Climate change impacts on the shrinkage of Lake Urmia

Climate change is among the most complex challenges that humans are faced with currently and in the future. Since the early 1980s, the remarkable warming of the Earth due to the rise of CO₂ and other greenhouse gases caused by human activities has been the main factor of climate change in the recent century. This will expose most areas of the Earth to temperature rise, changes in the precipitation regime, and the rise of CO₂ concentrations. Numerous studies have so far been conducted to evaluate the climate in future decades. Iran is among the countries affected by the adverse consequences of climate change (Pittock 1983; Cooley 1990; Higgins & Scheiter 2012; Adib et al. 2021a; Gholami et al. 2022). According to several studies, the most significant temperature rise and precipitation reduction have occurred in the western areas of Iran and this trend will continue in the future (e.g., Najafi & Moazami 2016; Ahmadi et al. 2018; Alizadeh-Choobari & Najafi 2018; Moghim 2018; Nazeri Tahroudi et al. 2019; Abbasian et al. 2021; Doulabian et al. 2021). Lake Urmia in northwestern Iran has also witnessed severe area reduction in recent decades. The shrinking process in the lake began in the mid-2000s and is now exposed to the danger of full drying. The evaluation of satellite images demonstrates that the lake lost 88% of its area in 2015 (Figure 1) (AghaKouchak et al. 2015; Abbasi et al. 2019; Biazar et al. 2020). The drying of the lake has been attributed to several reasons, including drought, construction of a highway on the lake, increasing use of water resources of the lake basin, and low precipitation and snowfall. Until 2012, more than 200 dams on river basins of the lake were ready for operation or in the last design stages. In October 2015, the water level of Lake Urmia was announced to be 1,270.04 m, showing a reduction by 40 cm compared to that of October 2014. In the case of drying of the lake, the regional temperate weather will change into hot weather along with salt storms, changing the regional environment. In addition to salt, many contaminations containing toxic heavy metals used in the industry and agricultural pesticides have penetrated the surface and subsurface water resources related to the lake. In the case of drying of the lake, most poisonous materials will become airborne, posing the risks of respiratory diseases for people and threatening the ecosystem of the region. In 2014, a national workgroup, entitled Lake Urmia Survival, began to address the crisis and performed measures whose positive effects emerged in the past 2 years. On August 14, 2020, however, the lake level was announced to be 1,271.41 m above mean sea level, with an equivalent area and a volume of 2,900 km² and 3.75 billion m³, respectively. Compared with the beginning of the Lake Urmia survival plan on August 14, 2014, it has risen by 126 cm, equivalent to 1,375 km² in surface and 2.9 billion m³ in.volume. Nevertheless, this condition is not stable, and the risk of crisis is still threatening the lake. In this regard, the results of several studies carried out on this issue are described in the following (Tabnak 2020).

