High-Resolution Urban Flood Susceptibility Mapping in Miami-Dade County: An AHP-Based GIS and Multi-Criteria Decision Analysis Approach

Abstract

Urban flooding is prevalent in low-lying, coastal regions, where subtle topographic variation, shallow groundwater, and impervious surfaces govern inundation dynamics. This study evaluates urban flood susceptibility across Miami-Dade County by integrating flood-conditioning factors, including elevation, slope, rainfall, land use/land cover, distance to roads and open water, stream power index (SPI), topographic wetness index (TWI), groundwater depth, and flow accumulation within an Analytical Hierarchy Process (AHP)-based weighted overlay framework. The AHP-derived weights demonstrated strong consistency (consistency ratio = 0.022) and were applied to reclassify each conditioning factor into five flood susceptibility classes—very low to very high. The model performance was evaluated using the Federal Emergency Management Agency (FEMA) flood zone, and the findings demonstrated that the AHP-based framework effectively differentiates flood susceptibility at a fine urban scale, achieving strong predictive performance; area under the Curve (AUC) = 0.85. The results also reveal pronounced spatial variability in flood susceptibility, with northeastern urbanized areas, particularly in Hialeah, Miami Gardens, Miami Lakes, and Downtown Miami, exhibiting higher susceptibility compared to the northwestern Everglades region. Overall, this study presents a robust urban flood susceptibility framework that supports improved flood risk assessment and decision-making in complex urban coastal environments.

1. Introduction

Global warming and rapid urbanization in recent decades have multiplied the risk of flooding worldwide [1,2]. It affects the frequency and intensity of rainfall patterns, resulting in intense flooding in many regions across the world [3,4,5]. The sixth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) emphasized that more land areas will experience flooding in the coming decades [6,7]. In addition, a global temperature rise of 1.5 °C is expected to intensify this issue further, with heavier rainfall, rising sea levels, frequent storm surges, widespread deforestation, and continued urban expansion all contributing to an increasingly dire flood scenario in the coming years [8]. Floods are the most frequent and costliest natural disasters in recent decades, and the severity and magnitude of flooding are escalating each year [9,10]. According to recent statistics, floods are responsible for 20% of losses caused by natural disasters worldwide [11,12,13]. Additionally, it has been projected that flood likelihood on a global scale will be threefold by 2050 [14]. As these impacts grow in scale and frequency, the implementation of proactive flood mitigation strategies has become more critical than ever [15,16]. Therefore, it is important to comprehensively understand the factors contributing to flooding, evaluate flood susceptibility mapping, and prioritize the identification of flood-susceptible zones to minimize future flood-related casualties worldwide.

Urban areas face tremendous flooding risks, and the challenges are intensifying due to rapid and widespread urban growth [17]. Urban flooding occurs when heavy rainfall, storm surges or inadequate drainage capacity lead to the accumulation of water in developed areas [18]. As cities continue to expand, according to United Nations projections, approximately 68% of the total global population will reside in urban areas by 2050 [19]. Alarmingly, research indicates that the pace of urban expansion is nearly twice the rate of population growth [20]. Therefore, urban flooding presents a particularly complex and rapidly intensifying challenge influenced by heavy rainfall, expanding impervious surfaces, and altered hydrological cycles. Urban areas are dominated by impervious surfaces such as asphalt, cement, and concrete, which work as a barrier and prevent water from seeping into the ground [21,22]. High population density and widespread urban expansion increase these surfaces, leading to greater surface runoff and water overflow during heavy rainfall [23,24]. Every year, the damage due to urban flooding increases exponentially due to urban growth in this low-lying region [25,26]. Therefore, the integration of urban flood susceptibility mapping and mitigation strategies is important for prioritizing flood loss mitigation [27]. As urban flooding is becoming more frequent in Miami-Dade County and will increase exponentially, urban flood susceptibility mapping is gaining more attention in the research community [28].

Flood susceptibility mapping involves assessing the likelihood of flood occurrences within a specific area [29,30]. These maps play a critical role in identifying regions that are highly susceptible to flooding [31,32]. Flood-susceptible regions need to be identified not only to implement effective strategies for flood management but also to minimize future flood-related losses [33]. Miami-Dade County is located in the southern part of the Conterminous USA. Its proximity to the coast, low-lying topography and rapid urbanization make it susceptible to urban flooding [34,35]. Moreover, the county’s flat neighborhood and shifting rainfall pattern present significant challenges for effective storm drainage management in tackling a large volume of stormwater during intense rainfall [36,37]. As urban flooding is a prominent disturbance to this region, and the major drivers are projected to increase in the coming years [38]; thus, tackling this problem requires a comprehensive urban flood susceptibility analysis to better understand the spatial extent of urban flooding.

Different approaches have been adopted to assess and understand urban flood susceptibility in different regions worldwide. A variety of approaches, such as SWAT (Soil and Water Assessment Tool) [39], bivariate statistical modeling [40,41], and multivariate logistic regression [42], have been employed in the literature to analyze urban flood susceptibility. The soil and water assessment tool is a rainfall runoff model that simulates how rainfall generates surface runoff by incorporating key processes such as infiltration, evapotranspiration, percolation and channel flow [43,44]. Given that floods are fundamentally spatial, geographical information systems (GIS) and remote sensing techniques play a vital role in identifying flood-prone areas [45]; these techniques have been widely used to determine flood-prone areas in different regions [46,47,48,49]. Various studies have highlighted the advantages of GIS and remote sensing in flood susceptibility mapping at different spatial scales [50], and these tools are extremely helpful in detecting the spatial aspect of floods for management activities [51,52]. Researchers have also utilized numerous training-based GIS methods [53,54] and data-driven models combined with GIS, such as frequency ratio [55,56,57], logistic regression [42,58,59,60], and fuzzy logic [40,61]. The majority of these methods are effective at detecting flood-prone areas; however, they have some limitations in determining the relationship between variables [62]. To address this problem, researchers have employed machine learning algorithms such as support vector machines [63,64,65,66,67], deep learning [68,69], decision trees [55,70], and artificial neural networks [64,71,72,73,74,75]. However, machine learning methods are complex and often demand high-performance computing resources [62,76]. In complex and data-scarce urban regions, a more efficient approach, such as the AHP combined with GIS, is necessary. AHP is regarded as a cost-effective, simple and understandable technique for flood susceptibility mapping [77,78,79,80]. In addition, remote sensing data such as LiDAR (Light Detection and Ranging) can penetrate vegetation cover and provide high-resolution bare ground digital elevation model (DEM) data, which facilitates precise urban flood mapping [81,82,83,84,85]. This study focuses on a high-resolution and data-intensive assessment of urban flood susceptibility by integrating diverse geospatial layers, including land use/land cover, surface morphology, groundwater depth and key hydrometeorological variables that capture the fine-scale, infrastructure-driven flooding dynamics that are increasing rapidly across urbanized coastal regions in Miami-Dade County. The research question for the study is as follows:

