Introduction

Droughts continue to be a significant natural hazard around the world and specifically in arid or semiarid areas like Iran. Recently, 10 out of the 28 provinces (35% of Iran) were affected by several droughts [10]. The most costly drought disaster was in 1999, with an estimated loss of $1605 million [21].

Increasing evidence of global warming is followed by a pressing question: will climate change exacerbate the risk of drought at a regional or local scale?

Recently, according to the Fifth Assessment Report produced by the Intergovernmental Panel on Climate Change (IPCC), droughts have become stronger and longer and have influenced larger areas since the 1970s. The land area (Iran) affected by drought is expected to expand and so availability of water resources by mid-century could decrease as much as 30% [1, 9].

Climate change can explain any alteration in climate over time, either because of natural variability or due to human activity. These changes can cause alterations in the statistical properties of distribution of the variable investigated, as changes in their mean values or variability in a specific range of values. Alteration in precipitation variability may cause more frequent and detrimental extreme events such as drought [18]. Climate change is anticipated to affect the frequency and severity of droughts principally [13]. Most of the research in Iran focuses on changes in the mean of the climate [22] rather than on changes in individual temporal which is important in drought characterisation. Also, no work proves the occurrence of future trends in precipitation and runoff at different time scales over the central area of Iran. Therefore, influences of climate change on drought characterisation remain unknown. Although General Circulation models (GCMs) have been improved for the application of modelling the Earth’s climate, there is still uncertainty in the climate projections which are required to determine drought risk [26]. The GCMs have complex mathematical formulations in order to explain how the climate works and how it would change if main factors of climate disturbances are known.

The difficulty lies in the fact that GCMs which are applied to project global climate change cannot sufficiently determine the main factors of climate disturbances that are likely to affect regional climates [2, 9].

Furthermore, the model outputs for extreme events at large scales are not as reliable as those of climate averages at smaller scales. Several methods have improved downscaling precipitation from GCMs [14, 30].

A few studies, such as [32]), defined the impact of climate change on worldwide drought on the basis of multiple drought indicators including SPI, potential evaporation anomaly, soil moisture anomaly and Palmer Drought Severity Index. However, there is a lack of research, which includes uncertainties of regional climate change into drought risk evaluations at the local level [31].

Even in the research at continental scale, there is uncertainty in the selection of GCMs. For example, the study by [20]) is an example of drought characterisation in Australia which applied climate variables from 14 GCMs randomly selected from the IPCC 4th assessment report.

The research of [16]) is another example of evaluation of potential impacts of climate change on SPI and RDI for the period of 2010–2039 in the west of Iran, which applied data of one GCM-run, selected randomly from the Third Assessment Report (TAR) based on the IPCC SRES scenarios. However, no study has focused on the future changes in drought risk in the centre of Iran based on analysis the CMIP5 models, which is essential for improvement planning for water resources and demand management. The CMIP5 can simulate a standard set of models in order to:

1) Evaluate how realistic the models are in simulating the recent past

2) Make projections of future climate change on two time scales: near term (out to about 2040) and long term (out to about 2100).

3) Understand factors responsible for differences in model projections, containing quantifying some key feedback such as those including clouds and the system of atmospheric carbon cycles [9]

The Zayandeh Rud basin (Fig. 1) is important in central Iran where agriculture is the dominant activity. The important crops are rice, wheat, potatoes and barley. About 90% of the basin’s land use is dedicated to agriculture. Also, one major urban area, the city of Esfahan, is located in the basin. There has been a large increase in water demands from surrounding agricultural, industrial and metropolitan areas, which have experienced water scarcity and strong droughts several times (1972, 1976, 1980, 1984, 1990, 1996, 1998, 1999 and 2000) over the past five decades [9]. However, there is lack of knowledge of the characteristics of drought and predication for future risks in this area.

Fig. 1
figure 1

Grid points across the Zayandeh Rud basin for the HadCM3 model

Full size image

This work is the first study that applies multiple indices (SPI and SRI) to assess historical and future drought in the Zayandeh Rud basin for use in future water planning models. Also, impact studies of climate change on hydrology at regional scales need grids or stations at finer resolutions. A method to approach this is downscaling of the GCM data and applying bias correction.

Also, the impacts of non-climatic factors such as human abstraction on drought have previously been neglected. For example, previous studies in Africa, North and South America and Australia only focused on the direct cause of drought such as climate change [3, 17, , 19, 28, 34].

Therefore, the aim of this chapter is to determine the future climate change impact on drought and the objectives are:

  1. 1.

