Next Article in Journal
Advancing Sustainable Learning in Education of Learners with Visual Impairment in Historically Disadvantaged Schools
Previous Article in Journal
A Framework for Building Sustainability Assessment for Developing Countries Using F-Delphi: Moroccan Housing Case Study
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Extreme Precipitation and Flood Hazard Assessment for Sustainable Climate Adaptation: A Case Study of Diyarbakır, Turkey

Department of Civil Engineering, Dicle University, Diyarbakir 21280, Turkey
*
Author to whom correspondence should be addressed.
Sustainability 2025, 17(20), 9339; https://doi.org/10.3390/su17209339
Submission received: 17 July 2025 / Revised: 30 September 2025 / Accepted: 5 October 2025 / Published: 21 October 2025

Abstract

This study investigates flood risk trends using rainfall data collected from 13 districts of Diyarbakır Province, Turkey, with a focus on supporting sustainability-oriented climate adaptation. Both annual and seasonal precipitation variations were examined, with particular emphasis on the role of maximum daily rainfall in driving flood potential. In addition, the analysis integrates extreme precipitation patterns with regional hazard characteristics to provide a more comprehensive flood risk assessment framework. Non-parametric statistical methods, including the Mann–Kendall trend test and Spearman’s Rho correlation, were applied to detect trends in annual and seasonal datasets. Flood magnitudes were estimated using the Generalized Extreme Value (GEV) and Peaks Over Threshold (POT) approaches. The dataset covers varying periods between 2009 and 2023, depending on station availability. The results show a statistically significant increase in both annual and winter precipitation at Bismil, and a significant winter increase at Çermik. Other stations displayed upward trends that were not statistically significant. Çüngüş, Lice, and Kulp were identified as particularly susceptible to extreme rainfall. Although the relatively short observation period poses a limitation, consistent patterns of intensified precipitation were detected. Previous studies in Turkey have demonstrated that such events often cause severe infrastructure damage and displacement of vulnerable communities. The findings of this study provide practical insights for national and regional authorities, including the Disaster and Emergency Management Authority (AFAD), the General Directorate of State Hydraulic Works (DSİ), and the Ministry of Environment, Urbanization, and Climate Change, to strengthen sustainable climate adaptation planning and disaster risk reduction strategies. Overall, this research highlights the importance of integrating extreme precipitation analysis into sustainable flood management, resilient infrastructure development, and long-term sustainability policies, thereby reinforcing the connection between hydrological risk assessment and sustainability science.

1. Introduction

Flash floods are among the most destructive and frequent natural hazards worldwide, posing significant threats to both human life and the environment. They occur suddenly, often leaving little time for preparedness or mitigation [1]. Each year, floods and flash floods affect millions of people globally, leading to displacement, loss of livelihoods, and substantial economic damage [2]. These intense hydrological events frequently result in widespread infrastructure damage and fatalities, particularly in regions with vulnerable terrain or inadequate early warning systems [3]. The increasing frequency and intensity of such events have been strongly associated with climate change, rapid urbanization, and deforestation, which collectively accelerate surface runoff and reduce natural infiltration [4]. Recent studies highlight the importance of integrated flood risk assessments that combine geomorphological, hydrological, and geospatial approaches to improve prediction and mitigation strategies [5]. For example, AlRifai et al. [6] conducted a GIS-based flood risk assessment in Saudi Arabia using morphometric parameters and GeoAI, emphasizing the critical role of terrain analysis in arid regions. Similarly, Durlević [7] examined the Mlava River Basin in Serbia and underscored the compounded risks posed by torrential floods and landslides.
Turkey is situated in a geologically complex and topographically diverse region, making it highly susceptible to natural hazards. The interaction of structural and meteorological factors often results in disasters such as floods, landslides, erosion, avalanches, droughts, and earthquakes [8]. While these events cannot be entirely prevented, their impacts can be mitigated through effective risk management strategies [9].
Floods, in particular, are among the most destructive hazards, generally occurring during intense rainfall when river discharge exceeds channel capacity and inundates adjacent areas. Globally, floods cause substantial economic losses and significant threats to human life. In Turkey alone, 2563 flood events were recorded between 1955 and 2014, leading to approximately 1500 deaths and an average annual economic loss of around 100 million USD [10].
Comprehensive risk analyses that integrate extreme precipitation and river levels in urban settings are essential to better understand flood dynamics and to plan adaptive infrastructure [11]. In the Valencia region, for instance, climate-driven flood risks are amplified by urbanization and land use changes, prompting the adoption of adaptive, nature-based strategies in line with the UN Sustainable Development Goals [12].
Accurate estimation of peak flood discharges is fundamental for predicting potential damages. Accordingly, flood risk assessments play a crucial role in enhancing pre-disaster planning, hazard mapping, and emergency response strategies [13,14]. Post-disaster recovery and rehabilitation efforts are equally important for restoring social and economic stability in affected regions [15,16].
Projections under global climate change scenarios suggest a marked increase in future flood risks. Alfieri et al. [17] reported that in a warming climate, both the frequency and severity of river floods are expected to rise worldwide, with existing infrastructural vulnerabilities further intensifying these risks.
Flood risk assessment requires the integration of multiple factors, including precipitation, topography, drainage networks, land use, and soil properties [18]. Geographic Information Systems (GISs) and hydrological/hydraulic modeling tools are indispensable for analyzing these complex interrelations. GIS-based platforms such as NetCAD facilitate spatial risk evaluations, while models like HEC-HMS and HEC-RAS simulate rainfall–runoff processes and floodplain dynamics [19,20].
GIS technologies (e.g., ArcGIS, QGIS, HEC-RAS) have become essential for reducing both time and cost in flood management while supporting data-driven decision making. Numerous regional studies across Turkey, including those conducted in Artvin [21,22], İzmir, and the Seyhan Basin [23,24] demonstrate the effectiveness of GIS-based flood risk assessments. Regions receiving intense precipitation, located near river channels, or situated on steep slopes are especially vulnerable. Anthropogenic drivers such as deforestation, rapid urbanization, and land use changes further exacerbate flood hazards [23].
Recent graduate and doctoral research has successfully employed these methods in different basins across Turkey. For instance, a study in the Göksu Stream Basin of Istanbul (2002–2023) found that while urbanization only slightly expanded floodplain areas, it substantially increased the number of buildings at risk [25]. In the Esmahanım Stream Basin of Düzce, two-dimensional modeling identified 1022 residents and 393 structures at risk under a 500-year flood scenario [26]. In the Dinsiz River Basin of Sakarya, industrial damage was estimated using HAZUS-MH functions, which yielded higher probabilities of damage compared to traditional Pistrika & Jonkman curves [27]. In the Söğütlü River Basin of Trabzon, the Analytic Hierarchy Process (AHP) was applied to weight physical and anthropogenic factors, revealing that 2.3% of the basin was classified as “high” or “very high” risk [28]. Similarly, one-dimensional hydraulic modeling in the Ağrı–Doğubayazıt basins combined with the Frequency Ratio method reported prediction accuracies ranging from 76% to 84% [29]. These findings underscore the value of GIS-assisted hydrodynamic simulations for accurately representing flood propagation and improving damage estimation when integrated with population and infrastructure data.
In flood risk assessments, the frequency and magnitude of extreme precipitation events are often analyzed using the Generalized Extreme Value (GEV) and Peaks Over Threshold (POT) approaches. GEV estimates flood discharges based on annual maxima, whereas POT incorporates all extreme events exceeding a predefined threshold, offering more detailed and reliable risk projections. Particularly in cases with limited data, POT provides greater flexibility and is advantageous for analyzing rare but high-impact events [30]. For this reason, both methods are widely used in flood risk studies from the perspectives of engineering, disaster management, and urban planning [31,32].
Cherif [33] applied GEV to assess extreme values of climatic variables in Mediterranean cities, while Kolaković et al. [34] used POT to calculate flood return values in river systems, reporting more reliable results than block maxima analyses. Bat-Erdene et al. [35] modeled annual maximum rainfall using GEV in sparsely gauged basins in Mongolia. Reinders and Munoz [36] demonstrated that GEV performed poorly in arid regions of the U.S., where LN3 and P3 distributions were more suitable. Berton and Rahmani [37] compared GEV and LP3, finding that GEV yielded more reliable estimates for return periods exceeding 50 years. Pan et al. [38] showed that POT produced less biased results than Annual Maxima methods, particularly for frequently recurring floods. In a later study, Pan et al. [39] developed a regional flood frequency analysis framework in Australia using a regionalized POT model and regularized regression, concluding that POT outperformed AMS.
While recent studies have advanced integrated flood risk frameworks by incorporating vulnerability metrics and decision-support techniques [40,41], the present work does not compute a Flood Vulnerability Index (FVI) and instead focuses specifically on the hazard dimension of flooding. By concentrating on extreme precipitation and terrain-driven accumulation potential, this study complements such vulnerability-oriented research with a clear hazard-based perspective.
Within this context, while most existing literature concentrates on individual basins or urban corridors, the present study adopts a broader perspective by evaluating precipitation trends across 13 hydrologically and topographically diverse districts in the Southeastern Anatolia Region of Turkey. The main objective is to analyze both annual and seasonal extreme precipitation patterns and to assess their spatial variability in relation to flood risk. In addition to identifying statistically significant trends, the study employs both frequency-based (GEV) and threshold-based (POT) methods to evaluate flood susceptibility. Moreover, it highlights the influence of region specific hydro-meteorological factors such as evaporation, terrain, and short-duration intense rainfall on flood hazard. By integrating statistical trend detection with localized extreme event modeling, this research provides a more nuanced and scalable framework for regional flood risk assessment under changing climatic conditions. Despite the limitation of relatively short data records, the study offers a rare province-level application that combines trend and frequency analyses, thereby contributing an original perspective on the climatic drivers of flood risk in Diyarbakır.
Diyarbakır Province was selected as the study area because of its diverse geographical and hydrometeorological conditions. The region encompasses both mountainous districts in the north and low-lying agricultural plains in the south, where the Tigris River and its tributaries increase flood susceptibility. Northern districts are more prone to intense precipitation events, whereas southern areas frequently experience poor drainage and low-slope conditions that exacerbate surface runoff. In addition, the availability of reliable rainfall records from 13 meteorological stations provides a strong basis for conducting a spatially comprehensive flood risk analysis.
This study positions itself as one of the few regional-scale assessments in Southeastern Turkey that integrates both trend detection and frequency analysis using station-level data. Unlike previous studies that often apply a single model or focus on basin-specific hydrology, this work combines both the Generalized Extreme Value (GEV) and Peaks Over Threshold (POT) approaches, enabling a comparative evaluation. The analysis particularly underscores the advantages of the POT method for short time series and critically examines the limitations of GEV, especially when highly negative shape parameters yield unrealistic quantile estimates. By synthesizing trend and extreme value analyses across 13 stations, this study provides a more detailed understanding of flood potential from both temporal and statistical perspectives.

