Flood Susceptibility Analysis with Integrated Geographic Information System and Analytical Hierarchy Process: A Multi-Criteria Framework for Risk Assessment and Mitigation

Abstract

Flooding is among the most destructive natural disasters globally, and it inflicts severe damage on both natural environments and human-made structures. The frequency of floods has been increasing due to unplanned urbanization, climate change, and changes in land use. Flood susceptibility maps help identify at-risk areas, supporting informed decisions in disaster preparedness, risk management, and mitigation. This study aims to generate a flood susceptibility map for Davidson County of Tennessee using an integrated geographic information system (GIS) and analytical hierarchical process (AHP). In this study, ten flood causative factors are employed to generate flood-prone zones. AHP, a form of weighted multi-criteria decision analysis, is applied to assess the relative impact weights of these flood causative factors. Subsequently, these factors are integrated into ArcGIS Pro (Version 3.3) to create a flood susceptibility map for the study area using a weighted overlay approach. The resulting flood susceptibility map classified the county into five susceptibility zones: very low (17.48%), low (41.89%), moderate (37.53%), high (2.93%), and very high (0.17%). The FEMA flood hazard map of Davidson County is used to validate the flood susceptibility map created from this approach. Ultimately, this comparison reinforced the accuracy and reliability of the flood susceptibility assessment for the study area using integrated GIS and AHP approach.

1. Introduction

Water-related natural hazards, including floods, droughts, and landslides, have become more frequent and severe due to the unpredictable changes in rainfall and runoff patterns caused by climate change and urbanization [1,2,3]. Flooding causes extensive damage to ecosystems and physical infrastructures. The Joint Economic Committee reported that flooding in the United States resulted in annual damages ranging from USD 179.8 billion to USD 496.0 billion, which accounts for 1% to 2% of the gross domestic product in 2023 [4]. This economic toll of flooding encompasses a range of damages, including infrastructure repair, loss of property, and economic disruption. Therefore, effective flood management methods, such as flood susceptibility maps, are crucial for reducing flood impacts by identifying flood likelihood, disaster severity, and high-risk areas to enhance prevention and control efforts [5].

The threat and hazard identification and risk assessment conducted by Office of Emergency Management, Nashville—Davidson County identified flooding as the highest-ranking hazard in Davidson County, with a total risk factor of 174.8, among other hazards such as tornadoes, winter storms, droughts, wildfires, thunderstorms, and hazardous materials incidents [6]. This result is bolstered by numerous flooding incidents in the county over the past century. The Great Flood of 1927, Nashville’s second most severe, flooded Shelby Park into a lake when the Cumberland River reached 56.2 feet on 1 January 1927, surpassing the flood stage by 16.2 feet [7]; 2 people lost their lives, and 10,400 were displaced from their homes. On 1 May 1975, the Cumberland River peaked at 47.6 feet, exceeding the flood stage by 7.6 feet, making it the highest recorded flood under regulated conditions in Metro Nashville with an 80-year recurrence. The river remained above flood stage for over six days, resulting in USD 6.6 million in damage [8]. Similarly, intense rainfall in February 1989 caused regional flooding in West and Middle Tennessee, impacting areas such as Lebanon (Wilson County), Obion (Obion County), and parts of Nashville and Antioch (Davidson County) [9].

The 2010 Nashville flood, which broke previous rainfall records for much of the south-central United States, is the most recent significant flooding event to affect Davidson County [10]. The flood caused approximately USD 2 billion in private property damage and USD 300 million in losses to public buildings and infrastructure. Metro Nashville saw extensive damage, impacting over 11,000 properties, including more than 9000 homes. About 5850 affected properties were outside the 100-year floodplain [11]. In Davidson County, the flood led to 10 deaths and USD 1.5 billion in property damage. Flash floods are a recurring issue in various regions of Davidson County, including Nashville, Hermitage, Antioch, Goodlettsville, Richland, and Little Creek. Noteworthy recent events include the flash flood in Hermitage on 7 July 2016, which resulted in approximately USD 250 million in property damage. On 1 September 2017, another significant flash flood occurred in Forest Grove, causing an estimated USD 200 million in property damage, particularly impacting the western two-thirds of Davidson County. On 7 November 2017, a flash flood in West Meade led to USD 50 million in property damage, primarily affecting parts of southern Davidson County. On 6 February 2019, a flash flood in Richland caused USD 1 billion in property damage in central Nashville, while another on 23 February 2019 in Little Creek resulted in USD 250 million in damage [8]. Furthermore, a flash flood on 27 March 2021, resulted in the loss of four lives in Davidson County [12]. The recurrence, spatial distribution, and substantial economic impacts of these flood events highlight the critical need for detailed assessments of flood management strategies across Davidson County.

Given that flooding cannot be entirely avoided, adopting flood mitigation measures is crucial to minimize disruptions to properties [13]. Flood mitigation measures can be categorized as structural and non-structural measures. The common structural flood mitigation measures in use for reducing flood risks include levees, flood walls, dams, and detention ponds. Similarly, non-structural flood mitigation measures can effectively reduce flood risk and damage by working with the natural flood dynamics within the floodplain, offering both short- term and long-term benefits at a lower cost. The non-structural measures that are commonly utilized in alleviating flood risk and damage include flood warning systems, land use regulations, flood emergency preparedness plans, and flood susceptibility mapping. Flood susceptibility mapping is a non-physical non-structural flood mitigation measure that involves creating maps that identify areas at risk of flooding based on various factors, such as topography, soil types, land use, and historical flood data. Flood susceptibility mapping is critical for early warning systems, facilitating flood risk management, and land use strategies to reduce future flood risks [14].

