Abstract

This study was conducted to assess the impacts of climate change on drought over the Lake Urmia basin, Iran. Drought events for 2011–2040, 2041–2070, and 2071–2100 were analyzed based on the Standardized Precipitation Index (SPI) and the Standardized Precipitation Evapotranspiration Index (SPEI) and were compared with the adopted baseline period (1976–2005). The SPI and SPEI were calculated using the precipitation and temperatures obtained from the second-generation Canadian Earth System Model (CanESM2) under Representative Concentration Pathway (RCP) 2.6 and RCP 8.5 as optimistic and pessimistic scenarios respectively. The results of SPI analyses revealed that under RCP 2.6 the frequency of droughts is almost constant while under RCP 8.5 drought frequency increased especially in the period 2071–2100. The calculated SEPI under both scenarios and during all future periods predict that the frequency and duration of droughts will increase. Generally, the difference between the SPI and SPEI is related to the input to each index. SPI is solely based on precipitation while the SPEI accounts for both precipitation and potential evapotranspiration (PET). Under global warming and changing climate, the significant role of PET was highlighted. It was concluded that the SPEI outperformed the SPI for drought studies under a changing climate.

INTRODUCTION

Drought can be defined as a lack of precipitation, stream flow, reservoir storage, and a decrease in groundwater level (Chitsaz & Hosseini-Moghari 2018) or it can be considered as a period of time in which the available water cannot supply water demands (Tsakiris et al. 2007). Drought is a natural disaster which occurs in every climate and has many adverse social, economic and environmental impacts. Statistics have shown that 22% of the total damage and 3% of deaths caused by natural disasters worldwide are related to drought events (Wilhite et al. 2007). Data from the Emergency Events Database disclose that over the period 1970 to 2007, drought caused over $29.5 billion of losses across Asia (Kallis 2008).

While currently the impacts of drought are significant, studies have shown that global warming and climate change have the potential to increase drought occurrence and frequency (Sharma & Mujumdar 2017). This has prompted increasing research into droughts and the effects of climate change on them (Staben et al. 2015; Anderson et al. 2018). In this regard, Hosseinizadeh et al. (2015) assessed the effect of climate change on drought over the Dezful basin in Iran using the Standardized Precipitation Index (SPI). This study considered outputs of 15 General Circulation Models (GCMs) under A2, B1, and A1B scenarios for the period 2020 to 2044. The results of the 2015 study concluded that the frequency of moderate and severe droughts would decrease under all scenarios and models. Yuan et al. (2017) studied drought projections under climate change over the middle and lower Jinsha River basin in China. They used the outputs of five GCMs under Representative Concentration Pathway (RCP) 4.5 for the period 2021 to 2050. The results of this 2017 study concluded that for the future period, the area that would experience drought would increase by 43.2%. Lee et al. (2014) evaluated the effect of climate change under A1B, A2, and B1 scenarios for the near future (2011–2040), middle future (2041–2070), and late future (2071–2100) on the Han River basin in Korea. They used output corresponding to the Coupled GCM (CGCM) V. 3.1 under A1B, A2 and B1 scenarios. Their results showed that under all scenarios and future periods, drought would be mitigated in terms of frequency and magnitude.

Dastorani et al. (2011) assessed the potential of climate-change impacts on drought across Yazd, Iran, using HadCM3 outputs under A2 and B2 scenarios for the period of 2010 to 2039. The delta method was used for downscaling the GCM outputs. While the 2011 results showed that under the A2 scenario the trend for the drought indices would be negative and the drought indices would experience a positive trend under the B2 scenario, the result indicated an increase in vulnerability to future droughts. Liu et al. (2013) analyzed drought projection during the 21st century across the Arkansas Red River basin in the USA under A1B and A2 scenarios using the outputs of 16 GCMs. To this end, Liu et al. (2013) used the SPI and Palmer Drought Severity Index (PDSI) to quantify droughts. Both drought indices under both of the emissions scenarios projected that droughts will occur with greater severity and frequency from the middle of the 21st century.

These previous studies indicate that the effects of the climate change of drought are different in every region. Therefore, there is need for drought and the effects of climate change on it to be assessed separately in each region. While many drought studies under climate change have been performed based on the SPI, rising temperatures under global warming also play an important role especially in arid and semi-arid areas. Therefore, in the present study, drought events were analyzed under changing climate across the Lake Urmia basin in Iran using both the SPI and Standardized Precipitation Evapotranspiration Index (SPEI). The assessment of the changes in drought frequency and duration was based on the findings of the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (Allen et al. 2014). The study area, data, and drought indices are described in detail as follows.