    To quantify the range of predicted changes in future climate conditions (precipitation and temperature) and compare them with historical observations from the Zayandeh Rud basin

  2. 2.

    To assess the contribution of human withdrawals of water versus climate impacts on the future stream flow (runoff) to quantify anthropogenic influence

  3. 3.

    To determine the impact of predicted climate change on drought severity, duration and frequency using SPI and SRI as basin-scale drought metrics and analyse the relationship between meteorological and hydrological drought indices without adaptation scenarios.

The climate projections are input into a calibrated water allocation model (WEAP) for the Zayandeh Rud river basin [9]. The outputs of this model are applied to determine the impacts of climate change on other water resources, water demands and also crop productions and finally understand the impact of climate change on hydrological drought which include human impacts. Understanding the propagation of climate change impacts by nonlinear water allocation model processes among drought indices could determine the risk behaviours across drought indices.

Materials and Methods

To assess the potential impact of future climate change on meteorological and hydrological droughts and effects of human abstraction on the hydrological droughts, the methodology used in this chapter is divided into four sections. At first, the future climatic model prediction (the ‘Climate change simulations’ section) which is necessary to provide precipitation data for meteorological drought is explained. The climate model generates the initial data (precipitation, temperature, evaporation, relative humidity and wind speed) to run the hydrological model and water allocation model (the ‘Hydrological model’ section and the ‘Water Allocation Model’ section). Because data available was in monthly scale, so the results are based on monthly scale. In the ‘Future drought indices’ section, the method that is applied to calculate the intensity of droughts and analysis of the frequency and duration of meteorological and hydrological droughts is presented.

Climate Change Simulations

Generally, climate change predictions made by models are not aligned with the ‘real’ natural environment because of uncertainties and data errors in the models. CMIP5 results tried to fill this gap recently with a finer resolution for the models and also with new climate change scenarios. In this research, different outputs from climate models were used (in monthly scale) from 38 GCM which participated in the CMIP5 model. These new models are more nuanced, more developed vis-a-vis the CMIP3. In addition to the CMIP5, new models for predicting climate change using different scenarios, such as ‘Representative Concentration Pathways’ (RCP) developed by Stöckle et al. [24] and Van Vuuren et al. [27] exist. This model can be used to predict GHG mitigation potential. [8].

Model scenarios applied in this study include historical simulations and future projections. The historical simulations were forced by observed natural and anthropogenic atmospheric composition changes spanning 1971–2005; they are applied to make a baseline against which to determine climate change in future projection. The future projection is obtained by forcing from the RCPs. Unlike the Special Report on Emission scenarios (SRES) that announced the climate projections for the previous CMIP experiment (CMIP3), the CO2 concentration in RCP2.6 is below B1, in RCP6.0 is a little above A1B and in RCP8.5 surpasses A2. In this study, the RCP8.5 scenario (which is the severest one) is applied for the years 2006–2100. The severest potential GHG path for the twenty-first century is selected to make the strongest planning adaptation to mitigate the potential climate change impacts on droughts, supply availability and water demands.

Multiple ensemble members are available for each CMIP5 scenario for the given model. Assuming that there are enough models in the ensemble to approach reliable estimates of a potential climate change signal; in this study, only one ensemble from each CMIP5 model (total 38 models) and scenario RCP8.5 is applied. The variables applied are: precipitation, temperature, relative humidity and wind speed. However, in the results section, only precipitation and temperature, the most important variables, are represented and analysed. The aim of providing 38 coupled GCMs in the scenario of RCP8.5 is to show the uncertainty in climate impacts growing from future climate modelling.

Moreover, biases in climate variables such as precipitation should be taken care of; otherwise, they will extend into the computations for subsequent years. Possible sources which cause errors and bias are:

  • Partial ignorance about geophysical processes

  • Assumptions for numerical modelling

  • Limited spatial resolution

  • Parameterization

  • Bias on resolved scales

  • Additional bias can occur on smaller scales (sub-grid/station).

In order to solve the resolution problems and possible errors in GCM outputs, they are downscaled statistically to each of the meteorological stations. However, to decrease the model’s error and increase the resolution precision, we use a simple downscaling technique to increase the accuracy of the model. Some downscaling techniques attempt to improve daily timescales. In this study, because the drought characteristic analysis cases and water evaluation and planning models are used on monthly resolution, just monthly average climate data are necessary and so resolving the high-frequency variability (the intent of more complex approaches) is not necessary.

In order to remove bias between the GCM and reality, monthly precipitation and temperature time series from GCM and observations for a specific location for the same reference period is needed, which is denoted by Xp, gcm and Xp,obs, respectively.