What are the key factors contributing to urban flood susceptibility in Miami-Dade County, and which areas are more susceptible to flooding?

Despite the widespread application of GIS-based AHP and MCDA in urban flood susceptibility mapping, many existing assessments remain constrained by regional-scale spatial resolution, generalized land-use classifications, and limited representation of urban infrastructure and shallow groundwater interactions [86,87]. In such settings, flooding is often driven not only by intense precipitation or fluvial overflow but also by complex interactions among microtopography, impervious surface connectivity, stormwater, and shallow groundwater dynamics [88]. Conventional flood susceptibility modeling approaches, particularly those relying on coarse-resolution regional-scale hydrological models, often oversimplify these localized processes, resulting in inadequate representation of microscale inundation and infrastructure-induced water accumulation [89,90]. Therefore, localized flood drivers and microscale infrastructure-driven inundation processes in low-lying coastal regions are often insufficiently captured [91,92,93,94]. Moreover, prior studies have largely emphasized watershed or basin-scale assessments, which may overlook fine-scale flood susceptibility in complex urban settings. As a result, a significant gap persists in fine-scale urban flood susceptibility research that explicitly incorporates spatial heterogeneity and the infrastructure-driven nature of flooding in densely developed regions.

To address this gap, the present study integrates rigorously validated AHP-based MCDA with high-resolution spatial data, which enhances both methodological reliability and spatial precision. The proposed framework also improves upon traditional regional-scale models by enabling a very high-resolution (3 m) LiDAR-derived DEM with spatially explicit groundwater depth, land-use intensity, and urban morphological characteristics. Moreover, it enables the identification of localized risk zones that support targeted flood mitigation and infrastructure planning. Beyond the Miami-Dade case study, the methodological contribution of this work lies in demonstrating a transferable and scalable methodology for adapting MCDA-based flood susceptibility mapping to low-relief, coastal urban cities, where traditional floodplain approaches and coarse-resolution models fail to resolve localized inundation dynamics. The novelty of this work lies in the comprehensive integration of a fine-scale topography, groundwater conditions, and urban infrastructure characteristics within a robust and rigorously validated GIS–MCDA framework. Moreover, the proposed framework provides a practical, data-efficient decision support tool for other rapidly urbanizing coastal regions facing similar flood management challenges. The key contribution of this research lies in incorporating fine-scale topographic, urban infrastructure, and other flood conditioning factors and the capability of effectively capturing micro-scale, infrastructure-driven flood susceptible zones in low-lying coastal environments.

The objectives of this study are to:

• Develop an urban flood susceptibility map for Miami-Dade County to evaluate small-scale spatial variability;
• Identify areas of very high and low flood susceptibility for targeted planning and mitigation.

2. Materials and Methods

2.1. Study Area

Miami-Dade County, located in the southeastern part of Florida, is located at a latitude of 25.5516° N and a longitude of 80.6327° W and covers an area of approximately 2431 square miles. The county is surrounded by the Atlantic Ocean to the east, Everglades National Park to the west, Broward County to the north and the Florida Keys to the south. It has a tropical monsoon climate characterized by dry and wet seasons. The wet season typically spans from May to October, and the county experiences high temperatures, high humidity and heavy rainfall during this time. The wet season also coincides with the hurricane season, which is from June to November and brings intense rainfall and storm surges to the region. The dry season usually spans from November through April, and the weather is generally cooler with less rainfall and lower humidity. The average annual rainfall in Miami-Dade County typically ranges from 55 to 65 inches. The county has 2.6 million people, which makes it one of the most populated regions across Florida and the seventh most populous county in the United States [95].

The region shown in Figure 1 experiences complex and multifaceted challenges influenced by a combination of natural and human factors, including low-lying neighborhoods, aggressive industrialization, and massive economic growth [96,97]. Besides, the region’s groundwater is close to the surface, and a high water table makes it easier for the land to become saturated with heavy rainfall in a fleeting period, contributing to flooding every year [98,99]. In addition, global warming combined with shifts in rainfall patterns has caused more frequent and intense rainfall events and prolonged inundation [100]. Furthermore, changes in land use/land cover and rapid urban development escalate the threat of developing impervious surfaces and overwhelmed drainage systems during heavy rainfall [101]. Therefore, understanding these unique interactions between natural and man-made factors influencing urban flood susceptibility is important for adapting effective mitigation strategies and enhancing urban flood resilience in communities and the environment.