MATERIALS AND METHODS

Case study catchment

This study analyses droughts across the Lake Urmia basin in Iran. Lake Urmia, which is the second largest hypersaline lake in the world, is the largest water body in Iran (Azarnivand et al. 2015). It is located in the north-west of Iran and has a catchment area of 51,800 km2. During the last two decades, the lake water level has decreased by 7 m. Under these conditions, the volume of renewable water cannot meet human and lake water demands. As a consequence, available static water resources over the basin are being consumed, which highlights the serious impact that drought can have on the economy, the community and the environment across the basin. It is vital that water resource planning for the future considers the impact of the climate change of droughts. This study aims to identify the potential impact of droughts under a changing climate to inform water resource planning in the Lake Urmia basin.

Drought indices

The most common way to monitor droughts is through the use of drought indices (Hosseini-Moghari et al. 2017). The SPI (McKee et al. 1993) and SPEI (Vicente-Serrano et al. 2010) are the two most well-known indices. These indices can be calculated in multi-time-scales (1-, 3-, 6-, 9-, and 12-month intervals up to 48 months), so they can be used to detect short-, mid- and long-term droughts (Hosseini-Moghari & Araghinejad 2015). In general, 1-month and 3-month time-scales are used to identify short-term droughts, 6-month and 9-month time-scales are used to identify mid-term droughts, and 12- and 24-month time-scales are used to identify long-term droughts (Mishra et al. 2007). The calculation steps for the indices can be described as follows:

  • (a)

    fitting the proper probability distribution to input data (P for SPI, P-PET for SPEI);

  • (b)

    calculating the probability related to each data using the fitted distribution;

  • (c)

    converting the calculated probability to a standard normal distribution; and

  • (d)

    the corresponding value to the transferred probability (on a standard normal distribution) shows the SPI or SPEI value.

Details on the SPI and SPEI are detailed by McKee et al. (1993) and Vicente-Serrano et al. (2010) respectively. Guttman (1999) and Stagge et al. (2015) have shown that using a default probability distribution (gamma distribution for SPI and log-logistic distribution for SPEI) can cause an error in the results. So in this study, the appropriate probability distribution was selected for each month in the SPI and SPEI calculation process based on the recommendations of Guttman (1999) and Stagge et al. (2015). Table 1 shows the drought category which corresponds to each drought index interval.

Table 1

Drought categories based on the SPI and the SPEI

Standardized index Description D-scale 
−0.49 to 0.49 Normal WD 
−0.50 to −0.79 Abnormally dry D0 
−0.80 to −1.29 Moderate drought D1 
−1.30 to −1.59 Severe drought D2 
−1.60 to −1.99 Extreme drought D3 
−2.0 or less Exceptional drought D4 
Standardized index Description D-scale 
−0.49 to 0.49 Normal WD 
−0.50 to −0.79 Abnormally dry D0 
−0.80 to −1.29 Moderate drought D1 
−1.30 to −1.59 Severe drought D2 
−1.60 to −1.99 Extreme drought D3 
−2.0 or less Exceptional drought D4 

POTENTIAL EVAPOTRANSPIRATION CALCULATION

On the basis that Lee et al. (2017) demonstrated that the following equations (called the Torrent White (TW) method) have merit when estimating PET during the SPEI calculation process, this method was used in the present study to calculate PET. The TW method only requires monthly-mean temperature data. The PET value is calculated using the TW method as follows: 
formula
(1)
 
formula
(2)
 
formula
(3)
where PET is potential evapotranspiration (mm), K is a correction factor which is computed based on the latitude and month, T is the monthly temperature (°C), I is the heat index and m is a coefficient which is computed based on I. Further details about the Torrent White method are given by Safavi et al. (2015).

Data

For baseline period we have used the daily precipitation and temperature data from WFDEI (WATCH Forcing Data methodology applied to ERA-Interim data; Weedon et al. (2014)) between 1976 and 2005. These data are available for free worldwide with a spatial resolution of 0.5° × 0.5° through https://esg.pik-potsdam.de/search/isimip/. Our analysis showed that the WFDEI data are accurate enough over Lake Urmia basin in comparison with in-situ precipitation and temperature from 59 stations at a monthly time-scale with Nash–Sutcliffe efficiency (NSE) about 0.943 and 0.976 for precipitation and temperature, respectably.

Three future periods were assessed also as follows: 2011–2040, 2041–2070 and 2071–2100. The precipitation and temperature data for these periods were projected based on the predictors obtained from the second generation Canadian Earth System Model (CanESM2) under RCP 2.6 (optimistic) and RCP 8.5 (pessimistic) scenarios. The CanESM2 predictors for statistical downscaling are available for free through http://climate-scenarios.canada.ca/?page = pred-canesm2. There are 26 predictors and their names can be found in Table 2.

