Leveraging the GEV Model to Estimate Flood Due to Extreme Rainfall in Ungauged Dry Catchments of the Gobi Region

Abstract

Extreme high flows can have negative economic, social, and ecological effects and are expected to become more severe in many regions due to climate change. Knowledge of maximum flow regimes and estimation of extreme rainfall is important, especially in ungauged dry regions, for planning and infrastructure development. In this study, we propose a regional method for estimating extreme flow regimes and modeled extreme rainfall using the extreme value theory, with examples from the Gobi region of Mongolia. The first step is to apply the Generalized Extreme Value (GEV) theory for the maximum rainfall data using 44-year observational data covering the period 1978–2022. Then, estimated rainfall with a 100-year return period is used for the empirical equation of the maximum flood calculation. As a result, most stations’ maximum rainfall follows a Fréchet distribution and 100-year return period rainfall values that range between 27.8–130.6 mm. The local reference value in the 100-year return period rainfall is defined as 90 mm for the whole Gobi region. Our results show that extremely high rainfall in the Gobi region has changed from −7% to 16%, leading to higher flood events. These findings further provide evidence for the maximum rainfall for flood calculation, climate change impact assessment, water resource planning, and management studies.

1. Introduction

Having information on the maximum flow of rivers is crucial for human security, as well as hydro techniques and water infrastructure planning, especially in scarcely gauged and ungauged basins [1]. The geographical location of Mongolia combined with the continental harsh climate favors the formation of water resources that are unequally distributed throughout the country. Perennial rivers exist in the northern part of the country, but the southern part is drier [2]. Weather conditions and surface flow are mostly dependent on rainfall, intensity, and the structure of soil and soil infiltration rate because, during heavy rains, there is a small flow of dry pebbles occurring in the study area, the Gobi region. This region is included in the Central Asian Internal drainage basin without outflow, and 1000–1500 mm of water evaporates from open water. Following the climate zone Köppen–Geiger classification [3], the southern part of Mongolia lies in the dry Bwk (cold desert climate) class, even though the local climate is dry and less precipitation occurs compared to the other regions of the country, with heavy rains and rainwater that contribute in reaching flood levels. Such floods are a major cause of damage and loss for the local community [4].

In the Gobi region, floods occur during the heavy rainfall in the summertime. In that region, rainfall flash floods and mixed mud floods occur as a result of high-intensity rainfall during July and August, but the water that feeds small rivers infiltrates into the soil or evaporates within 2–3 days. The floods have not been measured because of the limited local observational network. It is crucial to estimate floods for infrastructure planning and development projects, as well as to ensure community safety. Empirical methods are applied to estimate rainfall floods for the dry catchments in the region. Regardless of the empirical method applied, a key driver is the amount of maximum rainfall that is considered for flood calculation.

In recent years, heavy rainfall patterns and extreme precipitation magnitude have changed under a changing climate [5,6,7,8,9]. Moreover, several studies noted that hydroclimatic conditions and atmospheric circulation patterns over Eurasia have significantly changed since the mid-1990s [10,11,12,13]. Light rains and the number of rainy days have decreased while rain intensity has increased [14,15,16]. Recent studies also concluded that the maximum precipitation has occurred in a shorter duration, and very heavy precipitation events have increased [17,18]. Similar studies were pointed out from research conducted in Mongolia, which highlighted that annual precipitation has not changed much over time, but the duration of high-intensity rains increased significantly in the eastern and southeastern parts of the country [19]. The variability in rainfall patterns and increased air temperature tend to have more extreme events and floods in drylands [20]. These changes in the rainfall patterns are projected to continue in the future [21,22,23].

Extreme rainfall data and extreme value modeling are among the most critical processes in flood calculation. Plenty of analyses and studies on extreme precipitation changes have been conducted in different regions by employing different methods. Most precipitation data tend to have extreme values with a heavy-tailed distribution [24]. Modeling high rainfall using the fitted distribution of the actual data increases the possibilities of using input for the flood model development. The characteristics of extreme precipitation have been analyzed using statistical approaches such as probability distributions like the Generalized Extreme Value (GEV) distribution model [25,26,27,28,29,30] and frequency analysis [31]. Different regions in the world have adopted the GEV model for maximum rainfall distributions. GEV was recommended as a suitable model for describing the distribution of annual maximum precipitations in the Ontario region, Malaysia, China, and European countries [32,33,34,35,36]. Especially for the arid region, GEV reasonably predicts higher extreme rainfall intensity [31] and represents a reasonable distribution to reproduce when data is extremely imbalanced with a limited sample size [37]. However, past studies concluded that sampling data and geographical locations have different behavior in estimating extreme rainfall through the GEV model [34]. Thus, to recognize the local characteristics of extreme rainfall and to further calculate flood conditions, the GEV model was selected for this study.

