Spatial Assessment of Flood Susceptibility in the Abai Region, Kazakhstan

Abstract

Floods are among the most frequent and destructive natural hazards in Kazakhstan, particularly in the Abai Region, Kazakhstan, where topographic, hydrological, and climatic factors strongly influence flood occurrence. This study presents a comprehensive spatial assessment of flood susceptibility in the Abai Region using a multi-criteria Geographic Information System (GIS) approach. The analysis integrates twelve flood-conditioning factors representing hydrological, topographic, environmental, and anthropogenic variables. The relative importance of these factors was determined using the Analytical Hierarchy Process (AHP). The results indicate that distance to rivers (20%) and precipitation (16%) are the most influential drivers of flood susceptibility, followed by Height Above Nearest Drainage (HAND) (11%) and drainage density (9%). The resulting flood susceptibility map classifies the study area into five susceptibility levels. Approximately 56.6% of the study area falls within the moderate susceptibility class, while 25.0% is categorized as high susceptibility, mainly concentrated in low-lying floodplains and foothill regions. Low-susceptibility areas account for 18.1% of the region, whereas the very high and very low susceptibility classes together represent less than 1% of the territory. Model performance was evaluated using Receiver Operating Characteristic (ROC) analysis, yielding an Area Under the Curve (ROC–AUC) value of 0.893, indicating strong agreement between predicted susceptibility patterns and observed flood occurrences. Additional validation metrics derived from the confusion matrix show an overall accuracy of 83.3%, precision of 0.75, recall of 1.0, and a Kappa coefficient of 0.67, confirming reliable predictive performance. Sensitivity analysis with ±10% variation in factor weights further demonstrated the spatial stability of the susceptibility results. The resulting susceptibility map provides an important spatial basis for infrastructure planning, flood mitigation, and disaster preparedness in the Abai Region and offers a transferable framework for flood-susceptibility assessment in other semi-arid regions of Central Asia.

1. Introduction

In recent years, the frequency and magnitude of natural and technological disasters have increased. Floods rank among the most destructive natural hazards worldwide [1]. They cause severe damage to the natural environment and human infrastructure [2,3]. These impacts highlight the requirement for proactive flood risk assessment and efficient mitigation measures. Such measures help protect vulnerable communities and ecosystems [4]. Uncontrolled urbanization, climate change, land-use changes, rapid snowmelt, and heavy rainfall make floods more frequent and more severe [5,6]. Because of this, it becomes increasingly important to identify and map areas at risk of flooding. Assessing vulnerability also plays a key role in improving flood risk management [1,4]. According to statistical records, floods represent the most common natural hazard in Kazakhstan. They occur more frequently than earthquakes and cause greater economic losses [7]. Many settlements face flood risk, including about 750 rural and 850 urban areas. More than 3 million people live in these zones [8,9]. During severe flood years, up to 50,000 hectares of agricultural land are inundated, and around 7000 residential buildings are damaged, affecting a total area of 635,000 m². Floods also impair critical infrastructure, including roads, power supply, and communication networks. For example, the 1993 spring floods affected 669 settlements, resulted in six fatalities, and forced the evacuation of more than 12,700 people [10]. In 2024, major flooding occurred again in eastern Kazakhstan. These floods damaged over 300 houses and led to the evacuation of approximately 1200 residents [11]. In spring 2024, floods affected several regions of Kazakhstan, causing partial inundation of settlements and transport infrastructure [7,12]. River systems in the country are determined by the hydrological characteristics of their basins, and local climate and terrain strongly control river flow modes [13]. Extensive lowland areas allow water to spread over large surfaces, increasing flood risk during extreme hydrometeorological events [14].

Floods cannot be fully prevented; however, efficient mitigation measures can greatly reduce flood damage. Effective flood risk reduction begins with locating areas highly susceptible to flooding. Such information supports preparedness planning and helps reduce potential losses [1,4]. Flood hazard and flood risk maps are key tools in this process, as they guide land-use planning, emergency response, and long-term risk management [2,12]. Reliable risk assessment depends on the precise mapping of the main environmental and human factors that influence flood occurrence and spatial dispersion [4,12,15]. In this context, merging spatial datasets is essential for a complete assessment of elements affecting flood occurrence and extent. Multi-Criteria Decision Analysis (MCDA) provides an organized framework for combining multiple variables into a weighted flood-susceptibility model [4,16]. By evaluating several conditioning factors within a single analytical scheme, MCDA helps identify the most vulnerable areas.

One of the most widely applied MCDA approaches is the Weighted Linear Combination (WLC) method [17], where each factor is assigned a relative weight and combined to produce an overall susceptibility index [18]. The Analytical Hierarchy Process (AHP) is commonly used to obtain these weights via pairwise comparisons based on expert judgment and available evidence [19,20]. The combination of multiple spatial datasets within a Geographic Information Systems (GIS) environment provides a strong basis for accurate and practical flood risk mapping. GIS enables the overlay, processing, and spatial analysis of flood-related factors, supporting the creation of detailed flood-susceptibility and hazard maps. Many studies have successfully applied a combined GIS–AHP approach for flood-susceptibility assessment. For example, study [4] applied AHP combined with remote sensing and GIS data to assess flood risk in a data-scarce region. The analysis included key factors such as elevation, slope, Topographic Wetness Index (TWI), land-use/land cover (LULC), precipitation, distance to rivers and roads, and the Normalized Difference Vegetation Index (NDVI). The resulting five-class flood risk map showed values ranging from 8.71% to 30.99%. Assessment using the Area Under the Curve (AUC) of 0.86 demonstrated good predictive performance in the Dosso region. Similarly, study [2] developed a GIS–AHP flood-susceptibility model for Davidson County, Tennessee. Ten conditioning factors were weighted using AHP and integrated through a weighted overlay in ArcGIS Pro 3.0 (https://www.esri.com/). The final five-level susceptibility map showed high agreement with FEMA flood risk zones, confirming model dependability. Study [1] applied a geospatial MCDA approach to map flood-prone areas in Texas, USA. Factor weights were obtained using AHP, and a weighted overlay produced the final susceptibility map. ROC/AUC validation yielded a value of 0.90, indicating high agreement with historical flood records. The analysis showed that approximately 62% of Texas is exposed to elevated flood risk, particularly in major river basins.

Despite the extensive global application of GIS–AHP models, their adaptation to semi-arid continental environments influenced by snowmelt regimes remains insufficiently explored. Most existing studies focus on humid or monsoon-dominated climates where rainfall intensity is the primary driver of flooding.