Table 2

NCEP predictors and their correlation with temperature and precipitation over Lake Urmia basin during 1976–2005

Code Predictor Correlation (partial correlation)
 
Super predictors
 
Temperature Precipitation Temperature Precipitation 
tempgl Mean temperature at 2 m 0.98 (0.98) 0.16 YES NO 
mslpgl Mean sea-level pressure 0.7 (0.37) 0.10 YES NO 
p1_fgl 1000hPa Wind speed 1 0.11 0.12 NO NO 
p1_thgl 1000hPa Wind direction 0.28 0.06 NO NO 
p1_ugl 1000hPa Zonal wind component 0.28 0.21 NO NO 
p1_vgl 1000hPa Meridional wind component 0.35 0.01 NO NO 
p1_zgl 1000hPa Relative vorticity of wind 0.14 0.01 NO NO 
p1_zhgl 1000hPa Divergence 0.14 0.20 NO NO 
p5_fgl 500hPa Wind speed 0.20 0.22 NO NO 
p5_ugl 500hPa Zonal wind component 0.17 0.10 NO NO 
p5_vgl 500hPa Meridional wind component 0.28 0.37 (0.27) NO YES 
p5_zgl 500hPa Relative vorticity of wind 0.34 0.14 NO NO 
p500gl 500hPa Geopotential height 0.9 (0.27) 0.27 YES NO 
p5thgl 500hPa Wind direction 0.15 0.14 NO NO 
p5zhgl 500hPa Divergence 0.15 0.11 NO NO 
p8_fgl 850hPa Wind speed 0.08 0.19 NO NO 
p8_ugl 850hPa Zonal wind component 0.39 0.01 NO NO 
p8_vgl 850hPa Meridional wind component 0.12 0.31 (0.17) NO YES 
p8_zgl 850hPa Relative vorticity of wind 0.41 0.03 NO NO 
p850gl 850hPa Geopotential height 0.15 0.24 (0.15) NO YES 
p8thgl 850hPa Wind direction 0.31 0.00 NO NO 
p8zhgl 850hPa Divergence 0.31 0.22 NO YES 
prcpgl Total precipitation 0.25 0.69 (0.69) NO YES 
s500gl Specific humidity at 500hPa 0.48 0.20 NO NO 
s850gl Specific humidity at 850hPa 0.83 0.00 NO NO 
shumgl Surface specific humidity 0.88 (0.27) 0.04 YES NO 
Code Predictor Correlation (partial correlation)
 
Super predictors
 
Temperature Precipitation Temperature Precipitation 
tempgl Mean temperature at 2 m 0.98 (0.98) 0.16 YES NO 
mslpgl Mean sea-level pressure 0.7 (0.37) 0.10 YES NO 
p1_fgl 1000hPa Wind speed 1 0.11 0.12 NO NO 
p1_thgl 1000hPa Wind direction 0.28 0.06 NO NO 
p1_ugl 1000hPa Zonal wind component 0.28 0.21 NO NO 
p1_vgl 1000hPa Meridional wind component 0.35 0.01 NO NO 
p1_zgl 1000hPa Relative vorticity of wind 0.14 0.01 NO NO 
p1_zhgl 1000hPa Divergence 0.14 0.20 NO NO 
p5_fgl 500hPa Wind speed 0.20 0.22 NO NO 
p5_ugl 500hPa Zonal wind component 0.17 0.10 NO NO 
p5_vgl 500hPa Meridional wind component 0.28 0.37 (0.27) NO YES 
p5_zgl 500hPa Relative vorticity of wind 0.34 0.14 NO NO 
p500gl 500hPa Geopotential height 0.9 (0.27) 0.27 YES NO 
p5thgl 500hPa Wind direction 0.15 0.14 NO NO 
p5zhgl 500hPa Divergence 0.15 0.11 NO NO 
p8_fgl 850hPa Wind speed 0.08 0.19 NO NO 
p8_ugl 850hPa Zonal wind component 0.39 0.01 NO NO 
p8_vgl 850hPa Meridional wind component 0.12 0.31 (0.17) NO YES 
p8_zgl 850hPa Relative vorticity of wind 0.41 0.03 NO NO 
p850gl 850hPa Geopotential height 0.15 0.24 (0.15) NO YES 
p8thgl 850hPa Wind direction 0.31 0.00 NO NO 
p8zhgl 850hPa Divergence 0.31 0.22 NO YES 
prcpgl Total precipitation 0.25 0.69 (0.69) NO YES 
s500gl Specific humidity at 500hPa 0.48 0.20 NO NO 
s850gl Specific humidity at 850hPa 0.83 0.00 NO NO 
shumgl Surface specific humidity 0.88 (0.27) 0.04 YES NO 

Downscaling

Due to the coarse resolution of general climate models, their outputs need to be downscaled to a local scale. In this study, statistical downscaling was used based on the CanESM2 predictors and support vector regression (SVR; Vapnik (1995)) model (details of the SVR model are discussed in Ahmadebrahimpour et al. (2018)). To this end, the following steps were undertaken:

  • the selection of super predictors based on the partial correlation coefficient between observed predictand (daily temperature and precipitation) and observed predictors (National Centres for Environmental Prediction (NCEP) predictors);

  • training and evaluation of the regression model (SVR) based on selected predictors and given predictand; and

  • projections of future temperature and precipitation based on the calibrated SVR model and future predictors obtained from the CanESM2 model.