The model has three parameters, which are found through different methods such as maximum likelihood estimator, L-moments and probability-weighted moments, and penalized maximum likelihood estimator [38]. The location parameter is defined as a distribution that follows Gumbel, Fréchet, and Weibull. Investigations should take into account that changes in high rainfall are critical for the analysis of the maximum flood calculation, especially in the case of ungauged dry regions where data availability is limited, such as in the case of the Gobi region. Here, there is no information on flood occurrence in dry beds and channels, and studies on how runoff responds to rainfall patterns are unknown [39]. To overcome actual limitations and provide additional insights into the regional flood dynamics, the empirical method “Rainfall intensity method” for flood calculation was adopted using local parameters focusing on small rivers with catchment areas of less than 200 km², and considering a return period of 100 years. The empirical equation depends on various parameters, in particular the 100-year return period rainfall. Therefore, to improve the flood calculation, three goals were achieved in this study. The first is to identify the trends in high rainfall from the observed data and the second is to apply extreme precipitation occurrence using the GEV model in the ungauged dry catchments of the Gobi region. The last goal is to estimate maximum floods in these dry river catchments using the modelled 100-year return period rainfall. The outcomes of the study would be crucial for flood prevention and management in the region.

2. Materials and Methods

2.1. Study Area and Dataset

The study area is located in the Umnugobi province of southern Mongolia, Gobi region (see Figure 1). Due to the recent intensive development of the mining industry in this region, the human population and number of livestock are rapidly growing. Flood occurrences in the Gobi region are serious in the short term, and their research is crucial for infrastructure development and engineering design. The study area belongs to the desert and desert steppe region. The mean annual precipitation is 63–102 mm [40]. The most distinctive rainfall condition characteristic of the region is that over 90% of total precipitation falls during the summer season, and only about 7% comes down during the winter season. The time range of flood is, on average, 15 to 20 days, depending on rainfall intensity. The Gobi region of Mongolia has experienced notable changes in precipitation patterns and an increase in flood events influenced by climate change and human activities. The alteration in precipitation patterns has resulted in more frequent and intense flood events in the Gobi region. The increase in heavy rainfall leads to flash floods, which are exacerbated by the region’s arid conditions and sparse vegetation, reducing the land’s ability to absorb water. These floods pose significant challenges to local communities, particularly nomadic herders, by damaging infrastructure, degrading pastures, and threatening livestock [41]. These climatic shifts underscore the need for adaptive strategies to mitigate the adverse effects on the environment and livelihoods in the Gobi region. Since the 1940s, Mongolia’s average temperature has risen by approximately 2.25 °C, about three times the global average [23,42]. This warming has led to changed precipitation and altered rainfall patterns, with a shift from light, ground-absorbing rains to heavier downpours that cause rapid runoff and soil erosion. These changes have contributed to the expansion of desertification in the Gobi region [41].

The daily precipitation data from 10 meteorological stations/gauges and morphometric measurement data from 27 dry beds are included in the study. Precipitation data gathered from the National Agency of Meteorology and Environmental Monitoring refer to the period of 1978–2022 for four stations and 1985–2022 for six stations, respectively (

Table 1. Information of the meteorological stations used in the study.

No.	Station Name	Station Code	Elevation, m asl	Natural Zone
1	Bayan-Ovoo	S1	1179	Desert
2	Dalanzadgad	S2	1465	Desert
3	Gurvantes	S3	1726	Desert steppe
4	Khanbogd	S4	1115	Desert
5	Khurmen	S5	1701	Desert steppe
6	Mandal-Ovoo	S6	1084	Desert steppe
7	Noyon	S7	1960	Desert steppe
8	Saikhan	S8	1328	Desert steppe
9	Tsogt-Ovoo	S9	1287	Desert steppe
10	Tsogttsetsii	S10	1480	Desert steppe

).