2.2. Data Sources

The data sources include high-resolution (resolution = 3 m) digital elevation model (DEM) data obtained from the Light Detection and Ranging (LiDAR) website (https://www.usgs.gov/tools/lidarexplorer accessed on 14 February 2026). A digital elevation model (DEM) determines the elevation of the ground relative to any vertical datum and offers a good opportunity to extract topographic information that is important for flood susceptibility mapping [12,102]. Land use/land cover plays an important role in urban flooding, and the current land use/land cover data were obtained from the Miami-Dade Open Data Portal website (https://gis-mdc.opendata.arcgis.com/ accessed on 14 February 2026). Road network, groundwater and open water polygons were collected from the Miami.Gov website. Rainfall return period data were collected from 18 different stations inside Miami-Dade County and interpolated to obtain the rainfall scenario from the NOAA website (https://hdsc.nws.noaa.gov/pfds/ accessed on 14 February 2026).

The first step in urban flood susceptibility mapping requires the selection of flood conditioning factors depending on the topography and hydrological characteristics of the study area [103,104]. In this study, a set of flood conditioning factors was selected, and the datasets included elevation, slope, rainfall, stream power index (SPI), topographic wetness index (TWI), land use/land cover, groundwater depth, flow accumulation, distance from roads, and distance from open water. All the flood conditioning factors are presented in Figure 2, Figure 3, Figure 4, Figure 5 and Figure 6 and described in Table 1.

Table 1. Key flood conditioning factors.

Slope	Steeper slopes lead to faster runoff, reducing the available time for the soil to absorb water [105]. Conversely, flatter slopes lead to water accumulation and are more susceptible to flooding [106,107]. Most of the county’s slopes are below 4%, as illustrated in Figure 2.
Elevation	Elevation data have been derived from digital elevation models (DEMs), and elevation is negatively correlated with flooding [108,109]. As the elevation of Miami-Dade County is low, as illustrated in Figure 2, it is more susceptible to flooding.
Flow accumulation	Flow accumulation is the number of upstream cells that drain into a particular location in a digital elevation model [110,111]. Areas with high flow accumulation experience greater flow volume and pressure and are prone to flooding during heavy rainfall [112,113].
SPI	Higher SPI values indicate greater erosive power and are more susceptible to flooding [114,115]. The stream power index is calculated using the Miami-Dade County digital elevation model, as illustrated in Figure 3. The SPI was calculated using Equation (1) [116,117]:
	$\mathbf{S}\mathbf{P}\mathbf{I}={\mathbf{A}}_5\times \mathit{t}\mathit{a}\mathit{n}\left(\mathsf{\beta }\right)$	(1)
TWI	The TWI determines the water-saturated areas and spatial distribution of water on the surface and underground [118,119]. It can be calculated using Equation (2):
	TWI = Ln (As/tan β)	(2)
	where ${\mathbf{A}}_{\mathit{s}}$ = the upslope contributing area and $\mathsf{\beta }$ = the local slope angle. High TWI values represent favorable areas for water accumulation and higher susceptibility to flooding [120].
Land use/land cover	Urban expansion leads to more impervious surfaces, which reduce water infiltration capacity and increase surface runoff, thereby increasing flood risk [121,122]. In this study, land use/land cover was classified into five categories: urban/suburban, water/wetland, barren land, upland nonforest, and agricultural as illustrated in Figure 4.
Distance from road	The greater the impervious surface area is, the higher the risk of flooding [123,124,125]. The road network data were collected from the florida.gov website, and the road distance was also calculated using the Euclidean distance tool in ArcGIS Pro 3.5 software and classified into 5 different classes as illustrated in Figure 5.
Distance from water	When a flood starts to overflow, the areas closer to the waterbodies are considered highly susceptible to inundation [126,127,128]. The distance from the open water layer was determined by the Euclidean distance tool in ArcGIS Pro.
Rainfall	Flooding in low-lying areas is mostly caused by surface runoff from extensive rainfall [129,130]. We collected rainfall data from the NOAA precipitation depth duration curve. We considered rainfall recurrence periods over 1 year-24 h for flood susceptibility mapping. The rainfall maps were prepared using the station locations and rainfall data across the study area and interpolated via the IDW interpolation method as illustrated in Figure 6.
Groundwater depth	The groundwater table in Miami-Dade County is high and very close to the surface, which leaves less space for water absorption during heavy rainfall [131,132]. This limited vertical space for infiltration causes rain water to accumulate rapidly during heavy rainfall, resulting in reduced infiltration, higher surface runoff, and an increased likelihood of flooding.

Table 1. Key flood conditioning factors.

Slope	Steeper slopes lead to faster runoff, reducing the available time for the soil to absorb water [105]. Conversely, flatter slopes lead to water accumulation and are more susceptible to flooding [106,107]. Most of the county’s slopes are below 4%, as illustrated in Figure 2.
Elevation	Elevation data have been derived from digital elevation models (DEMs), and elevation is negatively correlated with flooding [108,109]. As the elevation of Miami-Dade County is low, as illustrated in Figure 2, it is more susceptible to flooding.
Flow accumulation	Flow accumulation is the number of upstream cells that drain into a particular location in a digital elevation model [110,111]. Areas with high flow accumulation experience greater flow volume and pressure and are prone to flooding during heavy rainfall [112,113].
SPI	Higher SPI values indicate greater erosive power and are more susceptible to flooding [114,115]. The stream power index is calculated using the Miami-Dade County digital elevation model, as illustrated in Figure 3. The SPI was calculated using Equation (1) [116,117]:
	$\mathbf{S}\mathbf{P}\mathbf{I}={\mathbf{A}}_5\times \mathit{t}\mathit{a}\mathit{n}\left(\mathsf{\beta }\right)$	(1)
TWI	The TWI determines the water-saturated areas and spatial distribution of water on the surface and underground [118,119]. It can be calculated using Equation (2):
	TWI = Ln (As/tan β)	(2)
	where ${\mathbf{A}}_{\mathit{s}}$ = the upslope contributing area and $\mathsf{\beta }$ = the local slope angle. High TWI values represent favorable areas for water accumulation and higher susceptibility to flooding [120].
Land use/land cover	Urban expansion leads to more impervious surfaces, which reduce water infiltration capacity and increase surface runoff, thereby increasing flood risk [121,122]. In this study, land use/land cover was classified into five categories: urban/suburban, water/wetland, barren land, upland nonforest, and agricultural as illustrated in Figure 4.
Distance from road	The greater the impervious surface area is, the higher the risk of flooding [123,124,125]. The road network data were collected from the florida.gov website, and the road distance was also calculated using the Euclidean distance tool in ArcGIS Pro 3.5 software and classified into 5 different classes as illustrated in Figure 5.
Distance from water	When a flood starts to overflow, the areas closer to the waterbodies are considered highly susceptible to inundation [126,127,128]. The distance from the open water layer was determined by the Euclidean distance tool in ArcGIS Pro.
Rainfall	Flooding in low-lying areas is mostly caused by surface runoff from extensive rainfall [129,130]. We collected rainfall data from the NOAA precipitation depth duration curve. We considered rainfall recurrence periods over 1 year-24 h for flood susceptibility mapping. The rainfall maps were prepared using the station locations and rainfall data across the study area and interpolated via the IDW interpolation method as illustrated in Figure 6.
Groundwater depth	The groundwater table in Miami-Dade County is high and very close to the surface, which leaves less space for water absorption during heavy rainfall [131,132]. This limited vertical space for infiltration causes rain water to accumulate rapidly during heavy rainfall, resulting in reduced infiltration, higher surface runoff, and an increased likelihood of flooding.

