Abstract

Drought incidents occur due to the fact that precipitation values are below average for many years. Drought causes serious effects in many sectors, such as agriculture, economy, health, and energy. Therefore, the determination of drought and water scarcity, monitoring, management, and planning of drought and taking early measures are important issues. In order to solve these issues, the advantages and disadvantages of five different meteorological drought indices were compared, and the most effective drought index was determined for monitoring drought. Accordingly, in the monthly, 3-month, and 12-month time period, covering the years between 1966 and 2017 (52 years), Standardized Precipitation Index (SPI), Statistical Z-Score Index (ZSI), Rainfall Anomaly Index (RAI), Standardized Precipitation Evapotranspiration Index (SPEI), and Reconnaissance Drought Index (RDI) were used. It was concluded that precipitation-based SPI and ZSI are similar patterns and precipitation, and temperature-based SPEI and RDI are similar patterns. Also, it has been determined that RAI is more effective than other indices in determining the periods of extreme drought or wet. Furthermore, SPEI and RDI have been found to be superior to other indices as they take into account the water consumption and climate effects caused by evapotranspiration.

HIGHLIGHTS

  • Performance of SPI, ZSI, YAI, SPEI and RDI values were compared by using regression analysis, correlation analysis, and time series plot.

  • Models were evaluated according to statistical criteria such as R, R2 and RMSE.

  • RAI has been found to be the most effective in determining extreme arid and wet periods.

  • It was found that SPI values gave results close to ZSI and SPEI values gave results close to RDI in the same time period.

  • Increasing evapotranspiration values in the 21st century make SPEI and RDI indices more important than other indexes.

Graphical Abstract

Graphical Abstract
Graphical Abstract

INTRODUCTION

Drought indices are vital as they are the basis for drought management plans by identifying and monitoring drought conditions, determining the timing and level of drought reactions, characterizing and comparing drought events, combining drought severity levels with drought responses (Wilhite 2005).

Drought indices use a variety of hydro-meteorological data, such as rainfall, streamflow, reservoir storage, soil moisture, groundwater, and water supply indicators (Sivakumar et al. 2011). In general, based on these physical datasets, drought indicators consist of three leading indicators: meteorological, agricultural, and hydrological drought indicators. Popular meteorological drought indicators include: Decile Index (DI), Palmer Drought Severity Index (PDSI), Self-calibrated PDSI (SC-PDSI), Palmer Z Index (PZI), Percent Departure from Normal (PDN), Effective Drought Index (EDI), Standard Precipitation Index (SPI), Reconnaissance Drought Index (RDI) Rainfall Variability Index (RVI), China-Z Index (CZI), and Standardized Precipitation Evapotranspiration Index (SPEI). Agricultural drought indicators include: Aridity Index (AI), Moisture Adequacy Index (MAI), and Standardized Soil Moisture Anomaly Index (SSMAI). Hydrological drought indicators include: Standardized Water Level Index (SWLI), Surface Water Supply Index (SWSI), Supply and Demand Drought Index (SDDI), Streamflow Drought Index (SDI), and Standardized Runoff Index (SRI).

In this study, drought indices based on monthly average temperature and precipitation, easy to calculate, have similar drought classification thresholds and which are the most widely used in the literature were preferred. EDI, DI, PDSI, PDN, RVI, and CZI indicators can be used as an alternative to meteorological drought assessment. However, they were not used in this study due to the lack of meteorological inputs used in the calculation of the indices and the classification difference of the indices. By using these indices, it could be determined which index could determine droughts more precisely according to different threshold levels.

There is not a single drought index that can explain all kinds of drought types (meteorological, agricultural, hydrological), climate regime, and the sectors affected by drought. There are many factors to determine which index is best for a particular need or application. Users need to answer the following questions to decide which indices are best suited for their current status (Wilhite & Pulwarty 2017).

  • Do the indices allow drought detection on time to initiate drought reduction actions?

  • Are the indices sensitive to climate, time, and space to determine the beginning and end of the drought?

  • Do the indices and various severity levels reflect the effects occurring on the ground for a particular location or region?

  • Are hybrid indicators used to take into account many factors and inputs?

  • Does the data source have a long recording period that can give planners and decision-makers a strong historical and statistical sign?

  • Are indices easy to apply? Do your users have enough resources?

  • Can indices be compared with other indices?

  • Do the indices give reliable and consistent results?