A combination of field measurements and GIS-based spatial analysis was employed. Within the study area, 27 small ephemeral streams selected to systematically analyze using topographic maps (1:100,000 scale) for initial watershed delineation. GIS was employed to extract key catchment characteristics, including drainage area, stream length, and slope (longitudinal gradient). Field measurements were then conducted to validate these parameters and provide additional cross-sectional data. According to

Table 2. Morphometric data of dry beds.

No.	Code of the Dry Beds	Coordinates		Catchment Area	Length	Slope
		Longitude	Latitude	S, km2	L, km	J, ‰
1	OP1	105.57018	43.64322	16.9	7.9	12.4
2	OP2	105.60599	43.64468	5.0	2.5	1.26
3	OP3	105.63950	43.63189	9.4	4.9	16.2
4	OP4	105.69949	43.57616	21.1	6.1	10.4
5	OP5	105.70716	43.56900	30.6	7.8	6.88
6	OP6	105.79048	43.52828	31.0	16.6	10.4
7	OP7	106.15165	43.08921	27.9	7.6	6.20
8	OP8	106.41030	43.02717	1.0	1.4	17.8
9	OP9	106.41524	43.02572	1.4	1.9	10.2
10	OP10	106.49008	43.01194	21.6	5.1	5.82
11	OP11	106.52481	43.01148	1.1	1.4	21.4
12	OP12	106.56545	43.01076	6.1	3.8	9.39
13	OP13	106.58033	43.01038	1.9	1.8	16.2
14	OP14	106.61200	43.01063	7.8	5.6	10.7
15	OP15	106.61437	43.01077	5.9	4.9	12.3
16	OP16	106.64359	43.01625	2.5	1.1	10.8
17	OP17	106.64596	43.01674	3.0	1.5	12.8
18	OP18	106.70157	43.01905	8.3	3.2	7.8
19	OP19	106.73463	43.00953	24.3	6.2	7.3
20	OP20	106.94933	42.93028	9.7	4.3	6.9
21	OP21	106.98352	42.84089	3.4	2.2	8.3
22	OP22	107.02945	42.77521	1.5	1.8	8.6
23	OP23	107.13343	42.66283	80	17.9	9.5
24	OP24	107.36268	42.48558	49.4	7.8	4.5
25	OP25	107.43750	42.46302	95.2	16.9	4.3
26	OP26	107.55582	42.44049	26.8	9.2	6.0
27	OP27	107.61814	42.41307	11.5	3.5	6.1

, the catchment area ranges from 1.0 to 95.2 km², while their lengths vary between 1.11 and 7.9 km. The bed slope ranges from 1.8 to 15.1‰. The Manning (roughness) coefficient is 0.067 s/m^1/3, while the Chezy’s coefficient varies between 6.93–12.2 for the selected 27 dry beds. The flow velocity ranges from 0.03 to 0.53 m/s, as determined during field measurements performed in 2013, while the depth of the riverbeds varies in the range 0.01–0.29 m.

2.2. Methods

Precipitation changes from 1978 to 2022 were analyzed using the Mann–Kendall test. Then, the GEV model was applied to define a 100-year return period rainfall, which is one of the important parameters for calculating maximum runoff.

2.2.1. Trend Analysis

The non-parametric Mann–Kendall (MK) trend analysis [43,44,45] was applied to the maximum rainfall for the ten stations of the study area. The purpose of the test is to assess statistically whether there is a monotonic upward or downward trend of the variable at a certain time series. The MK test works as the slope of the estimated linear regression line is non-zero. The MK test is a non-parametric test which does not require the residuals from the fitted regression line, and uses the following statistics for the time series ${\mathrm{x}}_1,\dots {\mathrm{x}}_{\mathrm{n}}$:

$$\mathrm{S}=\sum _{\mathrm{i}=1}^{\mathrm{n}-1}\sum _{\mathrm{j}=\mathrm{i}+1}^{\mathrm{n}}\mathrm{s}\mathrm{g}\mathrm{n}({\mathrm{x}}_{\mathrm{j}}-{\mathrm{x}}_{\mathrm{i}})$$

(1)