Furthermore, output from the GCM for some future period of the same length as the reference period, Xf, gcm, is needed. The question remains about how to best combine these three sources of information into the most robust projections of the unknown future observations, Xf, obs, to use as input for the hydrology model (WEAP). This study considered a general approach namely change factor. This is similar to delta change methods used for weather generators. However, the approach taken here is simpler, as a shifted and scaled version of the observed time series is applied for the future rather than a series taken from a weather generator.

The change factor methodology uses the observed monthly variability and changes the mean and monthly variance as simulated by the GCM. In the simplest case, this is the ‘delta method’, where the monthly variability is assumed to have the same magnitude in the future and reference periods, and the corrected monthly data is:

$$ {X}_{\mathrm{DEL}}\left(\mathrm{t}\right)={X}_{\mathrm{p},\mathrm{obs}}\left(\mathrm{t}\right)+\left({\overline{X}}_{\mathrm{f},\mathrm{gcm}}-{\overline{X}}_{\mathrm{p},\mathrm{gcm}}\right) $$
(1)

where the time mean is denoted by the bar above a symbol and the result of the bracket (\( \overline{X} \)f, gcm - \( \overline{X} \)p, gcm) known as climate signal which shows in Fig. 2.

Fig. 2
figure 2

Monthly climate change signal for the HadCM3 model

Full size image

However, in a more general case, considering changes in variance, is:

$$ {X}_{\left(f, obs\ m,y\right)}=\left[{\overline{X}}_{f,{gcm}_m}\right]+\left[{\overline{X}}_{p,{obs}_m}-{\overline{X}}_{p,{gcm}_m}\right]\times \left[\frac{{\overline{\sigma}}_{f,{gcm}_m}}{{\overline{\sigma}}_{p,{gcm}_m}}\right] $$
(2)

X(f, obs, m, y) represents the unknown future observations value of variable X for a given month, m, and period of years, y. The variables contain temperature, rainfall, relative humidity and wind speed;\( {\overline{X}}_{f,{gcm}_m} \) indicates the mean future simulation for a specific month and period of years (such as 2006 to 2040). \( {\overline{X}}_{p,{obs}_m} \) is the mean present-day observed climate for a specific month averaged across all years of the historical period (1971–2005), as measured from the meteorological stations in the study area; \( {\overline{X}}_{p,{gcm}_m} \) indicates the mean simulation from GCM for a specific location for the reference period (e.g. 1971–2005); \( {\overline{\sigma}}_{f,{gcm}_m}\mathrm{and}\ {\overline{\sigma}}_{p,{gcm}_m} \) represent the standard deviations of the raw model output for the future and present-day period for a specific month. The ratio of \( \frac{{\overline{\sigma}}_{f,{gcm}_m}}{{\overline{\sigma}}_{p,{gcm}_m}} \) is shown in Fig. 3.

Fig. 3
figure 3

Monthly ratios between future and present mean standard deviation of precipitation for the HadCM3 model

Full size image

Figure 4 shows precipitation anomaly relative to the monthly mean for all models for 1971–2005 (bracketed term 2 in Eq. 2). The figure explains to what extent the biases in the mean are seasonally dependent. The figure indicates variability in precipitation anomaly between the models for each month. Some models show negative and some of them represent positive precipitation anomaly. The models that show small anomaly for winter and spring (the seasons in which drought mostly occur) can better match with observation data. Generally, it seems for all models during the summer period (June to September) the anomaly is less compared to other seasons. In this figure model, HadCM3 has the least difference from the observation data.

Fig. 4
figure 4

Precipitation anomaly (bias) for the 38 models under the RCP8.5 scenario

Full size image

Hydrological Model

Calculation of streamflow is necessary to measure SRI. Therefore, it is necessary to run a hydrological model (WEAP model) to examine the streamflow. The module of hydrology in WEAP is spatially extended, with a study area constructed as a set of sub-basins which create the whole of the Zayandeh Rud basin. The spatial continuity of the WEAP application across a catchment provides all terrestrial factors for simulation of the hydrologic cycle. The calibrated WEAP model was selected as the WEAP platform with its integrated hydrology module which can run the model under the ensemble of future climate scenarios easily. Moreover, the WEAP model can simulate different assumption scenarios which are essential to determine the impact of climate change on meteorological, hydrological and socio-economic drought.

The climate forcing monthly data set of precipitation, temperature, relative humidity and wind speed are used for sub-catchments which are fractionally separated into land use/land cover classes. The baseline year range for the climate model is 1971–2005 and the future period is 2006–2100.