The contribution of this study lies in three main aspects:

• Regional specificity—providing the first comprehensive flood-susceptibility assessment for the newly established Abai Region and addressing an important local data gap;
• Hydrological context—considering flood generation controlled by the interaction of spring snowmelt and rapid temperature variability typical of the Central Asian steppe-to-mountain transition;
• Methodological refinement—integrating the Height Above Nearest Drainage (HAND) metric together with selected anthropogenic indicators to improve terrain-based representation of flood processes.

Flood-susceptibility assessment is an important tool in integrated water-resources management and climate adaptation, as it helps identify areas exposed to hydrological extremes under changing climate conditions. These assessments provide a scientific basis for proactive flood risk reduction, spatial planning, and adaptive water-management strategies, especially in data-scarce regions. However, region-specific GIS-based flood-susceptibility assessments remain limited for the Abai Region and the broader continental environments of Eastern Kazakhstan.

The primary goal of this study is to develop a region-specific flood-susceptibility assessment for the newly established Abai Region to support spatial decision-making in a data-scarce environment. The specific objectives are as follows:

• Characterize regional flood drivers by evaluating the interaction of cryospheric (snowmelt), topographic (e.g., HAND and slope), and anthropogenic factors across Eastern Kazakhstan;
• Apply a multi-criteria GIS–AHP framework adapted to continental hydroclimatic conditions through regionally derived factor weights;
• Delineate susceptibility zones for spatial planning, including the identification of priority areas for infrastructure protection and land-use regulation.

Unlike many GIS–AHP studies that mainly focus on rainfall intensity, this study addresses conditions typical of the Central Asian steppe-to-mountain region, where spring floods and ice-related processes create significant hazards. By integrating HAND with anthropogenic factors, we improve terrain-based flood representation and increase the practical value of the susceptibility map for regional planning and flood risk reduction.

2. Study Area

The Abai Region lies in northeastern Kazakhstan and borders the Russian Federation to the north and the People’s Republic of China to the southeast (Figure 1). Authorities recently established it as an independent administrative unit, with Semey as the regional center, following its separation from the former East Kazakhstan Region [11].

The Abai Region is characterized by a highly varied physical and geographical structure, notable for its pronounced structural-tectonic diversity. The area features gently rolling, park-like steppes that transition into western steppe landscapes [23], as well as mountain ranges that run from northeast to south. Among these are the Saur and Tarbagatai ranges, with elevations reaching up to 3000 m above sea level (Figure 1). In the extreme southern part of the region lie the desert-steppe plains of the Balkhash-Alakol intermontane basin [24]. The western and southwestern regions are characterized by the elevated plains that are typical of Central Kazakhstan. To the north and northwest, the hilly terrain of Central Kazakhstan gradually transitions into the vast lowlands of the West Siberian Plain near the Priirtysh-Semey area.

The climate in the Abai Region varies significantly from the mountainous southeast to the lowland northwest, both in temperature and precipitation patterns. Some parts of the region fall within arid and semi-arid climatic zones. The sharply continental climate prevalent in the desert and semi-desert areas becomes more moderate in the foothill and mountainous zones. January is the coldest month, with average temperatures ranging from −12 °C to −20 °C, although extreme lows can reach −45 °C to −49 °C. In contrast, July, the warmest month, sees average temperatures of 20–23 °C, with maximums rising up to +45 °C. Despite a long-term warming trend in the annual mean temperature, minor decreases are noted during the winter seasons. The average annual air temperature ranges from −3.6 to 3.0 °C in the lowland southwestern regions and near large reservoirs like Zaisan, to −6 to −7 °C in the high-mountain areas.

The distribution of precipitation throughout the region is quite irregular. The northeastern mountain and foothill areas receive the highest amounts, ranging from 400 to 650 mm annually, while the intermontane basins experience the least, with less than 200 mm per year. During the warm season, from March to October, precipitation is considerably higher than in the cold season, which lasts from December to March. The peak monthly rainfall usually occurs in June or July, particularly in mountainous regions [7].

The region boasts a wealth of water resources, encompassing over 800 rivers with a combined length surpassing 10,000 km. The Irtysh River serves as the main waterway in the Abai Region, accompanied by significant tributaries like the Shagan, Shulbinka, and Shar rivers. Additionally, the Urdzhar River, one of the three key rivers supplying Lake Alakol, traverses this area. The hydrology of the Irtysh is heavily affected by atmospheric precipitation, including snow, as well as groundwater from the Altai Mountains. The region is also home to significant lakes such as Alakol and Sasykkol, along with numerous smaller lakes and reservoirs, the largest being the Shulbinsk Reservoir [13,14,25].

The Abai Region encompasses a range of natural zones, including steppe, semi-desert, and desert landscapes. It features dry feather-grass–fescue steppes, desert vegetation complexes, and high ecological diversity, driven by distinct latitudinal zones, notably visible in the foothills and within expansive, deep basins. For instance, around 70% of the Zaisan Basin is characterized by sparse vegetation typical of desert and semi-desert areas. The region offers one of the most diverse arrays of vegetation and soil types in Kazakhstan, from the desert communities of the Kazakh Uplands to the taiga forests of the northeastern Altai and tundra ecosystems in the high-mountain zones [26,27].

The regional land-use/land-cover structure is dominated by natural steppe and semi-desert landscapes, interspersed with agricultural areas, forested zones in mountainous regions, and relatively sparse urban settlements concentrated along major river corridors. These surface characteristics strongly influence hydrological response by controlling infiltration capacity, soil-moisture retention, and the generation of surface runoff, thereby shaping spatial flood susceptibility across the Abai Region. A detailed classification of LULC categories and their hydrological roles is provided in Section 3.1.9 [25].

3. Data and Methods

The selection of twelve conditioning factors was informed by the unique hydrological and socio-economic characteristics of the Abai Region. Specifically, HAND and TWI were prioritized to represent local drainage structures and terrain-controlled water accumulation within the steppe environment. In addition, Precipitation and Elevation were weighted to reflect the significance of spring snowmelt. Furthermore, to address a research gap in regional planning, Population Density and Distance to Roads were included to improve the model’s relevance for civil protection and disaster risk reduction.

3.1. Data

A universally accepted set of criteria for creating a flood hazard layer does not exist [15,16]. Nonetheless, flood-susceptibility assessments often consider factors such as slope, proximity to rivers, land-use, and elevation [17]. Few studies have integrated precipitation into GIS-based multicriteria flood-susceptibility analyses [4,18]. Consequently, this study includes both total annual precipitation and effective precipitation to more accurately reflect their impact on flood dynamics.