Several studies have been conducted to create and validate flood susceptibility maps using different approaches, such as multi-criteria decision analysis [14,15,16,17,18,19,20,21], logistic regression [22,23,24,25], frequency ratio [25,26,27,28,29], weight of evidence [30,31,32], and statistical index method [27,33,34]. Multi-criteria decision analysis (MCDA) is a decision-making tool used to assess and rank numerous factors based on multiple criteria. In flood susceptibility mapping, MCDA helps analyze various factors contributing to flood risk, combining them into a single, comprehensive evaluation to identify the most vulnerable areas. The weighted linear combination method is typically used in MCDA to define these vulnerable zones [35]. In this approach, each conditioning factor is multiplied by its respective weight, and the sum of these weighted factors determines the final flood susceptibility [16]. The analytical hierarchy process (AHP), which uses pairwise comparisons to evaluate how one alternative rank relative to another based on specified criteria, is the preferred method for assigning weights [20,36].

A flood is a calamity with complex spatial and temporal dimensions so identifying the major causative factors for flooding in any region is critical for mapping flood hazard zones using integrated geographic information systems (GIS) and AHP approach. Flooding events are influenced by a range of factors, such as rainfall intensity, land use/landcover, soil type, and topography. For example, Kazakis et al. [17] applied GIS and AHP to delineate flood susceptibility zones in a region of Greece, the Rhodope–Evros, by analyzing seven factors: geology, flow accumulation, land use/land cover, slope, drainage distance, rainfall intensity, and elevation. Vojtek and Vojtekova [20] mapped flood susceptibility zones at a national scale in Slovakia, considering seven factors including flow accumulation, slope, lithology, river network density, curve number, distance from rivers, and elevation. These factors covered hydrologic, hydraulic, and topographical aspects that influence the flooding event in any region. Using same approach, Hammami et al. [16] used eight factors, elevation, rainfall, groundwater level, drainage density, slope, soil, land use/land cover, and lithology to identify flood vulnerable areas in Tunisia. Kazakis et al. [17] identified flow accumulation as the major flood causative factor in their study area, while Vojtek and Vojtekova [20] found slope as the main factor with a relative weight of 0.35, and Hammani et al. [16] highlighted land use/land cover as the primary factor with a weight of 0.23.

While the integrated GIS and AHP methodology has been extensively utilized for flood risk zone mapping globally [16,17,20,37], its application at the county level within the United States of America remains underexplored. This research focuses on Davidson County, Tennessee, addressing the critical need for localized flood susceptibility mapping at the administrative region scale. By concentrating on Davidson County, this research provides valuable insights into flood risk assessment at a scale that aligns with local governance structures and decision-making processes. Furthermore, a persistent issue in flood susceptibility mapping is validating results, particularly in areas with limited historical flood data. This study utilizes Federal Emergency Management Agency (FEMA) flood maps for validation, to address the challenge of limited historical data availability. The vector-based flood hazard data for Davidson County was obtained from FEMA’s Digital Flood Insurance Rate Map database (effective 20 June 2024). These data, with a spatial accuracy equivalent to a mapping scale of 1:6000, provided precise georeferencing and seamless integration into GIS-based analyses and was therefore used to validate the flood susceptibility map developed using the integrated GIS-AHP method.

Following a literature review, ten causative factors were selected to encompass topographic, hydrologic, and hydraulic aspects of flood for Davidson County, Tennessee. These factors include elevation, normalized difference vegetation index (NDVI), drainage density, slope, distance from river network, distance from road network, soil type, rainfall, topographic wetness index (TWI), and land use/land cover. The main objective of this paper is to analyze these ten key flood conditioning factors using an integrated GIS and AHP approach to identify flood-susceptible regions in Davidson County for flood risk assessment. Furthermore, this research seeks to achieve the following:

• Provide a detailed flood susceptibility map at the county level, addressing the limited application of GIS-AHP methodology at this scale within the United States.
• Leverage FEMA maps to overcome the challenge of limited historical flood data availability while providing a practical and accessible validation method.
• Establish a baseline for future studies to compare and evaluate advanced techniques like Fuzzy AHP or machine learning approaches in similar contexts.

2. Materials and Methods

2.1. Study Area

Davidson county is one among the ninety-five counties of Tennessee with an area of 503.5 square miles. It shares boundary with six other counties of Tennessee: Robertson, Sumner, Wilson, Rutherford, Williamson, and Cheatham. After Shelby County, Davidson County is the most populous county in Tennessee with a population of 715,884 [38]. The population of Davidson was 626,681 in 2010 and saw a growth of 14.26% between 2010 to 2020. Davidson county is home to major towns like Nashville, Bellevue, Belle Meade, Bordeaux, Hermitage, Antioch, Providence, Baker Town, Goodlettsville, and Forest Hills. The Cumberland River, the second largest tributary of Ohio River, drains all of Davidson County, dividing it from east to west. As it flows through, the Cumberland River intersects with the Stones River upstream of downtown Nashville. To the southwest, the Harpeth River runs along the county’s edge, though it merges with the Cumberland River in Cheatham County further downstream. Alongside these major rivers, several other creeks and small streams drain through Davidson County. The location map of the study area is shown in Figure 1.

