Parameters and methods used in flood susceptibility mapping: a review

Floods are one of the most damaging natural disasters affecting human life globally. Flood susceptibility mapping (FSM) is one of the critical tools for studies such as identifying flood-prone areas, land use planning, emergency preparedness, and disaster risk reduction (Shokouhifar et al. 2022). The accuracy and reliability of these maps require a broad understanding of the parameters and methods used in FSM. The choice of parameters and the applied method can affect the boundaries and sensitivity levels of the identified areas (Burgan & Icaga 2019). The choice of parameters depends on the characteristics of the study area and the method used for mapping, but their selection is subjective and may vary depending on the judgment of the expert, which can lead to differences in the results of flood susceptibility maps.

This article highlights the strengths and limitations of different approaches by examining the parameters and methods used in FSM. Understanding the strengths and limitations of the parameters and methods used by researchers and practitioners is essential to produce accurate and reliable flood susceptibility maps (Ilderomi et al. 2022; Shokouhifar et al. 2022). The formation, development, and resultant effects of the flood vary depending on the geographical parameters. However, selecting an appropriate method that accurately represents how the parameters affect calculations in nature can directly impact the accuracy of the results.

Slope, soil properties, land use, precipitation, and drainage are some of the common parameters used. Also, statistical analysis, fuzzy logic, hybrid methods, machine learning, and physically based modeling are common methods used for FSM. Some methods are simple and effective in identifying important parameters (e.g. statistical analysis), some methods have advantages in integrating multiple parameters and taking into account uncertainties in data (e.g. fuzzy logic), and some methods have the ability to learn from data and identify complex relationships between parameters (e.g. machine learning algorithms). Physically based models simulate the behavior of water according to the characteristics of the basin over different scenarios, but require extensive data and are computationally intensive. However, they are useful in modeling complex hydrological systems (Fenicia et al. 2014; Tehrany et al. 2014a; Tariq et al. 2021).

Many researchers in the literature have stated that each method used in FSM has some advantages and limitations. The choice of the method also depends on the availability of data available in the study area, the characteristics of the study area, and the purpose of the mapping. However, all methods contain uncertainties that may affect the accuracy and reliability of maps (Dimitriadis et al. 2016; Adib et al. 2019).

FSM requires the integration of multiple data sources, including topographic, hydrological, and meteorological data (Kuriqi & Hysa 2021). Identifying appropriate methods and parameters ensures that these data sources are effectively integrated and used to produce accurate flood susceptibility maps. By analyzing the methods and parameters used for FSM and understanding the regional characteristics of the parameters contributing to flooding, the accuracy and reliability of flood susceptibility maps can be improved. The parameters methods used in the studies of FSM in the literature were investigated in this study. In order to analyze the methods used in studies on flood susceptibility, scientific publications were searched on the Web of Science (WoS). Publications were searched as ‘flood susceptibility’ using the ‘title’ constraint. During the research, only articles were scanned from all the years and narrowed down by choosing the ‘WoS Core Collection (Web of Science Core Collection)’. After scanning, studies were eliminated based on their relevance to the subject, and evaluations were conducted. Within these criteria, 170 studies published between the year of the first published study in 2014 and the date of the last data update for this study, 30 June 2022, were examined in detail, and the parameters and methods used in FSM were analyzed.

As mentioned above, this article presents a comprehensive analysis of the methods used in FSM and their associated parameters, and discusses the strengths and weaknesses of each method. In this respect, it helps researchers to understand the complexity and diversity of methods, which are numerous and diverse, and to determine the most appropriate method and data according to the purposes of their studies, characteristics of the region and data sources.

Separating parameters into main groups provides a framework for easily comparing different methods and selecting convenient parameters. This also helps researchers identify gaps in the literature and design future studies to improve FSM.

Contributing to the FSM literature as it provides a comprehensive analysis of different methods and parameters, this study is a valuable resource for researchers, practitioners, and policy makers in the field of flood risk management, and it is intended to contribute to the development of understanding of FSM methods. This can also be particularly beneficial for decision-makers and practitioners who are responsible for managing flood risk and need to make informed choices about the most effective methods for their particular context. At the same time, specifying parameters as the main groups has made the complex landscape of FSM simpler and has made the overall properties of the parameters more accessible and comprehensible.