Equation (1) illustrates the number of the data $n$ for ${\mathrm{x}}_{\mathrm{j}}$ and ${\mathrm{x}}_{\mathrm{i}}$ which are values in years $\mathrm{j}$ and $\mathrm{i}$. Then, Z statistics are computed using the following formula:

$${\mathrm{Z}}_{\mathrm{m}\mathrm{k}}=\left\{\begin{array}{cc}{\displaystyle \frac{\mathrm{S}-1}{\sqrt{\mathrm{V}\mathrm{a}\mathrm{r}\left(\mathrm{S}\right)}}},\hfill & \mathrm{i}\mathrm{f} \mathrm{S}>0\hfill \\ 0,\hfill & \mathrm{i}\mathrm{f} \mathrm{S}=0\hfill \\ {\displaystyle \frac{\mathrm{S}+1}{\sqrt{\mathrm{V}\mathrm{a}\mathrm{r}\left(\mathrm{S}\right)}}},\hfill & \mathrm{if} \mathrm{S}<0\hfill \end{array}\right.$$

(2)

Positive and negative ${\mathrm{Z}}_{\mathrm{m}\mathrm{k}}$ explain the increasing and decreasing trend, respectively. The magnitude of the trend was analyzed through Sen’s slope estimator.

2.2.2. GEV Model for Extreme Precipitation

GEV distribution has been widely used [46,47] in many fields, such as economy, finance, engineering, climate, and natural sciences, to estimate the distribution of extreme values. The distinguishing feature of extreme value analysis is the objective to quantify the stochastic behavior of a process at unusually large—or small—levels. In particular, extreme value analyses usually require estimating the probability of events that are more extreme than any that have already been observed [48]. GEV has two principles for maximum value from the time series: block maxima and threshold criteria value. The distribution consists of three parameters such as shape (ξ), scale (σ), and location (µ).

$${\mathrm{H}}_{\xi ,\mu ,\sigma }=\left\{\begin{array}{c}\exp \left({-\left(1+{\displaystyle \frac{\xi \left(\mathrm{x}-\mu \right)}{\sigma }}\right)}^{-1/\xi }\right), \xi \ne 0\\ \exp \left(-{\mathrm{e}}^{-{\displaystyle \frac{\mathrm{x}-\mu }{\sigma }}}\right), \xi =0\end{array}\right.$$

(3)

ξ = 0 Gumbel, ξ > 0 Fréchet, and ξ < 0 Weibull, where, three parameters of the GEV model (namely, location ($\mu$), scale (δ) and shape $(\xi )$ parameters) are considered to compute the independent variable x.

To estimate parameters, the maximum likelihood estimator (MLE) is applied as described in the following equation.

$$\begin{array}{c}\left(\mu ,\sigma , \xi \right)=-m{log}_{\sigma }-\left(1+{\displaystyle \frac{1}{\xi }}\right){\displaystyle \sum _{i=1}^m}\log \left[1+\xi \left({\displaystyle \frac{z_i-\mu }{\sigma }}\right)\right]-{\displaystyle \sum _{i=1}^m}{\left[1+\xi \left({\displaystyle \frac{z_i-\mu }{\sigma }}\right)\right]}^{-{\displaystyle \frac{1}{\xi }}}\\ 1+\xi \left({\displaystyle \frac{z_i-\mu }{\sigma }}\right)>0, for i=1,\dots ,m\end{array}$$

(4)

where $z_i$ represents the independent variable.

The return level at 100 years of rainfall was identified for each station. After rainfall estimation, the maximum flood discharges for the dry channels were analyzed by means of an empirical model, namely the “Rainfall intensity method”. For this method, 1% precipitation is one of the important parameters [49,50] in the method. Since more than half of the annual precipitation occurs during the summer season [51], floods are primarily caused by rainfall during this time of year.

2.2.3. Rainfall Intensity Method for Flood Calculation

The method is adopted in catchments over Mongolia and is widely used, especially for the ungauged and unstudied catchments. Local parameters were applied to reference values. One of the inputs is maximum rainfall at a 100-year return level. Former studies [52] use the regional maximum rainfall value or log-normal/gamma distribution.

$$Q_{1\%}=q_{1\%}J{H}_{1\%}s{I}_{1\%}F$$

(5)

where $Q_{1\%}$ (m³/s) represents the flood flow with 1% probability. The coefficient of specific runoff with a 1% probability $q_{1\%}$ (l/s km²) is applied. Multipliers including flow multiplier $J$ and lake, forest, and swamp multipliers $s$ are used. Also, $I_{1\%}$ stands as a multiplier to change from 1% probability to other rates of probability. $H_{1\%}$ defines maximum rainfall with 1% probability (mm) and $F$ (km²) represents the catchment area.