More details about statistical downscaling can be found in Hashmi et al. (2009) and Rahimi et al. (2018).

RESULTS AND DISCUSSION

Table 2 shows the correlation between daily precipitation and temperature with NCEP predictors. Based on Table 2, four and five predictors were selected as super predictors for projections of future temperature and precipitation. For temperature, the ‘mean temperature at 2 m’ and for precipitation, the ‘total precipitation’ were superior to all other 25 predictors. Figure 1 illustrates the monthly average of observed and simulated precipitation and temperature for the baseline period. Based on this figure, it was concluded that the calibrated model simulated well both precipitation and temperature. The NSE performance criteria on a monthly scale between observed and simulated precipitation vary in the range of [0.61, 0.83] and for temperature vary in the range of [0.88, 0.98]. The calibrated model was used for precipitation and temperature downscaling in future periods.

Figure 1

Observed and simulated precipitation and temperature based on NCEP predictors and SVR model over Lake Urmia basin during 1976–2005.

Figure 1

Observed and simulated precipitation and temperature based on NCEP predictors and SVR model over Lake Urmia basin during 1976–2005.

In the first stage after downscaling, the changes in precipitation and temperature across the Lake Urmia basin were assessed under the various periods and scenarios. Figure 2 shows the monthly pattern of precipitation and temperature for the baseline and future periods. Under the RCP 2.6 projection, the annual precipitation was found to increase in the periods 2041–2070 and 2071–2100. For example, during 2041–2070, which corresponds to the maximum increase in precipitation, the annual precipitation is estimated to reach 397 mm, i.e. an increase of 1.1%. However, under the RCP 8.5 projection, the annual precipitation would decrease by 4.28%, 6.93% and 8.64% in the periods 2011–2040, 2041–2070 and 2071–2100, respectively. Under both scenarios in all future periods, the temperature is estimated to increase across the study basin. The average annual temperature under the RCP 2.6 projection would be increased by 0.81 °C, 1.09 °C, and 1.36 °C for the periods 2011–2040, 2041–2070, and 2071–2100 respectively while under the RCP 8.5 projection it would increase by 1.06 °C, 1.89 °C, and 2.46 °C respectively.

Figure 2

Monthly mean of precipitation and temperature for the baseline period and future periods.

Figure 2

Monthly mean of precipitation and temperature for the baseline period and future periods.

The next stage was to assess drought events using the SPI and SPEI. Figure 3 shows the number of different drought categories (frequency) detected in the baseline and future periods using the SPI. The SPI1 and SPI3 results for short-term droughts indicate that the frequency of normal conditions (WD class) will decrease in the future in most cases. Under the RCP 2.6 projection, the frequency of moderate and severe droughts is almost constant while under the RCP 8.5 projection it will increase. The number of exceptional droughts (D4) under both scenarios will increase in the future periods. For example, in the case of the SPI3, seven exceptional drought events were identified during the baseline period while 12, eight, and nine exceptional drought events were identified under the RCP 2.6 projection and 11, 12, and 15 exceptional droughts were identified under the RCP 8.5 projection during the periods 2011–2040, 2041–2070, and 2071–2100, respectively. Generally, under mid- and long-term droughts (SPI6, SPI9, SPI12, and SPI24) the WD classes slightly decrease and the number of drought events are almost unchanged based on the RCP 2.6 projection. In contrast and especially during 2071–2100, the frequency of drought events in all categories increases under the RCP 8.5 projection.

Figure 3

The frequency of different drought categories for the baseline period and future periods based on the SPI.

Figure 3

The frequency of different drought categories for the baseline period and future periods based on the SPI.

Figure 4 shows the frequency of drought events based on the SPEI. As shown in Figure 4, the WD frequency in all future periods and under both scenarios decreases significantly particularly under the RCP 8.5 projection and as a consequence the frequency of drought events will increase in all almost categories and scenarios. For example, based on SPEI9, 14 extreme droughts (D3) were identified in the baseline period with 19, 15, and 17 extreme droughts identified under the RCP 2.6 projection and 21, 22 and 28 extreme droughts identified under the RCP 8.5 projection, during the periods of 2011–2040, 2041–2070, and 2071–2100 respectively. The present results highlight that the SPEI identifies more droughts than the SPI. While under the RCP2.6 projection the precipitation increases the increase in temperature and the resulting increase in PET will reduce the available water. In the case of the RCP 8.5 projection the decreases in precipitation and increases in PET compound the impact of climate change and will increase the number of drought events and in particular the occurrence of exceptional droughts. These results are in agreement with Shadkam et al. (2016) who assessed inflows into Lake Urmia under the RCP 2.6 projection and the RCP 8.5 projection. Shadkam et al. (2016) estimated that the inflow into Lake Urmia will decrease about 10% under the RCP 2.6 projection and decrease about 27% under the RCP 8.5 projection. This highlights that, in arid and semi-arid regions, an assessment of precipitation alone may underestimate the impact of climate change on droughts.

