Drought occurrence under future climate change scenarios in the Zard River basin, Iran

Global warming affected by human activities causes changes in the regime of rivers. Rivers are one of the most vital sources that supply fresh water. Therefore, management, planning, and proper use of rivers will be crucial for future climate change conditions. This study investigated the monitoring of hydrological drought in a future period to examine the impact of climate change on the discharging flow of the Zard River basin in Iran. Zard River is an important supplier of fresh and agricultural water in a vast area of Khuzestan province in Iran. A continuous rainfall-runoff model based on Soil Moisture Accounting (SMA) algorithm was applied to simulate the discharge flow under 10 scenarios (obtained from LARS-WG.6 software) of future climate change. Then, the Stream-flow Drought Index (SDI) and the Standard Precipitation Index (SPI) were calculated for each climate change scenario for the future period (2041–2060). The results of the meteorological drought assessment showed that near normal and moderate droughts had higher proportions among other drought conditions. Moreover, the hydrological drought assessment showed the occurrence of two new droughts (severe and extreme) conditions for the future period (2041–2060) that has never happened in the past (1997–2016).


INTRODUCTION
Drought and flood are two major events in Iran's climate.
According to climatologists' points of view, the frequency and perpetuation of these events imply occurrence of climate change in Iran. Drought is a natural disaster which is not apparent until its final stage. In addition to precipitation patterns, other extreme climatic events such as severe temperature and low relative humidity often happen along with drought in many parts of the world, which can dramatically increase drought intensity. Drought is caused by a shortage of water in an area, leading to different effects in different parts, especially in the environment. Drought is not just a physical or natural phenomenon. The effect of drought on society is due to the interaction between a natural phenomenon (lower precipitation can be caused by natural climate change) and the people who need water. Therefore, the lack of attention to sustainable development in human activity, improper consumption and industrialized societies will exacerbate the effects of drought (Wilhite & Pulwarty ). Drought has been mainly studied in meteorological, hydrologic, agricultural and social-economical. In investigating meteorological drought, the frequency, duration, and intensity of low precipitation are studied. In the other studies, the effects of drought on rivers' discharge, changes in soil moisture and its human consequences are considered Two meteorological and hydrological droughts have been investigated in this paper. The hydrological drought occurs when the meteorological drought continues for a long time so that the rivers' flow or the underground water level decreases. This phenomenon occurs due to the lack of winter precipitation in the mid-latitudes (Yildiz ).
Therefore, any change in the precipitation pattern can influence the amount and intensity of droughts. Many studies have been conducted on droughts. For example, Jasim & Awchi () investigated regional meteorological drought in Iraq and employed the Standardized Precipitation Index (SPI). In this research, only the past period  was investigated. The results indicated that the mild drought has higher proportions among different types of meteorological droughts. Adib & Tavancheh ()  Their results showed that almost all the stations suffered from extreme droughts during the study period. All the above studies either only use past periods to evaluate droughts or just use indices related to the precipitation. In this research, not only past and future periods are investigated, but also meteorological and hydrological droughts are considered. Consequently, these studies prove that drought monitoring by using different indexes is suitable to detect the duration and intensity of hydrological and meteorological droughts. Additionally, local managers and policy-makers can apply it to their water resources planning.
The fifth report of the Intergovernmental Panel on Climate Change (IPCC ) showed that global warming causes changes in the water cycle due to increases in concentrations of greenhouse gases. Moreover, the precipitation pattern will change more severely in dry and semidry areas of the world. Both flood occurrence and winter precipitation decrease can be the results of these changes.   Table 1 presents the stations by their characteristics.

