Cumulative catchments: A novel approach for catchment-based regional flood susceptibility

Cumulative catchments: A novel approach for catchment-based regional flood susceptibility | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Cumulative catchments: A novel approach for catchment-based regional flood susceptibility Abdullah Akbas, Aydoğan AVCIOĞLU, Hasan OZDEMIR, Tolga GORUM, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8379220/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Catchments are a fundamental unit of hydrological and geomorphological systems. Although floods occur in rivers, they also result from processes within catchments. For this reason, catchment morphometry and parameters such as climatic, soil, and hydrological are widely used to define flood susceptibility in regions where data is scarce. However, floods mostly occur at inter-catchments. An inter-catchment is a transitional area between adjacent drainage basins where water flow is not confined to a single channel and may contribute to multiple basins. However, morphometric calculations often neglect spatial gradients between sub-catchments and inter-catchments in assessing flood susceptibility. In this study, we therefore developed the cumulative catchment approach, which delineates small sub-catchments that reflect the upstream drainage contribution while not neglecting the spatial gradients within sub-catchments. Based on this approach, many morphometric parameters have been calculated. Additionally, parameters related to climatic and hydrological conditions, as well as land-use/soil types, have been assigned to the catchments using zonal statistics. Although the aim of the study was not to compare machine learning methods, the models that performed best in flood prediction were compared, and flood susceptibility was obtained on a cumulative catchment basis using machine learning. The algorithms such as XGBoost, random forest, logistic regression, and support vector machine were used, and XGBoost was determined to be the best model for defining flood susceptibility based on the ROC curve using the inventory belong to FlooDOT (FlooD invetory Of Türkiye) dataset. Furthermore, a bivariate map was constructed between model parameters and susceptibility values to understand the impact of covariates. To the best of our knowledge, this study provides the pioneer flood susceptibility analysis regionally for Türkiye, based on cumulative catchments. Model results present flood probabilities in a more realistic and high-resolution manner, based on the cumulative characteristics of the catchments, in the form of sub-catchments (small catchments). This approach also offers an opportunity in terms of regional applicability, particularly in data-scarce areas. Cumulative catchments Flood susceptibility Catchment morphometry Türkiye Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1. Introduction Floods are one of the catastrophic hazards, causing many fatalities and devastating consequences worldwide every year. The document by the Center for Research on the Epidemiology of Disasters (CRED), based on EMDAT (International Disaster Database) data, reports that only floods caused 104.000 deaths and cost 651 billion dollars between 2000 and 2019 ( CRED and UNDRR, 2020 ). Although floods occur in rivers, they are the output of many processes and are affected by activities within the catchment area. On the one hand, catchments are the fundamental unit of hydrological systems where all fluxes such as energy and mass, interact ( Chorley & Kennedy,1971; Strahler, 1980 ). Consequently, all activities within the catchment can cause positive or negative feedback on the fluxes, thereby altering their quantity and direction. Climate, for instance, is one of the important reasons that can change the feedback mechanism in catchments. Due to climate change, the temporal and spatial properties of various climatic parameters, including precipitation, soil moisture, and snowmelt, may impact the feedback mechanisms of catchments. Additionally, changes in these conditions due to climate change may lead to different feedback mechanisms within hydro-climatological subsystems (Alfieri et al., 2015 ; Kreibich et al., 2022 ). Blöschl et al. ( 2020 ), for example, investigated changes in hydro-climatological variables, such as precipitation, snowmelt, and temperature, across Europe. They concluded that changes in the magnitudes of these parameters over time affect floods and their trends, with regional differences. Furthermore, many anthropogenic activities in catchments also drastically change the feedback mechanisms (Rentschler et al., 2023 ; Yang & Tian, 2009 ). Andreadis et al. ( 2023 ), for example, demonstrate that urbanization in floodplains since 1985 has dramatically increased, and this has led to an increase in the number of people affected by floods, even though the frequency of floods has not increased. On the other hand, catchment properties can affect fluxes such as runoff and peak flow. Chorley ( 2021 ) has stated that catchment properties can influence peak runoff. The first attempt to introduce the impact of catchment properties on hydrology and runoff characteristics was executed by Horton ( 1932 ). Horton explained the impact of catchment properties on flood characteristics using different morphometric indices such as form factor, and drainage density. From that day onward, numerous studies have investigated flood susceptibility using catchment morphometry (Bhat et al., 2019 ; Chissende et al., 2025 ; Patton and Baker, 1976 ; Florinsky, 2012 ; Ghasemlounia and Utlu, 2018; Islam and Deb Barman; 2020 ; Khodaei et al., 2025 ; Magilligan, 1992 ; Morel et al., 2023 ; Obeidat et al., 2021 ; Ozdemir and Bird 2009 ; Pike and Wilson, 1971 ; Rogelis et al., 2014; Rogelis and Werner, 2014 ; Tola and Shetty, 2022 g et al., 2022 ). Ghasemlounia and Utlu ( 2018 ) have attempted to prioritize the flood susceptibility using catchment morphometry, and they expressed that prioritizing flood susceptibility based on the ranking of parameters for catchment morphometry has also been the result of subjective or expert-based knowledge. On the other hand, Stolle et al. ( 2015 ) have investigated the role of catchment properties in predicting flows in fluvial and debris flow-dominated catchments using morphometric characteristics (catchment slope, area, etc.) via statistical methods. They concluded that morphometric characteristics are significant in determining flood and debris flow risk; when correctly selected and analyzed, their predictive power can be high. On the other hand, a study by Ozdemir and Akbas ( 2023a ) proved that the results of the flood hazard analysis using the hydrodynamic model (LISFLOOD-FP) and a morphometric flood susceptibility model for the catchments gave similar results, and that the morphometric flood susceptibility model can explain potential flood hazards in the catchments. Researchers have primarily used catchment morphometry parameters, such as drainage density, form factor, and time of concentration, for the interested catchment and attempted to understand the impact of these properties on hydrological conditions, including peak flow. However, since Horton ( 1932 ), all morphometric methodological explanations on flood-generating mechanisms and catchment morphometry have been built upon single and individual catchments, and sub-catchments have been ignored. For instance, Patton and Baker ( 1967 ) also used many parameters related to morphometric analyses in individual selected catchments and explained that flash floods are correlated with catchment morphometric parameters. Moreover, Costa ( 1987 ) compared the hydraulic and morphometric properties of single catchments to understand the correlation of flash flood occurrence in the conterminous United States using flood inventory data and concluded that the largest flash floods result from the combination of optimum conditions in terms of catchment morphometry, climate, and hydraulic properties. These kinds of morphometric analyses only consider small and individual catchments and ignore spatial gradients of sub-catchments and inter-catchments (see definitions in Fig. 2 ). Interestingly, flood inventories (for example, FlooDOT, used in this study by Akbas et al. ( 2025 , 2026 ) reveal that all floods occur in specific areas within sub-catchments, especially inter-catchments, which represent the upstream areas. Ozdemir and Akbaş ( 2023 ) addressed this issue in their study comparing flood modelling results with catchment morphometric results, but it has not yet been clarified. Hence, a new perspective and novel approach are required to overcome the limitations of the current representation of sub-catchments in morphometry. Therefore, we propose and utilize a novel cumulative catchment approach to define flood susceptibility at the regional level, as well as catchment-based high-resolution aspects (small catchments). Cumulative catchments are catchments that are divided into smaller sub-catchments according to a specified threshold value, but despite their small size, they reflect their upstream area. This developed catchment approach offers an innovative perspective that enables the spatial representation of flood susceptibility at high resolution (i.e., small catchments but represents the upstream) and the generation of a flood susceptibility map at a regional scale. Therefore, this study presents a novel cumulative basin approach to represent the basins where floods occur and, for the first time, reveals basin-based flood susceptibility at a regional scale. Hence, the aim of this study is a) to present the novel cumulative catchment approach and apply flood susceptibility regionally to catchments via machine learning algorithms, such as XGBoost, random forest, logistic regression, and support vector machine, and b) to obtain regional flood susceptibility mapping over Türkiye for a pioneering study. Although different machine learning models were used in this study, the primary aim is to identify the model that best reflects the cumulative basin approach, rather than comparing these algorithms. 2. Study area and historical floods Türkiye has distinct characteristics, in which topographic relief changes rapidly over short distances, creating a strong gradient between the coastal and inner parts (Fig. 1 ). This situation is particularly pronounced in the Anatolian plateau and in the transitions between the mountainous areas that form its margins. Additionally, seasonal climatic conditions associated with large-scale atmospheric processes can have a significant impact on Türkiye. In winter, for instance, mid-latitude cyclones originating from Icelandic low-pressure systems descend into the Mediterranean area, bringing abundant precipitation to Türkiye because of the Polar jet ( Tatlı et al., 2004 ). The most extreme precipitation occurs during this time, causing numerous disasters, including floods and landslides (Akbas, 2023 ; Gorum and Fidan, 2021 ; Yüksek et al., 2013 ). Also, cyclones have an intrinsic connection with topography in which coastal areas facing cyclones receive more precipitation, while the continental inner areas correspond to rain-shadow zones because of the topographic barrier (Aydoğan et al., 2016 ; Karaca et al., 2000 ). Unlike winter, during the summer season, the polar jet ascends to the north, and Monsoon low and Azores high pressures influence Türkiye, causing severe drought conditions (Tatli et al., 2004 ). However, even during summer, most of the Black Sea region is exposed to cyclones and experiences hazards due to extreme precipitation. On the other hand, the Central Anatolian plateau and Northeast Anatolia also experience similar hazard conditions due to convective precipitation (Türkeş, 1996 ). On the other hand, some researchers have stated that development and growth initiatives in Türkiye have had a significant impact on land use, resulting in a profound effect on hydrogeological conditions. Consequently, alongside climatic vulnerability, changes in land use within watersheds also creates a significant potential for generating disasters (Koç et al., 2020 ; Ozdemir and Elbaşı, 2015 ). The study area in this paper covers 25 major basins defined by the General Directorate of State Hydraulic Works (DSI) in Türkiye, and specifically the Meriç-Ergene, Marmara, Susurluk, Kuzey Ege, Gediz, Küçük Menderes, Büyük Menderes, Batı Akdeniz (Western Mediterranean), Antalya, Burdur Göller, Akarçay, Sakarya, Batı Karadeniz (Western Black Sea), Yeşilırmak, Kızılırmak, Konya, Doğu Akdeniz (Eastern Mediterranean), Seyhan, Asi, Ceyhan, Fırat-Dicle (Euphrates- Tigris), Doğu Karadeniz (Eastern Black Sea), Çoruh, Aras, Van Gölü, basins respectively (Fig. 1 ). These basins are used for operational activities of the DSI. Based on FlooDOT (Flood Inventory of Türkiye, Akbas et al. ( 2025 , 2026 ), Fig. 1 depicts a map of recent Turkish floods and their impacts, including fatalities, injuries, and losses. As can be seen from the map, floods occur in every part of Türkiye, resulting in many casualties. In this study, the main basins belonging to DSI were used to compare flood susceptibility results obtained from the cumulative catchments approach in order to better understand flood-generating mechanisms. 3. Materials and Methods 3.1. Construction of cumulative catchments and rivers Digital Elevation Models (DEMs) are one of the most important databases used to obtain hydrographic data such as river networks and catchments (Lehner et al., 2008 ). However, it is not possible to use DEMs directly in hydrological analyses because these data, which are created using different techniques ranging from synthetic aperture radar (SAR) to UAVs or stereo imagery, have the characteristics of a digital surface model (DSM) rather than a DEM. One of the most significant disadvantages of using this type of data in hydrological analysis is that it contains numerous artificial structures that can impede the flow of water or direct it in the wrong direction. Apart from this, the presence of forested areas is another limitation in terms of hydrological and geomorphic aspects (Gorum, 2019 ). Additionally, it may be challenging to obtain an accurate representation of the river’s network and channels. Therefore, DEMs to be used in hydraulic or hydrological studies must first be cleared of artificial structures ( Ozdemir and Akbaş, 2025; Yamazaki et al., 2017 ). For example, Yamazaki et al. ( 2017 ) applied various filtering techniques to SRTM (90m) data, utilizing databases associated with tree density and height, as well as reference ground elevation, to increase the accuracy of the data by approximately 40% and 60%. This type of filtering is especially important for flat areas and closed catchments where river channels are very difficult to obtain because the DEM may misrepresent the river network in these areas. For this reason, FABDEM (Forest And Buildings removed Copernicus 30m DEM), which is one of the most important DEM sources that removes the presence of obstacles like buildings and tree heights that prevent the establishment of a stream network for hydraulic and hydrological studies, was used in this study (Hawker et al., 2022 ). FABDEM uses a random forest algorithm through many covariates related to buildings, such as night lights, GHS Urban database, world settlement footprint, and forests, including land-cover, forest heights, etc., and provides a DEM for hydrological works. According to Ozdemir and Akbas ( 2023b ), FABDEM provides results similar to those of LIDAR data in terms of both flood inundation depth and river network production in catchment studies. After deciding to use FABDEM, the Archydro package was employed to create the river network and catchments in the study (Djokic et al., 2011 ; Maidment, 2022 ). This package is frequently preferred in hydrology thanks to its sub-packages that allow the correction of a conditioned digital elevation model for river networks and catchments, as well as the numerous computational flexibilities it provides. In this respect, the catchment construction and river network processes, as determined by FABDEM data, are illustrated in Fig. 2 . Since there are endorheic basins in many parts of Türkiye, it is essential to burn DEM in order to reflect the river network properly. Here, EU-Hydro data obtained from satellites were used as the benchmark river network while the DEM was burned ( CLMS, 2025 ). However, as EU-Hydro includes a different network in some areas, it has been checked and corrected again using 1/25.000 topographic maps. Although it is stated that FABDEM has been corrected based on the forest and settlement, filling operations have been performed to address potential gaps in some areas. Since the hydrologically conditioned DEM data were ready to generate a stream network, the first process, i.e., determining the flow direction, was carried out (Fig. 2 a). Although D-Infinity (DINF) and Multiple Flow Direction (MFD) methods are among the algorithms used to determine the flow direction, the D8 method was preferred in this study because it is one of the most frequently used and stable algorithms in the literature (Jenson and Domingue, 1988 ; O'Callaghan and Mark, 1984 ; Tarboton, 1997 ). This method determines the direction in which water will flow depending on gravity (slope) on a cell-by-cell basis and assigns eight directional information to the cells. After determining the direction in which the water will flow, the flow accumulation is started. Here, each cell is summed towards the outlet of the river. The following process is the construction of the river network and catchments. The main issue is to create catchments and rivers by selecting a threshold value for flow accumulation because this threshold value is very important as it will form a catchment according to the confluence of the rivers obtained after the river network is created (Fig. 2 b). There is no standard threshold for flow accumulation in the literature. However, there is one very important data for answering these questions: the FlooDOT inventory data. When the inventory is examined, it is obvious that even intermittent streams on topography maps have the potential to cause flooding. Therefore, a threshold value of 5000 was chosen for a detailed river network and watershed based on flow accumulation (the number of raster cells that drain into a given cell). According to this selected threshold value, the river construction process was carried out as shown in Fig. 2 c based on pour points. Many large and small river catchments were divided, and inter-catchments together with catchments are defined based on streamlink and pour points (Fig. 2 d). However, in a catchment system, the area between the confluence and the main tributary is shown as a small catchment (inter-catchment). These are actually quite large basins in terms of the area they contain behind them, and the FlooDOT inventory shows that flooding usually occurs in these confluence areas. Geomorphometric studies, whether for tectonic purposes or for flooding or landslides, are traditionally carried out behind the confluence/junction points that connect to the main tributary. In other words, the inter-catchments in Fig. 2 d are always ignored. In this study, for the first time, a cumulative catchment approach was introduced without ignoring inter-catchments (Fig. 2 e). The cumulative catchments are catchments that are divided into smaller sub-catchments according to a specified threshold value, but despite their small size, they reflect their upstream area. However, one of the important issues here is that if this study is carried out by taking into account the topological characteristics of the sub-catchments themselves, the hierarchical structure of the river systems and statistical values of the sub-catchments will be incorrect, as shown in Fig. 2f 1 . In this study, we introduce a cumulative catchment system to represent better geo-environmental predisposing factors of flood events in susceptibility models (Akbas et al., 2023 ; Fig. 2f 2 ). Otherwise, classical catchment morphometry ignores the spatial differences of sub-catchments as shown in Fig. 2f 3 . In order to achieve this task, the Archydro package was used to accumulate statistical properties of sub-catchments (Djokic et al., 2011 ; Maidment, 2022 ). 