3. Results

A temperature rise indeed accelerates the hydrological cycle, hence increasing the intensity and frequency of extreme rainfall events [53]. A projected increased summer rainfall intensity, as well as an increased frequency of heavy rainfall events, is likely to intensify the risk of river and flash floods in the region. The number of very heavy precipitation days is a climate index that indicates the annual number of days with a daily rainfall of over 20 mm. A seasonal increase in monthly precipitation occurs mainly in July–August [40].

The current analysis was done using rainfall data from 10 meteorological stations distributed across the study area and covering 38–44 complete years. Firstly, a time series of extreme values was made from observational datasets using block maxima. The daily maximum of the year at each station is summarized in Figure 2.

Daily maximum rainfall of the stations ranges between 0.4 mm to 90 mm for the period 1978–2022, as reported in Figure 3.

The Mann–Kendall test analyzed the trends for the maximum rainfall data, and the results are shown in

Table 3. Mann–Kendall test parameters.

Stations Code	Z-Value	Sen’s Slope	p-Value
S1	1.08	0.12	0.27
S2	−0.41	−0.03	0.68
S3	−0.88	−0.07	0.37
S4	1.42	0.09	0.14
S5	2.05	0.30	0.04
S6	1.68	0.19	0.09
S7	1.37	0.23	0.16
S8	1.04	0.12	0.29
S9	0.67	0.06	0.53
S10	−2.18	−0.62	0.02

. As visible, there are statistically insignificant increasing and decreasing trends in the data, with only two stations (S5 and S10) that have significant increasing and decreasing trends, respectively.

The estimated quantile and GEV parameters (location, shape, and scale) are shown in

Table 4. Estimated GEV parameters for each station.

Station Code	Maximum Livelihood
	Location (µ)	Scale (σ)	Shape (ξ)
S1	11.5	5.5	0.1
S2	27.9	5.1	−0.1
S3	33.7	7.3	−0.2
S4	27.9	5.1	−0.1
S5	11.3	8.1	0.2
S6	9.4	6.2	−0.2
S7	15.4	7.3	0.2
S8	16.1	8.0	0.1
S9	33.7	7.2	−0.2
S10	16.5	12.7	0.2

. The shape parameter ranges from −0.2 to 0.2 and the location parameter from 9.4 to 33.7, while the scale parameter ranges from 5.1 to 12.7.

Stations coded S1 (Gurvantes), S5 (Khurmen), S7 (Noyon), S8 (Saikhan), and S10 (Tsogttsetsii) are described as Fréchet distribution, while others follow the Weibull distribution. The goodness of fit of GEV was evaluated using the QQ plot, which is a useful tool to validate the similarity of the empirical distribution and critical distribution [26]. The results show that GEV has fitted the maximum rainfall for the stations well. Figure 4 shows the QQ plot (95% tolerance interval) for the fit of the GEV of annual maximum rainfall.

The 100-year return period value ranges from 27.8 mm to 130.6 mm (

Table 5. Estimated maximum rainfall at 1% (i.e., with a return period of 100 years).

No.	Station Code	Return Level 100-Years
1	S1	44.6
2	S2	63.1
3	S3	68.4
4	S4	46.5
5	S5	79.1
6	S6	27.8
7	S7	79.4
8	S8	71.1
9	S9	54.1
10	S10	130.6

). With the aim of improving the calculation of the flood, the 1% rainfall using the GEV model is reanalyzed for those stations. In fact, it is more reasonable to use extreme rainfall from the GEV than reference values of the observed maximum rainfall in the stations.

We applied the 1% rainfall from GEV model to calculate the maximum flow in the selected 27 catchments (Figure 5). The morphometric parameters of the various catchments range between 0.4–95.2 km², length 1.11–17.9 km, and slope 1.26–31.5‰, respectively. The results of the estimated maximum runoff range between 0.94–15.7 m³/s for the rainfall intensity method, using the former 100-year return period rainfall defined as 90 mm for the study area or the observed maximum value of the closest station. The re-estimated maximum rainfall through the GEV model is 130.6 mm. Using that value in calculating the maximum flood provides a range of 1.4–22.6 m³/s.

