figure 3c further illustrates that steep and very steep slopes, covering only 11% of the city’s total area, are primarily located in the western region, mainly dominated by the highlands.

3.1.4. Rainfall

Rainfall is also a major factor influencing flood risk in the city of Uvira (Table 2). Daily maximum rainfall ranges from 60 to 98 mm, with notable spatial variations: lower rainfall amounts are recorded in the northern part of the city and gradually increase toward the south. The areas that receive the highest (classified as high and very high) amount of daily rainfall, covering 16.9% of the city (

Table 7. Maximum daily rainfall in Uvira.

Maximum Daily Rainfall (mm)	Description	Area (km2)	%
60–68	Very low	14.2	48.9
68–75	Low	4.0	13.8
75–83	Medium	5.9	20.4
83–90	High	4.2	14.5
90–98	Very high	0.7	2.4
Total		29	100

) and represented on the map in orange and red (Figure 3d), are mainly located in the southern zones.

3.1.5. Drainage Density

In Uvira, drainage density, ranging from 0 to 2.7 km/km², plays a crucial role in the occurrence of floods. It is particularly high in areas adjacent to rivers (Figure 4a), thereby increasing flood susceptibility in those regions. However, zones characterized by very high and high drainage density cover only 1% and 9.3% of the city’s total area, respectively, whereas areas with moderate drainage density account for 7.9% (

Table 8. Drainage density in Uvira.

Drainage Density (km/km2)	Description	Area (km2)	%
0–0.5	Very low	22.0	75.9
0.5–1.1	Low	1.7	5.9
1.1–1.6	Medium	2.3	7.9
1.6–2.1	High	2.7	9.3
2.1–2.7	Very high	0.3	1.0
Total		29	100

). The remaining 81.8% of the city is situated in regions characterized by low or very low drainage density, where river networks are less concentrated. These areas of high drainage density are represented in red and orange on the map (Figure 4a), covering small areas across the city.

3.1.6. Soil Type

The city of Uvira is predominantly covered by Mollic FLUVISOLS, which occupy approximately 84.5% of its total land area (

Table 9. Soil types of Uvira.

Soil Type	Area (km2)	%
Humic Ferralsols	0.6	2.1
Mollic Fluvisols	24.5	84.5
Gleyic Solonchaks	3.9	13.4
Total	29	100

), and by Gleyic SOLONCHAKS, which extend across 13.4% in the northeastern region (Figure 4b). These findings indicate that nearly the entire city consists of soil types highly susceptible to flooding. Gleyic SOLONCHAKS exhibit very high flood susceptibility, while Mollic FLUVISOLS are also classified as highly susceptible. Only 2.1% of Uvira’s land area is composed of Humic FERRALSOLS, which are less susceptible to flooding. They are mainly located in the west, represented on the map in green (Figure 4b).

3.1.7. Topographic Wetness Index

The topographic wetness index (TWI) in the city of Uvira, ranging from 17.4 to 21.3 (

Table 10. Variation in topographic wetness index in the city of Uvira.

Topographic Wetness Index	Description	Area (km2)	%
17.4–18.2	Very low	1.2	4.1
18.2–19.0	Low	4.1	14.1
19.0–19.8	Medium	7.7	26.6
19.8–20.5	High	8.6	29.7
20.5–21.3	Very high	7.4	25.5
Total		29	100

), is generally high, with areas of elevated TWI values (classified as high and very high) covering approximately 55.2% of the city’s total land area. These areas represented in orange and red on the map (Figure 4c) typically extend from the east (Lake Tanganyika shoreline) to the central part of the city and are most dominant in the North. Additionally, 26.6% of Uvira’s land area falls within the medium TWI range, while the remaining 18.2% is characterized by low TWI values. Areas with moderate TWI values are mainly situated in the city’s center, whereas low TWI zones are concentrated in the western region.

3.1.8. Land Use/Land Cover

Human activities, including agriculture and urban development, occupy a substantial portion of Uvira’s land area (up to 63.1%), while the remaining 36.9% consists of vegetated areas, including forests (

Table 11. Land use/land cover classes in Uvira.