A suitable drought index for a region may not be applied specifically for other regions due to the natural complexity of drought events, hydro-climatic conditions, and watershed features (Smakhtin & Hughes 2007). Therefore, many researchers around the world have compared different drought indices to find the appropriate drought index for a particular region or basin. Wang et al. (2017) compared SPI, SPEI, and SC-PDSI in the arid region of northwestern China. It was found that all three forms of drought indices significantly correlated with each other. The drought severity of SPI and SPEI was found to be statistically similar. Adnan et al. (2018) compared SPI, SPEI, CZI, DI, MCZI, ZSI, PDN, RDI, SSMAI, SC-PDS, and RVI in Pakistan. The performance, efficiency, and importance of the indices were tested by applying different statistical tests. SPI, SPEI, and RDI results were found to be useful for monitoring the drought situation in Pakistan. Myronidis et al. (2018) compared SPI, SPEI, DI, EDI, PDSI, CZI, ZSI, RDI, PZI, and SC-PDSI on tracking the effect of drought on streamflow. It was revealed that the 3-month Z-score drought index has the best association with the four streamflow variables. Payab & Türker (2019) compared and evaluated the performance of SPI, RDI, ZSI, CZI, and SDDI for their suitability, dependability and effectiveness in the northern part of Cyprus. The study revealed that all the drought indices are highly and significantly correlated with each other (minimum 0.982 and maximum 0.998). Li et al. (2019) compared SPI and SPEI in the analysis of spatial and temporal changes of climatic drought events in the Yangtze River Basin. The results revealed that the SPEI calculated by the Penman–Monteith (PM) method performed better than the SPI and the SPEI based on the Thornthwaite (TH) method. Shamshirband et al. (2020) modeled and compared SPI, SSI, and SPEI. Based on Pearson correlation and cross-correlation, it was identified that although SPEI is more sensitive to climatic conditions, the drought index of SPI was better, and hydrological drought index predicted SSI with a lower error. In the study, the SPEI index contributes to the literature in terms of being a more effective index than SPI, ZSI, RAI, and RDI indices due to the increasing temperatures in the 20th century as a result of global warming effects. In addition, it was determined that RAI was the most effective index in determining the extremely arid and wet periods.

Drought indices are an important tool that can track climatic change and interpret meteorological, agricultural, and hydrological changes that will occur in a region. The most commonly used time scale for drought analysis is monthly or annual (Mishra & Singh 2010). It is possible to say that the analyses using a monthly time period are more suitable for agricultural and water supply problems (Panu & Sharma 2002). A 12-month time period can be used to identify hydrological droughts. Furthermore, Wu et al. (2001) stated that the reliability of the index decreased in the 36- and 48-month time periods selected in their reviews. In the drought index calculations, 1- and 3-month time scales can be used as meteorological, 6- and 9-month time scales as agricultural drought and 12- to 24-month time scales can be used as hydrological drought indices (Kapluhan 2013). In line with previous studies and reports, monthly, 3-month and 12-month periods, which are important meteorological and hydrological drought time periods, were used in the study. The use of these time periods is valuable in evaluating the relationship between meteorological and hydrological droughts and the lag time and response rate for future studies.

It is vital to determine the most appropriate drought index for the study area and to monitor droughts. This study aimed to determine the most suitable drought index by comparing five different meteorological drought indices at Erzincan meteorological observation station, and to determine the effect of meteorological parameters such as temperature and precipitation in the calculation of the indices. The comparison process was carried out by examining the linear regression line, which was adapted to the scattering diagrams of the indices, Pearson correlation analysis, and the change of indices over time in the monthly, 3-month, and 12-month time periods. As a result of the study, it was found that SPEI and RDI indices, which take into account potential evaporation (PET) values in drought calculations, are more effective in monitoring drought and RAI is the optimal index in determining the extreme events.

MATERIALS AND METHODS

Study area and data

In this study, monthly average rainfall and air temperatures covering the years between 1966 and 2017 belonging to the Erzincan meteorological observation station (MOS) in the Euphrates basin were used to calculate the drought indices. The drought indices of all stations in the study area were compared and the results were found to be similar. However, the station (Erzincan) that best expresses the results visually was used in the case study. The location map showing the location of the stations in the Euphrates basin is shown in Figure 1.

Figure 1

Location map of Erzincan meteorological observation station used in the study.

Figure 1

Location map of Erzincan meteorological observation station used in the study.

The Euphrates River has geopolitical importance as it flows through Turkey and Middle Eastern countries such as Syria and Iraq. For this reason, the drought indices of the meteorology stations (Table 1) located in the Euphrates basin, which has an impact on the Middle East water problem, were evaluated.