Case study and data usage
In Table 1, the historical period data (precipitation, runoff, temperature, and evaporation data) is provided by the Iran meteorological organization (a government agency) and the water utility company of Khuzestan province (a government agency). Also, soli map, land use-land cover and digital elevation maps were obtained from the water utility company of Khuzestan province.
The average annual precipitation in the studied drainage basin is about 580 mm and it varies from 402 to 792 among different stations. Table 2

LARS-WG 6 includes climate scenarios based on 5
GCMs. Table 3 shows the list of GCMs in the LARS-WG 6.
It covers the whole period of 2041-2060 studied in this research. Also, Figure 2 shows the whole process of producing daily site-specific climate scenarios.

Drought indices
Drought indices are derived from short-term data series of precipitation, runoff, soil moisture, and river flows, in order to present an understandable large sample. To make better use of raw data to be understandable and increase the decision-making ability of designers and planners, we expressed the aforementioned indicators only numerically.
We can use the obtained data of these indicators to detect the features of drought including duration, severity, and frequency. None of these indicators are superior compared to the other indicators, but some of them are more suitable for some applications (Barua et al. ). In this study, the

Standardized Precipitation Index (SPI) and Streamflow
Drought Index (SDI) were used in the interpretation of the meteorological and hydrological drought.

SPI index
This index is based on calculating the probability of rainfall for any timescale. SPI uses the monthly precipitation where α and β represent shape and scale, respectively, x represents the amount of precipitation and Γ(α) is the gamma function. We used the maximum likelihood method to estimate the parameters α and β.
x is the average of precipitation and n is the number of days with observed rainfall. Gamma function is undefined for the precipitation number of zero (x ¼ 0) since it is possible to have precipitation data equals to zero among the data. Therefore, the cumulative probability function that includes zero values is defined as follows: where q is the zero chance of precipitation. If m is the number of precipitations, whose magnitude in the series equals zero, so q can be estimated from the q ¼ m=n equation. Changing the shape of the gamma cumulative probability is based on a random variable of Z (standard precipitation) with zero mean and variance. For a given month and time scale, the cumulative probability G(X) of an observed amount of precipitation is given by: The drought happens when the amount of standard precipitation is continuously negative and becomes À1 or less, whereas positive values represent the termination of drought (Table 4). We can use the total negative values to analyze the characteristics of a standard rainfall drought (duration, magnitude, and intensity).

SDI index
To compute the SDI index, we used the equation below to make a data series from the mean of the monthly river discharge series (Qij) (Liu et al. ): where V i,k is cumulative flow discharge. In addition, i and j represent water year and the months of the water year, respectively. For example, If k ¼ 1, V i,k is related to the first three months of water year for the i th water year. SDI based on these data sets of river flows and for the water year base period of k related to the i th water year, SDI can be obtained from the following equation: (8) v k and S k are the mean total volume flow rate and standard deviation of cumulative flow volume, respectively, for the base period k in a long time. Table 5 shows different drought states in the SDI method (Nalbantis ).