2.2. Data

Multiple datasets representing hydrologic, topographic, and hydraulic elements that influence flooding in a region are used in this study to create a flood susceptibility map of Davidson County. The data types used in this research, along with their sources, are provided in

Table 1. Source, resolution, and temporal coverage of data for different flood causative factors.

Data	Source of Data	Original Resolution	Temporal Coverage
DEM	3DEP from United States Geological Survey (USGS) (https://apps.nationalmap.gov/)	10 m × 10 m	30 January 2023 and 7 April 2023 (Published Date)
Slope	Derived from DEM	10 m × 10 m	Same as of DEM
NDVI	Landsat 9 from USGS (https://earthexplorer.usgs.gov/)	30 m × 30 m	21 August 2023
LULC	Multi-Resolution Land Characteristics Consortium (https://www.mrlc.gov/data)	30 m × 30 m	NLCD 2023, updates annually
Soil Type	gSSURGO data from United States Department of Agriculture (USDA) (https://datagateway.nrcs.usda.gov/)	30 m × 30 m	October 2024, updates annually
Drainage Density	Derived from DEM	10 m × 10 m	Same as of DEM
Distance from Road	Topologically Integrated Geographic Encoding and Referencing (TIGER) from US Census Bureau	Not Applicable	September 2024, updates annually
Distance from River	Derived from DEM	10 m × 10 m	Same as of DEM
Rainfall	USGS Water Data for Nation (https://waterdata.usgs.gov/nwis)	Not Applicable	2012–2023 (Daily)
TWI	Derived from DEM	10 m × 10 m	Same as of DEM
FEMA Flood Map	FEMA Flood Map Service Center (https://msc.fema.gov/portal/home)	Not Applicable	20 June 2024 (Published Date)

. These data are preprocessed and analyzed within a GIS environment to create thematic maps that illustrate different flood conditioning factors.

Both raster-based data, including digital elevation model (DEM), land use/land cover (LULC), and Landsat 9 imagery, and vector-based data, such as soil type, rainfall, and TIGER road data, were processed to a uniform resolution of 30 m × 30 m and projected to the NAD 1983 UTM Zone 16N coordinate system. This standardization allowed for the creation of consistent thematic maps, which were subsequently reclassified and overlaid to produce a weighted overlay map using the analytical hierarchy process (AHP) approach [21]. For NDVI computation, atmospherically corrected surface reflectance data and cloud-free Landsat images were acquired and processed to minimize potential errors [39]. The raster data (DEM) is processed using various spatial analyst and geoprocessing tools in ArcGIS Pro (Esri, Redlands, CA, USA) to derive different flood conditioning factors, such as elevation, slope, drainage density, TWI, and distance from rivers. Figure 2 provides a schematic representation of the methodology used in this research.

2.3. GIS Processing of Flood Conditioning Factors

2.3.1. Elevation

Elevation is critical in determining flood susceptibility, with lower elevation regions more prone to flooding [17,40]. For this study, a 1/3-arc-second (10 m × 10 m) resolution digital elevation model (DEM) was obtained from the USGS and clipped to match the boundaries of the study area. The elevation within the study area is between 21 m and 353 m above mean sea level (masl), as illustrated in Figure 3a.

2.3.2. Slope

Surface slope plays a vital role in influencing flood susceptibility. Regions with gentle slopes are more vulnerable to flooding due to their tendency to retain water, whereas steeper slopes are less affected as water drains downslope rapidly [41]. A slope gradient map, created using the DEM and the slope function tool in ArcGIS Pro, is presented in Figure 3b.

2.3.3. Normalized Difference Vegetation Index (NDVI)

Vegetation plays a crucial role in reducing surface runoff, thus lowering the likelihood of flooding. NDVI is a remote sensing tool that utilizes reflectance in the visible and near-infrared (NIR) wavelengths to assess and monitor the health and coverage of green vegetation in any area. The NDVI is computed using an equation:

NDVI = (NIR − Red)/(NIR + Red)

(1)

The NDVI values range from −1 to 1, where negative values correspond to water, snow, or clouds; values near zero indicate bare soil or rocks; and higher values reflect dense and healthy vegetation [42]. In this study, the Landsat 9 Surface Reflectance product from USGS was processed using the NDVI raster function. Bands 4 (red) and 5 (near-infrared) of Landsat 9 were used to generate the NDVI map, which is presented in Figure 3c.

2.3.4. Land Use/Land Cover (LULC)

LULC significantly affects water management, storage, and movement across a landscape. The type and distribution of various LULC categories play a critical role in determining the likelihood and intensity of flooding. Flooding mechanism varies with land use, such as urban areas, vegetated regions, and those with soil or rocky textures [43]. Urbanized areas and regions with dense water networks tend to have a higher flood risk due to impervious surfaces and modified water flow patterns. These conditions increase surface runoff, limit water infiltration, and can overwhelm drainage systems, leading to rapid and severe flooding during heavy rainfall. For this study, LULC data were obtained from the Multi-Resolution Land Characteristics (MRLC) Consortium and clipped to align with the boundaries of Davidson County as shown in Figure 3d.