2. Materials and Methods

2.1. Study Area

Diyarbakır Province is situated along the foothills of the Southeastern Taurus Mountains and is characterized by highly variable topography. Among the districts, Kulp, Lice, and Hani are dominated by mountainous and forested terrain, whereas Bismil, Çınar, and Silvan primarily consist of alluvial plains and agricultural lands. The Tigris River and its tributaries form the principal fluvial network of the region and play a decisive role in local flood dynamics. The district of Dicle, which lies directly along the river, is particularly prone to flooding.
The regional climate is classified as semi-arid continental. Intense rainfall events frequently occur during the winter months, while in spring, rapid temperature increases coupled with snowmelt often lead to sudden changes in streamflow regimes. These conditions substantially elevate flood risk. In many areas, agricultural lands are located adjacent to stream channels, further amplifying the potential for economic losses during flood events.
Steep slopes and narrow valleys dominate upland districts such as Eğil and Dicle, whereas relatively flat terrain and wide plains prevail in Bismil and Çınar. These geomorphological contrasts influence both the direction of floodwater movement and the likelihood of water accumulation. In addition, anthropogenic factors such as settlement patterns, land use, drainage systems, and infrastructure quality significantly shape flood hazard across the province.
Elevation ranges from approximately 600 to 1200 m above sea level, depending on the district. In upland areas such as Kulp, Hani, and Lice, surface runoff is rapid due to steep gradients. Conversely, in low-lying districts like Bismil and Silvan, floodwaters are more likely to accumulate, thereby increasing inundation risk. Overall, the diverse topography, hydrological characteristics, and climatic conditions of Diyarbakır Province provide a suitable setting for conducting a comprehensive flood risk assessment. The availability of reliable rainfall data from 13 meteorological stations further strengthens the potential for a spatially detailed analysis of flood susceptibility.
This study encompasses 13 districts of Diyarbakır Province (Figure 1), located in the Southeastern Anatolia Region of Turkey: Bismil, Çermik, Çınar, Çüngüş, Dicle, Eğil, Ergani, Hazro, Hani, Kocaköy, Kulp, Lice, and Silvan. The study area lies within the Upper Tigris Basin and exhibits a diverse topographic structure, with mountainous terrain dominating the northern districts and relatively flat plains extending across the south. Elevation ranges from approximately 600 to 1200 m above sea level, strongly influencing hydrological behavior.
The regional climate is classified as semi-arid continental, with the majority of precipitation occurring during the winter and spring months. The Tigris River and its tributaries constitute the primary hydrological network, directly shaping flood dynamics across the basin. The districts display distinct geomorphological characteristics: Kulp, Lice, and Hani are situated in steep and elevated terrains prone to rapid runoff, while Bismil, Çınar, and Silvan consist largely of extensive agricultural lowlands that are highly susceptible to surface water accumulation and flooding.

2.2. Data and Analysis

Daily and monthly precipitation and evaporation data covering the period 2009–2023 were obtained from the Turkish State Meteorological Service. Evaporation records were available only for the Ergani station, which was considered representative of the Upper Tigris Basin due to its central location. Quality control procedures including range checks, consistency verification, and removal of suspicious values were applied to ensure data reliability.
Statistical analyses were performed in Python 3.11 using the pyMannKendall, Scipy.stats, and NumPy libraries, while pandas, matplotlib, and seaborn were employed for data management and visualization. Descriptive statistics were also calculated in Microsoft Excel 2019, which was used to prepare summary tables and charts. Spatial analyses, including rainfall and flood susceptibility mapping, were conducted in ArcMap 10.6.1 (ESRI, Redlands, CA, USA) using the Spatial Analyst toolbox (Fuzzy Membership, Weighted Overlay, and Hydrology Tools).
To assess temporal variability and sensitivity to extreme events, descriptive indicators such as mean (μ), standard deviation (σ), minimum and maximum values, coefficient of variation (Cv), skewness (Cs), and correlation coefficient (r) were calculated for each station.
These indicators provided insights into precipitation anomalies, irregular rainfall patterns, and potential flood triggers. For example, high coefficients of variation and positive skewness highlight irregular or extreme rainfall regimes that increase flood risk. The correlation coefficient (r) was used to evaluate the internal consistency between monthly and annual precipitation datasets across stations, ensuring robustness before applying trend and frequency analyses.
Trend analyses were conducted on both annual and seasonal series, with special emphasis on winter months due to their strong association with flood occurrence in the region. This approach allowed for the detection of seasonal changes in precipitation and a more reliable evaluation of flood potential.

2.3. Trend Analysis

To detect long-term changes in precipitation patterns, non-parametric statistical techniques were employed, including the Mann–Kendall (MK) trend test, Sen’s slope estimator, and Spearman’s Rho correlation analysis. These methods were selected because precipitation data often deviate from normal distribution and frequently include extreme values, which reduce the reliability of parametric tests.
The MK test was preferred since it does not require assumptions about data distribution and is effective in detecting monotonic trends. To complement this, Sen’s slope estimator was applied to quantify both the direction and magnitude of detected trends. Spearman’s Rho correlation coefficient was also used to assess non-linear relationships and strengthen the reliability of the trend analysis. Together, these approaches provided a robust evaluation by incorporating both median-based and rank-sensitive statistical measures.
A 95% confidence level (p < 0.05) was adopted as the threshold for statistical significance, which is standard in hydrometeorological studies.
Mann–Kendall Test: The MK test evaluates whether a time series exhibits a monotonic upward or downward trend. It does not require the data to be normally distributed or linear, but it assumes independence (no autocorrelation). The null hypothesis states that no trend exists, while the alternative hypothesis indicates the presence of either an upward or downward trend. In this study, the MK test was applied to both annual precipitation totals and seasonal maximum daily precipitation series. The test statistic (S) and standardized Z values were calculated as follows:
S =   i 1 n 1 j = i + 1 n s g n ( x j x i )
The Z statistic is:
Z =   s 1 v a r   ( S ) ,     S > 0     0 ,                         S = 0   s + 1 v a r   ( S ) ,     S < 0
  • S : The test statistic that determines the direction of the trend
  • x j ,   x i : Precipitation values in the time series.
  • n : Total number of observations (number of years).
  • s g n : Sign function of the difference between two data points.
  • V a r   S : Variance of the S statistic.
  • Z : Z-score indicating the statistical significance of the trend.
Positive Z values indicate increasing trends, while negative Z values indicate decreasing trends. Trends were considered statistically significant when p < 0.05.
Sen’s Slope Estimator: To measure the magnitude of trends identified by the MK test, Sen’s slope was applied. This estimator calculates the median rate of change (mm/year) between data pairs in the time series:
β = m e d y a n   x j x i j i                         ( 1 i < j n )
  • β : Annual slope coefficient (mm/year)
  • x j ,   x i : Precipitation values in the time series
  • j i : Time interval between years
  • M e d i a n : The median of all calculated slope values
  • n : Total number of years in the time series
Positive slopes represent increasing trends, whereas negative slopes represent decreasing trends.
Before applying the MK test, lag-1 autocorrelation coefficients were calculated for each precipitation series to verify the independence assumption. Most coefficients were weak (r1 < 0.3), indicating that serial correlation effects were minimal. Therefore, the MK results can be interpreted with confidence.
Spearman’s Rho Correlation Analysis: Spearman’s Rho is a rank-based correlation coefficient, suitable for identifying monotonic but non-linear trends. The coefficient was computed by ranking the raw data and applying the following formula:
p = 1 6 d i 2 n ( n 2 1 )
Positive and statistically significant correlations (p < 0.05) indicate increasing precipitation trends over time, while negative significant correlations suggest decreasing trends.

2.4. Seasonal Trend Analysis

In addition to the annual series, seasonal precipitation trends were analyzed for winter, spring, summer, and autumn using the Mann–Kendall test and Sen’s slope estimator. Particular emphasis was placed on the winter and spring seasons because of their stronger association with flood risk in the study area.
This seasonal analysis allowed for the identification of temporal variations that may not be visible in annual datasets. Winter precipitation is of particular concern, as intense rainfall combined with snowmelt often triggers rapid changes in streamflow and increases the likelihood of flooding. By applying non-parametric methods across different seasons, the analysis provides a more detailed understanding of the temporal distribution of extreme rainfall and its implications for flood susceptibility.