Figure 4

The frequency of different drought categories for the baseline period and future periods based on the SPEI.

Figure 4

The frequency of different drought categories for the baseline period and future periods based on the SPEI.

Tables 3 and 4 show the longest drought event in each period and scenario and correspond with average and maximum index value, and drought severity. Severity has been defined as summation of SPI or SPEI values during the detected drought event. The SPI results in Table 3 do not show a specific pattern. For example, the longest drought duration based on the SPI12 in the baseline period is equal to 52 months while in the period 2071–2100 under the RCP 8.5 projection it is equal to 44 months. In the case of the SPI24 the longest durations are equal to 89 and 93 months, respectively. However, the results of the SPEI show a significant pattern. Based on the SPEI, the longest drought duration will be unchanged or increase in severity in all future periods and under both scenarios (except for the SPEI1) it will be increased significantly. The worst conditions occur based on the SPEI12 under the RCP 8.5 projection for the period 2071–2100 where the longest drought duration will increase from 41 months (baseline) to 88 months and the average SPEI will decrease from −1.45 (D2 class) to −2.09 (D4 class). These results are in agreement with Stagge et al. (2017) which showed that the change of PET can have significant effect on droughts. The result confirms the remarks of Lee et al. (2017), who assessed the SPI and SPEI under future climate conditions and also showed that the increase in drought severity and duration based on the SPEI is greater than under the SPI.

Table 3

Characteristics of the longest drought event during each period based on the SPI

 SPI1
 
SPI3
 
SPI6
 
 Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity 
Baseline (1976–2005) −1.16 −2.29 −8.09 11 −1.14 −1.98 −12.52 30 −1.40 −2.21 −41.94 
RCP 2.6 (2011–2040) −0.94 −1.17 −2.81 −0.75 −1.84 −6.74 14 −1.34 −2.04 −18.75 
RCP 2.6 (2041–2070) −0.63 −1.11 −2.53 −0.47 −1.81 −3.79 16 −1.31 −1.99 −20.94 
RCP 2.6 (2071–2100) −0.76 −1.12 −2.28 10 −0.66 −1.85 −6.60 18 −1.33 −2.01 −23.89 
RCP 8.5 (2011–2040) −0.97 −1.15 −4.84 12 −0.86 −1.84 −10.35 14 −1.37 −2.07 −19.14 
RCP 8.5 (2041–2070) −1.15 −1.21 −3.45 12 −1.00 −1.88 −11.99 23 −1.40 −2.16 −32.09 
RCP 8.5 (2071–2100) −1.23 −2.02 −3.69 11 −1.03 −2.58 −11.37 22 −1.57 −2.65 −34.52 
 SPI9
 
SPI12
 
SPI24
 
 Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity 
Baseline (1976–2005) 35 −1.47 −2.39 −51.29 52 −1.40 −2.05 −72.71 89 −1.34 −2.17 −116.79 
RCP 2.6 (2011–2040) 36 −1.40 −2.30 −50.45 38 −1.47 −2.41 −55.88 47 −1.58 −2.59 −74.05 
RCP 2.6 (2041–2070) 38 −1.37 −2.25 −51.88 41 −1.44 −2.36 −59.07 53 −1.54 −2.53 −81.43 
RCP 2.6 (2071–2100) 37 −1.38 −2.27 −51.24 42 −1.46 −2.39 −61.20 54 −1.56 −2.57 −84.35 
RCP 8.5 (2011–2040) 35 −1.44 −2.35 −50.40 43 −1.51 −2.45 −64.80 48 −1.62 −2.66 −77.66 
RCP 8.5 (2041–2070) 39 −1.52 −2.45 −59.31 42 −1.60 −2.58 −67.41 65 −1.61 −2.83 −104.95 
RCP 8.5 (2071–2100) 37 −1.65 −2.71 −61.12 44 −1.76 −2.82 −77.63 93 −1.72 −3.09 −159.90 
 SPI1
 
SPI3
 
SPI6
 
 Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity 
Baseline (1976–2005) −1.16 −2.29 −8.09 11 −1.14 −1.98 −12.52 30 −1.40 −2.21 −41.94 
RCP 2.6 (2011–2040) −0.94 −1.17 −2.81 −0.75 −1.84 −6.74 14 −1.34 −2.04 −18.75 
RCP 2.6 (2041–2070) −0.63 −1.11 −2.53 −0.47 −1.81 −3.79 16 −1.31 −1.99 −20.94 
RCP 2.6 (2071–2100) −0.76 −1.12 −2.28 10 −0.66 −1.85 −6.60 18 −1.33 −2.01 −23.89 
RCP 8.5 (2011–2040) −0.97 −1.15 −4.84 12 −0.86 −1.84 −10.35 14 −1.37 −2.07 −19.14 
RCP 8.5 (2041–2070) −1.15 −1.21 −3.45 12 −1.00 −1.88 −11.99 23 −1.40 −2.16 −32.09 
RCP 8.5 (2071–2100) −1.23 −2.02 −3.69 11 −1.03 −2.58 −11.37 22 −1.57 −2.65 −34.52 
 SPI9
 