2.3.5. Soil Type

Soil texture and structure determine its permeability, drainage, and water retention capacity, which directly influences surface runoff and flood risk [44]. For this study, the gridded Soil Survey Geographic (gSSURGO) dataset from the USDA was used to map soil types within the study area. The gSSURGO data include symbols representing various soil types. Using the USDA’s map unit descriptions for Davidson County, Tennessee, these soil symbols were grouped into nine broad categories: silty clay loam, gravelly silt loam, urban land complex, cherty silt loam, rock outcrop, pits, water, silt loam, and sulfur complex. This categorization was achieved using attribute tables and the field calculator tool in ArcGIS Pro. The resulting thematic soil map, shown in Figure 3e, was converted to raster format and subsequently classified into five distinct classes.

2.3.6. Drainage Density (DD)

Drainage density is a key metric for assessing watershed behavior that affects water flow, flood risk, and erosion potential. It is defined as the total stream length per unit area (km/km²) [45]. Generally, regions with higher drainage density exhibit increased surface runoff, which elevates the likelihood of flooding [20]. In this study, a flow accumulation raster generated from the DEM served as the input for the line density analysis tool in ArcGIS Pro to create a drainage density map, depicted in Figure 3f.

2.3.7. Distance from Road (DRO)

Impervious materials such as asphalt, concrete, and bituminous macadam, commonly used in pavements, roads, and parking areas, increase surface runoff, thereby raising the risk of flooding. As a result, the distance of an area from the road network is a key factor influencing its flood vulnerability. The road network of the study area was mapped using the USDA’s Topologically Integrated Geographic Encoding and Referencing (TIGER) street data. The Euclidean distance tool in ArcGIS Pro’s spatial analyst was used to create a distance from road map [18], as shown in Figure 3g.

2.3.8. Distance from River (DRI)

Areas located close to river networks serve as floodplains during intense rainfall events, making proximity to the river network an important factor in flood risk assessment. In this study, a flow accumulation raster derived from the hydro DEM was used to delineate the stream network, with a threshold set at 0.1% of the maximum flow accumulation value. Given that the Cumberland rivers drain the county, a lower threshold was applied to include smaller creeks, which can contribute significantly to flash flooding. Euclidean distance tool in ArcGIS Pro was utilized to generate a distance from river map [18], as shown in Figure 3h.

2.3.9. Rainfall

Davidson County has experienced significant flooding events triggered by intense rainfall, including those in 1989 and 2010. This study utilized daily rainfall data from six USGS precipitation gauges, as detailed in

Table 2. USGS rain gauges used with locations and average annual rainfall data.

USGS Site Number	Latitude (Degree)	Longitude (Degree)	Location	Average Annual Rainfall for 12 Years (Inch)
03430550	36.009	86.702	Mill Creek near Nolensville	49.48
03431040	36.072	86.733	Seven Mile Creek at Blackman Rd	47.33
03431100	36.094	86.794	Glendale Lane at Nashville	49.00
03431655	36.102	86.868	Richland Creek Belle Meade	45.89
03426387	36.358	86.725	Mansker Creek at Millersville	51.65
03431530	36.274	86.817	Whites Creek at Old Hickory Blvd	54.59

, to obtain rainfall information for the study area from 2012 to 2023. These rainfall stations are located in regions representing diverse topographical features and land uses across Davidson County. This distribution helps to capture the heterogeneity of the county, ensuring that our analysis accounts for variations in rainfall patterns influenced by different landscape characteristics. Furthermore, selected stations provide long-term continuous rainfall data required for accurate flood susceptibility mapping. The average annual rainfall was computed, and the Inverse Distance Weighting (IDW) tool in ArcGIS Pro was used to create an average annual rainfall map for the county, as shown in Figure 3i.

2.3.10. Topographic Wetness Index (TWI)

TWI is a numerical measure that quantifies the potential for water accumulation in an area, based on its topographic features. A higher TWI value suggests an increased likelihood of flooding [32]. TWI is computed using an equation [46]:

TWI = ln (Specific Catchment Area/tan (slope))

(2)

where Specific Catchment Area = (Catchment Area/Unit Contour Length)

Catchment Area = (Flow Accumulation Raster + 1) × (Area of each cell)

tan (slope) = (Slope Raster/100) + 0.001

Flow accumulation raster and slope (percent rise) were used as input in raster calculator tool to generate TWI maps, which is shown in Figure 3j.

2.4. Analytical Hierarchy Process for Relative Weightage

AHP has been extensively used in various studies and regions [20,47,48], demonstrating its robustness and accuracy when combined with GIS for spatial analysis. While advanced methods such as Fuzzy AHP (FAHP) or machine learning algorithms offer additional capabilities, including handling uncertainty and non-linear relationships between variables [49,50], they also come with increased complexity [51]. FAHP, for instance, requires more computational resources and involves additional steps to model fuzziness in decision-making [52]. These methods are particularly useful in cases where data uncertainty is significant. However, in the context of this study, the available data are well structured and consistent, making the traditional AHP approach sufficient for determining relative weights of different flood conditioning factors.

