1 Introduction

Climate change, which is one of the most important environmental problems of recent years globally, shows its effect on precipitation as well as many hydrological/climatological variables. Precipitation, which is a variable climate element in time and space, directly affects natural water cycles and also provides information about the climate of a region. Extremes in the amount of precipitation cause an increase in natural disasters such as drought, flooding, soil erosion and forest fires (Gu et al. 2017). These natural disasters, which occur as a result of excessive precipitation events, can cause economic damage, even loss of life, as well as ecological. Potential changes in the distribution and amount of precipitation may further amplify the local effects of water scarcity (Farre and Faci 2006). In arid and semiarid regions, the water requirement in the agricultural sector constitutes approximately 80% of the total water use (Fereres and Soriano 2007). Therefore, the downward trend in precipitation could have serious consequences for the agricultural sector, one of the world's largest consumers of freshwater resources.

According to the IPCC (2007), more frequent, more severe and prolonged droughts, heat waves and forest fires are expected in Southern Europe, including Turkey, in the twenty-first century (Kreibich et al. 2017). It is estimated that this situation will increase water stress and cause more frequent and severe droughts, resulting in water shortages, increases in forest fires, loss of biodiversity and income loss in agriculture and tourism (Back et al. 2019). In addition, with the increase in the number of days with short but heavy downpours, significant increases in flash floods are predicted (Coscarelli and Caloiero 2012). It is stated by various researchers that precipitation characteristics and variability may also cause precipitation erosivity and soil loss and decrease in agricultural productivity (Nunes et al. 2016; Elagib 2011). The changes in the amount and intensity of precipitation, and the presence of extreme values effect soil erosion. The soil erosion by water, one of the leading threats to soil degradation, is a quite complex phenomenon (De Luis et al. 2010; Lukić et al. 2019). Precipitation erosion is the potential of atmospheric precipitate to reason soil erosion and indicates the climatological element in the overall water erosion process. Soil erosion by water is a major cause of land degradation and a critical environmental hazard in modern times. The amount of precipitation is not the only effective factor in the risk of soil erosion. At the same time, vegetation, soil properties and topography play an important role in soil erosion (Bezak et al. 2021), because a decrease in the amount of precipitation may adversely affect the development of vegetation, which serves as a source of soil organic matter and protection for the soil surface, and the decreased amount of vegetation and precipitation may cause a severe effect that increases soil erosion (Panagos et al. 2015). Another important issue affecting soil erosion by water is the precipitation concentration throughout the year. Precipitation concentration is one of the most important issues of climate. The distribution of precipitation also changes from year to year, as precipitation throughout the year is quite irregular, variable and unpredictable. Precipitation concentration should be taken into account in soil erosion, soil protection and land evaluation studies (Lukić et al. 2019). For these reasons, examining the seasonal/annual spatial and temporal variability of precipitation is of great importance to determine the erosive potential of precipitation and its effects on soil erosion, to evaluate climate-induced changes and to determine adequate future soil and water resources management strategies (Yin et al. 2021; Valdes Pineda et al. 2016).

In many countries, soil loss as a result of erosion is determined in relation to precipitation characteristics, maximum precipitation intensity, duration and kinetic energy. The limited number of measurement systems for precipitation intensity necessitates the determination of erosion severity values by using the monthly total precipitation amounts (Ries et al. 2009). For this purpose, various statistical procedures such as Precipitation Concentration Index (PCI), Seasonality Index (SI), and Modified Fournier Index (MFI) are widely used, and precipitation variability, its aggressiveness and concentration are investigated at both global and regional scales. Many studies have been carried out using these indices to assess possible changes in the water cycle and also to describe morphological and hydrological processes (Valdes Pineda et al. 2016). Some of these studies are: Turkey (Yeşilırmak and Atatanır 2016; Ustaoğlu et al. 2021; Koycegiz and Buyukyildiz 2021), Algeria (Ghorbani et al. 2021), Italy (Di Lena et al. 2021), Spain (Ramos and Martinez 2006), Iraq (Mohammed and Scholz 2019), China (Shi et al. 2015; Gu et al. 2017; Yin et al. 2021; Li et al. 2017; Du et al. 2020), Sudan (Elagib 2011), India (Doyle 2020), Ethiopia (Shawul and Chakma 2020), Portugal (Nunes et al. 2016) and Mexico (Núñez-González 2020).

Martin-Vide (2004) determined that the daily precipitation concentration values obtained for 1951–1990 period in the 32 meteorological stations of the peninsular Spain clearly divide the country into two regions. It revealed that the eastern facade of the country had at least 70% of the total precipitation and 25% of the rainiest days and was represented by a high concentration, while the rest present more regular precipitation. In the study conducted by Al-Shamarti (2016) using monthly precipitation data of 29 stations for the period 1980–2010 in Iraq, the PCI < 10 interval according to precipitation concentration values was not found in any station (uniform precipitation distribution). According to the range of PCI values obtained, the study area was divided into two zones as low concentration zone (PCI < 16) and high concentration zone (PCI > 16). According to the maximum frequencies, moderate precipitation prevails in 63.52% of the years (13 out of 29 stations) during the study period in the low concentration zone. Irregular and strongly irregular precipitation distribution prevails in 76.20% of the years (16 out of 29 stations) in the high concentration zone. Tolika (2019) investigated the precipitation concentration on an annual and seasonal basis using daily precipitation data from 22 stations in Greece during the 1958–2000 periods. On an annual basis, a moderately concentrated distribution was obtained in the north of Greece, while PCI values increased toward the south of the country. Whole Greece is characterized by a moderate concentrated distribution in winter, spring and autumn, while in summer the index values are much higher. Mondol et al. (2018) examined precipitation concentration for different temporal scales using precipitation data from 30 meteorological stations for the period 1980 and 2011 in Bangladesh. As a result of the analysis, annual and semi-annual PCI values for all regions have moderate and irregular precipitation distribution. Also, seasonal and quarterly PCI values were generally moderately irregular, while most PCI values for shorter two-month scales showed a uniform precipitation distribution.