2.5. Extreme Value Analysis

The Generalized Extreme Value (GEV) distribution is widely applied to model block maxima of hydrological data and is defined by location, scale, and shape parameters that control the tail behavior of extremes. The Peaks Over Threshold (POT) approach, using the Generalized Pareto Distribution (GPD), models exceedances above a selected threshold and is advantageous for capturing rare, high-intensity events in shorter records. Both methods are standard in hydrological frequency analysis to estimate return levels and inform flood risk assessment.
Generalized Extreme Value (GEV) Distribution: The GEV distribution estimates the probability of extreme events based on annual maxima. For each station, the highest daily precipitation value for every year was extracted, and the GEV parameters were estimated:
  • Location (μ/GEV_loc): central tendency of the distribution.
  • Scale (σ/GEV_scale): degree of variability,
Shape (γ/GEV_shape): tail behavior, which defines the type of distribution. From these parameters, precipitation values corresponding to 10-, 25-, and 50-year return periods (GEV_Q10y, GEV_Q25y, GEV_Q50y) were calculated. These estimates provide projections of the intensity of rare extreme rainfall events.
Peaks Over Threshold (POT) Method: The POT method is an alternative extreme value analysis technique that evaluates flood risk based on daily precipitation events exceeding a statistically defined threshold. In this study, a threshold value (POT_threshold) was determined for each station, and all precipitation events above this threshold were included in the analysis. For stations with three or more exceedances (Excess_N ≥ 3), the Generalized Pareto Distribution (GPD) was applied to estimate return period values for 10, 25, and 50 years (POT_Q10y, POT_Q25y, POT_Q50y).
The POT method is particularly useful for short data records because it incorporates multiple extreme events instead of relying solely on annual maxima. This approach provides greater flexibility and can generate more realistic flood scenarios by capturing rare but high-impact precipitation events.
For this study, the threshold values were selected from the upper percentile of the daily precipitation series at each station, following common practice in extreme value analysis. This procedure ensured that a sufficient number of exceedances were included while maintaining statistical robustness.
Model Evaluation: The adequacy of the fitted GEV and POT models was tested using the Kolmogorov–Smirnov (K–S) and Anderson–Darling goodness-of-fit tests. In addition, the Akaike Information Criterion (AIC) was used for model comparison. Both approaches confirmed that the selected models provided an adequate fit to the observed extremes, though limitations were noted for certain stations where short records or outliers influenced the results.
Implementation: All analyses were conducted in Python. The pyMannKendall, scipy.stats, and NumPy libraries were used for statistical modeling, while pandas, matplotlib, and seaborn supported data management and visualization. Spatial analyses were performed in ArcMap, which was also used to map rainfall and flood susceptibility.

