Abstract

Hydrological drought plays an important role in planning and managing water resources systems to meet increasing water demands due to population growth. In this study, the effects of climate change on the hydrological drought characteristics of the Gharasu basin, as one of the major sub-basins of the Karkheh river basin, are investigated. This river basin has experienced severe droughts, and floods, in recent years. The uncertainties in projected drought conditions are characterized based on a suite of 34 general circulation models (GCMs). Based on hydrological simulations over the historical period, 12 GCMs are selected to estimate projected runoff values and the corresponding streamflow drought index (SDI) in the future period. The ‘run theory’ is applied to evaluate the drought characteristics under Representative Concentration Pathways (RCPs) 4.5 and 8.5 emission scenarios. Results show that uncertainties of drought projection under RCP8.5 are higher than under RCP4.5, where among different drought characteristics, the maximum uncertainty is detected for drought severity and maximum drought duration. Moreover, the uncertainty of drought projection in wet periods is greater than that in dry periods.

HIGHLIGHTS

  • 34 different General Circulation Models (GCMs) are employed to calculate the runoff time series of Gharesu river basin in Iran.

  • The future drought projection and estimation of streamflow drought index (SDI) is carried out.

  • Applying the Run Theory, the drought characteristics are derived under two CMIP5 RCP scenarios.

  • Comparing the obtained results from different GCMs, the uncertainties in estimating drought characteristics are analyzed.

Graphical Abstract

Graphical Abstract
Graphical Abstract

INTRODUCTION

Drought is a natural and multidimensional phenomenon that can persist over several years with significant negative impacts on various aspects of human life such as the environment, industry, and economy. Droughts are broadly divided into four categories: meteorological, hydrological, agricultural, and socio-economic. While these types of droughts are closely related and interact with each other (Wilhite & Glantz 1985; Orville 1990; Adib et al. 2018), hydrological drought plays an important role in planning and managing water resources systems due to the increasing water demand and population growth (Nalbantis 2008; Shukla & Wood 2008; Forzieri et al. 2013; Tabari et al. 2013; Ashrafi & Mahmoudi 2019). Drought indices are generally used to assess drought conditions in an area and examine its various characteristics including severity, duration, and spatial expanse (Mishra & Singh 2010). Several studies have developed methodologies to assess the hydrological and meteorological drought indices (Tabari et al. 2013; Mondal & Mujumdar 2015; Yeh et al. 2015; Moazami et al. 2016; Gumus & Algin 2017; Trinh et al. 2017a, 2017b; Myronidis et al. 2018), including the streamflow drought index (SDI) developed by Nalbantis (2008).

Because of nonstationarity of the hydroclimate system due to climate change, prediction of future droughts and study of their intensity and duration are considered a challenge (Najafi et al. 2017). Successive decreases of water resources due to drought conditions make integrated water resources management more complicated. Therefore, decision-makers should develop strategies to balance available resources and existing demands to reduce system shortages (Ashrafi & Dariane 2017), mitigate environmental disasters (Bakhsipoor et al. 2019), and enhance firm energy production (Schaefli et al. 2007; Wang et al. 2014; Zallaghi et al. 2020). Numerous studies have shown increases in drought severity in response to global warming worldwide (Orlowsky & Seneviratne 2013; Van Loon & Laaha 2015; Trinh et al. 2017a, 2017b; Ishida et al. 2018; Zhang et al. 2018; Jiao & Yuan 2019; Jehanzaib et al. 2020). Prudhomme et al. (2014) projected an increase in the severity of hydrological drought in the late 21st century. They used five atmosphere–ocean general circulation models (GCMs) from the Coupled Model Intercomparison Project Phase 5 (CMIP5) archive under four emission scenarios. Tan et al. (2014) examined the impacts of climate change on streamflow in the Johor river basin in Malaysia. They highlighted the importance of analyzing various climate change scenarios for hydrological assessments. McDonald et al. (2008) found a potential linkage between radiation force, prolonged drought and the hydrological effects of future climate warming in southern and western California. Sheffield et al. (2012) found small changes in meteorological drought conditions at the global scale, while Dai (2013) pointed to the increase in meteorological droughts under global warming worldwide. Trenberth et al. (2014) predicted that changes in the global water cycle in the 21st century would not be uniform and climate change would raise the weather system temperature and, in many cases, this would dry out the water resource leading to droughts.

Several factors contribute to the high level of uncertainty in future drought estimations, including GCM structure and initial conditions, the assumptions made in downscaling (Fu et al. 2009; Blöschl & Montanari 2010), emission scenarios, methods used to bias correct GCM outputs and observed datasets (Ehret et al. 2012; Singh & Najafi 2020). The main objective of this study is to characterize the uncertainty of projected hydrological drought conditions of the Gharesu river basin under climate change. In order to assess the uncertainties resulting from climate change modeling, 34 GCMs are used in this research. The achieved results of all 34 GCMs are downscaled and the GCM models that can represent the historical basin runoff well are selected for future runoff simulations at basin scale. For this purpose, using the observational data over the base period, the 12 best models are selected for estimating runoff in future periods. The streamflow drought index (SDI) is then calculated using the runoff time series in the Gharasu watershed for the future period (2006–2030). The drought characteristic for this period are assessed using the results of different models and the range of changes and uncertainties of the characteristics of the drought index for the future period are examined. Frequency analysis is performed for different SDI values and the corresponding uncertainties are reported.

The Gharesu river basin is one of the main sub-basins of the Karkheh river basin. Analysis of the hydrological changes of this sub-basin is highly demanded due to the presence of numerous agricultural, industrial, and urban sites downstream of the Karkheh reservoir. As discussed by Ashrafi et al. (2017), the Karkheh river basin is one of the most important basins in Iran and plays a key role in the development of the western part of the country. Considering recent climatologic and hydrologic changes that have led to drought conditions, it is essential to investigate different probable future scenarios for this area for decision-makers of water resources systems of the basin.

The Gharesu watershed has experienced severe climate variability in recent years that have led to severe droughts and unexpected intense floods. The most recent severe flood in March 2019 inflicted great damage on Khuzestan Province. Therefore, targeted studies have started to investigate climate change in this watershed. There are several sources of uncertainty in projected changes in watershed runoff using GCMs. In addition to identifying the sources of these uncertainties, it is very important to estimate the levels of these uncertainties in order to obtain practical results. Therefore, this study investigates the uncertainties in estimating drought indices using results obtained from GCMs.