The study is also expected to contribute to the representation of complex processes and overcoming uncertainties for identifying flood-sensitive areas. FSM is a complex process that requires consideration of multiple parameters such as topography, soil, land use, and precipitation patterns, along with other natural data. These parameters can interact with each other in complex ways, and although their relationship is not always direct, their results can have a large impact. In this respect, analyzing, understanding, and comparing the different methods used in FSM contributes to identifying the most effective approaches to represent these complex processes. Also, dividing the parameters used in the methods into main groups to understand which natural processes that influence flooding provides important insights to identify the most critical parameters to be considered in FSM. In other words, FSM involves dealing with the uncertainties that may arise due to lack of data, incomplete information, and unforeseen events.

The concepts of ‘flood hazard’ and ‘flood risk’ are two basic concepts for holistic flood management studies, which are related to each other but should not be used interchangeably (Kaya 2017). Flood hazard refers to the environmental impact in the event of a possible flood event (Klipalo et al. 2022), taking into account the speed, depth, and magnitude of the flood, while flood risk refers to the combination of the negative consequences of the flood on human health, economic activities, and the environment (Flood Directive 2007/60/EC). Determination of flood hazard preparation, planning, management, etc., is one of the guiding bases for their studies and is generally classified as ‘low risk’, ‘medium risk’, and ‘high risk’ (Kaya 2018). As can be understood from the definitions, more data are needed for flood hazard analysis than flood susceptibility analysis. At the same time, in flood risk analyses, different from hazard and sensitivity analyses, detailed data on elements at risk and vulnerability are needed (Pangali Sharma et al. 2022). This makes the preparation of flood risk and hazard maps more difficult than flood susceptibility maps. Susceptibility maps, which can be produced with more limited data compared to hazard and risk analyses, are important data sources for basic approaches in preliminary studies against flood events that may occur in the future (Pangali Sharma et al. 2022). Considering the impact of changing climatic conditions and changes in land use on floods, it is clear that flood susceptibility analyses will have an important place in determining early warning systems and strategies to prevent future floods and/or reduce their damage (Vojtek & Vojteková 2019).

Natural disasters, such as floods, landslides, and cyclones, are mostly dependent on several conditions being present (Ali et al. 2019; Waqas et al. 2021). To evaluate flood susceptibility, it is necessary to investigate a series of flood-triggering and causal parameters, and their relationship with flooding (Radmehr & Araghinejad 2015; Sahana & Patel 2019; Ullah & Zhang 2020). In the reviewed studies, researchers used different flood-causative parameters. Choosing parameters to create flood susceptibility maps is often seen as a complicated challenge for FSM (Waqas et al. 2021; Tariq et al. 2022). There is no specific guideline for selecting flood-controlling factors that affect flood occurrence (Ullah & Zhang 2020). The selection of flood-controlling factors depends on the physical and natural characteristics of the study area and data availability (Kia et al. 2012; Liuzzo et al. 2019; Ullah & Zhang 2020).

The parameters frequently used in FSM are summarized below, respectively, based on the frequency of use determined by this study.

This parameter has a significant impact on flood formation and distribution. It is an indicator describing the sensitivity of the region to flooding (Youssef et al. 2011). Areas with a high slope (assuming that land cover and use are similar) are less exposed to flooding than areas with low slopes (Hammami et al. 2019). It was used in 165 (%97) of the 170 articles examined in this study (Table 1).

Topographical data like slope, SPI, aspect, TWI, etc., can be derived from DEM. Elevation can be derived from DEM as well as from height data in DEM. The data are effective factors both in the establishment of the hydrological model and in the strength and development of floods and their impact on land cover. In addition to using DEM and elevation data, in the production of flood susceptibility maps, relative elevation, elevation difference, and standard deviation of elevation parameters are also used in the literature. It was used in 153 (%90) of the 170 articles examined in this study (Table 1). Typically, many input data in flood modeling are generated from the DEM, but the quality of the DEM parameters used in the modeling is critical to the reliability of the flood probability map (Avand et al. 2022).

The most important parameter in the occurrence of floods is rainfall (Tehrany et al. 2015a, 2015b; Khosravi et al. 2018). The time–space relationships of precipitation are complex; however, they are the basic calculations for flood simulation models (Khosravi et al. 2018). The intensity and duration of the precipitation, as well as the amount of precipitation, are also effective for the development of the flood and flood duration, its area of influence, and possible damages to occur in the region. Mean monthly rainfall, annual rainfall, and rainfall anomaly data are used in the studies. They were used in 138 (%81) of the 170 articles examined in this study (Table 1).