A water balance model for each land use/land cover class divides water into surface runoff, infiltration, evapotranspiration, interflow and percolation and base flow factors. Each fractional area’s values within a sub-basin altogether show the response of the lumped hydrology. In this study, the Rainfall Runoff model (which runs at the monthly timescale) estimates evapotranspiration for irrigated and rain-fed crops applying crop coefficients. The remainder of rainfall which is not involved in evapotranspiration is simulated as runoff to a river, or may be proportioned through runoff to a river and flow to groundwater by catchment links.

This provide more details of the WEAP hydrologic module. Also, makes a more general justification for this type of simulation approach.

Water Allocation Model

Water demands for the different users in the Zayandeh Rud basin have been projected to the year 2100. Projected water demands were applied to simulate the future hydrology. The assumptions for future water demands were verified from reports of the Iranian Ministry of Energy, Ministry of Agriculture (Jahad Keshavarzi) and Esfahan Regional water authority for the study area. In addition, the assumption for the future population growth rate was collated from reports of Iran’s population census data 2005. The highest priority use for water is domestic and the second priority belongs to industry, while agricultural water use has the third priority.

Future Drought Indices

Both the meteorological and hydrological drought indices in the future periods use the cumulative probability of the precipitation, and runoff is converted to the z value of a normal distribution with zero mean and unit variance [30]. Therefore, the standardised precipitation index and standardised runoff index are applied to indicate the meteorological and hydrological droughts.

The Mann-Kendall test was used to estimate the trends of drought and the impact values of climate change projections on future drought characteristics such as intensity, duration and frequency.

Results

Firstly, the projection of climate variables for all models (38 models) is represented and compared with the historical simulation. Secondly, for the selection of a model (among 38 models) to use its climate data as an input to the hydrological model (to generate runoff values), the results of an empirical statistical downscaling technique are shown. The selection of a model is based on the CDF. With the CDF method, a model with the smallest bias in the raw precipitation can be selected. The main reason for this selection is based on the assumption that GCMs with a realistic current climate will have a realistic climate change signal.

Then, the runoff values generated by the WEAP model and affected by direct influences (climatic) and indirect influences (human water abstractions) are estimated. Finally, the future projection of meteorological and hydrological drought defined by the SPI and SRI are analysed.

Climate Projections and Impact on Drought

Figure 5 represents a comparison of the annual cycle of the historical GCM simulations compared with observations for temperature and precipitation data from four meteorological stations in the Zayandeh Rud basin. Despite the range of historical simulations for the observed annual cycle being wide, the ensemble mean catches the observed seasonal cycle and value of the temperature and rainfall. This gives more confidence that on average the climate projections’ models simulate the climate in Zayandeh Rud on this time scale. The ensemble mean is represented in order to capture the spread in outcomes produced by all ensembles.

Fig. 5
figure 5

Yearly fluctuations of monthly mean rainfall (a) and temperature (b) from 1971 to 2005 are shown. The data has been averaged for the four meteorological stations and compared with the ensemble of historical GCM simulations for the same 34-year period. The thick and thin purple lines show the mean and range of observed monthly values, respectively. The thick and thin red lines indicate the mean and range of the ensemble means from 38 GCMs, respectively. The vertical lines show ± standard deviation from the means for the observations and GCMs projections

Full size image

The figure explains both temperature and precipitation values of each ensemble member. Furthermore, the monthly precipitation simulation produced by the mean of ensembles shows a smaller bias with observed data.

Figure 6 indicates the RCP8.5 simulations for 2006–2100. In the future (2006–2100), temperature increases on average of about 4 °C are usually statistically significant (p < 0.01) compared to 1971–2005 in 8 out of 12 months. Also, generally, precipitation is projected to decrease in the future. The statistically significant (p < 0.01) changes occur during the period from January to May.

Fig. 6
figure 6

Annual cycle of monthly mean rainfall (a) and temperature (b) averaged for the four meteorological stations in Zayandeh Rud basin from 1971 to 2005 is shown in this figure. This has been compared with the ensemble of downscaled RCP 8.5GCM simulations done for the period 2006–2100. The thick and thin purple lines show the mean and range of observed monthly values, respectively. The red and black lines are the mean and range of the ensemble means from 38 GCMs, respectively. The vertical lines show standard deviation from the means for the observations and GCMs projections