Delju et al. (2013) studied the climate change of Lake Urmia in the period 1964–2005. Their results showed a decrease of 9.2% in the precipitation and an increase of 0.8 °C in the average maximum temperature. Seasonal changes were visible, especially in winter and spring. According to the Palmer Drought Index, severe drought covers Lake Urmia every 5 years, so that these droughts have become more severe and prolonged at the end of the study period. Alizadeh-Choobari et al. (2016) investigated the effect of climate change and human factors on the rapid decline of the water level of Lake Urmia. They calculated the trend of changes in the lake water level, temperature, precipitation, and evaporation of six stations (Urmia, Tabriz, Khoy, Mahabad, Maragheh, and Sahand), and the flow of three rivers of Mahabad, Nazlu, and Siminehrood, which are discharged into the lake. According to their results, the average temperature of six stations increased by 0.18 °C per decade from 1951 to 2013, which was 0.58 °C per decade in the Urmia station. The precipitation in the basin decreased by approximately 9 mm per decade, which is 17.2 mm per decade for the Urmia station. In addition, evaporation increased by 6.2 mm per decade and the lake water level decreased by more than 6 m from 1995 to 2009. Given these meteorological and socioeconomic droughts, they recommended implementing water resource management policies to improve the condition of the lake. Shadkam et al. (2016a, 2016b) suggested a rapid reduction of 40% of water used in agriculture to protect the Lake Urmia. They estimated the amount of inflow into the lake under RCP2.6 and RCP8.5 scenarios using the Variable Infiltration Capacity (VIC) model. According to their results, if climate change proceeds under the optimistic scenario of RCP2.6, the lake situation can be improved by reducing irrigation water allocated to agriculture. However, if climate change conditions continue under the pessimistic RCP8.5 scenario, the reduction of agricultural irrigation water will not be effective in saving the lake, which requires more drastic measures. Ahmadaali et al. (2018) investigated the effect of climate change on the rapid decline of the water level of Lake Urmia. Their results showed that the highest values of indices of environmental sustainability and agricultural sustainability were related to the scenario of combining the crop pattern change with improving the total irrigation efficiency under the B1 emission scenario. Alizade Govarchin Ghale et al. (2018) analyzed the water budget of Lake Urmia and the severity of drought in the basin from 1985 to 2010 and presented a new hypothesis to quantify the human and climatic impacts on the volume of Lake Urmia. They stated that the impact of human activities and the misuse of groundwater resources in the region were more effective than that of climate change in reducing the water level of the lake. Since 1998, a major decrease has been observed in the water level of Lake Urmia. Human impacts and climatic factors affect the drying of Lake Urmia by 80 and 20%, respectively. They presented two proposals to solve this crisis. The first step to recover Lake Urmia is to evaluate surface water management and operation policies of dams and groundwater resources, and the second step is to evaluate and classify the regional agricultural products in terms of water consumption to instruct residents the optimal irrigation methods. Tabari & Willems (2018) investigated the impact of the seasonally varying footprint of climate change on precipitation within the Middle East. They indicated that climate change signals and associated uncertainties over the middle East region remarkably vary with seasons. Sanikhani et al. (2018) investigated the impact of climate change on the runoff to Lake Urmia in the two basins of Aji-Chay and Mahabad-Chay. They used the LARS-WG model, including fourth assessment report (AR4) general circulation models (GCMs), to downscale minimum and maximum precipitation and solar radiation and estimated the future runoff using these variables by the gene expression programming (GEP) model over three 20-year periods (2011–2030, 2046–2065, and 2080–2099). Their results showed decreases in the peak flow of 50 and 55.9% in Aji-Chay and Mahabad basins, respectively, from 2080 to 2099 compared to the base period. Abbasi et al. (2019) monitored and forecasted meteorological drought in Lake Urmia from 1951 to 2009. They used the Standardized Precipitation Evapotranspiration Index (SPEI) drought index at 1, 3, 6, 9, 12, 24, and 48-month time scales to monitor drought and provided a relationship for drought prediction of the next step using the GEP model. They also assessed the accuracy of the GEP model predictions and improved overestimate and underestimate error values. According to their results, severe meteorological droughts occurred in two periods from 1959 to 1967 and 1998 to 2009, and the recent droughts have had a devastating effect on Lake Urmia. Khazaei et al. (2019) analyzed the climate change on Lake Urmia's runoff accounts for the relationships between atmospheric climate change and hydrological and vegetation cover changes within the landscape. Results show that precipitation, temperature, and soil moisture changes cannot explain the sharp decline in lake water levels since 2000. Schulz et al. (2020) have recently studied the components of the water balance in Lake Urmia and its changes in the last 50 years by measuring the components. Their results also showed the effect of climate change on the water level of Lake Urmia. They acknowledged that reduction of the water allocated to agriculture could lead to the stabilization of the lake, and the continuation of this uncontrolled harvest could lead to its complete collapse. Abbasian et al. (2021) studied the climate change of Lake Urmia within the period 2060–2080. Their results showed that a decrease of 8% within the precipitation and a rise of 4 °C within the average maximum temperature in 2060–2080 supported RCP8.5. Their results indicate that projected changes within the number of dry months in 2060–2080 under RCP4.5 and RCP8.5 relative to the historic period vary between 2.4–7.3 and 4.5–13.2%, respectively.

According to the mentioned studies, Lake Urmia in western Iran is among the most important ecosystems whose conditions influence the living of millions of people in the region; besides, the lake is the habitat for hundreds of species. Hence, investigating the drought condition in the region is of great importance. Given the studies mentioned in the literature review, no comprehensive research has been conducted on the regional meteorological drought in the future period affected by climate change models. The purpose of this study is to investigate the meteorological droughts (Standardized Precipitation Index (SPI) and SPEI) and its relationship with the shrinkage of Lake Urmia in the baseline period (1951–2005) and the future period (2025–2079).