Table 1

Meteorological stations in the Euphrates basin

Station Number/NameLatitudeLongitudeAltitude (m)
17762/Kangal 39.24 37.38 1,521 
17094/Erzincan 39.75 39.48 1,216 
17096/Erzurum 39.95 41.19 1,758 
17740/Hınıs 39.37 41.69 1,715 
17165/Tunceli 39.11 39.54 981 
17099/Ağrı 39.72 43.05 1,646 
17780/Malazgirt 39.14 42.53 1,540 
17204/Muş 38.75 41.50 1,322 
17776/Solhan 38.96 41.05 1,366 
17203/Bingöl 38.88 40.50 1,139 
17666/İspir 40.48 40.99 1,223 
17688/Tortum 40.30 41.54 1,576 
17692/Sarıkamış 40.33 42.60 2,102 
17690/Horasan 40.04 42.17 1,540 
17716/Zara 39.89 37.75 1,338 
17265/Adıyaman 37.75 38.28 672 
Station Number/NameLatitudeLongitudeAltitude (m)
17762/Kangal 39.24 37.38 1,521 
17094/Erzincan 39.75 39.48 1,216 
17096/Erzurum 39.95 41.19 1,758 
17740/Hınıs 39.37 41.69 1,715 
17165/Tunceli 39.11 39.54 981 
17099/Ağrı 39.72 43.05 1,646 
17780/Malazgirt 39.14 42.53 1,540 
17204/Muş 38.75 41.50 1,322 
17776/Solhan 38.96 41.05 1,366 
17203/Bingöl 38.88 40.50 1,139 
17666/İspir 40.48 40.99 1,223 
17688/Tortum 40.30 41.54 1,576 
17692/Sarıkamış 40.33 42.60 2,102 
17690/Horasan 40.04 42.17 1,540 
17716/Zara 39.89 37.75 1,338 
17265/Adıyaman 37.75 38.28 672 

Analysis of drought indices

Numerous climatological, meteorological, or hydrological indices are used to identify drought, which is a complex process. Drought indices can clearly define the emergence and development of arid conditions. Drought indices are important in terms of detecting the onset of drought conditions, measuring and monitoring drought incidents, and determining the magnitude of the drought disaster.

The drought index suitable for a region can be selected according to the geomorphological, climatological, hydrological, ecological, and soil characteristics of the region, and the types of drought (meteorological, hydrological, agricultural, socio-economic) and the type of economic activity. Some indices are more suitable for monitoring, and some are more suitable for analyzing historical drought events. For this reason, using several indices or indicators with some advantages and weaknesses compared to each other is more useful in terms of qualifying drought and comparing the results. In this study, SPI, ZSI, RAI, SPEI, and RDI, drought indices accepted in national and international scientific studies, were used. The drought classification of the indices used is shown in Table 2.

Table 2

Classification of drought indices (McKee et al. 1993; Keyantash & Dracup 2002; Tsakiris et al. 2007)

SPI-SPEI-ZSI-RDIRAICategory
0 ≤ Index 0 ≤ Index Wet 
−1 < Index ≤ 0 −1.2 < Index ≤ 0 Mild 
−1.5 < Index ≤ −1.0 −2.1 < Index ≤ −1,2 Medium 
−2 < Index ≤ −1.5 −3 < Index ≤ −2.1 Severe 
Index ≤ −2.0 Index ≤ −3 Extreme 
SPI-SPEI-ZSI-RDIRAICategory
0 ≤ Index 0 ≤ Index Wet 
−1 < Index ≤ 0 −1.2 < Index ≤ 0 Mild 
−1.5 < Index ≤ −1.0 −2.1 < Index ≤ −1,2 Medium 
−2 < Index ≤ −1.5 −3 < Index ≤ −2.1 Severe 
Index ≤ −2.0 Index ≤ −3 Extreme 

Standard Precipitation Index (SPI)

SPI is one of the most popular meteorological drought indices to identify and monitor drought using long-term monthly precipitation data. This index only needs monthly precipitation data as input and can monitor drought on multiple time scales. The long-term precipitation data are first adapted to a Gamma probability distribution and then converted to a standard normal distribution, thus obtaining the SPI values with an average of 0 and a standard deviation of 1 (McKee et al. 1993).

Standard Precipitation Evapotranspiration Index (SPEI)

SPEI is based on the combination of the PDSI's sensitivity to changes in evaporation demand caused by temperature fluctuations and trends and SPI's multiple time period features. The first step in calculating SPEI is to obtain PET values. Acquisition of pan evaporation by means of direct measurements can be an expensive and a tedious task, so the evaporation rate is routinely estimated by statistical regression and parametric methods such as the Thornthwaite and Hargreaves approaches (Jacobs et al. 1998; Ali Ghorbani et al. 2018). For this reason, the Thornthwaite equation was preferred in PET calculation in this study. The difference between the precipitation values (P) and the PET values calculated using the Thornthwaite (1948) method, which requires only the air temperature as the input variable, is calculated by Equation (1):
formula
(1)

Here, Di is a measure of the climatic water balance for a month. After fitting the three-parameter log-logistic distribution to this difference, the distribution is converted to the standard normal distribution, resulting in SPEI values (Vicente-Serrano et al. 2010). SPEI values obtained in this way are used to classify drought in a similar way to SPI.

Z-Score Index (ZSI)

The statistical Z score is a dimensionless coefficient obtained by subtracting the population mean () from a single precipitation value () and then dividing the difference by the population standard deviation (σ). This conversion process is called standardization or normalization. The Z score indicates how many standard deviations a precipitation value is above or below average. This index is not the same as SPI because it does not require the data to be adjusted to fit the Gamma or Pearson type III distribution (Tsakiris & Vangelis 2004; Akhtari et al. 2009). The calculation of ZSI values is given in Equation (2):
formula
(2)

Rainfall Anomaly Index (RAI)

This index was produced by Van Rooy (1965) to determine the deviation in precipitation. Equations (3) and (4) are used in the calculation of the index.