Since this parameter is effective on physiological and hydrological processes in flood modeling studies, it is very important in the susceptibility maps. Although the land cover and land use are used interchangeably in the literature, they are two different terminologies. Land use refers to the use of the basin, such as settlement, industry, and agriculture, which are shaped by socioeconomic activities. Land cover expresses the physical and biological conditions (forests, arid areas, water bodies, etc.) that make up the nature of the basin. LULC data cover the environmental impacts of anthropogenic activities and the full distribution of natural cover (Waleed & Sajjad 2022). LULC data is commonly extracted from satellite images. It was used in 136 (%80) of the 170 articles examined in this study (Table 1).

The spread of the flood in the basin is related to the distance of the location to the drainage system (Elkhrachy 2015). While less flooding occurs in areas far from the drainage system, it occurs more in closer areas (Samanta et al. 2018). In this respect, drainage network density is one of the important parameters for flood sensitivity analysis. For this reason, it is one of the parameters frequently used in studies and was used in 130 (76%) of 170 articles examined within the scope of this research.

TWI is also one of the parameters obtained from the secondary derivative of DEM and defines the hydrological settings of the region. TWI is a widely used parameter for determining the location and size of water-saturated regions. Topography is a key parameter on the spatial distribution of water on the surface and underground, so the parameter is generally calculated to measure the topographical effect on water processes. TWI was used in 121 (%71) of the 170 articles examined in this study (Table 1).

The literature has shown that formations with geologically impermeable surface properties are more sensitive to floods (Islam & Sado 2000). At the same time, geology is significantly influential on the formation of the drainage pattern, which is associated with the accumulation processes of water and the processes affecting the overflow capacity (Bui et al. 2019). Therefore, geology plays an important role in the mapping of flood susceptibility maps. It was used in 115 (%68) of the 170 articles examined in this study (Table 1).

Soil types are another parameter commonly used in susceptibility mapping. The soil type of the basin controls the amount of water infiltrating the soil and thus the precipitation mass that will cause the flood (indirectly the development rate and magnitude of the flood). It was used in 95 (%56) of the 170 articles examined in this study (Table 1).

In addition to morphometric properties, drainage density, it demonstrates the nature of the ground and its geotechnical features. It was used in 83 (%49) of the 170 articles examined in this study (Table 1).

Aspect is a main factor for susceptibility analysis like slope and it defines the direction of the slope. Slope direction has a strong influence on hydrological processes (especially through frontal precipitation direction and evapotranspiration), and as a result it also affects vegetation, even more so in dry environments (Sidle & Ochiai 2006). In the articles examined, it was determined as one of the parameters with a high effect (Talukdar et al. 2020; Chakrabortty et al. 2021; Nguyen 2022a). It was used in 80 (%47) of the 170 articles examined in this study (Table 1).

The SPI is geomorphic in origin and is generated from DEM. In other words, it is one of the parameters obtained (Tehrany et al. 2015a, 2015b). It is a metric value of flow strength that shows the corrosive effect of water. Some studies in the literature have shown that the secondary derivative of DEM demonstrates the hydrological characteristics of the region and highlights the significant role of SPI in the production of flash flood maps, as well as its correlation with structural zones and land use (Hang et al. 2021). It was used in 78 (%46) of the 170 articles examined in this study (Table 1).

This parameter is indicative of processes related to accumulation, flow velocity, and erosion (Mirzaei et al. 2021). Curvature affects the flow and it also influences the occurrence of floods (Pradhan 2010), and floods mostly occur in areas with flat curvature (Tehrany et al. 2015a, 2015b). Some studies show that elevation and curvature are the best predictive factors for estimating the flood occurrences (Tehrany et al. 2014b). They were used in 71 (%42) of the 170 articles examined in this study (Table 1). Moreover, it seems in Table 1 that plan curvature was used in 38 of the 170 articles, but profile curvature was used in 27 of them.

Many studies have also shown that there is a negative relationship between the occurrence of floods and land cover (Tehrany et al. 2014b). Land use is taken into account in the production of flood susceptibility maps because the characteristics of the land cover (vegetation density, residential areas, forested areas, etc.) affect the runoff and NDVI is one of the important parameters in flood modeling (Khosravi et al. 2016, 2018). It was used in 57 (%34) of the 170 articles examined in this study (Table 1).