The novel aspects of this research are the selection of superior climatic models with the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) technique, the combination of those models with the K-nearest neighbors (KNN) method, and the quantification of the uncertainty values, and also discovered a relationship between meteorological drought and lake levels that has not been previously investigated.

Lake Urmia is the largest inland lake in Iran and the sixth largest endorheic lake in the world (Hammer 1986; Dudgeon 2020). The water of this lake is very salty and it is fed from Zarrineh, Simineh, Gader, Baranduz, Shahrchai, Nazlu, and Zola rivers. The basin area of Lake Urmia is 51,876 km², which is 3% of the total area of Iran. The plains of Tabriz, Urmia, Maragheh, Mahabad, Miandoab, Naqadeh, Salmas, Piranshahr, Azarshahr, and Oshnavieh are located in this basin. Therefore, it is one of the most valuable agricultural and livestock centers in Iran. At the Urmia station, the average of minimum temperature is 5.26 °C, the average of maximum temperature is 17.7 °C, and the average rainfall is 365 mm. Moreover, it is also classified as Dsa based on Köppen-Geiger's climate classification (Lotfirad et al. 2022).

This study used daily minimum temperature, maximum temperature, and precipitation data collected by the Urmia Synoptic Station, located west of Lake Urmia, from 1951 to 2005 (Figure 2). These data were obtained from the Iran Meteorological Organization (IRIMO 2019). Figure 3 shows the main steps of the study.

GCMs simulate the Earth's climate with mathematical equations that describe atmospheric, oceanic, and biological processes as well as interactions and feedback. The output of CMIP5 GCMs with 0.5° spatial size was used in this paper. Data were downloaded from http://gdo-dcp.ucllnl.org.

The specifications of the GCMs used here are listed in Table 1. Historical data of precipitation and minimum and maximum temperatures of the GCM cell corresponding to the Urmia synoptic station were compared in the baseline period (1951–2005).

In the equations above, X_obs and X_cal are the observational and calculated variables, respectively, is the mean value of the observational variable, and R is the linear correlation coefficient between the observational and calculated variables. In S_taylor criteria, k values in the temperature and precipitation were considered 4 and 2, respectively (Zamani & Berndtsson 2019; Farajpanah et al. 2020). The ideal value of RSR and MAPE indices is 0 and that of NS, R and S_taylor indices is 1.

Hwang & Yoon (1981) provided a technique for ranking the performance based on similarity to the ideal solution. This method includes the shortest distance from the ideal solution and the farthest distance from the non-ideal solution. TOPSIS has been used in many climate changes and hydrometeorological studies (e.g., Zamani & Berndtsson 2019; Farajpanah et al. 2020; Adib et al. 2021c; Esmaeili-Gisavandani et al. 2022). In this study, the performance criteria were calculated and the TOPSIS method was used to select GCMs with more consistency with the observational data in the baseline period. In this paper, the top 10 GCMs were selected in simulating precipitation and minimum and maximum temperatures of the baseline period, and these GCMs were used to project the future period. In the TOPSIS method, the Shannon entropy method was used for weighting (Lin 1991). The steps in TOPSIS are as follows:

Step 7: Ranking preference order according to the descending order of ⁠, that is the best rank and is the worst rank.

These stochastic methods are used to generate daily climatic variables, such as precipitation, minimum and maximum temperatures, humidity, etc., which are in accordance with the averages, monthly, or annual values. Weather generators produce long series and reduce the uncertainty of climate fluctuations (Semenov et al. 1998; Lotfirad et al. 2019). In this paper, the LARS-WG model was used as one of the most well-known weather generators. This model is based on semi-empirical distribution functions that are able to simulate wet and dry periods and declare the accuracy of this data generation by providing Kolmogorov–Smirnov (KS), t, and F tests. In this program, Fourier series are used to estimate temperature. Minimum and maximum daily temperatures are modeled as random processes with daily average and deviations that are dependent on the wet or dry condition of the day (Semenov et al. 1998; Semenov & Stratonovitch 2010; Lotfirad et al. 2021). A third-order Fourier series is used to simulate the average and standard deviation of seasonal temperature. The production of data by the LARS-WG includes calibration, evaluation, and creation of meteorological data (Nikakhtar et al. 2020). In this model, changes in minimum and maximum temperatures and precipitation under RCP4.5 and RCP8.5 scenarios are introduced as a scenario file to the LARS-WG model and daily data are generated for the future period (2025–2079), considering the 55-year observational data of the baseline period (1951–2005).