2.3. Methodology

Researchers and decision makers have adopted different mitigation strategies to prevent flood loss. One strategy introduces the idea of determining flood-prone areas by evaluating all the driving factors that are responsible for flooding for a specific region [133,134]. In addition, mapping flood-susceptible areas is useful for preparing early warning systems, damage mitigation and future flood adaptation strategies [135,136].

To obtain all the flood conditioning layers, LiDAR DEM data were collected, and the resolution was 3 m. Multiple thematic layers were prepared using the LiDAR DEM (digital elevation model). These datasets were collected from various sources and resized to 3 m. All the conditioning factors were processed and reclassified in ArcGIS Pro, after which all the factors were ranked using AHP, as illustrated in Table 2, Table 3 and Table 4. AHP is a pairwise ranking matrix, and researchers have frequently used it to rank the importance of the parameters [137]. In this process, each factor was assigned an arithmetic number between 1 and 5 based on its association with the other factor. Following this, a consistency ratio was computed to evaluate the reliability of the pairwise comparisons. After that, all the parameters were classified into 5 different classes from 1 to 5: 1—very low susceptible, 2—low susceptible, 3—medium susceptible, 4—highly susceptible, and 5—very highly susceptible, as demonstrated in Table 4. All these classifications were based on the Jenkins natural break, which is a popular method for classifying flood conditioning parameters [138]. This method is used for grouping data based on their intrinsic distribution to determine the best value settings in different classes [139,140,141]. It minimizes the average deviation from the class mean while maximizing the deviation of each class from the other groups [142]. In addition, this method also seeks to limit the variance within classes and maximize the variance between classes [143]. In comparison with equal interval classification, which divides the data ranges into uniform class widths and may oversimplify highly heterogeneous datasets. In contrast, quantile classification assigns an equal number of observations to each class regardless of their underlying variability, which may result in grouping dissimilar susceptibility values together while separating similar ones into different categories.

After that, all flooding factors were multiplied by their weight and summed in a raster calculator to obtain the final layer, as illustrated in Figure 7. Although several variables, including elevation, slope, flow accumulation, SPI, and TWI, were derived from the same LiDAR-based DEM, each parameter represents a distinct physical process governing flood occurrence. Elevation highlights regional drainage potential, slope reflects runoff velocity, flow accumulation indicates preferential flow pathways, SPI governs erosive capacity, and TWI emphasizes soil moisture and saturation tendencies. Therefore, these variables provide complementary rather than redundant information on infrastructure-driven flood processes. While some degree of correlation is expected among topographic variables, their combined application enables a more comprehensive representation of surface hydrological dynamics in low-lying urban landscapes. Moreover, the AHP weighting framework, supported by a low consistency ratio (CR = 0.022), ensured balanced contributions among interrelated variables and minimized the dominance of any single factor. Moreover, similar multi-parameter topographic applications have also been widely adopted in previous urban flood susceptibility studies, further validating the methodological design.

The formula for weighted overlay in urban flood susceptibility mapping is illustrated in Equation (3):

Flood susceptibility = R_elevation × W_elevation + R_slope × W_slope + R_{flow accumulation} × W_{flow accumulation} + R_rainfall × W_rainfall + R_{groundwater depth} × W_{groundwater depth} + R_LULC × W_LULC + R_{distance from road} × W_{distance from road} + R_{distance from open water} × W_{distance from open water} + R_SPl × W_SPI + R_TWI × W_TWI

(3)

where R is the ranking value and W is the weighted layer.

The final flood susceptibility map was validated using FEMA flood zone data to determine how effectively it captured the spatial extent of known flood-prone areas. In addition, the model’s predictive performance was further evaluated using the area under the curve (AUC) metric to ensure the reliability and robustness of the generated susceptibility map.

To assess urban flood susceptibility in Miami-Dade County, it is important to evaluate the relationship between flood contributing factors and flood events. Numerous studies have shown that urban flood susceptibility is a complex phenomenon and is potentially influenced by topographical, hydrological, and geographical contributing factors [144,145]. Given the county’s low-lying flat topography, dense urban infrastructures, and monsoon climate, certain parameters become especially important, including precipitation, land use/land cover, and groundwater depth. In this study, 10 different flood conditioning factors were selected based on previous research, as illustrated in Figure 3, based on the effectiveness of the parameters and their relevance with respect to the study area. Each of these selected factors was analyzed to generate the current urban flood susceptibility map.