SPI12
 
SPI24
 
 Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity 
Baseline (1976–2005) 35 −1.47 −2.39 −51.29 52 −1.40 −2.05 −72.71 89 −1.34 −2.17 −116.79 
RCP 2.6 (2011–2040) 36 −1.40 −2.30 −50.45 38 −1.47 −2.41 −55.88 47 −1.58 −2.59 −74.05 
RCP 2.6 (2041–2070) 38 −1.37 −2.25 −51.88 41 −1.44 −2.36 −59.07 53 −1.54 −2.53 −81.43 
RCP 2.6 (2071–2100) 37 −1.38 −2.27 −51.24 42 −1.46 −2.39 −61.20 54 −1.56 −2.57 −84.35 
RCP 8.5 (2011–2040) 35 −1.44 −2.35 −50.40 43 −1.51 −2.45 −64.80 48 −1.62 −2.66 −77.66 
RCP 8.5 (2041–2070) 39 −1.52 −2.45 −59.31 42 −1.60 −2.58 −67.41 65 −1.61 −2.83 −104.95 
RCP 8.5 (2071–2100) 37 −1.65 −2.71 −61.12 44 −1.76 −2.82 −77.63 93 −1.72 −3.09 −159.90 
Table 4

Characteristics of the longest drought event during each period based on the SPEI

 SPEI1
 
SPEI3
 
SPEI6
 
 Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity 
Baseline (1976–2005) −1.08 −1.43 −5.38 −1.27 −1.96 −11.39 29 −1.31 −2.13 −37.85 
RCP 2.6 (2011–2040) −0.82 −1.12 −4.92 11 −1.23 −1.85 −13.54 30 −1.34 −2.11 −34.88 
RCP 2.6 (2041–2070) −0.88 −1.21 −7.02 12 −1.23 −1.83 −14.71 31 −1.36 −2.11 −33.89 
RCP 2.6 (2071–2100) −1.06 −1.42 −6.36 13 −1.24 −1.86 −16.14 30 −1.36 −2.13 −40.94 
RCP 8.5 (2011–2040) −0.85 −1.16 −4.26 12 −1.24 −1.85 −14.86 41 −1.35 −2.17 −55.46 
RCP 8.5 (2041–2070) −1.27 −1.68 −8.87 23 −1.62 −2.53 −37.22 42 −1.59 −2.52 −66.82 
RCP 8.5 (2071–2100) −1.55 −2.08 −13.94 21 −2.10 −3.34 −44.06 54 −1.82 −3.13 −98.07 
 SPEI9
 
SPEI12
 
SPEI24
 
 Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity 
Baseline (1976–2005) 40 −1.43 −2.32 −57.23 41 −1.45 −2.28 −59.64 63 −1.14 −2.41 −71.64 
RCP 2.6 (2011–2040) 41 −1.49 −2.36 −61.15 42 −1.54 −2.44 −63.22 65 −1.43 −2.57 −92.86 
RCP 2.6 (2041–2070) 43 −1.51 −2.38 −65.05 41 −1.57 −2.49 −61.39 75 −1.46 −2.62 −109.39 
RCP 2.6 (2071–2100) 40 −1.52 −2.38 −60.61 56 −1.47 −2.51 −82.49 67 −1.47 −2.63 −98.49 
RCP 8.5 (2011–2040) 40 −1.53 −2.41 −61.32 52 −1.48 −2.52 −76.93 72 −1.49 −2.65 −107.19 
RCP 8.5 (2041–2070) 67 −1.59 −2.73 −106.25 76 −1.67 −2.98 −126.59 85 −1.82 −3.11 −154.61 
RCP 8.5 (2071–2100) 87 −1.96 −3.39 −170.67 88 −2.09 −3.56 −183.86 96 −2.25 −3.74 −216.32 
 SPEI1
 
SPEI3
 
SPEI6
 
 Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity 
Baseline (1976–2005) −1.08 −1.43 −5.38 −1.27 −1.96 −11.39 29 −1.31 −2.13 −37.85 
RCP 2.6 (2011–2040) −0.82 −1.12 −4.92 11 −1.23 −1.85 −13.54 30 −1.34 −2.11 −34.88 
RCP 2.6 (2041–2070) −0.88 −1.21 −7.02 12 −1.23 −1.83 −14.71 31 −1.36 −2.11 −33.89 
RCP 2.6 (2071–2100) −1.06 −1.42 −6.36 13 −1.24 −1.86 −16.14 30 −1.36 −2.13 −40.94 
RCP 8.5 (2011–2040) −0.85 −1.16 −4.26 12 −1.24 −1.85 −14.86 41 −1.35 −2.17 −55.46 
RCP 8.5 (2041–2070) −1.27 −1.68 −8.87 23 −1.62 −2.53 −37.22 42 −1.59 −2.52 −66.82 
RCP 8.5 (2071–2100) −1.55 −2.08 −13.94 21 −2.10 −3.34 −44.06 54 −1.82 −3.13 −98.07 
 SPEI9
 
