Assessment of drought using different tests and drought indices in Elazig, Turkey

Water is one of the most critical elements for all living creatures to survive and is essential for their life needs. The water needs for living things are met from surface and groundwater resources. To allocate water resources, it is necessary to determine the efficiency of water resources. Historical changes in river flow are vital for water management and water resources planning. The concept of drought is one of the biggest issues related to water availability under climate change. Therefore, hydrometeorological measurement is crucial in the development and planning of the region to assess the drought. Changes in hydrometeorological parameters (precipitation, temperature, streamflow, evaporation, etc.) affect the region's climate change.

Drought is considered a major natural hazard that humanity has encountered since ancient times. Yet, it affects more people than any other hazard and has the most complex structure of all natural hazards (Ashraf et al. 2021; Wilhite 2021). Unlike other extreme events (cyclones, hurricanes, floods, etc.), drought evolves in a slow and steady progression, and its onset and end are not easy to determine (WMO 2023). Droughts are notably more common and severe in arid and semi-arid regions and can last weeks, months, years, decades, or centuries. The frequency and intensity of drought events are on the rise and significantly affect the global climate's continued warming, human survival, and the sustainable development of society (Thomas & Prasannakumar 2016). Wilhite & Glantz (1985) stated that drought generally forms when a region does not get enough rainfall for an extended period, leading to water scarcity. In a zone of persistent severe drought, the factors and physical mechanisms affecting drought are highly complex, mainly due to: first, the underdevelopment of water conservation projects and the inadequate resilience to disasters; second, the ecological excessive cutting of forest resources, resulting in severe soil erosion and destruction of vegetation, which alters the natural order of the water cycle system, in particular the scale, timing, and frequency of the water cycle; third, abnormal atmospheric circulation leading to irregular precipitation (Min et al. 2011). Drought is also caused by water scarcity, high water use, and unplanned use of water resources (Zou et al. 2023). Earlier studies have demonstrated that drought is primarily the result of insufficient rainfall and too extreme an increase in temperature (Liu et al. 2021). Thus, drought characteristics in a specific region and time scale can be assessed based on precipitation and temperature, and relevant measures can be adopted to minimize the frequency and intensity of drought (Wang et al. 2015). Drought generally relates to a persistent below-average water deficit that can last for a season or longer (Dai 2011; Aghakouchak & Nakhjiri 2012; Constas et al. 2014). For example, researchers such as Aghakouchak & Nakhjiri (2012) and Wilhite & Glantz (1985) have categorized drought into hydrological, meteorological, socioeconomic, and agricultural according to causal factors. They noted that rainfall deficiencies commonly cause drought types. Among them, the main factors causing an increased probability of meteorological drought are hydrological and agricultural droughts (Zhan et al. 2016; WMO 2023).