The selection of flood conditioning factors was guided by input from five experts, including four from academics with expertise in hydrology, GIS, and flood risk management, and one professional from the water industry. The process involved a thorough literature review to identify commonly used factors, followed by group discussions to refine the list. This ensured the inclusion of relevant factors while excluding others less applicable to the specific context of Davidson County. The influence of each flood conditioning factor varies [16]. AHP, also known as the pairwise comparison technique [36], can be used to compute the relative weights for each factor. This approach involves constructing a square matrix where each factor is compared with every other factor based on its relative importance. The relative weights for the ten selected conditioning factors were assigned based on empirical knowledge and previous studies [14,18,20]. In this study, the relative importance of each flood conditioning factor ranges from 1 (most important) to 10 (least important). For example, if one factor is judged to be four times more important than another, the corresponding value in the matrix is 4, while the reciprocal value 1/4 is assigned to the reverse comparison.

Once all pairwise comparisons are completed, the principal eigenvector of the matrix is calculated to derive the relative weights of each factor. This process involves normalizing the comparison matrix by dividing each element by its column total, then calculating the average of each row in the normalized matrix. The resulting eigenvector represents the relative weights of each factor in the flood susceptibility model. The relative weights were calculated by deriving the principal eigenvector of a 10 × 10 square reciprocal matrixes, with criteria ratings estimated through expert judgment and literature review using a nine-point continuous scale [36], as shown in

Table 3. Saaty scale (1980).

Scale	Numerical Rating	Reciprocal
Equally Important	1	1
Equally to Moderately Important	2	1/2
Moderately Important	3	1/3
Moderately to Strongly Important	4	1/4
Strongly Important	5	1/5
Strongly to Very Strongly Important	6	1/6
Very Strongly Important	7	1/7
Very Strongly to Extremely Important	8	1/8
Extremely Important	9	1/9

.

The consistency of the relative weights obtained from pairwise comparison matrix was checked using a numerical index known as consistency ratio (CR). It can be calculated using an equation:

CR = CI/RI

(3)

where CI is the consistency index, and RI is the random inconsistency index. CI was calculated using an equation:

CI = (λ − n)/(n − 1)

(4)

where λ is the average value of consistency vector and n is the number of factors. Saaty [36] has provided values for RI for different number of factors that are listed in

Table 4. Random inconsistency index.

n	1	2	3	4	5	6	7	8	9	10
RI	0	0	0.58	0.90	1.12	1.24	1.32	1.41	1.45	1.49

.

A CR ratio above 0.1 suggests the judgments may lack reliability due to inconsistency, while a CR of 0 indicates perfect consistency [15,36,53]. The formula for the weighted aggregation method is given by the following equation:

FSM = ∑w_j × y_j

(5)

where FSM is the flood susceptibility map, w_j is the weight of flood conditioning factor j, and y_j is the susceptibility class for each factor j.

2.5. Sensitivity Analysis

The flood susceptibility assessment process may be influenced by the variability of the selected flood conditioning factors, potentially introducing biases [37]. To address this concern, we investigated to evaluate how uncertainties in the assigned factor weights affect the flood susceptibility assessment. The error Δε_i, caused by independent errors Δw_j in the weighting coefficient values, is given by the following equation:

$$\Delta {\epsilon }_i=\sqrt{\sum _{j=1}^n{\left(\Delta w_j\cdot r_{ij}\right)}^2}$$

(6)

where Δεᵢ represents the error caused by uncertainty in each grid cell, r_ij is the rating factor, and n denotes the number of factors. To obtain Δwᵢ, the final relative weight of each factor was adjusted by 20% from the original values [37,47,54,55,56,57]. The changes in weight values for each factor are as follows: 0.016 for elevation, 0.022 for slope, 0.008 for NDVI, 0.042 for LULC, 0.058 for soil type, 0.010 for drainage density, 0.004 for distance from road, 0.006 for distance from river, 0.004 for rainfall, and 0.030 for TWI. A map representing the error ∆εᵢ at each grid cell was generated and then added to and subtracted from the basic FSM map, creating two extreme scenarios, FSM_max and FSM_min, which correspond to the maximum and minimum values, respectively.

3. Results

The final weights and the normalized pairwise comparison matrix were derived using the approximation method [58].

Table 5. Pairwise comparison matrix.

Factors	Soil	LULC	TWI	Slope	Elevation	DD	NDVI	DRI	Rainfall	DRO
Soil	1	2	3	4	5	6	7	8	9	10
LULC	0.5	1	2	3	4	5	6	7	8	9
TWI	0.33	0.5	1	2	3	4	5	6	7	8
Slope	0.25	0.33	0.5	1	2	3	4	5	6	7
Elevation	0.20	0.25	0.33	0.5	1	2	3	4	5	6
DD	0.17	0.20	0.25	0.33	0.5	1	2	3	4	5
NDVI	0.14	0.17	0.20	0.25	0.33	0.5	1	2	3	4
DRI	0.13	0.14	0.17	0.20	0.25	0.33	0.5	1	2	3
Rainfall	0.11	0.13	0.14	0.17	0.20	0.25	0.33	0.5	1	2
DRO	0.10	0.11	0.13	0.14	0.17	0.20	0.25	0.33	0.5	1

represents the final hierarchical order selected for ten flood conditioning factors. The values in each row represent the comparison of the importance between two parameters. For example, soil type is considered the most important criterion, while distance from the road is the least important factor for this study area. Similarly, LULC is the second most important factor, and a value of 0.5 is assigned when establishing the relationship between LULC and soil type in the second row.

