Introduction

Drought is defined as dry conditions over a specific period. Drought impacts are not immediately obvious but develop slowly in comparison with natural disasters, such as floods, storms, and tornadoes (Bayissa et al. 2018; Ding et al. 2021). Drought damages occur in different aspects of the environment and economic, such as drying up of rivers, forest fires, organism and crop losses, and water shortage (Wilhite et al. 2007; Wang et al. 2021). Drought’s characteristics must be investigated for reducing damages, early warning preparedness, and contingency planning (Wilhite 1993; Kogan 2000; Wilhite and Svoboda 2000; Arab and Elyasi 2010). Drought is mainly grouped into four major categories: meteorological, agricultural, hydrological, and socioeconomic, generally occurring in a particular order (Wilhite and Glantz 1985; Mishra and Singh 2010; Tu et al. 2018; Zuo et al. 2022). Meteorological drought which is induced by precipitation deficiency or water shortage (Tu et al. 2018) impacts soil moisture, i.e., agricultural drought (Wu et al. 2020), and consequently, low recharge from soil to streams, lakes, and groundwater causes hydrological drought (Mishra and Singh 2010; Zargar et al. 2011). Increasing the gap between the demand and supply of goods and commodities due to a water shortage refers to a socioeconomic drought (Wilhite and Glantz 1985; Zseleczky and Yosef 2014; Guo et al. 2019). Many indices have been developed and applied by researchers for monitoring drought. Details on different drought indices can be found in drought reviews (e.g. Mishra and Singh 2010; Zargar et al. 2011; Xu et al. 2015) and drought index books (Svoboda and Fuchs 2016; Dalezios et al. 2017). Some researchers have investigated drought characteristics by a single index (Morid et al. 2007; Miah et al. 2017; Shi et al. 2018; Yao et al. 2018; Forootan et al. 2019; Simsek 2021), and some studies have focused on the relationship between different drought indices (Jain et al. 2015; Bayissa et al. 2018; Mohammad et al. 2018; Zhao et al. 2019). Wang et al. (2019) studied correlation between meteorological and hydrological droughts in a basin in China and concluded that hydrological drought occurred later than meteorological droughts. Due to the complexity of the drought events and variation in meteorological and physical characteristics of any region, drought indices may vary from place to place (Jain et al. 2015). Therefore, it is necessary to understand the occurrence of drought at different temporal scales and recognize an appropriate drought index for interested region, which helps to alleviate drought-related disasters (Zhang et al. 2021). Drought events have intermittently occurred in different parts of Iran and have affected the country harmfully, including water resources and ecological, biological, agricultural, social, and economic aspects over the past decades (Rahmaninan 2000). Geographical and seasonal changes of precipitation amount in different parts of Iran, varying from 50 to 1500 mm, led to various drought phenomena (Abbaspour and Sabetraftar 2005). Due to this variety of the spatio-temporal distribution of precipitation and the frequent occurrence of drought in Iran, it is necessary to study this phenomenon to manage water resources. This necessity is vital in the southwestern Iran, including Dez catchment, where most of the water demands are supplied from surface water resources. Therefore, this study aims at evaluating and comparing the performance of six drought indices to acquire a better knowledge of drought propagation. The results can help to inform future early warning and monitoring systems for hydrological and agricultural droughts.

Data and methods

Study area

The study area is upper Dez basin located in the southwestern of Iran, which spreads over three provinces, with an area of 16,150 km2. Upper Dez is located between 48° 20′ to 50° 20′ N and 32° 30′ to 34° 7′ E (Fig. 1). Climate conditions vary from north to south and west to east due in part to the difference between minimum (−31 °C) and maximum average temperatures (47.8 °C). The average annual evaporation is 1985 mm, and the average annual sunshine of 3020 h. The altitude ranges from 490 to 4260 m above sea level (a.s.l) with an annual precipitation of 500 mm. Cropland (32.5%), forest (32.5%), and pastures (35.0%) mainly cover the area. There are two main aquifers in the region, named Droud-Borujerd aquifer, where the average depth to groundwater table ranges from 0.5 to 46 m, and the Azna-Aligudarz aquifer, where groundwater depth ranges from 1 to 53 m. Aligudarz, Azna, Darud, and Borujard are the region’s most important cities, with a population of 713,430 people. The main activity in this region is agriculture, which is mainly affected by drought.