The selection of flood conditioning factors was based on the relevance of urban flood processes, regional hydrological characteristics, data availability, and consistency with previous research. Several flood conditioning factors, including elevation, slope, rainfall, land use/land cover, groundwater depth, roads/open water distance, and flow accumulation, were prioritized due to their strong influence on surface runoff generation, water accumulation, and drainage efficiency in highly urbanized environments. Soil type was excluded due to the region’s relatively uniform limestone geology and highly permeable surficial aquifer system. Besides, drainage data are partially represented by land use/land cover, road density and topographic attributes. The inclusion and exclusion of the factors thereby reflect both the hydrological characteristics of the study area and practical data considerations.

The higher weight assigned to rainfall and land use/land cover in the AHP ranking reflects the county’s sensitivity to heavy rainfall events and widespread urban expansion. This prioritization is consistent with recent research, which similarly identifies these factors as dominant drivers of flood susceptibility [9,96,146,147,148]. Rainfall return period-derived data from NOAA stations highlighted localized gradients across the county, indicating that specific neighborhoods are more likely to experience short-duration heavy rainfall. This also aligns with recent climate projection studies reporting that extreme rainfall intensities in South Florida are expected to increase under warming conditions [149,150,151].

2.4. Analysis

The AHP is a flexible mathematical model and has been reconstructed by researchers because of its several advantages in critical decision-making [152,153]. It consists of decision stages in which the values are assigned and alternative values are determined according to the criteria chosen to manage the decision-making process. The first stage in AHP is to determine the hierarchical structure and create a pairwise comparison matrix, where each criterion is compared with every other criterion [154]. A consistency ratio is also calculated using the random index to determine whether the comparisons are logically consistent [155]. To fulfill the criteria, AHP utilizes a scale from 1 to 9, as illustrated in Table 2, where 1 means equal importance and 9 means extreme importance of one criterion over another. Using this scale, the pairwise comparison matrix is calculated.

Table 2. Scales for pairwise comparison in AHP [156,157,158].

Importance (Scores)	Definition
1	Equal importance
3	Moderate importance of one over another
5	Strong importance of one over another
7	Very strong importance of one over another
9	Extreme importance of one over another
2, 4, 6, 8	Intermediate values
Reciprocals	Reciprocals for inverse comparison

Table 2. Scales for pairwise comparison in AHP [156,157,158].

Importance (Scores)	Definition
1	Equal importance
3	Moderate importance of one over another
5	Strong importance of one over another
7	Very strong importance of one over another
9	Extreme importance of one over another
2, 4, 6, 8	Intermediate values
Reciprocals	Reciprocals for inverse comparison

Let $A=\left[aij\right]$ be the $n\times n$ pairwise comparison matrix, where n is the number of criteria and each element $a$_ij expresses how important criterion $\mathrm{i}$ is relative to criterion $\mathrm{j}$. The matrix takes the form

$$\mathrm{A}=\left[\begin{array}{ccc}{a}_{11}& a_{12}\dots & a_{1n}\\ a_{21}& a_{22}\dots & a_{2n}\\ a_{n1}& a_{n2}\dots & a_{nn}\end{array}\right]$$

(4)

The diagonal elements are always equal to 1.

$$a_{\mathrm{i}\mathrm{i}}=1 \mathrm{for} \mathrm{all} i$$

The matrix is also reciprocal, meaning that if criterion i is judged to be more important than criterion j, their inverse relationship is automatically defined:

$$a_{ji}={\displaystyle \frac{1}{a_{ij}}}$$

(5)

The weight, w, is obtained from the principal eigenvector of the matrix $A$:

$${\mathrm{A}}_w={\mathsf{\lambda }}_{\mathrm{m}\mathrm{a}\mathrm{x}} \mathrm{w}$$

(6)

where ${\mathsf{\lambda }}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ is the maximum eigenvalue of A, and w is the corresponding eigenvector. The eigenvector is then normalized so that the weight sums to 1.

Human judgments hold imperfection, so AHP evaluates the consistency of pairwise comparisons. First, the consistency ratio is computed as [159], as shown in Equation (7):

$$CI ={\displaystyle \frac{\left({\lambda }_{\left\{max\right\}}- n\right)}{\left(n - 1\right)}}$$

(7)

where ${\lambda }_{\left\{max\right\}}$ is the maximum eigenvalue of the comparison matrix and n is the number of criteria.

Here, λ_max ≈ 10.2960, and n = 10

After that, the consistency ratio is calculated by comparing the CI with the random index (RI), which is the average CI of randomly generated reciprocal matrices of size n:

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

(8)

According to Saaty [160], the matrix is considered acceptably consistent if:

$$CR<0.10$$

In this study, n = 10, and we utilized Saaty’s random index RI = 1.49; therefore, the calculated values are

$$CI\approx 0.03289; CR\approx 0.02207$$

(9)

Since the CR value is 0.022, the pairwise comparisons are consistent, and the derived weights are statistically reliable for use in flood susceptibility mapping.

The AHP pairwise comparison matrix in Table 3 represents the relative importance of ten flood conditioning factors influencing urban flood susceptibility in Miami-Dade County. The pairwise comparisons presented in this study are based on a combination of expert knowledge in hydrology and urban flood mapping, together with an extensive literature review of relevant studies on urban flood susceptibility in coastal environments. Besides, the weighting process also accounted for local hydrological characteristics, including shallow groundwater conditions, low-lying flat terrain, and dense urban infrastructure. The relative importance of each conditioning factor was evaluated through systematic pairwise comparison using Saaty’s nine-point scale, considering their influence on surface imperviousness, groundwater interactions, and infrastructure constraints in Miami-Dade County. To ensure logical consistency, the resulting comparison matrix was carefully examined and validated using the consistency ratio (CR), which yielded a value of 0.022, emphasizing excellent internal coherence.

Table 3. Pairwise comparison matrix for flood conditioning factors.