The occurrence of floods is significantly influenced by local terrain characteristics, soil properties, land-use patterns, and current soil moisture conditions [1]. This study utilized 12 conditioning factors to develop a detailed flood susceptibility map for the Abai Region, encompassing hydrological, topographic, geological/geomorphological, ecological, and human-related components.

These factors include the following:

• Digital Elevation Model (DEM) [22];
• Slope (S) [28];
• Height Above Nearest Drainage (HAND) [29,30];
• Drainage Density (DD) [31];
• Distance from Rivers (DRI) [21];
• Precipitation (P) [7,24];
• Topographic Wetness Index (TWI) [32];
• Soil Texture Classes (ST) [33];
• Land-use/Land Cover (LULC) [34];
• Normalized Difference Vegetation Index (NDVI) [4];
• Distance from Roads (DRO) [21];
• Population Density (PD) [35].

Traditional flood susceptibility assessments focus on terrain characteristics that influence surface water accumulation. This study takes a more integrated approach, combining topographic, hydrological, and climatic variables, particularly precipitation. Human-related factors such as population density and proximity to roads are also included, as human activities often intensify flood impacts. The methodology follows recent research trends: environmental processes and socio-economic pressures are analyzed together to improve spatial risk assessments. Figure 2 schematically illustrates the methodology used in this research.

3.1.1. Elevation (DEM)

Elevation diversity plays a crucial role in determining flood susceptibility. It influences the direction, speed, and collection of surface runoff. Areas at lower elevations are particularly vulnerable to flooding, as gravity tends to gather water in low-lying depressions and valley floors. Higher elevations show less direct flood susceptibility. However, runoff from upland slopes can increase downstream flood hazards [5,15]. Elevation also controls spatial patterns of snow accumulation and the timing of spring snowmelt. These factors, in turn, influence the magnitude and routing of runoff in snowmelt-dominated hydrological regimes [36].

Accurate terrain data are vital for reliable flood modeling. Choosing suitable digital elevation data is therefore critical. The ALOS PALSAR DEM (12.5 m resolution) performs well in flat and lowland areas. However, previous studies show that differences between ALOS and SRTM depend on terrain complexity, vegetation density, and surface attributes. Comparative studies indicate that both SRTM and ALOS produce similar elevation errors relative to LiDAR data. Still, SRTM is more robust to radar artifacts and is frequently preferred for regional assessments in uneven terrains [22].

Taking these factors into account, we use the Shuttle Radar Topographic Mission (SRTM) DEM with a 30 m spatial resolution. This dataset provides consistent coverage and stable quality across the study area. Elevation in the Abai Region ranges from 130 to 2983 m (Figure 3). This variation lets us evaluate how topography influences flood susceptibility [37].

3.1.2. Slope

Slope is a critical topographic factor in flood susceptibility assessment because it determines the velocity, direction, and concentration of surface runoff. Gentle slopes facilitate water accumulation, which increases flood susceptibility. In contrast, steep slopes accelerate water flow, reduce local water accumulation, and may increase downstream flood peaks. Steep gradients also enhance erosion and sediment transport, further affecting flood dynamics. The slope layer was derived from the digital elevation model (DEM) using the Slope tool in ArcGIS 10.6. The resulting slope values ranged from 0 to 60.9° (Figure 3B). These values were applied to classify terrain gradients and to identify areas with higher flood susceptibility [38].

3.1.3. Height Above Nearest Drainage (HAND)

Height Above Nearest Drainage (HAND) is the vertical distance from a location to the nearest drainage channel [29]. Researchers widely use HAND in flood-susceptibility studies because it indicates whether an area is likely to store or convey water [30]. Low HAND values usually indicate valley bottoms and floodplains, which flood more often, whereas high values represent upland areas with lower flood probability.

Compared with traditional topographic indicators such as elevation or slope, the HAND index provides a more direct representation of hydrologically connected terrain positions. While elevation describes absolute terrain height, slope reflects local gradients that control runoff velocity, and HAND identifies areas topographically close to the drainage network, which are therefore more prone to water accumulation. This capability improves the delineation of flood-prone zones, particularly in landscapes with extensive plains and wide river valleys [30].

In the Abai Region, HAND values range from 0 to 744.0 m (Figure 3C), capturing strong topographic variation throughout the area. To create the HAND layer, we derived it from the SRTM DEM using a standard hydrological workflow: calculating flow direction and accumulation, extracting the drainage network, and computing elevation differences relative to the channels [29]. This dataset thus supports the identification of low-lying flood-prone areas and the mapping of natural drainage pathways within the regional river system.

3.1.4. Drainage Density (DD)

Drainage density (DD) refers to the total length of river channels per unit area. This metric reflects the efficiency with which a watershed conveys runoff. High DD values indicate a dense river network that rapidly transfers water, potentially resulting in intense flooding. In contrast, low DD values correspond to sparse channel networks, where water accumulates and may cause more widespread flooding.

In the Abai Region, drainage density ranges from 0 to 16.8 km/km² (Figure 3D). DD was calculated by dividing the total river-network length by the corresponding unit area [30]. The river network was delineated using a hydrologically corrected flow-accumulation raster. This spatial variability in DD facilitates the evaluation of differences in hydrological response across the study area.

3.1.5. Distance from Rivers (DRI)

Floods most commonly occur in the vicinity of river channels; therefore, proximity to rivers constitutes a critical determinant of flood susceptibility. Areas situated within or near active floodplains are especially prone to inundation during episodes of intense rainfall or rapid snowmelt.

The river network was derived from OpenStreetMap vector data. The Euclidean Distance tool in ArcGIS 10.6 was then used to produce a continuous raster representing distances from rivers. The resulting Distance to River Index (DRI) map (Figure 3E) displays values ranging from 0 to 154,163.0 m. This map illustrates spatial patterns of proximity to watercourses and supports the identification of areas with heightened flood susceptibility [22].

3.1.6. Precipitation (P)

Precipitation is one of the primary drivers of flood generation, as it directly influences runoff formation. The amount, duration, and intensity of precipitation control soil saturation levels, while extreme events may exceed infiltration capacity and generate substantial surface runoff, particularly in areas with limited drainage infrastructure. In the hydrological context of the Abai Region, precipitation interacts with seasonal snow accumulation and spring snowmelt processes, which together govern the timing and magnitude of channel overflow. This combined rainfall–snowmelt influence makes precipitation a key variable in regional flood hazard assessment.