3. Results

This section presents the findings obtained from descriptive statistics, trend analyses, and seasonal variation assessments of precipitation data collected from the selected meteorological stations. The key characteristics of each station are first summarized, followed by tables reporting monthly and daily average precipitation values along with their minimum, maximum, and standard deviation metrics. Subsequently, non-parametric statistical tests, including the Mann–Kendall and Spearman’s Rho, were applied to detect significant trends. For stations exhibiting noteworthy results, time-series graphs of maximum daily precipitation were produced to illustrate the temporal evolution of extreme events. In addition, seasonal analyses were performed to evaluate the concentration and intensification of extreme rainfall episodes within specific periods of the year.
Table 1 summarizes annual precipitation statistics across the 13 stations, showing clear spatial variability. Mean annual precipitation ranges from 355.7 mm at Bismil to 779.0 mm at Çüngüş. Çüngüş, Kulp and Lice record both high mean values and large standard deviations, reflecting irregular rainfall and higher flood susceptibility. Bismil and Silvan show lower and more stable totals. Coefficients of variation (Cv) above 0.30 at Çüngüş, Eğil and Kocaköy also indicate greater interannual variability. Most stations have positive skewness, suggesting occasional extreme events; negative skewness at Silvan and Hazro points to more balanced but variable rainfall. Correlation coefficients (r) are mostly weak but upward; Bismil is the only site with a statistically significant positive trend.
To evaluate whether the differences in annual precipitation among the stations are statistically significant, the non-parametric Kruskal–Wallis H-test was applied to the data series. The results (Table 2) indicate statistically significant spatial differences in precipitation values between some stations (p < 0.05). These differences suggest that local topographic, climatic, orographic, and land use factors may influence precipitation distribution across the region.
Table 3 presents mean monthly precipitation values and variability across the 13 stations, showing pronounced spatial differences. Çüngüş (779 mm) and Kulp (764 mm) have the highest monthly means and very large standard deviations, indicating irregular rainfall and higher flood potential. Lice and Hani also show elevated precipitation, while Bismil (355 mm) remains lowest and more stable (SD = 73 mm). Stations such as Hani, Kocaköy and Kulp display SD values above 200 mm, reflecting abrupt changes and intense rainfall episodes. High minimum precipitation at Lice, Kulp and Çüngüş implies persistent runoff potential, whereas Hazro, Çınar and Bismil show lower and more uneven rainfall distribution through the seasons.
As shown in Table 4, to evaluate the intra-annual distribution and concentration of precipitation, two indices were employed: the Concentration Index (CI) and the Intra-Annual Coefficient of Variation (IACV). Due to the unavailability of continuous monthly datasets for all years, representative monthly maximum values were used to construct a seasonal profile for each station. Although this approach does not capture inter-annual variability, it provides a valuable first-order assessment of seasonal rainfall patterns. The CI and IACV were selected as widely applied indicators for evaluating precipitation distribution and variability. The CI is particularly suitable for representing rainfall seasonality in cases of short data records or limited station coverage, whereas the IACV provides a simple yet effective measure of intra-annual variability. These indices were preferred in this study due to their simplicity, interpretability, and broad acceptance in climatological and hydrological research.
The CI values ranged from 0.51 to 0.64, indicating a moderate to high concentration of precipitation in certain months. Stations such as Kulp, Çüngüş, and Lice exhibited the highest CI values (>0.60), suggesting that a significant proportion of annual rainfall occurs within a few months, typically in late winter and early spring. Similarly, IACV values were relatively high at all stations, exceeding 45%. This suggests substantial month-to-month variability, with rainfall peaks occurring abruptly rather than being evenly distributed throughout the year. These findings reinforce the role of seasonal concentration in flood risk, especially in regions such as Kulp and Hani, where both CI and IACV values were elevated. Such regions may be particularly susceptible to flash floods and surface runoff events due to the clustered nature of precipitation.
Table 5 summarizes daily precipitation statistics across the 13 stations, showing strong contrasts in rainfall intensity and variability. Çermik, Lice, Ergani and Kulp record the highest daily means (>580 mm), indicating frequent rapid runoff potential, while Çınar and Bismil remain lowest and relatively stable. Maximum single-day rainfall peaks above 1000 mm at Çüngüş, Çermik, Lice, Hani and Dicle confirm susceptibility to intense hydrometeorological hazards. Many stations also have days with no rainfall, highlighting irregular annual distribution. Standard deviation values are highest at Çüngüş (≈480 mm), Kulp (≈450 mm) and Hani (≈410 mm), reflecting highly erratic events that can overwhelm drainage and infiltration systems, whereas Bismil and Ergani show comparatively lower variability and steadier daily patterns.
As shown in Figure 2, both monthly and daily average precipitation values exhibit a generally increasing trend across the study area, with spatial variability among stations. Çüngüş, Kulp, and Lice display the highest precipitation values in both metrics, highlighting their susceptibility to extreme rainfall events. Conversely, Bismil and Çınar record the lowest averages, yet Bismil’s statistically significant upward trend underscores its latent flood hazard due to low-slope topography and accumulation potential. The pronounced gap between monthly and daily values at most stations suggests irregular rainfall regimes characterized by short-duration, high-intensity events, which further elevates the risk of surface runoff and flash floods in susceptible areas.
The data presented in Table 6 reflect the monthly average evaporation values for the Ergani district during the period 2009–2023, along with the corresponding minimum, maximum, and standard deviation values. Notably, the average monthly evaporation observed at the Ergani station was remarkably high, at 1019.22 mm. This elevated value suggests that evaporation constitutes a dominant component of the regional water cycle, indicating substantial water loss, particularly during the summer months. Regardless of precipitation intensity, such high evaporation rates can lead to reduced surface water retention and a rapid decline in soil moisture. Consequently, this may result in hydrological imbalances, with potential deficits in the water budget and an increased likelihood of surface runoff and flooding due to insufficient soil saturation during subsequent rainfall events. The minimum evaporation value of 0.50 mm likely corresponds to the winter months, when evaporation is significantly limited, implying a higher potential for surface water retention during this period. This seasonal variation highlights the increased flood potential of winter precipitation, as limited evaporative loss during colder months may lead to soil saturation and rapid surface runoff. By contrast, the maximum evaporation value, recorded at an extreme 3874.20 mm, points to periods of intense evaporation likely driven by exceptionally high temperatures, low humidity, and strong winds. The very high standard deviation (SD = 1514.11 mm) further indicates substantial intra-annual variability in evaporation rates.
Before applying the Mann–Kendall test, lag-1 autocorrelation coefficients (r1) were calculated for each station to verify the independence assumption of the test (Table 7).
Lag-1 autocorrelation coefficients (r1) were calculated for each station based on maximum monthly precipitation time series. These values help assess the independence assumption required by the Mann–Kendall trend test. The interpretation column indicates whether the autocorrelation is likely to affect the reliability of trend detection. As shown in Table 7, the majority of the stations exhibited low or negligible lag-1 autocorrelation (r1 < 0.3), indicating that the independence assumption of the Mann–Kendall test is generally satisfied. Only the Bismil station showed a high autocorrelation coefficient (r1 = 0.767), suggesting that the detected trend at this station should be interpreted with caution. Overall, the results support the statistical validity of the applied trend analysis across most stations.
The findings presented in Table 8 summarize the results of the Mann–Kendall (MK) trend analysis applied to precipitation data observed at 13 stations during the period 2009–2023. In the analysis, the direction and statistical significance of precipitation trends were evaluated for each station based on the calculated Z-value, p-value, and Theil–Sen slope.
One of the key outputs of the trend analysis, the Z-value, is used to determine the direction of the trend; positive values indicate an increasing trend, while negative values indicate a decreasing trend. As shown in the table, all stations exhibit positive Z-values, suggesting an overall increasing trend in precipitation across the region. However, the primary criterion for determining whether these trends are statistically significant is the p-value.
The Bismil station is the only location that exhibits a statistically significant increasing trend at the 95% confidence level, with a Z-value of 2.4406 and a p-value of 0.0147. This indicates that precipitation in the Bismil region has shown a consistent upward trend over the years, and that this increase is not random but reflects a meaningful pattern. The Theil–Sen slope was calculated to be approximately 38.31 mm/year, suggesting an average annual increase of around 38 mm in precipitation. Such an increase may contribute to higher flood risk by affecting soil saturation levels and surface runoff dynamics.
Although the Z-values for all other stations are positive, the corresponding p-values exceed 0.05, indicating that the observed trends are not statistically significant. For example, stations such as Çermik (Z = 0.9342, p = 0.3502) and Dicle (Z = 1.2015, p = 0.2296) exhibit relatively high slope values (47.3 and 78.96 mm/year, respectively) and positive Z-scores, yet these trends do not meet the threshold for statistical significance. This suggests that, while a potential upward trend in precipitation may be present in these regions, the data series does not provide sufficient statistical evidence to confirm its existence. In particular, stations such as Çüngüş (slope = 110.51 mm/year) and Kocaköy (slope = 91.1 mm/year) reveal notably high slope values; however, their p-values exceeding 0.05 imply that such increases may result from short-term fluctuations or extreme events rather than from a consistent trend. Therefore, unless supported by long-term monitoring, these patterns cannot be considered meaningful trends for informing policy or planning efforts.
The results of the Mann–Kendall (MK) trend analysis indicate that all stations included in the study exhibit upward trends in precipitation. However, among these, only the increasing trend observed at the Bismil station is statistically significant. For the remaining stations, the detected trends likely fall within the bounds of random variability. Therefore, in areas such as Bismil, where the trend has been found to be statistically significant, the increasing precipitation pattern should be considered in relation to potential hydrological risks, including flooding, soil saturation, and surface runoff.
Table 9 presents the results of Spearman’s Rho rank correlation test applied to 13 stations within the scope of the study. This test is used to detect the presence of a monotonic trend by considering the ordinal structure of changes over time. The obtained Rho coefficient indicates the direction of the trend (positive or negative), while the p-value determines its statistical significance. According to Spearman’s Rho analysis, a statistically significant increasing trend was identified only at the Bismil station (Rho = 0.7248, p = 0.0116). For all other stations, although the Rho coefficients are positive, the p-values exceed 0.05, indicating that the observed trends are not statistically significant. For instance, stations such as Dicle (Rho = 0.5714, p = 0.1802) and Hani (Rho = 0.5714, p = 0.1802) reveal moderate positive correlations, yet fail to reach statistical significance. This suggests that, while there may be a potential increasing trend in these regions, the available data do not provide sufficient statistical confidence to confirm it. At stations such as Ergani (Rho = 0.0818, p = 0.811), the correlation coefficient is very low, indicating no apparent trend in precipitation. Similarly, at stations such as Çınar, Kulp, Lice, and Silvan, the high p-values imply that the observed variations in precipitation over time are likely due to random fluctuations. Therefore, overall, the results of Spearman’s Rho test reveal that a statistically significant increasing trend exists only at the Bismil station, while the trends observed at other stations are not statistically meaningful.
Figure 3 illustrates the spatial distribution of trend analysis results based on the Mann–Kendall (MK) and Spearman’s Rho (SR) tests across the study area. The left map indicates the Z-values derived from the MK test, while the right map presents the corresponding results from the SR test. In both maps, the Bismil district appears as a hotspot with the highest Z-values, indicating a statistically significant increasing trend in precipitation. Other districts exhibit mostly low to moderate Z-values, implying non-significant or weak trends. The spatial patterns in both tests are broadly consistent, reinforcing the reliability of the findings. These maps visually highlight areas where precipitation trends may contribute to hydrological risks and where further investigation or mitigation planning may be necessary.
Table 10 presents the trend analysis of maximum daily precipitation during the winter season for the stations included in the study and evaluates the statistical significance of these trends using the Mann–Kendall test. This analysis plays a critical role in identifying seasonal changes in precipitation and assessing whether flood potential has increased, particularly in the winter months. The analysis is limited to the winter season due to data insufficiency in other seasons. Notably, statistically significant increasing trends in maximum daily winter precipitation were identified at the Bismil (Z = 3.093; p = 0.002) and Çermik (Z = 2.18; p = 0.0293) stations. The Theil–Sen slope was calculated as 12.625 mm/year for Bismil and 19.58 mm/year for Çermik. These findings indicate that extreme rainfall events during the winter season are becoming more intense at these two stations, highlighting them as areas of concern in terms of rising flood risk. The particularly high Z-value at Bismil suggests a very strong increasing trend.
Figure 4 illustrates the annual maximum monthly precipitation values observed in Bismil between 2009 and 2023, along with the corresponding trend line. The upward slope of the trend line indicates a general increase in extreme monthly precipitation events over the analyzed period, supporting the results of the Mann–Kendall and Spearman’s Rho analyses, which revealed a statistically significant upward trend in the region. The error bars shown in Figure 4 represent the standard deviations of annual maximum monthly precipitation values, indicating the degree of interannual variability and uncertainty in the observed data.
Figure 5 displays the annual maximum monthly precipitation values for Çermik between 2009 and 2023, along with the fitted trend line. The upward trend line suggests a potential increase in extreme monthly precipitation events during this period; however, the statistical significance of this trend remains limited, as supported by the Mann–Kendall and Spearman’s Rho analyses. The vertical error bars in Figure 5 represent the standard deviations of maximum monthly precipitation values for each year, providing insight into the variability and uncertainty of extreme precipitation events in Çermik between 2009 and 2023.
Although the Z-values are positive for all other stations, the p-values exceed 0.05, indicating that the observed trends are not statistically significant. For instance, stations such as Çınar, Çüngüş, Dicle, and Eğil display Z-values close to 1.8 and relatively high slope values, yet their p-values were found to be 0.0715, just above the threshold for statistical significance. By contrast, the Ergani station shows a negative trend (Z = −0.156; p = 0.8763; slope = −3.45 mm/year). This result suggests that there is no significant change in the maximum daily precipitation at this station, and while winter precipitation may indicate a decreasing tendency, the trend is weak and statistically insignificant. Similarly, at stations such as Silvan, Hani, Hazro, Kocaköy, Kulp, and Lice, the p-values were considerably high, and the detected trends did not reach statistical significance.
Before applying the extreme value models, the goodness-of-fit was assessed using the Kolmogorov–Smirnov (K–S) test and Akaike Information Criterion (AIC). High K–S p-values and relatively low AIC values indicate that both the GEV and GPD models adequately fit the precipitation data (Table 11).
GEV_shape, GEV_loc, and GEV_scale represent the shape, location, and scale parameters of the Generalized Extreme Value distribution, respectively. GEV_Q10y, GEV_Q25y, and GEV_Q50y indicate the estimated precipitation values for 10-, 25-, and 50-year return periods.
Table 12 presents the results of the flood risk analysis based on the Generalized Extreme Value (GEV) distribution for the stations included in the study. At many stations, an increase in the return period corresponds to higher expected maximum precipitation values. This trend theoretically confirms that the GEV model effectively represents extreme events. For instance, at the Bismil station, the flood quantile for a 10-year return period (Q10y) is 118.53 mm, increasing to 124.47 mm for 25 years (Q25y) and 127.08 mm for 50 years (Q50y). Similarly, at the Çınar station, Q10y is calculated as 134.05 mm, Q25y as 143.96 mm, and Q50y as 149.38 mm. At these stations, the GEV shape parameter is positive, indicating that flood magnitudes grow progressively with increasing return periods in a controlled manner.
By contrast, stations such as Kulp (γ = 1.1044), Hani (γ = 0.3977), and Dicle (γ = 0.263) also exhibit a consistent increase in flood quantiles with longer return periods. For example, at the Hani station, Q10y is 290.93 mm, Q25y is 320.78 mm, and Q50y reaches 336.69 mm. These increases are both moderate and practically meaningful in engineering terms. The GEV model appears to perform well at these stations, with the data fitting the distribution adequately. However, at certain stations, flood estimates derived from the GEV model far exceed realistic thresholds. For instance, at the Çermik station, Q10y is 85.519 mm, but Q25y rises to 17,241,683 mm and Q50y to 884,000,000 mm. Similarly, Q50y estimates at Eğil, Hazro, and Kocaköy reach extreme values of approximately 3.4 billion mm, 7.29 billion mm, and 105 billion mm, respectively. Such anomalies are typically caused by highly negative shape parameters. Indeed, the shape parameters were calculated as γ = –5.5966 for Çermik, γ = –5.9947 for Hazro, and γ = –6.7487 for Kocaköy. These results indicate that, at these stations, the GEV model fails to produce reliable estimates, likely due to factors such as limited data, the influence of outliers, or poor distribution fit. Consequently, the GEV distribution is not considered a suitable method for flood frequency analysis at these locations.
The failure of the GEV model at these stations is closely related to the highly negative shape parameters (γ < 0), which indicate a bounded distribution with a finite upper limit. In such cases, even slight deviations or outliers in the annual maximum rainfall series can cause the model to drastically overestimate flood quantiles, as observed in Çermik, Eğil, Hazro, and Kocaköy. The negative shape parameter essentially forces the distribution to become heavily right-skewed, and in short time series, the tail behavior becomes extremely sensitive to rare extremes. Therefore, under these circumstances, the GEV distribution is not a suitable method for flood frequency analysis. This limitation is well documented in the literature, especially for small datasets or regions with abrupt topographic or climatic changes.
In certain seasons, the GEV model yielded shape parameter values and return levels that deviated considerably from expected hydrological behavior. This aberrant performance may be attributed to the limited sample size within seasonal subsets, the high variability of extreme precipitation events, or the influence of outlier years. Such factors can lead to unstable parameter estimation and inflated return levels. Therefore, seasonal results with extreme parameter values should be interpreted with caution, even though the overall model fit, as confirmed by the K–S test and AIC values, remains satisfactory.
Table 13 presents the results of the analysis based on the Peak Over Threshold (POT) method for the stations included in the study. The analysis utilized extreme precipitation observations that exceeded station-specific threshold values (ranging between 50% and 70%), with at least three exceedance events used per station to estimate flood quantiles. One of the most notable outcomes of this method was observed at the Bismil and Silvan stations. At Bismil, using a 70 mm threshold and only three extreme rainfall events, the POT analysis yielded a particularly high flood estimate for the 50-year return period (710.6 mm). In contrast, the estimates for the 10- and 25-year return periods were more reasonable, calculated as 104.37 mm and 108.8 mm, respectively. Considering that Bismil is a region occasionally affected by sudden summer downpours, the 10- and 25-year return levels appear consistent with the climatic reality of the area. The extremely high 50-year projection, however, may be interpreted as an upper-bound scenario that could be useful for scenario-based planning and risk management of extreme events.
For the stations of Çermik, Çınar, Eğil, Ergani, Hani, Hazro, Kocaköy, Kulp, and Lice, the POT analysis yielded identical precipitation estimates for the 10-, 25-, and 50-year return periods. For instance, all three return periods resulted in 347.2 mm for Çermik, 143.5 mm for Çınar, and 316.5 mm for Kulp. This outcome indicates a lack of variation, suggesting that the POT model did not capture any increase in precipitation magnitude with increasing return periods at these stations.
The Çüngüş station stands out with a value of 499.9 mm, which remains constant across all return periods. Despite data limitations, this estimate suggests a high flood potential. Considering the orographic structure of the Çüngüş area and its tendency to experience sudden rainfall events, this projection appears consistent with the region’s natural conditions. Similarly, the estimates obtained for the Dicle station, ranging between 205.5 and 205.6 mm, align with previously recorded flood events in the area, indicating that the model outputs may reasonably agree with real-world observations.
The Silvan station exhibits a highly variable pattern in terms of potential flood magnitude. While the 10-year return period shows a relatively low estimate of 149.91 mm, the values increase significantly to 179.2 mm for the 25-year period and 554.2 mm for the 50-year period. This indicates that extreme rainfall events in Silvan are rare but impactful, and that more severe flood scenarios may be possible over longer return periods. The high values projected by the model, particularly for long-term scenarios, appear consistent with the region’s geographical and meteorological characteristics.
To validate the reliability of the extreme value models used in this study, two statistical methods were applied: the Kolmogorov–Smirnov (K–S) test and the Akaike Information Criterion (AIC). These metrics assess how well the Generalized Extreme Value (GEV) distribution and the Peak Over Threshold (POT) method fit the observed or derived precipitation values.