The variable D, which is the difference between the precipitation and PET, follows the log-logistic probabilistic distribution function; hence, the log-logistic distribution was used to calculate the SPEI.

SPI and SPEI uses probability distribution functions, while other meteorological indices such as Percent of Normal Index (PN), China-Z index (ZSI), Deciles index (DI), China-Z index (CZI), Effective drought index (EDI), and Modified CZI (MCZI) do not use these functions. Therefore, SPI and SPEI can show features of meteorological drought well. These two indices are conventional indices for studying meteorological drought in different regions of the world. The SPI considers only the precipitation variable, but the SPEI, in addition to precipitation, also uses the temperature variable, which is an important factor in the climate. These indices have been used in similar studies such as Bazrafshan et al. (2017) and Abbasi et al. (2019) in Iran.

To evaluate the meteorological drought situation of Lake Urmia for the 55-year baseline and future periods, the SPEI and the SPI were calculated on time scales of 3, 6, 12, 24, and 48 months. In short-term time scales, both indices have large fluctuations, which decrease with an increasing temporal scale (Abbasi et al. 2019). The run theory was used to determine the severity, duration, and peak of each drought event (Mishra & Singh 2010). Drought indices are time series whose values represent the intensity of drought. In this study, a zero threshold value was used to detect drought events, extracting the characteristics of each drought event, such as severity, duration, and peak. Drought indices are time series whose values represent the intensity of drought. In this study, a zero threshold value was used to detect drought events, extracting the characteristics of each drought event, such as severity, duration, and peak. There are various methods for determining the drought threshold, including −1 determining the threshold, but selecting zero is one of the most usable and simplest methods for determining the drought threshold. Moyé et al. (1988), Şen (1991), Mishra & Singh (2010), and Sharafati et al. (2020) applied this drought threshold. Also, values below zero SPI and SPEI indicate that drought and compatibility of the zero threshold with this issue can simplify the diagnosis of drought.

At first, the historical data of 23 GCMs were compared with the observed data of the Urmia station in the period 1951–2005. The performance of GCMs in the baseline period was evaluated by the five efficiency criteria of MAPE, RSR, NS, R, and S_taylor (Tables 2–4). Each of the GCMs in the simulation of the variables used in the baseline period was ranked using the TOPSIS method. Table 5 shows the weights of the performance criteria used in the TOPSIS, which were obtained based on the Shannon entropy method (Chen 2019; Adib et al. 2021b). The top 10 GCMs were chosen for projecting the precipitation, minimum and maximum temperatures, as well as their corresponding values in the future period. Table 6 lists the 10 chosen GCMs with higher ranks based on the TOPSIS ranking. In the baseline period, these GCMs had better conformity to the observed data. Therefore, it was assumed that the GCMs would have better performance in the future period than in the other ones.

The top 10 GCMs chosen using the TOPSIS were weighted by the KNN method. Greater weights were allocated to the GCMs with better performance in the baseline period. This could reduce the uncertainty in the projections of the future period (Ashofteh et al. 2015; Moghadam et al. 2019; Lotfirad et al. 2021). Table 7 demonstrates the weights determined for each GCM in each month. For instance, the GCMs FGOAL-g2 and IPSL-CM5A-MR had the best performance and, consequently, the greatest weights in predicting the precipitation in January and February, respectively. Tables 8 and 9 show the top 10 GCMs projecting minimum and maximum temperatures, respectively.

By combining the top 10 GCMs and forming a hybrid model, the coefficient of variation was determined. The coefficient of variation is a simple method to measure the reliability of a model, and the lower the coefficient of variation, the higher its reliability and, in fact, the lower its uncertainty. Table 10 shows that the hybrid model has the least uncertainty compared with the top 10 GCMs.

The LARS-WG model measures the goodness of fit of the semi-empirical distributions applied to daily time series of precipitation and minimum and maximum temperatures using the KS test. Lower KS statistics, along with p-values closer to one, indicate the performance accuracy of the LARS-WG model in fitting the experimental distribution. According to Table 11, the LARS-WG model has a very good performance.