If P >, it is anomaly positive and index value:
formula
(3)
If P <, it is anomalic negative and index value:
formula
(4)

Here, is average of precipitation series, is average of the highest 10 values, and is average of the smallest ten values.

Reconnaissance Drought Index (RDI)

This is based on the ratio between cumulative precipitation values and PET. The initial value of RDI is obtained by calculating an ‘ak’ ratio between precipitation in a given area and total potential evapotranspiration for each consecutive period of each month in a year. It is expressed mathematically as:
formula
(5)
Here, Pij and PETij are the precipitation and potential evapotranspiration value of the i'th year of the j'th month, respectively. Standardized RDI values are calculated as in Equation (6) by fitting log-normal distribution to RDIak values (Tigkas 2008):
formula
(6)

Here, yi is calculated as ln(αki), and is the arithmetic mean, and σy is the standard deviation.

Comparison of drought indices

The selection of the appropriate drought index in any area is vital for reducing drought conditions and developing effective strategies. In this study, it was investigated to determine the performance, advantages, and disadvantages of five meteorological drought indices, namely, SPI, ZSI, SPEI, RAI, and RDI. In this study, comparing the different drought indices has provided more detailed and reliable results of the drought analysis.

The indices used in the study were preferred because they have a variable time scale that allows the determination of drought dynamics for various meteorological, agricultural, and hydrological applications; they are easy to compare since standardization is performed, and they provide calculation convenience because they can be calculated with a small number of statistical parameters such as precipitation and temperature. Due to insufficient parameters such as solar radiation, relative humidity, ground humidity, and temperature and rainfall as the main parameters representing drought, only monthly temperature, and precipitation-based drought indices were used in this study.

It was determined by McKee et al. (1993) that drought severity can be determined and evaluated in long and short time steps such as 1, 3, 6, 9, 12, or 24 months. In addition, Wu et al. (2001) propounded that the reliability of the index decreased in the 36- and 48-month periods. Therefore, in the study, drought indices were calculated at monthly, 3-month and 12-month time steps to compare and evaluate short- or long-term droughts which are important meteorological and hydrological drought time periods.

Evaluation of models

In this study, the Pearson correlation coefficient (R), determination coefficient (R2), and root mean square error (RMSE) criteria, which are among the strong statistical criteria, were used. RMSE is a statistical measure that measures the predictive accuracy by determining the differences between the predicted and the observed values. The determination coefficient (R2) is a statistical measure that shows how close the data are to the fitted regression line. These statistical calculations can be made with the help of Equations (7)– (9), respectively:
formula
(7)
formula
(8)
formula
(9)

Here, Di1 is the first drought index selected, Di2, the second drought index selected, and N the number of data. R2 ranges from 0 to 1, with higher values indicating less error variance, and typically values greater than 0.5 are considered acceptable (Santhi et al. 2001; Van Liew et al. 2003). In addition, it is commonly accepted that the lower the RMSE, the better the model performance (Moriasi et al. 2007).

RESULTS AND DISCUSSION

Correlation analysis of drought indices

Correlation analysis is a method used to determine the direction and strength of the relationship between the two variables. Correlation analysis can be applied by Pearson's r, Spearman's rs, and Kendall's τ coefficients. Since Pearson correlation analysis is a parametric statistical test, it is based on the normality assumption. This assumption is provided because the calculation of the selected drought indices is based on standardization. In other words, when the skewness and kurtosis coefficients of all indices are analyzed, they are in the range [−1, +1]. For this, Pearson correlation analysis was used in this study while conducting correlation analysis. If the variables did not obey the assumption of normality, the relationships of the index could be interpreted by Spearman or Kendall τ rank correlation tests. The analysis was carried out at the level of 0.01 significance and is shown in Table 3. Moreover, in this study, r ≥ 0.8 values were chosen as critical values indicating a strong positive relationship.