METHODOLOGY AND STUDY AREA

The historical data from 34 GCMs are first extracted for the base period (1981–2005). These data are then downscaled using the BCSD statistical downscaling method (Najafi et al. 2017; Li et al. 2018). The basin runoff is calculated for the base period using the calibrated rainfall-runoff (IHACRES) model in the base period. Study of the performance exhibited by different models in simulating the runoff of the Gharesu river basin led to the selection of the best 12 models. Although the correlation coefficient is a useful indicator to assess the goodness of fit for linear models, it can be a poor estimator of goodness of fit for nonlinear models because of model bias (McCuen et al. 2006). Considering such limitations, the Nash–Sutcliffe efficiency coefficient has become more popular in hydrological modeling and performed research which adopted GCM modeling (Najafi et al. 2011; Saghafian & Mehdikhani 2014; McMahon et al. 2015). Therefore, the best models are selected based on the values of the Nash–Sutcliffe efficiency coefficient (Nash & Sutcliffe 1970).

Future climate data on temperature and precipitation for the selected models are next extracted under two, RCP4.5 and RCP8.5, emission scenarios for the future period (2006–2030). The BCSD statistical method is utilized for downscaling the extracted precipitation and temperature data from different GCMs and the change factor (CF) method is applied for correcting the errors in the future period. After calibrating and validating the IHACRES rainfall-runoff model, it is implemented to simulate and project the future runoff (from October 2006 to September 2030) under the two emission scenarios RCP4.5 and RCP8.5. The streamflow drought index (SDI) is then utilized at 3-, 6-, 9- and 12-month timescales to investigate future drought conditions. Finally, employing the ‘run theory’ (Yevjevich 1967), the characteristics of the predicted droughts are derived, and the results are analyzed. Applying the results of different models and the range of variations, we are able to consider the uncertainties of the proposed modeling and determine the most and less reliable results. Figure 1 presents the methodology of this study schematically.

Figure 1

The flowchart of the research.

Figure 1

The flowchart of the research.

Figure 2

The geographical position of the Gharesu river basin in Iran.

Figure 2

The geographical position of the Gharesu river basin in Iran.

As shown in Figure 1, after calculating SDI in 3, 6, 9 and 12 months for RCP4.5 and RCP8.5 scenarios, the run theory is applied to derive the characteristics of the predicted drought events in different scenarios. Comparing the results achieved from different GCM models, the uncertainties of drought projection become more visible.

Study area

The Gharesu sub-basin is located in the northwest of the Karkheh river basin (Figure 2). It is one of the five main sub-basins of the Karkheh river basin that is situated in Kermanshah Province and is classified as a semi-arid basin. The average elevation and temperature of the basin are 1,555 m and 14.4 °C, respectively, and its peak annual discharge is generated from snowmelt in spring (Samadi et al. 2013a, 2013b).

The Gharesu sub-basin covers an area of approximately 5,656 km2 with elevation ranging from 1,237 to 3,350 masl (with the average of 1,555 masl). The mean, minimum, and maximum annual temperatures in this sub-basin are 14.4, −5.8, and 29.6 °C, respectively. The Gharesu river is influenced by heavy winter precipitation, which results in peak annual runoff in spring (Samadi et al. 2013b). In this study, daily data obtained from six rain gage stations, three synoptic meteorological stations, and one hydrometric station during a period of 25 years (1981–2005) are used. The time series for the observed monthly runoff, which is calculated by taking the average of the daily observed runoff for the base period (1981–2005), is employed to calibrate and validate the rainfall-runoff model.

The observed and calculated data

The daily data from six rain gage stations, three synoptic weather stations and one hydrometric station for 25 years (1981–2005) are used (Figure 2) in this study. The data from all rain gages and synoptic stations for each variable are transformed into a single time series using the Thiessen-polygon method.

The outputs of the temperature and precipitation variables from the 34 coupled GCMs in the CMIP5 are also applied. Since the outputs of the GCMs are in macro-scale, the BCSD statistical method is implemented to downscale them to 0.5° × 0.5° resolution so that they could be used for regional assessments (Najafi et al. 2017; Li et al. 2018). Table 1 summarizes the characteristics of all considered GCM models.