Figure 4 demonstrates changes in the minimum and maximum temperatures, mean temperature and precipitation, along with the water balance (P-PET). Under both RCP4.5 and RCP8.5 scenarios, the temperature increases in all months of the future period. The precipitation increases in some months and decreases in the other ones, but the reduction percentage is more than the rising percentage. Under the RCP4.5 scenario, the largest precipitation reduction, equal to 13.8 mm, occurs in November, while the largest precipitation rise, equal to 4.6 mm, belongs to July. Under the RCP8.5 scenario, the largest precipitation decrease is in March (9.5 mm), while the largest precipitation increase is in July (3.3 mm). In the minimum temperature, the largest rises under the RCP4.5 and RCP8.5 scenarios are, respectively, 3.3 and 4.3 °C, occurring in January. In the mean temperature, the largest rise under the RCP4.5 scenario is 4.1 °C in January and August, and it is 3.1 °C in the same months under the RCP8.5 scenario. The amount of water balance (P-PET) under the RCP4.5 scenario has decreased in all months except January, with the largest decrease in August by more than 23 mm. Also, the amount of water balance (P-PET) under the RCP8.5 scenario has decreased in all months, with the largest decrease in August by more than 34 mm.

The SPI and SPEI were computed in 3, 6, 12, 24, and 48-month scales. Figures 5–9 demonstrate the SPEI variations in the Urmia station during the 55-year statistical baseline (1951–2005) and the future (2025–2079) periods. According to Figures 5–9, the SPEI time series has many fluctuations in short-term time scales, which decrease with the increase in the time scale (Abbasi et al. 2019). In other words, the occurrence number of droughts decreases, while their duration grows at longer time scales. According to Figure 9, which is based on SPEI-48, long droughts occurred in two periods of 1958–1968 and 1998–2005 in.the baseline period. In the future period under both RCP4.5 and RCP8.5 scenarios, they will occur in the periods of 2047–2059 and 2072–2078. Graphs of SPI changes in 3-, 6-, 9-, 12-, 24-, and 48-month scales are provided in the appendix.

The droughts that occurred in the baseline period and are projected for the future period are monitored in Table 12. The droughts become longer and more severe with increasing the time scale. In the baseline period, the most severe drought occurred at the end of the studied statistical period (1998–2005). The peak and severity of this drought on the 48-month scale are 1.97 and −125.6, respectively. According to climatic projections, the droughts are intensified in the future period under both scenarios due to the significant temperature rise and precipitation variations, which are (mostly) decreasing. The peak and severity of this drought in the 48-month scale under RCP4.5 scenario are 1.97 and −125.6, respectively. Also, the peak and severity of drought in the 48-month scale under RCP8.5 scenario are –2.10 and −77, respectively. According to Table 12, the number of droughts increases, while their duration decreases in the future period compared to the baseline period.

Figure 10 indicates the correlation between the SPI and the SPEI in the 3-, 6-, 12-, 24-, and 48-month scales, which means that the SPI, which only deals with the precipitation data, properly demonstrates the drought conditions. The interesting point is the remarkable rise in the correlation between the SPI and the SPEI in the future period caused by the reduced precipitation, the increased temperature, and consequently, the grown evapotranspiration, which plays a fundamental role in the SPEI. Figure 11 shows the variations in the water level of Lake Urmia from January 1966 to July 2020. Based on the Mann–Kendall trend test, this time series witnesses a very decreasing trend, so that its Z-MK and Sen Slope values are estimated at −17.63 and −0.009 at 95% confidence level.

Figure 12 illustrates the correlation between the SPEI and the SPI and the monthly water level of Lake Urmia. According to the figure, long-term droughts (e.g., the 48-month scale) has shown more correlations with the changes in the water level of the lake, since the correlation between the SPEI and the lake water level is more than that that between the SPI and the water level.

Figure 13 shows the linear regression between SPI-48 and SPEI-48 versus lake water level; Figure 13 also explicitly shows that SPEI drought, which includes both precipitation and temperature, has a robust relationship with the water level than the SPI, which considers only precipitation. This relationship is significant in the long term (Figures 12 and 13).

Lake Urmia is among the most important water bodies of Western Asia. The extraordinary salinity of the lake water has formed a special ecosystem. The stress applied to the lake has disrupted the living conditions of Artemia, one of the valuable living organisms of this ecosystem, which feeds from the lake algae. Therefore, the reduced number of Artemia increases the algae, changing the lake into a swamp.