Land Use/Land Cover Classes	Area (km2)	%
Built up	7.6	26.2
Agriculture and barren area	10.7	36.9
Green areas	8.0	27.6
Forest	2.7	9.3
Total	29	100

). Tree cover is primarily concentrated in the northern part of the city, especially in the northeast (Figure 4d), whereas built-up areas dominate the central region, extending from the eastern zone (Lake Tanganyika’s shoreline) to the west. Agricultural zones are mainly located in the northwest, west, and southern parts of the city. The analysis of Sentinel 2 images produced an LULC map of the city of Uvira, which was validated with an overall accuracy of 0.995 and a Kappa coefficient of 0.993.

3.2. Flood Susceptibility in the City of Uvira

This subsection presents the main result of the study, the flood susceptibility map, which provides valuable insights into the spatial distribution of flood susceptibility across the city of Uvira by identifying flood-prone areas. That map was generated using a GIS, RS, and the AHP, a Multi-Criteria Decision Making approach. Eight key factors influencing flood susceptibility were selected and analyzed: distance to water bodies, elevation, slope, rainfall intensity, drainage density, soil type, the topographic wetness index, and land use/land cover. Each factor was reclassified, assigned values ranging from 1 to 5 for normalization, and then given weighted scores reflecting their relative importance, as determined through pairwise comparisons. These weighted factors were finally combined using a weighted linear combination method to produce the final susceptibility map.

The results presented in

Table 12. Flood susceptibility class area in Uvira.

Flood Susceptibility	Area (km2)	%
Very high	4.9	16.9
High	20.4	70.4
Medium	3.4	11.7
Low	0.3	1.0
Total	29	100

indicate that 16.9% of Uvira’s total land area is exposed to very high flood susceptibility. Most of these zones are located near the rivers and are dispersed throughout the city (Figure 4). Figure 4 further illustrates that a substantial portion of these areas lies along the three city’s largest rivers, Mulongwe, Kavimvira and Kalimabenge, which frequently overflow during intense rainfall events. Proximity to rivers and high drainage density are key contributors to flood susceptibility in these zones. Moreover, an overlay with the land use/land cover map revealed that some human settlements are established in these zones, directly exposing the population to potential flood impacts.

Table 12 further indicates that 70.4% of the city’s land area is classified as having high flood susceptibility, spanning various areas between rivers and along the Lake Tanganyika coastline. The flood susceptibility map (Figure 5) highlights a continuous belt of high-susceptibility zones extending from the northern to the southern parts, demonstrating that flood susceptibility in Uvira is not localized but widespread. In total, approximately 87.3% of Uvira’s land area falls within flood-prone zones, classified as high and very high susceptibility, with the majority of human settlements also located within these areas.

Furthermore, 11.7% of the city’s area, primarily in the western and northern regions, is classified as having moderate flood susceptibility, while only 1.0%, located in the northwest, is considered to have low susceptibility.

3.3. Sensitivity Analysis

This section presents the results of the sensitivity analysis, conducted using a map removal approach to assess the consistency of the AHP-derived weights and validate the flood susceptibility map. This involved systematically excluding each flood susceptibility factor and regenerating the susceptibility map accordingly. The impact of removing each factor was then measured using the mean absolute difference.

Table 13. Flood susceptibility area coverages per scenario.

Scenario	Factor Weight	Flooded Area (km2)
		Very High	High	Medium	Low	Very Low	Total
Scenario 1: Base case (equal weights)	0.125	2.1	21.7	4.7	0.4	0.1	29
Scenario 2: Distance removed	0.142	1	21.1	6	0.8	0.1	29
Scenario 3: Elevation removed	0.142	1.3	19.2	7.7	0.6	0.2	29
Scenario 4: Slope removed	0.142	1	19.3	8.3	0.4	0	29
Scenario 5: Rainfall removed	0.142	5	20.4	3.1	0.4	0.1	29
Scenario 6: Drainage density removed	0.142	7.8	18.2	2.7	0.3	0	29
Scenario 7: Soil type removed	0.142	2.1	19.8	6.3	0.8	0	29
Scenario 8: TWI removed	0.142	2.1	21.5	5.1	0.3	0	29
Scenario 9: LULC removed	0.142	2.6	21.7	4.2	0.4	0.1	29

summarizes the results issued of the sensitivity analysis, which assessed the impact of removing individual factors on the overall flood susceptibility map. The base scenario, in which equal weights were assigned to all factors, served as a reference. The other scenarios (2 to 9) represent the exclusion of one factor at a time with corresponding recalculations of flood susceptibility level areas for each level.