The flood susceptibility map for the study area was generated by overlaying thematic maps of different flood conditioning factors, using the weightings for each factor as outlined in

Table 6. Normalized factor weights and final relative weights.

Factors	Soil	LULC	TWI	Slope	Elevation	DD	NDVI	DRI	Rainfall	DRO	Wt (%)
Soil	0.34	0.41	0.39	0.35	0.30	0.27	0.24	0.22	0.20	0.18	29
LULC	0.17	0.21	0.26	0.26	0.24	0.22	0.21	0.19	0.18	0.16	21
TWI	0.11	0.10	0.13	0.17	0.18	0.18	0.17	0.16	0.15	0.15	15
Slope	0.09	0.07	0.06	0.09	0.12	0.13	0.14	0.14	0.13	0.13	11
Elevation	0.07	0.05	0.04	0.04	0.06	0.09	0.10	0.11	0.11	0.11	8
DD	0.06	0.04	0.03	0.03	0.03	0.04	0.07	0.08	0.09	0.09	5
NDVI	0.05	0.03	0.03	0.02	0.02	0.02	0.03	0.05	0.07	0.07	4
DRI	0.04	0.03	0.02	0.02	0.02	0.01	0.02	0.03	0.04	0.05	3
Rainfall	0.04	0.03	0.02	0.01	0.01	0.01	0.01	0.01	0.02	0.04	2
DRO	0.03	0.02	0.02	0.01	0.01	0.01	0.01	0.01	0.01	0.02	2

. From these relative weights between flood conditioning factors, a consistency index of 0.06 and a consistency ratio of 0.04 were obtained, which satisfied the consistency check [53].

The relative weights were multiplied by the corresponding reclassified factor map and then overlaid to generate a flood susceptibility map within the GIS environment. To refine the results, neighborhood analysis was performed on the flood susceptibility map using a 5 × 5 rectangular window. The flood susceptibility map was classified into five distinct categories based on flood risks, ranging from very low to very high. Figure 4a depicts the baseline influence of the ten flood conditioning factors on flood occurrence in Davidson County. Figure 4b,c represent the maximum and minimum scenarios, respectively, accounting for weight uncertainty by applying a ±20% variation to the original weights. This sensitivity analysis provides a range of potential outcomes, enhancing the robustness of flood susceptibility assessment.

The flood susceptibility map generated was then compared with the FEMA flood hazard map for Davidson County to validate the identified flood risk zones. Figure 4d shows a FEMA flood hazard map for Davidson County that highlights regions prone to flooding based on probability assessments, including the 1% annual chance flood hazard (Special Flood Hazard Area or SFHA), the 0.2% annual chance flood hazard, minimal flood zones, and designated floodways. FEMA’s methodology incorporates hydrological, topographical, and historical flood data to assess flood risk and is periodically updated to reflect changes in land use, climate patterns, and hydrological conditions. Figure 5 presents the outcome of combining the flood susceptibility map, created using the integrated AHP and GIS approach, with the FEMA flood hazard map.

4. Discussion

4.1. Spatial Distribution of Flood Susceptibility

A flood susceptibility map for Davidson County was developed by using the relative weights obtained from AHP in a weighted overlay analysis within ArcGIS Pro. The resulting map categorizes the county’s flood risk into five distinct classes: very low, low, moderate, and very high susceptibility. Figure 4a illustrates that areas with high and very high flood susceptibility are concentrated near the Cumberland River and other water bodies. These vulnerable zones are characterized by low elevation, gentle slopes, and increased drainage density, aligning with findings from multiple studies [20,37,48]. The northwestern and southwestern sections of Davidson County, featuring higher elevations, steeper slopes, greater distance from rivers, and forest cover, exhibit very low to low flood susceptibility. Meanwhile, the central portion of the county, with its moderate elevation, low slope, and developed land use, predominantly falls within the moderate flood susceptibility category.

The classification revealed that 17.48% of the total area falls under the very low-risk category, 41.89% under the low-risk category, 37.53% under the moderate-risk category, 2.93% under the high-risk category, and 0.17% under the very high-risk category as presented in

Table 7. Area coverage of different flood susceptibility classes.

Flood Susceptibility	Area (km2)	Area (%)
Very Low	237.81	17.48
Low	569.93	41.89
Moderate	510.66	37.53
High	39.90	2.93
Very High	2.26	0.17

. Although the high flood risk zone accounts for only approximately 3% of the total area in Davidson County, it is important to note that this zone includes many urban areas with significant infrastructure and population, such as Nashville, Goodlettsville, Oak Hill, and Bordeaux. This situation underscores the critical need for comprehensive flood mitigation strategies in these areas.

The sensitivity analysis of the flood susceptibility mapping, as presented in

Table 8. Percentage area coverage of different flood susceptibility classes for three flood susceptibility maps (FSMs).