SPEI12
 
SPEI24
 
 Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity Duration (months) Mean Peak Severity 
Baseline (1976–2005) 40 −1.43 −2.32 −57.23 41 −1.45 −2.28 −59.64 63 −1.14 −2.41 −71.64 
RCP 2.6 (2011–2040) 41 −1.49 −2.36 −61.15 42 −1.54 −2.44 −63.22 65 −1.43 −2.57 −92.86 
RCP 2.6 (2041–2070) 43 −1.51 −2.38 −65.05 41 −1.57 −2.49 −61.39 75 −1.46 −2.62 −109.39 
RCP 2.6 (2071–2100) 40 −1.52 −2.38 −60.61 56 −1.47 −2.51 −82.49 67 −1.47 −2.63 −98.49 
RCP 8.5 (2011–2040) 40 −1.53 −2.41 −61.32 52 −1.48 −2.52 −76.93 72 −1.49 −2.65 −107.19 
RCP 8.5 (2041–2070) 67 −1.59 −2.73 −106.25 76 −1.67 −2.98 −126.59 85 −1.82 −3.11 −154.61 
RCP 8.5 (2071–2100) 87 −1.96 −3.39 −170.67 88 −2.09 −3.56 −183.86 96 −2.25 −3.74 −216.32 

It should be noted that climate-change studies are always associated with uncertainties. These uncertainties can emerge in different parts of calculations. The GCM structure and different scenarios can be considered as the main source of calculation uncertainties. So, the results of climate-change studies may be too coarse for small-scale planning. The results only showed the potential for water resources in a given region to decrease or increase over time. The uncertainty regarding the decreases and increases in precipitation and temperature which are estimated based on GCMs under different RCPs is high. However, the results can indicate the future trend of available water resources. In the current study, the results indicate that the available water in the basin is decreasing. This has major consequences for the viability of any new water development projects in the basin, which is a very important result for decision-makers.

CONCLUSIONS

This study was conducted to assess the impacts of climate change on drought based on the findings of the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Drought events for 2011–2040, 2041–2070, and 2071–2100 were analyzed based on the SPI and SPEI and were compared with the adopted baseline period (1976–2005). The SPI and SPEI were calculated using the precipitation and temperatures obtained from CanESM2 under RCP 2.6 and RCP 8.5 as optimistic and pessimistic scenarios respectively. The results of SPI analyses revealed that under RCP 2.6 the frequency of droughts is almost constant while under RCP 8.5 drought frequency increases especially in the period 2071–2100. The calculated SEPI under both scenarios and during all future periods predicts that the frequency and duration of droughts will increase significantly. Generally, there is a notable difference between the SPI and SPEI. This difference is related to the input to each index. SPI is solely based on precipitation while the SPEI accounts for both precipitation and potential evapotranspiration (PET). Under global warming and changing climate, the significant role of PET was highlighted. It was concluded that the SPEI outperformed the SPI for drought studies under a changing climate.

While the uncertainty regarding the decreases and increases in precipitation and temperature which are estimated based on GCMs under different RCPs is high, the results of this study can indicate the future trend of available water resources in the Lake Urmia basin. The results of this study indicate that the available water in the basin is decreasing. It is concluded that this has major consequences for the viability of any new water development projects in the basin, which is a very important result for decision-makers.