Over the past two decades, the water level and the area of this lake have decreased dramatically. This stress has been imposed on the lake by two main reasons. The first factor is human activities, which includes increasing use of water supply resources of the lake for the development of industry and agriculture, improper cultivation patterns, and the construction of a highway dividing the lake into northern and southern parts. The second factor, which is less obvious than the first one, is climate change and the resulting droughts. In order to confront the dryness crisis of Lake Urmia, the national work group for the survival of Lake Urmia was founded in 2014. The implemented plans by this group and very good precipitation in the past 2 years in the basin of the Lake Urmia have slightly alleviated the crisis in the lake; however, this improvement is not stable. Given the two main stressing factors, this paper deals with climate change and the occurrence of meteorological droughts in the baseline and future periods.

By considering the daily statistics of the 55-year period (1951–2005) of the synoptic station in Urmia, climate change in the future period was projected under both RCP4.5 and RCP8.5 scenarios using 23 CMIP5 GCMs. Top 10 GCMs simulating minimum and maximum temperatures and precipitation of the baseline period were chosen using the TOPSIS method and used for the future period. In order to reduce the uncertainty in the projections, the KNN method was used to weight the chosen GCMs. Changes in the precipitation and minimum and maximum temperatures were introduced to the LARS-WG model to generate daily data for the 55-year future period (2025–2079). The use of weather generators resulted in considering dry and wet periods in the projection, which is also effective in reducing the projection uncertainties. According to the results, all seasons experience precipitation rises, while spring and autumn experience reduced precipitation. The largest rise in the minimum temperature (3.9 °C) in winter is under the RCP8.5 scenario, while the largest increase in the maximum temperature (4.1 °C) in summer occurs under the RCP8.5 scenario. The largest precipitation reduction (27.8%) occurs in autumn. Therefore, the temperature rise and precipitation reduction in the Urmia station from 1951 to 2013 will continue in the future, as mentioned by Alizadeh-Choobari et al. (2016).

Also, the amount of long-term annual water balance (P-PET) under RCP4.5 and RCP8.5 scenarios decreased compared to the baseline period by 50 and 68%, respectively. In general, the results obtained from this study on climate change and its impact on increasing temperature and decreasing rainfall are consistent with the results of studies (Delju et al. 2013; Alizadeh-Choobari et al. 2016; Shadkam et al. 2016a, 2016b; Ahmadalipour et al. 2017; Sanikhani et al. 2018; Tabari & Willems 2018; Schulz et al. 2020).

Afterward, the drought indices of SPI and SPEI were calculated for the baseline and future periods based on the SPEI considering such properties as severity, duration, peak, and monitoring of droughts. Compared to the baseline period, the duration of drought occurrences generally decreases, while their occurrence number, severity, and peak increase in the future period. As a result, the droughts that occurred under the RCP8.5 scenario have higher severity, duration, and peak than those under the RCP4.5 scenario. Furthermore, there is a higher correlation between the SPEI and the water level of Lake Urmia than the SPI; a regression equation for both indices with the water level was presented. With respect to the mentioned points, Lake Urmia will undergo more stress from the climate change perspective. Therefore, management measures should be taken to reduce drought damages and protect the environment of this ecosystem. The national work group of the Lake Urmia survival has proposed some plans for stabilization of the current condition and its improvement in the future years. These include leading 600 MCMs of the water of the Zab and Kani dams to the lake and operating large urban sewage treatment plants in the cities of Tabriz and Urmia, resulting in the addition of almost 250 MCM water to the Lake Urmia.

According to the results obtained in this research, climate change and the occurrence of meteorological droughts will result in the general reduction of runoffs into Lake Urmia and the occurrence of flash floods in summers. Hence, it is suggested to execute watershed management and aquifer management operations in the natural resources of the basin to reduce the damages caused by climate change and human activities in the basin of Lake Urmia. On the other hand, a considerable temperature rise in the region will increase the evapotranspiration and crop water requirement. It is recommendable that climate studies be conducted using a combination of GCMs. Also, the old, outdated irrigation systems should be replaced with smart new ones. Moreover, the cultivation pattern should be modified, and water accounting should be employed in the basin of the lake.

The authors have no conflicts of interest to declare that are relevant to the content of this manuscript.

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