3.2. Parameter selection for flood susceptibility modelling After construction of cumulative catchments, nine morphometric parameters (Table 1 ), three climatological and hydrological parameters, and one parameter that contains both surface and vegetation parameters (Curve Number) were obtained to use in susceptibility analyses (Fig. 3 ). For catchment morphometry, we determined parameters based on Ozdemir and Akbas ( 2023 ). Table 1 The parameters used in catchment morphometry and flood susceptibility analysis. Name of Parameters Definition, Importance, and Data Sources Equations Topographic Wetness Index ( \(\:TWI\) ) Specific catchment area divided by slope gradient in radians (Beven and Kirkby, 1979 ). This index quantifies the role of topography on hydrological conditions. FADEM is the data source for this and following morphometric parameters \(\:TWI=ln\left(\frac{{A}_{s}}{tan\:\beta\:}\right)\) (1) Strahler Orders (Sth) Merging orders with the same code only increases order power. n + n = n + 1 (2) Area (A) Shows the cumulative area of the catchments. \(\:A=\:\sum\:_{i=1}^{n}A\) (3) Drainage Density ( \(\:{D}_{d}\) ) Drainage length in the catchment divided by the total basin area (Horton, 1945 ). \(\:{D}_{d}=\:\sum\:_{i=1}^{n}L/{A}_{b}\) (4) Form Factor( \(\:{R}_{f}\) ) Catchment area divided by the square of the catchment length (Horton, 1945 ) \(\:{R}_{f}=\frac{A}{{{L}_{b}}^{2}}\) (5) Melton ruggedness number ( \(\:{mR}_{n}\) ) The relief of a catchment is divided by the square root of the catchment area (Melton, 1965 ). \(\:H=\:H/\sqrt{{A}_{b}}\) (6) Time of concentration ( \(\:{T}_{c}\) ) Refers to the time between when the water starts to flow and when it collects in the outlet area ( Giandotti,1934, from Ravazzani et al., 2019 ). \(\:{T}_{c}=\:\frac{4\sqrt{{A}_{b}}+1.5\:L}{0.8\sqrt{{H}_{mean-out}}}\) (7) Stream Power Index ( \(\:SPI)\) The catchment area is divided by a slope gradient in radians (Florinsky, 2012 ; Moore et al., 1991 ). \(\:SPI=ln\left(\frac{{A}_{b}}{tan\:\beta\:}\right)\) (8) Hypsometric Integral ( \(\:{H}_{i}\) ) It refers to the area under the hypsometric curve (Pike and Wilson, 1971 ). \(\:{H}_{i}=\:\frac{\stackrel{̿}{H}-{H}_{min}}{{{H}_{mak}-H}_{min}}\) (9) 100-year precipitation (P100 year) Daily precipitation amount characterized by a 100-year return period computed over 30 years (1989–2018) by the ERA5 model via the GEV model. \(\:G\left(\chi\:\right)=\text{exp}\left\{-{\left[1+\xi\:\left(\frac{\chi\:-\mu\:}{\sigma\:}\right)\right]}^{-1/\xi\:}\right\}\) (10) Aridity The aridity index is a practical indicator of the water balance of catchments. It is a ratio of precipitation and evaporation (Zomer et al., 2022 ). \(\:AI=\:\frac{P}{{E}_{T0}}\) (11) SCS-Curve Number II Curve Number is a combination of both land-use/land-cover (therefore vegetation) and soil. \(\:{CN}_{II}=\:f\left(land-use,\:soil\:group,\:slope\right)\) (12) 100-year discharge (Q100 year) Daily discharge amount characterized by a 100-year return period by the EFAS model via the Gumbel extreme value distribution using L-moments \(\:F\left(\chi\:\right)=exp\:\left[-\text{exp}\left(-\frac{\chi\:-\mu\:}{\sigma\:}\right)\right]\) (13) Here, \(\:\:{A}_{P}\) =Catchment perimeter length (km), \(\:{N}_{u1}\) = Total number of 1st orders according to Strahler order, \(\:\:L\) = Length of river drainage (km), \(\:{A}_{b}\) =Catchment Area (km 2 ), \(\:{L}_{b}\) = Catchment drainage length, \(\:H\) = Relief, \(\:\:{H}_{mean-out}\) =Catchment average-outlet elevation, \(\:{A}_{s}\) = Specific area, \(\:\:tan\:\beta\:\) = slope (in radian), \(\:\mu\:\) = location, \(\:\sigma\:\) =scale and \(\:\xi\:\) =shape, \(\:P\) = Precipitation, \(\:{E}_{T0}\) = Reference Evaporation Although there are many approaches to calculate the Tc parameter, the method by Giandotti ( 1934 ) was used because a methodological study by Ravazzani et al. ( 2019 ) demonstrates that this method more reliably estimates the time of concentration than other methods. On the other hand, extreme precipitation data were obtained from ERA5-Land, which is sufficient to represent the original climate conditions for Türkiye (Akbas and Ozdemir, 2024 ), allowing for the calculation of the 100-year precipitation event. Furthermore, 100-year discharge data were directly obtained from EFAS data ( EFAS, 2024 ). Annual maximum series (AMAX) was decided for extreme precipitation definition (Akbas, 2023 ; Bayliss and Jones, 1993 ; Jones et al., 2014 ), and 100-year precipitation was calculated via Generalized Extreme Value Distribution/GEV distribution (Coles et al, 2001 ). The maximum likelihood (MLE) method has been recommended as the best approach in hydrological studies (Coles et al., 2001 ; Gilleland and Katz, 2016 ), although there are many methods for calculating location, shape, and scale parameters. On the other hand, studies conducted by Seckin et al. ( 2011 ) and Akbas ( 2023 ) focused on the extreme characteristics of precipitation and runoff data for the entire Türkiye. In both studies, it was emphasized that the distribution that best fits the extreme flow and precipitation data is the GEV. Therefore, GEV was used in this study. Aridity was directly obtained from Zomer et al. ( 2022 ). Curve Number II values were generated by overlaying the 2018 Corine landuse and national-scale soil maps (Akbas et al., 2025 ). For more detailed information about the SCS Curve Number methodology, please check Cronshey ( 1986 ). Nine morphometric parameters were calculated using catchment topological properties, except for the topographic wetness index and stream power index. SAGA GIS was employed to calculate these indices. On the other hand, TWI and SPI, CNII, Aridity, and Q100 years were aggregated into catchments using zonal statistics. Consequently, the database is prepared for susceptibility modelling, in which we employed several machine learning techniques. 3.3. Machine learning properties for flood susceptibility analyses The purpose of using machine learning in this study is to generate flood susceptibility maps based on parameters derived from a cumulative catchment analysis. From this perspective, flood susceptibility maps were obtained using some of the most widely used and reportedly most successful models, as discussed by Waled and Sajjad ( 2025 ). In order to obtain a more accurate representation of the flood susceptibility models, we have utilized various frequently used classification models, including Random Forest, Support Vector Machine, XGBoost, and Logistic Regression. The robustness of the performance of these models has been acknowledged, so far, with a wide perspective in earth-science applications, such as landslide (Loche et al., 2022 ), wildfire (Zhang et al., 2019 ), soil erosion (Vergari, 2015 ), and flood studies (Seleem et al., 2022 ). Therefore, we structured the model architecture to train, predict, validate, and evaluate the performance of these models on top of the scikit-learn library (Pedregosa et al., 2011 ) built in Python, which enables us to achieve these analyses through an end-to-end machine learning pipeline. It is essential to recognize that, as we are not directly addressing the performances of many models, the decision of the final model will be determined by expert interpretation in accordance with model performances. Model training for classification We have used the catchments delineated by our cumulative approach as a training dataset, which includes environmental covariate information. The binary classification is essential for susceptibility modelling. Therefore, we classified flooded catchments as 1 and non-flooded basins as 0 using FlooDOT inventories (Akbas et al., 2025 , 2026 ). This resulted in 2300 flooded and 95,000 non-flooded catchments. We employed the “RandomUnderSampler” function (Lemaître et al., 2017 ), a method for addressing imbalanced datasets by randomly selecting a portion of data from the target classes. Subsequently, we applied the Hold-out method, partitioning the dataset into two mutually exclusive subsets: a training set and a test set. As a rule of thumb, we split our dataset into 70% training and 30% independent testing datasets. For the purpose of model optimization, the “Randomized Search with Cross-Validation” method is used within scikit-learn (Pedregosa et al., 2011 ). This optimization, referred to as hyperparameter tuning, involves multiple hyperparameters and seeks to identify the optimal combination using a cross-validation method, which we selected as five in our model training. Each model has been employed to classify the susceptibility classes of unseen catchments by model, with the likelihood of proximity to 1 expressed as a percentage. Model performance evaluation The assessment of various model performances has been achieved using common evaluation metrics, including the ROC-AUC curve, confusion matrix, precision, recall, and F1-score. When integrated, these measurements provide a comprehensive view of a model's performance, offering insights into its overall accuracy, error distribution, and the trade-offs between mitigating false positives and false negatives. Further, overfitting has been evaluated by considering notable disparities between evaluation matrices. This method necessitates a reciprocal evaluation of model performance datasets during both training and testing, wherein we observe no significant indications of overfitting ( Table S1 and Figure S2 ). The importance assessment of covariates Revealing the importance of model inputs to model performance is crucial as machine learning algorithms often encounter black-box issues that impede model interpretability. Therefore, we utilize the SHAP (SHapley Additive exPlanations) package (Lundberg and Lee, 2017 ), which enhances the explainability of model outputs by providing comprehensive insights through the quantification of input contributions. In addition to evaluating feature relevance, Seleem et al. ( 2022 ) indicated that SHAP also assesses whether a feature favours or adversely impacts the expected values. 4. Results and Discussion 4.1. A short glance at model results We used Random Forest, Support Vector Machine, XGBoost, and Logistic Regression models in this study. The purpose of using these models is not to compare models, but to better explain the cumulative catchment model and flood reasons based on the best model. When compared to other algorithms, XGBoost and Random Forest produced the highest results in performance metrics (Table S1 ), demonstrating their relative performance across various accuracy criteria. However, due to its constant superior performance across all assessment criteria on the test set, we selected XGBoost as the final model for the interpretation of flood susceptibility in Türkiye. The model evaluation metrics revealed overall robust results, with an accuracy of 0.72, a macro-average F1-score of 0.72, and a ROC-AUC of 0.80, demonstrating the generalization capacity of the XGBoost model. Furthermore, a study by Waleed and Sajjad (2024) was conducted to compare 14 machine learning models. They have suggested the XGBoost model for flood susceptibility, which proved usable in our study. Our final flood susceptibility map with XGBoost illustrates flood probability ranging from 0 to 1, presented through a continuous gradient scale, which preserves the full variability of model outputs rather than using the natural breaks (Jenks) approach ( Jenks, 1967 ). Each probability score represents a distinct cumulative catchment that we defined to be greater than 97,908 across Türkiye. Even though we maintain our susceptibility values as they were originally determined, the results can be categorized into ten groups based on 0.1 intervals ranging from 0 to 1: very low (0–0.1), low (0.1–0.2), low-moderate (0.2–0.3), moderate (0.3–0.4), moderate-high (0.4–0.5), high (0.5–0.6), high-very high (0.6–0.7), very high (0.7–0.8), extreme (0.8–0.9), and very extreme (0.9–1.0). Overall results demonstrate that approximately 24.8% of 97,908 catchments fall into very low and low susceptibility (9.02% + 15.8%), 33.4% into low-moderate and moderate (17.2% + 16.2%), 24.2% into moderate-high and high (13.7% + 10.5%), 12.5% into high-very high and very high (7.33% + 5.14%), and 5.1% into extreme and very extreme flood susceptible areas (3.72% + 1.36%) 4.2. Flood susceptibility results based on cumulative catchments Flood susceptibility maps have been generated using selected machine learning methods in this study. However, one map stems from the XGBoost method has been presented here based on the best scores of the methods (Fig. 4 ). For other maps, please check the supplementary file. According to Fig. 4 , flood susceptibility has distinct spatial characteristics. One of them is the susceptibility gradient in catchments between the coastal and the inner part of Türkiye. The catchments, along with the coastal part of Türkiye, have the highest flood susceptibility, with values that almost reach 1. However, the coastal part of Türkiye has a distinct gradient as well. For instance, the East and West Black Sea have the highest values, while the middle of this area has the lowest part compared to all coastal parts of Black Sea. Fatal floods are abundant in this area compared to other parts of Türkiye (Fig. 1 ). Moreover, another spatial characteristic is plains, where flood susceptibility has the highest values compared to other geomorphic units. The flood susceptibility values highlight the importance of cumulative catchments, indicating that susceptibility to flooding is higher at outlets compared to the catchment origins. For instance, catchments located in close proximity to the outlets of the longest rivers—Kızılırmak, Sakarya, and Yeşilırmak—have resulted in higher susceptibility to flooding compared to the origin of catchments. Figure 4 illustrates the histograms belonging to the overall flood susceptibility with respect to the catchment’s flood susceptibility. The mean flood susceptibility of Türkiye, indicated by the orange line, is approximately 0.38 (moderate category). Some basins, such as Asi (mean value: 0.52), Gediz (0.46), Küçük Menderes (0.58), Kuzey Ege (0.55), Marmara (0.6), Meriç-Ergene (0.61), and Susurluk (0.47), have a negatively skewed distribution (high values of flood susceptibility) compared to overall susceptibility. These basins contain high flood susceptibility values in almost every part of the area. For example, the average flood susceptibility of the Meriç-Ergene is 0.61, which is significantly higher than the average of the whole country. Nevertheless, some basins, such as Akarçay (0.34), Aras (0.31), Burdur (0.33), Çoruh (0.30), Konya Kapalı (0.26), Van Gölü (0.28), have a positively skewed distribution in which negative values (low values) are most dominant with respect to other catchments. The Çoruh major basin has higher relief (check the Melton map in Fig. 2 ). On the other hand, Doğu Karadeniz, which has the most fatal floods compared to other areas, has a bimodal distribution. Flood susceptibility is higher at the outlets of individual catchments than in their upper parts in the Doğu Karadeniz basin. This condition is also supported by Elbası ( 2022 ), who executed a hydrodynamic model for the Doğu Karadeniz, and found that the flood hazard is higher at the outlet of the Doğu Karadeniz basin. Furthermore, certain basins almost exactly represent the flood susceptibility distribution belonging to Türkiye. These area, Doğu Akdeniz (0.37), Sakarya (0.37), Seyhan (0.35), Yeşilırmak (0.38), Kızılırmak (0.35), and Fırat-Dicle (0.32). In these areas, very extreme flood susceptibility is observed, but these values are only prevalent in the river course, the immediate surroundings of the river, and its outlet. For example, the map in Fig. 4 shows the Kızılırmak, which is the second-largest basin after the Fırat-Dicle, and susceptibility values in the histogram follow a similar distribution to that of the overall country. However, in the lower reaches of the river, in the sub-catchments located near the main outlet, flood sensitivity has been observed to reach extremely high values. This observation underscores the crucial importance of adopting a cumulative catchment-based approach in flood risk assessments, particularly for the Sakarya River, despite its large-scale basin characteristics. This situation shows that it is more effective to examine the effects of flooding by dividing the catchment area into smaller sections than to consider the entire catchment morphology. Understanding the flood-generating mechanism is substantial for flood susceptibility. The cumulative catchment-based approach yields information on both the properties of the catchments and may provide insight into how these properties influence flood-generating mechanisms using tools of machine learning techniques. Figure 5 illustrates the bivariate map between susceptibility values and parameters, together with SHAP graphs to understand the impact of parameters on flood susceptibility. High values of time concentration (see SHAP), for instance, are essential for model construction and high susceptibility. Because the model has all high-high groups overlapping in the bivariate map, except for some lower time of concentration values in floodplains, where susceptibility is higher. In hydrology, it is the most important and widely used parameter for designing discharge