Loss module
One of the most complex parts of hydrological modeling is the Loss module. This is because of a large number of activities that are computed simultaneously in the module; thus, the component values are required to be well defined.
In this study, the algorithm of SMA was used. This algorithm is used to simulate the long-term relationship among precipitation, runoff, storage, evapotranspiration and soil loss in the basin. This algorithm divides the surface of the basin into five parts (Figure 3), which are separately explained below.
Since the formulas related to the algorithm of the soil moisture are unexplainable in their entirety in this research, only the main ones are discussed. These formulas include phenomena such as infiltration, percolation, lateral movement of groundwater, etc. The soil infiltration in the calculating algorithm of the soil moisture is computed by potential soil infiltration (PSI) (mm/hour), which is calculated by using the following equation (Bennett ): where MSI is the maximum infiltration of soil (mm/hour); SOS t is the maximum capacity of the soil (mm) and SOS m is the amount of water volume in the soil (mm). This  equation implies that the potential soil infiltration cannot be more than maximum soil infiltration and it is linearly related to the water volume of the soil. If soil is without water or with a low amount of water, the potential soil infiltration can be equal to the maximum soil infiltration. Actual soil infiltration (ASI) at time t is calculated by the following equation: where AW t is the amount of permeable water. It is inferred from Equation (10) that actual soil infiltration cannot be more than current soil permeability. The percolation is similar to the permeability phenomenon, which is the water amount that reaches the groundwater via soil layers. Potential soil percolation (PSP) is calculated by the following equation (mm/hour): where MSP is the maximum soil percolation, SOS t is the amount of water volume in the soil (mm), SOS m is the capacity of the soil (mm), GWS t is the storage of underground water for the current layer of water (mm) and GWS m is the maximum storage of underground water (mm). Equation (11) shows that the potential soil percolation depends on the amount of water in the soil as well as the underground water. If the storage of underground water is in saturated or saturation condition, the large amount of water cannot permeate the groundwater aquifers.
The output current of underground water layers indicates groundwater flow in a peripheral mode (a flow that is ultimately added as base flow to surface flow). The algorithm of soil moisture calculation of this peripheral current can be computed by the following equation: GWF t and GWF tþ1 are the groundwater flow rates (mm) at the beginning of the time interval t and t þ 1, respectively; PGWP t is potential soil percolation of underground water (mm 3 /hour); K is the coefficient of underground water storage layer(hour); t is the simulation time step.

Evaluating the hydrological model
For evaluating the model, we used the following equations: Root Mean Square Error (RMSE) (Peak): Mean Absolute Error (MAE) (Peak): Coefficient of Determination (R 2 ): Nash-Sutcliffe Efficiency (NSE): where Q O , Q O and Q S are observed, averaged and simulated flows rates in m 3 =s, respectively.

Rainfall-runoff modeling
We employed Continuous Hydrological Modeling to simulate future climate projections of runoffs and calibrated the continuous model of rainfall-runoff by using data of six water years (1975-76 to 1980-81 Table 7. These results illustrate a close agreement between the simulated streamflow and observed flow. Also, Figure 4 shows that the model fitted very well in daily simulating the streamflow, base flows, and peak flows. For validation of the model, the rainfall and runoff data of three water years (1998-1999 to 2000-2001) are used.
Evaluation of the validated model is shown in Table 8.
The results imply that there is a close agreement between the simulated streamflow and observed flow for the validation period. Figure 5 shows that the model can simulate daily runoff.
It is worth mentioning that the sensitivity parameter of the SMA algorithm was also performed. The sensitivity analy-    and tension storage are the most sensitive parameters, and groundwater 2 and groundwater 2 percolation are the least sensitive parameters in the loss module. Additionally, from sensitivity analysis, it can be concluded that the moisture condition of the upper soil layers affects sensitivity analysis more than the groundwater layers and this may be because of significant changes in the soil moisture conditions compared to the groundwater conditions. Also, it is worth mentioning that the baseflow method can also reduce the sensitivity of the groundwater parameters in some cases. The parameters' sensitivity ranking is presented in Table 9.

The meteorological drought and rainfall change
In this research, to assess annual meteorological drought, we calculated the SPI index for a 12-month interval for all selected rain gauge stations in both periods of the future and past. Figure 6 shows the temporal distribution of SPI based on 10 climate change scenarios in the future period compared to the baseline period. As is shown in Figure 6 From a comparison of the previous and future periods, we can conclude that the extreme drought in the past period was higher (7-11%) than the future period. Besides, the severe and moderate droughts in the past period were usually lower than the future period. Changes in the nearnormal condition vary in different models of the future period compared to the past period.
An investigation of the annual rainfalls shows that the precipitation increases in the future period compared to the past period. The models under the RCP85 scenario show more  to a maximum of 13.2 and 17%, respectively. Figure 7 shows the annual rainfall for all 13 stations. Overall, it can be concluded that the models under the RCP85 scenario show a more intensive change in precipitation compared to the models under the RCP45 scenario.