4. Discussion

The annual maxima of daily time series in rainfall data for the period 1978–2022 were analyzed to force the GEV model to calculate the 100-year return period rainfall using 10 meteorological stations evenly distributed over the study area. This is important in the Gobi region, where the estimation of flood flows for the dry catchments depends on an accurate extreme rainfall estimation. Due to the arid region, flood observations are rare and rainfall is the main driver of the floods. Extreme rainfall is a key input parameter reflected on flood flows in arid and semi-arid region where catchments respond with rapid runoff due to limited infiltration capacity [54]. The ability of estimated return period rainfall is therefore critical for flood risk assessments and infrastructure design. The practice of using historical reference rainfall values for flood investigations is common but its reliability should be accurately assessed due to the impacts of climate variability and long-term changes.

According to the Mann–Kendall (MK) test applied here to the series of annual maxima for the study period, there are no significant trends at most stations, except for two. This is normal for arid and semi-arid regions where interannual rainfall variability is often high and long-term trends are difficult to detect and similar heterogeneity in trends has been observed in neighboring arid regions. Different trends in extreme rainfall have been observed in arid regions, such as decreasing in Inner Mongolia [55,56] and increasing trends in North China and the Yellow River basin, while other regions including the Yangtze River basin, Southeast Coast, South China, Inner Mongolia, Northwest China, and Tibetan Plateau over the period 1961–2009 [56]. Such spatial contrasts highlight the complex interaction between regional climate drivers, such as the East Asian monsoon and westerlies, which influence precipitation patterns across the region.

The block maxima approach, using annual maxima from daily time series, was used to collect extreme rainfall events over time. The GEV model was then applied, defined its flexible three-parameter formulation that allows it to represent different types of extreme value distribution following the Gumbel, Fréchet, and Weibull distribution. In this study, the shape parameter indicated that the Fréchet distribution provided the best fit to the observed rainfall extremes. This aligns with previous global analyses [48,57], which found that heavy-tailed Fréchet distributions often characterize extreme rainfall in arid and semi-arid regions where isolated convective storms produce highly localized, intense rainfall events. However, some studies conducted in different climate zones such as humid temperate have found that the Gumbel distribution offers a better fit, especially in areas with more regular and less intense rainfall patterns [58]. Other authors [59,60] obtained similar results, with 80% of the rainfall data corresponding to a Fréchet distribution, considering a global data survey. However, other local regions have different results, with the Gumbel probability distribution performing better, adjusting to 87.5% of cases [31]. Based on the literature evidence and the results presented here, the GEV model has the potential to predict extremes connected to climate variations; therefore, time series duration and characteristics of the data should be further considered in the future.

The results presented in the current study, combined with findings from the literature, indicate that the GEV model is not only useful for characterizing past extremes but also offers a foundation for developing climate-informed flood estimates when combined with climate model projections. Another important factor highlighted by this study is the influence of time series length and data quality on the reliability of extreme value analysis. The 44-year period analyzed here provides a reasonable basis for estimating 100-year rainfall return periods; however, longer records would improve confidence, particularly given the high variability inherent to arid climate systems. Future work should consider integrating paleo-hydrological data, satellite-derived rainfall products, and reanalysis datasets to extend the temporal coverage and reduce uncertainty.

5. Conclusions

This study aims to estimate the rainfall maximum flood along the small creeks and dry beds in the Gobi region of southern Mongolia, Umnugobi province. For that purpose, modeling the 100-year return period rainfall is the main input for calculating the maximum magnitude of the floods. Because of a lack of heavy rainfall analysis, the GEV model was tested for this dry region. The maximum rainfall was analyzed and modeled with GEV in the Gobi region of Mongolia considering the period from 1978–2022, and this pointed out changes that span between −7% and 16%, depending on the natural characteristics of each zone where a station is located. The heavy rainfall from the two stations has a significant increasing and decreasing trend as the MK test, but others are insignificant.

Following the GEV analysis, most stations’ maximum rainfall has a Fréchet distribution and 100-year values that range between 27.8–130.6 mm. The local value in the 100-year return period rainfall is defined as 90 mm for the whole Gobi region.

The geographical distribution of the maximum rainfall shows that the western and northern part of the region has a high potential, meaning that heavy floods are more likely to happen here. The catchment area of the selected creeks and dry beds varies from 0.4 to 95.2 km² with a length of 17.9 km. Field surveys were carried out on 27 dry beds and morphological and cross-sectional data were measured.