Table 3

Pearson correlation coefficient matrix of drought indices of Erzincan station

IndexSPI-1SPI-3SPI-12SPEI- 1SPEI- 3SPEI-12ZSI-1ZSI-3ZSI-12RAI-1RAI-3RAI- 12RDI- 12
SPI-1 0.53 0.23 0.90 0.48 0.19 0.97 0.53 0.23 0.75 0.38 0.20 0.21 
SPI-3  0.44 0.53 0.89 0.36 0.55 0.98 0.44 0.42 0.70 0.42 0.41 
SPI-12   0.23 0.41 0.81 0.26 0.45 0.99 0.19 0.32 0.96 0.93 
SPEI-1    0.60 0.30 0.89 0.53 0.23 0.69 0.38 0.21 0.29 
SPEI-3     0.50 0.50 0.88 0.41 0.37 0.63 0.40 0.49 
SPEI-12      0.21 0.37 0.82 0.14 0.26 0.79 0.96 
ZSI-1       0.56 0.26 0.74 0.40 0.23 0.25 
ZSI-3        0.46 0.42 0.70 0.43 0.42 
ZSI-12         0.19 0.33 0.95 0.93 
RAI-1          0.59 0.20 0.17 
RAI-3           0.26 0.30 
RAI-12            0.90 
RDI-12             
IndexSPI-1SPI-3SPI-12SPEI- 1SPEI- 3SPEI-12ZSI-1ZSI-3ZSI-12RAI-1RAI-3RAI- 12RDI- 12
SPI-1 0.53 0.23 0.90 0.48 0.19 0.97 0.53 0.23 0.75 0.38 0.20 0.21 
SPI-3  0.44 0.53 0.89 0.36 0.55 0.98 0.44 0.42 0.70 0.42 0.41 
SPI-12   0.23 0.41 0.81 0.26 0.45 0.99 0.19 0.32 0.96 0.93 
SPEI-1    0.60 0.30 0.89 0.53 0.23 0.69 0.38 0.21 0.29 
SPEI-3     0.50 0.50 0.88 0.41 0.37 0.63 0.40 0.49 
SPEI-12      0.21 0.37 0.82 0.14 0.26 0.79 0.96 
ZSI-1       0.56 0.26 0.74 0.40 0.23 0.25 
ZSI-3        0.46 0.42 0.70 0.43 0.42 
ZSI-12         0.19 0.33 0.95 0.93 
RAI-1          0.59 0.20 0.17 
RAI-3           0.26 0.30 
RAI-12            0.90 
RDI-12             

Note: Bold character shows a strong positive relationship.

Where SPI-1 represents the drought index calculated over a time period of 1 month, SPI-3 over a time period of 3 months, and SPI-12 over a time period of 12 months, when the correlation matrices were examined, the strongest relationship was observed among the indices in the same time periods. As time difference increases (monthly to yearly), the relationship between variables has weakened. Among the indices, the strongest correlation coefficient (0.99) was observed between SPI-12 and ZSI-12, and the lowest correlation coefficient (0,14) was between RAI-1 and SPEI-12 (Table 3).

Linear regression analysis

Regression analysis, correlation analysis, time series plot, outer circles and Rescale range test and its Hurst index value can be used to compare drought indices. However, in this study, regression analysis, correlation analysis, and time series plot were preferred in order to provide sufficient visuality and effective results and to be easy. In order to compare drought indices, scatter diagrams of indices were drawn and statistically evaluated. For this, R2 and the RSME were used. RSME is the square root of the variance of residues and shows the model's compatibility with the data. That is, it shows how close the observed data points are to the predicted values of the model. R2 is the variable that shows to what extent the regression line fits the observations. Furthermore, the functional relationship between the equation of the line and the two variables can be expressed. For example, scattering diagrams of the drought indices of Erzincan station are shown (Figure 2).

Figure 2

Scatter diagrams of various drought indices of Erzincan station: (a) SPI-1 and SPEI-1, (b) SPI-1 and ZSI-1, (c) SPI-1 and RAI-1, (d) SPI-3 and SPEI-3, (e) SPI-3 and ZSI-3, (f) SPI-3 and RAI-3, (g) SPI-12 and SPEI-12, (h) SPI-12 and ZSI-12, (i) SPI-12 and RAI-12, (j) SPI-12 and RDI-12, (k) SPEI-12 and RDI-12.

Figure 2

Scatter diagrams of various drought indices of Erzincan station: (a) SPI-1 and SPEI-1, (b) SPI-1 and ZSI-1, (c) SPI-1 and RAI-1, (d) SPI-3 and SPEI-3, (e) SPI-3 and ZSI-3, (f) SPI-3 and RAI-3, (g) SPI-12 and SPEI-12, (h) SPI-12 and ZSI-12, (i) SPI-12 and RAI-12, (j) SPI-12 and RDI-12, (k) SPEI-12 and RDI-12.

In Figure 2, scatter diagrams of all drought indices belonging to Erzincan station, equation of the lines, and the coefficient of determinations are shown. When the diagrams are examined, the points that show the strongest fit, that are along a linear line, are the values of SPI-12 and ZSI-12, whereas the weakest fitting points are scattered between SPI-3 and RAI-3 values. When the results of the statistical indicators (R2 and RMSE) were examined, it was found that the values of SPI give close results with ZSI and SPEI gives close results with RDI in the same time period. R2 values close to 1 and RMSE values close to 0 indicate that the indices give close results.

Comparison of SPI and ZSI values

In this study, the time series plot of the drought indices are used to compare the five drought indices. In all selected time periods, when the values of SPI and ZSI were examined, almost the same result was found in both indices. However, SPI shows more severe values in dry periods, while ZSI shows more severe values in wetlands (Figure 3). ZSI, in terms of ease of account and SPI in terms of reliability, can be preferred to observe droughts. However, in terms of index reliability and being the most preferred index in the literature, using the SPI provides more accurate drought characteristics.