Matrix	Slope	Elevation	Distance from Water	Distance from Road	Rainfall	Groundwater Depth	SPI	TWI	Flow Accumulation	Land Use Land Cover
Slope	1	1	1/3	1/3	1/9	1/7	1/3	1/3	1/3	1/9
Elevation	1	1	1/3	1/3	1/9	1/7	1/3	1/3	1/3	1/9
Distance from water	3	3	1	1	1/7	1/5	1	1	1	1/7
Distance from road	3	3	1	1	1/7	1/5	1	1	1	1/7
Rainfall	9	9	7	7	1	3	7	7	7	1
Groundwater depth	7	7	5	5	1/3	1	5	5	5	1/3
SPI	3	3	1	1	1/7	1/5	1	1	1	1/7
TWI	3	3	1	1	1/7	1/5	1	1	1	1/7
Flow accumulation	3	3	1	1	1/7	1/5	1	1	1	1/7
Land use/land cover	9	9	7	7	1	3	7	7	7	1

Table 3. Pairwise comparison matrix for flood conditioning factors.

Matrix	Slope	Elevation	Distance from Water	Distance from Road	Rainfall	Groundwater Depth	SPI	TWI	Flow Accumulation	Land Use Land Cover
Slope	1	1	1/3	1/3	1/9	1/7	1/3	1/3	1/3	1/9
Elevation	1	1	1/3	1/3	1/9	1/7	1/3	1/3	1/3	1/9
Distance from water	3	3	1	1	1/7	1/5	1	1	1	1/7
Distance from road	3	3	1	1	1/7	1/5	1	1	1	1/7
Rainfall	9	9	7	7	1	3	7	7	7	1
Groundwater depth	7	7	5	5	1/3	1	5	5	5	1/3
SPI	3	3	1	1	1/7	1/5	1	1	1	1/7
TWI	3	3	1	1	1/7	1/5	1	1	1	1/7
Flow accumulation	3	3	1	1	1/7	1/5	1	1	1	1/7
Land use/land cover	9	9	7	7	1	3	7	7	7	1

Several factors, including distance to roads, SPI, TWI, and flow accumulation, received closely similar weights, highlighting their comparable roles in influencing water accumulation and drainage patterns in the Miami-Dade region. These similarities emphasize balanced contributions among interrelated variables, and are consistent with previous urban flood susceptibility assessments in low-lying coastal environments [161,162].

Among all criteria, rainfall and land use/land cover were assigned the highest relative importance, highlighting their dominant role in driving urban flooding in this low-lying, highly urbanized environment. These factors were consistently ranked highest compared to topographic variables such as slope and elevation in low-lying urban environments [163]. The groundwater depth also received a higher rank, representing the critical role of the shallow water table in limiting infiltration during rainfall events. The literature also supports this evidence [164,165,166]. Intermediate importance was assigned to distance from roads and water bodies, stream power index, topographic wetness index, and flow accumulation, representing their contribution to localized drainage behavior. In contrast, slope and elevation were ranked with comparatively lower weights, which is consistent with similar studies [167]. The weights and susceptibility levels are illustrated in Figure 4.

Table 4. Flood conditioning factors with their weighted value based on AHP, reclassified proposed weight and susceptibility levels.

Parameters	Level	Susceptibility Level	Proposed Weight	Weight Based on AHP
Elevation (meter)	(−3.961)–(−0.663)	Very high	5	1.94%
	(−0.662)–0.628	High	4
	0.629–1.488	Medium	3
	1.489–2.241	Low	2
	2.242–5.182	Very Low	1
Slope (%)	0.001–0.882	Very high	5	1.94%
	0.883–2.294	High	4
	2.295–4.471	Medium	3
	4.472–8.412	Low	2
	8.413–15	Very low	1
Rainfall (inches)	3.962–4.174	Very Low	1	28.74%
	4.175–4.288	Low	2
	4.289–4.395	Medium	3
	4.396–4.509	High	4
	4.51–4.64	Very High	5
Groundwater depth (meter)	2.08–5.48	Very high	5	16.76%
	5.49–7.69	High	4
	7.70–10.59	Medium	3
	10.6–14.81	Low	2
	14.82–27.64	Very low	1
Distance from road (meter)	0.01–2190.1	Very high	5	4.38%
	2190.11–5727.96	High	4
	5727.97–9855.46	Medium	3
	9855.47–14,488.36	Low	2
	14,488.37–21,479.84	Very Low	1
LULC	Urban and Suburban	Very High	5	28.74%
	Water/Wetlands	High	4
	Barren land/rangeland	Medium	3
	Upland Nonforest	Low	2
	Agriculture/Upland Forest	Very Low	1
SPI (meter)	(−4.2)–(−3.63)	Very Low	1	4.38%
	(−3.64)–(−2.5)	Low	2
	(−2.51)–(−1.31)	Medium	3
	(−1.32)–(−0.7)	High	4
	(−0.71)–0.25	Very High	5
TWI (meter)	0.915–2.215	Very Low	1	4.38%
	2.216–2.642	Low	2
	2.643–3.332	Medium	3
	3.333–3.962	High	4
	3.963–6.096	Very High	5
Flow Accumulation (meter)	0–0.09	Very Low	1	4.38%
	0.1–0.18	Low	2
	0.19–0.35	Medium	3
	0.36–0.57	High	4
	0.57–2.42	Very High	5
Distance from open water (meter)	0.01–1188.12	Very high	5	4.38%
	1188.13–3132.31	High	4
	3132.32–5670.56	Medium	3
	5670.57–8694.86	Low	2
	8694.87–13,771.36	Very Low	1

Table 4. Flood conditioning factors with their weighted value based on AHP, reclassified proposed weight and susceptibility levels.