Meteorological drought lasts for a short period, followed by drought-induced disasters due to inadequate precipitation. However, severe meteorological drought is a situation where the yearly precipitation is below 25% of the normal precipitation for that region (Pal et al. 2000). Since drought is a serious climatic event, it is important for drought analysis to study the impact of climate change on climatological quantities such as precipitation, temperature, evapotranspiration, and its reflections on drought. For this purpose, when looking at the literature, trends in climatological quantities can be detected by a number of parametric (F-test, T-test, linear regression) and non-parametric (Sen's Slope Test, Mann–Kendall, Standard Homogeneity Test) methods (Gumus et al. 2021; Mersin et al. 2022a, 2022b). Since drought is a recurring climatic event, its consequences come out in major impact to agricultural manufacturing, causing reduction of energy production, water supply, mass migration, and loss of life (Masih et al. 2014). Researchers such as Degefu & Bewket (2014) and Sheffield & Wood (2008) have shown in different ways that the causes of drought and its effects on the environment, which are spatially or temporally dependent, are determined by their characteristics such as frequency, magnitude, and intensity. In recent years, drought studies have been common (Chen et al. 2009; Dogan et al. 2012; Malik et al. 2021; Moghimi & Zarei 2021; Tayfur 2021; Yuce & Esit 2021). In addition, studies have been conducted to compare different methods and indexes for drought assessment. Several drought-related indices have been proposed to correctly recognize and assess drought events (magnitude, severity, duration, etc.). The standard precipitation index (SPI) has been commonly used in various climatic zones to describe and compare (McKee et al. 1993; Mersin et al. 2022a). However, numerous studies have agreed on the increasing global temperature, which causes the increase in water demand due to evapotranspiration (Heim 2017). Therefore, SPEI was improved by considering precipitation and potential evapotranspiration (PET) (Vicente-Serrano et al. 2010; Liu et al. 2021). Numerous methods are used to analyze trends in drought assessment. For example, Şen (2012) proposed the innovative trend analysis (ITA). A lot of studies were conducted using different methods for different regions (Kahya & Kalayci 2004; Gocic & Trajkovic 2013; Wu & Qian 2017; Gedefaw et al. 2018; Gumus 2019; Danandeh Mehr & Vaheddoost 2020; Gumus et al. 2021; Tayfur 2021; Avsaroglu & Gumus 2022). An innovative trend significance test (ITST) was developed by Şen (2017) to obtain a trend at a certain significance level. Moreover, Sen et al. (2019) proposed the innovative polygon trend analysis (IPTA) to obtain the variations in successive periods in the time series. IPTA method is a new trend analysis method. There are limited studies on this method (Sen et al. 2019; Achite et al. 2021; Şan et al. 2021; Gumus et al. 2022; Hırca et al. 2022). Jain et al. (2015) conducted a comparison of effective drought index (EDI), SPI, Chinese Z index (CZI), rainfall decile based on drought index (RDDI), statistical Z-score, and rainfall deviation (RD) for their applicability in drought-prone regions of the Ken River Basin in India. They stated that drought indices have a high correlation at the same time steps and can be utilized as alternatives.

Frequency analysis is a widespread tool in the analysis of hydrometeorological data, in particular floods and droughts. It is inadequate to use the frequency of drought phenomena alone unless it is numerically associated with duration, intensity, or intensity in the form of intensity–duration–frequency (IDF) curves from which can be deduced any drought of a certain duration, intensity, and return period (Dalezios et al. 2000; Rahmat et al. 2015). Gupta et al. (2020) developed severity–duration–frequency (SDF) curves based on SPEI that indicate an increase in severity with increasing drought duration. The selected drought indices are vital due to climate variability from one region to another. Eris et al. (2020) utilized several indices to examine the spatial-temporal analysis of meteorological drought over the Küçük Menderes Basin in the Aegean Region of Turkey. Dabanlı et al. (2017) analyzed long-term spatial-temporal drought variability in Turkey using SPI data. Danandeh Mehr et al. (2020) evaluated the impacts of climate change on meteorological drought using SPI and SPEI. There are many small- and large-scale studies on Turkey (Cavus & Aksoy 2019; Gumus et al. 2021; Yuce & Esit 2021; Katipoğlu et al. 2022; Seker & Gumus 2022; Sarıgöl & Katipoğlu 2023). Elazig is one of the crucial provinces in the Eastern Anatolia region, which is rich in the presence of water resources, natural beauties, and biodiversity. The presence of dam lakes around Elazig and the richness of its water resources require sustainable use of water resources as well as accurate prediction of disasters caused by climatic events and taking precautions in advance. In this study, a detailed drought analysis of Elazig located in the Euphrates basin is carried out by using monthly precipitation, temperature and evaporation data, SPI, SPEI, and CZI indices recorded between 1980 and 2022 from the meteorological stations of Agin, Baskil, Elazig, Maden, Keban, Karakocan, Palu, and Sivrice stations in Elazig, considering monthly and annual time scales. To analyze the effects of climate change, parametric (linear trend) and non-parametric (Sen's slope, Mann–Kendall, and Run tests) tests were applied for precipitation, temperature, and evaporation data. Moreover, precipitation IDF curves for the return intervals of 2, 5, 10, 25, 50, and 100 years were obtained using precipitation data for log-normal distribution.

This study aims: (1) to obtain SPI, SPEI and CZI drought indices at different time scales (1- and 12-month scales), (2) to compare the performances of SPI, SPEI, and CZI, (3) to obtain IDF, (4) to apply IPTA for the trend analysis.