REFERENCES

REFERENCES
Allen
M. R.
,
Barros
V. R.
,
Broome
J.
,
Cramer
W.
,
Christ
R.
,
Church
J. A.
,
Clarke
L.
,
Dahe
Q.
,
Dasgupta
P.
,
Dubash
N. K.
&
Edenhofer
O.
2014
IPCC Fifth Assessment Climate Change 2014: Synthesis Report
. IPCC, Geneva, Switzerland.
Anderson
B.
,
Manouseli
D.
&
Nagarajan
M.
2018
Estimating scenarios for domestic water demand under drought conditions in England and wales
.
Water Science and Technology: Water Supply
18
(
6
),
2100
2107
.
Azarnivand
A.
,
Hashemi-Madani
F. S.
&
Banihabib
M. E.
2015
Extended fuzzy analytic hierarchy process approach in water and environmental management (case study: Lake Urmia Basin, Iran)
.
Environmental Earth Sciences
73
(
1
),
13
26
.
Dastorani
M. T.
,
Massah Bavani
A. R.
,
Poormohammadi
S.
&
Rahimian
M. H.
2011
Assessment of potential climate change impacts on drought indicators (case study: Yazd station, Central Iran)
.
Desert
16
(
2
),
159
167
.
Guttman
N. B.
1999
Accepting the standardized precipitation index: a calculation algorithm
.
JAWRA Journal of the American Water Resources Association
35
(
2
),
311
322
.
Hashmi
M. Z.
,
Shamseldin
A. Y.
&
Melville
B. W.
2009
Statistical downscaling of precipitation: state-of-the-art and application of bayesian multi-model approach for uncertainty assessment
.
Hydrology and Earth System Sciences Discussions
6
(
5
),
6535
6579
.
Hosseini-Moghari
S. M.
&
Araghinejad
S.
2015
Monthly and seasonal drought forecasting using statistical neural networks
.
Environmental Earth Sciences
74
(
1
),
397
412
.
Hosseini-Moghari
S.-M.
,
Araghinejad
S.
&
Azarnivand
A.
2017
Drought forecasting using data-driven methods and an evolutionary algorithm
.
Modeling Earth Systems and Environment
3
(
4
),
1675
1689
.
Hosseinizadeh
A.
,
SeyedKaboli
H.
,
Zareie
H.
,
Akhondali
A.
&
Farjad
B.
2015
Impact of climate change on the severity, duration, and frequency of drought in a semi-arid agricultural basin
.
Geoenvironmental Disasters
2
(
1
),
23
.
Kallis
G.
2008
Droughts
.
Annual Review of Environment and Resources
33
,
85
118
.
Liu
L.
,
Hong
Y.
,
Looper
J.
,
Riley
R.
,
Yong
B.
,
Zhang
Z.
,
Hocker
J.
&
Shafer
M.
2013
Climatological drought analyses and projection using SPI and PDSI: case study of the Arkansas Red River Basin
.
Journal of Hydrologic Engineering
18
(
7
),
809
816
.
McKee
T. B.
,
Doesken
N. J.
&
Kleist
J.
1993
The relationship of drought frequency and duration to time scales
. In:
Proceedings of the 8th Conference on Applied Climatology
.
American Meteorological Society, Boston
,
MA, USA
, pp.
179
183
.
Mishra
A. K.
,
Desai
V. R.
&
Singh
V. P.
2007
Drought forecasting using a hybrid stochastic and neural network model
.
Journal of Hydrologic Engineering
12
(
6
),
626
638
.
Safavi
M.
,
Asareh
A.
&
Zaynivand
N.
2015
Evaluating estimation methods of potential evaporation-transpiration in Deylam city
.
WALIA Journal
31
,
80
84
.
Shadkam
S.
,
Ludwig
F.
,
van Vliet
M. T. H.
,
Pastor
A.
&
Kabat
P.
2016
Preserving the world second largest hypersaline lake under future irrigation and climate change
.
Science of the Total Environment
559
,
317
325
.
Staben
N.
,
Nahrstedt
A.
&
Merkel
W.
2015
Securing safe drinking water supply under climate change conditions
.
Water Science and Technology: Water Supply
15
(
6
),
1334
1342
.
Stagge
J. H.
,
Tallaksen
L. M.
,
Gudmundsson
L.
,
Van Loon
A. F.
&
Stahl
K.
2015
Candidate distributions for climatological drought indices (SPI and SPEI)
.
International Journal of Climatology
35
(
13
),
4027
4040
.
Stagge
J. H.
,
Kingston
D. G.
,
Tallaksen
L. M.
&
Hannah
D. M.
2017
Observed drought indices show increasing divergence across Europe
.
Scientific Reports
7
,
14045
.
Tsakiris
G.
,
Pangalou
D.
&
Vangelis
H.
2007
Regional drought assessment based on the reconnaissance drought index (RDI)
.
Water Resources Management
21
(
5
),
821
833
.
Vapnik
V.
1995
The Nature of Statistical Learning Theory
.
Springer Verlag
,
New York
,
USA
.
Vicente-Serrano
S. M.
,
Beguería
S.
&
López-Moreno
J. I.
2010
A multiscalar drought index sensitive to global warming: the standardized precipitation evapotranspiration index
.
Journal of Climate
23
(
7
),
1696
1718
.
Weedon
G. P.
,
Balsamo
G.
,
Bellouin
N.
,
Gomes
S.
,
Best
M. J.
&
Viterbo
P.
2014
The WFDEI meteorological forcing data set: WATCH forcing data methodology applied to ERA-Interim reanalysis data
.
Water Resources Research
50
(
9
),
7505
7514
.
Wilhite
D. A.
,
Svoboda
M. D.
&
Hayes
M. J.
2007
Understanding the complex impacts of drought: a key to enhancing drought mitigation and preparedness
.
Water Resources Management
21
(
5
),
763
774
.
Yuan
Z.
,
Xu
J.
,
Chen
J.
,
Huo
J.
,
Yu
Y.
,
Locher
P.
&
Xu
B.
2017
Drought assessment and projection under climate change: a case study in the middle and lower Jinsha River Basin
.
Advances in Meteorology
2017
,
5757238
.