1. Introduction
Floods can be characterized as the natural disaster with the highest frequency, the largest number of people affected and the third biggest cause of economic losses, behind storms and earthquakes
[1][2][3][1,2,3]. Between 2000 and 2021, floods represented 40% of all natural disasters, reaching 50% of the annual amount in 2006 and 2021. Asian countries are the most affected, followed by South America
[3][4][3,4]. At the end of 2021, approximately 56% of the world’s population lived in urbanized areas. In Brazil, this percentage is around 87%
[5]. The main local impacts associated with the urbanization process in the urban water cycle are the reduction in vegetal interception, evapotranspiration, and infiltration, and runoff increases in volume and velocity, mainly as a result of vegetation removal, soil paving, interventions in the natural drainage and the construction of artificial drainage systems
[6].
Although flood events are a natural phenomenon usually responsible for a series of ecosystem services
[7], their interactions with urban aspects, such as population increase in susceptible areas and urban sprawl, tend to explain the escalation of associated negative impacts.
In developing countries, cities may continue to occupy floodplains without proper urban planning and reproducing a hygienist approach in stormwater systems
[8]. Failures in drainage systems functioning in urban areas directly affect daily routines and may promote disruption and losses in a series of urban systems such as housing, energy, water supply, sewage, mobility, community facilities and economy, among others, creating a cascade of effects
[8][9][8,9].
It is difficult to produce sustainable urban tissue when flooding is capable of affecting almost all urban systems, directly or indirectly, being able to degrade the urban environment and impoverish the local population.
Considering the context of the Flood Risk Management (FRM) approach, an anticipated risk analysis is a strategic demand for more resilient cities in current and future adverse scenarios of increasing urbanization and climate change
[10]. The identification and classification of flood-prone areas comprise a fundamental step in the FRM approach, providing subsidies for land use planning, floodproofing policies, mitigation measures and early warning systems, among others
[11].
However, robust flood risk analysis can be a highly demanding task, mainly in developing countries where available data related to socioeconomic, environmental and field aspects face problems like low-quality information, low spatial coverage and short time series. The lack of specialized technical teams also aggravates this framework. In addition, even when considering an ideal situation of data availability (more likely to occur in developed countries), before detailing the risk analysis it may be useful to have a preliminary tool to hierarchize priority areas to be studied first.
In this sense, preliminary mapping tools can provide subsidies for urban planning and for detailed urban water management using the few variables available. Flood Hazard Mapping (FHM) aims to represent at least one flooding characteristic among flood depth, spatial flooding extent, flooding duration, damages and losses associated with a return period in the region of interest It can be generated from knowledge on recent historical floods, from geological and geomorphic evidence, aerial photography, satellite imagery and running hydrologic–hydrodynamic models
[11][12][11,12] and by using quantitative and semi-quantitative Multi-Criteria Decision Making (MCDM) methods
[13][14][13,14].
2. GIS and MCDM in Flood Risk Management
2.1. Flood Risk Management
Considering the definition given by UNESCO, Flood Risk can be defined as a combination of the flood hazard and its associated consequences reaching a socioeconomic system
[10]. Hazard can be defined as the physical component of risk, given by a defined rainfall and its transformation process into runoff through the floodway along the watershed. Physical characteristics of the watershed, such as slope, land use, drainage density and urbanization, among others, imply flood characteristics such as water depth, flooding extent, flow velocity and flood duration, for example
[15][16][15,16]. The consequences are related to the exposure and vulnerability aspects of socioeconomic systems
[17]. Vulnerability can be divided into economic value, susceptibility to damage and resilience
[18][19][18,19].
In this sense, studies that do not include flood characteristics should not be named as Flood Hazard Mapping (FHS). Within the same reasoning, a Flood Susceptibility Mapping (FSM), as a preliminary step, should not include flood characteristics.
Unlike other natural disasters, such as hurricanes and earthquakes, many types of floods allow a set of measures to be adopted in the physical environment (watershed) in order to modify the runoff generation process, allowing flood patterns to be changed in time and space. In this sense, it is possible to act to reduce the consequent hazard (inundation) for the same triggering event (precipitation). Rainfall characteristics are common FCFs used in flood indexes. However, considering rainfall in a static analysis introduces a complex issue if a flooded area does not necessarily have a direct relation to where the precipitation itself falls. Water flows through a path in the direction of valley zones or sinks, promoting floods where the topography favors them and not exactly in the precipitation area.