Full size image

Selection Climate Model to Input to the Hydrological Model

In order to use the variables from the GCM model as an input for the hydrologic model (WEAP), the best model needs to be selected. Although the downscaling method removes the GCM bias in the mean and the variance, selecting the best model with CDF is desirable, because it is assumed that a model with a realistic current climate has a more realistic climate change signal. Comparison between observed and simulated climate change to select the best model is not easy as climate for both observed and simulated conditions varies as a result of increasing atmospheric greenhouse gas concentration or other forcing changes (forced signal) plus natural variability (natural variability noise) under these conditions. The cumulative distribution function for mean monthly precipitation was used to select the best model. CDF is an empirical statistical technique which shows how GCM simulated values match with observed values at the same period of time (1971–2005) [15, 25]. Internal variability simulated and observed time series do not match in time; however, for the CDF, this does not matter, as it describes only the distribution regardless of when the individual values were observed or simulated. As shown in Fig. 7, GCM outputs have deviations from the observed data for 1971–2005. Significant differences between the CDFs are derived with different GCM models; however, only one model has a CDF similar to the observed data as shown in Fig. 8.

Fig. 7
figure 7

Cumulative distribution function of the 38 models for mean monthly precipitation

Full size image
Fig. 8
figure 8

Cumulative distribution function of the best model for mean monthly precipitation

Full size image

River Flow Simulation and Impact of Human Water Abstraction on Reduction of Flow

To obtain the total impact of human water use on the drought, the flow upstream and downstream of the Zayandeh Rud basin has been compared with total water consumption (for agriculture, domestic and industry) (Fig. 9). With respect to the assumption for increasing population and measuring irrigation demands, generally, the water demands are expected to increase by about 4% per year (2006–2100). It should be noted that because of a lack of groundwater, the water users mostly depend on surface water. The figure shows that the flow downstream is too low and the flow upstream is higher; however, upstream, the river flow is not able to cover all water demands during 2006 to 2100.

Fig. 9
figure 9

Comparison of average water supply and demand in the Zayandeh Rud basin over the period 2006–2100

Full size image

The Zayandeh Rud river flow with and without human abstraction has been compared in Fig. 10. Annual river flow and peak flow declined due to increased evapotranspiration as a result of irrigation water use when compared with the flow without human abstraction. Human water abstraction exceeds the effects of low flow during the given drought period.

Fig. 10
figure 10

Comparison of measured flow and flow with demand abstraction over the period of 2006–2100

Full size image

The reservoir regulating measures in Zayandeh Rud are not sufficient to compensate for the abstraction and therefore the low flow regime changes to even drier conditions. The following sections show the drought severity and vulnerability which includes both climate change and human abstractions’ impacts. The figure also shows that for some years (such as 2083, 2084, 2087 and 2088) there is a significant flow reduction due to consecutive dry years with low precipitation and high evapotranspiration.

Simulated stream flows under climate change and human abstractions are shown for the four stations and overall basin respectively in Fig. 11. Stream flows for all sub-basins show a decreasing trend. The Mann-Kendall test was applied on the simulated stream flow, which confirmed the presence of a decreasing trend in Figs. 11 and 12. At a 5% significant level (α), the null hypothesis stating there is no trend in the simulated stream flow was rejected.

Fig. 11
figure 11

Future simulated stream flows for the four stations in the Zayandeh Rud river basin

Full size image
Fig. 12
figure 12

Future simulated stream flows for the whole Zayandeh Rud river basin

Full size image

A mean monthly stream flow analysis was carried out on the 34-year historical stream flow recorded at the four gauges applied to calculate overall basin performance in comparison with the simulated stream flow statistics (Fig. 13).

Fig. 13
figure 13

The box plots of monthly stream flow for observed data (1971–2005) and for future simulation data (2006–2100). In each box, the central points is the mean value; the lower and upper edges of the box are the 25th and 75th percentiles, respectively and the whiskers extend to the min and max data points

Full size image

The figure shows the historical mean monthly stream flows are higher than the simulated values in the future. This can be attributed to the fact that rainfall is predicted to be less in the future.

Projected Change of Intensity-Duration and Frequency of Drought

Precipitation data from the climate model was used to project future meteorological drought. In order to determine hydrological drought, the WEAP model was applied to project future runoff given the precipitation, temperature, relative humidity and wind speed. For drought projections, ensemble mean monthly precipitation should not be applied since the averaging process decreases the monthly variation of precipitation and may give misleading outputs. One of the 38 models was chosen as it matched the 1970–2005 observational periods with the most similarities from a statistical point of view (by using Student’s test for significance between the observed climatic variables and the historical simulation models). The severe droughts on the SPI are projected to be more severe than those on the SRI. However SRI is similar to SPI in terms of onset and timing of droughts. Comparing the time series for the SPI and SRI also shows that the variability of SPI is more significant with weaker persistence compared to SRI, because the rainfall variability was filtered by the hydrologic system of vegetation, topography and soil.