Table 1

The considered CMIP5 GCMs

Modeling centerResolution (latitude × longitude)Model name
Commonwealth Scientific and Industrial Research Organization and Bureau of Meteorology, Australia 1.25° × 1.87° ACCESS1-0 
Commonwealth Scientific and Industrial Research Organization and Bureau of Meteorology, Australia 1.25° × 1.87° ACCESS1-3 
Beijing Climate Center 2.77° × 2.81° bcc-csm1-1 
Beijing Climate Center 1.12° × 1.12° bcc-csm1-1-m 
College of Global Change and Earth System Science, Beijing Normal University 2.8° × 2.8° BNU-ESM 
Canadian Center for Climate Modelling and Analysis 2.77° × 2.81° CanESM2 
National Center for Atmospheric Research, Canada 1.25° × 0.94° CCSM4 
Community Earth System Model Contributors 1.25° × 0.94° CESM1-BGC 
Community Earth System Model Contributors 1.25° × 0.94° CESM1-CAM5 
Euro-Mediterranean Center on Climate Change, Italy 0.74° × 0.75° CMCC-CM 
National Center for Meteorological Research/European 1.4° × 1.4° CNRM-CM5 
Australia's Commonwealth Scientific and Industrial 1.8° × 1.8° CSIRO-Mk3-6-0 
EC-Earth Consortium 1.125° × 1.125° EC-EARTH 
Atmospheric Sciences and Geophysical Fluid Dynamics/Institute of Atmospheric 2.°° × 2.8° FGOALS-g2 
Atmospheric Sciences and Geophysical Fluid Dynamics/Institute of Atmospheric Physics, Chinese Academy of Sciences, China 1.65° × 2.81° FGOLS-S2 
The First Institute of Oceanography, SOA, China 2.81° × 2.81° FIO-ESM 
NOAA Geophysical Fluid Dynamics Laboratory, USA 2.0° × 2.5° GFDL-CM3 
NOAA Geophysical Fluid Dynamics Laboratory, USA 2.02° × 2.00° GFDL-ESM2G 
NOAA Geophysical Fluid Dynamics Laboratory, USA 2.02° × 2.50° GFDL-ESM2M 
NASA Goddard Institute for Space Studies, USA 2.02° × 2.50° GISS-E2-R 
Met Office Hadley Centre, UK 1.25° × 1.25° HadGEM2-CC 
Met Office Hadley Centre, UK 1.25° × 1.25° HadGEM2-ES 
Institute for Numerical Mathematics, Russia 1.5° × 2.0° inmcm4 
Institute Pierre-Simon Laplace, France 1.89° × 3.75° IPSL-CM5A-LR 
Institute Pierre-Simon Laplace, France 1.26° × 2.00° IPSL-CM5A-MR 
Institute Pierre-Simon Laplace, France 1.89° × 3.75° IPSL-CM5B-LR 
Atmosphere and Ocean Research Institute (The University of Tokyo) 1.4° × 1.4° MIROC5 
Atmosphere and Ocean Research Institute (The University of Tokyo) 2.79° × 2.81° MIROC-ESM 
Atmosphere and Ocean Research Institute (The University of Tokyo) 2.79° × 2.81° MIROC-ESM-CHEM 
Max Planck Institute for Meteorology, Germany 1.86° × 1.87° MPI-ESM-LR 
Max Planck Institute for Meteorology, Germany 1.86° × 1.87° MPI-ESM-MR 
Meteorological Research Institute, Japan 1.12° × 1.12° MRI-CGCM3 
Norwegian Climate Center, Norway 1.89° × 2.50° NorESM1-M 
Norwegian Climate Center, Norway 1.89° × 2.50° NorESM1-ME 
Modeling centerResolution (latitude × longitude)Model name
Commonwealth Scientific and Industrial Research Organization and Bureau of Meteorology, Australia 1.25° × 1.87° ACCESS1-0 
Commonwealth Scientific and Industrial Research Organization and Bureau of Meteorology, Australia 1.25° × 1.87° ACCESS1-3 
Beijing Climate Center 2.77° × 2.81° bcc-csm1-1 
Beijing Climate Center 1.12° × 1.12° bcc-csm1-1-m 
College of Global Change and Earth System Science, Beijing Normal University 2.8° × 2.8° BNU-ESM 
Canadian Center for Climate Modelling and Analysis 2.77° × 2.81° CanESM2 
National Center for Atmospheric Research, Canada 1.25° × 0.94° CCSM4 
Community Earth System Model Contributors 1.25° × 0.94° CESM1-BGC 
Community Earth System Model Contributors 1.25° × 0.94° CESM1-CAM5 
Euro-Mediterranean Center on Climate Change, Italy 0.74° × 0.75° CMCC-CM 
National Center for Meteorological Research/European 1.4° × 1.4° CNRM-CM5 
Australia's Commonwealth Scientific and Industrial 1.8° × 1.8° CSIRO-Mk3-6-0 
EC-Earth Consortium 1.125° × 1.125° EC-EARTH 
Atmospheric Sciences and Geophysical Fluid Dynamics/Institute of Atmospheric 2.°° × 2.8° FGOALS-g2 
Atmospheric Sciences and Geophysical Fluid Dynamics/Institute of Atmospheric Physics, Chinese Academy of Sciences, China 1.65° × 2.81° FGOLS-S2 
The First Institute of Oceanography, SOA, China 2.81° × 2.81° FIO-ESM 
NOAA Geophysical Fluid Dynamics Laboratory, USA 2.0° × 2.5° GFDL-CM3 
NOAA Geophysical Fluid Dynamics Laboratory, USA 2.02° × 2.00° GFDL-ESM2G 
NOAA Geophysical Fluid Dynamics Laboratory, USA 2.02° × 2.50° GFDL-ESM2M 
NASA Goddard Institute for Space Studies, USA 2.02° × 2.50° GISS-E2-R 
Met Office Hadley Centre, UK 1.25° × 1.25° HadGEM2-CC 
Met Office Hadley Centre, UK 1.25° × 1.25° HadGEM2-ES 
Institute for Numerical Mathematics, Russia 1.5° × 2.0° inmcm4 
Institute Pierre-Simon Laplace, France 1.89° × 3.75° IPSL-CM5A-LR 
Institute Pierre-Simon Laplace, France 1.26° × 2.00° IPSL-CM5A-MR 
Institute Pierre-Simon Laplace, France 1.89° × 3.75° IPSL-CM5B-LR 
Atmosphere and Ocean Research Institute (The University of Tokyo) 1.4° × 1.4° MIROC5 
Atmosphere and Ocean Research Institute (The University of Tokyo) 2.79° × 2.81° MIROC-ESM 
Atmosphere and Ocean Research Institute (The University of Tokyo) 2.79° × 2.81° MIROC-ESM-CHEM 
Max Planck Institute for Meteorology, Germany 1.86° × 1.87° MPI-ESM-LR 
Max Planck Institute for Meteorology, Germany 1.86° × 1.87° MPI-ESM-MR 
Meteorological Research Institute, Japan 1.12° × 1.12° MRI-CGCM3 
Norwegian Climate Center, Norway 1.89° × 2.50° NorESM1-M 
Norwegian Climate Center, Norway 1.89° × 2.50° NorESM1-ME 

The selected models among all considered GCMs are ACCESS1-0, ACESS1-3, CESM1-BGC, CESM1-CAM5, CNRM-CM5, FGOLS-S2, IPSL-CM5A-LR, IPSL-CM5A-MR, MIROC-ESM, MIROC-ESM-CHEM, MRI-CGCM3, and NorESM1-M, determined in Table 1 in bold. These models were selected based on their performances in simulating the monthly runoff in the Gharesu river basin throughout the base period (1981–2005) where the achieved results were compared to the historical data based on the values of the Nash–Sutcliffe efficiency (NSE) coefficient. The threshold value of NSE coefficient for GCM selection has been chosen as 0.5, based on the characteristics of problem phenomenon, basic definition of NSE, and the comprehensive suggestions provided by McMahon et al. (2015).

The rainfall-runoff model

The IHACRES model introduced by Jakeman & Hornberger (1993) is a lumped, semi-conceptual rainfall-runoff model for simulating runoff throughout a basin based on the corresponding rainfall. It has been used for various types of small to large basins with acceptable results. In this study, the IHACRES2.1 version is used. It has applications for basins with continuous rainfall, temperature, and runoff data. According to Figure 3, this model consists of two nonlinear and linear modules. Rainfall and temperature are first converted into effective rainfall at each time step using the non-linear module and then into runoff in the same time step by employing the linear unit hydrograph module (Jakeman et al. 1990; Jakeman & Hornberger 1993).