The results indicate that removing drainage density (scenario 6) resulted in a significant increase in the area classified as “very high” susceptibility compared to the baseline scenario. This finding aligns with the AHP-derived flood susceptibility maps, which showed very high flood susceptibility in areas of high drainage density, confirming the strong influence of this factor. Conversely, the exclusion of distance from waterbodies (scenario 2), elevation (scenario 3) and slope (scenario 4) resulted in a decrease in areas classified as “very high” susceptibility compared to the baseline. These observations further support the weights assigned to these factors in determining flood susceptibility in Uvira.

Other scenarios such as the removal of soil type, the topographic wetness index and land use/land cover resulted in “very high” flood-prone areas that were either identical or slightly different from the baseline scenario. This outcome indicates their relatively moderate influence on flood susceptibility, consistent with their assigned weights in the AHP analysis.

The findings presented in

Table 14. AHP and sensitivity analysis: flood susceptibility class area.

Flood Susceptibility	AHP-Derived		Sensitivity-Derived
	Area (km2)	%	Area (km2)	%
Very high	4.9	16.9	1.6	5.5
High	20.4	70.4	18.1	62.4
Medium	3.4	11.7	8.8	30.4
Low	0.3	1.0	0.5	1.7
Total	29	100	29	100

indicate that both the AHP and the sensitivity analysis highlight extensive flood-prone zones. The sensitivity analysis reinforces the AHP findings, confirming that a significant portion of Uvira’s land area is prone to flooding, covering approximately 87.3% and 67.9% of the city, according to the AHP and sensitivity analyses, respectively. However, the sensitivity analysis reveals a broader extent of areas classified as moderately susceptible, which were not fully captured by the AHP method relying on expert-based judgment. Both approaches identify zones of “very high susceptibility” along the major rivers, emphasizing the critical role of proximity to waterbodies in influencing flood susceptibility in Uvira (Figure 6).

Further supporting the AHP-derived map, the sensitivity analysis demonstrates that flood susceptibility gradually decreases westward. The most flood-prone areas are primarily located in the eastern part of the city, along the Lake Tanganyika shoreline, and extend progressively across the urban landscape (Figure 6).

4. Discussion

This study finds that flood susceptibility in Uvira results from a combination of topographic, hydrological, climatic, and human factors. Among the eight factors analyzed, proximity to water bodies emerged as the most influential contributor to flood susceptibility. Over 63% of Uvira is located within approximately 486 m of rivers and Lake Tanganyika, significantly increasing the risk of both riverine and lake-based flooding. This finding aligns with the earlier work of [15], which reported a 2 m rise in Lake Tanganyika’s water level since 2019, exacerbating shoreline flooding risk.

Elevation and slope greatly influence the spatial distribution of flood-prone areas. The predominance of low-lying land (over 90% of the city lies below 851 m) creates natural runoff accumulation zones. This supports the findings of [54] who emphasized the importance of elevation in directing water flow, as gravity drives water from higher to lower terrain. Similarly, ref. [55] observed that flood events frequently occur in low-elevation or downstream areas. Furthermore, 75.5% of the city’s land area lies within flat to gently sloping terrain, exacerbating flood susceptibility. In comparison, ref. [17] reported that 78.2% of Uvira’s land area is prone to flooding when slope is taken into consideration.

The analysis also reveals that areas with high drainage density near rivers correspond to zones of the highest flood susceptibility, highlighting the strong relationship between drainage patterns and flood risk. Elevated drainage density and stream frequency can accelerate surface water flow during intense rainfall events, thereby increasing flood likelihood [18]. Notably, the areas surrounding the Kavimvira River exhibit the highest drainage density and are frequently affected by flooding due to river overflow. For instance, following the torrential rainfall on 12 December 2020, several cases of damage were reported, including four fatalities, two missing persons, two serious injuries, and numerous homes destroyed, particularly in the northern districts where the Kavimvira River overflowed its banks [56].

The city of Uvira receives a significant amount of rainfall, ranging from 985 to 1302 mm annually. However, total annual rainfall may have less impact on flooding compared to the intensity and frequency of extreme rainfall events, which significantly contribute to triggering flash floods. Their contribution to lake overflow flooding, however, is limited, as such events are more closely associated with the slow and steady rise in Lake Tanganyika’s water level. For instance, a high-intensity rainfall event following a wetter-than-usual rainy season caused the severe flash flood of April 2020, which resulted in 43 fatalities, 200 injuries, and the destruction of over 5500 homes and left at least 70,000 people homeless [13]. Additionally, ref. [57] emphasizes that intense rainfall events are among the leading causes of flash floods, often producing catastrophic impacts in affected regions.