The most significant dry years are extracted from Figs. 14 and 15; then, the minimum index, driest month and maximum duration for both meteorological and hydrological are analysed and are shown in Tables 1 and 2. The red in the tables indicates the longest duration (10 to 12 months) and the blue colour illustrates the shortest duration (1 to 3 months). The tables also indicate the intensity of the most significant meteorological droughts is more variable compared to hydrological drought. Table 1 shows that the driest month in all meteorological drought events is in winter or spring in the near future. However, the driest month in all hydrological drought events is in summer or autumn. Maximum duration for both meteorological and hydrological drought events is between 5 to 12 months for the near future. The minimum index for both meteorological and hydrological drought in the year of 2018 is the most significant.

Fig. 14
figure 14

Projected time series variation of SPI and SRI for four stations of the Zayandeh Rud basin for the near future (2005–2040) in the left and for the far future (2041–2075) in the right

Full size image
Fig. 15
figure 15

Projected time series variation of SPI and SRI for the near future (2005–2040) in the left and for the far future (2041–2100) in the right for the Zayandeh Rud basin

Full size image
Table 1 Characteristics of droughts at 12-month timescale for the period of 2005–2040
Full size table
Table 2 Characteristics of droughts at 12-month timescale for the period of 2041–2075
Full size table

Table 2 represents the driest month in all meteorological drought events is in winter or spring for the far future. While the driest month in all hydrological drought events is in summer or autumn. Maximum duration for both meteorological and hydrological droughts is between 4 to 12 months. Furthermore, the minimum index for both meteorological and hydrological drought in the year of 2056 is the most significant in the period of 2041–2075.

Comparison between characteristics of droughts between baseline (Table 3) and future time (Tables 1 and 2) shows that minimum index for both meteorological and hydrological drought for the near and far future time is greater than those in the baseline. The maximum duration for meteorological drought was between 3 and 7 months for 1971–2005 (Table 3) which is shorter than those predicted in future times. However, the maximum duration for hydrological drought was between 4 and 12 months.

Table 3 Characteristics of droughts at 12-month timescale for the period of 1971–2005
Full size table

Analysing drought frequency (Fig. 16) demonstrates that in general the number of drought events declines with the rise of drought duration for both drought indices. Comparison between baseline and future projections shows that the difference of the frequency varies with drought indices. For a given drought duration, the number of drought events enhances from baseline to the future. For example, for SPI with a duration of 6 months, the frequency increases from 17 under the baseline to 83 and 33 under the near and far future. The frequency for both droughts with the maximum duration (e.g. 12 months) is increased during the near and far future.

Fig. 16
figure 16

Frequency of meteorological drought occurrence in the left and hydrological drought occurrence in the right for baseline period (1971–2005) and the near (2006–2040) and far future (2041–2075)

Full size image

Projected Change of Extreme Drought Events

Intensity, duration and frequency of droughts are applied to analyse extreme drought events, which in practice can be more important for drought management [30]. Table 4 represents the longest duration for drought intensity less than − 1 with the two indices. The longest duration rises from baseline to the future. For example, the longest duration for SPI grows from 40 months under baseline to 61 months in the near future and 94 months in the far future. The longest duration for SRI in the near and far future is 71 and 103 months, which are 1.18 and 1.71 times more than the baseline.

Table 4 The longest duration (by Total Months) of drought events (drought events here are determined by thresholds of intensity I < − 1 and are estimated by SPI (Standardized Precipitation Index) and SRI (Standardized Runoff Index)
Full size table

Table 5 indicates the change of drought intensity under baseline and future time scale. It also signifies that the impact on the most severe hydrological droughts is greater than for the meteorological drought. So the intensity of the severe drought increases from − 1.96 to − 2 for SPI for the near future, from − 1.96 to − 2.1 for the far future. The intensity also rises from − 1.46 to − 1.61 for SRI for the near future and − 1.45 to − 1.66 for the far future.

Table 5 Minimum values in SPI and SRI time series
Full size table

Extension of Drought from Meteorological to Hydrological Systems

The climate change impact on meteorological and hydrological drought differs under baseline and future time [30]. The change in the total number of drought events and frequency of drought occurrence in terms of both drought indices under the RCP8.5 scenario is not small. Figure 17 indicates the number of drought events with different intensity and duration during baseline and the near and far future time scales. SPI projects drought frequency to increase with durations of 5, 6 and 7 months and SRI projects drought frequency to increase with durations of 6, 8 and 12 months, while the frequency decreases for durations of 1, 2, 3 and 4 months for both SPI and SRI. Hydrological droughts are more sensitive to climate change than meteorological drought (SPI) due to the nonlinear response of soil moisture and runoff to the precipitation and temperature changes [30].The precipitation change decreases from January to May and November to December significantly. The most significant temperature change occurs between February and June.