Elazig province is located in the southwest of the Eastern Anatolia Region, in the Upper Euphrates Region. With an area of 9,153 km², it constitutes 0.12% of Turkey's territory. Situated between 40 0 21′ and 380 30′ east longitudes and 38 0 17′ and 39 0 11′ north latitudes, the region is surrounded by Bingol to the east, Tunceli to the north (via Keban Dam Lake), Malatya to the west and southwest (via Karakaya Dam Lake) and Diyarbakir to the south.

The study was evaluated based on the following sections. Data analysis, run test, linear trend analysis, Mann–Kendall test, rainfall IDF, Sen's slope method, and drought indices (IPTA, SPI, SPEI, and CZI) are given in 3.1, 3.2, 3.3, 3.4, 3.5, 3.6, and 3.7 sections, respectively.

Meteorological data were obtained from the General Directorate of Meteorology (MGM) as daily data, which were converted into monthly data. Encountering data irregularities in meteorological data is common. Therefore, several methods are usually used to obtain missing data. Almost all stations in Elazig province have complete meteorological data (temperature, precipitation). However, the homogeneity method was used to complete the missing data in some stations. In the study, station numbers were determined as eight. Station names, altitude, latitude (°) and longitude (°) are shown in Table 1. Some meteorological stations' data periods range from 1955 to 2022. However, to determine the trends, the years between 1980 and 2022 were selected by utilizing the common year's scale measured at all stations.

The temporal randomness of the time series is determined by the Run test. Random behavior is known to be troublesome due to the difficulty of assessing the event. The test is a statistical technique that investigates the randomness of a data set within a given distribution and determines whether a random process deduces it. The method was developed by Yevjevich (1967) and is called the running theory or threshold level method. The method can be used for daily, monthly, and annual data. As such, it is critical to determine whether an experiment's result is really random, particularly where random and serial data are of interest for further theorizing and analyses.

Then, a two-sided confidence interval checks the estimated quantity Z against the standard normal distribution table. Thus, if |Z| > Z_1−α⁄2, H_o is rejected, H_a is accepted, and there is a valuable trend. However, if the condition |Z| > Z_1−α⁄2 is not met, H_o is accepted, and H_a is rejected, i.e., no statistical trend exists. In this study, it is significant that a significance level of 5% is used for the analysis since Z_1−α⁄2 = 1.96.

The IDF curve shows a calculation of the intensity of precipitation as a mathematical function of its duration and frequency. Precipitation IDF curves have an active role in hydrology and water management. The IDF curves are representations of the graphical probability of a certain average precipitation intensity occurring at a specific time. IDF curves can have different mathematical expressions that are empirically or theoretically fitted to the precipitation data observed. The first IDF curve was derived as early as 1932. Since then, various IDF relationships have been configured for a large number of regions worldwide. Precipitation assessment is often performed using IDF curves for various plans concerning water resources. IDF curves are commonly utilized tools in different engineering projects in water management, design, planning, and operation of water resources projects or against floods (Nhat et al. 2006). They are commonly used in water resources engineering, especially for the simultaneous characterization of design storm events of various magnitude, duration, and return periods. Statistical analyses of historical precipitation phenomena (intensity, duration and return period) are utilized to design flood protection structures and numerous other civil engineering structures containing hydrological flows (Prodanovic & Simonovic 2007; McCuen 2016).

Precipitation IDF curves are selected for eight meteorological stations in Elazig. The curves are station-specific and cannot be valid for a different station or a different hydrological basin. A methodology is needed to derive an intensity-SDF curve that can be used as a key curve for a region or hydrological basin. Thus, a regional drought IDF curve would be useful especially for unmeasured points with missing data.

Sen (1968) introduced a non-parametric test for detecting the slope of a trend. When the data time series has a linear trend, this method is applicable to compute the variation of the unit time (Yu et al. 1993). The median of all slopes between all pairs of data in the same series is considered the slope of the trend. The method's detail was given by Sen (1968).