Parameters	Level	Susceptibility Level	Proposed Weight	Weight Based on AHP
Elevation (meter)	(−3.961)–(−0.663)	Very high	5	1.94%
	(−0.662)–0.628	High	4
	0.629–1.488	Medium	3
	1.489–2.241	Low	2
	2.242–5.182	Very Low	1
Slope (%)	0.001–0.882	Very high	5	1.94%
	0.883–2.294	High	4
	2.295–4.471	Medium	3
	4.472–8.412	Low	2
	8.413–15	Very low	1
Rainfall (inches)	3.962–4.174	Very Low	1	28.74%
	4.175–4.288	Low	2
	4.289–4.395	Medium	3
	4.396–4.509	High	4
	4.51–4.64	Very High	5
Groundwater depth (meter)	2.08–5.48	Very high	5	16.76%
	5.49–7.69	High	4
	7.70–10.59	Medium	3
	10.6–14.81	Low	2
	14.82–27.64	Very low	1
Distance from road (meter)	0.01–2190.1	Very high	5	4.38%
	2190.11–5727.96	High	4
	5727.97–9855.46	Medium	3
	9855.47–14,488.36	Low	2
	14,488.37–21,479.84	Very Low	1
LULC	Urban and Suburban	Very High	5	28.74%
	Water/Wetlands	High	4
	Barren land/rangeland	Medium	3
	Upland Nonforest	Low	2
	Agriculture/Upland Forest	Very Low	1
SPI (meter)	(−4.2)–(−3.63)	Very Low	1	4.38%
	(−3.64)–(−2.5)	Low	2
	(−2.51)–(−1.31)	Medium	3
	(−1.32)–(−0.7)	High	4
	(−0.71)–0.25	Very High	5
TWI (meter)	0.915–2.215	Very Low	1	4.38%
	2.216–2.642	Low	2
	2.643–3.332	Medium	3
	3.333–3.962	High	4
	3.963–6.096	Very High	5
Flow Accumulation (meter)	0–0.09	Very Low	1	4.38%
	0.1–0.18	Low	2
	0.19–0.35	Medium	3
	0.36–0.57	High	4
	0.57–2.42	Very High	5
Distance from open water (meter)	0.01–1188.12	Very high	5	4.38%
	1188.13–3132.31	High	4
	3132.32–5670.56	Medium	3
	5670.57–8694.86	Low	2
	8694.87–13,771.36	Very Low	1

3. Results

The urban flood susceptibility map in Figure 8 represents spatial variability in flood-prone areas across Miami-Dade County. The flood susceptibility index shown in the map derived from GIS-based weighted overlay analysis ranges from low (green) to high (red), capturing fine-scale spatial variability influenced by topography, land use changes and hydrometeorological factors. High susceptibility clusters are primarily concentrated in the central and eastern parts of the county, indicated by orange to red zones. These areas consist of densely urbanized neighborhoods, extensive impervious cover, and shallow groundwater tables. All of these factors are responsible for enhancing surface runoff accumulation during intense rainfall events. Notably, elevated susceptibility is also observed toward the southeastern part alongside the coast, where a higher concentration of urban infrastructures and impervious surfaces elevates the risk of urban flooding. In contrast, the northwestern part of the county, particularly in the Everglades area, exhibits predominantly low flood susceptibility. These zones potentially consist of lower urban development and greater natural cover, which collectively mitigate flood potential.

To evaluate the performance of the continuous flood susceptibility index, we employed receiver operating characteristic (ROC) curve analysis. This approach evaluates how well the model distinguishes between flooded and nonflooded areas using continuous susceptibility values [168,169]. We have generated an equal number of random points within the FEMA Special Flood Hazard Area (SFHA) and assigned a flood occurrence value of 1 for flood-prone and 0 for nonflood-prone, which also reduces the bias and improves classifier evaluation reliability. The continuous flood susceptibility raster was employed to extract the predicted susceptibility values at each validation point. In ArcGIS Pro, using the Extract multi values to points tool, these raster values were added directly to the attribute table of the validation points. The output includes both the observed and predicted susceptibility values (0 = no flood, 1 = flood) for each point.

The validation dataset was then exported to a csv file for statistical analysis, and ROC curve analysis was conducted using Python 3.9. The ROC curve in Figure 9 plots the true positive rate against the false positive rate across a range of susceptibility thresholds. The AUC was then calculated to evaluate the model’s ability to differentiate between flood and nonflood locations.

The vertical axis of the curve shows the true positive rate, and the horizontal axis of the curve shows the false positive rate. True positive determines the count of pixels accurately identified as a flood occurrence, and false positive determines the count of pixels incorrectly identified as a flood event [170,171]. Higher AUC values represent greater model accuracy. An AUC value close to 1 indicates a perfect model with the highest accuracy, and values less than or equal to 0.5 indicate that the model is not worthy of analysis [172,173].

The ROC approach is particularly well suited for validating urban flood susceptibility models because the GIS-based MCDA framework generates a continuous susceptibility index rather than a binary flood/nonflood map [174]. ROC analysis assesses the model’s ability to distinguish between flooded and nonflooded areas across all possible thresholds, eliminating dependence on subjective class boundary selection [175]. Furthermore, ROC–AUC offers a threshold-independent, scale-invariant metric that is robust to class imbalance, making it effective for validating spatial models using point-based flood inventories derived from FEMA [176,177]. The use of equal flooded and nonflooded points further mitigated sampling bias and enhanced evaluation reliability.

Finally, ROC–AUC has been widely applied and recommended in previous flood susceptibility and hazard-mapping studies, allowing objective comparison with existing studies and ensuring methodological consistency [171,178,179,180]. The ROC-based validation method is particularly important because it considers predictive skill rather than only relying on visual agreement with FEMA flood zones [181]. FEMA flood zones particularly represent regulatory and 100-year flood plains that are particularly designed to provide support for insurance, zoning, and compliance purposes [35,182]. Besides, they depict riverine flooding with standardized return periods and are not intended to represent microscale urban flooding caused by intense rainfall, shallow groundwater fluctuations, or impervious surface expansion [183,184]. These maps are often derived from generalized hydraulic modeling and coarse spatial resolutions and therefore fail to capture recent land-use modifications, changes in urban infrastructure, and fine-scale topographic features that strongly influence localized flood behavior in dense urban environments [185,186]. Therefore, the higher ROC values presented in this research provide a statistically rigorous and widely accepted validation framework for evaluating the predictive performance of continuous flood susceptibility models.