4. Discussion

This study contributes to the growing body of research on regional-scale flood risk in the context of changing precipitation regimes. By applying non-parametric trend detection methods (Mann–Kendall and Spearman’s Rho) alongside extreme value analyses (GEV and POT), it provides a comprehensive assessment of seasonal and annual rainfall trends in the Diyarbakır Province of southeastern Turkey. These methods have proven effective in previous hydrometeorological studies, and their combined use strengthens the reliability of the findings, particularly in regions with limited discharge data.
The identification of statistically significant increasing trends at the Bismil and Çermik stations signals a heightened seasonal flood risk in these areas. This is consistent with findings from other semi-arid and topographically variable regions, where even modest shifts in precipitation intensity can disproportionately increase flood susceptibility. In particular, districts such as Lice, Kulp, and Çüngüş, despite not showing statistically significant trends, exhibited elevated maximum daily rainfall values, suggesting that the absence of significance may be due to data length limitations rather than a lack of risk.
A key strength of this study lies in its multi-station approach, which allows for the evaluation of spatial heterogeneity in flood drivers within a single province. This regional focus provides critical insights for localized risk reduction strategies and can inform early warning systems, infrastructure planning, and watershed-based climate adaptation policies. Moreover, by integrating both frequency- and threshold-based extreme value models, the study captures a wider range of hydrological behavior than methods relying solely on annual maxima. A key limitation of this study is the relatively short length of the data series (2009–2023) and the limited spatial coverage of evaporation records, which may affect the robustness and generalizability of the results.
Despite these contributions, several limitations must be acknowledged. The relatively short observation period (2009–2023) may not adequately capture long-term climatic shifts, and the absence of hydrological validation through streamflow or flood event data limits model calibration. Additionally, the exclusion of land use change, soil characteristics, and infiltration capacity data may lead to the underestimation or overestimation of runoff generation in certain districts. Future research should consider integrating remote sensing-derived flood extents, land surface models, or hydrodynamic simulations to enhance predictive capabilities.
The incorporation of spatial significance testing and seasonality metrics, such as CI and IACV, has improved the robustness of regional precipitation interpretation and flood risk assessment.
Overall, the findings emphasize the importance of high-resolution, station-based analyses for understanding emerging flood patterns in climate-sensitive and infrastructure-vulnerable regions such as southeastern Turkey.
The variability observed in the time series and the increase in extreme precipitation events can have significant impacts on local infrastructure and vulnerable communities. For instance, inadequate drainage systems and unplanned urbanization in regions such as Çermik may exacerbate flood risks. In particular, sensitive groups living in low-lying areas or informal settlements may be disproportionately affected by floods. Therefore, adaptation strategies such as improving urban planning, strengthening critical infrastructure, and developing early warning systems are of vital importance for reducing flood risks.
Our results align with regional evidence from semi-arid Mediterranean and Middle Eastern basins. Over the Iberian sector of the Mediterranean, daily rainfall often concentrates into a limited number of events, as captured by concentration indices [42]. Similarly, Algeria in the southern Mediterranean exhibits high daily precipitation concentration, indicating that a few rainy days contribute a large share of annual totals [43]. Across the broader Mediterranean, daily concentration and its variability have been shown to be modulated by atmospheric teleconnections [44]. In the Middle East, multi-decadal PCI analyses in the Levant (1970–2018) reveal spatial heterogeneity and locally increasing annual concentration, consistent with our CI/IACV signals [45]. Moreover, regional reviews emphasize seasonally varying precipitation changes and extremes in the Eastern Mediterranean, supporting our interpretation that concentrated rainfall enhances flood susceptibility in semi-arid settings [46].

5. Preventive Measures

Flood control infrastructure should be strengthened in stations with statistically significant increasing trends, such as Bismil and Çermik. The increasing precipitation trends in these areas may lead to severe flooding, particularly when surface runoff exceeds channel capacity. Therefore, stormwater drainage systems should be expanded, the proportion of permeable surfaces in urban areas should be increased (e.g., green roofs, permeable pavements), and development within riverbeds must be strictly prohibited.
In regions with high variability in maximum daily precipitation (e.g., Çüngüş: 1362 mm/day, Lice: 1117 mm/day), the establishment of early warning systems is essential. Automated weather stations should be deployed more densely, and real-time data should be used to update flood risk maps. Local disaster preparedness plans should be revised in accordance with projected increases in precipitation, and public alert systems (e.g., SMS or mobile application notifications) should be developed to improve awareness and preparedness.
In low-slope areas with high water accumulation potential, particularly Bismil, Silvan, and Çınar, agricultural drainage systems and flood levees should be established. Controlled drainage should be implemented in farmlands, and both crop types and planting schedules should be reconsidered to reduce hazard to flooding.