in various methods, such as the SCS curve number and the rational method ( Chow et al., 1988 ). These results demonstrate that higher values of time of concentration, as determined by cumulative catchments, contribute to flood generation through the accumulation of water in channels. Furthermore, 100-year runoff also reveals the importance of main river courses and the patterns where higher aridity values (i.e., positive water balance) are higher. Additionally, SHAP values highlight the high values of 100-year runoff in the model setup, demonstrating their role in the flood-generation mechanism in cumulative catchments, particularly in main river courses and areas with high precipitation values. Furthermore, higher values of the bivariate map between Strahler and flood susceptibility are only dominant in the main river courses, unlike 100-year runoff values. Climatic parameters, such as aridity and 100-year precipitation, are important in certain basins, including Doğu Karadeniz, Batı Karadeniz, Seyhan, Ceyhan, and Susurluk (see Fig. 4 ), with specific values influencing flood-generating mechanisms, as illustrated by SHAP values and bivariate maps. SHAP graph highlights the importance of a higher values aridity (positive water balance) index than the 100-year precipitation. Guo et al. ( 2014 ) have expressed that the aridity index is first-order responsible for the shape of the flood frequency curve. Therefore, some areas, such as Doğu Karadeniz, have the highest flood probabilities despite catchment characteristics (such as time of characteristics, Melton ruggedness number) that are not suitable for floods. The Melton ruggedness number is one of the critical parameters that control or affect flood characteristics, which are defined as flood and debris floods, as well as debris flows, by catchment properties (Wilford et al., 2004 ; Church and Jakob, 2020 ). SHAP values mark higher values of these parameters. When Melton ruggedness numbers increase, it indicates the debris flood characteristics. SHAP values demonstrate that, in the construction of the XGBoost model, debris characteristic catchments have a greater influence than flood characteristic catchments. Additionally, a bivariate map between Melton ruggedness numbers and flood susceptibility highlights the debris characteristics of catchments, where high flood susceptibility values correspond to catchments with debris floods. Stolle et al. ( 2015 ) expressed that the Melton ruggedness number can only explain debris susceptibility. The topographic wetness index and stream power index are the most commonly used and critical parameters in flood susceptibility assessments ( Pourzangbar et al., 2025 ). High topographic wetness index values indicate that water remains in the flat areas of the catchment for an extended period, and the catchments are prone to flooding. SHAP values emphasize the lower values of the stream power index, while higher values of the topographic wetness index are more influential in model construction; however, lower values are also shown to be influential. A bivariate map between the topographic wetness index and flood susceptibility shows the importance of flat areas. Despite the catchments of the Central Anatolian plateau, such as Konya, Akarçay, and Burdur, being completely flat and having high topographic wetness index values, it is observed that flood susceptibility values are low. Moreover, the stream power index has patterns almost similar to Melton's ruggedness numbers, in which debris characteristic catchments are important. On the other hand, area is an important parameter for flood analyses because it is an irreplaceable parameter of regional flood frequency analyses (Hosking and Wallis 1997 ; Kumar and Chatterjee, 2005 ). However, as can be seen from the data obtained from the SHAP graph, it has not been observed to have a significant importance in model construction. It is assumed that this is likely due to other morphometric parameters, including characteristics of the area. However, the bivariate map reveals the importance of the area on the main river courses. Moreover, SHAP values indicate that form factor has no influence on model construction, which is also observed in the bivariate map. Drainage density follows this pattern as well. SHAP values and the bivariate map show a similar, less impactful effect compared to the form factor. The SCS Curve number is an important and frequently used parameter in hydrology for predicting excess rainfall and, consequently, surface runoff ( Chow et al., 1988; Cronshey, 1986 ). High Curve Number values indicate an empirical relationship showing that precipitation falling into the catchment area is likely to turn into surface runoff. In this study, higher values of curve number influence model construction, as illustrated in the SHAP figure. However, this relationship is not similar in every catchment of Türkiye. Some DSI main basins such as Doğu Karadeniz, Batı Karadeniz, Seyhan, Ceyhan, and the main river courses of Büyük/Küçük Menderes support this idea. High Curve number values in other large DSI major basins have not affected flood sensitivity. The overall results illustrate that catchments with high flood susceptibility in Türkiye are clearly defined in areas where all parameters (climate, morphometry, and hydrology) are optimized, and the cumulative catchment makes these even more pronounced. 6. Conclusion This study presents a novel approach to assessing catchment morphometry in relation to flood susceptibility. Classical morphometric approaches, which belong to catchment morphometry, ignore the sub-catchments, and this does not represent proper catchment morphometry, as it cannot explain flood-generating mechanisms. However, the flood inventories indicate that floods are concentrated in inter-catchments. Therefore, we propose a novel approach, termed cumulative catchments, which are derived from small sub-catchments and capture both the effects of upstream catchments and the contributing area behind them. This approach not only accounts for the influence of inter-catchments on catchment morphometry but also allows for a more detailed spatial representation of flood susceptibility (i.e., high resolution in terms of sub-catchment size) based on morphometric characteristics. Using hydrography data from FABDEM and flood inventory belonging to Türkiye, we explained the flood susceptibility via a cumulative catchment approach based on morphometric indices and climate and hydrology through the XGBoost machine learning method. Results showed that the cumulative catchment approach does not ignore the inter-catchments and provides a more detailed and reliable susceptibility distribution map. In this perspective, for the first time, a regional flood susceptibility result for Türkiye is obtained, and it is ready for flood management and hazard management. For instance, in Türkiye, it was determined that the Meriç within the DSI basins has the highest probability of flooding, and the spatial distribution of high flood sensitivity values was also identified. However, this approach can be applied to any region or continent worldwide. On the other hand, the flood-generating mechanism and the importance of these parameters, which are related to catchment morphometry, climate, and hydrology, were identified. It has been concluded that higher flood sensitivity values correspond to optimal conditions for catchment morphometry and the hydro-climatological characteristics of catchments. Additionally, we recognized that although SHAP values indicate the relative importance of parameters when constructing a model, bivariate maps have demonstrated that this importance varies spatially. This study presents a novel approach to catchment morphometry, despite utilizing widely employed machine learning methods. Therefore, it is also possible to compare machine learning methods with various network-based and algorithm-based machine learning models in watershed morphometry and flood susceptibility, and to conduct a separate study on the most meaningful results in flood susceptibility. On the other hand, considering the climatic conditions and changes in land use in the catchment areas, adopting a dynamic-based cumulative catchment approach, such as climate-change conditioned flood susceptibility, is feasible. The Cumulative catchments approach could be beneficial in determining the flood potential of basins in data-scarce regions where hydrodynamic modelling data is unavailable. Furthermore, cumulative basin outcomes may be helpful to policymakers in developing flood-related policies. Declarations Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Code availability: Not applicable Ethics approval: Not applicable Acknowledgment: This study was funded by TUBITAK-Scientific and Technological Research Council of Türkiye with 3501 Career Development Program (CAREER) (Project No: 121Y578). Data Availability Statement: The data that support the findings of this study are available from the corresponding author upon reasonable request. Authors' contributions Conceptualization : Abbreviations: [AA: Abdullah AKBAS, AAVCI: Aydogan AVCIOGLU, HO: Hasan OZDEMİR, TG: Tolga GORUM, PB: Paul BATES], Methodology: [AA, AAVCI, HO, TG, PB]; Software and Validation: [AA, AAVCI]; Formal analysis and investigation: [AA]; Writing - original draft preparation: [AA, AAVCI]; Writing - review and editing: [AA, AAVCI, HO, TG, PB]; Supervision: [AA, AAVCI, HO, TG, PB]. All authors read and approved the final manuscript. References Addor, N., Nearing, G., Prieto, C., Newman, A. J., Le Vine, N., & Clark, M. P. (2018). A ranking of hydrological signatures based on their predictability in space. Water Resources Research, 54(11), 8792-8812. Alfieri, L., Burek, P., Feyen, L., & Forzieri, G. (2015). Global warming increases the frequency of river floods in Europe. Hydrology and Earth System Sciences, 19(5), 2247-2260. Akbas, A. (2023). Seasonality, persistency, regionalization, and control mechanism of extreme rainfall over complex terrain. Theoretical and Applied Climatology, 152(3-4), 981-997. Akbas, A., Gorum, T., Ozdemir, H., Doğan, E., & Çorapcı, F. (2023). Hierarchical basins: A new approach to quantify geomorphometric flood susceptibility in the regional aspect. In Proceedings of the Regional Conference of Geomorphology Cappadocia 2023 (pp. 51). Nevşehir, Türkiye. Akbaş, A., Görüm, T., & Özdemir, H. (2024). Türkiye’de Taşkınları Oluşturan Mekanizmaların ve Taşkın Tehlikenin Alansal ve Zamansal Dinamiklerinin Ortaya Çıkarılması. 121Y578 Nolu Tübitak Projesi Sonuç Raporu. Akbas, A., & Ozdemir, H. (2024). Comparing Satellite, Reanalysis, Fused and Gridded (In Situ) Precipitation Products Over Türkiye. International Journal of Climatology, 44(16), 5873-5889. Akbas, A., Görüm, T., & Özdemir, H., (2025). FlooDOT-based Analysis of Floods in Turkey and Spatial Comparison of Various Data Sources. TURKISH JOURNAL OF CIVIL ENGINEERING, 36(6). Akbas, A., Gorum, T., and Ozdemir, H. (2026). FlooDOT (FlooD inventory Of Türkiye): A comprehensive flood inventory and its spatio-temporal analyses, J. Flood Risk Manag., under review. Andreadis, K. M., Wing, O. E., Colven, E., Gleason, C. J., Bates, P. D., & Brown, C. M. (2022). Urbanizing the floodplain: global changes of imperviousness in flood-prone areas. Environmental Research Letters, 17(10), 104024. Aydoğan, D., Kankal, M., & Önsoy, H. (2016). Regional flood frequency analysis for Çoruh Basin of Turkey with L‐moments approach. Journal of Flood Risk Management, 9(1), 69-86. Bayliss, A. C., & Jones, R. C. (1993). Peaks-over-threshold flood database: Summary statistics and seasonality (IH Report No. 121). Institute of Hydrology, Wallingford, UK. Retrieved from http://nora.nerc.ac.uk/id/eprint/6075/1/IH_121.pdf Beven, K. J., & Kirkby, M. J. (1979). A physically based, variable contributing area model of basin hydrology/Un modèle à base physique de zone d'appel variable de l'hydrologie du bassin versant. Hydrological Sciences Journal, 24(1), 43-69. Bhat, M. S., Alam, A., Ahmad, S., Farooq, H., & Ahmad, B. (2019). Flood hazard assessment of upper Jhelum basin using morphometric parameters. Environmental Earth Sciences, 78(2), 54. Blöschl, G., Hall, J., Viglione, A., Perdigão, R. A., Parajka, J., Merz, B., ... & Živković, N. (2019). Changing climate both increases and decreases European river floods. Nature, 573(7772), 108-111. Chissende, A., Mehmood, M. S., Huo, A., Lou, P., & Wang, Z. (2025). Riverine flood susceptibility assessment using drainage basin morphometry in the Bahe River. International Journal of Disaster Risk Reduction, 117, 105181. Chorley, R.J. and Kennedy, B.A., 1971. Physical geography: a systems approach. London: Prentice-Hall. . Chorley, R.J. (2021). The drainage basin as the fundamental geomorphic unit. In Water, earth, and man: A synthesis of hydrology, geomorphology, and socio-economic geography. Routledge. https://doi.org/10.4324/9781003170181 Chow, V. T., D. R. Maidment, and L. W. Mays. 1988. Applied hydrology. Singapore: McGraw-Hill. Church, M., & Jakob, M. (2020). What is a debris flood?. Water resources research, 56(8), e2020WR027144. CLMS-Copernicus Land Monitoring Service, 2025, EU-Hydro, Database [Available online: https://doi.org/10.2909/393359a7-7ebd-4a52-80ac-1a18d5f3db9c] Cronshey, R. (1986). Urban hydrology for small watersheds (No. 55). US Department of Agriculture, Soil Conservation Service, Engineering Division. CRED & UNDRR. (2020). The Human Cost of Disasters: An Overview of the Last 20 Years (2000-2019). UN Office for Disaster Risk Reduction. Access link: https://www.undrr.org/publication/human-cost-disasters-overview-last-20-years-2000-2019 Coles, S., Bawa, J., Trenner, L., & Dorazio, P. (2001). An introduction to statistical modeling of extreme values (Vol. 208, p. 208). London: Springer. Djokic, D., Ye, Z., & Dartiguenave, C. (2011). Arc hydro tools overview. Redland, Canada, ESRI, 5. EFAS (2023), European Flood Awareness System (EFAS) Data, DOI: 10.24381/cds.e3458969 Elbası, A. (2022). Producing flood hazard maps using regional flood frequency [PhD thesis], Istanbul University (In Turkish) Florinsky, I.V. (2012) Digital Terrain Analysis in Soil Science and Geology. Academic Press is an imprint of Elsevier. The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK, 379p. Gao, H., Fenicia, F., & Savenije, H. H. (2023). HESS Opinions: Are soils overrated in hydrology?. Hydrology and Earth System Sciences, 27(14), 2607-2620. Ghasemlounia, R., & Utlu, M. (2021). Flood prioritization of basins based on geomorphometric properties using principal component analysis, morphometric analysis and Redvan’s priority methods: A case study of Harşit River basin. Journal of Hydrology, 603, 127061. Giandotti, M. 1934. Previsione delle piene e delle magre dei corsi d’acqua, 107. Rome: Memorie e studi idrografici, Servizio Idrografico Italiano Gilleland E, Katz RW (2016) extRemes 2.0: an extreme value analysis package in R. J Stat Softw 72(8):1–39 Gorum, T. (2019). Landslide recognition and mapping in a mixed forest environment from airborne LiDAR data. Engineering Geology, 258, 105155. Gorum, T., & Fidan, S. (2021). Spatiotemporal variations of fatal landslides in Turkey. Landslides, 18(5), 1691-1705. Hawker, L., Uhe, P., Paulo, L., Sosa, J., Savage, J., Sampson, C., & Neal, J. (2022). A 30 m global map of elevation with forests and buildings removed. Environmental Research Letters, 17(2), 024016 Horton, R. E. (1932). Drainage-basin characteristics. Eos, Transactions American Geophysical Union, 13(1), 350–361. https://doi.org/10.1029/tr013i001p00350 Horton, R. E. (1945). Erosional development of streams and their drainage basins; hydrophysical approach to quantitative morphology. Geological society of America bulletin, 56(3), 275-370. https://doi.org/10.1130/0016-7606(1945)56[275:EDOSAT]2.0.CO;2 Hosking JRM, Wallis JR (1997) Regional frequency analysis: an approach based on L-moments. Cambridge University Press, Cambridge, p 240 Islam, A., & Deb Barman, S. (2020). Drainage basin morphometry and evaluating its role on flood-inducing capacity of tributary basins of Mayurakshi River, India. SN Applied Sciences, 2(6), 1087. Jenson, S., & Domingue, J. (1988). Extracting topographic structure from digital elevation data for geographic ınformation system analysis. Photogrammetric Engineering And Remote Sensing, 54(11), 1593–1600 Jones, M. R., Blenkinsop, S., Fowler, H. J., & Kilsby, C. G. (2014). Objective classification of extreme rainfall regions for the UK and updated estimates of trends in regional extreme rainfall. International Journal of Climatology, 34(3), 751-765. O'Callaghan, J. F., & Mark, D. M. (1984). The extraction of drainage networks from digital elevation data. Computer vision, graphics, and image processing, 28(3), 323-344. Karaca, M., Deniz, A., & Tayanç, M. (2000). Cyclone track variability over Turkey in association with regional climate. International Journal of Climatology: A Journal of the Royal Meteorological Society, 20(10), 1225-1236. Khodaei, H., Nasiri Saleh, F., Nobakht Dalir, A., & Zarei, E. (2025). Future flood susceptibility mapping under climate and land use change. Scientific Reports, 15(1), 12394. Koç, G., Petrow, T., & Thieken, A. H. (2020). Analysis of the most severe flood events in Turkey (1960–2014): which triggering mechanisms and aggravating pathways can be identified? Water, 12(6), 1562. Kreibich, H., Van Loon, A. F., Schröter, K., Ward, P. J., Mazzoleni, M., Sairam, N., ... & Di Baldassarre, G. (2022). The challenge of unprecedented floods and droughts in risk management. Nature, 608(7921), 80-86. Kumar, R., & Chatterjee, C. (2005). Regional flood frequency analysis using L-moments for North Brahmaputra region of India. Journal of Hydrologic Engineering, 10(1), 1-7. Lemaître, G., Nogueira, F., & Aridas, C. K. (2017). Imbalanced-learn: A Python toolbox to tackle the curse of imbalanced datasets in machine learning. Journal of Machine Learning Research, 18(17), 1–5. Lehner, B., Verdin, K., & Jarvis, A. (2008). New global hydrography derived from spaceborne elevation data. Eos, Transactions American Geophysical Union, 89(10), 93-94. Loche, M., Alvioli, M., Marchesini, I., Bakka, H., & Lombardo, L. (2022). Landslide susceptibility maps of Italy: Lesson learnt from dealing with multiple landslide types and the uneven spatial distribution of the national inventory. Earth-Science Reviews, 232, 104125. Lundberg SM, Lee SI. 2017. A unified approach to interpreting model predictions. In: Proceedings of the 31st international conference on neural information processing systems, p. 4768–4777. Maidment, D. R., & Morehouse, S. (2002). Arc Hydro: GIS for water resources. ESRI. Inc., Redlands, CA. Magilligan, F. J. (1992). Thresholds and the spatial variability of flood power during extreme floods. Geomorphology, 5(3–5), 373–390. https://doi.org/10.1016/0169-555x(92)90014-f Melton, M. A. (1965). The geomorphic and paleoclimatic significance of alluvial deposits in southern Arizona. The Journal of geology, 73(1), 1-38. https://www.jstor.org/stable/30066379 Moore, I. D., & Grayson, R. B. (1991). Terrain‐based catchment partitioning and runoff prediction using vector elevation data. Water Resources Research, 27(6), 1177-1191. Morel, M., Piton, G., Kuss, D., Evin, G., & Le Bouteiller, C. (2023). Statistical modeling of sediment supply in torrent catchments of the northern French Alps. Natural Hazards and Earth System Sciences, 23(5), 1769-1787. Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., ... & Duchesnay, É. (2011). Scikit-learn: Machine learning in Python. the Journal of machine Learning research, 12, 2825-2830. Pike, R. J., & Wilson, S. E. (1971). Elevation-relief ratio, hypsometric integral, and geomorphic area-altitude analysis. Geological Society of America Bulletin, 82(4), 1079-1084. https://doi.org/10.1130/0016-7606(1971)82[1079:ERHIAG]2.0.CO;2 Rogelis, M. C., & Werner, M. (2014). Regional debris flow susceptibility analysis in mountainous peri-urban areas through morphometric and land cover indicators. Natural Hazards and Earth System Sciences, 14(11), 3043-3064. Seleem, O., Ayzel, G., de Souza, A. C. T., Bronstert, A., & Heistermann, M. (2022). Towards urban flood susceptibility mapping using data-driven models in Berlin, Germany. Geomatics, Natural Hazards and Risk, 13(1), 1640-1662. Seckin, N., Haktanir, T., & Yurtal, R. (2011). Flood frequency analysis of Turkey using L‐moments method. Hydrological Processes, 25(22), 3499-3505. Stolle, A., Langer, M., Blöthe, J. H., & Korup, O. (2015). On predicting debris flows in arid mountain belts. Global and Planetary Change, 126, 1-13. Strahler, A. N. (1980). Systems theory in physical geography. Physical Geography, 1(1), 1-27. Obeidat, M., Awawdeh, M., & Al‐Hantouli, F. (2021). Morphometric analysis and prioritisation of watersheds for flood risk management in Wadi Easal Basin (WEB), Jordan, using geospatial technologies. Journal of Flood Risk Management, 14(2), e12711. O'Callaghan, J. F., & Mark, D. M. (1984). The extraction of drainage networks from digital elevation data. Computer vision, graphics, and image processing, 28(3), 323-344. Ozdemir, H., & Bird, D. (2009). Evaluation of morphometric parameters of drainage networks derived from topographic maps and DEM in point of floods. Environmental geology, 56(7), 1405-1415. Ozdemir, H., & Elbaşı, E. (2015). Benchmarking land use change impacts on direct runoff in ungauged urban watersheds. Physics and Chemistry of the Earth, Parts A/B/C, 79, 100-107. Ozdemir, H., & Akbas, A. (2023a). Is there a consistency in basin morphometry and hydrodynamic modelling results in terms of the flood generation potential of basins? A case study from the Ulus River Basin (Türkiye). Journal of Hydrology, 625, 129926. Ozdemir, H. & Akbas, A. (2023b). The Effects of Spatial Resolution Variability of Digital Elevation Models on Flood Hazard Analysis. JOURNAL OF GEOGRAPHY, (46), 137-156. https://doi.org/10.26650/JGEOG2023-1177718 Ozdemir, H., & Akbas, A. (2025). Evaluating the capability of UAV-Derived DEMs for flood hazard analysis. Physical Geography, 46(3), 175-195. Patton, P. C., & Baker, V. R. (1976). Morphometry and floods in small drainage basins subject to diverse hydrogeomorphic controls. Water Resources Research, 12(5), 941–952. https://doi.org/10.1029/wr012i005p00941 Ravazzani, G., Boscarello, L., Cislaghi, A., & Mancini, M. (2019). Review of time-of-concentration equations and a new proposal in Italy. Journal of Hydrologic Engineering, 24(10), 04019039. Rentschler, J., Avner, P., Marconcini, M., Su, R., Strano, E., Vousdoukas, M., & Hallegatte, S. (2023). Global evidence of rapid urban growth in flood zones since 1985. Nature, 622(7981), 87-92. Yamazaki, D., Ikeshima, D., Tawatari, R., Yamaguchi, T., O'Loughlin, F., Neal, J. C., ... & Bates, P. D. (2017). A high‐accuracy map of global terrain elevations. Geophysical Research Letters, 44(11), 5844-5853. Yüksek, Ö., Kankal, M., & Üçüncü, O. (2013). Assessment of big floods in the Eastern Black Sea Basin of Turkey. Environmental monitoring and assessment, 185(1), 797-814. Wilford, D. J., Sakals, M. E., Innes, J. L., Sidle, R. C., & Bergerud, W. A. (2004). Recognition of debris flow, debris flood and flood hazard through watershed morphometrics. Landslides, 1(1), 61-66. Tarboton, D. G. 1997. "A new method for the determination of flow directions and upslope areas in grid digital elevation models." Water Resources Research 33(2): 309-319. Tatli, H., Nüzhet Dalfes, H., & Sibel Menteş, Ş. (2004). A statistical downscaling method for monthly total precipitation over Turkey. International Journal of Climatology: A Journal of the Royal Meteorological Society, 24(2), 161-180. Tola, S. Y., & Shetty, A. (2022). Flood susceptibility modeling based on morphometric parameters in Upper Awash River basin, Ethiopia using geospatial techniques. Sustainable Water Resources Management, 8(2), 49. Türkeş, M. (1996). Spatial and temporal analysis of annual rainfall variations in Turkey. International Journal of Climatology: A Journal of the Royal Meteorological Society, 16(9), 1057-1076. Waleed, M., & Sajjad, M. (2025). Advancing flood susceptibility prediction: A comparative assessment and scalability analysis of machine learning algorithms via artificial intelligence in high-risk regions of Pakistan. Journal of Flood Risk Management, 18(1), e13047. https://doi.org/10.1111/jfr3.13047 Vergari, F. (2015). Assessing soil erosion hazard in a key badland area of Central Italy. Natural Hazards, 79(Suppl 1), 71-95. Víg, B., Fábian, S. Á., Czigány, S., Pirkhoffer, E., Halmai, Á., Kovács, I. P., ... & Lóczy, D. (2022). Morphometric analysis of low mountains for mapping flash flood susceptibility in headwaters. Natural Hazards, 114(3), 3235-3254. Zhang, G., Wang, M., & Liu, K. (2019). Forest fire susceptibility modeling using a convolutional neural network for Yunnan province of China. International Journal of Disaster Risk Science, 10(3), 386-403. Zomer, R. J., Xu, J., & Trabucco, A. (2022). Version 3 of the global aridity index and potential evapotranspiration database. Scientific Data, 9(1), 409. Supplementary Files SupplementaryFile.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 21 Jan, 2026 Reviewers invited by journal 21 Jan, 2026 Editor invited by journal 13 Jan, 2026 Editor assigned by journal 16 Dec, 2025 First submitted to journal 16 Dec, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8379220","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":578161561,"identity":"1d194cdb-8511-4f21-acdb-da9cab62da86","order_by":0,"name":"Abdullah Akbas","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0003-2024-0565","institution":"Bursa Uludağ University","correspondingAuthor":true,"prefix":"","firstName":"Abdullah","middleName":"","lastName":"Akbas","suffix":""},{"id":578161562,"identity":"6ec2096e-3830-41e0-bfff-a71235111838","order_by":1,"name":"Aydoğan AVCIOĞLU","email":"","orcid":"","institution":"BRGM: Bureau de Recherches Geologiques et Minieres","correspondingAuthor":false,"prefix":"","firstName":"Aydoğan","middleName":"","lastName":"AVCIOĞLU","suffix":""},{"id":578161563,"identity":"a37ae54d-89c3-482f-86fb-423767bb1fe6","order_by":2,"name":"Hasan OZDEMIR","email":"","orcid":"","institution":"Bursa Uludag University: Bursa Uludag Universitesi","correspondingAuthor":false,"prefix":"","firstName":"Hasan","middleName":"","lastName":"OZDEMIR","suffix":""},{"id":578161564,"identity":"59bb35ea-6d9f-425e-b0d0-78ec2651cde2","order_by":3,"name":"Tolga GORUM","email":"","orcid":"","institution":"Istanbul Technical University - Ayazaga Campus: Istanbul Teknik Universitesi","correspondingAuthor":false,"prefix":"","firstName":"Tolga","middleName":"","lastName":"GORUM","suffix":""},{"id":578161565,"identity":"74dde280-a39f-4e4b-ada5-457528cdfbb5","order_by":4,"name":"Paul Bates","email":"","orcid":"","institution":"Bristol University","correspondingAuthor":false,"prefix":"","firstName":"Paul","middleName":"","lastName":"Bates","suffix":""}],"badges":[],"createdAt":"2025-12-16 19:00:45","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8379220/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8379220/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":102177771,"identity":"276df1e2-7fa0-4f5b-b70b-e6a107211a56","added_by":"auto","created_at":"2026-02-09 06:42:29","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":11273149,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of floods and flood causalities in Türkiye. Figure demonstrate the fatalities with changing dot along with bivariate maps with number of injured and lost person due to floods.\u003c/p\u003e","description":"","filename":"Figure1.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8379220/v1/a4641132321a03ed518258be.jpg"},{"id":102177827,"identity":"046ceaaf-638d-4059-ad0d-1cf898f3cfb8","added_by":"auto","created_at":"2026-02-09 06:42:54","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":4650942,"visible":true,"origin":"","legend":"\u003cp\u003eFlow chart of this study. In this study a) Flow direction b) flow accumulation c) streamlink d) catchments and inter-catchments e) sub-catchments, and f) cumulative catchments.\u003c/p\u003e","description":"","filename":"Figure2.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8379220/v1/967bf85867e3209f1b31afa0.jpg"},{"id":102177744,"identity":"09aeb1a7-ce83-47a5-b9be-9c36dfe01307","added_by":"auto","created_at":"2026-02-09 06:42:22","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":19569270,"visible":true,"origin":"","legend":"\u003cp\u003eCovariates that used in machine learning. The maps show Topographic Wetness Index (TWI), Catchment Area (A), Stream Power Index (SPI), Hypsometric Integral (Hi), 100-year return level of precipitation event (P100yr), 100-year return level of runoff event (Q100yr), Curve Number II (CNII), Time of concentration (Tc), Form factor (Rb), Drainage density (Dd), Melton topographic ruggedness index (Melton), UN Aridity index (Aridity), Strahler river order (Strahler), and subcatments, respectively.\u003c/p\u003e","description":"","filename":"Figure3.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8379220/v1/42c2a910bd4e97d8a9d838b8.jpg"},{"id":102177772,"identity":"34ff1d0e-cf0d-429a-8652-1e4be0011bb1","added_by":"auto","created_at":"2026-02-09 06:42:29","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":11045824,"visible":true,"origin":"","legend":"\u003cp\u003eThe patterns of flood susceptibility obtained from XGBoost model. Figure depicts the distribution of susceptibility, while histograms illustrate flood susceptibility for each catchments compared to overall flood susceptibility results. Flu areas show 25\u003csup\u003eth\u003c/sup\u003e and 75\u003csup\u003eth\u003c/sup\u003e percentiles while solid blue line demonstrate mean value of catchment flood susceptibility. Orange dashed line shows the mean value of overall flood susceptibility of Türkiye.\u003c/p\u003e","description":"","filename":"Figure4.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8379220/v1/aeaeccdccaa8f85e228ad273.jpg"},{"id":102177736,"identity":"0c220c78-ac91-4138-a4be-578c711ca375","added_by":"auto","created_at":"2026-02-09 06:42:17","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":15165328,"visible":true,"origin":"","legend":"\u003cp\u003eIllustration of bivariate maps and SHAP graphs. Bivariate map constructed between flood susceptibility and 13 parameters used in model. Also SHAP values demonstrate relative importance of each value parameters on model construction.\u003c/p\u003e","description":"","filename":"Figure5.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8379220/v1/d1f2e39cc3fc09c66ddbb9fb.jpg"},{"id":102296723,"identity":"ed0a72b4-7df4-4c2c-badd-83644bd527c4","added_by":"auto","created_at":"2026-02-10 10:21:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":62712954,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8379220/v1/9997fa32-04b6-4dc3-840e-1cc35e9d7036.pdf"},{"id":102177677,"identity":"6766d73d-a848-4423-9cee-55fcd811c62b","added_by":"auto","created_at":"2026-02-09 06:42:06","extension":"docx","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":24000807,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFile.docx","url":"https://assets-eu.researchsquare.com/files/rs-8379220/v1/12c2dfd56eb4f555570e256a.docx"}],"financialInterests":"","formattedTitle":"Cumulative catchments: A novel approach for catchment-based regional flood susceptibility","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eFloods are one of the catastrophic hazards, causing many fatalities and devastating consequences worldwide every year. The document by the Center for Research on the Epidemiology of Disasters (CRED), based on EMDAT (International Disaster Database) data, reports that only floods caused 104.000 deaths and cost 651\u0026nbsp;billion dollars between 2000 and 2019 (\u003cb\u003eCRED and UNDRR, 2020\u003c/b\u003e). Although floods occur in rivers, they are the output of many processes and are affected by activities within the catchment area.\u003c/p\u003e \u003cp\u003eOn the one hand, catchments are the fundamental unit of hydrological systems where all fluxes such as energy and mass, interact (\u003cb\u003eChorley \u0026amp; Kennedy,1971;\u003c/b\u003e Strahler, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e1980\u003c/span\u003e). Consequently, all activities within the catchment can cause positive or negative feedback on the fluxes, thereby altering their quantity and direction. Climate, for instance, is one of the important reasons that can change the feedback mechanism in catchments. Due to climate change, the temporal and spatial properties of various climatic parameters, including precipitation, soil moisture, and snowmelt, may impact the feedback mechanisms of catchments. Additionally, changes in these conditions due to climate change may lead to different feedback mechanisms within hydro-climatological subsystems (Alfieri et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Kreibich et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Bl\u0026ouml;schl et al. (\u003cb\u003e2020\u003c/b\u003e), for example, investigated changes in hydro-climatological variables, such as precipitation, snowmelt, and temperature, across Europe. They concluded that changes in the magnitudes of these parameters over time affect floods and their trends, with regional differences. Furthermore, many anthropogenic activities in catchments also drastically change the feedback mechanisms (Rentschler et al., \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; \u003cb\u003eYang \u0026amp; Tian, 2009\u003c/b\u003e). Andreadis et al. (\u003cb\u003e2023\u003c/b\u003e), for example, demonstrate that urbanization in floodplains since 1985 has dramatically increased, and this has led to an increase in the number of people affected by floods, even though the frequency of floods has not increased.\u003c/p\u003e \u003cp\u003eOn the other hand, catchment properties can affect fluxes such as runoff and peak flow. Chorley (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) has stated that catchment properties can influence peak runoff. The first attempt to introduce the impact of catchment properties on hydrology and runoff characteristics was executed by Horton (\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e1932\u003c/span\u003e). Horton explained the impact of catchment properties on flood characteristics using different morphometric indices such as form factor, and drainage density. From that day onward, numerous studies have investigated flood susceptibility using catchment morphometry (Bhat et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Chissende et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Patton and Baker, \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e1976\u003c/span\u003e; Florinsky, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; \u003cb\u003eGhasemlounia and Utlu, 2018;\u003c/b\u003e Islam and Deb Barman; \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Khodaei et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Magilligan, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e1992\u003c/span\u003e; Morel et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Obeidat et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Ozdemir and Bird \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Pike and Wilson, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e1971\u003c/span\u003e; \u003cb\u003eRogelis et al., 2014;\u003c/b\u003e Rogelis and Werner, \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Tola and Shetty, \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e2022\u003c/span\u003eg et al., \u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Ghasemlounia and Utlu (\u003cb\u003e2018\u003c/b\u003e) have attempted to prioritize the flood susceptibility using catchment morphometry, and they expressed that prioritizing flood susceptibility based on the ranking of parameters for catchment morphometry has also been the result of subjective or expert-based knowledge. On the other hand, Stolle et al. (\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) have investigated the role of catchment properties in predicting flows in fluvial and debris flow-dominated catchments using morphometric characteristics (catchment slope, area, etc.) via statistical methods. They concluded that morphometric characteristics are significant in determining flood and debris flow risk; when correctly selected and analyzed, their predictive power can be high. On the other hand, a study by Ozdemir and Akbas (\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2023a\u003c/span\u003e) proved that the results of the flood hazard analysis using the hydrodynamic model (LISFLOOD-FP) and a morphometric flood susceptibility model for the catchments gave similar results, and that the morphometric flood susceptibility model can explain potential flood hazards in the catchments.