2.2. Types of Floods
Floods can be defined as “the temporary covering by water of land not normally covered by water”, and their driving forces can vary significantly
[20]. For example, coastal floods are usually caused by storm surges, wave overtopping or tsunamis. Fluvial floods are induced by rainfall and exceed the main channels’ capacity, in which their ranges can vary gradually from lowland floods to flash floods; pluvial floods are caused by rainfall in urban areas, and groundwater floods by exfiltration. Floods can also be caused by snow-melt and dam failure, among other things
[10][16][21][22][23][10,16,21,22,23]. Despite having the same element as a triggering factor, the scope of mitigation measures and land restrictions imposed on an area susceptible to slow-rise flooding will be considerably different from those subjected to frequent flash floods.
By defining aspects associated with different types of floods, actors that participate in the decision-making process can readily distinguish specific tools and available management options to deal with flood risks
[17]. According to
[21], a more comprehensive evaluation should consider more than one factor, including the spatial patterns of the causative factors of flooding as well as atmospheric and catchment conditions.
2.3. GIS and MCDM in Flood Risk Management
Studies utilizing the combination of GIS and MCDM for flood applications have been published since the 1990s
[11][24][25][11,24,25]. MCDM applications require criteria to be grouped, standardized and weighted, enhancing work interaction between scientific fields
[7].
In a recent review involving flood indexes
[13], the authors supported the view that MCDM methods have shown the “capability of integrating stakeholder’s input with less complex processing, less input data requirement, accurate results, and decreased uncertainty”.
Physical features such as slope, geology, distance to the main river network, land use and land cover (LULC), terrain elevation, geology and runoff generation are commonly used as indicators in flood susceptibility indexes
[26][27][28][29][30][31][26,27,28,29,30,31]. Other studies have explored earth observation and remote sensing to detect frequently flooded areas
[12][31][12,31]. Flood indexes can also be developed to characterize flood types (flash floods or riverine floods) using hydrograph results from hydrodynamic catchment modeling
[32]. The combination of hazards and consequences of a given flood, creating a Flood Risk Index, is also a common application
[33][34][33,34], and more recently flood resilience analyses and indexes have been gaining space in the literature
[18][19][35][36][37][38][18,19,35,36,37,38].
Kazakis et al.
[27] presented a weight review of a Flood Hazard Index (FHI) which incorporated a sensitivity analysis and renamed the index FHI-S. Using the same seven FCFs as in FHI—flow accumulation, geology, land use, elevation, slope, distance from network drainage (DFND) and rainfall (FIGUSED-S method)—in an Analytic Hierarchy Process (AHP), the results obtained showed differences in the choice of the weights. On a 0–10 scale, elevation became the most important factor, instead of flow accumulation, changing its weight from 2.1 to 3.0. Distance for drainage network sustained second place, but changed weight from 2.1 to 2.5, followed by slope, in third place, changing its weight from 0.5 to 1.6, the highest percentage change, occupying the place previously given to elevation. Flow accumulation changed from 3.0 to 1.2 and the other indicators received weights lower than 1. For example, the geology parameter received a weight of 2% in FIGUSED and 4% in FIGUSED-S.
Tehrany et al. (2013; 2014; 2015) developed different methodologies to assess flood hazard using a decision tree (DT)
[28], a bivariate and multivariate statistical model
[29] and machine learning Kernel type
[30]. In the first method, ten FCFs were used: elevation, slope, curvature, stream power index (SPI), topographic wetness index (TWI), distance from network drainage (DFDN), geology, rainfall, land use/cover and soil and surface runoff. The last incorporated a support vector machine (SVM) technique in four Kernel types and the Frequency Ratio (FR) method with eleven FCFs. As a result, the parameters of elevation and slope factors were considered the most influential factors in all Kernel types.
In all mentioned studies
[27][28][29][30][27,28,29,30], flood inventory was also given by points indicating historical events. They also highlighted the flood susceptibility assessment as a preliminary step in flood risk management, considering the importance of hydrodynamic modeling as a next step.
Using ten FCFs, Mahmoud and Gan
[39] presented a flood susceptibility index with flow accumulation, runoff and soil type as the most influential factors. These results differed from other studies mentioned and, after a sensitivity analysis, the authors concluded that flood susceptibility maps should include more than six FCFs, while other studies suggest that a reduced number of independent FCFs can achieve accurate results
[40].
The fact that historically observed flood points are located in high slope zones calls for attention. Pham et al. (2020) discussed different methods and performed a classification between a flash flood and a non-flash flood event before developing a GIS-based approach for flash flood susceptibility assessment
[41].
An example of an index with only four FCFs was developed for a stretch of river in Iran
[39]. The authors applied an Analytical Hierarchy Process (AHP) and compared the results with flood extent and depths obtained by hydrological–hydrodynamic modeling using HEC-RAS for 50-year and 100-year rainfall events. The FCFs chosen were slope, distance from drainage network (DFDN), LULC and altitude/elevation, and they received weights of 0.138, 0.232, 0.546 and 0.084, respectively. The final index was composed of a sum of normalized weights and normalized ratings for each cell of a 30 m horizontal resolution raster layer.