6. Conclusions

According to the results of the Mann–Kendall and Spearman’s Rho tests conducted in this study, statistically significant positive precipitation trends were identified particularly at the Bismil and Çermik stations. The Mann–Kendall analysis revealed an annual increasing trend of approximately 38.31 mm/year at the Bismil station (Z = 2.4406; p = 0.0147), while for the winter season specifically, this value was calculated as Z = 3.093 (p = 0.002), with a Theil–Sen slope of 12.625 mm/year. Similarly, at the Çermik station, a significant increasing trend of 19.58 mm/year was observed in the winter months (Z = 2.18; p = 0.0293).
These findings indicate that extreme precipitation events are increasing in both frequency and intensity in the identified regions, particularly in Bismil and Çermik. As a result, surface runoff is likely to intensify, and the risk of flooding is projected to rise in the future. In particular, the wide and low-sloped plain topography of Bismil creates favorable conditions for water accumulation, making the area more vulnerable to flood events. Furthermore, Spearman’S Rho analysis supports this upward trend for Bismil, with a correlation coefficient of Rho = 0.7248 and a p-value of 0.0116, reinforcing the significance of the observed pattern. In the remaining 11 stations, the observed trends were not statistically significant. For instance, although stations such as Çüngüş (Theil–Sen slope = 110.5 mm/year) and Kocaköy (91.1 mm/year) exhibit relatively high positive slope values, the corresponding p-values exceed 0.05, and thus these trends cannot be considered significant. This outcome can be attributed to the limited length of the data series (11 years for most stations), seasonal variability, and regional topographic influences. Additionally, the location of some stations in high-altitude, steep-sloped areas (e.g., Kulp, Lice, Hani) may lead to rapid surface runoff into drainage systems, preventing the accumulation effects typically associated with increasing precipitation trends.
In the winter season analysis, the Ergani station exhibited a negative trend (Z = −0.156; slope = −3.45 mm/year); however, this trend was not statistically significant. This may be attributed to various environmental factors such as local microclimatic conditions, wind effects, and high evaporation rates. Within the scope of the study, evaporation data were available only for the Ergani station. The average monthly evaporation was recorded as 1019.22 mm, with a maximum value reaching an extreme of 3874.20 mm. These high values indicate a significant reduction in soil moisture, particularly during the summer months, suggesting that surface conditions become considerably dry. Evaporation processes may potentially contribute to a reduction in surface water, thereby playing a short-term mitigating role in flood risk. However, the data indicate that the intensity and maximum daily values of precipitation—such as the extreme daily rainfall recorded at Çüngüş (1362.10 mm/day)—far exceed the influence of evaporation. This suggests that soil saturation is reached rapidly, resulting in excess water being quickly converted into surface runoff. In other words, the severity of rainfall outweighs the mitigating effects of evaporation, thereby enhancing flood potential across the region.
The research findings clearly indicate a statistically significant increase in extreme precipitation trends, particularly in the Bismil and Çermik regions. If the current trends persist, this suggests a likely rise in both the frequency and severity of seasonal flood events in the future. Considering the annual slope values obtained (e.g., Bismil: 38 mm/year; Çermik: 19.5 mm/year), it is estimated that by 2030 total precipitation may increase by approximately 10–20% at these stations. This could lead to heightened risks related to urbanization, agricultural areas, and drainage infrastructure. Moreover, in regions such as Çüngüş, Hani, and Lice, where maximum daily precipitation values are already intense, the current p-values may not yet indicate statistical significance; however, if the observed trends continue, the potential for future flood events is likely to increase in these areas as well. By contrast, the Bismil and Çermik stations already exhibit statistically significant increases in precipitation, underscoring their heightened hazard to extreme rainfall events. The GEV and POT methods employed in this study enabled the estimation of flood magnitudes based on station-specific precipitation time series. In particular, the consistency and plausibility of both models’ outputs for stations such as Bismil, Çınar, Hani, and Kulp indicate that reliable future flood scenarios can be modeled for these regions. Furthermore, the alignment between the increasing flood projections derived from the POT method and the Mann–Kendall trend results, especially in stations like Bismil and Silvan, demonstrates the sensitivity and applicability of the model to local conditions. Although the GEV distribution produced extreme values at some stations, this highlights the method’s parametric sensitivity and the critical importance of data length and quality. In this context, the establishment of early warning systems, the modeling of precipitation–soil moisture interactions, and the implementation of advanced projection analyses based on climate change scenarios are of great importance for effective flood risk management. In this study, the Mann–Kendall trend analyses conducted for stations in the Diyarbakır region revealed statistically significant increases in extreme precipitation events, particularly at the Bismil and Çermik stations. Similar trends have also been reported in various geographical regions in the international literature. For instance, POT-based analyses conducted on the Iberian Peninsula identified positive trends in annual maximum precipitation at certain locations and noted that the POT method produced more stable results than the GEV model, particularly in short data series [47]. In a GEV-based study involving 27 stations in Senegal, increasing trends were observed at only five stations, while decreasing trends in extreme precipitation were found at most locations [48]. In South Korea, Wi et al. [49] conducted a comprehensive analysis comparing the Mann–Kendall test with both POT and GEV methods, reporting that the results varied across regions and that the GEV model was particularly sensitive to the distribution of extreme values. In this context, the failure of the GEV model to produce realistic results at certain stations in the present study (e.g., Eğil, Kocaköy, Hazro) is consistent with similar concerns noted in the international literature. By contrast, the more stable performance of the POT method in short-term series and threshold-based evaluations provided a more realistic modeling of flood risk in areas such as Bismil and Silvan.
Beyond statistical trend detection, it is crucial to recognize the interactions between precipitation dynamics and topographic and land use characteristics, particularly in the context of flood risk management. In regions such as Bismil and Çermik, steep slopes, impervious urban surfaces, and altered land-cover patterns can exacerbate runoff generation in response to extreme precipitation events. Therefore, an integrated multivariate approach that combines precipitation trends with geospatial variables (e.g., slope, drainage density, land use, and soil type) is essential for accurately assessing spatial flood susceptibility. The findings of this study underscore the importance of incorporating such multivariate analyses into territorial planning. This would enhance the predictive power of future flood risk assessments and support the development of actionable strategies, including infrastructure reinforcement, land use regulation, and early warning systems, particularly in vulnerable districts of the Upper Tigris Basin.
The findings of this study are applicable to local and regional flood risk management, infrastructure planning, and climate adaptation strategies, particularly in semi-arid and hydrologically sensitive areas. What distinguishes this work is its regional-scale, multi-station approach combined with both trend and extreme value analyses, offering a more detailed and scalable framework for assessing flood hazard in data-limited environments.
Despite the valuable insights provided by this study, several limitations should be acknowledged. These include the relatively short observation period, the lack of evaporation data for most stations, and inconsistencies in the performance of the GEV model at certain locations. Addressing these limitations in future studies will enhance the robustness and applicability of the findings.
In addition to advancing the understanding of extreme precipitation dynamics and flood risk in semi-arid regions, this study provides critical insights for sustainable flood management, resilient infrastructure planning, and climate adaptation policies. By linking hydrological risk assessment with sustainability science, the findings contribute to long-term strategies that support disaster risk reduction and sustainable urban development in vulnerable regions.

Author Contributions

All authors contributed to this study. Material preparation and data collection were performed by R.Ç. The final draft of the manuscript was written by B.K. This article has been produced from B.K.’s ongoing doctoral dissertation. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data are contained within the article.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
GEVGeneralized Extreme Value
POTPeaks Over Threshold
GISGeographic Information Systems
SDStandard Deviation
HECHydrologic Engineering Center
MKMann–Kendall
HMSHydrologic Modeling System
RASRiver Analysis System
AHPAnalytic Hierarchy Process