Fig. 1
figure 1

Location map of study area

Full size image

Datasets

Stream flow discharge, precipitation, groundwater level records, and soil moisture were the hydrological and meteorological data used to analyze the drought indices. Daily discharge data from hydrometry stations and monthly groundwater level records were obtained from National Iranian Water Resources Management Company archive (https://data.wrm.ir). Also, the daily precipitation was obtained from both National Water Resources Management Company and Meteorological Organization of Iran data banks (https://www.irimo.ir/far/index.php). Drought indices, variables, and dataset are summarized in Table 1. For monitoring standardized precipitation index (SPI) and standardized stream flow index (SQI), 35 years of monthly time series on discharge and climatological parameters were supplied (Table 1). The spatial distribution of drought was investigated at nine selected climatic stations, which involved the same time series length as the stream flow gauging data. Daily base flow and direct runoff data were separated and used to investigate the standardized surface runoff index (SRI) and base flow index (SBFI). Soil moisture data (SM) obtained from the GLDAS CLM dataset at 1° × 1° per month (Rodell and Beaudoing 2007) was applied to monitor the standardized soil moisture index (SSI). The standardized groundwater level index (SGI) was monitored using data obtained from 75 observation wells during 2003–2017 (Fig. 1). Time series of groundwater levels in each observation well were used to generate the groundwater unit hydrograph (GUH) for both aquifers. Then SGI was calculated in terms of GUH fluctuations.

Table 1 Drought indices, variables, and time series of dataset
Full size table

Methodology

Drought indices

The drought indices applied depict three of drought categories: meteorological, agricultural, and hydrological droughts. The drought indices were the standardized precipitation index (SPI) (McKee et al. 1995), standardized runoff index (SRI) (Shukla and Wood 2008), standardized stream flow index (SQI) (Nalbantis and Tsakiris 2009), standardized base flow index (SBFI), standardized soil moisture index (SSI) (Xu et al. 2018), and standardized groundwater level index (SGI) (Kubicz 2018b; Halder et al. 2020). The drought indices were selected based on data availability, simple algorithm, and applicability to monitoring drought propagation. Drought indices were calculated using Eq. (1):

$${DI}_{\textrm{i},\textrm{k}}=\frac{x_{\textrm{i},\textrm{k}}-{\overline{x}}_{\textrm{k}}}{S_{\textrm{k}}}$$
(1)

where DIi, k is the standardized drought index, xi, k stands for the variable values, \({\overline{x}}_{\textrm{k}}\) is for the mean, and Sk is for the standard deviation of variables at year i, and kth period, respectively. Equation (1) is applied to all drought indices based on the concept used by McKee et al. (1993) for SPI, who suggested applying the procedure to other drought-related variables. Drought indices were calculated for different time scales, including 1-, 3-, 6-, 9-, 12-, 18-, and 24-month periods and spatial distribution over the region. The drought classification based on DI is shown in Table 2 (McKee et al. 1993).

Table 2 SPI values corresponding to different categories of drought severity (after McKee et al. 1993)
Full size table

Drought correlation analysis

The Pearson correlation method was used to analyze the relationship between the three major drought categories. The Pearson coefficient (Eq. 2) is a criterion that shows linearity between two datasets and has been used in many similar studies (Jain et al. 2015; Ding et al. 2021). The correlation coefficient (CC) was calculated between all drought indices, and the maximum coefficient was selected to show differences in response to spatial and temporal variation in the drought types.

$$r=\frac{1}{n}\sum \limits_{i=1}^n\left(\frac{x_{\textrm{i}}-\overline{x}}{S_{\textrm{x}}}\right)\left(\frac{y_{\textrm{i}}-\overline{y}}{S_{\textrm{y}}}\right),$$
(2)

where n stands for the number of dataset values, xi and yi are the values of the arrays, \(\overline{x}\) and \(\overline{y}\) are the mean of two arrays, and Sx and Sy are the standard deviation of two arrays.