Climatic data for this study were obtained from meteorological stations of RSE “Kazhydromet” and KazNIIMOS. After climatological processing, the spatial distribution of precipitation was interpolated using the Inverse Distance Weighting (IDW) method. The resulting precipitation map reveals marked spatial variability, with annual values ranging from 202 to 485 mm (Figure 3F), reflecting atmospheric heterogeneity across the study area (

Table 1. Weather stations located in the Abai Region and their average annual precipitation (2000–2024).

Station Name	Average Precipitation (mm)
SEMIJARKA	231.3
SEMIPALATINSK	312.9
DMITRIEVKA	355.2
KAINAR	225.6
SHALABAY	343.2
ZHALGYZTOBE	308.1
BARSHATAS	222.6
KARAUYL	241.9
KOKPEKTY	311.8
AKTOGAY	202.6
AYAGOZ	295.4
URZHAR	485.2
AKSUAT	217.2
BAKTY	308.6

) [7,39].

3.1.7. Topographic Wetness Index (TWI)

The Topographic Wetness Index (TWI) is a widely used terrain parameter for evaluating flood susceptibility because it represents the potential for water accumulation and persistent soil moisture. TWI integrates upslope contributing area and slope, allowing assessment of how water distributes and concentrates across the landscape [32].

TWI was calculated using the following formula:

$$TWI=ln\left({\displaystyle \frac{Specific Catchment Area}{tan\left(slope\right)}}\right)$$

where

• Specific Catchment Area = (Catchment Area/Unit Contour Length);
• Catchment Area = (Flow Accumulation Raster + 1) × (Area of Each Cell);
• $tan\left(slope\right)=(Slope Raster/100)+0.001$.

Flow-accumulation and percent-slope rasters were integrated in ArcGIS Pro 3.0 (https://www.esri.com/) using the Raster Calculator. The final TWI map (Figure 3J) displays values ranging from 2.5 to 28.3, clearly delineating zones with high moisture concentration and elevated flood potential.

3.1.8. Soil Texture (ST)

Soil texture significantly influences flood susceptibility because it controls infiltration capacity and soil water retention. Clay-rich soils generally have low permeability, which increases surface runoff and enhances flood susceptibility. In contrast, sandy soils promote higher infiltration rates and, therefore, reduce runoff generation.

In this study, soil texture data were obtained from the SoilGrids 250 m dataset (2017) and classified according to the USDA soil–texture system. The dominant soil types in the study area include clay loam, silty clay loam, sandy clay loam, silt, silt loam, and loam. These classes were used to evaluate how spatial variability in soil properties influences flood-susceptibility patterns across the Abai Region [26,33,40].

3.1.9. Land-Use/Land Cover (LULC)

Land-use and land cover (LULC) strongly affect runoff generation, infiltration, and local water storage. Urban areas with impervious surfaces reduce infiltration and significantly increase surface runoff. As a result, heavy rainfall can trigger rapid and intense flooding. In contrast, vegetated areas such as forests, rangelands, and croplands improve water absorption and reduce flood severity.

In this study, we analyzed eight LULC categories: water (1.57%), forest (1.95%), flooded vegetation (0.40%), cropland (4.37%), built-up areas (0.27%), bare ground (0.11%), snow/ice (0.002%), and rangeland (91.33%). Rangeland dominates the Abai Region. The spatial distribution of these categories influences runoff patterns because each land-cover type affects water movement differently. For example, water bodies and flooded vegetation store water, while bare ground increases runoff and erosion. Therefore, LULC analysis helps us better understand flood-susceptibility patterns across the region [34].

3.1.10. Normalized Difference Vegetation Index (NDVI)

Vegetation reduces surface runoff by increasing infiltration and, therefore, lowers flood susceptibility. The Normalized Difference Vegetation Index (NDVI) measures vegetation density and condition using reflectance values from visible and near-infrared bands. Researchers widely apply NDVI in remote-sensing studies to evaluate vegetation cover and its role in hydrological processes.

NDVI is calculated as follows:

$$NDVI={\displaystyle \frac{NIR-Red}{NIR+Red}}$$

Values range from −1 to 1, where negative values indicate water, snow, or clouds; values near zero represent bare soil; and higher values denote dense, healthy vegetation [2,3].

In this study, surface reflectance data from the Landsat 9 satellite (USGS) were processed using the Raster Calculator tool. Bands 4 (Red) and 5 (NIR) were used to generate the NDVI map, with values ranging from 0.4 to 0.7 (Figure 3J). Atmospheric correction and cloud-free Landsat imagery were applied to improve accuracy [41].

3.1.11. Distance from Roads (DRO)

Impervious surfaces such as asphalt, concrete, and pavement increase surface runoff and accelerate flood development. Therefore, distance to roads is an important variable in flood hazard modeling.

In this study, we obtained the road network for the Abai Region from OpenStreetMap. We then calculated Euclidean distance using the Spatial Analyst tools in ArcGIS 10.6. The resulting DRO raster (Figure 3K) shows distances from 0.0 to 71,602.0 m. This layer helps us assess how proximity to transportation infrastructure influences local flood exposure [2,21].

3.1.12. Population Density

Population density indicates the estimated number of people per grid cell. In this study, we used a 3-arc-second (~100 m) resolution GeoTIFF dataset produced by WorldPop. The dataset aligns national population totals with official United Nations estimates. “NoData” values represent uninhabited areas identified using the Built-Settlement Growth Model (BSGM) [35].

Researchers generated this dataset using a random forest–based spatial disaggregation method [35]. Bondarenko and Kerr (WorldPop) performed the subsequent modeling under the guidance of Sorichetta. In the Abai Region, population density ranges from 0 to 925.1 people/km². These data help us assess human exposure to flood hazards.

3.2. Methods

We applied a GIS-based multi-criteria approach to assess flood susceptibility [38,41,42]. Twelve conditioning factors cover hydrological, topographic, environmental, and anthropogenic characteristics. These were integrated within a spatial analysis framework. The Analytical Hierarchy Process (AHP) [4], a common Multi-Criteria Decision Analysis (MCDA) method, helped derive the relative weights and support flood-susceptibility assessment in the Abai Region. We combined remote sensing (RS) data with Geographic Information Systems (GIS) for spatial processing and overlay analysis of selected variables [43,44,45].

Building on this framework, we select hazard criteria based on factors that strongly influence riverine flooding and assess social vulnerability using population density data [46].