Elagib (2011) calculated the SI, PCI and MFI values for the extreme arid region of Sudan and found that the precipitation characteristics were not significantly homogeneous in terms of erosivity and seasonality and that there were temporal and spatial contrasts in the trends of these indices. Du et al. (2020) examined the long-term concentration index (LCI) and annual PCI values in ten main basins with different climatic and geographical characteristics in China. Shawul and Chakma (2020) found irregular precipitation distribution in the Upper Awash Basin in Ethiopia according to the annual average and dry season PCI values, while they found lower precipitation concentration in the rainy season PCI values. Nunes et al. (2016) showed an overall increasing trend in precipitation, PCI and MFI values for Portugal in autumn and summer months while Lukic et al. (2019) found an increasing trend in potential erosivity based on FI, MFI and PCI across the Pannonia basin (Hungary) according to the MK trend test. Valdes Pineda et al. (2016) examined the spatiotemporal patterns, aggressiveness and concentration of annual precipitation along the southwestern coast of South America from 1930 to 2006 at an annual and multi-decade scale using data from 107 stations. As a result of the analyses, in the last 50 years of the analyzed period, insignificant positive trends in annual precipitation and aggressiveness were observed for the region between 36° and 44°S, and significant negative trends was observed between 44° and 49°S. Garcia-Marin et al. (2017) investigated spatial variations and trends in annual precipitation, seasonality and precipitation erosivity using SI, PCI and MFI with monthly precipitation data of 225 stations in Western Andalusia for the period 1945–2005. According to the results of the analysis, precipitation is characterized by a marked seasonality, while a small area has a high erosivity index. The results of trend test showed that there was a decreasing trend in spring–summer precipitation in 72% of the analyzed precipitation measurements.

According to international publications and IPCC reports, Turkey, located in the south of the Mediterranean Region, is one of the most vulnerable countries to be exposed to drought as a result of severe warming (Şen 2022; IPCC 2007; Turkes 2020). As a natural result of the Mediterranean climate, a hot and dry summer season is dominant in a large area of Turkey except the north. Global and regional climate changes due to other human activities such as increased greenhouse effect and deforestation cause decreasing water resources and increasing the frequency and severity of drought events in Turkey, through the degradation of ecosystems and the weakening of biodiversity.

Seyhan Basin, located in the south of Turkey and hosting the most fertile agricultural lands of Turkey and Europe, is located in the Mediterranean Basin, which is considered by the IPCC as one of the regions to be affected by climate change (IPCC 2007). Researchers focus on the Seyhan Basin for drought or other issues (Cavus and Aksoy 2019; Koycegiz and Buyukyildiz 2022; TUBİTAK-Marmara Research Center 2010; Dikici 2022; Topçu et al. 2022; Gumus 2019).

Trend analyses have also been studied by many researchers in order to determine the effects on precipitation of climate change. In trend analysis, nonparametric trend tests such as Spearman's Rho (SRho) and the Mann–Kendall (MK) are widely used (Shawul and Chakma 2020; Du et al. 2020; Ghorbani et al. 2021; Ashraf et al. 2021; Borah et al. 2022; Saplıoğlu and Güçlü 2022). However, these trend tests applied to the entire time series are purely statistical and create a lack of information about sub-trends (or short-term latent trends) in the data. In addition, if the changes in the data are not supported by the results obtained from the graphical investigation, in some cases they may give the meaningless results (Onyutha 2021). For this reason, Onyutha's test, a graphical–statistical method, has been proposed by Onyutha (2016a, b, 2021) to identify sub-trends and analyze separately in order to support the nature of the change in hydrological time series.

In this study, precipitation data of seven meteorological stations belonging to the Seyhan Basin, located in the Mediterranean Region, which is shown as the most sensitive and vulnerable region to global warming, for the period 1970–2019 were used.

The main purpose of this article is:

  • 1. To examine the characteristics of the PCI, SI and MFI indices on an annual and seasonal scale based on monthly precipitation for 50 years in the Seyhan Basin,

  • 2. To investigate the spatial and temporal variation of both precipitation and the studied indices,

  • 3. To examine changes of the obtained index values the statistically with the Mann–Kendall test and also statistical–graphical with the Onyutha's Test.

Although it has been studied by some researchers on various subjects, as far as we know, there is no such study reported in the literature dealing comprehensively with examination of the temporal and spatial changes of the precipitation amount, concentration, seasonality and aggressiveness characteristics in the Seyhan Basin. In addition, the absence of a study on the sub-trends analysis of both precipitation data and of the used indices using the Onyutha's test, which is a fairly new method in the Seyhan Basin, is considered as the other novelty of this study. The results to be obtained are also important in terms of measures that can be taken against the negative effects of climate change in the Seyhan Basin, where agricultural activities are quite intense and where different climate structures are observed.

2 Study area and data

The Seyhan Basin is located between 36°30' and 39°15' north latitudes and 34°45' and 37°00' east longitudes in the Eastern Mediterranean (Fig. 1). It covers 2.82% of Turkey with an area of 22,042 km2 and has the second largest river drainage basin in the Eastern Mediterranean after the Nile. It extends to Ceyhan in the east, Konya and Berdan in the west, Develi Basin and Kulmaç Mountains in the north, and the Mediterranean Sea in the south. The basin includes the catchment areas of the Seyhan River and its tributaries Göksu and Zamantı. The Seyhan River, which empties into the Mediterranean, has a length of 560 km in the western part of Çukurova (SYGM 2019). The elevation and slope in the basin decrease from north to south. Although plains dominate the southern part of the basin, it generally has a mountainous appearance in terms of landforms. It extends from zero elevation on the Mediterranean coast in the south to the heights of 3500 m in the Taurus Mountains in the north. In the basin, there is a mountain system called the Middle Taurus, which is located in the northern and northeastern parts of Adana province and starts in the middle part of the basin. In the basin, which has the most fertile agricultural lands (about 41% of the basin) in Turkey and Europe, 78% of the agricultural lands are dry farming and the rest is irrigated agriculture (TUBITAK-Marmara Research Center 2010). Depending on the topographic characteristics of the Seyhan Basin, the Mediterranean climate is commanding in the lower part of the basin and the continental climate is dominant in the middle and upper parts of the basin. Annual precipitation in the basin varies between 400 and 1000 mm (Gumus 2019).

Fig. 1
figure 1

The location of the Seyhan Basin in Turkey and the location of the meteorological stations used on the Seyhan Basin

Full size image

The monthly precipitation data used in this study belong to 7 stations operated by the State Meteorological Service and cover the period 1970–2019. The recordings of the seven stations used were selected according to data length and the absence of missing data. The longest period without missing data is the period 1979–2020. In addition, there was no change of location at the stations during this period. The location of the stations used in the basin is shown in Fig. 1. In Table 1, basic information about the stations used and some statistical features of monthly precipitation data for the period 1970–2019 are given. The density ridgeline plots of the monthly precipitation data of the stations are also shown in Fig. 2. Due to the difference in its topography, there are different climatic characteristics in the Seyhan Basin. The humid air masses brought by the frontal systems entering the Seyhan Basin precipitation area from the south and southwest leave a significant amount of precipitation to this region with the orographic effect of the Taurus Mountains it encounters while moving over the area. Therefore, the amount of precipitation in the southern part of the precipitation area is higher than in the other parts. As the humid air that crosses the Taurus Mountains moves toward the north, it leaves less precipitation to the northern parts of the precipitation area due to the decreased moisture it carries. The Mediterranean climate is dominant in the south of the basin (Adana–Karaisalı–Karataş) where the altitude is low (coastal), and the continental climate is dominant in the middle and upper parts of the basin with high altitudes (partly mountainous). Depending on this situation, it is observed that there is a decrease in precipitation amounts in the basin from south to north (Table 1, Fig. 2). The region to the southwest of the Seyhan Basin (Ulukışla-17906) has a microclimate under the influence of the Mediterranean climate (SYGM 2019).

Table 1 Descriptive statistics of monthly precipitation data
Full size table
Fig. 2
figure 2

Density ridgeline plots for monthly total precipitation

Full size image

As illustrated in Table 1, monthly total precipitation in the basin varies between 0 and 639.7 mm for the period of 1970–2019. The average, maximum and standard deviation values are higher at Karataş, Adana and Karaisalı stations in the south of the basin compared to other stations in the basin (Table 1). According to Table 1, the stations with the highest number of non-rainy months (non-rainy days) are Adana and Karataş stations with the values of 10.83% (79%) and 15% (80%), respectively.

Skewness and actual excess kurtosis values are also given in Table 1 for each station. Skewness and kurtosis are the most important indicators of the distribution of the data set. Skewness measures how much the distribution deviates from the symmetry around the mean. That is, it shows the asymmetry of the probability distribution. The skewness of the normal distribution is 0. Negative (positive) skewness indicates left (right) skewed distribution. If the skewness is between ±1, it is interpreted that the distribution does not deviate significantly from the normal distribution (Bono et al. 2019; Xu and Shang 2018; Aawar et al. 2019; Hawkins 1980). On the other hand, kurtosis is an indicator that expresses the tailedness of the distribution of the data set compared to the normal distribution, that is, how often outliers occur. There are three types of kurtosis: mesokurtic, leptokurtic and platykurtic. The normal distribution has a kurtosis value of 3 (actual excess kurtosis is 0) and is called mesokurtic. Leptokurtic is a distribution type with a kurtosis value >3 (actual excess kurtosis >0). It indicates that the data are heavy-tailed (more outlier value) and longer center peak relative to the normal distribution. Platykurtic is a distribution type with a kurtosis value <3 (actual excess kurtosis <0). There is a distribution with less-tailed (outliers are rare) and a lower center peak than the normal distribution (Aawar et al. 2019; Larasati et al. 2018). From the above explanations, as can be seen from Table 1, the skewness values of the monthly total precipitation values observed at all stations vary between 0.81 and 2.38, and there is positive skewness in all of them. Actual excess kurtosis values are also positive (between 0.28 and 9.55), and the monthly total precipitation values for all stations have a leptokurtic distribution. In particular, the excess kurtosis values of Karataş, Adana and Karaisalı stations in the south of the basin are quite high and there are heavy-tailed. On the other hand, monthly total precipitation values of Tomarza station shows the closest distribution to the normal distribution with skewness =0.81 and actual excess kurtosis =0.28. The high skewness and high actual excess kurtosis in the monthly total precipitation of Karataş, Adana and Karaisalı stations in the south of the basin (Table 1) is also clearly seen in the ridgeline graphs in Fig. 2. The fact that the monthly average precipitation values (and maximum precipitation values) at these stations are higher than the other stations indicates that there is a risk of instantaneous flooding in the regions where these stations are located.

3 Methodology

The methodological framework of this study is presented in Fig. 3.

Fig.3
figure 3

The methodological framework of the study

Full size image

4 Precipitation concentration index

The precipitation concentration index (PCI), first proposed by Oliver (1980) and later developed by de Luis et al. (2011), is used to evaluate the distribution of precipitation time series. PCI, which is used to determine the temporal distribution of precipitation, is also a powerful indicator for estimating drought (or flood) risk and determining precipitation erosivity. Annual PCI and seasonal PCI (SPCI) are calculated with the following equations.

$${\text{PCI}} = \frac{{\mathop \sum \nolimits_{i = 1}^{12} P_{i}^{2} }}{{\left( {\mathop \sum \nolimits_{i = 1}^{12} P_{i} } \right)^{2} }} \times 100$$
(1)
$${\text{SPCI}} = \frac{{\mathop \sum \nolimits_{i = 1}^{3} P_{i}^{2} }}{{\left( {\mathop \sum \nolimits_{i = 1}^{3} P_{i} } \right)^{2} }} \times 25$$
(2)

Here Pi represents the monthly precipitation amount in the ith month. For analysis at the seasonal scale monthly precipitation data are partitioned as follows: Winter season months (December, January, February), spring season months (March, April, May), summer season months (June, July, August) and autumn season months (September, October, November). The classification range of precipitation according to PCI values is given in Table 2.

Table 2 Classification range of PCI values (Oliver 1980)
Full size table

5 Seasonality index

Seasonality Index (SI), which is used to analyze the seasonality of precipitation and proposed by Walsh and Lawler (1981), is a function of monthly and annual total precipitation and is calculated by Eq. 3.

$${\text{SI}}_{i} = \frac{1}{{P_{i} }}\mathop \sum \limits_{n = 1}^{12} \left| {X_{ni} - \frac{{P_{i} }}{12}} \right|$$
(3)

where Xni is the precipitation for n months of year i and Pi is the annual total precipitation for year i. The classification of precipitation regimes according to the SI values obtained is given in Table 3.

Table 3 Seasonality index (SI) classes (Walsh and Lawyer 1981)
Full size table

The relative seasonality of precipitation indicates the degree of variation in monthly precipitation throughout the year and considers seasonal variations in precipitation, rather than months being "dry" or "wet" in absolute terms (Walsh and Lawyer 1981). The determination of changes in seasonal precipitation regime and changes in SI is essential for agricultural planning.

6 Modified Fournier index

Precipitation erosivity is the erosive force of precipitation and depends on precipitation intensity and amount (Wischmeier and Smith 1978). Knowing the spatial and temporal variability of precipitation is important for the development of methods for estimating erosion-related risks (Di Lena et al. 2021). The Fournier index, which was first proposed by Fournier (1960) to examine the erosive effect of precipitation, was later developed by Arnoldus (1980) as the Modified Fournier Index (MFI). MFI is an index that considers the relationship between transported material, climate data and topographic features. On an annual and seasonal scale, MFI is calculated by Eqs. (4) and (5), respectively.

$${\text{MFI}} = \frac{{\mathop \sum \nolimits_{i = 1}^{12} P_{i}^{2} }}{{\mathop \sum \nolimits_{i = 1}^{12} P_{i} }}$$
(4)
$${\text{SMFI}} = \frac{{\mathop \sum \nolimits_{i = 1}^{3} P_{i}^{2} }}{{\mathop \sum \nolimits_{i = 1}^{3} P_{i} }}$$
(5)

According to the calculated MFI values, the precipitation erosivity was classified as follows: MFI < 60 (markedly low), MFI = 60 – 90 (low), MFI = 90 – 120 (moderate), MFI = 120 – 160 (high), and MFI ≥ 160 (very high (VH) (Nunes et al. 2016). Higher MFI values indicate greater aggressiveness, while lower MFI values indicate lower aggressiveness of precipitation (Elagib 2011).

7 Mann–Kendall test

The Mann–Kendall (MK) test is a nonparametric test and recommended by the World Meteorological Organization (WMO) (Mann 1945; Kendall 1975). In the MK test, which can also be used in the presence of missing data and does not have to comply with a certain distribution of the data, the order of the data is taken as the basis rather than the size (Kahya and Kalayci 2004). The MK statistic S is calculated by the following equation.

$$S = \sum\limits_{i = 1}^{n - 1} {\sum\limits_{j = i + 1}^{n} {{\text{sgn}} (x_{j} - x_{i} )} }$$
(6)

where n is the number of data. sgn is the sign function and it is found by comparing the values of xj and xi as follows.

$${\text{sgn}} (x_{j} - x_{i} ) = \left\{ {\begin{array}{*{20}c} { + 1} & {{\text{if}}} & {(x_{j} - x_{i} ) > 0} \\ 0 & {{\text{if}}} & {(x_{j} - x_{i} ) = 0} \\ { - 1} & {{\text{if}}} & {(x_{j} - x_{i} ) < 0} \\ \end{array} } \right.$$
(7)

The variance of S is calculated by Eq. (8),

$$Var(S) = \frac{{n(n - 1)(2n + 5) - \sum\nolimits_{i = 1}^{k} {t_{i} (t_{i} - 1)(2t_{i} + 5)} }}{18}$$
(8)

where k is the tied groups representing observations having the same value, ti is the number of data values n the ith group and symbol (∑) represents the summation of all the tied groups. The standardized MK statistic (ZMK) value is calculated by Eq. (9),

$$Z_{{{\text{MK}}}} = \left\{ {\begin{array}{*{20}c} {\frac{S - 1}{{\sqrt {{\text{Var}}(S)} }}} & {{\text{if}}} & {S > 0} \\ 0 & {{\text{if}}} & {S = 0} \\ {\frac{S + 1}{{\sqrt {{\text{Var}}(S)} }}} & {{\text{if}}} & {S < 0} \\ \end{array} } \right\}$$
(9)

The ZMK test statistic is compared with the critical value (zcritical) at the selected α significance level to determine whether there is a trend in the time series. If |z ≤ zcritical, then there is no significant trend in the time series, otherwise there is a significant trend. The positive (negative) ZMK value represents the increasing (decreasing) trend (Özel et al. 2004; Kahya and Kalaycı 2004; Bhatti et al. 2020; Shawul and Chakma 2020). In this study, the existence of a trend in time series are examined at α = 0.05 significance level (zcritical = 1.96). The presence of a significant trend in the data set is also made by comparing the P-value in the standard normal distribution table, which corresponds to the calculated Z value, with the α significance level. If the data set's p-value is less (greater) than α significance level, the trend is significant (not significant) and H1 (Ho) is accepted.

7.1 Detect serial correlation and pre-whitening procedure

Serial correlation is the situation in which consecutive values in the time series affect each other. In particular, the presence of a positive serial correlation causes a more significant trend at a certain level of significance than it actually should be (Van Storch and Navarra 1999). Prior to the Mann–Kendall test can be applied, it should be checked whether there is a serial dependency in the series. If there is a serial correlation at certain significance level, it should be eliminated, and the Mann–Kendall test should be applied later (Van Storch and Navarra 1999). Autocorrelation analysis is performed to determine the serial dependency of the data set. The lag-1 autocorrelation coefficient (r1) is calculated with the following equation:

$$r_{1} = \frac{{\frac{1}{n - 1}\mathop \sum \nolimits_{i = 1}^{n - 1} \left( {x_{i} - \overline{x}} \right)\left( {x_{i + 1} - \overline{x}} \right)}}{{\frac{1}{n}\mathop \sum \nolimits_{i = 1}^{n} \left( {x_{i} - \overline{x}} \right)^{2} }}$$
(10)

where xi is the series values, \(\overline{x }\) is the sample mean, and n is the sample size. The serial correlation of time series is determined using following equations:

$$\frac{{ - 1 - 1.645\sqrt {\left( {n - 2} \right)} }}{n - 1} \le r_{1} \le \frac{{ - 1 + 1.645\sqrt {\left( {n - 2} \right)} }}{n - 1}$$
(11)

If the r1 value is inside the above interval, there is no serial correlation in the time series; otherwise, is serial correlation in the observations. Various approaches such as trend-free pre-whitening (Yue et al. 2002) and pre-whitening (Von Storch and Navarra 1999) have been proposed to eliminate the serial dependency that is significant at the α significance level in the data sets. In this study, the pre-whitening was applied to time series with serial correlation (Eq. 12). The new time series (yi) obtained by applying the pre-whitening procedure has one less data than the original time series.

$$y_{i} = x_{i + 1} - r_{1} x_{i}$$
(12)

8 Onyutha’s trend test

Consider that dy is the rescaled version of the given data Y such that dy = n–w–2t where n is the sample size, w denotes the number of times a data point appears within the data set and t is the number of times a given data point is exceeds other observations (Onyutha 2021). The standard rescaled series (ey) can be given by

$$e_{y,i} = d_{y,i} x\sqrt {\left( {n - 1} \right)x\left( {C_{y} } \right)^{ - 1} }$$
(13)

where

$$C_{y} = \mathop \sum \limits_{i = 1}^{n} \left( {d_{y,i} } \right)^{2}$$
(14)
$$T = \sum\limits_{j = 1}^{n} {\sum\limits_{i = 1}^{j} {e_{y,i} } }$$
(15)
$$V(T) = \frac{{n(n^{2} - 1)}}{12}$$
(16)

T is the test statistic of Onyutha's trend test and V(T) is expressed as the variance of the T statistic. In order to detect sub-trends in the data set, it is necessary to obtain the aj series. The temporal variability of the aj series is similar to a Markov process. The aj series is obtained by Eq. 17. When the obtained series (aj) is plotted against the observation period, the temporal break points (sub-trend plot) of the data set can be observed.

$$a_{j} = \mathop \sum \limits_{i = 1}^{j} e_{y,i} \, {\text{for}} \, 1\, \le \,j\, \le \,n$$
(17)

In order to examine the significance levels of the trends, qk for graphical analysis and cumulative sum of difference Statistic Z (CSDz) series for statistical and graphical analysis were obtained. If the qk series falls below 100(1–α) % confidence interval limits (CILs), there is no sub-trend Eqs. (18–20). However, if the CILs are exceeded, there is a sub-trend. In this study, the time window was determined as 10 years. A window of 10 years is vital to analyze decadal variability in the climatic conditions. Accordingly, CSDz statistics were calculated.

$$q_{k} = \mathop \sum \limits_{i = 1}^{k} a_{j} \;{\text{for}}\; \, 1\, \le \,k\, \le \,n$$
(18)
$${\text{CIL}}_{s} = \mp \left| {Z_{{{\raise0.7ex\hbox{$\alpha $} \!\mathord{\left/ {\vphantom {\alpha 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}}}} } \right|x\sqrt {V^{F} \left( T \right)}$$
(19)
$$Z = \frac{T}{{\sqrt {V\left( T \right)} }}$$
(20)

The standardized version of the T statistic is expressed as Z. If T > 0 there is a positive trend, if T < 0 there is a negative trend. VF(T) is the version of V(T) which is corrected from the influence of persistent fluctuations in the data (Onyutha 2021). The CSD-based variability analyses tool (CSD–VAT) was used to implement the Onyutha test, a relatively new and powerful trend method (Onyutha 2020). See references for more information and access to CSD–VAT (Onyutha 2020, 2021, 2022).

9 Results and discussion

9.1 Spatial variability and trend analysis of annual total precipitation

The spatial distribution of the annual average precipitation in the basin for the period 1970–2019 is given in Fig. 4a. It is seen from Fig. 4a that the annual average precipitation has the highest values in the lower part of the basin and the lowest values in the other parts of the basin (partly mountainous). The statistical characteristics ​​of the annual total precipitation data are available in Fig. 4b. The distribution of these values in the basin also decreases from south to north, similar to the precipitation distribution (Fig. 4b).

Fig. 4
figure 4

a Spatial distribution of annual average precipitation, b statistical characteristics of annual total precipitation and c annual total precipitation time series

Full size image

According to Fig. 4b, annual average precipitation values in the basin vary between 320 and 770 mm. The annual total precipitation time series are given in Fig. 4c. In the precipitation time series for the period 1970–2019 in Fig. 4c, it can be clearly observed that the precipitation amounts at the stations located in the south are both higher and have a more skewed distribution compared to the others. According to Fig. 4b, the amount of precipitation at station 17936 (located on the southern slopes of the Taurus Mountains) is higher than at stations 17351 and 17981, which are closer to the southern coast. This situation may have been caused by the humid air mass coming from the Mediterranean to the region where station 17936 is located, rising along the mountain slope and leaving orographic precipitation.

Before applying the MK test to the annual total precipitation data of seven stations, it was examined whether there was a serial correlation. For this purpose, the autocorrelation coefficients (r1) calculated for each station remains between the limit values (−0.253 ± 0.212) calculated with Eq. 11. As a result, there was no serial correlation in any station data, and the MK test was applied directly to the annual precipitation totals. The results of the MK and Onyutha's trend tests are given in Table 4 and Fig. 5a–b, respectively. Trend analyses were evaluated at α = 0.05 significance level. The results obtained show significant spatial differences depending on the variable topographic structure of the basin. As shown in Table 4, the MK results revealed both decreasing and increasing trends in annual total precipitation. Positive values (S and ZMK) are more dominant at low altitude stations (on or near the beach) in the south of the basin, which is under the influence of the Mediterranean climate regime, but these increase trends are no statistically significant. Of the stations where an insignificant increase trend is obtained, only 17906 (Ulukışla) is located in a high altitude and mountainous region, and this station has a microclimate under the influence of the Mediterranean climate. There is a decreasing trend in the annual precipitation data of the three stations located in the northern part of the basin with a partly mountainous topography. Among these decreasing trends, a significant negative trend (ZMK = |−2.08| >Zcritical = 1.96, pvalue = 0.037 < 0.05) was found at the level of 0.05 (the bold numbers) in the annual precipitation data for only Sarız (No: 17840) station. The linear regression equations given in Fig. 4c are also similar to the MK test results. According to the regression equations, there is an increase trend in the first four stations and a decreasing trend in the remaining stations for the period studied.

Table 4 MK trend results of annual total precipitation
Full size table
Fig. 5
figure 5

The sub-trend graphs of annual precipitation time series in terms of a the cumulative effect of temporal variation and b the importance of sub-trends

Full size image

The results given in Table 4 for the period 1970–2019 are compared with another study (SYGM 2019) conducted for the period 1970–2016 in the basin. The MK test statistics of both studies show that there is a general increase trend in annual precipitation totals in the last three years (2017–18–19). The fact that the insignificant decreasing trend of precipitation obtained in Karataş and Karaisalı in SYGM (2019) caused an insignificant increasing trend in this study, and the significant decreasing trend in Pınarbaşı causing an insignificant decreasing trend is an indicator of the increase in annual precipitation totals in the basin in the last three-year period. In Fig. 4c, the increase in annual total precipitation after 2016, especially in the first three stations, is clearly seen.

In Fig. 5, sub-trend graphs are given in terms of the cumulative effect and significance level of the temporal variation of the total precipitation of seven stations in the Seyhan basin. In Fig. 5a, the presence of the sub-trend is graphically detected. According to Onyutha (2021), the point where the cumulative effect of temporal variation changes slope indicates the existence of a sub-trend. When Fig. 5a is analyzed from this point of view, the change point is determined as 1997 for 17351, 1994 for 17802, 1995 for 17837, 1996 for 17840, 1995 for 17906, 1995 for 17936 and 1990 for 17981, respectively. Figure 5b is drawn considering the fact that the year 1995 was determined as the common breaking point throughout the basin. According to Fig. 5b, MK and Onyutha trend test showed similar behavior in general. However, while the significance status changed in some sub-trends, the second term (after 1995) sub-trend was found to be insignificantly negative in contrast to the general trend at station 17936. 17351 and 17840 stations reflected the behavior of the MK test in both sub-periods. While the direction of the trend is the same at other stations, the significance differs. Based on the information obtained, it is possible to say that sub-trend analysis allows more temporal inferences. While the negative trend is dominant in the north of the basin, the positive trend is dominant in the south of the basin. This situation may cause the reservoirs of the water structures in the north to be adversely affected. In addition, there is a risk of flooding in the low-altitude southern parts of the basin.

10 Spatial variability and characteristics of the PCI, SI and MFI at annual scale

The statistical characteristics of the annual PCI, SI and MFI values obtained from monthly precipitation data of the stations studied for the period 1970–2019 are shown in Fig. 6, and the spatial distribution of these indices is shown in Fig. 7. As shown in Fig. 6a, annual PCI values vary between 10.16 and 30.57%. The minimum PCI values are close to each other in almost all stations and have a moderate precipitation distribution. Maximum PCI values show strong irregular precipitation distribution at the stations in the south (Karataş-17981, Adana-17351, Karaisalı–17936) and southwest (Ulukışla-17906) of the basin, and irregular precipitation distribution at the other stations (north of the basin). Average PCI values have irregular precipitation distribution in the north of the basin and moderate precipitation in the other part (Figs. 6a, 7a). In Fig. 6a, it is seen that the PCI values at the stations in the south of the basin have higher standard deviation values than the other regions in the examined period.

Fig. 6
figure 6

The statistical characteristics of a PCI, b SI and c MFI calculated for 1970–2019 period

Full size image
Fig. 7
figure 7

Spatial distribution of annual PCI, SI and MFI values

Full size image

Annual SI values vary between 0.33 and 1.21 (Fig. 6b). The behavior of the annual SI and PCI values in the basin is similar at almost all stations (Fig. 7a, b). As with PCI values, SI values are also higher at stations in the south of the basin (coastal region) and lower at other stations (partly mountainous region). The SI mean values (as illustrated in Fig. 6b) show a "markedly seasonal with a long dry season" precipitation regime at two of the seven stations (17981 and 17351) and a "seasonal" precipitation regime at the other stations.

In Fig. 6c, it is seen that the annual MFI values vary between 22.29 and 386.01. According to the minimum MFI values, precipitation at all stations shows a significantly lower risk of erosive. MFI values of stations 17981, 17351 and 17936 located in the south of the basin have very high standard deviation and these stations are in a very high erosivity risk class with maximum MFI values >160. As illustrated in both Figs. 6c and 7c, according to the average MFI values, there is a high erosivity risk in the south of the basin (except 17351) and a moderate and low erosivity risk in other regions.

Table 5 shows the number of years in the defined class ranges of the annual PCI, SI and MFI indices in the 1970–2019 period (50 years). The number of years given in Table 5 clearly reveals the effect of the topographic structure of the basin on PCI, SI and MFI values. The year numbers of PCI, SI and MFI are concentrated in the higher class range in the low altitude regions of the basin compared to the high altitude regions. According to the Oliver (1980) classification range, there is no uniform precipitation distribution at any station in the annual PCI values obtained over the 50-year period studied. According to annual PCI values, precipitation distributions generally show an increase from moderate to strong irregular precipitation distribution at low altitude stations (south of the basin) depending on the topography of the basin during the 50-year observation period. On the other hand, in the north and southwest stations of the basin, where the altitude is around 1500 m, there prevails mostly moderate and irregular precipitation distribution. At high altitude stations (except 17906) there is no strong irregular distribution of precipitation in any year. There is a strong irregular precipitation distribution in only one year (1986) at station 17906.

Table 5 Number of years in the class range of annual PCI, SI and MFI
Full size table

As illustrated in Table 5, there are hardly any years with SI values below 0.40. The year numbers for the SI in Table 5 show that precipitation regimes of coastal region are generally “seasonal” and “markedly seasonal with a long dry season.” In the mountainous regions of the basin, the SI values show that the precipitation regimes are more dominant in the "Rather seasonal with a short drier season" and "seasonal" character. In the stations in partly mountainous regions, there is no precipitation regime of “Most precipitation in <3 months” and “Extreme seasonality, with almost all precipitation in 1–2 months” in any year. Station 17981 is the station where SI values between 1.1 and 1.2 (most precipitation in < 3 months) are the most (In 12 years of the 50-year period). The extreme precipitation regime (SI ≥ 1.20), in which almost all precipitation occurs within one to two months in the basin, was observed only at stations 17981 (in 2010) and 17351 (in 2004). The southern parts of the basin, where the stations with higher SI values are obtained, are the regions where agricultural activities are intense. Considering that high SI values represent the precipitation regime in which most precipitation occurs within a few months (2–3 months), precipitation in the southern part of the basin is likely to pose a risk for agricultural activities.

The MFI results show (as illustrated in Table 5) that the aggressiveness of precipitation in the Seyhan Basin is strongly influenced by the annual amount of precipitation depending on the geographical feature of the basin. MFI values ​​increase with both annual total precipitation and concentration and seasonality. In this context, there are more years in which MFI values ​​are in the high class range in the stations in the south of the basin, which have higher precipitation, PCI and SI values ​​compared to other regions. Annual MFI values ​​obtained during the 1970–2019 observation period at stations located in the upper part of the basin (altitude approximately 1500 m) are generally categorized as markedly low (MFI < 60) and low (MFI = 60–90) (Fig. 7c, Table 5). MFI values were generally obtained as moderate (90–120), high (120–160) and very high (>160) at low altitude stations with high precipitation in the south of the basin. Precipitation of stations 17981 and 17936 potentially carries a high risk of erosion. Annual MFI values for station 17981 are high in 34% of the 50-year period and in the very high erosion risk class in 40%. For the station 17936, they are in the high erosion risk class in 26% and very high in 28% (Table 5).

11 Trend analysis of the PCI, SI and MFI at annual scale

It was determined that there was no serial correlation for any station data in the serial correlation analysis performed to apply the MK test to annual PCI, MFI and SI values. MK trend analysis results of annual PCI, MFI and SI indices are given in Table 6. According to α = 0.05 significance level, there is an insignificant positive trend (IPT) in the PCI values at stations 17981, 17351, 17936 located in the south of the basin and station 17840 in the north. Insignificant negative trends (INT) were obtained at other stations. It can be seen from Table 6 that the trends of the annual SI values show similar patterns with the annual PCI trends. As in annual PCI, the MK results in SI revealed insignificant decreasing and increasing trends at the 0.05 level. While the most positive trend in PCI (SI) was obtained at station 17981 with a value of z = 1.94 (1.47), the most negative trend was observed at station 17802 with a value of z =  −1.41 (−1.11). Both high precipitation totals and high PCI (SI) increase the likelihood of heavy precipitation and the resulting flood risk (Li et al. 2017). In line with this explanation, the fact that there is a statistically insignificant increase trend in annual precipitation and annual PCI values at stations 17981, 17351 and 17981 (Fig. 5a, b) where the annual precipitation in the basin is high (Table 4 and Table 6) poses a risk in terms of flooding in these regions. In station 17840, where an insignificant increase trend is obtained in PCI and SI, has a lower monthly (42.22 mm) and annual precipitation average (507 mm) compared to the stations with other insignificant positive trends (Table 1 and Fig. 4a–b). Although a significant decrease trend of 0.05 level was obtained in the annual total precipitation values of this station (as shown in Table 4), the fact that there was an insignificant increase trend in PCI and SI indicates that more dry periods may be seen in this region.

Table 6 MK trend results of annual PCI, SI and MFI values
Full size table

According to the MK trend results given in Table 6 for the annual MFI, there is a decreasing trend at the stations in the north of the basin and an increasing trend at the remaining stations. None of the positive trends are statistically significant. Negative trends obtained only in annual MFI values of stations 17840 and 17802 are significant at 0.05 significance level. At these stations, MK test statistics were obtained as z =  −2.10 and z =  −2.23, respectively. Among the stations located in high altitude regions, only station 17906 showed an insignificant positive trend in annual MFI values. The MK results for the annual PCI, SI and MFI reveal a clear difference in the behavior of stations near the coast in the south of the basin and stations in the rest of the basin. There is an altitude difference of approximately 1500 m between the stations in the south and north of the basin. This difference in the topography of the basin brings with it climatic differences. Because the south of the basin is under the influence of the Mediterranean climate and the north is under the influence of the continental climate. Even though Ulukışla (station 17906) is at a high altitude, it has a microclimate feature due to its proximity to the coastal region which has the Mediterranean climate regime. This situation shows itself in the trend results obtained for annual precipitation, PCI, SI and MFI given in Table 4 for station 17906.

In Fig. 8, sub-trend graphs of annual PCI, SI and MFI time series are given. PCI did not show any behavior that exceeded the level of significance throughout the study period. However, the difference in the behavior of the stations located in the north and south of the basin draws attention. While the stations located in the north of the basin (17837, 17840 and 17802) generally have a decreasing trend, the stations located in the south show the opposite trend. The general trend aspects of the sub-trend analysis are in agreement with the MK test. However, when the time series is examined, there are trends in the opposite direction of the general behavior, albeit in short periods. The fact that the Z values are between CILs indicates that there is no significant trend. The annual PCI sub-trend time series revealed that precipitation became more and more irregular in the coastal parts of the basin, while it became more regular in the northern parts where the continental climate is dominant. However, the decrease in precipitation in the north may have caused the PCI to show a more regular precipitation regime (Fig. 5b). Compared to the annual SI sub-trend time series MK test, the two methods deny the existence of a significant trend. It is noteworthy that between the years 1990–1995, stations other than 17837 station showed insignificant decreasing trend behavior. The sudden increase trend that follows indicates that the annual total precipitation falls to the ground in a shorter time. Similarly, in recent years, SI has been in an increasing trend at all stations except 17802. This situation poses a significant danger to the basin. When the annual MFI sub-trend time series is examined, 17840, similar to the MK test, showed a significant decreasing trend. In addition, 17802 showed decreasing trend behavior in both methods. However, while it was significant in the MK test, it was detected as a nonsignificant trend in the Onyutha test. From the beginning of the study period to 1985, there is a decreasing trend in the basin. While the Z statistic oscillated around zero until the 2010s, then a sudden increase draws attention. It has been observed that the precipitation in the basin has been more aggressive in recent years.

Fig. 8
figure 8

Sub-trends of annual PCI, SI and MFI values

Full size image

12 Spatial variability and characteristics of the PCI and MFI at seasonal scale

The spatial distribution of the SPCI values for the 1970–2019 precipitations in the Seyhan Basin is shown in Fig. 9. The variation range of SPCI values in each season is as follows: 11.25–14.28 for autumn, 9.74–11.47 for winter, 9.60–12.64 for spring and 15.36–19.76 for summer. As seen in Fig. 9, the lowest SPCI values are seen in winter. In general, while SPCI has lower values in winter than in autumn, it increases again from spring to summer. As shown in Fig. 9, the SPCI values for the autumn season are moderate throughout the basin, while for the winter and spring seasons, a uniform precipitation distribution in the north of the basin and a moderate precipitation in the south is dominant. The summer season SPCI values show that there is an irregular precipitation distribution (16 ≤ SPCI ≤ 20) in the basin in this season. SPCI values spatially decrease also from south to north of the basin in all four seasons.

Fig. 9
figure 9

Spatial variation of SPCI values for each season

Full size image

The spatial distribution of Seyhan Basin SMFI values for the period 1970–2019 is given in Fig. 10. SMFI values in the basin show an opposite behavior compared to SPCI values in some seasons. While the winter season with the lowest SPCI values was the season with the highest values in SMFI, the summer season with the highest SPCI values was the season with the lowest SMFI values. SMFI values for autumn in the basin range from 34.04 to 132.01. According to Fig. 10, only the precipitation of station 17981 in the south of the basin poses a high risk of erosion (120–160), while the precipitation of other stations have generally markedly low and low erosion risk in this season. Although there was an increase in winter SMFI values compared to the autumn season, there was no change in the erosion class (markedly low and low) represented by the precipitation of stations located in high altitude regions. The most significant change in winter compared to autumn was obtained at stations 17936 and 17351 located in the south of the basin. The precipitations of stations no 17936 and 17351, which have low erosion risk in autumn, reached very high and high erosion risk levels in winter, respectively. The precipitation of station 17981, which poses a high erosion risk in autumn, is in the category of very high erosion risk in winter. As shown in Fig. 10, SMFI values are generally markedly low and at low erosivity level in the basin in the spring season. Only at station 17936 precipitation erosivity is moderate. In summer, there is a significantly low (<60) precipitation erosivity in the whole basin. The spatial distribution of SMFI values in the basin generally shows lower values in the north of the basin and higher values in the south for each season.

Fig. 10
figure 10

Spatial variation of SMFI values for each season

Full size image

13 Trend analysis of the PCI and MFI at seasonal scale

As a result of the serial correlation analyses of SPCI and SMFI values, autocorrelation values (r1) of summer PCI (r1 = −0.255) at station 17981, autumn PCI (r1 = −0.259) and winter MFI (r1 = 0.309) at station 17837 and winter MFI (r1 = 0.284) at station 17840 were found outside of the limit values (−0.253 ± 0.212). Therefore, there is a serial correlation in these index values. Pre-whitening procedure was applied using Eq. 12 to the relevant index values and then they were subjected to the MK test. MK trend test was applied to SPCI and SMFI values in the period 1970–2019 and the results are given in Table 7. The trend results were evaluated at the level of α = 0.05. The bold values in Table 7 represent that the null hypothesis (Ho) is rejected at α = 0.05 significance level, and there is a significant trend as statistically. The first 3 stations in Table 7 are the ones located in the low altitude areas in the south of the basin (close to the coastline), and the other stations are located in the high altitude areas of the basin. As shown in Table 7, about 57% of the trend results for SPCI in all seasons show a decreasing trend and the rest show an increasing trend. However, according to the results obtained, there is a significant decreasing trend (ZMK = −2.65) at the level of α  = 0.05 in autumn only at Karaisalı (No: 17936) station in SPCI. In autumn, SPCI values generally have a decreasing trend in the south of the basin (except 17981), while both increasing and decreasing trends are determined in the north of the basin. In winter, insignificantly increasing trend in Karaisalı and insignificantly decreasing trend was found in other stations. In the spring, insignificant trends were determined, generally increasing in the north (except Pınarbaşı) and decreasing in the south (except Karataş) of the basin. In the summer months, insignificant increasing or decreasing trends were found in approximately half of the stations in SPCI.

Table 7 MK trend results of SPCI and SMFI values
Full size table

According to the results given in Table 7, a decreasing trend was found in approximately 61% of the SMFI values. None of the trends obtained in autumn were significant and generally decreasing trend in the south of the basin and increasing trend were found in the north. Positive trends were found in SMFI at all stations except for station 17840 in winter. However, there is a significant positive trend at the level of α = 0.05 for only 17906 (ZMK = 3.16) in winter season. Negative trends were found in the SMFI values of all stations in the spring, and the trend obtained only for stations 17837 (ZMK =  −2.24) and 17840 (ZMK =  −2.04) is significant at the level of α = 0.05. There is no significant trend in SMFI at any of the stations during the summer months. However, there is an insignificant increasing trend in SMFI values only at station 17840 for summer season and insignificant decreasing trend in other stations.

SPCI sub-trend time series graphs are given in Fig. 11. Autumn PCI, which gave a significant decreasing trend result in the MK test, is also consistent in the Onyutha test. In recent years, precipitation irregularity has increased in all seasons. However, between 1995 and 2005, summer PCI showed a significant increase. According to the autumn PCI sub-trend graph, the basin-wide decrease in 1980 showed a sudden increase after the trend. This spike similarly occurred in the spring PCI sub-trend chart between 1980 and 1990. When SPCI sub-trends were examined, no significant difference was found between the north and south of the basin.

Fig. 11
figure 11

Sub-trends of SPCI

Full size image

SMFI sub-trend graphs are given in Fig. 12. In the Onyutha test of SMFI, a significant (negative) trend was detected only in the spring MFI. The general trend behaviors detected in the MK test are largely consistent with Onyutha trend test. Trend variability in summer is less compared to other seasons. In addition, the increasing trend is striking, consistent with the information obtained from other indices throughout the basin. This shows that precipitation has an increasingly aggressive regime in recent years.

Fig. 12
figure 12

Sub-trends of SMFI

Full size image

14 Discussion and conclusions

In this study, precipitation, precipitation concentration, seasonality and erosivity were investigated as spatiotemporal in the Seyhan Basin in the south of Turkey for the period 1970–2019. Monotonic and sub-trends of precipitation and precipitation indices were examined using the Mann–Kendall and Onyutha's Trend tests, respectively. In the basin, which has a sensitive hydro-meteorological balance due to its geographical conditions, the highest index values in both annual and seasonal scales also exhibit a behavior similar to the distribution of precipitation, and were generally obtained in the south of the basin. In this context, there is an irregular precipitation distribution in the south of the basin and a moderate precipitation distribution in the other parts of the basin in the annual PCI. According to the annual average SI, "seasonal" and "Markedly seasonal with a long dry season" precipitation regimes dominate the basin. As in PCI and SI, annual MFI values are higher in the south of the basin and there is a high risk of erosion. The highest SPCI and SMFI values were also observed in the south of the basin, in all seasons. Summer PCI values have an irregular precipitation distribution. This shows the importance of this season in the annual precipitation regime and the effect of convective processes. However, precipitation in the entire basin in the summer months carries a markedly low risk of erosivity in the SMFI. In SMFI, there is very high erosivity in the south of the basin, especially in winter.

Generally, there is no significant trend as annual and seasonal in precipitation amount and the indices examined during the 50 years according to the MK results. On the other hand, generally, there is a decreasing trend in annual total precipitation and index values in the north of the basin, which has a continental climate, while there is an increasing trend in other regions. The increasing trend in annual precipitation, PCI and SI values indicates a trend toward more seasonal conditions and dry periods (low precipitation and long dry season) (Li et al. 2017). The southern part of the basin, where this situation is observed, is a region where intensive agricultural activities are carried out. Therefore, this situation will cause a decrease in soil water in these regions and the desired yield from harvests will not be obtained. On the other hand, the decreasing trend in precipitation and MFI values, especially in the northern regions of the basin, should not be interpreted as a decrease in the risk of soil erosion in these regions (Vrieling et al. 2010).

The decreasing trend in precipitation in the northern part of the basin will also cause a decrease in water resources. This may require water transfer between basins. The increase in precipitation in the southern part of the basin, where urbanization and agricultural activities are intense, may lead to flood disasters with the increase of urbanization in the future. The tendency of the annual total precipitation to occur in a shorter time in the basin indicates that heavy precipitation may occur. This situation increases the erosion effect of precipitation to dangerous levels.

Onyutha's trend test, which is an innovative method, shows similar behavior to the Mann–Kendall method when considering the entire study period. However, it is vital that Onyutha's trend test provides information about sub-trends. Accordingly, the variation observed in monthly total precipitation around 1995 can be considered as a breaking point. In addition, the sudden decrease and subsequent sudden increase in the concentration and erosivity of precipitation in the 1980s is quite remarkable.

Based on the results obtained, it is very important that agricultural activities, which are carried out intensively in almost the entire the basin, are carried out with efficient water strategies. With this point of view, water potentials in neighboring basins should also be evaluated. It is thought that minimizing water losses and using the existing water efficiently will enable the sustainable use of the precipitation whose regime has become irregular. In addition, the flood risk should be taken into account in the southern regions of the basin where urbanization is more intense. Urbanization should be planned considering this threat.

As a result, examining and interpreting the precipitation changes in a region from various aspects is of great importance in terms of assessment of flood and drought risk, use of water resources, soil and water protection, erosion control studies, water supply and hydraulic structures.