Results and discussion

Relation between meteorological and hydrological droughts

The maximum correlation coefficient (MCC) was used to investigate propagation between meteorological (SPI) and hydrological droughts (SQI, SRI, and SFBI) for 1-, 3-, 6-, 12-, 18-, 24-month temporal scales. Figure 2 shows the heat map of the correlation coefficient between the SPI and SQI for the selected meteorological stations. From Fig. 2, as the time scale was lengthened, the MCC responded strongly, and the MCC for SQI reached when both indices were compared for the 24-month time scale (CC=0.74). Because the SQI is based on total stream flow, so there might be a specific lag time with the SPI. To find the general lag time between SPI and SQI, the MCC was extracted by averaging each row of the heat map grid (Table 3). Results showed the MCC ranged from 0.05 to 0.53, which is dominantly related to 12-month SPI. This behavior is in convenient with the results of the previous study (Bayissa et al. 2018). As expected, the lowest CC belonged to SPI-1 with SQI-24. This discrepancy might be related to the attenuation of the 1-month precipitation effect on the 24-month discharge of the basin. Additionally, the MCC value was spatially related to the Sorkhab flow gauging station, located downstream of the basin. Conversely, upstream stations showed the lowest correlation, such as Gale Rood and Kazem Abad (Fig. 2). It suggests that low-discharge rivers may not be a good indicator for hydrological drought assessment.

Fig. 2
figure 2

Pearson correlation coefficient between SPI and SQI for 1-, 3-, 6-, 12-, 18-, and 24-month time scales at different sub-basins: Azna Cham (a), Tire Dorood (b), Absabzeh (c), Sorkhab (d), Kazem Abad (e), Kamandan (f), Gale Rood (g), Sekaneh (h), and Silakhor (i)

Full size image
Table 3 Average correlation coefficient for different SPI time scales (1-, 3-, 6-, 12-, 18-, and 24-month) for each period of hydrological drought indices
Full size table

Correlation between surface runoff index (SRI) and SPI showed that the best CC was reached when both indices had the same time scales (Fig. 3). For example, in most of the sub-basins, SPI-1 correlated better with SRI-1 or SPI-3 correlated better with SRI-3 and so on. Consequently, SRI responded faster to the propagation of meteorological drought than SQI because SRI was only based on direct runoff. These findings support the results of research by Li et al. (2016), who reported a slight time lag between SPI and SRI. The MCC was found downstream in the Sorkhab region (Fig. 3d), possibly due to the accumulation of runoff concentrates toward downstream and interflow delay effects. The SRI index at the 1-month time scale explained better hydrological drought among the other time scales. As a result, the SRI can be used as a useful complement to the SPI on monthly to seasonal time scales (Shukla and Wood 2008).

Fig. 3
figure 3

Pearson correlation coefficient between SPI and SRI for 1-, 3-, 6-, 12-, 18-, and 24-month time scales at different sub-basins: Azna Cham (a), Tire Dorood (b), Absabzeh (c), Sorkhab (d), Kazem Abad (e), Kamandan (f), Gale Rood (g), Sekaneh (h), and Silakhor (i)

Full size image

The correlation coefficient between standardized base flow index (SBFI) and SPI increased as the time scale for SPI lengthened (Fig. 4). Generally, the trend of relation between SBFI and SPI was similar to the SQI and SPI relation. Base flow in most of the sub-basins correlated better with more extended time scales of the SPI. However, the SBFI index was expected to respond with a lag time to meteorological droughts. The correlation coefficient was different spatially because SBFI depends on many catchment factors, such as permeability of soil layers, catchment area, and location of rivers in snow-dominated regions (Sutanto and Van Lanen 2022).

Fig. 4
figure 4

Pearson correlation coefficient between SPI and SBFI for 1-, 3-, 6-, 12-, 18-, and 24-month time scales at different sub-basins: Azna Cham (a), Tire Dorood (b), Absabzeh (c), Sorkhab (d), Kazem Abad (e), Kamandan (f), Gale Rood (g), Sekaneh (h), and Silakhor (i)

Full size image

Mean values of hydrological drought index were calculated by averaging the correlation coefficient of all the SPI time scales (1-, 3-, 6-, 12-, 18-, and 24-month) corresponding to each of the hydrological index time scales. In the other words, the values of each row in the heat map were averaged for each sub-basin and presented in Table 3. Results showed that SRI had the weakest averaged correlation coefficient over the region. This does not mean that SRI is not correlated to meteorological drought. It reveals several factors, for example, seasonal water harvesting and consuming for agricultural demands, independent from precipitation and affecting the SRI. However, the SRI showed drought propagation faster than SQI and SBFI, although it had low CC. The most significant averaged correlation coefficient was reached using SQI, and in the region, SQI better described propagation of meteorological drought than SRI and SBFI.