\u003c/p\u003e \u003cp\u003eResearchers have primarily used catchment morphometry parameters, such as drainage density, form factor, and time of concentration, for the interested catchment and attempted to understand the impact of these properties on hydrological conditions, including peak flow. However, since Horton (\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e1932\u003c/span\u003e), all morphometric methodological explanations on flood-generating mechanisms and catchment morphometry have been built upon single and individual catchments, and sub-catchments have been ignored. For instance, Patton and Baker (\u003cb\u003e1967\u003c/b\u003e) also used many parameters related to morphometric analyses in individual selected catchments and explained that flash floods are correlated with catchment morphometric parameters. Moreover, Costa (\u003cb\u003e1987\u003c/b\u003e) compared the hydraulic and morphometric properties of single catchments to understand the correlation of flash flood occurrence in the conterminous United States using flood inventory data and concluded that the largest flash floods result from the combination of optimum conditions in terms of catchment morphometry, climate, and hydraulic properties. These kinds of morphometric analyses only consider small and individual catchments and ignore spatial gradients of sub-catchments and inter-catchments (see definitions in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Interestingly, flood inventories (for example, FlooDOT, used in this study by Akbas et al. (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2025\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) reveal that all floods occur in specific areas within sub-catchments, especially inter-catchments, which represent the upstream areas. Ozdemir and Akbaş (\u003cb\u003e2023\u003c/b\u003e) addressed this issue in their study comparing flood modelling results with catchment morphometric results, but it has not yet been clarified. Hence, a new perspective and novel approach are required to overcome the limitations of the current representation of sub-catchments in morphometry. Therefore, we propose and utilize a novel cumulative catchment approach to define flood susceptibility at the regional level, as well as catchment-based high-resolution aspects (small catchments). Cumulative catchments are catchments that are divided into smaller sub-catchments according to a specified threshold value, but despite their small size, they reflect their upstream area. This developed catchment approach offers an innovative perspective that enables the spatial representation of flood susceptibility at high resolution (i.e., small catchments but represents the upstream) and the generation of a flood susceptibility map at a regional scale.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTherefore, this study presents a novel cumulative basin approach to represent the basins where floods occur and, for the first time, reveals basin-based flood susceptibility at a regional scale. Hence, the aim of this study is a) to present the novel cumulative catchment approach and apply flood susceptibility regionally to catchments via machine learning algorithms, such as XGBoost, random forest, logistic regression, and support vector machine, and b) to obtain regional flood susceptibility mapping over T\u0026uuml;rkiye for a pioneering study. Although different machine learning models were used in this study, the primary aim is to identify the model that best reflects the cumulative basin approach, rather than comparing these algorithms.\u003c/p\u003e"},{"header":"2. Study area and historical floods","content":"\u003cp\u003eT\u0026uuml;rkiye has distinct characteristics, in which topographic relief changes rapidly over short distances, creating a strong gradient between the coastal and inner parts (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e1\u003c/span\u003e). This situation is particularly pronounced in the Anatolian plateau and in the transitions between the mountainous areas that form its margins. Additionally, seasonal climatic conditions associated with large-scale atmospheric processes can have a significant impact on T\u0026uuml;rkiye. In winter, for instance, mid-latitude cyclones originating from Icelandic low-pressure systems descend into the Mediterranean area, bringing abundant precipitation to T\u0026uuml;rkiye because of the Polar jet (\u003cb\u003eTatlı et al., 2004\u003c/b\u003e). The most extreme precipitation occurs during this time, causing numerous disasters, including floods and landslides (Akbas, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Gorum and Fidan, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Y\u0026uuml;ksek et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Also, cyclones have an intrinsic connection with topography in which coastal areas facing cyclones receive more precipitation, while the continental inner areas correspond to rain-shadow zones because of the topographic barrier (Aydoğan et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Karaca et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2000\u003c/span\u003e). Unlike winter, during the summer season, the polar jet ascends to the north, and Monsoon low and Azores high pressures influence T\u0026uuml;rkiye, causing severe drought conditions (Tatli et al., \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e2004\u003c/span\u003e). However, even during summer, most of the Black Sea region is exposed to cyclones and experiences hazards due to extreme precipitation. On the other hand, the Central Anatolian plateau and Northeast Anatolia also experience similar hazard conditions due to convective precipitation (T\u0026uuml;rkeş, \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e1996\u003c/span\u003e). On the other hand, some researchers have stated that development and growth initiatives in T\u0026uuml;rkiye have had a significant impact on land use, resulting in a profound effect on hydrogeological conditions. Consequently, alongside climatic vulnerability, changes in land use within watersheds also creates a significant potential for generating disasters (Ko\u0026ccedil; et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Ozdemir and Elbaşı, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2015\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe study area in this paper covers 25 major basins defined by the General Directorate of State Hydraulic Works (DSI) in T\u0026uuml;rkiye, and specifically the Meri\u0026ccedil;-Ergene, Marmara, Susurluk, Kuzey Ege, Gediz, K\u0026uuml;\u0026ccedil;\u0026uuml;k Menderes, B\u0026uuml;y\u0026uuml;k Menderes, Batı Akdeniz (Western Mediterranean), Antalya, Burdur G\u0026ouml;ller, Akar\u0026ccedil;ay, Sakarya, Batı Karadeniz (Western Black Sea), Yeşilırmak, Kızılırmak, Konya, Doğu Akdeniz (Eastern Mediterranean), Seyhan, Asi, Ceyhan, Fırat-Dicle (Euphrates- Tigris), Doğu Karadeniz (Eastern Black Sea), \u0026Ccedil;oruh, Aras, Van G\u0026ouml;l\u0026uuml;, basins respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e1\u003c/span\u003e). These basins are used for operational activities of the DSI. Based on FlooDOT (Flood Inventory of T\u0026uuml;rkiye, Akbas et al. (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2025\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2026\u003c/span\u003e), Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e1\u003c/span\u003e depicts a map of recent Turkish floods and their impacts, including fatalities, injuries, and losses. As can be seen from the map, floods occur in every part of T\u0026uuml;rkiye, resulting in many casualties. In this study, the main basins belonging to DSI were used to compare flood susceptibility results obtained from the cumulative catchments approach in order to better understand flood-generating mechanisms.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"3. Materials and Methods","content":"\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Construction of cumulative catchments and rivers\u003c/h2\u003e \u003cp\u003eDigital Elevation Models (DEMs) are one of the most important databases used to obtain hydrographic data such as river networks and catchments (Lehner et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). However, it is not possible to use DEMs directly in hydrological analyses because these data, which are created using different techniques ranging from synthetic aperture radar (SAR) to UAVs or stereo imagery, have the characteristics of a digital surface model (DSM) rather than a DEM. One of the most significant disadvantages of using this type of data in hydrological analysis is that it contains numerous artificial structures that can impede the flow of water or direct it in the wrong direction. Apart from this, the presence of forested areas is another limitation in terms of hydrological and geomorphic aspects (Gorum, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Additionally, it may be challenging to obtain an accurate representation of the river\u0026rsquo;s network and channels. Therefore, DEMs to be used in hydraulic or hydrological studies must first be cleared of artificial structures (\u003cb\u003eOzdemir and Akbaş, 2025;\u003c/b\u003e Yamazaki et al., \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). For example, Yamazaki et al. (\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) applied various filtering techniques to SRTM (90m) data, utilizing databases associated with tree density and height, as well as reference ground elevation, to increase the accuracy of the data by approximately 40% and 60%. This type of filtering is especially important for flat areas and closed catchments where river channels are very difficult to obtain because the DEM may misrepresent the river network in these areas.\u003c/p\u003e \u003cp\u003eFor this reason, FABDEM (Forest And Buildings removed Copernicus 30m DEM), which is one of the most important DEM sources that removes the presence of obstacles like buildings and tree heights that prevent the establishment of a stream network for hydraulic and hydrological studies, was used in this study (Hawker et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). FABDEM uses a random forest algorithm through many covariates related to buildings, such as night lights, GHS Urban database, world settlement footprint, and forests, including land-cover, forest heights, etc., and provides a DEM for hydrological works. According to Ozdemir and Akbas (\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2023b\u003c/span\u003e), FABDEM provides results similar to those of LIDAR data in terms of both flood inundation depth and river network production in catchment studies. After deciding to use FABDEM, the Archydro package was employed to create the river network and catchments in the study (Djokic et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; \u003cb\u003eMaidment, 2022\u003c/b\u003e). This package is frequently preferred in hydrology thanks to its sub-packages that allow the correction of a conditioned digital elevation model for river networks and catchments, as well as the numerous computational flexibilities it provides. In this respect, the catchment construction and river network processes, as determined by FABDEM data, are illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Since there are endorheic basins in many parts of T\u0026uuml;rkiye, it is essential to burn DEM in order to reflect the river network properly. Here, EU-Hydro data obtained from satellites were used as the benchmark river network while the DEM was burned (\u003cb\u003eCLMS, 2025\u003c/b\u003e). However, as EU-Hydro includes a different network in some areas, it has been checked and corrected again using 1/25.000 topographic maps.\u003c/p\u003e \u003cp\u003eAlthough it is stated that FABDEM has been corrected based on the forest and settlement, filling operations have been performed to address potential gaps in some areas. Since the hydrologically conditioned DEM data were ready to generate a stream network, the first process, i.e., determining the flow direction, was carried out (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). Although D-Infinity (DINF) and Multiple Flow Direction (MFD) methods are among the algorithms used to determine the flow direction, the D8 method was preferred in this study because it is one of the most frequently used and stable algorithms in the literature (Jenson and Domingue, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e1988\u003c/span\u003e; O'Callaghan and Mark, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e1984\u003c/span\u003e; Tarboton, \u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e1997\u003c/span\u003e). This method determines the direction in which water will flow depending on gravity (slope) on a cell-by-cell basis and assigns eight directional information to the cells. After determining the direction in which the water will flow, the flow accumulation is started. Here, each cell is summed towards the outlet of the river. The following process is the construction of the river network and catchments. The main issue is to create catchments and rivers by selecting a threshold value for flow accumulation because this threshold value is very important as it will form a catchment according to the confluence of the rivers obtained after the river network is created (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003eb).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThere is no standard threshold for flow accumulation in the literature. However, there is one very important data for answering these questions: the FlooDOT inventory data. When the inventory is examined, it is obvious that even intermittent streams on topography maps have the potential to cause flooding. Therefore, a threshold value of 5000 was chosen for a detailed river network and watershed based on flow accumulation (the number of raster cells that drain into a given cell). According to this selected threshold value, the river construction process was carried out as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003ec based on pour points. Many large and small river catchments were divided, and inter-catchments together with catchments are defined based on streamlink and pour points (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003ed). However, in a catchment system, the area between the confluence and the main tributary is shown as a small catchment (inter-catchment). These are actually quite large basins in terms of the area they contain behind them, and the FlooDOT inventory shows that flooding usually occurs in these confluence areas. Geomorphometric studies, whether for tectonic purposes or for flooding or landslides, are traditionally carried out behind the confluence/junction points that connect to the main tributary. In other words, the inter-catchments in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003ed are always ignored. In this study, for the first time, a cumulative catchment approach was introduced without ignoring inter-catchments (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003ee). The cumulative catchments are catchments that are divided into smaller sub-catchments according to a specified threshold value, but despite their small size, they reflect their upstream area. However, one of the important issues here is that if this study is carried out by taking into account the topological characteristics of the sub-catchments themselves, the hierarchical structure of the river systems and statistical values of the sub-catchments will be incorrect, as shown in \u003cb\u003eFig.\u0026nbsp;2f\u003c/b\u003e\u003csub\u003e\u003cb\u003e1\u003c/b\u003e\u003c/sub\u003e. In this study, we introduce a cumulative catchment system to represent better geo-environmental predisposing factors of flood events in susceptibility models (Akbas et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; \u003cb\u003eFig.\u0026nbsp;2f\u003c/b\u003e\u003csub\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sub\u003e). Otherwise, classical catchment morphometry ignores the spatial differences of sub-catchments as shown in \u003cb\u003eFig.\u0026nbsp;2f\u003c/b\u003e\u003csub\u003e\u003cb\u003e3\u003c/b\u003e\u003c/sub\u003e. In order to achieve this task, the Archydro package was used to accumulate statistical properties of sub-catchments (Djokic et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; \u003cb\u003eMaidment, 2022\u003c/b\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Parameter selection for flood susceptibility modelling\u003c/h2\u003e \u003cp\u003eAfter construction of cumulative catchments, nine morphometric parameters (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), three climatological and hydrological parameters, and one parameter that contains both surface and vegetation parameters (Curve Number) were obtained to use in susceptibility analyses (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e3\u003c/span\u003e). For catchment morphometry, we determined parameters based on Ozdemir and Akbas (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe parameters used in catchment morphometry and flood susceptibility analysis.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eName of Parameters\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDefinition, Importance, and Data Sources\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEquations\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopographic Wetness Index (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:TWI\\)\u003c/span\u003e\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSpecific catchment area divided by slope gradient in radians (Beven and Kirkby, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e1979\u003c/span\u003e). This index quantifies the role of topography on hydrological conditions. FADEM is the data source for this and following morphometric parameters\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:TWI=ln\\left(\\frac{{A}_{s}}{tan\\:\\beta\\:}\\right)\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStrahler Orders (Sth)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMerging orders with the same code only increases order power.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en\u0026thinsp;+\u0026thinsp;n\u0026thinsp;=\u0026thinsp;n\u0026thinsp;+\u0026thinsp;1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eArea (A)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eShows the cumulative area of the catchments.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:A=\\:\\sum\\:_{i=1}^{n}A\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDrainage Density (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{D}_{d}\\)\u003c/span\u003e\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDrainage length in the catchment divided by the total basin area (Horton, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e1945\u003c/span\u003e).\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{D}_{d}=\\:\\sum\\:_{i=1}^{n}L/{A}_{b}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eForm Factor(\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{f}\\)\u003c/span\u003e\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCatchment area divided by the square of the catchment length (Horton, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e1945\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{f}=\\frac{A}{{{L}_{b}}^{2}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMelton ruggedness number (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{mR}_{n}\\)\u003c/span\u003e\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe relief of a catchment is divided by the square root of the catchment area (Melton, \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e1965\u003c/span\u003e).