Fig. 17
figure 17

The number of meteorological droughts (in the left) and hydrological droughts (in the right) with intensity I < − 1 and the different duration

Full size image

Discussion

The current GCM models have small potential resolution in their projections of future rainfall and temperature. Most of these projections are of global climate change; the regional level projection may differ within a large range [12, 15].

With lack of regional climate models, for better water resources’ planning in the developing world river basins need to combine with effects of climate change on extreme events, which are based on the downscaled GCM projections [5]. Applying particularised downscaling methods perhaps is as data exhaustive as providing regional climate models. Therefore, the use of uncomplicated statistical methods (such as the method (the ‘Climate change simulations’ section) which was used in this study) is more suitable for projections in river basins of developing areas [14]. The results in this study showed that there are some differences between mean values of climate model simulations and observations (Figs. 5 and 6), so a bias correction was performed.

In this study, the potential force of future climate alteration on drought was estimated by using two drought indices (e.g. SPI and SRI). Climate alteration projections on the basis of statistical downscale through GCM climate models are applied to calculate drought frequency, severity and period by extension from meteorological to hydrological drought. At first, in order to evaluate uncertainties in the projections, a multi-model ensemble from 38 monthly GCM simulations with the severest emission scenario (RCP 8.5) from CMIP5 was applied. The scenario of RCP 8.5 is selected in order to show the severe drought conditions under climate change. Previous research [23] stated that developing the climate change scenarios is necessary in order to better understand the changes in relation to the amount of CO2 concentration and radiative forcing that influence climate response. Thus, in this study, with selecting the highest CO2 emission scenario (e.g. RCP8.5), some uncertainty remains in the projection of future drought.

None of the SRES simulations [11] in previous studies considered projecting changes in temperature and precipitation extremes as in RCP 8.5, which has the largest radiative forcing among the scenarios. Finally, the best model (between 38 models) which matched with the observation data was selected based on the CDF method. Although the presented method for selecting the best model aims to show the strengths of the model to simulate observed climate change (specially for historical period) and assumed that the best simulation of the past can be more realistic to simulate climate change signal for future period, few limitations remain:

  1. 1)

    There is uncertainty in the features of the future precipitation dynamics and no single or group of ensemble model cannot clearly stand out in performance of the features.

  2. 2)

    The best model was selected in this study based on the skill of the model in simulating past climate and so it may eliminate the range of possible projections for unknown type of climate change.

  3. 3)

    The selected model is not a sample that captures the structural uncertainty in the model itself; for example, the model did not consider uncertainty in the structure of numerical method used.

Previous approaches only use one or some randomly selected GCMs. Others use minimum, maximum or mean ensembles without analysing climatic anomalies for intra-annual variability to consider the best model for climate projection.

In this study, the impact of both possible climate change and human water use (based on assumptions) on the projected hydrological drought characteristic for 2006–2100 has been considered (Fig. 16). Obtained future drought simulation results (in terms of intensity, duration and driest month) were compared to the historical period (Tables 1, 2 and 3). The integration of climate models, emission scenarios and human water abstraction obtains future hydrological drought predictions.

A previous study which focused on global scale only used a period of future time (e.g. 2070–2099) instead of a continuous period, to predict future hydrological drought.

Other indices such as EDI (Effective Drought Index) or PDSI (Palmer Drought Severity Index) were not used for this study as they need daily precipitation and data of soil layers which were not available.

The negative impact of climate change on the flow regime with and without human water abstraction (Figs. 10 and 11) is projected. This projection in this study is based on hydrological model of WEAP (which include some uncertainties) to determine future flow projection.

Due to the increased temperature and decreased precipitation as well as the nonlinear hydrological responses to precipitation and temperature change, they are the most important factors in the change of drought characterisation. In addition, high wind speed and high evaporation increases the drought severity in the basin. The sensitivity of the indices to changes in precipitation and flow was tested. The results in this study show that both meteorological and hydrological droughts may increase in the future (Fig. 17), primarily as a result of increased temperatures (Fig. 6). Intensity, duration and frequency of drought are likely to rise across all time periods due to climate change and human influences (Tables 4 and 5).

According to a previous study [33], the future projected changes of temperature and precipitation in the Asian and Middle East regions may cause significant intensification of surface heat fluxes; however, this effect is not directly estimated in this research. As discussed in previous research [18], decreasing precipitation and longer dry spells probably cause drought.