Figure 3

Variation of SPI and ZSI values of Erzincan station in time: (a) monthly, (b) 3 monthly, (c) 12 monthly.

Figure 3

Variation of SPI and ZSI values of Erzincan station in time: (a) monthly, (b) 3 monthly, (c) 12 monthly.

Comparison of SPI and RAI values

RAI and SPI values are obtained by normalization of rainfall data and give close results in both indices (Figure 4). However, RAI is simpler, as the calculation procedure does not need to be fitted to any theoretical distribution of data, according to SPI. Furthermore, it is more sensitive in detecting extreme drought and wetlands since it is fluctuating in a wider range.

Figure 4

Variation of SPI and RAI values of Erzincan station in time: (a) monthly, (b) 3 monthly, (c) 12 monthly.

Figure 4

Variation of SPI and RAI values of Erzincan station in time: (a) monthly, (b) 3 monthly, (c) 12 monthly.

Comparison of SPI and SPEI values

It is seen that there is a good relationship between the indices for the Erzincan MOS in three time scales (1, 3, and 12 months). However, it was observed that the difference between the indices widened with the increase in the time period. Furthermore, while the values of SPEI were higher than the values of SPI between 1966 and 1997, the values of SPI were higher than the values of SPEI between 1997 and 2017. It has been determined that the increase in average temperatures and PET values after 1997 caused excessive water consumption (Yao et al. 2018) (Figures 5 and 6).

Figure 5

Erzincan MOS annual average temperature and PET values calculated with the Thornthwaite equation in time.

Figure 5

Erzincan MOS annual average temperature and PET values calculated with the Thornthwaite equation in time.

Figure 6

Variation of SPI and SPEI values of Erzincan station in time: (a) monthly, (b) 3 monthly, (c) 12 monthly.

Figure 6

Variation of SPI and SPEI values of Erzincan station in time: (a) monthly, (b) 3 monthly, (c) 12 monthly.

Rainfall is a critical meteorological factor of drought, but rising temperatures play an increasingly stronger role in influencing drought severity (Xu et al. 2015). Therefore, SPEI is a better indicator than SPI in terms of climate warming, water balance, and water use evaluation, as it is temperature-sensitive.

The second factor that determines the formation of drought is temperature. The higher the evaporation occurring on the surface of the ground due to the high temperature, the more effective the drought. The increase in temperature causes excessive water consumption by increasing the PET and thus low SPEI value. In arid regions, water consumption occurs mainly by evapotranspiration because there is enough energy (radiation) for evaporation. In the 21st century, the drought severities defined by SPEI are higher than the severities determined by SPI due to the increasing amount of evaporation.

When the 3-month and 12-month indices are analyzed, it is seen that the gap between indices has gradually widened in recent years. Furthermore, before 1997, SPI shows drier periods, while after 1997, SPEI shows drier periods. It was concluded that this situation was caused by evaporation and transpiration losses that occurred with the effect of increasing temperature after 1997 (Figure 6(b) and 6(c)).

Comparison of SPI and RDI values

SPI is obtained by fitting the cumulative precipitation series into a probability distribution and converting them into a standard normal distribution. The calculation of the index is achieved only by using precipitation values. RDI is an index calculated from the ratio of precipitation values to PET values. RDI gives more precise results as it considers both precipitation and average air temperatures. However, since PET values may be zero in short time periods, the index may be ambiguous. Figure 7 shows the comparison of these two indices.

Figure 7

Variation of the 12-monthly SPI and RDI values of Erzincan station in time.

Figure 7

Variation of the 12-monthly SPI and RDI values of Erzincan station in time.

When the index values are compared, it is seen that both indices behave in the same way. However, the difference between the index values is due to PET. Since meteorological factors such as temperature, sunshine, evaporation, and transpiration are taken into account in the RDI calculation, they can indicate the water balance in a region, and therefore the severity of the drought, more precisely than SPI. However, PET calculations can cause some errors because only temperature values are used. Monthly scales may not react rapidly, especially in rapidly developing drought events. As a result, it is seen that RDI performs better in drought monitoring for water resources planning and management in the study area.

Comparison of SPEI and RDI values

SPEI and RDI calculated using precipitation and PET values show similar results in determining drought (Figure 8). However, it can be said that SPEI, which is effective in all time periods, is somewhat superior since RDI calculation may cause ambiguous results in short time periods.

When five meteorological drought indices used in the study were compared, similar results were obtained. It is also concluded that each of these indices can be used for drought monitoring and water resource management. However, since the PET values are also taken into account for the SPEI and RDI index calculations, it gives some better results than the other indices. It is because temperature rising trends in the study area, which are affected by climate change, should also be taken into account. As the increase in air temperatures, evaporation, and transpiration are greatly affected, it triggers the formation of droughts.