The drought indices are typically calculated numerical representations of drought severity that are considered using climatic or hydrometeorological inputs. These indices are assessed for a specific time to quantify the characteristic state of drought in different locations.

In which p is the exceeding probability of computed values of D_i.

The general objective of this study is to make a detailed drought analysis of Elazig province at different time scales by considering different indices. Because Elazig province is considered a center of attraction for developing vegetable and fruit growing, field crops, aquaculture, and livestock sub-sectors due to its natural resources such as soil, climate, topography, and hydraulic structures. Therefore, the drought analysis has been carried out by considering the period between 1980 and 2022 with SPI (only precipitation parameter), SPEI (both precipitation and temperature) and CZI indices.

The study analyzed Elazig's monthly temperature (max, min, mean), precipitation, and evaporation data for 1980–2022, applying Sen's slope, Mann–Kendall, and run tests, as shown in Table 3. The decreasing trend was detected in precipitation data for Agin, Baskil, Karakocan, Keban, Maden, Palu, and Sivrice monitoring stations. In contrast, an upward trend was detected for Elazig at a 95% confidence interval. These trends were measured by Sen's slope method. The precipitation data for all stations were not randomly distributed. There is no trend in the precipitation data for Baskil, Elazig, Keban, Maden, Palu, and Sivrice stations, while there is a trend in data for Agin and Karakocan based on the Mann–Kendall test.

Test interpretation for run test: H₀: Data are randomly distributed. H_a: Data are not randomly distributed. As the computed p-value is lower than the significance level alpha = 0.05, one should reject the null hypothesis H₀ and accept the alternative hypothesis, H_a.

Drought analysis of Elazig province between 1980 and 2022 was calculated according to three critical indices. The main reason for using three indices is that they consider different parameters. Drought indices are compared according to different drought categories. Among the drought categories, normal drought was the most common. Very severe drought has the most negligible share. The situation in the moderate drought category is the same as in the very severe drought category. According to these results, it is concluded that SPI is more sensitive than the SPEI index in the very severe drought category. In the moderate severe drought, 26 months were realized in the SPI index, while 54 months were calculated according to the SPEI index and 35 months were calculated according to the CZI index. In the severe drought, these values are 15, 25, and 18 months for SPI and SPEI, CZI, respectively. Similarly, extreme drought values were 12, 7, and 4 months for SPI, SPEI, and CZI, respectively. Table 4 shows the monthly and yearly drought and wet periods. In addition, Table 4 shows the drought and wet years based on the SPI, SPEI and CZI indices.

First, the climate change effects on temperature, precipitation, and evaporation parameters were analyzed. A decreasing trend was detected in precipitation data, while an increasing trend was detected in temperature and evaporation data based on a 95% confidence interval. Second, the relationship between the two indices was analyzed in detail for Elazig province. Although normal drought has the highest share among the drought categories, very severe drought has the lowest share.

In contrast, more severe droughts were observed in the regions close to dams (Keban) based on the analysis. In this context, the highest SPI-1 value was 3.191 at Sivrice station, but the lowest SPI-1 value was −4.253 at Elazig station. Similarly, the driest SPI-12 value was −3.156 at Karakocan, but the wettest SPI-12 value was 2.855 at Baskil station. As a result, dry and wet periods were observed at all stations between 1980 and 2020, both SPI-1 and SPI-12.

Monthly and annual drought assessments were obtained by calculating the SPEI-1 and SPEI-12 values (Figure 5). The data were plotted to clarify the wettest and driest periods throughout the time series. Finally, wetter periods were observed in the regions close to mountain areas (Maden), while more severe droughts were observed in the regions close to dams (Keban) based on the analysis. In the study, the highest SPEI-1 value was 2.802 at Baskil station, but the lowest SPEI-1 value was −2.814 at Palu station. Similarly, the driest SPEI-12 value was −2.263 at Karakocan, but the wettest SPEI-12 value was 2.668 at Elazig station. As a result, dry and wet periods were observed at all stations between 1980 and 2020, both SPEI-1 and SPEI-12. Katipoğlu et al. (2022) reported that SPEI is more sensitive than other indices in determining drought trends since it also considers PET values, which may indicate the upward climate change impacts in the last years.