The rainfall in Iran is generated by Mediterranean synoptic systems, which flow eastward along with western winds in the cold season. Variability in annual rainfall is made by synoptic systems and year to year change in the number of passing cyclones. Frontal Mediterranean cyclones co-operating with the western airflows make rainfall in late autumn and especially in winter in Iran. It is noteworthy that the Mediterranean system precipitation and also the possible influence of ENSO, which can increase drought risk in future in Iran, is not well simulated in future climate models [7]. However, the GCMs used in this study have improved compared with previous models [35]. They now contain the likeness of the ocean, atmospheric chemistry, vegetation, carbon cycle, land surface, aerosol and sea ice at finer spatial resolution [].

Potential risk from drought in the future should be estimated keeping in mind the increased demands of water for domestic, industrial and agricultural uses in addition to the forecasted changes in the climatic variables, such as precipitation, temperature, humidity and wind, especially in the rapidly growing regions of Asia, the Middle East and North America [29]. The anthropogenic impact on drought is less well known and such influences have seldom been surveyed.

The results in this study show that low peak flows and increased abstraction of water by humans causes more risk of hydrological drought events.

Insufficient and poor data related to land use [6] and non-inclusion of projected population and associated demands of water have not been addressed in earlier studies. Furthermore, high uncertainties in climate change prediction models give wide-ranging results. Crop pattern values in this study remain the same as historical data, but changes in population and water demands are projected. Therefore, this study is limited in terms of potential future land use changes. This study attempts to give some preliminary ideas of possible changes in risk of drought combining climatic changes and changes in anthropogenic demands in case of business as usual, and this has not been studied for the Zayendah Rud basin of Iran.

Without awareness of possible future events, it is hard to improve the conception about future adaptation to change in the risk of drought impacts associated to climate change and water abstractions by humans. For this reason, decision makers in national and regional governments require facts regarding the potential vulnerability of drought impacts associated with climate change and how it could be decreased.

Conclusion

Projecting the potential impact of climate change on meteorological-hydrological drought is necessary, particularly for regions like Iran where the projected climate change impacts are investigated rarely for water planning and water management.

This paper addressed the important objective of water resource management under climate change impacts on drought conditions by applying a statistical method for climate change studies to assess the potential impact of future climate change on drought characterisation. An ensemble of statistically downscaled variables from GCM projections engaged in the CMIP5 was applied as analytical variables.

Changes in climate extremes by selecting the highest CO2 emission scenario (RCP8.5) may indicate larger impacts on droughts.

The analysis represents a reduction of rainfall over the basin in the future scenario. Following this change, the stream flow values will decrease and will influence the water availability for users.

Stream flows are set to reduce in the future. This has been verified from the plots and as well as applying the Mann-Kendall test. The magnitude of future maximum monthly stream flow is expected to decrease relative to observed maximum monthly stream flow.

Findings from this study showed a significant potential future growth in drought cases, mainly due to warming climate and human water abstraction.

Significant meteorological drought is expected to occur in the winter and spring months of January to June. However, the driest month for hydrological drought is in the summer and autumn (July to December) (e.g. no changes in seasonality of droughts compared to historic period).

The human influences on projected hydrological drought, previously missed in many projections for future hydrological drought, have been outlined in this study. This study confirms the previous study [4] which stated that human influences can account for future hydrological drought in areas of Asia, the Middle East and the Mediterranean.

Therefore, in the Zayandeh Rud basin, as an example of these regions, low flows are expected to be even lower in the future and drought will likely rise dramatically. Better scenarios of future human water demand can provide great skillful projections for the twenty-first century. Nevertheless, they are not available yet, as a result of the lack of comprehensive future socio-economic and land use projections that are dependent on each other.

Currently, human water abstraction has an additional influence on hydrology and water resources and so it is necessary to involve it in hydrological models applied for projections of future hydrological drought.

Determination of this study is based on a future modelled climate simulation that may not represent full reality. The estimation was completed assuming that some, but not all non-climatic factors which may influence the future dynamics of drought, will stay constant. For example, the irrigation infrastructure, cropping patterns, irrigation and farming techniques, the water distribution and water use efficiencies for the future period are speculated to be same as in the baseline (historical) period.

However, any change in the irrigation demand in the future is associated to crop evapotranspiration demand forced by climate change. Evapotranspiration demand is expected to rise as a result of temperature effects. In addition, increasing future industrial and domestic water demands will affect the deficit of the river flow and hydrological drought.