When studies on flood sensitivity are examined in the literature, although it is seen that many different methods are used in sensitivity analysis, certain methods are preferred more than others in practice. However, although there is no general acceptance by scientists working in this field that commonly used methods are superior to other methods, the commonly used methods in the evaluation of flood sensitivity in the literature are, respectively, multi-criteria decision-making (MCDM) methods, physically based hydrological models, statistical methods, and various soft computing methods. These methods, which are suggested by researchers to create more objective flood susceptibility maps, differ from each other in terms of whether they need expert opinion and ease of application.

Physically based hydrological models are effective in flood modeling (Dimitriadis et al. 2016; Kaya 2022); one-dimensional models such as Mike 11, ISIS, and HEC-RAS, and two-dimensional models such as TELEMAC-2D, RMA2, SRH-2D, and Hydro_AS-2D are frequently used (Knebl et al. 2005; Lavoie & Mahdi 2017; Tehrany et al. 2019). Physically based models require fieldwork for data collection, a large budget (Fenicia et al. 2014; Tehrany et al. 2014a), and significant computational resources and are not suitable for large-scale studies (Tehrany et al. 2019). However, it is also stated that other methods cannot replace traditional hydraulic modeling (Sole et al. 2013; Rahmati et al. 2016b), but they can be used especially in large-scale analyses or in developing countries (Albano & Sole 2018; Vojtek & Vojteková 2019). In this study, studies using physically based hydrological models and flood susceptibility analysis were not included in the evaluation due to the difference in the method applied.

Decision-making is the process of choosing among various alternatives. MCDM is a process in which many criteria can be evaluated together and values can be assigned to alternatives in complex problems such as disasters, and MCDM methods are methods that allow the best choice to be selected from more than one criterion applied at the same time (Malczewski 1999). The Multiple Criteria Decision Analysis (MCDA) provides a rich collection of technical procedures for structuring decision problems and designing, evaluating, and prioritizing alternative decisions (Malczewski 1999). In other words, it is one of the methods that allow decision-makers to make the most appropriate decision according to the problem and factors, taking into account the effectiveness of a large number of independent variables (Arslankaya & Göraltay 2019). On the other hand, they are methods that can be evaluated by presenting inferences in a common language in cases where the criteria conflict with each other or where criteria cannot be expressed with a numerical value (Hamurcu & Eren 2015; Arslankaya & Göraltay 2019). Due to its simple structure, MCDM methods are widely used in flood susceptibility analysis (e.g. analytical hierarchy process (AHP) (Al-Abadi et al. 2016; Mahmoud & Gan 2018; Bera et al. 2022), analytical network process (ANP) (Dano et al. 2019; Balogun et al. 2020), decision-making trial and evaluation laboratory (DEMATEL) (Wang et al. 2018; Ali et al. 2020), weighted linear combination (WLC) (Tang et al. 2018; Stavropoulos et al. 2020), etc.). However, MCDM methods are largely based on expert opinion and give subjective results (Chowdary et al. 2013). However, the strength of the method is that it can take into account a potential effect that has not emerged until the date of analysis.

Statistical methods are indirect methods that are commonly used to evaluate the correlation between flood triggers and floods based on mathematical expressions (Dai & Lee 2002; Chen & Wang 2007; Wubalem et al. 2022). The most commonly used statistical methods in flood susceptibility assessments are bivariate statistical analysis (BSA) and multivariate statistical analysis (MSA). The frequency ratio (FR) method, one of the BSA methods, is one of the most widely used methods to measure the effect of each factor class on flood (Jebur et al. 2014). Logistic regression (LR), a typical MSA method, determines the effect of each factor affecting flood formation (Jebur et al. 2014). It has been stated in studies that these statistical analysis methods have very good performance in flood susceptibility assessment (Rahmati et al. 2016b; Liu et al. 2022). However, while statistical methods rely on predicted variables based on linear assumptions, they often have a nonlinear nature due to the complex mechanism of flooding (Tehrany et al. 2015a, 2015b; Costache & Bui 2019; Liu et al. 2022). The complex nature of floods has encouraged researchers to switch from traditional/exact/rigid computing methods to soft computing methods using heuristic approaches that better reflect real life.