Subsequently, all analyses are performed in ArcGIS 10.6. All datasets are standardized to a common projection (EPSG: 32644, WGS 84/UTM Zone 44N) [47] and a spatial resolution of 30 × 30 m, ensuring consistency across all thematic layers. Finally, the layers are reclassified and integrated using a weighted overlay technique based on the AHP method [48,49].

3.2.1. MCDA

To ensure consistent processing in flood-susceptibility modeling, we classified all spatial layers according to their characteristics and international cartographic standards. For continuous geophysical variables such as DEM, HAND, drainage density (DD), distance to rivers (DRI), distance to roads (DRO), and population density (PD), we applied quantile classification. This method distributes values evenly across classes.

For other continuous variables, including slope, precipitation, TWI, and NDVI, we used the Natural Breaks (Jenks) method to identify natural groupings within the data. We applied bilinear or cubic interpolation to these layers to create smooth transitions between raster cells and better represent gradual environmental gradients.

For categorical datasets such as LULC and soil texture, we classified values by type. We used the nearest neighbor resampling method to preserve discrete class values and avoid mixing during spatial processing. This approach maintains thematic integrity and ensures accurate integration within the MCDA-based flood risk model [16].

3.2.2. AHP

The Analytic Hierarchy Process (AHP) is widely applied in spatial decision-making research [2,4] and has demonstrated high reliability and accuracy when integrated with GIS for environmental assessments. GIS-based AHP is particularly effective at synthesizing large volumes of heterogeneous data and deriving factor weights for complex multi-parameter evaluations [48].

Criterion weights were derived using the conventional crisp Analytical Hierarchy Process (AHP) based on Saaty’s 1–9 pairwise comparison scale. Pairwise comparison matrices were constructed to evaluate the relative importance of flood-conditioning factors, and the Consistency Ratio (CR) was calculated to verify the reliability of expert judgments. No fuzzy AHP formulation or fuzzy membership functions were applied in this study.

The pairwise comparison matrices were completed by seven domain experts selected according to clearly defined professional criteria and relevant expertise in hydrology, flood risk management, climate analysis, and GIS-based spatial modeling. All experts possessed substantial professional or research experience in water-related environmental assessment and spatial hazard analysis. The final pairwise comparison matrix was obtained by aggregating individual expert judgments using the geometric mean method.

Flood-susceptibility assessment follows several steps. First, we structure the problem into a hierarchy that reflects relationships among factors influencing flood hazards. Experts then complete pairwise comparison matrices to evaluate the relative importance of each factor. The final pairwise comparison matrix was obtained by aggregating individual expert judgments using the geometric mean method.

Next, a weighted linear combination method was applied. Each reclassified thematic layer was multiplied by its corresponding weight derived from the pairwise comparisons [4,19]. The comparison matrix was normalized, and the principal eigenvector was calculated to obtain the final factor weights. The reliability of the weighting scheme was confirmed when the Consistency Ratio met the accepted threshold (CR ≤ 0.1), indicating consistent expert evaluations (

Table 2. Saaty scale (1987) [19,20,50].

Numerical Rating	Reciprocal
1	1
2	1/2
3	1/3
4	1/4
5	1/5
6	1/6
7	1/7
8	1/8
9	1/9

).

In the hazard comparison matrix, each pair of criteria was evaluated such that the better option received a score ranging from 1 (equally important) to 9 (extremely more important), while the less favorable option was assigned the reciprocal value. The ratings were then averaged and adjusted. Most experts assessed the relative importance of one indicator compared to another. A total of seven domain experts contributed to the construction of the pairwise comparison matrices.

We assess the consistency of these judgements by calculating the Consistency Index (CI) and Consistency Ratio (CR) [19,20]. The Consistency Index quantifies the deviation of the comparison matrix from a perfectly consistent state and is calculated as follows:

$$CI={\displaystyle \frac{{\lambda }_{max}-n}{n-1}}$$

where $n$ is the number of criteria and ${\lambda }_{max}$ is the maximum eigenvalue of the comparison matrix. For a perfectly consistent matrix, ${\lambda }_{max}=n$, yielding $CI=0$.

The Consistency Ratio provides an overall measure of the reliability of expert judgments and is computed as follows:

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

where $RI$ is the Random Index corresponding to the matrix size $n$. Commonly accepted thresholds are as follows:

• $CR<0.10$: the matrix is consistent;
• $0.10\le CR\le 0.20$: acceptable but may require improvement;
• $CR>0.20$: the matrix is inconsistent.

AHP enables the calculation of factor weights and priorities and supports structured evaluation of complex environmental processes within spatial decision-making frameworks [16,50,51].

3.2.3. Model Validation and Robustness Assessment

To validate the GIS–AHP flood-susceptibility model, independent flood-occurrence data from historical records and documented inundation areas in the Abai Region were used. The flood inventory included settlement locations marked as flood-affected (1) or non-flood (0) and was excluded from AHP weight derivation.

Predictive performance was measured by ROC analysis and the Area Under the Curve (AUC). This method tests how mapped susceptibility matches observed floods without relying on a threshold. Susceptibility-index values from the GIS–AHP raster at inventory sites were matched with binary flood labels to form the ROC curve and AUC metric.

ROC/AUC calculations were completed in R 4.6 (https://www.r-project.org/) with the pROC package (https://rpubs.com/Wangzf/pROC, 8 February 2026). AUC near 1 means strong separation of flood and non-flood sites; 0.5 is random [52,53].

As GIS–AHP is expert-driven, sensitivity analysis assessed the robustness of the weighting scheme. Key conditioning-factor weights were changed by ±10%, renormalized, and the susceptibility map recalculated. Model stability was examined by comparing High and Very High susceptibility zones across scenarios. Limited variation showed robust results despite moderate uncertainty in expert weights.

4. Results and Discussion

The results integrate global flood hazard principles with region-specific hydrological characteristics of the Abai Region, where continental climate, complex terrain, and rapid runoff processes intensify flood susceptibility.

4.1. Reclassification of Flood Hazard Components

We reclassified all twelve conditioning factors into five ordinal classes (1–5), ranging from very low to very high susceptibility (

Table 3. Flood hazard thresholds.