\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:H=\\:H/\\sqrt{{A}_{b}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(6)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTime of concentration (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{c}\\)\u003c/span\u003e\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRefers to the time between when the water starts to flow and when it collects in the outlet area (\u003cb\u003eGiandotti,1934, from\u003c/b\u003e Ravazzani et al., \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{c}=\\:\\frac{4\\sqrt{{A}_{b}}+1.5\\:L}{0.8\\sqrt{{H}_{mean-out}}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStream Power Index (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:SPI)\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe catchment area is divided by a slope gradient in radians (Florinsky, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; \u003cb\u003eMoore et al., 1991\u003c/b\u003e).\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:SPI=ln\\left(\\frac{{A}_{b}}{tan\\:\\beta\\:}\\right)\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(8)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHypsometric Integral (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{H}_{i}\\)\u003c/span\u003e\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIt refers to the area under the hypsometric curve (Pike and Wilson, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e1971\u003c/span\u003e).\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{H}_{i}=\\:\\frac{\\stackrel{̿}{H}-{H}_{min}}{{{H}_{mak}-H}_{min}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(9)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e100-year precipitation (P100\u0026nbsp;year)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDaily precipitation amount characterized by a 100-year return period computed over 30 years (1989\u0026ndash;2018) by the ERA5 model via the GEV model.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:G\\left(\\chi\\:\\right)=\\text{exp}\\left\\{-{\\left[1+\\xi\\:\\left(\\frac{\\chi\\:-\\mu\\:}{\\sigma\\:}\\right)\\right]}^{-1/\\xi\\:}\\right\\}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(10)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAridity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe aridity index is a practical indicator of the water balance of catchments. It is a ratio of precipitation and evaporation (Zomer et al., \u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:AI=\\:\\frac{P}{{E}_{T0}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(11)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSCS-Curve Number II\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCurve Number is a combination of both land-use/land-cover (therefore vegetation) and soil.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{CN}_{II}=\\:f\\left(land-use,\\:soil\\:group,\\:slope\\right)\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(12)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e100-year discharge (Q100\u0026nbsp;year)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDaily discharge amount characterized by a 100-year return period by the EFAS model via the Gumbel extreme value distribution using L-moments\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:F\\left(\\chi\\:\\right)=exp\\:\\left[-\\text{exp}\\left(-\\frac{\\chi\\:-\\mu\\:}{\\sigma\\:}\\right)\\right]\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(13)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003eHere,\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:{A}_{P}\\)\u003c/span\u003e\u003c/span\u003e=Catchment perimeter length (km), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{N}_{u1}\\)\u003c/span\u003e\u003c/span\u003e= Total number of 1st orders according to Strahler order, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:L\\)\u003c/span\u003e\u003c/span\u003e= Length of river drainage (km), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{A}_{b}\\)\u003c/span\u003e\u003c/span\u003e=Catchment Area (km\u003csup\u003e2\u003c/sup\u003e), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{L}_{b}\\)\u003c/span\u003e\u003c/span\u003e= Catchment drainage length, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:H\\)\u003c/span\u003e\u003c/span\u003e= Relief, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:{H}_{mean-out}\\)\u003c/span\u003e\u003c/span\u003e=Catchment average-outlet elevation, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{A}_{s}\\)\u003c/span\u003e\u003c/span\u003e= Specific area, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:tan\\:\\beta\\:\\)\u003c/span\u003e\u003c/span\u003e= slope (in radian), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\mu\\:\\)\u003c/span\u003e\u003c/span\u003e = location, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e=scale and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\xi\\:\\)\u003c/span\u003e\u003c/span\u003e=shape, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:P\\)\u003c/span\u003e\u003c/span\u003e= Precipitation, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{E}_{T0}\\)\u003c/span\u003e\u003c/span\u003e= Reference Evaporation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAlthough there are many approaches to calculate the Tc parameter, the method by Giandotti (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e1934\u003c/span\u003e) was used because a methodological study by Ravazzani et al. (\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) demonstrates that this method more reliably estimates the time of concentration than other methods. On the other hand, extreme precipitation data were obtained from ERA5-Land, which is sufficient to represent the original climate conditions for T\u0026uuml;rkiye (Akbas and Ozdemir, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), allowing for the calculation of the 100-year precipitation event. Furthermore, 100-year discharge data were directly obtained from EFAS data (\u003cb\u003eEFAS, 2024\u003c/b\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAnnual maximum series (AMAX) was decided for extreme precipitation definition (Akbas, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Bayliss and Jones, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1993\u003c/span\u003e; Jones et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), and 100-year precipitation was calculated via Generalized Extreme Value Distribution/GEV distribution (Coles et al, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2001\u003c/span\u003e). The maximum likelihood (MLE) method has been recommended as the best approach in hydrological studies (Coles et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Gilleland and Katz, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), although there are many methods for calculating location, shape, and scale parameters. On the other hand, studies conducted by Seckin et al. (\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) and Akbas (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) focused on the extreme characteristics of precipitation and runoff data for the entire T\u0026uuml;rkiye. In both studies, it was emphasized that the distribution that best fits the extreme flow and precipitation data is the GEV. Therefore, GEV was used in this study. Aridity was directly obtained from Zomer et al. (\u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Curve Number II values were generated by overlaying the 2018 Corine landuse and national-scale soil maps (Akbas et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). For more detailed information about the SCS Curve Number methodology, please check Cronshey (\u003cb\u003e1986\u003c/b\u003e). Nine morphometric parameters were calculated using catchment topological properties, except for the topographic wetness index and stream power index. SAGA GIS was employed to calculate these indices. On the other hand, TWI and SPI, CNII, Aridity, and Q100 years were aggregated into catchments using zonal statistics. Consequently, the database is prepared for susceptibility modelling, in which we employed several machine learning techniques.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Machine learning properties for flood susceptibility analyses\u003c/h2\u003e \u003cp\u003eThe purpose of using machine learning in this study is to generate flood susceptibility maps based on parameters derived from a cumulative catchment analysis. From this perspective, flood susceptibility maps were obtained using some of the most widely used and reportedly most successful models, as discussed by Waled and Sajjad (\u003cb\u003e2025\u003c/b\u003e). In order to obtain a more accurate representation of the flood susceptibility models, we have utilized various frequently used classification models, including Random Forest, Support Vector Machine, XGBoost, and Logistic Regression. The robustness of the performance of these models has been acknowledged, so far, with a wide perspective in earth-science applications, such as landslide (Loche et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), wildfire (Zhang et al., \u003cspan citationid=\"CR82\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), soil erosion (Vergari, \u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), and flood studies (Seleem et al., \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Therefore, we structured the model architecture to train, predict, validate, and evaluate the performance of these models on top of the scikit-learn library (Pedregosa et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) built in Python, which enables us to achieve these analyses through an end-to-end machine learning pipeline. It is essential to recognize that, as we are not directly addressing the performances of many models, the decision of the final model will be determined by expert interpretation in accordance with model performances.\u003c/p\u003e \u003cp\u003e \u003cb\u003eModel training for classification\u003c/b\u003e \u003c/p\u003e \u003cp\u003eWe have used the catchments delineated by our cumulative approach as a training dataset, which includes environmental covariate information. The binary classification is essential for susceptibility modelling. Therefore, we classified flooded catchments as 1 and non-flooded basins as 0 using FlooDOT inventories (Akbas et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2025\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2026\u003c/span\u003e). This resulted in 2300 flooded and 95,000 non-flooded catchments. We employed the \u0026ldquo;RandomUnderSampler\u0026rdquo; function (Lema\u0026icirc;tre et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), a method for addressing imbalanced datasets by randomly selecting a portion of data from the target classes. Subsequently, we applied the Hold-out method, partitioning the dataset into two mutually exclusive subsets: a training set and a test set. As a rule of thumb, we split our dataset into 70% training and 30% independent testing datasets. For the purpose of model optimization, the \u0026ldquo;Randomized Search with Cross-Validation\u0026rdquo; method is used within scikit-learn (Pedregosa et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). This optimization, referred to as hyperparameter tuning, involves multiple hyperparameters and seeks to identify the optimal combination using a cross-validation method, which we selected as five in our model training. Each model has been employed to classify the susceptibility classes of unseen catchments by model, with the likelihood of proximity to 1 expressed as a percentage.\u003c/p\u003e \u003cp\u003e \u003cb\u003eModel performance evaluation\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe assessment of various model performances has been achieved using common evaluation metrics, including the ROC-AUC curve, confusion matrix, precision, recall, and F1-score. When integrated, these measurements provide a comprehensive view of a model's performance, offering insights into its overall accuracy, error distribution, and the trade-offs between mitigating false positives and false negatives. Further, overfitting has been evaluated by considering notable disparities between evaluation matrices. This method necessitates a reciprocal evaluation of model performance datasets during both training and testing, wherein we observe no significant indications of overfitting (\u003cb\u003eTable \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e and Figure S2\u003c/b\u003e).\u003c/p\u003e \u003cp\u003e \u003cb\u003eThe importance assessment of covariates\u003c/b\u003e \u003c/p\u003e \u003cp\u003eRevealing the importance of model inputs to model performance is crucial as machine learning algorithms often encounter black-box issues that impede model interpretability. Therefore, we utilize the SHAP (SHapley Additive exPlanations) package (Lundberg and Lee, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), which enhances the explainability of model outputs by providing comprehensive insights through the quantification of input contributions. In addition to evaluating feature relevance, Seleem et al. (\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) indicated that SHAP also assesses whether a feature favours or adversely impacts the expected values.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Results and Discussion","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e4.1. A short glance at model results\u003c/h2\u003e \u003cp\u003eWe used Random Forest, Support Vector Machine, XGBoost, and Logistic Regression models in this study. The purpose of using these models is not to compare models, but to better explain the cumulative catchment model and flood reasons based on the best model. When compared to other algorithms, XGBoost and Random Forest produced the highest results in performance metrics (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e), demonstrating their relative performance across various accuracy criteria. However, due to its constant superior performance across all assessment criteria on the test set, we selected XGBoost as the final model for the interpretation of flood susceptibility in T\u0026uuml;rkiye. The model evaluation metrics revealed overall robust results, with an accuracy of 0.72, a macro-average F1-score of 0.72, and a ROC-AUC of 0.80, demonstrating the generalization capacity of the XGBoost model. Furthermore, a study by Waleed and Sajjad (2024) was conducted to compare 14 machine learning models. They have suggested the XGBoost model for flood susceptibility, which proved usable in our study.\u003c/p\u003e \u003cp\u003eOur final flood susceptibility map with XGBoost illustrates flood probability ranging from 0 to 1, presented through a continuous gradient scale, which preserves the full variability of model outputs rather than using the natural breaks (Jenks) approach (\u003cb\u003eJenks, 1967\u003c/b\u003e). Each probability score represents a distinct cumulative catchment that we defined to be greater than 97,908 across T\u0026uuml;rkiye. Even though we maintain our susceptibility values as they were originally determined, the results can be categorized into ten groups based on 0.1 intervals ranging from 0 to 1: very low (0\u0026ndash;0.1), low (0.1\u0026ndash;0.2), low-moderate (0.2\u0026ndash;0.3), moderate (0.3\u0026ndash;0.4), moderate-high (0.4\u0026ndash;0.5), high (0.5\u0026ndash;0.6), high-very high (0.6\u0026ndash;0.7), very high (0.7\u0026ndash;0.8), extreme (0.8\u0026ndash;0.9), and very extreme (0.9\u0026ndash;1.0). Overall results demonstrate that approximately 24.8% of 97,908 catchments fall into very low and low susceptibility (9.02% + 15.8%), 33.4% into low-moderate and moderate (17.2% + 16.2%), 24.2% into moderate-high and high (13.7% + 10.5%), 12.5% into high-very high and very high (7.33% + 5.14%), and 5.1% into extreme and very extreme flood susceptible areas (3.72% + 1.36%)\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e4.2. Flood susceptibility results based on cumulative catchments\u003c/h2\u003e \u003cp\u003eFlood susceptibility maps have been generated using selected machine learning methods in this study. However, one map stems from the XGBoost method has been presented here based on the best scores of the methods (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e4\u003c/span\u003e). For other maps, please check the supplementary file. According to Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e4\u003c/span\u003e, flood susceptibility has distinct spatial characteristics. One of them is the susceptibility gradient in catchments between the coastal and the inner part of T\u0026uuml;rkiye. The catchments, along with the coastal part of T\u0026uuml;rkiye, have the highest flood susceptibility, with values that almost reach 1. However, the coastal part of T\u0026uuml;rkiye has a distinct gradient as well. For instance, the East and West Black Sea have the highest values, while the middle of this area has the lowest part compared to all coastal parts of Black Sea. Fatal floods are abundant in this area compared to other parts of T\u0026uuml;rkiye (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Moreover, another spatial characteristic is plains, where flood susceptibility has the highest values compared to other geomorphic units. The flood susceptibility values highlight the importance of cumulative catchments, indicating that susceptibility to flooding is higher at outlets compared to the catchment origins. For instance, catchments located in close proximity to the outlets of the longest rivers\u0026mdash;Kızılırmak, Sakarya, and Yeşilırmak\u0026mdash;have resulted in higher susceptibility to flooding compared to the origin of catchments.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e4\u003c/span\u003e illustrates the histograms belonging to the overall flood susceptibility with respect to the catchment\u0026rsquo;s flood susceptibility. The mean flood susceptibility of T\u0026uuml;rkiye, indicated by the orange line, is approximately 0.38 (moderate category). Some basins, such as Asi (mean value: 0.52), Gediz (0.46), K\u0026uuml;\u0026ccedil;\u0026uuml;k Menderes (0.58), Kuzey Ege (0.55), Marmara (0.6), Meri\u0026ccedil;-Ergene (0.61), and Susurluk (0.47), have a negatively skewed distribution (high values of flood susceptibility) compared to overall susceptibility. These basins contain high flood susceptibility values in almost every part of the area. For example, the average flood susceptibility of the Meri\u0026ccedil;-Ergene is 0.61, which is significantly higher than the average of the whole country.