Real-life problems contain natural uncertainties (Derin Cengiz 2020). Different solutions based on heuristic approaches have been proposed by researchers to solve complex real-life problems related to uncertainties. Soft computing is an approach to design systems with numerical intelligence that can change the analysis environment and learn to produce better results, presenting the reasons for their decisions, with the equivalent of human expertise in certain respects (Kabalcı 2022). Zadeh (1994), who developed the fuzzy set theory, expressed soft computing as follows:

Soft computing, which is tolerant of fuzziness, ambiguity, partial accuracy, and approximations as opposed to rigid computing, is based on techniques such as fuzzy logic, genetic algorithms, artificial neural networks (ANNs), machine learning, and expert systems (Doğan 2016). Soft computing offers a combination of computation and intelligent methods for solving complex, uncertain, real-life problems (Baskir 2018). Zadeh (1994) stated that soft computing is not a single method but rather a combination of several methods such as fuzzy logic, neural networks (NNs), and genetic algorithms. All these methods complement each other instead of competing with each other and can be used together to solve a specific problem (Buckley & Hayashi 1994; Doğan 2016). Soft computing methods, which are used in many different fields, have also been frequently used in landslide and flood susceptibility analyses in recent years.

As mentioned above, the methods used in flood susceptibility analyses have evolved over the years from traditional expert opinions to statistical methods based on big data and machine learning methods (Li et al. 2019; Liu et al. 2022). The ever-evolving machine learning algorithms have been increasingly applied in flood sensitivity estimation (Kia et al. 2012; Pradhan 2012; Madhuri et al. 2021; Liu et al. 2022; Nguyen 2022b). Random forest (RF) (Wang et al. 2015), ANN (Li et al. 2013), support vector machines (SVM) (Tehrany et al. 2015a, 2015b), and decision tree (Tehrany et al. 2013) are the most widely used machine learning algorithms in studies and are frequently used in flood susceptibility analysis (Liu et al. 2022).

Metaheuristic algorithms used in nonlinear modeling are an active research area that is rapidly becoming widespread in the literature, especially used to capture the changes in the nature of large-scale optimization problems (Bui et al. 2020). Metaheuristic optimization algorithms offer very good solutions for real-world optimization problems (Neumüller et al. 2011). Metaheuristic optimization algorithms (e.g. particle swarm optimization (PSO), gray wolf optimization (GWO), bat algorithm (Bat), etc.) that are increasingly used in flood susceptibility assessments give very good results, especially in hybrid models created by machine learning methods (e.g. Bui et al. 2020; Rahmati et al. 2020; Arora et al. 2021). In hybrid models where two or more methods are used together, various statistical or machine learning methods can be used in integration with different statistical, MCDM, metaheuristic algorithms, or machine learning methods (e.g. Tehrany et al. 2014a; Ali et al. 2020; Bui et al. 2020; Rahmati et al. 2020; Arora et al. 2021). Pham et al. (2018) stated that hybrid models improve and increase the prediction accuracy of bivariate statistical models such as LR, FR, evidential belief function, and weights of evidence (WoE) (Shahabi et al. 2021). When the flood susceptibility studies conducted in recent years are examined, especially hybrid models come to the fore among the methods used.

FSM is a vital component of flood risk management in terms of identifying areas most likely to be affected by floods and assessing the level of flood susceptibility. There are many methods and many parameters used in FSM in the literature. This article highlights the strengths and limitations of different approaches by examining the parameters and methods used in FSM. The grouping of parameters in susceptibility maps offers researchers and practitioners a broad perspective for evaluating input parameters and selecting the most accurate ones based on the region being studied. Understanding the strengths and limitations of the parameters and methods used by researchers and practitioners is very valuable in order to perform sensitivity analyses that are closest to the truth. In addition, the advantages and disadvantages of each of the various methods used for FSM are emphasized and the direction of the trend over the years has been determined. In this article, while the most commonly used methods for FSM and their strengths and limitations are examined, it is determined which parameters are in which input data type to represent natural features that increase flood susceptibility. Knowing which physical processes the parameters represent is very important in terms of choosing the most accurate and sufficient number of input parameters for the study area. This directly determines the reliability and accuracy of sensitivity analyses. The study identified topographical parameters, geological parameters, hydrography, LULC, socioeconomic parameters, meteorological and climatic parameters, and urban and environmental parameters as the main groups of input data types. However, there are also hundreds of subinput parameters identified in the literature. Although the input parameters differ, choosing at least one data type from each of the main groups is a valuable approach in terms of accurately determining flood susceptibility. Table 1 presents which parameters represent which natural process constituting flood susceptibility.