Level	Class	DEM (m)	S (Degree)	HAND (m)	DD (km/km2)	DRI (m)	P (mm)	TWI	ST	LULC	NDVI	DRO (m)	PD (Peop/km2)
Very high	5	130.0–350.0	0.0–2.0	0.0–2.0	7.6–16.8	0.0–3600.0	380.1–485.0	10.1–28.3	Silty clay loam (SiCL)	Built area, Bare ground	−0.4–0.0	0.0–1200.0	18.1–925.1
High	4	350.1–500.0	2.1–5.5	2.1–4.0	5.4–7.5	3.600.1–8400.0	330.1–380.0	8.1–10.0	Clay loam (CL), Silt (Si), Silt loam (SiL)	Crops, Rangeland	0.01–0.1	1200.1–3300.0	7.1–18.0
Medium	3	500.1–650.0	5.6–11.0	4.1–6.0	3.6–5.3	8400.1–15,100.0	290.1–330.0	7.1–8.0	Sandy clay loam (SCL)	Forest, Flooded vegetation	0.1–0.2	3300.1–7000.0	3.6–7.0
Low	2	650.1–800.0	11.1–20.0	6.1–8.0	1.6–3.5	15,100.1–26,600.0	260.1–290.0	6.1–7.0	Loam (L)	Natural wetlands, Flooded vegetation	0.2–0.3	7000.1–13,750.0	0.004–3.5
Very low	1	800.1–2983.0	20.1–60.9	8.1–744.0	0–1.5	26,600.1–154,163.0	202.0–260.0	2.5–6.0	-	Water bodies	0.3–0.7	13,750.1–71,602.0	0.0–0.003

). This approach follows widely applied multi-criteria flood hazard frameworks [54,55]. Reclassification ensures consistent comparison among different types of datasets, including continuous, categorical, and distance-based variables. It also standardizes the inputs for multi-criteria analysis [16].

We then applied the Weighted Overlay tool to integrate all layers into a comprehensive flood-susceptibility map (Figure 4). The resulting map shows clear spatial variability across the study area. It highlights strong physiographic contrasts between the mountainous eastern region and the lowland floodplains.

These spatial contrasts provide an important hydrological basis for interpreting regional flood-generation processes, as runoff accumulation and flood susceptibility are primarily concentrated within low-lying valley systems and river-adjacent terrains.

4.2. Matrix Comparison

The flood-susceptibility model is founded on the AHP pairwise comparison matrix (

Table 4. Pairwise comparison matrix of flood-conditioning factors based on the AHP method.

	DRI	Precipitation	HAND	DD	DEM	S	TWI	ST	LULC	DRO	NDVI	PD
DRI	1.0	1.0	2.0	2.0	2.0	2.0	2.0	3.0	7.0	5.0	9.0	9.0
Precipitation	1.0	1.0	1.0	2.0	2.0	2.0	2.0	2.0	7.0	5.0	7.0	4.0
HAND	0.5	1.0	1.0	1.0	2.0	1.0	1.0	2.0	2.0	2.0	7.0	3.0
DD	0.5	0.5	1.0	1.0	1.0	1.0	1.0	2.0	2.0	2.0	5.0	2.0
DEM	0.5	0.5	0.5	1.0	1.0	1.0	1.0	2.0	2.0	2.0	4.0	2.0
S	0.5	0.5	1.0	1.0	1.0	1.0	1.0	1.0	2.0	2.0	4.0	2.0
TWI	0.5	0.5	1.0	1.0	1.0	1.0	1.0	1.0	2.0	2.0	4.0	2.0
ST	0.33	0.5	0.5	0.5	0.5	1.0	1.0	1.0	1.0	1.0	2.0	3.0
LULC	0.14	0.14	0.5	0.5	0.5	0.5	0.5	1.0	1.0	1.0	2.0	3.0
DRO	0.2	0.2	0.5	0.5	0.5	0.5	0.5	1.0	1.0	1.0	2.0	3.0
NDVI	0.11	0.14	0.14	0.2	0.25	0.25	0.25	0.5	0.5	0.5	1.0	3.0
PD	0.11	0.25	0.33	0.5	0.5	0.5	0.5	0.33	0.33	0.33	0.33	1.0

), which quantifies the relative influence of the twelve conditioning factors according to Saaty’s scale [19,20], ranging from 0.11 (very weak influence) to 9.0 (extremely strong influence). The resulting matrix ensures methodological transparency and supports reproducibility in multi-criteria decision-making analyses [16]. The weighting coefficients were derived from pairwise comparisons provided by seven domain experts, as described in Section 3.2.2 [1,2]. The classical AHP approach was considered appropriate due to the structured and internally consistent nature of the input datasets.

The final AHP-derived weights (

Table 5. Final AHP-derived weights of flood-conditioning factors.

No.	Criterion	Geometric Mean (GMi)	Weight (wi)	Weight, %
1	Distance from Rivers	2.8373	0.1966	19.66%
2	Precipitation	2.3697	0.1642	16.42%
3	HAND	1.5326	0.1062	10.62%
4	Drainage Density	1.2836	0.0889	8.89%
5	Elevation	1.1892	0.0824	8.24%
6	Slope	1.1892	0.0824	8.24%
7	TWI	1.1892	0.0824	8.24%
8	Soil Texture	0.8402	0.0582	5.82%
9	LULC	0.6268	0.0434	4.34%
10	Distance from Roads	0.6652	0.0461	4.61%
11	NDVI	0.3416	0.0237	2.37%
12	Population Density	0.3707	0.0255	2.55%

) reveal a clear hierarchical ranking of the conditioning factors. Distance from rivers (19.66%) and precipitation (16.42%) represent the most influential variables in the study region. This pattern is consistent with previous hydrological studies identifying river proximity and rainfall intensity as key flood-generating mechanisms across diverse climatic settings, including continental and semi-arid regions [24,41,56,57]. In continental environments such as Kazakhstan and northern China, fluvial flooding may be influenced by seasonal snowmelt in combination with intense rainfall events [49,58]. The high weight for precipitation reflects the vulnerability of semi-arid regions to intense, sporadic rainfall events that generate rapid surface runoff due to limited infiltration capacity [59].

Topo-hydrological factors (HAND, Drainage Density, Elevation, Slope, and TWI) received moderate but important weights (8–11% each). Their combined contribution highlights the strong influence of terrain on flood processes. HAND effectively represents local drainage potential [39] and is particularly relevant in the undulating foothills and plains of the study area. The moderate weight assigned to TWI aligns with studies from similar landscapes in northern China, where it identifies saturation-prone areas but remains less influential than direct hydrological factors [60].

Soil Texture, LULC, and NDVI received lower weights (2–6%). These variables act as secondary modifiers rather than primary drivers. Although soil properties affect infiltration, intense precipitation events may override their influence in this regional context.