\u003c/p\u003e \u003cp\u003eNevertheless, some basins, such as Akar\u0026ccedil;ay (0.34), Aras (0.31), Burdur (0.33), \u0026Ccedil;oruh (0.30), Konya Kapalı (0.26), Van G\u0026ouml;l\u0026uuml; (0.28), have a positively skewed distribution in which negative values (low values) are most dominant with respect to other catchments. The \u0026Ccedil;oruh major basin has higher relief (check the Melton map in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003e\u003cb\u003e).\u003c/b\u003e On the other hand, Doğu Karadeniz, which has the most fatal floods compared to other areas, has a bimodal distribution. Flood susceptibility is higher at the outlets of individual catchments than in their upper parts in the Doğu Karadeniz basin. This condition is also supported by Elbası (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), who executed a hydrodynamic model for the Doğu Karadeniz, and found that the flood hazard is higher at the outlet of the Doğu Karadeniz basin.\u003c/p\u003e \u003cp\u003eFurthermore, certain basins almost exactly represent the flood susceptibility distribution belonging to T\u0026uuml;rkiye. These area, Doğu Akdeniz (0.37), Sakarya (0.37), Seyhan (0.35), Yeşilırmak (0.38), Kızılırmak (0.35), and Fırat-Dicle (0.32). In these areas, very extreme flood susceptibility is observed, but these values are only prevalent in the river course, the immediate surroundings of the river, and its outlet. For example, the map in Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e4\u003c/span\u003e shows the Kızılırmak, which is the second-largest basin after the Fırat-Dicle, and susceptibility values in the histogram follow a similar distribution to that of the overall country. However, in the lower reaches of the river, in the sub-catchments located near the main outlet, flood sensitivity has been observed to reach extremely high values. This observation underscores the crucial importance of adopting a cumulative catchment-based approach in flood risk assessments, particularly for the Sakarya River, despite its large-scale basin characteristics. This situation shows that it is more effective to examine the effects of flooding by dividing the catchment area into smaller sections than to consider the entire catchment morphology.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eUnderstanding the flood-generating mechanism is substantial for flood susceptibility. The cumulative catchment-based approach yields information on both the properties of the catchments and may provide insight into how these properties influence flood-generating mechanisms using tools of machine learning techniques. Figure\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e5\u003c/span\u003e illustrates the bivariate map between susceptibility values and parameters, together with SHAP graphs to understand the impact of parameters on flood susceptibility. High values of time concentration (see SHAP), for instance, are essential for model construction and high susceptibility. Because the model has all high-high groups overlapping in the bivariate map, except for some lower time of concentration values in floodplains, where susceptibility is higher. In hydrology, it is the most important and widely used parameter for designing discharge in various methods, such as the SCS curve number and the rational method (\u003cb\u003eChow et al., 1988\u003c/b\u003e). These results demonstrate that higher values of time of concentration, as determined by cumulative catchments, contribute to flood generation through the accumulation of water in channels. Furthermore, 100-year runoff also reveals the importance of main river courses and the patterns where higher aridity values (i.e., positive water balance) are higher. Additionally, SHAP values highlight the high values of 100-year runoff in the model setup, demonstrating their role in the flood-generation mechanism in cumulative catchments, particularly in main river courses and areas with high precipitation values. Furthermore, higher values of the bivariate map between Strahler and flood susceptibility are only dominant in the main river courses, unlike 100-year runoff values.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eClimatic parameters, such as aridity and 100-year precipitation, are important in certain basins, including Doğu Karadeniz, Batı Karadeniz, Seyhan, Ceyhan, and Susurluk (see Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e4\u003c/span\u003e), with specific values influencing flood-generating mechanisms, as illustrated by SHAP values and bivariate maps. SHAP graph highlights the importance of a higher values aridity (positive water balance) index than the 100-year precipitation. Guo et al. (\u003cb\u003e2014\u003c/b\u003e) have expressed that the aridity index is first-order responsible for the shape of the flood frequency curve. Therefore, some areas, such as Doğu Karadeniz, have the highest flood probabilities despite catchment characteristics (such as time of characteristics, Melton ruggedness number) that are not suitable for floods.\u003c/p\u003e \u003cp\u003eThe Melton ruggedness number is one of the critical parameters that control or affect flood characteristics, which are defined as flood and debris floods, as well as debris flows, by catchment properties (Wilford et al., \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; Church and Jakob, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). SHAP values mark higher values of these parameters. When Melton ruggedness numbers increase, it indicates the debris flood characteristics. SHAP values demonstrate that, in the construction of the XGBoost model, debris characteristic catchments have a greater influence than flood characteristic catchments. Additionally, a bivariate map between Melton ruggedness numbers and flood susceptibility highlights the debris characteristics of catchments, where high flood susceptibility values correspond to catchments with debris floods. Stolle et al. (\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) expressed that the Melton ruggedness number can only explain debris susceptibility. The topographic wetness index and stream power index are the most commonly used and critical parameters in flood susceptibility assessments (\u003cb\u003ePourzangbar et al., 2025\u003c/b\u003e). High topographic wetness index values indicate that water remains in the flat areas of the catchment for an extended period, and the catchments are prone to flooding. SHAP values emphasize the lower values of the stream power index, while higher values of the topographic wetness index are more influential in model construction; however, lower values are also shown to be influential. A bivariate map between the topographic wetness index and flood susceptibility shows the importance of flat areas.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eDespite the catchments of the Central Anatolian plateau, such as Konya, Akar\u0026ccedil;ay, and Burdur, being completely flat and having high topographic wetness index values, it is observed that flood susceptibility values are low. Moreover, the stream power index has patterns almost similar to Melton's ruggedness numbers, in which debris characteristic catchments are important.\u003c/p\u003e \u003cp\u003eOn the other hand, area is an important parameter for flood analyses because it is an irreplaceable parameter of regional flood frequency analyses (Hosking and Wallis \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e1997\u003c/span\u003e; Kumar and Chatterjee, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). However, as can be seen from the data obtained from the SHAP graph, it has not been observed to have a significant importance in model construction. It is assumed that this is likely due to other morphometric parameters, including characteristics of the area. However, the bivariate map reveals the importance of the area on the main river courses. Moreover, SHAP values indicate that form factor has no influence on model construction, which is also observed in the bivariate map. Drainage density follows this pattern as well. SHAP values and the bivariate map show a similar, less impactful effect compared to the form factor.\u003c/p\u003e \u003cp\u003eThe SCS Curve number is an important and frequently used parameter in hydrology for predicting excess rainfall and, consequently, surface runoff (\u003cb\u003eChow et al., 1988; Cronshey, 1986\u003c/b\u003e). High Curve Number values indicate an empirical relationship showing that precipitation falling into the catchment area is likely to turn into surface runoff. In this study, higher values of curve number influence model construction, as illustrated in the SHAP figure. However, this relationship is not similar in every catchment of T\u0026uuml;rkiye. Some DSI main basins such as Doğu Karadeniz, Batı Karadeniz, Seyhan, Ceyhan, and the main river courses of B\u0026uuml;y\u0026uuml;k/K\u0026uuml;\u0026ccedil;\u0026uuml;k Menderes support this idea. High Curve number values in other large DSI major basins have not affected flood sensitivity.\u003c/p\u003e \u003cp\u003eThe overall results illustrate that catchments with high flood susceptibility in T\u0026uuml;rkiye are clearly defined in areas where all parameters (climate, morphometry, and hydrology) are optimized, and the cumulative catchment makes these even more pronounced.\u003c/p\u003e \u003c/div\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003eThis study presents a novel approach to assessing catchment morphometry in relation to flood susceptibility. Classical morphometric approaches, which belong to catchment morphometry, ignore the sub-catchments, and this does not represent proper catchment morphometry, as it cannot explain flood-generating mechanisms. However, the flood inventories indicate that floods are concentrated in inter-catchments. Therefore, we propose a novel approach, termed cumulative catchments, which are derived from small sub-catchments and capture both the effects of upstream catchments and the contributing area behind them. This approach not only accounts for the influence of inter-catchments on catchment morphometry but also allows for a more detailed spatial representation of flood susceptibility (i.e., high resolution in terms of sub-catchment size) based on morphometric characteristics.\u003c/p\u003e \u003cp\u003eUsing hydrography data from FABDEM and flood inventory belonging to T\u0026uuml;rkiye, we explained the flood susceptibility via a cumulative catchment approach based on morphometric indices and climate and hydrology through the XGBoost machine learning method. Results showed that the cumulative catchment approach does not ignore the inter-catchments and provides a more detailed and reliable susceptibility distribution map. In this perspective, for the first time, a regional flood susceptibility result for T\u0026uuml;rkiye is obtained, and it is ready for flood management and hazard management. For instance, in T\u0026uuml;rkiye, it was determined that the Meri\u0026ccedil; within the DSI basins has the highest probability of flooding, and the spatial distribution of high flood sensitivity values was also identified. However, this approach can be applied to any region or continent worldwide.\u003c/p\u003e \u003cp\u003eOn the other hand, the flood-generating mechanism and the importance of these parameters, which are related to catchment morphometry, climate, and hydrology, were identified. It has been concluded that higher flood sensitivity values correspond to optimal conditions for catchment morphometry and the hydro-climatological characteristics of catchments. Additionally, we recognized that although SHAP values indicate the relative importance of parameters when constructing a model, bivariate maps have demonstrated that this importance varies spatially.\u003c/p\u003e \u003cp\u003eThis study presents a novel approach to catchment morphometry, despite utilizing widely employed machine learning methods. Therefore, it is also possible to compare machine learning methods with various network-based and algorithm-based machine learning models in watershed morphometry and flood susceptibility, and to conduct a separate study on the most meaningful results in flood susceptibility. On the other hand, considering the climatic conditions and changes in land use in the catchment areas, adopting a dynamic-based cumulative catchment approach, such as climate-change conditioned flood susceptibility, is feasible. The Cumulative catchments approach could be beneficial in determining the flood potential of basins in data-scarce regions where hydrodynamic modelling data is unavailable. Furthermore, cumulative basin outcomes may be helpful to policymakers in developing flood-related policies.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eCompeting Interests:\u0026nbsp;\u003c/strong\u003eThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability:\u0026nbsp;\u003c/strong\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval:\u0026nbsp;\u003c/strong\u003e Not applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgment:\u0026nbsp;\u003c/strong\u003eThis study was funded by TUBITAK-Scientific and Technological Research Council of T\u0026uuml;rkiye with 3501 Career Development Program (CAREER) (Project No: 121Y578).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability Statement:\u0026nbsp;\u003c/strong\u003eThe data that support the findings of this study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions Conceptualization\u003c/strong\u003e: Abbreviations: [AA: Abdullah AKBAS, AAVCI: Aydogan AVCIOGLU, HO: Hasan OZDEMİR, TG: Tolga GORUM, PB: Paul BATES], Methodology: [AA, AAVCI, HO, TG, PB]; Software and Validation: [AA, AAVCI]; Formal analysis and investigation: [AA]; Writing - original draft preparation: [AA, AAVCI]; Writing - review and editing: [AA, AAVCI, HO, TG, PB]; Supervision: [AA, AAVCI, HO, TG, PB]. All authors read and approved the final manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAddor, N., Nearing, G., Prieto, C., Newman, A. J., Le Vine, N., \u0026amp; Clark, M. P. (2018). A ranking of hydrological signatures based on their predictability in space. Water Resources Research, 54(11), 8792-8812.\u003c/li\u003e\n\u003cli\u003eAlfieri, L., Burek, P., Feyen, L., \u0026amp; Forzieri, G. (2015). Global warming increases the frequency of river floods in Europe. Hydrology and Earth System Sciences, 19(5), 2247-2260.\u003c/li\u003e\n\u003cli\u003eAkbas, A. (2023). Seasonality, persistency, regionalization, and control mechanism of extreme rainfall over complex terrain. Theoretical and Applied Climatology, 152(3-4), 981-997.\u003c/li\u003e\n\u003cli\u003eAkbas, A., Gorum, T., Ozdemir, H., Doğan, E., \u0026amp; \u0026Ccedil;orapcı, F. (2023). Hierarchical basins: A new approach to quantify geomorphometric flood susceptibility in the regional aspect. In Proceedings of the Regional Conference of Geomorphology Cappadocia 2023 (pp. 51). Nevşehir, T\u0026uuml;rkiye.\u003c/li\u003e\n\u003cli\u003eAkbaş, A., G\u0026ouml;r\u0026uuml;m, T., \u0026amp; \u0026Ouml;zdemir, H. (2024). T\u0026uuml;rkiye\u0026rsquo;de Taşkınları Oluşturan Mekanizmaların ve Taşkın Tehlikenin Alansal ve Zamansal Dinamiklerinin Ortaya \u0026Ccedil;ıkarılması. 121Y578 Nolu T\u0026uuml;bitak Projesi Sonu\u0026ccedil; Raporu.\u003c/li\u003e\n\u003cli\u003eAkbas, A., \u0026amp; Ozdemir, H. (2024). Comparing Satellite, Reanalysis, Fused and Gridded (In Situ) Precipitation Products Over T\u0026uuml;rkiye. International Journal of Climatology, 44(16), 5873-5889.\u003c/li\u003e\n\u003cli\u003eAkbas, A., G\u0026ouml;r\u0026uuml;m, T., \u0026amp; \u0026Ouml;zdemir, H., (2025). FlooDOT-based Analysis of Floods in Turkey and Spatial Comparison of Various Data Sources. TURKISH JOURNAL OF CIVIL ENGINEERING, 36(6).\u003c/li\u003e\n\u003cli\u003eAkbas, A., Gorum, T., and Ozdemir, H. (2026). FlooDOT (FlooD inventory Of T\u0026uuml;rkiye): A comprehensive flood inventory and its spatio-temporal analyses, J. Flood Risk Manag., under review.\u003c/li\u003e\n\u003cli\u003eAndreadis, K. M., Wing, O. E., Colven, E., Gleason, C. J., Bates, P. D., \u0026amp; Brown, C. M. (2022). Urbanizing the floodplain: global changes of imperviousness in flood-prone areas. Environmental Research Letters, 17(10), 104024.\u003c/li\u003e\n\u003cli\u003eAydoğan, D., Kankal, M., \u0026amp; \u0026Ouml;nsoy, H. (2016). Regional flood frequency analysis for \u0026Ccedil;oruh Basin of Turkey with L‐moments approach. Journal of Flood Risk Management, 9(1), 69-86.\u003c/li\u003e\n\u003cli\u003eBayliss, A. C., \u0026amp; Jones, R. C. (1993). Peaks-over-threshold flood database: Summary statistics and seasonality (IH Report No. 121). Institute of Hydrology, Wallingford, UK. Retrieved from http://nora.nerc.ac.uk/id/eprint/6075/1/IH_121.pdf\u003c/li\u003e\n\u003cli\u003eBeven, K. J., \u0026amp; Kirkby, M. J. (1979). A physically based, variable contributing area model of basin hydrology/Un mod\u0026egrave;le \u0026agrave; base physique de zone d\u0026apos;appel variable de l\u0026apos;hydrologie du bassin versant. Hydrological Sciences Journal, 24(1), 43-69.\u003c/li\u003e\n\u003cli\u003eBhat, M. S., Alam, A., Ahmad, S., Farooq, H., \u0026amp; Ahmad, B. (2019). Flood hazard assessment of upper Jhelum basin using morphometric parameters. Environmental Earth Sciences, 78(2), 54.\u003c/li\u003e\n\u003cli\u003eBl\u0026ouml;schl, G., Hall, J., Viglione, A., Perdig\u0026atilde;o, R. A., Parajka, J., Merz, B., ... \u0026amp; Živković, N. (2019). Changing climate both increases and decreases European river floods. Nature, 573(7772), 108-111.\u003c/li\u003e\n\u003cli\u003eChissende, A., Mehmood, M. S., Huo, A., Lou, P., \u0026amp; Wang, Z. (2025). Riverine flood susceptibility assessment using drainage basin morphometry in the Bahe River. International Journal of Disaster Risk Reduction, 117, 105181.\u003c/li\u003e\n\u003cli\u003eChorley, R.J. and Kennedy, B.A., 1971. Physical geography: a systems approach. London: Prentice-Hall. . \u003c/li\u003e\n\u003cli\u003eChorley, R.J. (2021). The drainage basin as the fundamental geomorphic unit. In Water, earth, and man: A synthesis of hydrology, geomorphology, and socio-economic geography. Routledge. https://doi.org/10.4324/9781003170181\u003c/li\u003e\n\u003cli\u003eChow, V. T., D. R. Maidment, and L. W. Mays. 1988. Applied hydrology. Singapore: McGraw-Hill.\u003c/li\u003e\n\u003cli\u003eChurch, M., \u0026amp; Jakob, M. (2020). What is a debris flood?. Water resources research, 56(8), e2020WR027144.\u003c/li\u003e\n\u003cli\u003eCLMS-Copernicus Land Monitoring Service, 2025, EU-Hydro, Database [Available online: https://doi.org/10.2909/393359a7-7ebd-4a52-80ac-1a18d5f3db9c]\u003c/li\u003e\n\u003cli\u003eCronshey, R. (1986). Urban hydrology for small watersheds (No. 55). US Department of Agriculture, Soil Conservation Service, Engineering Division.