Bivariate statistical methods are used to measure the relationship between flood formation and independent variables (Tehrany et al. 2014b). By using a two-variable probability model, the spatial relationship between flooded areas and each of the variables contributing to flood formation is determined (Tehrany et al. 2014b). The greater the bivariate probability, the stronger the relationship between flood occurrence and the variable (Lee & Talib 2005; Lee & Pradhan 2007).

In the reviewed studies, fuzzy logic, which is one of the widely used soft computing methods, machine learning algorithms (Hong et al. 2018; Costache 2019), metaheuristic optimization algorithms (Termeh et al. 2018; Arora et al. 2021), statistical methods (Hong et al. 2018), and MCDM methods (Ali et al. 2020; Tella & Balogun 2020) are used to produce flood susceptibility maps by creating a hybrid model. In particular, AHP, ANP, DEMATEL, TOPSIS, etc., are used to remove the subjectivity of MCDM methods based on expert opinion and to obtain more objective results. The use of hybrids with MCDM methods draws attention (Ali et al. 2020; Tella & Balogun 2020).

Evolutionary algorithms and swarm intelligence-based optimization algorithms are metaheuristic optimization algorithms used in the reviewed articles. Among the evolutionary algorithms, genetic algorithm, biogeography-based optimization (BBO), and differential evolution (Arora et al. 2021; Roy et al. 2021; Saleh et al. 2022); among the swarm intelligence-based optimization algorithms, PSO, GWO, bat algorithm, social spider optimization, grasshopper optimization algorithm, and bee algorithms (BA) are the most used algorithms (Table 3). In addition, the heuristic search algorithm K-Star (Kstar) is among the preferred algorithms (Siam et al. 2021; Ruidas et al. 2022; Table 3).

While each of the approaches mentioned in the previous section and within this section have strengths, they also have certain weaknesses that can produce various uncertainties in the FSM (Vojtek & Vojteková 2019). Therefore, the methodology chosen for flood susceptibility analyses should adequately represent the spatially continuous and cumulative nature of the influence of parameters on flood-producing mechanisms (Vojtek & Vojteková 2019). In addition, the selection of an appropriate methodology for FSM should also depend on the spatial scale (local, regional, national, or global) (Vojtek & Vojteková 2019).

When the literature is examined, it is seen that more than 20 parameters are used for flood susceptibility maps and researchers prefer to use a maximum of 10 parameters as the number of parameters in the preparation of the maps. While some researchers showed that rainfall, soil, and geology were the most effective parameters as a result of their study (Ouma & Tateishi 2014; Seejata et al. 2018), some researchers showed that LULC and altitude data were the most influential parameters (Kourgialas & Karatzas 2011). Rahmati et al. (2016a) and some researchers have shown that slope is the most influential parameter (Samanta et al. 2018). Lee et al. (2017) showed that distance from the river, geology, and DEM were the effective parameters. However, when the relationship between the examined methods and the types and number of parameters was examined, no significant result was reached.

Based on the studies reviewed, the choice of parameters used in FSM differs among researchers. Some studies used a limited number of parameters, while others used a more comprehensive set of parameters to represent a large number of risk-enhancing features. The parameters used in the preparation of flood susceptibility maps differ among researchers. However, in the studies, at least one/several parameters were chosen to represent data groups representing environmental, hydrological, topographic, and sociodemographic characteristics. Although different parameters were used in the studies, the effective parameters differed in the studies in which the common parameters were predominant. The reason for this may be the differences in the characteristics of the data used, as well as the differences in the natural characteristics of the regions. At the same time, another reason for this may be that the susceptibility analyses do not fully represent a holistic approach that further incorporates the presence of hydrological processes and combines them with the vulnerable features of the regions.

In general, it should be noted that current approaches to FSM cannot fully cover the complex relationships and mechanisms between hydrological processes and regional sensitivities. In this respect, it is important for future research to explore the potential of holistic approaches that combine different perspectives to increase the accuracy and reliability of FSM.

The authors thank Prof Dr Çağdaş Hakan Aladağ for offering helpful suggestions and comments to improve the paper.

The authors have not received any funding to perform this research work.

The authors have contributed equally to this work.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.