The modest weight for LULC, often higher in urbanized watersheds, suggests that in the Abai Region, natural topography and hydrology currently outweigh anthropogenic land cover changes in flood generation, a finding noted in similar continental regions [26,27]. Population density and distance from roads received the lowest weights and were therefore interpreted as anthropogenic spatial modifiers of flood susceptibility rather than direct indicators of social vulnerability. Their inclusion reflects potential influences of settlement distribution and infrastructure on runoff concentration and exposure patterns, while maintaining the conceptual distinction between susceptibility (hazard potential) and risk, as adopted in modern flood-assessment frameworks [35].

The pairwise comparison matrix (Table 4) and resulting AHP weights (Table 5) demonstrate high internal consistency, with a Consistency Ratio of CR = 0.0206, well below the accepted threshold of 0.10 (

Table 6. Random Index (RI) values for AHP pairwise comparison matrices of order n.

n	1	2	3	4	5	6	7	8	9	10	11	12
RI	0	0	0.58	0.90	1.12	1.24	1.32	1.41	1.45	1.49	1.51	1.54

). This reliable hierarchy of flood-controlling factors provides a robust hydrological basis for spatial flood risk interpretation and supports evidence-based land-use planning and integrated water-management strategies in the Abai Region.

To evaluate potential multicollinearity among terrain-derived predictors, Pearson correlation coefficients were calculated for the Digital Elevation Model (DEM), slope, and Topographic Wetness Index (TWI). The results indicate moderate relationships (DEM-slope r = 0.599; DEM-TWI r = −0.361; slope-TWI r = −0.457), suggesting partial association but no strong redundancy among variables (

Table 7. Pearson correlation coefficients among terrain variables.

Variable	DEM	Slope	TWI
DEM	1.000	0.599	−0.361
Slope	0.599	1.000	−0.457
TWI	−0.361	−0.457	1.000

).

These findings confirm that DEM, slope, and TWI represent complementary terrain-controlled hydrological processes-topographic position, runoff velocity, and moisture accumulation-supporting their joint inclusion in the GIS–AHP flood-susceptibility model without violating independence assumptions commonly required in multi-criteria decision analysis [61,62].

4.3. Spatial Distribution of Flood Susceptibility

The final susceptibility map classifies the region into five distinct categories (

Table 8. Area distribution of flood susceptibility classes.

Flood Susceptibility	Area (km2)	Area (%)
Very Low	149.2	0.1
Low	33 011.4	18.1
Moderate	102 972.7	56.6
High	45 369.8	25.0
Very High	315.3	0.2

). Approximately 25.0% of the territory corresponds to High susceptibility and 0.2% to Very High susceptibility, primarily concentrated in low-lying areas adjacent to major river channels and in zones of elevated drainage density. This spatial configuration reflects characteristic floodplain inundation dynamics and is consistent with regional flood-mapping studies across Central Asia, where riverine corridors represent the most hazardous environments [5].

The extensive Moderate susceptibility class (56.6%) predominantly occupies transitional slopes and foothill landscapes, indicating areas where flooding may occur under extreme precipitation or localized runoff accumulation. Such spatial dominance is consistent with the combined influence of precipitation variability and topographic control observed in continental foothill systems, including the northern Tien Shan and comparable environments in Xinjiang, China [56,58].

Similar spatial patterns have been reported in recent regional investigations of soil-moisture variability and hydroclimatic indices in East Kazakhstan, where antecedent saturation and intense precipitation jointly generate hazardous runoff conditions [27]. The broad extent of Moderate susceptibility, therefore, suggests that flood susceptibility in the Abai Region is not confined solely to major floodplains but represents a landscape-scale hydrological characteristic, emphasizing the importance of integrated watershed management and early-warning preparedness for surface and flash-flood processes.

In contrast, Low and Very Low susceptibility zones (18.2% combined) correspond to higher elevations and steeper slopes that promote rapid drainage and reduced water accumulation. This inverse relationship between elevation, slope, and flood susceptibility is a globally recognized geomorphological pattern and is clearly expressed within the study area [54].

Overall, the spatial correspondence between modeled high-susceptibility zones, known floodplain extents, and documented historical flood locations supports the physical plausibility of the AHP-based model and demonstrates its suitability for flood-susceptibility assessment in semi-arid continental environments such as the Abai Region (Figure 5). These results provide an important spatial foundation for evidence-based land-use planning, disaster-risk reduction, and integrated water-resources management under variable hydroclimatic conditions.

4.4. Model Validation and Sensitivity Analysis

Model validation was conducted using an independent flood inventory compiled from documented historical flood occurrence records at settlement locations within the Abai Region. Each settlement was assigned a binary validation label (1 = flood observed; 0 = non-flood/control), and the corresponding GIS–AHP susceptibility class was extracted from the final flood susceptibility map. The settlement-based flood inventory provides an intuitive “ground check” of the susceptibility map by comparing (1) observed flood occurrence at settlement locations with (2) the susceptibility class assigned by the GIS-AHP model at those same locations. The compiled inventory contains 44 settlements in total: 33 flood-affected locations (class = 1) and 11 non-flood/control locations (class = 0), with the modeled susceptibility classes predominantly falling into High and Very High for flood-affected settlements (

Table 9. Historical flood inventory settlements and modeled flood susceptibility classes [63,64,65].

Settlement Name	Flood Inventory Class (1/0)	Flood Susceptibility
Ayagoz city	1	High
Begen village	0	Moderate
Semey city	1	High
Belokamenka village	1	High
Beskaragay village	0	Moderate
Birlik village	1	High
Cheremushka village	1	High
Dolon village	1	High
Eginsu village	1	Very High
Glukhovka village	1	High
Grachi village	1	High
Karabas village	0	Moderate
Karaul village	0	Moderate
Kokzhyra village	1	High
Kopa village	1	High
Krasny Yar village	0	Moderate
Krivinka village	0	Moderate
Kumkol village	0	Moderate
Kyzyl Zhuldyz settlement	1	High
Kyzyl-Kesik village	1	High
Malkeldy village	0	Moderate
Mamyrsu village	1	High
Mirny settlement	1	High
Mukyr village	1	High
Novopokrovka village	1	High
Orlovka village	0	Moderate
Oyshilik village	0	Moderate
Peremenovka village	1	High
Rechnoye village	1	High
Sarjal village	1	High
Steklyanka village	1	High
Stepnoy settlement	1	High
Tana Myrza village	1	High
Tasaryk village	1	High
Uan settlement	1	High
Ukilikyz (Vozdvizhinka) village	0	Moderate
Vostochny settlement	1	High
Yesim settlement	1	High
Zhanatilek village	1	High
Zhantikey village	1	High
Zhetizhar village	1	High
Zhogary Eginsu village	1	Very High
Zholamanovka settlement	1	High
Znamenka village	1	High

Note: “Flood inventory class” is the independent validation label (1 = historically flood-affected; 0 = non-flood/control).