Hydrological drought index and runoff change
We evaluated the hydrologic drought using the SDI index.
In order to evaluate the whole hydrological drought, we obtained and drew the index by using 10 climate scenarios for time intervals of 3, 6, 9, and 12 months in the period of 2041 to 2060 in the future and 1997 to 2016 in the past (Figures 8 and 9).
The biggest wet-state (no-drought) in 3, 6, 9, and 12-month time scales in the future is related to the GFDL-CM3-RCP45, MIROC5-RCP45, HADGEM2-RCP45, HADGEM2-RCP45 models respectively. The biggest severe drought in 6, 9, and 12-month time scales in the future is related to the HADGEM2-RCP45 model. The biggest wetstate in 3, 6, 9 and 12-month time scales in the past period is seen in years of 1999, 1997, 1997, and 1997, respectively. The biggest moderate drought in the past period in 3, 6, 9, and 12-month time scales is related to the year 2013. In general, contrary to the base period, severe drought will happen in the future that never happened before. Also, the RCP45 scenarios show more intensive droughts compared to the In the historical period, only mild and moderate droughts happened. But, in the future period, severe and extreme droughts will also happen. Moreover, the percentage of the occurrence of years without drought will be higher compared to the historical period. Thus, 4 to 9% will be added to the occurrence of years without droughts related to 3 to 12-month time scales. In the future, the percentage of the occurrence of drought will be decreased, but the type of drought will alter from moderate to severe. The mild droughts in the historical period will change from 40-55% to 35-43% in the future. The occurrence of moderate droughts in the future depending on the time scale of 3 to 12 months will be between 7-15%. The occurrence of the severe drought will increase from 0 in the historical period to 5% in the future period. The occurrence of extreme droughts in the future depending on the time scale of 3 to 12 months will be between 0-1%. Also, by increasing the time scale of 3 to 12 months, the fluctuation of occurrence of droughts will decrease; as, the occurrence of droughts of 9 and 12 months will be similar to each other.
However, from the 3 to 6-month time scale, the difference in the percentage of occurrence of different types of drought and the no-drought conditions will be more obvious, which is related to the interannual fluctuations and changing of the seasons. It is obvious that the extreme condition of the RCP85 scenario caused more intense changes in drought conditions compared to the historical period as well as the RCP45 scenario.
Annual runoff analysis shows that the runoff will increase in the future compared to the historical period. Figure 13 shows the annual runoff for both future and historical periods.
The results confirm that the models under the RCP85 scenario show more increase in runoff compared to the models under the RCP45 scenario. The runoff will increase by about 8.8 and 13% under the RCP45 and RCP85 scenarios respectively.

DISCUSSION
In this study, we used the new LARS-WG6 stochastic weather generator (one of the first studies or even the first study that

CONCLUSIONS
The analysis of the SPI index showed that near-normal and moderate droughts have higher proportions among different types of meteorological droughts. A comparison between the historical period and future period indicates that the frequency and pattern of extreme drought will change in the future. These results can indicate that the future change in droughts can move outside the historical envelope due to climate change. An analysis of the SDI index showed that the number of droughts in the past period was more than the future period. However, the intensity of the drought is higher in the future period compared to the past period.
These results can indicate that the RCP45 and RCP85 scenarios may cause higher temperature change and a warming atmosphere causes more evaporation. That is, more water is available for precipitation; thus, it may be the reason for more intense changes in drought conditions in the future period. Also, severe and extreme droughts will occur in the future period that have not happened in the past period. Based on the intensity of the occurrence of droughts in the future period, we recommend more investigations into adaptation strategies guidance for water supply and food production to meet severe droughts. Our suggestions for water resources managers to deal with the adverse effects of climate change are as follows: • Extending efficient systems for water use • Providing certain technologies for the proper management of farms • Preserving the existing soil and ground water resources • Utilizing popular media, including radio and TV, to manage education and awareness campaigns FUNDING This research received no external funding.