There are several studies on performance evaluation of drought indices in the international literature. The results of these studies support our study to a great extent. Morid et al. (2006) compared the performance of seven indices (DI, PN, SPI, CZI, MCZI, ZSI, EDI) for drought monitoring in the Tehran province of Iran. As a result of the study, it was determined that SPI, CZI, and ZSI performed similarly in terms of drought detection, which is consistent with the results of our study. Adnan et al. (2018) determined the applicability and comparison of drought indices in Pakistan by evaluating the performance of 15 drought indices (SPI, SPEI, CZI, MCZI, ZSI, RVI, SSMAI, PDN, SC-PDSI, RDI, Weighted Anomaly Standardized Precipitation Index (WASPI), Composite ındex (CI), Percentage Area Weighted Departure (PAWD)). As a result of the study, it was determined that the performance of precipitation-based drought indices (MCZI, Z-score, CZI, WASPI, and RVI) is similar to SPI. These results coincide because the SPI and ZSI, which are among the precipitation-based indices used in our study, are similar, but they do not coincide because the precipitation-based SPI and RAI results are not similar. Furthermore, as a result of the study, it was determined that SPEI and RDI are the best drought indices as these two use evapotranspiration and precipitation, which is the same as the results of our study. Wable et al. (2019) assessed the performance of five drought indices (PDN, EDI, SPI, RDI, and SPEI) for a semi-arid basin located in western India. It was concluded that SPEI-9 is the most suitable drought index for monitoring drought conditions, which is similar to the results of our study. As a result of the literature review and the study, it was concluded that the SPEI index is the most efficient index.

CONCLUSIONS

In this study, correlation analysis, linear regression analysis, and variations of the indices over time were compared to determine the similarities, advantages, and disadvantages of meteorological drought indices. As a result of the analysis, it was concluded that meteorological indices generally have high statistical parameters (R2, RMSE) except for SPI-RAI and SPI-SPEI.

Similar results were obtained when the changes of meteorological drought indexes over time. However, it can be preferred for SYI, reliability, ZSI, ease of calculation, RAI, effectiveness in determining extreme arid and wet periods. SPEI and RDI, on the other hand, can be preferred because they are more sensitive to changing climatic conditions than other meteorological indices because they take into account evaporation and transpiration. It is also inevitable that SPEI and RDI, which depend on precipitation and PET, perform better than SPI, ZSI, and RAI, which depend only on precipitation. In addition, SPEI and RDI indicators are superior in terms of taking into account the climatic effects as they take temperature data into account. In observing droughts, RDI can be undefined in very cold regions (monthly average below zero) in monthly and 3-monthly time periods. For this reason, it is recommended to use the SPEI in drought monitoring systems for water resources planning and management in the Euphrates basin because it is more convenient and sensitive.

Figure 8

Variation of the 12-monthly SPEI and RDI values of Erzincan station in time.

Figure 8

Variation of the 12-monthly SPEI and RDI values of Erzincan station in time.

In order to develop the study, it is recommended to calculate PET values with the PM equation, to use more meteorology stations in a longer time period, and to take into account different drought indices. In addition, in future studies, it is recommended to examine the effects of meteorological parameters such as snowfall, wind speed, humidity, radiation, and atmospheric oscillations on various drought indices and drought characteristics.

ACKNOWLEDGEMENTS

The authors thank the General Directorate of Meteorology for the observed monthly total precipitation and temperature data provided, the Editor and the anonymous reviewers for their contributions to the content and development of this paper.