4. Discussion

One of the major findings of this paper is high flood susceptibility in the central and eastern parts of the county, including areas of Miami Beach, Downtown Miami, Little Haiti, Sweetwater, and other densely urbanized neighborhoods, reflecting the convergence of multiple reinforcing flood drivers, including shallow groundwater table and extensive impervious surfaces [96,187,188]. In comparison, these spatial patterns observed in Miami-Dade County align closely with findings from other low-lying, coastal urban regions such as New Orleans and Houston, where shallow groundwater tables, limited drainage capacity, and high impervious surfaces strongly govern flood susceptibility during extreme rainfall events [189,190,191,192]. In these environments, shallow groundwater tables restrict infiltration potential, while extensive impervious surfaces accelerate surface runoff during heavy rainfall events. These findings support the growing recognition that flooding in coastal urban environments is increasingly driven by land use and infrastructure rather than solely controlled by fluvial or coastal processes [91,193]. Comparable trends have also been documented in other rapidly urbanizing coastal regions, further reinforcing the generalizability of these mechanisms across metropolitan landscapes [194,195,196]. Research from New Orleans similarly indicates that rainfall-driven flooding often occurs beyond designated regulatory floodplains, underscoring the importance of susceptibility-based approaches that explicitly capture pluvial flooding processes not represented in conventional flood hazard maps [197]. The strong consistency with the present result and prior urban flood susceptibility studies supports the broader applicability of the proposed GIS–MCDA framework across coastal urban landscapes with comparable geomorphic and urban characteristics [198,199].

In contrast, the findings of this research also highlighted that western and northwestern areas of Miami-Dade County, located closer to the Everglades, exhibit comparatively lower flood susceptibility, due to reduced development density, lower surface runoff, and subtle microtopographic variations that promote surface water redistribution. The high-resolution 3-m LiDAR DEM provided subtle microtopographic variations, such as stream flows, depressions, and elevation differences, that are often overlooked in traditional coarser resolution DEMs [200]. The integration of high-resolution DEMs also aligns with previous studies that employed fine-scale topographic data to improve the accuracy of urban flood susceptibility, reinforcing the significance of fine-scale DEMs in analyzing flood dynamics across complex urban environments [201,202].

The concentration of higher flood susceptibility in urban environments also raises critical issues related to social equity and vulnerability. Many of the highly susceptible neighborhoods include socially and economically vulnerable populations and critical infrastructure systems that have limited capacity to adapt to frequent flooding [203]. Heavy flooding in these areas can disproportionately disrupt transportation networks, housing stability, emergency response times, and public health outcomes. As such, the susceptible regions identified in this study provide a valuable foundation to support equity-informed flood mitigation planning by identifying areas where targeted interventions could generate the greatest social and infrastructure resilience benefits. From a practical standpoint, it also serves as an effective decision support tool for urban planners, emergency managers, and local authorities.

Despite the strong performance of the research findings, several shortcomings should be acknowledged. First, the AHP method depends on expert judgement when assigning weights, which may introduce some subjectivity, although the very low consistency ratio (CR = 0.022) indicates excellent and robust internal coherence. Second, the study presents a static assessment of flood susceptibility based on current environmental and urban conditions and does not integrate future climatic and urban expansion scenarios, both of which may influence flood risk patterns over time. Third, the analysis focuses on urban flood susceptibility, rather than flood hazard, and therefore does not quantify flood depth, duration or flow velocity, which are critical for infrastructure damage or risk reduction assessment.

Future research could address these limitations by incorporating climate change projections, dynamic land use and urban growth scenarios, and hydrodynamic modeling approaches to assess flood depth and duration, thereby enhancing the practical applicability of urban flood risk assessments. Additionally, while the weighting criteria are consistent with local hydrological conditions and previous literature knowledge, future research could integrate data-driven sensitivity analyses to reduce subjectivity in the weighting process.

5. Conclusions

The study provides an in-depth understanding of urban flood susceptibility across Miami-Dade County by incorporating diverse geospatial and environmental factors into a comprehensive spatial modeling framework. The AHP employed for weighting the flood-conditioning factors demonstrated strong internal consistency, with a low consistency ratio of 0.022, determining the reliability and robustness of the pairwise comparisons. In addition, the resulting flood susceptibility maps provide a detailed spatial representation of highly vulnerable areas, particularly in the central and eastern regions of the county, as well as zones of relatively low susceptibility in the northwestern part. These maps serve as effective decision-support tools for urban planners and emergency managers, supported by strong model performance. Model validation using FEMA flood zones demonstrated strong predictive performance (AUC = 0.85), underscoring high accuracy in delineating flood susceptibility within a complex and densely urbanized environment. The findings further support the identification of highly flood-prone zones, enabling more efficient prioritization of resources during emergency response and mitigation planning. Moreover, this fine-scale analysis facilitates more cost-effective and equity-informed allocation of flood mitigation investments in vulnerable communities. The key contribution of this study lies in its capability to capture fine-scale spatial variability in urban flood susceptibility, enabling more targeted mitigation strategies and informed flood risk management. The novelty of this work lies in the integration of high-resolution LiDAR-derived topography, groundwater dynamics, and urban infrastructure characteristics within a rigorously validated GIS–MCDA framework. Moreover, the proposed urban flood susceptibility framework is transferable and can be applied to other coastal and low-lying urban landscapes globally, particularly areas experiencing similar flooding challenges.