\u003c/li\u003e\n\u003cli\u003eCRED \u0026amp; UNDRR. (2020). The Human Cost of Disasters: An Overview of the Last 20 Years (2000-2019). UN Office for Disaster Risk Reduction. Access link: https://www.undrr.org/publication/human-cost-disasters-overview-last-20-years-2000-2019\u003c/li\u003e\n\u003cli\u003eColes, S., Bawa, J., Trenner, L., \u0026amp; Dorazio, P. (2001). An introduction to statistical modeling of extreme values (Vol. 208, p. 208). London: Springer.\u003c/li\u003e\n\u003cli\u003eDjokic, D., Ye, Z., \u0026amp; Dartiguenave, C. (2011). Arc hydro tools overview. Redland, Canada, ESRI, 5.\u003c/li\u003e\n\u003cli\u003eEFAS (2023), European Flood Awareness System (EFAS) Data, DOI: 10.24381/cds.e3458969\u003c/li\u003e\n\u003cli\u003eElbası, A. (2022). Producing flood hazard maps using regional flood frequency [PhD thesis], Istanbul University (In Turkish)\u003c/li\u003e\n\u003cli\u003eFlorinsky, I.V. (2012) Digital Terrain Analysis in Soil Science and Geology. Academic Press is an imprint of Elsevier. The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK, 379p.\u003c/li\u003e\n\u003cli\u003eGao, H., Fenicia, F., \u0026amp; Savenije, H. H. (2023). HESS Opinions: Are soils overrated in hydrology?. Hydrology and Earth System Sciences, 27(14), 2607-2620.\u003c/li\u003e\n\u003cli\u003eGhasemlounia, R., \u0026amp; Utlu, M. (2021). Flood prioritization of basins based on geomorphometric properties using principal component analysis, morphometric analysis and Redvan\u0026rsquo;s priority methods: A case study of Harşit River basin. Journal of Hydrology, 603, 127061.\u003c/li\u003e\n\u003cli\u003eGiandotti, M. 1934. Previsione delle piene e delle magre dei corsi d\u0026rsquo;acqua, 107. Rome: Memorie e studi idrografici, Servizio Idrografico Italiano\u003c/li\u003e\n\u003cli\u003eGilleland E, Katz RW (2016) extRemes 2.0: an extreme value analysis package in R. J Stat Softw 72(8):1\u0026ndash;39\u003c/li\u003e\n\u003cli\u003eGorum, T. (2019). Landslide recognition and mapping in a mixed forest environment from airborne LiDAR data. Engineering Geology, 258, 105155.\u003c/li\u003e\n\u003cli\u003eGorum, T., \u0026amp; Fidan, S. (2021). Spatiotemporal variations of fatal landslides in Turkey. Landslides, 18(5), 1691-1705.\u003c/li\u003e\n\u003cli\u003eHawker, L., Uhe, P., Paulo, L., Sosa, J., Savage, J., Sampson, C., \u0026amp; Neal, J. (2022). A 30 m global map of elevation with forests and buildings removed. Environmental Research Letters, 17(2), 024016\u003c/li\u003e\n\u003cli\u003eHorton, R. E. (1932). Drainage-basin characteristics. Eos, Transactions American Geophysical Union, 13(1), 350\u0026ndash;361. https://doi.org/10.1029/tr013i001p00350\u003c/li\u003e\n\u003cli\u003eHorton, R. E. (1945). Erosional development of streams and their drainage basins; hydrophysical approach to quantitative morphology. Geological society of America bulletin, 56(3), 275-370. https://doi.org/10.1130/0016-7606(1945)56[275:EDOSAT]2.0.CO;2\u003c/li\u003e\n\u003cli\u003eHosking JRM, Wallis JR (1997) Regional frequency analysis: an approach based on L-moments. Cambridge University Press, Cambridge, p 240\u003c/li\u003e\n\u003cli\u003eIslam, A., \u0026amp; Deb Barman, S. (2020). Drainage basin morphometry and evaluating its role on flood-inducing capacity of tributary basins of Mayurakshi River, India. SN Applied Sciences, 2(6), 1087.\u003c/li\u003e\n\u003cli\u003eJenson, S., \u0026amp; Domingue, J. (1988). Extracting topographic structure from digital elevation data for geographic ınformation system analysis. Photogrammetric Engineering And Remote Sensing, 54(11), 1593\u0026ndash;1600\u003c/li\u003e\n\u003cli\u003eJones, M. R., Blenkinsop, S., Fowler, H. J., \u0026amp; Kilsby, C. G. (2014). Objective classification of extreme rainfall regions for the UK and updated estimates of trends in regional extreme rainfall. International Journal of Climatology, 34(3), 751-765.\u003c/li\u003e\n\u003cli\u003eO\u0026apos;Callaghan, J. F., \u0026amp; Mark, D. M. (1984). The extraction of drainage networks from digital elevation data. Computer vision, graphics, and image processing, 28(3), 323-344.\u003c/li\u003e\n\u003cli\u003eKaraca, M., Deniz, A., \u0026amp; Tayan\u0026ccedil;, M. (2000). Cyclone track variability over Turkey in association with regional climate. International Journal of Climatology: A Journal of the Royal Meteorological Society, 20(10), 1225-1236.\u003c/li\u003e\n\u003cli\u003eKhodaei, H., Nasiri Saleh, F., Nobakht Dalir, A., \u0026amp; Zarei, E. (2025). Future flood susceptibility mapping under climate and land use change. Scientific Reports, 15(1), 12394.\u003c/li\u003e\n\u003cli\u003eKo\u0026ccedil;, G., Petrow, T., \u0026amp; Thieken, A. H. (2020). Analysis of the most severe flood events in Turkey (1960\u0026ndash;2014): which triggering mechanisms and aggravating pathways can be identified? Water, 12(6), 1562.\u003c/li\u003e\n\u003cli\u003eKreibich, H., Van Loon, A. F., Schr\u0026ouml;ter, K., Ward, P. J., Mazzoleni, M., Sairam, N., ... \u0026amp; Di Baldassarre, G. (2022). The challenge of unprecedented floods and droughts in risk management. Nature, 608(7921), 80-86.\u003c/li\u003e\n\u003cli\u003eKumar, R., \u0026amp; Chatterjee, C. (2005). Regional flood frequency analysis using L-moments for North Brahmaputra region of India. Journal of Hydrologic Engineering, 10(1), 1-7.\u003c/li\u003e\n\u003cli\u003eLema\u0026icirc;tre, G., Nogueira, F., \u0026amp; Aridas, C. K. (2017). Imbalanced-learn: A Python toolbox to tackle the curse of imbalanced datasets in machine learning. Journal of Machine Learning Research, 18(17), 1\u0026ndash;5.\u003c/li\u003e\n\u003cli\u003eLehner, B., Verdin, K., \u0026amp; Jarvis, A. (2008). New global hydrography derived from spaceborne elevation data. Eos, Transactions American Geophysical Union, 89(10), 93-94.\u003c/li\u003e\n\u003cli\u003eLoche, M., Alvioli, M., Marchesini, I., Bakka, H., \u0026amp; Lombardo, L. (2022). Landslide susceptibility maps of Italy: Lesson learnt from dealing with multiple landslide types and the uneven spatial distribution of the national inventory. Earth-Science Reviews, 232, 104125.\u003c/li\u003e\n\u003cli\u003eLundberg SM, Lee SI. 2017. A unified approach to interpreting model predictions. In: Proceedings of the 31st international conference on neural information processing systems, p. 4768\u0026ndash;4777. \u003c/li\u003e\n\u003cli\u003eMaidment, D. R., \u0026amp; Morehouse, S. (2002). Arc Hydro: GIS for water resources. ESRI. Inc., Redlands, CA.\u003c/li\u003e\n\u003cli\u003eMagilligan, F. J. (1992). Thresholds and the spatial variability of flood power during extreme floods. Geomorphology, 5(3\u0026ndash;5), 373\u0026ndash;390. https://doi.org/10.1016/0169-555x(92)90014-f\u003c/li\u003e\n\u003cli\u003eMelton, M. A. (1965). The geomorphic and paleoclimatic significance of alluvial deposits in southern Arizona. The Journal of geology, 73(1), 1-38. https://www.jstor.org/stable/30066379\u003c/li\u003e\n\u003cli\u003eMoore, I. D., \u0026amp; Grayson, R. B. (1991). Terrain‐based catchment partitioning and runoff prediction using vector elevation data. Water Resources Research, 27(6), 1177-1191.\u003c/li\u003e\n\u003cli\u003eMorel, M., Piton, G., Kuss, D., Evin, G., \u0026amp; Le Bouteiller, C. (2023). Statistical modeling of sediment supply in torrent catchments of the northern French Alps. Natural Hazards and Earth System Sciences, 23(5), 1769-1787.\u003c/li\u003e\n\u003cli\u003ePedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., ... \u0026amp; Duchesnay, \u0026Eacute;. (2011). Scikit-learn: Machine learning in Python. the Journal of machine Learning research, 12, 2825-2830.\u003c/li\u003e\n\u003cli\u003ePike, R. J., \u0026amp; Wilson, S. E. (1971). Elevation-relief ratio, hypsometric integral, and geomorphic area-altitude analysis. Geological Society of America Bulletin, 82(4), 1079-1084. https://doi.org/10.1130/0016-7606(1971)82[1079:ERHIAG]2.0.CO;2\u003c/li\u003e\n\u003cli\u003eRogelis, M. C., \u0026amp; Werner, M. (2014). Regional debris flow susceptibility analysis in mountainous peri-urban areas through morphometric and land cover indicators. Natural Hazards and Earth System Sciences, 14(11), 3043-3064.\u003c/li\u003e\n\u003cli\u003eSeleem, O., Ayzel, G., de Souza, A. C. T., Bronstert, A., \u0026amp; Heistermann, M. (2022). Towards urban flood susceptibility mapping using data-driven models in Berlin, Germany. Geomatics, Natural Hazards and Risk, 13(1), 1640-1662. \u003c/li\u003e\n\u003cli\u003eSeckin, N., Haktanir, T., \u0026amp; Yurtal, R. (2011). Flood frequency analysis of Turkey using L‐moments method. Hydrological Processes, 25(22), 3499-3505.\u003c/li\u003e\n\u003cli\u003eStolle, A., Langer, M., Bl\u0026ouml;the, J. H., \u0026amp; Korup, O. (2015). On predicting debris flows in arid mountain belts. Global and Planetary Change, 126, 1-13.\u003c/li\u003e\n\u003cli\u003eStrahler, A. N. (1980). Systems theory in physical geography. Physical Geography, 1(1), 1-27.\u003c/li\u003e\n\u003cli\u003eObeidat, M., Awawdeh, M., \u0026amp; Al‐Hantouli, F. (2021). Morphometric analysis and prioritisation of watersheds for flood risk management in Wadi Easal Basin (WEB), Jordan, using geospatial technologies. Journal of Flood Risk Management, 14(2), e12711.\u003c/li\u003e\n\u003cli\u003eO\u0026apos;Callaghan, J. F., \u0026amp; Mark, D. M. (1984). The extraction of drainage networks from digital elevation data. Computer vision, graphics, and image processing, 28(3), 323-344.\u003c/li\u003e\n\u003cli\u003eOzdemir, H., \u0026amp; Bird, D. (2009). Evaluation of morphometric parameters of drainage networks derived from topographic maps and DEM in point of floods. Environmental geology, 56(7), 1405-1415.\u003c/li\u003e\n\u003cli\u003eOzdemir, H., \u0026amp; Elbaşı, E. (2015). Benchmarking land use change impacts on direct runoff in ungauged urban watersheds. Physics and Chemistry of the Earth, Parts A/B/C, 79, 100-107.\u003c/li\u003e\n\u003cli\u003eOzdemir, H., \u0026amp; Akbas, A. (2023a). Is there a consistency in basin morphometry and hydrodynamic modelling results in terms of the flood generation potential of basins? A case study from the Ulus River Basin (T\u0026uuml;rkiye). Journal of Hydrology, 625, 129926.\u003c/li\u003e\n\u003cli\u003eOzdemir, H. \u0026amp; Akbas, A. (2023b). The Effects of Spatial Resolution Variability of Digital Elevation Models on Flood Hazard Analysis. JOURNAL OF GEOGRAPHY, (46), 137-156. https://doi.org/10.26650/JGEOG2023-1177718\u003c/li\u003e\n\u003cli\u003eOzdemir, H., \u0026amp; Akbas, A. (2025). Evaluating the capability of UAV-Derived DEMs for flood hazard analysis. Physical Geography, 46(3), 175-195.\u003c/li\u003e\n\u003cli\u003ePatton, P. C., \u0026amp; Baker, V. R. (1976). Morphometry and floods in small drainage basins subject to diverse hydrogeomorphic controls. Water Resources Research, 12(5), 941\u0026ndash;952. https://doi.org/10.1029/wr012i005p00941\u003c/li\u003e\n\u003cli\u003eRavazzani, G., Boscarello, L., Cislaghi, A., \u0026amp; Mancini, M. (2019). Review of time-of-concentration equations and a new proposal in Italy. Journal of Hydrologic Engineering, 24(10), 04019039.\u003c/li\u003e\n\u003cli\u003eRentschler, J., Avner, P., Marconcini, M., Su, R., Strano, E., Vousdoukas, M., \u0026amp; Hallegatte, S. (2023). Global evidence of rapid urban growth in flood zones since 1985. Nature, 622(7981), 87-92.\u003c/li\u003e\n\u003cli\u003eYamazaki, D., Ikeshima, D., Tawatari, R., Yamaguchi, T., O\u0026apos;Loughlin, F., Neal, J. C., ... \u0026amp; Bates, P. D. (2017). A high‐accuracy map of global terrain elevations. Geophysical Research Letters, 44(11), 5844-5853.\u003c/li\u003e\n\u003cli\u003eY\u0026uuml;ksek, \u0026Ouml;., Kankal, M., \u0026amp; \u0026Uuml;\u0026ccedil;\u0026uuml;nc\u0026uuml;, O. (2013). Assessment of big floods in the Eastern Black Sea Basin of Turkey. Environmental monitoring and assessment, 185(1), 797-814.\u003c/li\u003e\n\u003cli\u003eWilford, D. J., Sakals, M. E., Innes, J. L., Sidle, R. C., \u0026amp; Bergerud, W. A. (2004). Recognition of debris flow, debris flood and flood hazard through watershed morphometrics. Landslides, 1(1), 61-66.\u003c/li\u003e\n\u003cli\u003eTarboton, D. G. 1997. \u0026quot;A new method for the determination of flow directions and upslope areas in grid digital elevation models.\u0026quot; Water Resources Research 33(2): 309-319.\u003c/li\u003e\n\u003cli\u003eTatli, H., N\u0026uuml;zhet Dalfes, H., \u0026amp; Sibel Menteş, Ş. (2004). A statistical downscaling method for monthly total precipitation over Turkey. International Journal of Climatology: A Journal of the Royal Meteorological Society, 24(2), 161-180.\u003c/li\u003e\n\u003cli\u003eTola, S. Y., \u0026amp; Shetty, A. (2022). Flood susceptibility modeling based on morphometric parameters in Upper Awash River basin, Ethiopia using geospatial techniques. Sustainable Water Resources Management, 8(2), 49.\u003c/li\u003e\n\u003cli\u003eT\u0026uuml;rkeş, M. (1996). Spatial and temporal analysis of annual rainfall variations in Turkey. International Journal of Climatology: A Journal of the Royal Meteorological Society, 16(9), 1057-1076.\u003c/li\u003e\n\u003cli\u003eWaleed, M., \u0026amp; Sajjad, M. (2025). Advancing flood susceptibility prediction: A comparative assessment and scalability analysis of machine learning algorithms via artificial intelligence in high-risk regions of Pakistan. Journal of Flood Risk Management, 18(1), e13047. https://doi.org/10.1111/jfr3.13047\u003c/li\u003e\n\u003cli\u003eVergari, F. (2015). Assessing soil erosion hazard in a key badland area of Central Italy. Natural Hazards, 79(Suppl 1), 71-95.\u003c/li\u003e\n\u003cli\u003eV\u0026iacute;g, B., F\u0026aacute;bian, S. \u0026Aacute;., Czig\u0026aacute;ny, S., Pirkhoffer, E., Halmai, \u0026Aacute;., Kov\u0026aacute;cs, I. P., ... \u0026amp; L\u0026oacute;czy, D. (2022). Morphometric analysis of low mountains for mapping flash flood susceptibility in headwaters. Natural Hazards, 114(3), 3235-3254.\u003c/li\u003e\n\u003cli\u003eZhang, G., Wang, M., \u0026amp; Liu, K. (2019). Forest fire susceptibility modeling using a convolutional neural network for Yunnan province of China. International Journal of Disaster Risk Science, 10(3), 386-403.\u003c/li\u003e\n\u003cli\u003eZomer, R. J., Xu, J., \u0026amp; Trabucco, A. (2022). Version 3 of the global aridity index and potential evapotranspiration database. Scientific Data, 9(1), 409.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":true,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"natural-hazards","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"nhaz","sideBox":"Learn more about [Natural Hazards](https://www.springer.com/journal/11069)","snPcode":"11069","submissionUrl":"https://submission.nature.com/new-submission/11069/3","title":"Natural Hazards","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Cumulative catchments, Flood susceptibility, Catchment morphometry, Türkiye","lastPublishedDoi":"10.21203/rs.3.rs-8379220/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8379220/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eCatchments are a fundamental unit of hydrological and geomorphological systems. Although floods occur in rivers, they also result from processes within catchments. For this reason, catchment morphometry and parameters such as climatic, soil, and hydrological are widely used to define flood susceptibility in regions where data is scarce. However, floods mostly occur at inter-catchments. An inter-catchment is a transitional area between adjacent drainage basins where water flow is not confined to a single channel and may contribute to multiple basins. However, morphometric calculations often neglect spatial gradients between sub-catchments and inter-catchments in assessing flood susceptibility. In this study, we therefore developed the cumulative catchment approach, which delineates small sub-catchments that reflect the upstream drainage contribution while not neglecting the spatial gradients within sub-catchments. Based on this approach, many morphometric parameters have been calculated. Additionally, parameters related to climatic and hydrological conditions, as well as land-use/soil types, have been assigned to the catchments using zonal statistics. Although the aim of the study was not to compare machine learning methods, the models that performed best in flood prediction were compared, and flood susceptibility was obtained on a cumulative catchment basis using machine learning. The algorithms such as XGBoost, random forest, logistic regression, and support vector machine were used, and XGBoost was determined to be the best model for defining flood susceptibility based on the ROC curve using the inventory belong to FlooDOT (FlooD invetory Of T\u0026uuml;rkiye) dataset. Furthermore, a bivariate map was constructed between model parameters and susceptibility values to understand the impact of covariates. To the best of our knowledge, this study provides the pioneer flood susceptibility analysis regionally for T\u0026uuml;rkiye, based on cumulative catchments. Model results present flood probabilities in a more realistic and high-resolution manner, based on the cumulative characteristics of the catchments, in the form of sub-catchments (small catchments). This approach also offers an opportunity in terms of regional applicability, particularly in data-scarce areas.\u003c/p\u003e","manuscriptTitle":"Cumulative catchments: A novel approach for catchment-based regional flood susceptibility","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-09 06:39:50","doi":"10.21203/rs.3.rs-8379220/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"","date":"2026-01-21T13:58:35+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-21T13:43:19+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"Natural Hazards","date":"2026-01-13T17:02:25+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-12-17T02:14:34+00:00","index":"","fulltext":""},{"type":"submitted","content":"Natural Hazards","date":"2025-12-16T14:00:11+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"natural-hazards","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"nhaz","sideBox":"Learn more about [Natural Hazards](https://www.springer.com/journal/11069)","snPcode":"11069","submissionUrl":"https://submission.nature.com/new-submission/11069/3","title":"Natural Hazards","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"8ff8eeb0-c76c-4285-950d-af4037734b9b","owner":[],"postedDate":"February 9th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-02-09T06:39:50+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-09 06:39:50","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8379220","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8379220","identity":"rs-8379220","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}