The assessment of monthly and annual drought values was determined by obtaining the calculated CZI-1 and CZI-12 values (Figure 6). Therefore, it was determined that historical extreme drought events could be observed using the lowest and highest index values (Figure 6). Finally, wetter periods were observed in the regions close to mountain area (Maden). In contrast, more severe droughts were observed in the regions close to dams (Keban) based on the analysis similar to SPI and SPEI. The highest CZI-1 value was 2.745 at Baskil station, but the lowest CZI-1 value was −3.271 at Karakocan station. Similarly, the highest CZI-12 value was 2.573 at Baskil, but the lowest CZI-12 value was −2.796 at Karakocan station. As a result, dry and wet periods were observed at all stations between 1980 and 2020, both CZI-1 and CZI-12.

Similarly, precipitation data have low values for July and August for all stations. Moreover, the precipitation phenomenon in this region does not reflect homogeneous and isotropic behavior, and January has a strong increasing trend in the monthly precipitation tendencies except for Baskil station. Nine months (June, October, February, December, May, April, March, November, and July) nearly remain below the 1:1 line. Since there is an irregular and stable variation of successive months as a polygon in all stations, as shown in Figure 8, the changes are non-systematic from month to month as Elazig's precipitation is unstable. This complex precipitation data set can be interpreted by considering the orographic factor and geographical location. Furthermore, Katipoglu et al. (2020) found an increase in drought trends in January, February, May, June, August, November, December, and annual periods. They also determined that the Euphrates Basin has a significant drought risk. Moreover, Dabanli (2018) that Elazig is one of the provinces with the highest drought risk in Turkey. The results for drought evaluation of Elazig are compatible with the literature (Dabanli 2018; Katipoglu et al. 2020).

Precipitation durations are selected from 2, 5, 10, 20, 30, 50, and 100-year returns. For each intensity duration, frequencies are determined. The Weibull, generalized extreme value, gamma (Pearson Type III), and exponential and log-normal distributions were then fitted to the intensity values of a selected precipitation duration. Once the best-fit probability distribution of a selected period is determined, the intensity of any return period can be calculated. Precipitation IDF curve is then obtained.

The general objective of this study is to make a detailed drought analysis of Elazig for different time scales by considering different indices. Because Elazig is considered as a center of attraction for the development of vegetable and fruit growing, field crops, aquaculture, and animal husbandry sub-sectors due to its natural resources such as soil, climate, topography, hydraulic structures (i.e., dams, lakes, etc.), which are suitable for production. For this reason, drought analysis was performed by considering the period between 1980 and 2022 with SPEI, CZI, and SPI indices, IPTA methods, Sen's slope, Mann–Kendall and run tests. First, the effects of climate change on temperature (max, min, average), precipitation, and evaporation parameters were examined in detail. A downward trend was detected in precipitation data, while an upward trend was detected in temperature and evaporation data based on a 95% confidence interval. Second, the relationship between the three indices was analyzed in detail for Elazig. In the moderately dry ⁠), 26 months were realized in the SPI index, while 54 months were calculated according to the SPEI index and 35 months were calculated according to the CZI index. In the severe dry (⁠⁠, these values are 15, 25, and 18 months for SPI and SPEI, CZI, respectively. Similarly, extremely dry values (⁠ were 12, 7 and 4 months for SPI and SPEI, CZI, respectively. Although normal drought has the highest share among drought categories (⁠⁠), extremely dry has the lowest share⁠, it is determined that SPI gives more sensitive results in the very severe drought category than the SPEI index. Thus, this drought analysis for Elazig can be used by local authorities or scientific institutions with different drought indices, data sets, or drought prediction methods around Elazig. It is expected to shed light on the drought analyses to be made in the field.

Special thanks to the General Directorate of Meteorology (MGM) for providing the database used in this study.

Veysi KARTAL contributed to conceptualization, methodology, validation, multiple regression analysis, investigation, resources, writing – original draft preparation, and visualization. The authors read and agreed to the published version of the manuscript.

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

The authors declare there is no conflict.