DATA AVAILABILITY STATEMENT

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

REFERENCES

Adnan
S.
Ullah
K.
Shuanglin
L.
Gao
S.
Khan
A. H.
Mahmood
R.
2018
Comparison of various drought indices to monitor drought status in Pakistan
.
Climate Dynamics
51
(
5–6
),
1885
1899
.
Akhtari
R.
Morid
S.
Mahdian
M. H.
Smakhtin
V.
2009
Assessment of areal interpolation methods for spatial analysis of SPI and EDI drought indices
.
International Journal of Climatology: A Journal of the Royal Meteorological Society
29
(
1
),
135
145
.
Ali Ghorbani
M.
Kazempour
R.
Chau
K. W.
Shamshirband
S.
Taherei Ghazvinei
P.
2018
Forecasting pan evaporation with an integrated artificial neural network quantum-behaved particle swarm optimization model: a case study in Talesh, Northern Iran
.
Engineering Applications of Computational Fluid Mechanics
12
(
1
),
724
737
.
Jacobs
A.
Heusinkveld
B.
Lucassen
D.
1998
Temperature variation in a class A evaporation pan
.
Journal of Hydrology
206
(
1–2
),
75
83
.
Kapluhan
E.
2013
Drought and drought in Turkey effect of agriculture
.
Marmara Geographical Review
27
,
487
510
.
Keyantash
J.
Dracup
J. A.
2002
The quantification of drought: an evaluation of drought indices
.
Bulletin of the American Meteorological Society
83
(
8
),
1167
1180
.
Li
X.
Sha
J.
Wang
Z.-L.
2019
Comparison of drought indices in the analysis of spatial and temporal changes of climatic drought events in a basin
.
Environmental Science and Pollution Research
26
(
11
),
10695
10707
.
McKee
T. B.
Doesken
N. J.
Kleist
J.
1993
The relationship of drought frequency and duration to time scales
. In
Proceedings of the 8th Conference on Applied Climatology
,
Boston, MA, USA
, pp.
179
183
.
Mishra
A. K.
Singh
V. P.
2010
A review of drought concepts
.
Journal of Hydrology
391
(
1–2
),
202
216
.
Moriasi
D. N.
Arnold
J. G.
Van Liew
M. W.
Bingner
R. L.
Harmel
R. D.
Veith
T. L.
2007
Model evaluation guidelines for systematic quantification of accuracy in watershed simulations
.
Transactions of the ASABE
50
(
3
),
885
900
.
Morid
S.
Smakhtin
V.
Moghaddasi
M.
2006
Comparison of seven meteorological indices for drought monitoring in Iran
.
International Journal of Climatology: A Journal of the Royal Meteorological Society
26
(
7
),
971
985
.
Myronidis
D.
Fotakis
D.
Ioannou
K.
Sgouropoulou
K.
2018
Comparison of ten notable meteorological drought indices on tracking the effect of drought on streamflow
.
Hydrological Sciences Journal
63
(
15–16
),
2005
2019
.
Panu
U.
Sharma
T.
2002
Challenges in drought research: some perspectives and future directions
.
Hydrological Sciences Journal
47
(
1
),
S19
S30
.
Santhi
C.
Arnold
J. G.
Williams
J. R.
Dugas
W. A.
Srinivasan
R.
Hauck
L. M.
2001
Validation of the SWAT model on a large river basin with point and nonpoint sources
.
Journal of the American Water Resources Association
37
(
5
),
1169
1188
.
Shamshirband
S.
Hashemi
S.
Salimi
H.
Samadianfard
S.
Asadi
E.
Shadkani
S.
Kargar
K.
Mosavi
A.
Nabipour
N.
Chau
K. W.
2020
Predicting standardized streamflow index for hydrological drought using machine learning models
.
Engineering Applications of Computational Fluid Mechanics
14
(
1
),
339
350
.
Sivakumar
M. V.
Motha
R.
Wilhite
D.
Wood
D.
2011
Agricultural Drought Indices
. In
Proceedings of an Expert Meeting
,
2–4 June, 2010
.
WMO
,
Murcia, Spain
.
Smakhtin
V. U.
Hughes
D. A.
2007
Automated estimation and analyses of meteorological drought characteristics from monthly rainfall data
.
Environmental Modelling & Software
22
(
6
),
880
890
.
Thornthwaite
C. W.
1948
An approach toward a rational classification of climate
.
Geographical Review
38
(
1
),
55
94
.
Tigkas
D.
2008
Drought characterisation and monitoring in regions of Greece
.
European Water
23
(
24
),
29
39
.
Tsakiris
G.
Vangelis
H.
2004
Towards a drought watch system based on spatial SPI
.
Water Resources Management
18
(
1
),
1
12
.
Tsakiris
G.
Pangalou
D.
Tigkas
D.
Vangelis
H.
2007
Assessing the areal extent of drought
. In:
Water Resources Management, New Approaches and Technologies
(G. P. Karatzas, M. P. Papadopoulou & K. P. Tsagarakis, eds).
European Water Resources Association
,
Chania, Crete, Greece
,
14–16 June 2007
.
Van Liew
M. W.
Arnold
J. G.
Garbrecht
J. D.
2003
Hydrologic simulation on agricultural watersheds: choosing between two models
.
Transactions of the ASAE
46
(
6
),
1539
1551
.
Van Rooy
M.
1965
A rainfall anomally index independent of time and space
.
Notos
14
,
43
48
.
Vicente-Serrano
S. M.
Beguería
S.
López-Moreno
J. I.
2010
A multiscalar drought index sensitive to global warming: the standardized precipitation evapotranspiration index
.
Journal of Climate
23
(
7
),
1696
1718
.
Wable
P. S.
Jha
M. K.
Shekhar
A.
2019
Comparison of drought indices in a semi-arid river basin of India
.
Water Resources Management
33
(
1
),
75
102
.
Wilhite
D. A.
2005
Drought and Water Crises: Science, Technology, and Management Issues
.
CRC Press
,
Boca Raton, FL
,
USA
.
Wilhite
D.
Pulwarty
R. S.
2017
Drought and Water Crises: Integrating Science, Management, and Policy
.
CRC Press
,
Boca Raton, FL, USA
.
Wu
H.
Hayes
M. J.
Weiss
A.
Hu
Q.
2001
An evaluation of the standardized precipitation index, the China-Z index and the statistical Z-Score
.
International Journal of Climatology: A Journal of the Royal Meteorological Society
21
(
6
),
745
758
.
Xu
C.
Li
J.
Zhao
J.
Gao
S.
Chen
Y.
2015
Climate variations in northern Xinjiang of China over the past 50 years under global warming
.
Quaternary International
358
,
83
92
.