Figure 3

The conceptual flowchart of IHACRES model.

Figure 3

The conceptual flowchart of IHACRES model.

Applying the 25-year time series of temperature, rainfall, and runoff observations within the base period (1981–2005), the IHACRES rainfall-runoff model was set up and calibrated. The developed model was calibrated by the observed data in the first 15 years and was validated within the last 10 years of the base period.

The change factor (CF) method for error correction

The CF method is a commonly used error correction method that is often applied to reduce the error between the outputs of GCMs and the observed data (Chen et al. 2011; Ouyang et al. 2015). Before using the data on climate variables for future periods, the errors have to be corrected. In this study, the long-term monthly averages of temperature and rainfall are applied based on the CF method. Equations (1) and (2) are implemented to calculate the amount of long-term average of temperature change and the long-term average of rainfall change ratio for each month in future (2006–2030) time horizon. These calculations are performed for each scenario based on each GCM model data.
formula
(1)
formula
(2)
where, and represent the scenario-based monthly average precipitation change ratio and monthly average temperature change related to the model in the , respectively. and is the simulated long-term average precipitation and temperature of the model in the for future period, and and stand for simulated long-term average precipitation and temperature of the model in the month of the base period (Anandhi et al. 2011). The values of temperature and rainfall variables in different future periods are achieved by implementing the corresponding change values into the observed data in the base period based on Equations (3) and (4):
formula
(3)
formula
(4)
where, and are the predicted temperature and precipitation for the model in the time step (sequential month) in the future period, respectively, and and are the temperature and precipitation of the time step within the base (historical) period (Diaz-Nieto & Wilby 2005; Anandhi et al. 2011).

Calculation of the SDI and drought characteristics

Nalbantis & Tsakiris (2009) developed the SDI to monitor hydrological conditions based on a computational approach similar to that for SPI. The SDI is calculated based on the monthly cumulative flow volume based on the following equation (Nalbantis & Tsakiris 2009):
formula
(5)
where, is the cumulative flow volume for the hydrological year and the reference period, the monthly flow for the hydrological year and the month, k = 1, …, 4 the reference period representing October–December (3 months), October–March (6 months), October–June (9 months), and October–September (12 months), respectively, n the year and j the month. The SDI values are calculated based on the following equation:
formula
(6)
where, and represent the mean and standard deviation for the cumulative flow volume, respectively. The drought classes for the SDI are as mentioned in Table 2 (Tigkas et al. 2015).
Table 2

The hydrological drought classes based on SDI (Tigkas et al. 2015)

StateDescriptionCriterion
Non-drought SDI ≥ 0 
Mild drought − 1 ≤ SDI < 0 
Moderate drought − 1.5 ≤ SDI < −1 
Severe drought − 2 ≤ SDI < −1.5 
Extreme drought SDI ≤ −2 
StateDescriptionCriterion
Non-drought SDI ≥ 0 
Mild drought − 1 ≤ SDI < 0 
Moderate drought − 1.5 ≤ SDI < −1 
Severe drought − 2 ≤ SDI < −1.5 
Extreme drought SDI ≤ −2 

In order to assess the drought, its characteristics should be evaluated (Wong et al. 2009). Various studies have attempted to define and apply the drought characteristics based on their major purposes (e.g., Wong et al. 2009; Gupta & Jain 2018; Kang & Jiang 2018; Sohoulande Djebou 2019; Surendran et al. 2019; Vazifehkhah et al. 2019). In this study, the characteristics of predicted drought were estimated for the time series of runoff of 12 selected models based on the run theory proposed by Yevjevich (1967). According to the run theory, a specific function or threshold level is used to divide the flow time series obtained from different models into wet and dry events and then the drought characteristics are calculated for them. Each wet or dry event has a start time and an end time, and the periods are defined according to a specific time step that can be day, week, month, or year (Yevjevich 1967). The present study used the monthly period. The applied duration (D) component can be assumed as a prolonged period (day, month, etc. based on the time step of the study) when the implemented index drops below the threshold level (Hong et al. 2015). In this study, the maximum drought duration (MD) is used as the drought characteristic for the different time series obtained from the selected climate models. Another characteristic estimated in this study for drought is the drought severity (S), which is by definition the cumulative absolute value of the dimensionless SDI during a continuous drought event. The peak of drought (P) is another studied characteristic that is considered to be the maximum absolute value of the SDI below the threshold level for the occurrence of drought, which is estimated in a climate scenario with a specific timescale by each of the climate models (Saghafian & Mehdikhani 2014). Finally, the number of drought events (EN) that occurred in the time series was obtained from each specific climate model. Moreover, in this study, relative number of drought steps (RNDS), relative number of severe droughts (RNSD), and relative number of extreme droughts (RNED) are calculated as the secondary characteristics.

RESULTS

Figure 4 shows the diagram of the calculated mean, minimum, and maximum values of the SDI for the future period (2006–2030) obtained from the 12 different GCMs under the RCP4.5 scenario. This diagram also presents the range of changes of the achieved SDIs. The ranges indicate the modeling uncertainty associated with calculating the SDI. The widest range of changes is observed for the 12-month SDI followed by the 9-month SDI and the narrowest for the 6-month and 3-month SDIs. This declares the high uncertainty of GCM projections of SDI index values over longer timescales. To better understand the uncertainties in different timescales, the proportion of total time steps with range value greater than 0.5 is presented here as the RG0.5. The value of RG0.5 was achieved equal to 0.29, 0.42, 0.52, and 0.48 for 3-, 6-, 9- and 12-month timescales, respectively. It can be concluded that increasing within the timescale results in enhancing the model uncertainties. Surendran et al. (2019) achieved the same conclusion for the projection of various hydrological indices in different timescales for arid and semi-arid regions. They found more uncertain results in longer timescales. Touma et al. (2015) stated that applying the total runoff at longer timescales results in achieving drought characteristics with more uncertainty. The largest range values of 3-month, 6-month, 9-month, and 12-month SDI are 1.18, 1.09, 1.23, and 1.24 for the 82nd 3-month, the 14th 6-month, the 9th 9-month, and the 9th year, respectively. Since the range values are proportional to the results' relative error of GCM models, the maximum range value for various timescales occurs at different time steps (Mujumdar & Ghosh 2008).