Flood Susceptibility	FSM (%)	FSMmax (%)	FSMmin (%)
Very Low	17.48	17.90	17.37
Low	41.89	41.62	42.49
Moderate	37.53	36.86	36.80
High	2.93	3.38	3.18
Very High	0.17	0.24	0.16

, offers valuable insights into the model’s robustness and the potential range of flood risk scenarios in Davidson County. When comparing the FSM_max scenario to the baseline FSM, small increases are observed in very high-, high-, and very low-susceptibility areas, with small decreases in moderate- and low-susceptibility areas. Similarly, the FSM_min scenario shows small increases in low- and high-susceptibility areas, and small decreases in very low-, moderate-, and very high-susceptibility regions. Notably, in both extreme conditions, high-susceptibility areas have increased while moderate-susceptibility areas have decreased. The relatively small variations observed across the three scenarios (FSM, FSM_max, and FSM_min) suggest that the flood susceptibility model is reasonably robust. The consistency in the overall distribution of susceptibility classes, even under extreme weighting scenarios, lends credibility to the baseline assessment. However, the observed changes, particularly in the high- and very high-susceptibility categories, underscore the importance of considering uncertainty in flood risk assessments. For instance, the potential increase in high-risk areas from 2.93% to 3.38% in the FSM_max scenario, while seemingly small, could have significant implications for urban planning and emergency preparedness, especially if these additional high-risk areas encompass critical infrastructure or densely populated zones.

4.2. Spatial Distribution of Flood Susceptibility by Land Cover Type

To analyze the distribution of flood susceptibility levels in relation to land use within the study area, we intersected the generated flood susceptibility map with the 2023 MRLC land use map.

Table 9. Area distribution of MRLC land cover classes corresponding to the designated flood susceptibility levels.

NLCD Classes	Very Low		Low		Moderate		High		Very High
	Area (km2)	Area (%)	Area (km2)	Area (%)	Area (km2)	Area (%)	Area (km2)	Area (%)	Area (km2)	Area (%)
Open Water	0.01	0.02	0.88	1.54	15.23	26.57	38.92	67.91	2.26	3.95
Developed, Open Space	3.27	1.67	92.48	47.20	100.08	51.09	0.08	0.04	0.00	0.00
Developed, Low Intensity	0.33	0.17	55.68	28.57	138.83	71.22	0.07	0.04	0.00	0.00
Developed, Medium Intensity	0.15	0.11	28.06	21.55	101.86	78.22	0.15	0.12	0.00	0.00
Developed, High Intensity	0.01	0.01	7.45	10.58	62.50	88.75	0.46	0.66	0.00	0.00
Barren Land	0.02	0.52	2.09	61.03	1.27	36.95	0.05	1.49	0.00	0.00
Deciduous Forest	221.54	52.41	187.14	44.27	13.99	3.31	0.02	0.00	0.00	0.00
Evergreen Forest	0.60	2.10	26.07	91.55	1.80	6.31	0.01	0.03	0.00	0.00
Mixed Forest	7.52	8.92	70.10	83.19	6.64	7.89	0.00	0.00	0.00	0.00
Shrub/Scrub	0.65	19.36	2.41	71.35	0.31	9.27	0.00	0.03	0.00	0.00
Herbaceous	1.23	15.63	5.36	67.98	1.28	16.22	0.01	0.17	0.00	0.00
Pasture/Hay	2.48	1.66	88.94	59.78	57.34	38.54	0.02	0.01	0.00	0.00
Cultivated Crops	0.00	0.05	2.73	30.78	6.14	69.17	0.00	0.00	0.00	0.00
Woody Wetlands	0.00	0.00	0.33	16.98	1.58	80.28	0.05	2.74	0.00	0.00
Emergent Herbaceous Wetlands	0.00	0.04	0.19	9.43	1.81	88.63	0.04	1.90	0.00	0.00

presents the areal distribution of National Land Cover Database (NLCD) classes across each flood susceptibility category. The results show distinct distributions of land use classes across different flood risk zones. For instance, open water areas are predominantly in high-risk zones (67.91%) and very high-risk zones (3.95%). Developed areas, such as open space, and low-, medium-, and high-intensity developments, are mostly in moderate-risk zones, with increasing percentages as development intensity rises. Specifically, 51.09% of the land classified as developed, open space falls within the moderate flood risk zone, along with 71.22% of developed, low intensity; 78.22% of developed, medium intensity; and 88.75% of developed, high intensity. Similarly, barren lands are primarily in low-risk (61.03%) and moderate-risk (36.95%) zones. Natural areas like deciduous forests, evergreen forests, mixed forests, and shrub/scrub are generally in low to very low-risk zones. For instance, 52.41% of deciduous forest falls within the very low-risk zone, while 44.27% is in the low-risk zone. Evergreen forests have 91.55% of their area under low flood risk, and mixed forests have 83.19% in the same category. Shrub/scrub areas have 71.35% in the low flood risk zone. Herbaceous regions show a mixed profile: 67.98% low-, 16.22% moderate-, and 15.63% very low-risk. Pasture/hay lands are split between low- (59.78%) and moderate- (38.54%) risk zones. In contrast, cultivated crops, woody wetlands, and emergent herbaceous wetlands are primarily in moderate-risk zones, with 69.17%, 80.28%, and 88.63% of their areas, respectively, in this zone. The remaining portions of these land cover classes are primarily in the low flood risk zone, accounting for 30.78%, 16.98%, and 9.43%, respectively. This spatial analysis highlights the need for tailored flood management strategies that consider both land use patterns and flood risk levels.