) [53].

To quantitatively assess predictive performance, susceptibility index values were extracted at inventory locations from the final raster and evaluated using Receiver Operating Characteristic (ROC) analysis. ROC curves provide a threshold-independent assessment of discrimination skill by summarizing the sensitivity–specificity trade-off across all possible cut-off values. The ROC curve (Figure 6) produced an Area Under the Curve (AUC) value of 0.893, indicating strong agreement between modeled susceptibility patterns and independent flood observations. In line with commonly used interpretation guidelines, an AUC within the 0.8–0.9 range reflects excellent discrimination capability of the model [66].

In addition to ROC–AUC analysis, model performance was further evaluated using a confusion matrix approach. Classification metrics, including overall accuracy, precision, recall, and Cohen’s Kappa coefficient, were calculated using the caret package in R (https://cran.r-project.org/web/packages/caret/index.html, 8 February 2026). The results indicate good predictive capability of the model, with an overall accuracy of 83.3%, precision of 0.75, recall of 1.0, and a Kappa coefficient of 0.667, indicating substantial agreement between predicted flood-susceptibility zones and observed flood locations.

To evaluate model robustness, the weights of the dominant conditioning factors were perturbed by ±10%, and the flood susceptibility map was recalculated using identical classification thresholds. The resulting changes in the spatial extent of susceptibility classes are summarized in

Table 10. Sensitivity analysis of flood susceptibility classes under ±10% perturbation of dominant AHP weights.

Flood Susceptibility Class	Baseline (km2)	−10% (km2)	Change (%)	+10% (km2)	Change (%)
Very Low	149.2	91.2	−38.9%	278.5	+86.7%
Low	33 011.5	30 769.6	−6.8%	37 102.0	+12.4%
Moderate	102 972.7	102 996.1	+0.02%	100 752.5	−2.2%
High	45 369.7	47 690.6	+5.1%	43 218.4	−4.7%
Very High	315.3	270.9	−14.1%	466.8	+48.1%

.

The sensitivity analysis indicates that the High susceptibility class varied within a relatively narrow range (−4.7% to +5.1%) compared to the baseline scenario. The Moderate class exhibited minimal variation (within ±2.2%), suggesting strong structural stability of the model. Although more pronounced percentage changes were observed for the Very High and Very Low classes, these categories represent small spatial proportions of the study area and are inherently more sensitive to boundary shifts. Larger relative changes observed in the Very High and Very Low classes are attributed to their limited spatial extent and do not indicate structural instability of the model.

Importantly, the spatial configuration of major flood-prone corridors along primary river systems remained consistent across all perturbation scenarios. No systematic displacement of hotspot areas was observed. These findings confirm that the GIS–AHP model is not overly sensitive to reasonable variations in expert-derived weights and demonstrates satisfactory robustness for regional-scale flood susceptibility assessment [66,67].

4.5. Limitations and Future Research Directions

Although the GIS–AHP approach produced a clear and interpretable flood-susceptibility map, the method has several limitations. First, AHP relies on expert judgment. Even though the Consistency Ratio (CR) confirms acceptable consistency, the method still includes subjectivity. Future studies could apply sensitivity analysis or combine AHP with data-driven techniques such as Random Forest or Maximum Entropy (MaxEnt). Recent studies show that such hybrid approaches improve objectivity in factor weighting [1,2,15].

Second, data resolution limits model accuracy. The 30 m SRTM DEM does not fully capture micro-topographic features that control local surface flow. In data-scarce regions such as Central Asia, researchers could use higher-resolution global datasets (e.g., ALOS or Copernicus DEM) or conduct UAV surveys in critical areas [22,23]. In addition, incorporating dynamic variables such as snowmelt timing—an important flood driver in Central Asia and northern China [5,6]—and higher temporal resolution precipitation data would improve model realism in climate-sensitive regions [26].

Third, this study assesses flood susceptibility, which reflects the natural tendency of an area to flood. A complete flood risk assessment should integrate the hazard map with detailed exposure data (e.g., infrastructure and agricultural land) and social vulnerability indicators (e.g., age and poverty). This integrated approach aligns with the Sendai Framework and advanced flood risk methodologies [3,4].

Despite these limitations, this study provides a foundational assessment of flood susceptibility in the Abai Region. It establishes a reproducible methodology, identifies flood-prone spatial corridors, and offers a scientifically grounded tool for prioritizing land-use planning, early warning systems, and targeted flood-mitigation strategies. Future research should focus on validating the susceptibility map using historical flood inventories, incorporating climate-change projections to assess potential shifts in flood susceptibility, and progressing toward an integrated socio-environmental flood risk assessment framework.

5. Conclusions

Analysis of twelve conditioning factors revealed that distance to rivers (20%) and precipitation (16%) are the principal drivers of flood susceptibility, highlighting the importance of river proximity and seasonal flow dynamics. Topo-hydrological variables, including HAND (11%) and drainage density (9%), further improve representation of terrain-controlled runoff accumulation. The final susceptibility map classifies 25.0% of the region as high susceptibility and 56.6% as moderate susceptibility, primarily concentrated in low-lying floodplains and foothill areas of the Irtysh River basin. Model reliability is confirmed by a low Consistency Ratio (CR = 0.0206) and good predictive performance based on ROC–AUC validation (AUC = 0.893). Sensitivity analysis further demonstrated that the model is stable under reasonable variations in expert-derived weights.

Despite these strengths, the static susceptibility framework is limited by medium-resolution input data, expert-driven weighting, and the exclusion of dynamic hydrological drivers. Future research should incorporate higher-resolution datasets, climate projections, and integrated vulnerability indicators to improve predictive capability.

Overall, the resulting susceptibility map provides a useful spatial basis for disaster-risk reduction, integrated water-resources management, climate-adaptation planning, and evidence-based land-use regulation in the Abai Region under evolving hydroclimatic conditions. Although the present study focuses on flood susceptibility, hydroclimatic extremes in continental Central Asia also include recurrent drought conditions. Future research should therefore extend the proposed GIS–AHP spatial framework toward integrated flood–drought risk assessment and adaptive water-resources management under increasing climate variability.