Figure 4

Projected 3-, 6-, 9-, and 12-month SDI values using the selected 12 GCMs in the future period (2006–2030) under RCP4.5 emission scenarios.

Figure 4

Projected 3-, 6-, 9-, and 12-month SDI values using the selected 12 GCMs in the future period (2006–2030) under RCP4.5 emission scenarios.

Based on the basic definition of the SDI, the positive values refer to the wet periods while the negative values indicate the drought time steps. Hence, to consider the critical conditions, the extreme values of the index should be extracted. Throughout the considered time horizon, the achieved values of SDI vary between −2.94 and 1.97 for 3-month, between −2.68 and 2.44 for 6-month, between −2.47 and 2.25 for 9- and 12-month timescales.

All drought indices (e.g., SPI, SDI, etc.) illustrate various hydrological components' deviation from the mean values over specified timescales (Surendran et al. 2019). Therefore, it is expected to achieve different values of SDI for different timescales as the duration and mean values of hydrological components vary in time. Van Lanen et al. (2013) and Keshavarz et al. (2014) revealed that the variations in drought indices (e.g., SPI) over different timescales can be related to the fluctuations of hydrological cycle components compared to the mean values in different durations. The SDI index measures how much streamflow for a given period has deviated from the normal mean or long-term average value. For this index, the number of effective hydrological components is more compared to SPI while the cumulative precipitation, temperature, evapotranspiration, infiltration, etc. can affect the runoff values over a specific timescale (Abdulai & Chung 2019).

Figure 5 shows a diagram of the calculated mean, minimum, and maximum values of SDI in the future period (2006–2030) obtained from the 12 different GCMs under the RCP8.5 scenario. This diagram also presents the range of changes in SDI. The maximum value of SDI was achieved as 2.05, 2.24, 2.11, and 2.11 for 3-month, 6-month, 9-month, and 12-month timescales.

Figure 5

Predicted 3-, 6-, 9-, and 12-month SDI values of runoff using the 12 selected GCMs in the future period (2006–20306) under RCP8.5 emission scenarios.

Figure 5

Predicted 3-, 6-, 9-, and 12-month SDI values of runoff using the 12 selected GCMs in the future period (2006–20306) under RCP8.5 emission scenarios.

Throughout the considered time horizon under RCP8.5 the variation of achieved SDI is between −3.19 and 2.05 for 3-month, between −2.70 and 2.24 for 6-month, between −2.28 and 2.11 for 9-month, and between −2.48 and 2.11 for 12-month timescales. Again, the widest range of changes is observed for the 12-month SDI and the highest uncertainty of GCM projections is achieved over the longest timescales. The value of RG0.5 is achieved equal to 0.48, 0.66, 0.68, and 0.72 for 3-, 6-, 9-, and 12-month timescales, respectively. As discussed before, the modeling uncertainty for SDI projection enhances by increasing within the timescale. Comparing the achieved results for RCP4.5 and RCP8.5 scenarios, it is clearly observed that the calculated values of the SDI and consequently the variability range values under the RCP8.5 scenario are higher than those under the RCP4.5 scenario.

Figure 6 shows the range of changes in drought characteristics, including S, P, MD, and EN, estimated by the different GCM models for the 3-, 6-, 9-, and 12-month timescales under the RCP4.5 and RCP8.5 scenarios. This figure presents the uncertainty in estimating drought characteristics.

Figure 6

Characteristics of the predicted SDI for the various 3-, 6-, 9-, and 12-month timescales.

Figure 6

Characteristics of the predicted SDI for the various 3-, 6-, 9-, and 12-month timescales.

As shown in Figure 6, the lowest levels of variation are observed in the estimates for the peak of drought and the number of drought events. The level of uncertainty is much higher for the estimated values of drought severity and maximum drought duration. Different drought characteristics provide different information about drought (Abdulai & Chung 2019), thus, considering the drought characteristics is important to water resources managers. For instance, the number of drought events and frequency provides no information on the event intensity or duration. On the other hand, the number of drought events depends on the chosen timescale (Vicente-Serrano & López-Moreno 2005). Moreover, despite the duration which strongly correlates to severity (Bonacci 1993; Woo & Tarhule 1994; Tarhule & Woo 1997), the peak value is not correlated to the severity or duration significantly (Andreadis et al. 2005; Yu 2013). Therefore, the level of uncertainty in projection of severity and maximum duration should be consistent while the uncertainty of peak estimation can be at a lower level. The results achieved by Yao et al. (2020) confirm our findings. They tried to estimate the drought characteristics in China, applying multiple GCMs. They found less uncertainty in peak projection compared to the severity and duration estimation.

Internal variability of climate system, model uncertainty, and scenario uncertainty can be considered as three dominant sources of uncertainty in drought projections (Hawkins & Sutton 2009, 2011; Lu et al. 2019). As proposed by Lu et al. (2019), those three uncertainties can be assumed independent. As the applied GCMs are identical in different scenarios, the scenario uncertainty in projecting drought characteristics can be estimated as the range of results changes achieved by different models in each scenario. It can be concluded from Figure 6(a) that the level of uncertainty in estimating drought severity under RCP4.5 is much lower than under RCP8.5. The mean values of drought severity increased from the 3-month to the 12-month timescale and its highest values were observed for the 12-month timescale. Regarding the definition of the applied drought severity (S), it is quite obvious that the value of severity as a cumulative index increases by increasing the timescale. Hawkins & Sutton (2009) revealed that the model uncertainty and scenario uncertainty varies by lead time while the model spread is much larger than the scenario spread.

The range of peak estimation has the highest value for the 3-month timescale while its values for the other timescales are almost identical. The uncertainty in estimating the drought peak and maximum drought duration under the RCP8.5 scenario is also higher than that under the RCP4.5 scenario. In other words, the projection uncertainty increases as the radiative forcing increases, which is indicated by some other researchers (Mujumdar & Ghosh 2008; Dubrovsky et al. 2009; Hawkins & Sutton 2009, 2011; Mishra & Singh 2009; Touma et al. 2015; Lu et al. 2019).

If the mean values of the results are used, these differences cannot be detected (Saghafian & Mehdikhani 2014). The largest values for the number of drought events are obtained for the 3-month timescale. The longer the timescale is the more the value of this index decreases so that its lowest value is observed for the 12-month timescale. Moreover, the levels of uncertainty in estimating the values of this index for the different timescales did not differ very much from each other.