4.3. Validation of FSM with FEMA Flood Hazard Maps

As shown in Figure 5, the flood hazard zone classification derived from this study exhibits a strong correspondence with FEMA’s flood hazard map. Specifically, areas identified as high and very high flood susceptibility lie within the 1% annual chance flood hazard zone. However, the agreement is not absolute; not all areas designated as SFHA are classified as high- or very high-risk by the present approach. For example, portions of the floodplain along the Cumberland River identified as SFHA are categorized as moderate susceptibility. In the eastern region of the county, areas near Hermitage, Old Hickory, and Lakewood, the spatial extent of SFHA on the FEMA map differs from that of the high- and very high-susceptibility zones on the flood susceptibility map.

Furthermore, areas within the 0.2% annual exceedance zone (500-year floodplain) defined by FEMA, representing less frequent but potentially severe flood events, are generally classified as moderate susceptibility, which is reasonable. The areas around Nashville, Bordeaux, Bellevue, and Belle Meade, characterized by a moderate risk of flooding, predominantly fall within the 0.2% annual chance zones. Additionally, areas with very low and low susceptibility generally align with areas of minimal flood risk on the FEMA maps.

Discrepancies are also observed along streams near Whites Creek, Oak Hills, and Berry Hills, where FEMA identifies 1% annual chance flood hazards, while the flood susceptibility map classifies these areas as moderate susceptibility. Such inconsistencies underscore the importance of using flood susceptibility maps as a complement to, rather than a replacement for, FEMA flood hazard maps. While flood susceptibility maps offer a broader perspective on potential flood risks, FEMA maps provide a more detailed and reliable assessment of flood hazards, especially near waterways. Nevertheless, the demonstrated effectiveness of integrating GIS and AHP in delineating flood hazard zones, as validated by the similarity between the generated map and the FEMA flood hazard map, suggests that the flood susceptibility map can serve as a valuable baseline for the quick identification of potential flood zones, urban planning, risk management, and emergency preparedness efforts.

4.4. Limitations and Potential Directions for Future Research

The overlap between high-risk zones in the flood susceptibility maps and SHFA highlights the effectiveness of the GIS-AHP method; however, it is important to note that it has some potential limitations. To map flood susceptibility zones, this work uses ten flood conditioning factors, which are the same number of factors utilized in studies by Mahmoud and Gan [59], Haghizadeh et al. [60], and Khosravi et al. [61]. Vojtek and Vojtekova [20] and Elkhrachy [15] utilized seven factors, but Varra et al. [37], and Hammami et al. [16] used eight, while Mukherjee and Singh [18] used nine. Only four criteria were included in several investigations, including those by Rahmati et al. [62] and Samanta et al. [63]. This shows the non-uniformity in the number of flood conditioning factors that should be used while using this approach. While there is no fixed number of factors to be considered, Mahmoud and Gan [59] recommend using at least six to prevent a single factor from dominating the weighting process, which could lead to overemphasizing certain flood-contributing factors. The local hydrological, land use and geomorphological variables of the study area have influenced the number of flood conditioning factors among studies [20]. Despite these differences, some flood conditioning factors consistently appear in research, including elevation, slope, TWI, LULC, and drainage density, all of which are incorporated in this study. A key limitation of this approach is its reliance on the precision and resolution of the input data, which might introduce errors, particularly in regions with diverse land cover or complex topography [37]. Although LiDAR and other high-resolution data can improve map accuracy, they are not always accessible, which could cause complex landscapes to be oversimplified. Furthermore, different land cover types have variable effects on flood susceptibility; thus, current data are required. There are issues with susceptibility map scales as well because different applications could require different scales [20,37].

These elements emphasize how crucial it is to properly select appropriate flood conditioning factors and evaluate data resolution and quality when analyzing and using flood susceptibility maps. Future studies may thus use nested models that smoothly switch between local and regional scales, offering risk assessments that are more thorough and adaptable. Moreover, the incorporation of socio-demographic factors, such as infrastructure vulnerability, and population density, into the examination of flood susceptibility in future models could give a more holistic picture of flood risk.

5. Conclusions

An integrated GIS and AHP methodology were used in this research to perform a comprehensive flood susceptibility assessment for Davidson County. The primary objective was to analyze flood conditioning factors using an integrated GIS and AHP approach to identify flood-susceptible regions in Davidson County for flood risk assessment. Ten flood conditioning factors representing hydrological, hydraulic, topographical, and geomorphological aspects of the study area were considered. The conditioning factors were preprocessed in GIS, and the AHP technique was used to calculate factor weights based on Saaty’s nine-point scale. The AHP analysis revealed soil type as the primary flood conditioning factor with a 29% relative weight, followed by land use/land cover (LULC) at 21%, topographic wetness index (TWI) at 15%, and slope at 11% for the study area. These relative weights were incorporated into a weighted overlay method in the GIS to generate a flood susceptibility map.

The resulting map was classified into five distinct categories based on flood risk, ranging from very low to very high-risk zones. This study found that most of the study areas fall into the low (41.89%) and moderate (37.53%) flood risk categories, with smaller portions classified as very low (17.48%), high (2.93%), and very high (0.17%). The map was subsequently validated through comparison with the FEMA flood hazard map for the study area. The flood zoning provided by this analysis can serve as a valuable tool for stakeholders in planning critical flood mitigation measures to reduce potential loss of life and property in future flood events.