References

  1. Jonkman, S.N. Global perspectives on loss of human life caused by floods. Nat. Hazards 2005, 34, 151–175. [Google Scholar] [CrossRef]
  2. UNDRR. Global Assessment Report on Disaster Risk Reduction; United Nations Office for Disaster Risk Reduction: Geneva, Switzerland, 2022; Available online: https://www.undrr.org/gar2022 (accessed on 1 September 2025).
  3. Paprotny, D.; Sebastian, A.; Morales-Nápoles, O.; Jonkman, S.N. Trends in flood losses in Europe over the past 150 years. Nat. Commun. 2018, 9, 1985. [Google Scholar] [CrossRef]
  4. Blöschl, G.; Hall, J.; Parajka, J.; Perdigão, R.A.P.; Merz, B.; Arheimer, B.; Aronica, G.T.; Bilibashi, A.; Bonacci, O.; Borga, M.; et al. Changing climate shifts timing of European floods. Science 2017, 357, 588–590. [Google Scholar] [CrossRef]
  5. Rahmati, O.; Falah, F.; Dayal, K.S.; Deo, R.C.; Mohammadi, F.; Naghibi, S.A.; Melesse, A.M.; Tien Bui, D. Flood risk assessment using machine learning and GIS: A global systematic review. Water 2022, 14, 620. [Google Scholar] [CrossRef]
  6. AlRifai, M.H.; Kafy, A.A.; Altuwaijri, H.A. Quantitative assessment of flood risk through multi parameter morphometric analysis and GeoAI: A GIS-based study of Wadi Ranuna Basin in Saudi Arabia. Water 2025, 17, 2108. [Google Scholar] [CrossRef]
  7. Durlević, U. Assessment of torrential flood and landslide susceptibility of terrain: Case study—Mlava River Basin (Serbia). Glas. Srp. Geogr. Drust. 2021, 101, 49–75. [Google Scholar] [CrossRef]
  8. Yaylak, M.M. Flood Risk Analysis of Bitlis Stream Using Geographic Information Systems (GIS). Master’s Thesis, Council of Higher Education Thesis Center, Bitlis Eren University, Bitlis, Turkey, 2016. [Google Scholar]
  9. Beyazıt, M.; Önöz, B. Flood and Drought Hydrology; Nobel Publishing: Ankara, Turkey, 2008. [Google Scholar]
  10. Baykal, T.; Şener, E.; Terzi, Ö. Application of analytic hierarchy process in flood risk analysis: A case study of the Küçük Aksu Basin (Antalya, Turkey). Iran J. Sci. Technol. Trans. Civ. Eng. 2023, 47, 2449–2466. [Google Scholar] [CrossRef]
  11. Li, W.; Wang, C.; Mo, J.; Hou, S.; Dang, X.; Shi, H.; Gong, Y. Compound Flood Risk Assessment of Extreme Rainfall and High River Water Level. Water 2025, 17, 841. [Google Scholar] [CrossRef]
  12. Cortiços, N.D.; Duarte, C.C. Climate Resilience and Adaptive Strategies for Flood Mitigation: The Valencia Paradigm. Sustainability 2025, 17, 4980. [Google Scholar] [CrossRef]
  13. De Moel, H.; Van Alphen, J.; Aerts, J.C.J.H. Flood maps in Europe: Methods, availability and use. Nat. Hazards Earth Syst. Sci. 2009, 9, 289–301. [Google Scholar] [CrossRef]
  14. Grünthal, G.; Thieken, A.H.; Schwarz, J.; Radtke, K.S.; Smolka, A.; Merz, B. Comparative risk assessments for the city of Cologne: Storms, floods, earthquakes. Nat. Hazards 2006, 38, 21–44. [Google Scholar] [CrossRef]
  15. Geyikli, M.S. Determining Flood Risk Maps Using GIS Software. Master’s Thesis, İnönü University, Graduate School, Malatya, Turkey, 2015. [Google Scholar]
  16. Shirzad, M.R. Evaluation of Flood Risk Using Remote Sensing and GIS Technologies: A Case Study of Kocaeli Maden Stream. Master’s Thesis, Kocaeli University, Kocaeli, Turkey, 2017. [Google Scholar]
  17. Alfieri, L.; Bisselink, B.; Dottori, F.; Naumann, G.; de Roo, A.; Salamon, P.; Feyen, L. Global projections of river flood risk in a warmer world. Earth’s Future 2017, 5, 171–182. [Google Scholar] [CrossRef]
  18. Demir, V.; Keskin, A.Ü. Evaluation of the economic damages of floods: A case study of the Samsun–Mert River Basin. Int. J. Eng. Res. Dev. 2022, 14, 663–678. [Google Scholar] [CrossRef]
  19. Onuşluel Gül, G.; Gül, A. Determining flood zones in the Bostanlı Basin of İzmir using HEC-HMS and HEC-RAS models. In Proceedings of the National Flood Symposium, İzmir, Turkey, 22–24 March 2010; pp. 267–275. [Google Scholar]
  20. Aydın, M.E.; Kılıç, M.S.; Aksoy, A.T. Use of GIS and HEC-RAS in flood risk analysis: Seyhan River Basin case study. In Proceedings of the 5th National GIS Congress, İzmir, Turkey, 19–22 October 2017. [Google Scholar]
  21. Oğuz, K.; Oğuz, E.; Coşkun, M. Identifying flood risk areas using GIS: Case of Artvin. In Proceedings of the 4th National Flood Symposium, Rize, Turkey, 23 November 2016; pp. 793–803. [Google Scholar]
  22. Özdemir, A.G.; Aksoy, A.I.; Kılıç, M.S. Use of NETCAD and GIS in natural disaster risk analysis: A case study of Lake Van Basin. Int. J. Soc. Sci. 2016, 13, 23–40. [Google Scholar]
  23. Aydın, M.E.; Kılıç, M.S.; Aksoy, A.T. Determining flood risk maps using GIS software. Harit. Tek. Derg. 2013, 29, 11–20. [Google Scholar]
  24. Aydın, M.E.; Kılıç, M.S.; Aksoy, A.T. Flood risk analysis using GIS and HEC-HMS: Case of Gediz River Basin. In Proceedings of the 3rd National GIS Congress, Ankara, Turkey, 22–25 October 2014. [Google Scholar]
  25. Aslami, K.A. Investigation of Flood Risk Maps Under the Impact of Urbanization in Floodplains: Case Study of Göksu Stream, Istanbul. Master’s Thesis, Istanbul Gelişim University, Graduate School of Education, Istanbul, Turkey, 2024. [Google Scholar]
  26. Kamuş, A.O. Flood Modeling and Risk Analysis of the Esmahanım Stream Basin, Düzce. Master’s Thesis, Sakarya University, Graduate School of Social Sciences, Sakarya, Turkey, 2023. [Google Scholar]
  27. Spor, P. Two-Dimensional Flood Modeling in Streams and the Development of Flood Hazard, Risk, and Damage Maps. Ph.D. Thesis, Sakarya University, Institute of Science, Sakarya, Turkey, 2023. [Google Scholar]
  28. Saltoğlu, F. Floods and Flood Risk Analysis in the Söğütlü Stream Basin (Trabzon) with Effective Factors. Master’s Thesis, Samsun University, Graduate School of Education, Samsun, Turkey, 2022. [Google Scholar]
  29. Toprak, A. The Impact of Geographical Factors on Flood and Inundation Formation and Flood Risk Analysis in Ağrı and Doğubayazıt Basins. Ph.D. Thesis, Fırat University, Institute of Social Sciences, Elazığ, Turkey, 2021. [Google Scholar]
  30. Wehner, M.F.; Duffy, M.L.; Risser, M.; Paciorek, C.J.; Stone, D.A.; Pall, P. On the uncertainty of long-period return values of extreme daily precipitation. Front. Clim. 2024, 6, 1343072. [Google Scholar] [CrossRef]
  31. Krakauer, N.Y. Extending the blended generalized extreme value distribution. Discov. Civ. Eng. 2024, 1, 97. [Google Scholar] [CrossRef]
  32. Caruso, M.F.; Marani, M. Extreme-coastal-water-level estimation and projection: A comparison of statistical methods. Nat. Hazards Earth Syst. Sci. 2022, 22, 1109–1128. [Google Scholar] [CrossRef]
  33. Cherif, S. A generalized extreme value approach for the analysis of stationary climatic covariate in a Mediterranean city. Atmos. Sci. Lett. 2025, 26, e1291. [Google Scholar] [CrossRef]
  34. Kolaković, S.; Mandić, V.; Stojković, M.; Jeftenić, G.; Stipić, D.; Kolaković, S. Estimation of large river design floods using the peaks over threshold (POT) method. Sustainability 2023, 15, 5573. [Google Scholar] [CrossRef]
  35. Bat-Erdene, M.; Zorigt, M.; Dambaravjaa, O.; Dechinlkhundev, D.; Sumiya, E.; Nones, M. Leveraging the GEV model to estimate flood due to extreme rainfall in ungauged dry catchments of the Gobi Region. Sustainability 2025, 17, 2500. [Google Scholar] [CrossRef]
  36. Reinders, J.B.; Munoz, S.E. Accounting for hydroclimatic properties in flood frequency analysis procedures. Hydrol. Earth Syst. Sci. 2024, 28, 217–227. [Google Scholar] [CrossRef]
  37. Berton, R.; Rahmani, V. Differences in flood quantiles estimate of disturbed and undisturbed watersheds in the United States. Hydroecol. Eng. 2024, 1, 10002. [Google Scholar] [CrossRef]
  38. Pan, X.; Rahman, A.; Haddad, K.; Ouarda, T.B. Peaks-over-threshold model in flood frequency analysis: A scoping review. Stoch. Environ. Res. Risk Assess. 2022, 36, 2419–2435. [Google Scholar] [CrossRef]
  39. Pan, X.; Yildirim, G.; Rahman, A.; Haddad, K.; Ouarda, T.B. Peaks-over-threshold-based regional flood frequency analysis using regularised linear models. Water 2023, 15, 3808. [Google Scholar] [CrossRef]
  40. Devi, K.; Reddy, C.C.; Rahul, K.; Khuntia, J.R.; Das, B.S. A Holistic Methodology for Evaluating Flood Vulnerability, Generating Flood Risk Map and Conducting Detailed Flood Inundation Assessment. Sci. Rep. 2025, 15, 28253. [Google Scholar] [CrossRef] [PubMed]
  41. Efraimidou, E.; Spiliotis, M. A GIS-Based Flood Risk Assessment Using the Decision-Making Trial and Evaluation Laboratory Approach at a Regional Scale. Environ. Process. 2024, 11, 9. [Google Scholar] [CrossRef]
  42. Martin-Vide, J. Spatial distribution of a daily precipitation concentration index in peninsular Spain. Int. J. Climatol. 2004, 24, 959–971. [Google Scholar] [CrossRef]
  43. Benhamrouche, A.; Boucherf, D.; Martin-Vide, J.; Meddi, M.; Dakhia, K.; Boukeloua, M. Spatial distribution of the daily precipitation concentration index in Algeria (south Mediterranean). Nat. Hazards Earth Syst. Sci. 2015, 15, 617–625. [Google Scholar] [CrossRef]
  44. Mathbout, S.; Lopez-Bustins, J.A.; Martin-Vide, J.; Bech, J.; Rodrigo, F.S. Spatiotemporal variability of daily precipitation concentration and links to teleconnections in the Mediterranean. Int. J. Climatol. 2020, 40, 5931–5950. [Google Scholar] [CrossRef]
  45. Salameh, A.A.M. Using the precipitation concentration index for characterizing the rainfall distribution in the Levant (1970–2018). J. Water Clim. Change 2024, 15, 1945–1960. [Google Scholar] [CrossRef]
  46. Zittis, G.; Almazroui, M.; Alpert, P.; Ciais, P.; Cristofanelli, P.; Elguindi, N.; Fnais, M.; Kaskaoutis, D.G.; Krichak, S.; Saaroni, H.; et al. Climate Change and Weather Extremes in the Eastern Mediterranean and Middle East. Rev. Geophys. 2022, 60, e2021RG000762. [Google Scholar] [CrossRef]
  47. Acero, F.J.; García, J.A.; Gallego, M.C. Peaks-over-threshold study of trends in extreme rainfall over the Iberian Peninsula. J. Clim. 2011, 24, 1089–1105. [Google Scholar] [CrossRef]
  48. Sarr, M.; Moussa, M.; Deme, E.; Diop, B. Trend and return level analysis of extreme rainfalls in Senegal. J. Water Resour. Prot. 2022, 14, 221–237. [Google Scholar] [CrossRef]
  49. Wi, S.; Park, J.; Jung, S.; Bae, D.H.; Kwon, H.-H. Spatial and temporal structure of extreme rainfall trends in South Korea. Water 2017, 9, 809. [Google Scholar] [CrossRef]
Figure 1. The study area.
Figure 1. The study area.
Sustainability 17 09339 g001