Furthermore, nearly the entire city of Uvira is underlain by soil types highly susceptible to flooding, with 84.5% of Mollic Fluvisols and 13.4% of Gleyic Solonchaks. Fluvisols are particularly common in riverine zones, as they typically occur in periodically flooded environments such as alluvial plains, river fans, valleys, and tidal marshes [58]. According to [36], both Fluvisols and Solonchaks are characterized by low permeability, which contributes to elevated flood susceptibility. In contrast, only 2.1% of Uvira’s land area consists of Humic Ferralsols, which are known for their higher permeability and thus lower flood susceptibility [36]. These soils are located in the western part of the city, where elevation is higher, consistent with the observation by [59] that Ferralsols are typically found on stable, flat to gently undulating plateaus.

In Uvira, the areas of elevated TWI values cover up to 55.2% of the city’s total land area. This significantly increases flood susceptibility in these regions, consistent with [60], who reported that higher TWI values are strongly associated with increased flood susceptibility.

The predominance of built-up and agricultural areas significantly increases Uvira’s flood susceptibility. As highlighted by [61], large-scale land reclamation, climate change, and intensified human activities have led to more frequent flood disasters. Ref. [62] further emphasizes that the rapid expansion of urban areas has increased flood vulnerability by expanding impermeable surfaces, which reduce natural infiltration and enhance surface runoff. In contrast, forest cover plays a mitigating role in flood prevention by intercepting stormwater and promoting soil moisture deficits, thereby improving the land’s capacity to absorb rainfall [63].

This study demonstrates that multiple factors contribute to the high frequency of flood events in Uvira, including its geographic location along the Lake Tanganyika shoreline, the presence of a dense river network, and its topographic characteristics. Land use change also significantly influences the severity of flood impacts. For example, ref. [16] notes that the effects of the 2020 flood in Uvira were exacerbated by urban expansion and climate change. Effective planning and mitigation strategies have to integrate multi-factorial assessments to enhance urban resilience in Uvira and other flood-prone cities within the Great Lakes region.

5. Conclusions

This study successfully assessed flood susceptibility in the city of Uvira, located in the eastern Democratic Republic of Congo, using the AHP as an MCDM approach. It integrated eight influential factors identified as key drivers of flooding in the city: distance from water bodies (0.319), elevation (0.083), slope (0.125), rainfall (0.083), drainage density (0.184), soil type (0.118), the topographic wetness index (0.032), and land use/land cover (0.056). Both the individual and combined impacts of these factors on flood susceptibility were evaluated. The findings indicate that proximity to water exerts the strongest influence on flood susceptibility in Uvira, followed by drainage density and slope, despite the relevance of all considered factors.

Despite challenges encountered during the research process (the lack of historical data and relevant preliminary research directly addressing the topic), this study successfully identified flood-prone areas in Uvira, thereby filling a critical knowledge gap. However, future studies should incorporate infrastructure and socio-economic variables to enable a more comprehensive flood risk assessment, including the evaluation of population vulnerability and exposure. Such an integrated approach would enhance Uvira’s urban resilience by informing the development of effective mitigation strategies and reflecting the adaptive capacities of local communities.

The study was conducted in a data-scarce context, which is a common limitation in the Democratic Republic of Congo. Although multiple flood events have occurred in Uvira, no official records were available to validate the study’s results. Future research should aim to incorporate such data if they become available. Furthermore, researchers working in collaboration with local authorities and organizations should establish systematic flood event documentation, including spatial mapping, to build a local database in a city where flash floods are both frequent and destructive.

While the results of this study should be supplemented by further analysis, some preliminary recommendations can be proposed. First, the construction of residential housing should be strictly regulated in areas adjacent to rivers, where flood susceptibility is extremely high. Establishing buffer zones and implementing mitigation measures such as reforestation and the introduction of mangroves in selected areas near rivers and along the Lake Tanganyika shoreline should be prioritized. Second, urban expansion should be directed toward areas with lower flood susceptibility, offering potential relocation zones for portions of the city’s population. However, thorough preliminary assessments are essential to identify the most suitable sites for such development.

This study provides valuable insights into the factors driving flood occurrence in Uvira and the locations most at risk. The findings offer a practical foundation for urban planners, policymakers, decision-makers, and local communities to design and implement evidence-based flood mitigation strategies.