Table 3 shows the values of the drought characteristics and the numbers of severe and extreme drought events exceeding the threshold level for each of the 3-, 6-, 9-, and 12-month timescales under the RCP4.5 and RCP8.5 scenarios of future emissions based on the mean values of the SDI obtained from the 12 GCMs. The values reported in this table were calculated based on the run theory, and using the time series of the mean estimated values. In this table, the RNDI represents the relative number of drought events, which was calculated based on the threshold level for determining drought and on the values of the SDI and was equal to the ratio of the time periods with drought events to the total number of the studied time periods. Using the same definition, we can define RNSDI and RNEDI as the ratios of the time periods with severe drought events and with extreme drought events to the total number of the studied time periods.

Table 3

Mean values of drought characteristics in each of the 3-, 6-, 9-, and 12-month timescales under the RCP4.5 and RCP8.5 scenarios for future emissions based on the results obtained from the 12 GCMs

TimescaleScenarioRNDIRNSDIRNEDIPeakSeverityDuration
3-month RCP4.5 46% 9% 5% 2.77 6.92 8.00 
RCP8.5 45% 9% 5% 2.87 6.78 7.83 
6-month RCP4.5 46% 10% 5% 2.49 6.81 7.25 
RCP8.5 49% 10% 5% 2.52 6.29 6.92 
9-month RCP4.5 49% 8% 7% 2.26 6.94 6.92 
RCP8.5 51% 8% 7% 2.28 6.61 6.58 
12-month RCP4.5 50% 8% 7% 2.29 7.26 7.17 
RCP8.5 50% 8% 7% 2.29 6.68 6.58 
TimescaleScenarioRNDIRNSDIRNEDIPeakSeverityDuration
3-month RCP4.5 46% 9% 5% 2.77 6.92 8.00 
RCP8.5 45% 9% 5% 2.87 6.78 7.83 
6-month RCP4.5 46% 10% 5% 2.49 6.81 7.25 
RCP8.5 49% 10% 5% 2.52 6.29 6.92 
9-month RCP4.5 49% 8% 7% 2.26 6.94 6.92 
RCP8.5 51% 8% 7% 2.28 6.61 6.58 
12-month RCP4.5 50% 8% 7% 2.29 7.26 7.17 
RCP8.5 50% 8% 7% 2.29 6.68 6.58 

We can infer from the results presented in Table 3 that the ratio of the number of time steps with drought to the total number of time steps (RNDI) increased with the timescale increase from 3 months to 12 months. However, this difference was not very noticeable. In general, drought duration decreased and the relative drought severity increased in the longer timescales. Based on the average values obtained from the different models, it can be said that the peak of droughts decreased at longer timescales. The changes in the drought characteristics for the mean data were not consistent with the results obtained from many of the individual studied models. In general, it is clear that the values obtained for the drought characteristics based on the mean values were significantly different from those based on the values obtained from the individual models. Therefore, due to the high level of uncertainty in the obtained results, the mean values of the data cannot be referred to and used in water resources planning and it is imperative that the values of all results obtained from the various models be also considered.

Figure 7 shows the diagrams of percentage frequency distributions for the various values of the SDI in the 3-, 6-, 9-, and 12-month timescales under the RCP4.5 and RCP8.5 scenarios. It is quite clear in these diagrams that the distributions of the results obtained from the different models for estimates of the percentages of wet periods (SDI > 0.0) were much wider than those for the dry periods. This analysis indicated that the uncertainty in estimating the wet periods is greater than that of estimating the dry periods. We can also see that the distributions of the results and their uncertainties for the 9- and 12-month timescales were much wider and greater, respectively, than those for the 3- and 6-month time timescales. In general, the lowest level of uncertainty was found in estimating the frequency percentage of SDI values larger than 2.5 and smaller than −3. Moreover, the uncertainty in estimating the SDI values of the dry periods (SDI < 0) under the RCP4.5 scenario was lower than that under the RCP8.5 scenario. This may be due to seasonal differences in flow rates: under scenarios in which the seasonal variation in watershed runoff is not great the differences between the results of various climate models increase resulting in reduced levels of certainty in the results obtained from the models.

Figure 7

Diagram showing frequency percentages of the various SDI values for the 3-, 6-, 9-, and 12-month timescales for the future period (2006–2030) under the RCP4.5 and RCP8.5 scenarios.

Figure 7

Diagram showing frequency percentages of the various SDI values for the 3-, 6-, 9-, and 12-month timescales for the future period (2006–2030) under the RCP4.5 and RCP8.5 scenarios.

CONCLUSION

Southwestern Iran has a broad range of agricultural, industrial, and municipal water demands, which necessitates the identification of future drought characteristics to plan for effective adaptation and mitigation strategies. This study evaluates the uncertainties in the future drought projections of Gharesu as one of the main sub-basins of Karkheh river basin. Based on the results of hydrological simulations over the historical periods, 12 GCMs were selected to assess projected drought uncertainties under the RCP4.5 and RCP8.5 emission scenarios. We evaluated the uncertainties in the estimated SDI index and its corresponding drought characteristics. The comparison results showed that the uncertainties in estimating SDI indicator under RCP8.5 are higher than those under RCP4.5. Moreover, the modeling uncertainty increases by enhancement within the timescale. In order to have a good understanding of future drought conditions, the variabilities of drought indicators and characteristics achieved from various GCMs should be considered instead of individual results or mean values. In this study, the maximum variabilities correspond to the drought severity and maximum duration. Both of them are critical parameters in water resources management for drought conditions. Furthermore, it was concluded that the uncertainty in estimating the drought characteristics depends on the length of timescales and the considered emission scenario. Moreover, it can be indicated that the uncertainty in drought projection for wet periods is greater than that for dry periods. This achievement is a key point in developing drought monitoring systems and should be of interest to water resources managers in arid and semi-arid regions.

ACKNOWLEDGEMENTS

The authors would like to thank the Shahid Chamran University of Ahvaz for financial support under the grant number SCU.EC98.31254.

DATA AVAILABILITY STATEMENT

All relevant data are included in the paper or its Supplementary Information.