Figure 2. Monthly average precipitation (mm) and daily average precipitation (mm).
Figure 2. Monthly average precipitation (mm) and daily average precipitation (mm).
Sustainability 17 09339 g002
Figure 3. Spatial analysis maps of the Mann–Kendall (M-K) and Spearman’s Rho (S-R) tests.
Figure 3. Spatial analysis maps of the Mann–Kendall (M-K) and Spearman’s Rho (S-R) tests.
Sustainability 17 09339 g003
Figure 4. Annual maximum monthly precipitation and trend line for Bismil (2009–2023).
Figure 4. Annual maximum monthly precipitation and trend line for Bismil (2009–2023).
Sustainability 17 09339 g004
Figure 5. Annual maximum monthly precipitation and trend line for Çermik (2009–2023).
Figure 5. Annual maximum monthly precipitation and trend line for Çermik (2009–2023).
Sustainability 17 09339 g005
Table 1. Descriptive statistics of precipitation data for the selected stations.
Table 1. Descriptive statistics of precipitation data for the selected stations.
DistrictStation CodePeriodMean (mm)SD (mm)CvCsr
Bismil181652012–2023355.7073.740.2070.480.64
Çermik178742009–2023601.70152.760.2541.060.54
Çınar183212013–2023368.63113.220.3072.104.66
Çüngüş183222013–2023779.01299.550.3851.331.45
Dicle183232013–2023589.78182.490.3090.23−1.67
Eğil183242013–2023573.75202.750.3531.432.31
Ergani178472009–2023575.76187.250.3251.401.01
Hazro183262014–2023630.68203.840.323−0.20−1.34
Hani183252013–2023689.50213.560.3100.56−1.35
Kocaköy183282013–2023596.58212.480.3560.980.32
Kulp183292013–2023764.30204.040.2670.03−2.76
Lice181632012–2023751.56169.750.226−0.12−0.97
Silvan181642012–2023552.73106.770.193−0.29−1.23
SD: Standard Deviation, Cv: Coefficient of Variation, Cs: Skewness, r: Correlation Coefficient.
Table 2. Kruskal–Wallis test results for annual precipitation among the stations.
Table 2. Kruskal–Wallis test results for annual precipitation among the stations.
TestH Statisticp-ValueSignificant at α = 0.05
Kruskal–Wallis H19.71250.0727No
Table 3. Mean monthly precipitation (mm), minimum, maximum, and standard deviation (by station).
Table 3. Mean monthly precipitation (mm), minimum, maximum, and standard deviation (by station).
DistrictStation CodePeriodMean (mm)Min. (mm)Max. (mm)SD (mm)
Bismil181652012–2023355.70260.00474.3073.74
Çermik178742009–2023601.70435.00858.40152.76
Çınar183212013–2023368.63284.50592.00113.22
Çüngüş183222013–2023779.01514.001310.40299.55
Dicle183232013–2023589.78385.30843.90182.49
Eğil183242013–2023573.75392.80943.90202.75
Ergani178472009–2023575.76426.40904.60187.25
Hazro183262014–2023630.68339.60855.40203.84
Hani183252013–2023689.50474.301009.90213.56
Kocaköy183282013–2023596.58393.40953.90212.48
Kulp183292013–2023764.30543.501007.70204.04
Lice181632012–2023751.56549.20987.50169.75
Silvan181642012–2023552.73420.00693.20106.77
Table 4. Calculated Concentration Index (CI) and Intra-Annual Coefficient of Variation (IACV) for each station.
Table 4. Calculated Concentration Index (CI) and Intra-Annual Coefficient of Variation (IACV) for each station.
StationCIIACV (%)
Bismil0.5446.1
Cermik0.5748.4
Cinar0.5345.7
Cungus0.6352.2
D-Bakir0.5143.8
Dicle0.6050.1
Egil0.5847.9
Ergani0.5646.5
Hani0.6251.8
Hazro0.5949.0
Kocakoy0.6150.7
Kulp0.6453.6
Lice0.6352.1
Table 5. Mean daily precipitation (mm), minimum, maximum, and standard deviation (by station).
Table 5. Mean daily precipitation (mm), minimum, maximum, and standard deviation (by station).
DistrictStation CodePeriodMean (mm)Min. (mm)Max. (mm)SD (mm)
Bismil181652012–2023265.080.00648.70194.97
Çermik178742009–2023588.5728.201107.20301.14
Çınar183212013–2023234.100.00615.90206.52
Çüngüş183222013–2023574.110.001362.10479.85
Dicle183232013–2023477.420.001060.70385.47
Eğil183242013–2023417.700.00939.50342.30
Ergani178472009–2023596.29342.601006.80197.18
Hazro183262014–2023473.110.001044.00384.39
Hani183252013–2023513.920.001063.60410.46
Kocaköy183282013–2023436.010.00948.80352.74
Kulp183292013–2023580.670.001142.00452.32
Lice181632012–2023588.730.001117.30384.88
Silvan181642012–2023286.240.00866.80286.29
Table 6. Monthly average evaporation (mm), minimum, maximum, and standard deviation.
Table 6. Monthly average evaporation (mm), minimum, maximum, and standard deviation.
DistrictStation CodePeriodMean (mm)Min. (mm)Max. (mm)SD (mm)
Ergani *178472009–20231019.220.503874.201514.11
* Evaporation data are available only for the Ergani region and not for the other districts.
Table 7. Lag-1 Autocorrelation Coefficients by Station.
Table 7. Lag-1 Autocorrelation Coefficients by Station.
StationLag-1 Autocorrelation (r1)Interpretation
Bismil0.767High autocorrelation
Cermik0.108Low or negligible autocorrelation
Cinar0.625Moderate autocorrelation
Cungus0.186Low or negligible autocorrelation
Dicle0.2Low or negligible autocorrelation
Egil0.044Low or negligible autocorrelation
Ergani−0.188Low or negligible autocorrelation
Hani0.138Low or negligible autocorrelation
Hazro−0.302Low or negligible autocorrelation
Kocakoy−0.073Low or negligible autocorrelation
Kulp0.213Low or negligible autocorrelation
Lice−0.046Low or negligible autocorrelation
Silvan−0.147Low or negligible autocorrelation
Table 8. Mann–Kendall trend results (by station).
Table 8. Mann–Kendall trend results (by station).
Station* Z Value* Trend* p* Slope* TrendDirection
Bismil2.4406Increasing0.014738.3143Increasing-
Çermik0.9342Increasing0.350247.3None-
Çınar0.6008Increasing0.54836.2667None-
Çüngüş0.9011Increasing0.3675110.5167None-
Dicle1.2015Increasing0.229678.9667None-
Eğil0.9011Increasing0.367586.65None-
Ergani0.1557Increasing0.87635.74None-
Hazro0.3757Increasing0.707135.66None-
Hani1.2015Increasing0.229683.35None-
Kocaköy0.9011Increasing0.367591.1None-
Kulp0.6008Increasing0.54887.075None-
Lice0.6186Increasing0.536250.9607None-
Silvan0.6186Increasing0.536231.5036None-
* Z value indicates trend direction; * p < 0.05 denotes statistical significance; * slope represents annual change; * trend is interpreted as increasing, decreasing, or none.
Table 9. Spearman’s Rho trend results (by station).
Table 9. Spearman’s Rho trend results (by station).
Station* Spearman Rhop-ValueInterpretation
Bismil0.72480.0116Increasing
Çermik0.32730.3259Insignificant
Çınar0.28570.5345Insignificant
Çüngüş0.46430.2939Insignificant
Dicle0.57140.1802Insignificant
Eğil0.46430.2939Insignificant
Ergani0.08180.811Insignificant
Hazro0.25710.6228Insignificant
Hani0.57140.1802Insignificant
Kocaköy0.32140.4821Insignificant
Kulp0.28570.5345Insignificant
Lice0.28570.4927Insignificant
Silvan0.23810.5702Insignificant
* If Rho is positive and p < 0.05: Increasing trend/If Rho is negative and p < 0.05: Decreasing trend/If p ≥ 0.05: Statistically insignificant trend.
Table 10. Winter season Mann–Kendall results for maximum daily precipitation.
Table 10. Winter season Mann–Kendall results for maximum daily precipitation.
StationSeasonzpTheil–Sen Slope (mm/yıl)Trend
BismilWinter3.0930.00212.625Increasing
ÇermikWinter2.180.029319.58Increasing
ÇınarWinter1.8020.071513.04No trend
ÇüngüşWinter1.8020.071548.833No trend
DicleWinter1.8020.071538.6No trend
EğilWinter1.8020.071534.9No trend
ErganiWinter−0.1560.8763−3.45No trend
HazroWinter1.5020.133138.45No trend
HaniWinter1.1270.259721.3No trend
KocaköyWinter1.5020.133138.65No trend
KulpWinter1.5020.133141.3No trend
LiceWinter1.6080.107816.24No trend
SilvanWinter0.6190.53625.033No trend
Table 11. Goodness-of-fit results for GEV and GPD models based on synthetic precipitation data.
Table 11. Goodness-of-fit results for GEV and GPD models based on synthetic precipitation data.
ModelShapeLocationScaleK–S DK–S p-ValueAIC
GEV0.3304116.459.930.10220.970657.51
(POT)0.35670.00129.060.23310.774741.89
Table 12. Flood risk analysis results using the Generalized Extreme Value (GEV) distribution.
Table 12. Flood risk analysis results using the Generalized Extreme Value (GEV) distribution.
Stationr-Max NGEV_ShapeGEV_LocGEV_ScaleGEV_Q10yGEV_Q25yGEV_Q50y
Bismil80.625877.633.9118.53124.47127.08
Çermik11−5.596614.51.685,519.0117,241,6838.84 × 108
Çınar70.367290.728.3134.05143.96149.38
Çüngüş70.0437146.3120403.51504.53576.82
Dicle70.263129.985.6275.31315.07338.78
Eğil7−5.68940.24.4282,219.662,112,1283.4 × 109
Ergani11−5.4148100.60.830,231.615,113,7982.31 × 108
Hani70.3977153.492.5290.93320.78336.69
Hazro6−5.994781.83365,849.11.08 × 1087.29 × 109
Kocaköy7−6.748720.52.61,509,1509.07 × 1081.05 × 1011
Kulp71.1044223.8102.4308.78313.79315.25
Lice8−5.421119.70.725,277.784,295,1401.95 × 108
Silvan80.7001147.931.7183.75188.29190.16
Table 13. Flood risk analysis results based on the Peak Over Threshold (POT) method.
Table 13. Flood risk analysis results based on the Peak Over Threshold (POT) method.
StationPOT_ThresholdExcess_NPOT_Q10yPOT_Q25yPOT_Q50yTrend
Bismil703104.37108.8710.6Increasing
Çermik703347.2347.2347.2No trend
Çınar603143.5143.5143.5No trend
Çüngüş603499.9499.9499.9No trend
Dicle603205.5205.51205.6No trend
Eğil603286.9286.9286.9No trend
Ergani703328.8328.8328.8No trend
Hani603323.7323.7323.7No trend
Hazro503266266266No trend
Kocaköy603294.2294.2294.2No trend
Kulp603316.5316.5316.5No trend
Lice703274.4274.4274.4No trend
Silvan703149.9179.2554.2Increasing
POT_threshold: threshold value selected for each station. Excess_N: number of threshold exceedances. POT_Q10y, POT_Q25y, and POT_Q50y represent the estimated precipitation values for 10-, 25-, and 50-year return periods, respectively, derived from the POT method.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Kaya, B.; Çelik, R. Extreme Precipitation and Flood Hazard Assessment for Sustainable Climate Adaptation: A Case Study of Diyarbakır, Turkey. Sustainability 2025, 17, 9339. https://doi.org/10.3390/su17209339

AMA Style

Kaya B, Çelik R. Extreme Precipitation and Flood Hazard Assessment for Sustainable Climate Adaptation: A Case Study of Diyarbakır, Turkey. Sustainability. 2025; 17(20):9339. https://doi.org/10.3390/su17209339

Chicago/Turabian Style

Kaya, Berfin, and Recep Çelik. 2025. "Extreme Precipitation and Flood Hazard Assessment for Sustainable Climate Adaptation: A Case Study of Diyarbakır, Turkey" Sustainability 17, no. 20: 9339. https://doi.org/10.3390/su17209339

APA Style

Kaya, B., & Çelik, R. (2025). Extreme Precipitation and Flood Hazard Assessment for Sustainable Climate Adaptation: A Case Study of Diyarbakır, Turkey. Sustainability, 17(20), 9339. https://doi.org/10.3390/su17209339

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop