Investigation of precipitation and extreme indices spatiotemporal variability in Seyhan Basin, Turkey Open Access

Precipitation is one of the meteorological events most affected by climate change. Human intervention in water resources and land cover is also increasing with the rising population. This situation causes significant temporal and spatial changes in precipitation (Rysman et al. 2013). The spatial variation of precipitation leads to flood and drought problems in some regions. It also has a significant impact on the economies of countries. The temporal variability of precipitation negatively affects many sectors, especially agriculture. Investigating precipitation on a spatial and temporal scale is vital for the development of sustainable management strategies (van der Pol et al. 2015; Barbosa & Lakshmi Kumar 2016; Mehta & Yadav 2021). The variability of precipitation negatively affects not only eco-hydrological processes but also people's socio-economic lives (IPCC 2021).

Climate change causes severe droughts and extreme natural events around the world. The increasing irregularity of precipitation on a temporal and spatial scale hinders the efficient operation of basin-based planning. Since precipitation variability is very high in arid and semi-arid climatic regions, drought events in these climatic regions may cause greater ecological/economic losses.

It is necessary to observe and evaluate the variability of precipitation to take precautions against disasters and to use precipitation efficiently. At this point, extreme precipitation indices, which are frequently used in determining the characteristics of precipitation, come to the fore (Sillmann et al. 2013; Gao & Shi 2016; Jiang et al. 2016; Khedhaouiria et al. 2020; Saddique et al. 2020). Because of there usefulness, extreme precipitation indices are frequently preferred in hydro-meteorological studies (Zhang et al. 2011; Tongal 2019; Bhatti et al. 2020; Shawul & Chakma 2020). These indices assist in the effective analysis of large-scale precipitation data. In addition, the indices provide very useful information about the trend, variability and other statistical properties of precipitation (Alexander et al. 2019).

Statistical methods are frequently used to determine the characteristics of hydro-meteorological events. Trend analysis is widely used in studies such as the determination of short- or long-term changes among these methods. Non-parametric trend methods such as modified Mann–Kendall (MMK), Mann–Kendall (MK), Spearman's rho (SR), and Sen's T are frequently used in the analysis of hydrological phenomena (Tosunoğlu 2017; Pour et al. 2020; Zhai et al. 2020; Elzopy et al. 2021; Esmaeilpour et al. 2021; Oscar-Júnior 2021). In addition, these methods are a powerful and effective tool in studies examining the variability in hydrological events.

Change point detection is very important in spatial and temporal variability analyses in hydrology (Tongal 2019). With change point detection, it is possible to examine periods with different behaviors that occur over time. Thus, current trends can be clearly examined. In the literature, there are many studies on change point analysis (Pettitt, standard normal homogeneity test, buishand range test, von neumann ratio test etc.) (Jaiswal et al. 2015; Abiy et al. 2019; Boluwade 2020; Ryberg et al. 2020).

Bhatti et al. (2020) analyzed the spatial and temporal changes in extreme precipitation indices in Pakistan using different statistical approaches (MK, sequential MK, Sen's slope estimator, Student's T-test, linear regression). As a result of the analysis, an increase in the spatial distribution of extreme precipitation indices was obtained as a whole. On the other hand, it was determined that extreme precipitation indices tended to decrease in humid regions affected by monsoon and westerlies during the study period. Mehta & Yadav (2021) investigated the trend of rainfall in the northwest region of India with the MK method, and the magnitude of the trends with the Sen's slope estimator. For the 102-year study period (1901–2002), an increasing trend was determined at α = 0.05 significance level in annual, post-monsoon, pre-monsoon and southwest monsoon rainfall, while a decreasing trend was obtained in winter precipitation. Pour et al. (2020), using MMK, robust ITA, Sobol's method and sequential MK, investigated the changes in the reference evapotranspiration (ET_o) data of ten meteorological stations in peninsular Malaysia and the data of six meteorological parameters affecting it. Abiy et al. (2019) analyzed rainfall trends in southeast Florida with linear regression and MK tests, and the presence of a sudden rainfall change with the Pettitt test. Ros et al. (2016) examined the short- and long-term trends of rainfall data for the 1948–2011 period obtained from 50 stations in the Kelantan River Basin in the northeast Malaysian Peninsula with the MK test. In addition, the homogeneity of rainfall time series was investigated by using four absolute homogeneity tests, namely the Buishand range test, standard normal homogeneity test, Pettitt test and von Neumann ratio test.

The Seyhan Basin, which is used as the study area, is located in the south of Turkey. In addition to agricultural activities, industry also plays an important role in the socio-economics of the basin. A large part of the water used in the basin is allocated to drinking and utility water. In addition, irrigation water is allocated from the Seyhan Basin to the irrigation areas outside the basin (General Directorate of Water Management 2017). In the literature, there are studies indicating that the Seyhan Basin is under threat of drought (Topçu & Seçkin 2016; Tuncok 2016; Gumus & Algin 2017; Cavus & Aksoy 2019; Keskiner et al. 2019; Altın et al. 2020). In addition, significant decreasing trends in precipitation, snow water equivalent and flow have been determined in the Seyhan Basin (Gokmen 2016). Keskiner et al. (2019) examined the meteorological drought in the Seyhan Basin with the Standardized Precipitation Index (SPI) and Percent of Normal Index (PM) drought indices using monthly and annual rainfall data for the 1950–2006 period. Altın et al. (2020) conducted a hydrological drought analysis with the Streamflow Drought Index (SDI) in three-, six-, nine- and 12-month periods using monthly streamflow data of eight stations in the Seyhan and Ceyhan Basins. In another study, spatial drought was examined by calculating SPI on a 12-month time-scale using the data of 19 meteorology stations in the Seyhan Basin, and it was determined that different severities of drought were experienced in the basin (Cavus & Aksoy 2019).

The Seyhan Basin is the second largest basin after the Nile among the basins in the Eastern Mediterranean Basin. The Seyhan Basin, which has the most fertile lands in Turkey and Europe, is one of the richest regions in the world in terms of biodiversity, and dry and irrigated agriculture is practiced in most of the basin (Talu & Özüt 2011). The hydrology of the Seyhan Basin is also of great importance for the surrounding basins, as there is water allocation from the Seyhan Basin to the neighboring basins (Kızılırmak Basin and Ceyhan Basin) (General Directorate of Water Management 2017). According to the IPCC (2021), the Mediterranean Basin, where the Seyhan Basin is located, is among the basins that will be most affected by climate change. The spatial and temporal variability of precipitation due to climate change will significantly affect the socio-economic structure and biodiversity of the basin and will cause floods and droughts in the basin.

According to the literature review, as far as we know, it has been determined that there is no study on the Seyhan Basin that comprehensively examines the spatiotemporal variability of precipitation and extreme precipitation indices. Therefore, a comprehensive statistical study examining precipitation is very important for the basin.

The novelty of this study, in addition to the ones mentioned above, lies in the determination of a change point at the basin scale. The change point detection methods used within the scope of the study are generally preferred in the literature for the examination of a station or time series. However, in this study, it is aimed to determine a hydrological change point at the basin scale based on these methods. Thus, it is aimed to provide an up-to-date perspective on the use of single change point determination methods in the literature. In addition, it is thought that examining the hydrological characteristics of the determined sub-periods will add scientific depth to hydrological time series analyses at the point of obtaining information that cannot be obtained from monotonic time series analyses. With such a perspective, it is thought that the investigation of precipitation and extreme precipitation indices for the study area will make important contributions to the science of hydrology in understanding the complex physics of precipitation.

The aim of this study is to determine the change points at the basin scale with change point detection methods, to analyze the precipitation and extreme precipitation indices with the modified Mann–Kendall and Spearman's rho trend tests, and to map and interpret the results. In summary, it is aimed to make a comprehensive statistical analysis of precipitation and extreme precipitation indices in the Seyhan Basin. It is thought that the results obtained may have utility for local administrators and policy makers in developing effective water-use strategies against climate change for the Seyhan Basin.

In this study, daily precipitation data from seven meteorology stations in the Seyhan Basin during 1970–2019 were used. Some information regarding the daily precipitation statistics of the stations used is given in Table 1. According to Table 1, the station with the lowest daily mean precipitation is Ulukışla at 0.877 mm, and the station with the highest is Karaisalı (No:17936) at 2.397 mm. The stations with the lowest and highest daily maximum precipitation are Tomarza (No:17837) and Karaisalı (No:17936) stations at 46.1 and 231 mm values, respectively. The lowest and highest standard deviation values are obtained at Ulukışla (No:17906) and Karaisalı (No:17936) stations with the values of 2.981 and 8.711, respectively. Kayseri/Pınarbaşı station has the lowest skewness value at 4.871, while Adana station (No:17351) has the highest skewness value at 7.430. It is observed that there is a decrease in the amount of precipitation in general from the south to the north of the basin (Table 1).

It is very important to determine the change point in the analysis of time series. The presence of a change point indicates a sudden and significant change in the data generation process. In this study, three methods, namely the Pettitt test (PT) (Pettitt 1979), Buishand rank test (BRT) (Buishand 1982) and standard normal homogeneity test (SNHT) (Alexandersson 1986), were used for change point detection. These three tests are capable of identifying the year in which a break is likely in a time series. SNHT has a characteristic feature in detecting breaks near the beginning and the end of a time series (Wijngaard et al. 2003). In the non-parametric PT, the corresponding order of the data is taken into account instead of the value. The sequencing approach makes the PT less sensitive than the other tests. Unlike SNHT, PT and BRT are more sensitive to detecting breaks in the middle of time series. SNHT and BRT assume that the data are normally distributed, while PT does not require this assumption. In these three methods, the null hypothesis (H₀) indicates that the series has independent and random distribution, while the alternative hypothesis (H₁) indicates that there is a sudden change. The mathematical formulation of SNHT, BRT and PT and their critical values for test statistics are not given here because they are available in many studies in the literature (Pettitt 1979; Buishand 1982; Alexandersson 1986; Ros et al. 2016).

If is larger than ⁠, then the null hypothesis (H₀) is rejected and so, H₁ is accepted. In this test, H₀ represents no-trend while H₁ indicates that the time series has meaningful increasing or decreasing trends. If the Z value is negative (positive), there is a decreasing (an increasing) trend.

If the |z| value is greater than the z_α value (|z| >z_α) determined from the standard normal distribution table at the chosen significance level α, the H₀ hypothesis, which is based on the fact that the observation values do not change over time, is rejected and it is concluded that there is a certain trend (Lehman & D'Abrera 1975; Sneyers 1990).

The Expert Team on Climate Change Detection (ETCCD) defined 27 extreme climate indices to analyze the observed or modeled precipitation and temperature parameters. Eleven of these indices are related to precipitation parameters and eight of these indices (RX1day, RX5day, SDII, CDD, CWD, R1mm, R10mm and R20mm) were examined within the scope of this study (Table 2). RX1day, RX5day and SDII are used to determine extreme precipitation events, while CDD and CWD are used to examine the seasonality of precipitation (Tao et al. 2018). Threshold indices R1mm, R10mm and R20mm are also used to categorize precipitation days (Sillmann et al. 2013). Thus, in addition to the extreme precipitation days, information can be obtained about the number of heavy precipitation and wet days. The indices used were calculated manually in Excel on an annual scale. The descriptions and formulas for each index are found in Zhang et al. (2011).

Within the scope of the study, three different methods were used (PT, BRT, SNHT) to obtain reliable results in the determination of the change point. The application of a single method in the detection of the change point may cause inaccuracies in the detection of the change point (Tao et al. 2018). In station-based change point determination studies, it is not possible to examine behavior at the basin scale. In the examined period, the years in which the change was determined in seven stations according to all three methods were taken into account and the results are given in Table 3.

Table 3

^aFor example, three statistical tests at station 17802 revealed that it was the change point for 1981. The years (1981 and 2002) to which the shaded values belong represent the change point years.

View Large

The values in Table 3 represent the number of statistical methods in which the change is determined in the relevant station and year. For the period 1970–2019, there is no change point in all three statistical methods in the years not given in Table 3. For example, the year 1973 was obtained as the change point in two and one statistical methods at stations 17351 and 17981, respectively. However, there is no change in other stations for this year. At stations 17802, 17906 and 17936, the year 1981 was obtained as the change point in three, one and one statistical methods, respectively. Similarly, the year 2002 was obtained as the change point in one, two and one statistical methods, respectively, at stations 17840, 17936 and 17981. As a result of the evaluation made in this way, the years with the highest change were determined as 1981 and 2002. In line with the results obtained, the period 1970–2019 was divided into three sub-periods: first period 01.01.1970–31.12.1981; second period 01.01.1982–31.12.2002; third period 01.01.2003–31.12.2019.

According to Figure 3, the stations with the highest monthly mean precipitation for all three sub-periods are 17936, 17981 and 17351 (south of the basin), and the stations with the lowest are 17906 (west of the basin), 17837, 17840, 17802 (north of the basin) (Figure 1). It is thought that this difference in the monthly mean precipitation data of the stations is due to the terrestrial climate in the north and west of the basin and the Mediterranean climate in the south. If an evaluation is made for the monthly mean precipitation data in terms of the change between all three periods, stations 17906 and 17981 generally show similar behavior to each other, and the other stations show similar behavior to each other. In five of the seven stations (except 17906 and 17981), there is a decrease in monthly mean precipitation from Period 1 to Period 2. The stations with the highest decrease are stations 17802 and 17840 (Figure 3).

The SR and MMK trend analysis results given in Table 4 for the period 1970–2019 are quite close to each other. According to the trend analysis results obtained, there is a generally insignificant decreasing trend in most of the precipitation indices at the stations in the north of the basin (17802, 17837, 17840) compared with both SR and MMK methods. A significant decreasing trend is obtained in SDII at station 17837 and SDII, R1mm and R20mm indices at station 17840. Increasing trends in precipitation indices are generally determined at station 17906 located in the west of the Seyhan Basin and at stations 17351 and 17981 in the south, and most of them are insignificant increasing trends. There is an insignificant decreasing trend in the CDD and CWD indices at station 17351, and an insignificant increasing trend in the other indices. Significant uptrends are found only in the indices of RX1day, RX5day, SDII and R20mm at station 17906 in the west of the basin. In this station, there is an insignificant decreasing trend in the R1mm index, while an insignificant increasing trend is found in the other indices. According to Table 4, the RX1day, RX5day and R10mm indices generally show similar behavior in the 1970–2019 period. An insignificant decreasing trend is determined at the 0.05 significance level at four stations in the RX1day index and five stations in the RX5day and R10mm indices. The index with the highest increasing trend is R20mm, and there is a decreasing trend in this index only at stations 17802 (insignificant) and 17840 (significant) in the north of the basin, and there is an increasing trend at the other stations.

The results obtained for the three sub-periods examined according to the SR and MMK trend analyses are given in Table 5, and the results based on the number of stations depending on the trend status are given in Figure 5. According to the trend analysis results of the three sub-periods, there is a mostly insignificant increasing trend (62%) in all periods and indices (Table 5 and Figure 5). Also, a 1.5% significant decreasing trend, 28% insignificant decreasing trend and 8.5% significant increasing trend were obtained (Figure 5). Only in the CDD index, a decreasing trend is determined in the seven stations examined in Period 2 and Period 3. RX5day, CDD, SDII, R10mm and R20mm indices show an increasing trend at all stations in Period 1. Significant decreasing trends are obtained in the CDD index (station 17802) in Period 2, and in the CDD and CWD indices (station 17837) in Period 3.

According to Figure 5, most of the significant uptrends are found in Period 1. There is no significant increasing trend in any of the stations and indices in Period 3. Significant or insignificant increasing trends are detected in R1mm and R10mm indices at all stations (Figure 5 and Table 5). Also, an increasing trend is determined in all indices and periods at stations 17936 and 17981. The highest decreasing trend is found at stations 17837 and 17840 located in the north of the basin (Table 5).

The spatial distribution of eight extreme precipitation indices used to determine the effect of climate change on precipitation variability in the Seyhan Basin was investigated. The study was carried out for the whole period and three sub-periods. The inverse distance weight (IDW) method, which is frequently used in the literature, is used to convert point data into spatial information (Fisher et al. 1987).

According to the maps given for the sub-periods (Figure 10(a)–10(c)), CDD values show a decrease from Period 1 to Period 2, and a general increase in CDD values is observed in Period 3 (in regions outside the south of the basin). This is clearly seen both in the box plot for CDD (Figure 6(a)) and in the trend results (Table 5). There is a general decrease in CDD values from Period 1 to Period 3 in the south of the basin. The high CDD values of the basin in the third period indicate agricultural droughts. The increase in precipitation intensity in the basin, where agriculture is intense, not only makes it difficult to control water resources, but also complicates agricultural planning strategies.

According to the CWD map given in Figure 9(b) for the whole period, CWD values vary between 4.56 and 6.52 days, the lowest CWD values are obtained at station 17906 and the highest CWD values are obtained at station 17981. According to Figure 10(d)–10(f), and when looking at the basin as a whole, there is a gradual decrease in general over the three periods. The eastern side of the basin has more consecutive precipitation days compared with the western side. CWD is crucial for maintaining soil moisture and flow dynamics in the watershed (Tongal 2019).

According to the spatial variation of the RX5day index for the sub-periods (Figure 14(d)–14(f)), the values in the northern region of the basin increased from the first period to the third period (191 mm, 203 mm, 222 mm). The larger RX5day values in the southern part of the basin increased from Period 1 (600 mm) to Period 2 (617 mm) and decreased from Period 2 to Period 3 (553 mm), while the value of RX5day at station 17981 increased from the first period to the third period, and at station 17936 it increased a little in Period 2 and decreased again in the third period.

When the spatial variation of the SDII index for the sub-periods is examined, low SDII values are dominant in the north of the basin, and high SDII values in the south, as in RX1day and RX5day (Figure 14(g)–14(i)). The most obvious changes are seen at stations 17840, 17936 and 17981, albeit slightly. Other stations almost preserve their SDII values in all three periods. There is a slight decrease in the SDII value of station 17840 in the second period compared with the first period. At stations 17936 and 17981 in the north of the basin, an increase is observed in SDII values from the first period to the third period (Figure 14(g)–14(i)).

Seyhan Basin is one of the most important basins of Turkey with its rich biodiversity and socio-economic opportunities. In this study, the spatiotemporal variability of precipitation in the Seyhan Basin was investigated with daily data from seven stations for the period 1970–2019. Pettitt, Buishand rank and standard normal homogeneity tests were used to determine the breaking points in the study period with a basin-scale approach. According to the results obtained, 1981 and 2002 were determined as the breaking years.

RX1day, RX5day, SDII, CDD, CWD, R1mm, R10mm and R20mm extreme indices were also investigated in the study as well as precipitation. Then, modified Mann–Kendall and Spearman's rho trend tests were applied to precipitation and extreme indices for all periods and three sub-periods. Significant trends were detected at stations 17837, 17840 and 17906. In the temporal analysis of precipitation, the second period (01.01.1982–31.12.2002) was determined to be a period in which a decrease was generally detected in precipitation and extreme indices.

Finally, the spatial variability of precipitation was investigated with extreme indices. While the number of rainy days decreases in the south of the basin, the intensity of precipitation increases. This situation increases the frequency of sudden heavy rains. Thus, the risk of flooding is increasingly coming to the fore. Looking at the north of the basin, a general decrease in precipitation and an increase in dry days were observed. It is thought that this situation will adversely affect agricultural activities.

The potential for short-term sudden rains to occur is increasing throughout the basin. In this case, a sensitive planning study should be carried out to use water resources efficiently. A significant decrease in the number of consecutive wet days was detected in the west of the basin. Considering the general decreasing trend of precipitation, it is a negative situation in terms of water resources. When we looked at the basin in general, it was determined that the flood risk increased in the southern parts, while the drought risk came to the fore in the north.

In line with the results obtained, examining the response of the basin to precipitation variability is to be considered in future studies. By modeling the land cover and soil characteristics, important information can be obtained in understanding the rainfall–runoff relationship. Decision-makers need to develop strategies by taking into account the hydrological variability in the basin.

We would like to thank the Turkish State Meteorological Service for providing the precipitation data.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.