REFERENCES

REFERENCES
Adib
A.
Mirsalari
S. B.
Ashrafi
S. M.
2018
Prediction of meteorological and hydrological phenomena by different climatic scenarios in the Karkheh watershed (south west of Iran)
.
Scientia Iranica
.
doi:10.24200/sci.2018.50953.1934
.
Anandhi
A.
Frei
A.
Pierson
D. C.
Schneiderman
E. M.
Zion
M. S.
Lounsbury
D.
Matonse
A. H.
2011
Examination of change factor methodologies for climate change impact assessment
.
Water Resources Research
47
(
3
),
W03501
.
Andreadis
K.
Clark
E.
Wood
A.
Hamlet
A.
Lettenmaier
D.
2005
Twentieth-century drought in the conterminous United States
.
Journal of Hydrometeorology
6
(
6
),
985
1001
.
Ashrafi
S. M.
Dariane
A. B.
2017
Coupled operating rules for optimal operation of multi-reservoir systems
.
Water Resources Management
31
(
14
),
4505
4520
.
Ashrafi
S. M.
Mahmoudi
M.
2019
Developing a semi-distributed decision support system for great Karun water resources system
.
Journal of Applied Research in Water and Wastewater
6
(
1
),
16
24
.
Ashrafi
S. M.
Ashrafi
S. F.
Moazami
S.
2017
Developing self-adaptive melody search algorithm for optimal operation of multi-reservoir systems
.
Journal of Hydraulic Structures
3
(
1
),
35
48
.
Bakhsipoor
I. E.
Ashrafi
S. M.
Adib
A.
2019
Water quality effects on the optimal water resources operation in Great Karun River Basin
.
Pertanika Journal of Science & Technology
27
(
4
),
1881
1900
.
Blöschl
G.
Montanari
A.
2010
Climate change impacts – throwing the dice?
Hydrological Processes
24
(
3
),
374
381
.
Bonacci
O.
1993
Hydrological identification of drought
.
Hydrological Processes
7
(
3
),
249
262
.
Chen
J.
Brissette
F. P.
Leconte
R.
2011
Uncertainty of downscaling method in quantifying the impact of climate change on hydrology
.
Journal of Hydrology
401
(
3–4
),
190
202
.
Dubrovsky
M.
Svoboda
M. D.
Trnka
M.
Hayes
M. J.
Wilhite
D. A.
Zalud
Z.
Hlavinka
P.
2009
Application of relative drought indices in assessing climate-change impacts on drought conditions in Czechia
.
Theoretical and Applied Climatology
96
(
1–2
),
155
171
.
Ehret
U.
Zehe
E.
Wulfmeyer
V.
Warrach-Sagi
K.
Liebert
J.
2012
HESS opinions. ‘Should we apply bias correction to global and regional climate model data?’
.
Hydrology and Earth System Sciences
16
(
9
),
3391
3404
.
Forzieri
G.
Feyen
L.
Rojas
R.
Flörke
M.
Wimmer
F.
Bianchi
A.
2013
Ensemble projections of future streamflow droughts in Europe
.
Hydrology and Earth System Sciences Discussions
10
(
8
),
10719
10774
.
Fu
G.
Charles
S. P.
Yu
J.
Liu
C.
2009
Decadal climatic variability, trends, and future scenarios for the North China Plain
.
Journal of Climate
22
(
8
),
2111
2123
.
Hawkins
E.
Sutton
R.
2009
The potential to narrow uncertainty in regional climate predictions
.
Bulletin of the American Meteorological Society
90
(
8
),
1095
1108
.
Hong
X.
Guo
S.
Zhou
Y.
Xiong
L.
2015
Uncertainties in assessing hydrological drought using streamflow drought index for the upper Yangtze River basin
.
Stochastic Environmental Research and Risk Assessment
29
(
4
),
1235
1247
.
Jakeman
A.
Hornberger
G. J.
1993
How much complexity is warranted in a rainfall-runoff model?
Water Resources Research
29
(
8
),
2637
2649
.
Jehanzaib
M.
Sattar
M. N.
Lee
J. H.
Kim
T. W.
2020
Investigating effect of climate change on drought propagation from meteorological to hydrological drought using multi-model ensemble projections
.
Stochastic Environmental Research and Risk Assessment
34
(
1
),
7
21
.
Keshavarz
M. R.
Vazifedoust
M.
Alizadeh
A.
2014
Drought monitoring using a Soil Wetness Deficit Index (SWDI) derived from MODIS satellite data
.
Agricultural Water Management
132
,
37
45
.
Li
G.
Zhang
X.
Cannon
A. J.
Murdock
T.
Sobie
S.
Zwiers
F.
Anderson
K.
Qian
B. J.
2018
Indices of Canada's future climate for general and agricultural adaptation applications
.
Climatic Change
148
(
1–2
),
249
263
.
McCuen
R. H.
Knight
Z.
Cutter
A. G.
2006
Evaluation of the Nash–Sutcliffe efficiency index
.
Journal of Hydrologic Engineering
11
(
6
),
597
602
.
McMahon
T. A.
Peel
M. C.
Karoly
D. J.
2015
Assessment of precipitation and temperature data from CMIP3 global climate models for hydrologic simulation
.
Hydrology and Earth System Sciences
19
,
361
377
.
Mishra
A. K.
Singh
V. P.
2010
A review of drought concepts
.
Journal of Hydrology
391
(
1–2
),
202
216
.
Moazami
S.
Abdollahipour
A.
Zakeri Niri
M.
Ashrafi
S. M.
2016
Hydrological assessment of daily satellite precipitation products over a basin in Iran
.
Journal of Hydraulic Structures
2
(
2
),
35
45
.
Mondal
A.
Mujumdar
P. P.
2015
Return levels of hydrologic droughts under climate change
.
Advances in Water Resources
75
,
67
79
.
Myronidis
D.
Ioannou
K.
Fotakis
D.
Dörflinger
G.
2018
Streamflow and hydrological drought trend analysis and forecasting in Cyprus
.
Water Resources Management
32
(
5
),
1759
1776
.
Najafi
M. R.
Moradkhani
H.
Jung
I. W.
2011
Assessing the uncertainties of hydrologic model selection in climate change impact studies
.
Hydrological Processes
25
(
18
),
2814
2826
.
Najafi
M. R.
Zwiers
F. W.
Gillett
N. P. J.
2017
Attribution of observed streamflow changes in key British Columbia drainage basins