Relation between SPI and SGI droughts

The spatial average of the SPI was calculated for the corresponding period of the SGI time scales to show the effects of meteorological drought propagation on groundwater level decline. Statistically, the correlation coefficient values increased when the SPI time scale extended from 1 to 24 months which is consistent with the study by Kubicz (2018a). When the same time scales were compared, longer time scales showed weaker correlation coefficient (Fig. 5). For example, the maximum correlation coefficient was achieved between SPI-24 and SGI-03 for both aquifers (MCC=0.63 and 0.39) which means that seasonal decreasing in groundwater table correlated stronger with more extended meteorological drought. Comparing SGI and SBFI for investigating groundwater drought, the SGI showed a stronger correlation. The simultaneous MCC between SPI and SGI is presented in Fig. 6. As it can be found, several periods with normal, moderate, and severe drought events have taken place during the long-term time series of precipitation. A sever meteorological drought was monitored from 2008 to 2009, reflecting slightly to considerable groundwater level delays. The lag time varied from 3 to 9 months in Droud-Borujerd and Azna-Aligudarz aquifers, respectively. Recorded precipitation deficits from 2008 reflected a continuous declining trend of groundwater level trend in the study area. The results of propagation drought from meteorological to groundwater showed that if it is impossible to monitor groundwater drought, the SPI-24 indicator can be used to investigate the risk of groundwater resource drought.

Fig. 5
figure 5

Pearson correlation coefficient between SPI and SGI for 1-, 3-, 6-, 12-, 18-, and 24-month time scales in Droud-Borujerd (a) and Azna-Aligudarz (b) aquifers

Full size image
Fig. 6
figure 6

Plots of SPI-24 versus SGI-3: Droud-Borujerd (a) and Azna-Aligudarz (b) aquifers

Full size image

Relation between SPI and SSI droughts

The standardized soil moisture index (SSI) was calculated based on the GLDAS CLM dataset per month, which corresponded to different soil layers, including 0–10 cm, 10–40 cm, 40–100 cm, and 100–200 cm. Figure 7 shows the correlation coefficient between the spatially average SPI 1- to 24-month time scales and corresponding SSI for different soil depths. At the topsoil (0–10 cm), MCC (r > 0.7) was found for time scales 18- and 24-month (Fig. 7a). The correlation between the SSI-1 and all the SPI time scales was relatively constant and low (r <0.12). The fluctuation of soil moisture at the ground surface is affected by many factors, such as root water uptake and evaporation. Therefore, the topsoil moisture conditions cannot be a good indicator for agricultural drought. For the middle soil depths (10–40 and 40–100 cm), results of correlation coefficient were almost similar (Fig. 7b and c), showing the MCC = 0.71 for SPI time scales 1-, 18-, and 24-month. At deeper depths, correlations tend to be constant at longer time scales of SSI. For the deepest soil layer (100–200 cm), MCC = 0.85 corresponded to SSI-12 and SPI-1 (Fig. 7d). It might be related to the reason that the soil moisture drought occurs more continuously, whereas meteorological drought occurs frequently but in short period.

Fig. 7
figure 7

Correlation coefficient between spatially average SPI for different time scales with SSI at 10 cm (a), 40 cm (b), 100 cm (c), and 200 cm (d) depths

Full size image

Conclusion

Understanding their relationship is essential to investigate drought propagation among meteorological, hydrological, and agricultural draughts. The SPI, SRI, SQI, SBFI, SSI, and SGI were applied to investigate the drought types in the upper Dez basin southwest of Iran. The correlation and propagation among the drought types were analyzed. The main conclusions reached as follows:

  1. (1)

    Results of comparing the SPI and SQI showed longer time scale of SPI correlated better with SQI and the highest correlation was reached when both indices had the same time scale of 24 months (CC=0.74). Comparing the SPI and SRI showed that the best CC was found when both indices had the same time scales, and conversely, the CC was increased between SBFI and SPI as the time scale for SPI increased.

  2. (2)

    Correlation between SPI and SGI resulted in the MCC between SPI-24 and SGI-03: MCC = 0.63 and MCC = 0.39 for Droud-Borujerd and Azna-Aligudarz aquifers, respectively. This result revealed seasonal depletion in the groundwater table correlated stronger with more extended meteorological drought. In deficit of groundwater drought monitoring, the SPI-24 indicator can be used to investigate the risk of groundwater resource drought.

  3. (3)

    Correlation between SPI and SSI was better as time scales increased, and at deeper soil depths, correlations tend to be constant. For the deepest soil layer (100–200 cm), MCC = 0.85 corresponded to SSI-12 and SPI-1. Correlations between SPI-SQI and SPI-SSI were stronger to reveal propagation from climatological drought.

Our results have important implications for adequate drought warning and prevention, and future studies should combine drought indices to extract an optimal index for predicting water resources situation.