.
Geophysical Research Letters
44
(
21
),
11012
11020
.
Nalbantis
I.
2008
Evaluation of a hydrological drought index
.
European Water
23
(
24
),
67
77
.
Nalbantis
I.
Tsakiris
G.
2009
Assessment of hydrological drought revisited
.
Water Resources Management
23
(
5
),
881
897
.
Orlowsky
B.
Seneviratne
S. I.
2013
Elusive drought: uncertainty in observed trends and short-andlong-term CMIP5 projections
.
Hydrology and Earth System Sciences
17
(
5
),
1765
1781
.
Orville
H. D.
1990
AMS statement on meteorological drought
.
Bulletin of the American Meteorological Society
71
(
7
),
1021
1025
.
Ouyang
F.
Zhu
Y.
Fu
G.
H.
Zhang
A.
Yu
Z.
Chen
X. J.
2015
Impacts of climate change under CMIP5 RCP scenarios on streamflow in the Huangnizhuang catchment
.
Stochastic Environmental Research and Risk Assessment
29
(
7
),
1781
1795
.
Prudhomme
C.
Giuntoli
I.
Robinson
E. L.
Clark
D. B.
Arnell
N. W.
Dankers
R.
Fekete
B. M.
Franssen
W.
Gerten
D.
Gosling
S. N.
2014
Hydrological droughts in the 21st century, hotspots and uncertainties from a global multimodel ensemble experiment
.
Proceedings of the National Academy of Sciences
111
(
9
),
3262
3267
.
Saghafian
B.
Mehdikhani
H.
2014
Drought characterization using a new copula-based trivariate approach
.
Natural Hazards
72
(
3
),
1391
1407
.
Samadi
S.
Wilson
C. A.
Moradkhani
H.
2013a
Uncertainty analysis of statistical downscaling models using Hadley Centre Coupled Model
.
Theoretical and Applied Climatology
114
(
3–4
),
673
690
.
Samadi
S.
Carbone
G. J.
Mahdavi
M.
Sharifi
F.
Bihamta
M. J.
2013b
Statistical downscaling of river runoff in a semi-arid catchment
.
Water Resources Management
27
(
1
),
117
136
.
Sheffield
J.
Wood
E. F.
Roderick
M. L.
2012
Little change in global drought over the past 60 years
.
Nature
491
(
7424
),
435
.
Shukla
S.
Wood
A. W.
2008
Use of a standardized runoff index for characterizing hydrologic drought
.
Geophysical Research Letters
35
(
2
),
L02405
.
Surendran
U.
Anagha
B.
Raja
P.
Kumar
V.
Rajan
K.
Jayakumar
M.
2019
Analysis of drought from humid, semi-arid and arid regions of India using DrinC model with different drought indices
.
Water Resources Management
33
(
4
),
1521
1540
.
Tabari
H.
Nikbakht
J.
Talaee
P. H.
2013
Hydrological drought assessment in Northwestern Iran based on streamflow drought index (SDI)
.
Water Resources Management
27
(
1
),
137
151
.
Tigkas
D.
Vangelis
H.
Tsakiris
G.
2015
DrinC: a software for drought analysis based on drought indices
.
Earth Science Informatics
8
(
3
),
697
709
.
Touma
D.
Ashfaq
M.
Nayak
M. A.
Kao
S. C.
Diffenbaugh
N. S.
2015
A multi-model and multi-index evaluation of drought characteristics in the 21st century
.
Journal of Hydrology
526
,
196
207
.
Trenberth
K. E.
Dai
A.
Van Der Schrier
G.
Jones
P. D.
Barichivich
J.
Briffa
K. R.
Sheffield
J.
2014
Global warming and changes in drought
.
Nature Climate Change
4
(
1
),
17
.
Trinh
T.
Kavvas
M. L.
Ishida
K.
Carr
K.
Ohara
N.
2017b
Projected 21st century climate change on snow conditions over Shasta Dam watershed by means of dynamical downscaling
.
Hydrological Processes
31
(
16
),
2887
2901
.
Van Lanen
H. A.
Wanders
N.
Tallaksen
L. M.
Van Loon
A. F.
2013
Hydrological drought across the world: impact of climate and physical catchment structure
.
Hydrology and Earth System Sciences
17
,
1715
1732
.
Vazifehkhah
S.
Tosunoglu
F.
Kahya
E.
2019
Bivariate risk analysis of the droughts using a nonparametric multivariate standardized drought index and copulas
.
Journal of Hydrologic Engineering
24
(
5
),
05019006
.
Wilhite
D. A.
Glantz
M. H.
1985
Understanding: the drought phenomenon: the role of definitions
.
Water International
10
(
3
),
111
120
.
Wong
G.
Lambert
M. F.
Leonard
M.
Metcalfe
A. V.
2009
Drought analysis using tri-variate copulas conditional on climatic states
.
Journal of Hydrologic Engineering
15
(
2
),
129
141
.
Woo
M.-K.
Tarhule
A.
1994
Streamflow droughts of northern Nigerian rivers
.
Hydrological Sciences Journal
39
(
1
),
19
34
.
Yao
N.
Li
L.
Feng
P.
Feng
H.
Li Liu
D.
Liu
Y.
Jiang
K.
Hu
X.
Li
Y.
2020
Projections of drought characteristics in China based on a standardized precipitation and evapotranspiration index and multiple GCMs
.
Science of the Total Environment
704
,
135245
.
Yeh
C. F.
Wang
J.
Yeh
H. F.
Lee
C. H.
2015
SDI and Markov chains for regional drought characteristics
.
Sustainability
7
(
8
),
10789
10808
.
Yevjevich
V. M.
1967
Objective Approach to Definitions and Investigations of Continental Hydrologic Droughts
.
Hydrology Papers 23
,
Colorado State University
,
Fort Collins, CO
,
USA
.
Yu
R. M. S.
2013
European Droughts Under Climate Change: Projections and Uncertainties
.
Doctoral dissertation
,
University of East Anglia
,
Norwich
,
UK
.
Zallaghi
E.
Akhoond-Ali
A. M.
Ashrafi
S. M.
2020
Dataset on sensitivity of water-energy nexus to Dez Dam power plant operation
.
Data in Brief
30
,
105454
.
Zhang
D.
Zhang
Q.
Qiu
J.
Bai
P.
Liang
K.
Li
X.
2018
Intensification of hydrological drought due to human activity in the middle reaches of the Yangtze River, China
.
Science of the Total Environment
637
,
1432
1442
.