GCMs simulation-based assessment for the response of the Mediterranean Gaza coastal aquifer to climate-induced changes

Arid and semi-arid areas, home to around 2000 million people or roughly 30% of the world's population who consume about 1,500,000 million cubic meters of fully tapped groundwater, are experiencing catastrophic water shortages and severe climate aridity (Burek et al. 2013; Dogrul et al. 2016; Anderson 2017; Hussain et al. 2019; Al-Najjar et al. 2020). The coastal Mediterranean aquifers are unique and the only water resource used for water supply services in the Middle East and North Africa where the threats of climate disturbances and overuse practices led to crucial deficiencies in terms of water and food security where the groundwater surveillance systems refer to a dangerous drop in the groundwater table to levels below the mean sea level, which, in turn, result in excessive seawater invasion and prolonged salinization (Ye et al. 2015; Dogrul et al. 2016; Zekri et al. 2017; Otkin et al. 2018; Schiermeier 2018; Gopalakrishnan et al. 2019; Hussain et al. 2019; Al-Najjar et al. 2021a, 2021b). As a result, new nontraditional water resources provided through seawater desalination and wastewater treatment industries are considerably identified as promising practical options to alleviate the continuous deterioration in the groundwater balance and close up the gap in the water budget (Al-Najjar et al. 2021a). Globally, seawater desalination technologies provide 32,500 million cubic meters of freshwater per year, or about 0.7% of global water needs. Wastewater treatment and reuse are widely used for agricultural irrigation and groundwater recharge in Australia, China, the United States (California and Florida), and the Arabian Gulf, with an annual treatment capacity of 280, 14,000, 1750, and 2800 million cubic meters, respectively. Specifically, setting a modeling basis for analyzing climate and human-induced effects on coastal aquifer underground water resources is critical for an effective groundwater management system that aims to adopt efficient policies and mitigation measures to remediate any hazardous effects on groundwater sustainable practices (Ye et al. 2015; Karimi et al. 2019). Climate models, which are primarily represented by global circulation models (GCMs), are spatially downscaled to a regional scale based on scenarios of greenhouse emissions by using local data acquired from ground meteorological stations to predict the future tendency of the climate. The fifth phase project of the coupled model intercomparison project (CIMP5) was developed based on four climate scenarios demonstrated by representative concentration pathways (RCPs) for the greenhouse gas concentrations trajectory adopted by the intergovernmental panel on climate change (IPCC), which are RCP 2.6, 4.5, 6.0, and 8.5 (IPCC 2019). Overcoming the shortcomings and complexity of groundwater numerical models, data-driven models are merit simulators for incorporating the various aspects that influence a groundwater system dynamic response (Butler et al. 2013; Yeh & Chang 2013; Singh 2014; Djurovic et al. 2015; Gladden & Park 2016; Thangarajan & Singh 2016). A stochastic serial dependent analysis is used to simulate the climate and groundwater time series to anticipate the future trend based on the inherent tendency of historical observations to design a representative statistical data model (Adamowski et al. 2012; Sahoo & Jha 2013; Kumbuyo et al. 2014; Mirzavand & Ghazavi 2015; Mogaji et al. 2015; Patle et al. 2015; Yan & Ma 2016; Zhou et al. 2017; Al-Najjar et al. 2020, 2021b). The intelligent artificial data-driven algorithmic approaches, which can describe the complex relationship between multi-input and output parameters, have been used to accurately detect the nonlinear behavior between climatic parameters and the response of the groundwater level (Emamgholizadeh et al. 2014; Afan et al. 2015; Chang et al. 2016; Sun et al. 2016; Ebrahimi & Rajaee 2017). The stochastic autoregressive integrated moving average (ARIMA) is a famous statistical model established based on the autoregression concept, which allows for the extraction of information from time-series data through autocorrelation analysis functions (Box & Jenkins 1970; Koopmans 1974; Kashyap & Rao 1976; Box et al. 2008; Chatfield 2008; Kumbuyo et al. 2014; Al-Najjar et al. 2020, 2021b). Numerous studies in the field of water and climate modeling have addressed the stochastic ARIMA models (Bazrafshan et al. 2015; Mirzavand & Ghazavi 2015; Djerbouai & Souag-Gamane 2016; Khorasani et al. 2016; Myronidis et al. 2018; Takafuji et al. 2018; Sakizadeh et al. 2019,Al-Najjar et al. 2020, 2021a, 2021b). Artificial neural networks (ANNs) are data-driven networks with neurons that receive, process, and transmit information from the input layer to the output layer. The logistic sigmoid function network of multi-layer feed-forward perceptron (MLP) with a single hidden layer is normally advised as a model for predicting water and climate interactions by numerous studies (Maier et al. 2010; Mohanty et al. 2010; Adamowski & Chan 2011; Chen et al. 2011; Jalalkamali et al. 2011; Nourani et al. 2011; Trichakis et al. 2011; Yoon et al. 2011; Rakhshandehroo et al. 2012; Taormina et al. 2012; Moosavi et al. 2013; Sahoo & Jha 2013; Shirmohammadi et al. 2013; Emamgholizadeh et al. 2014; He et al. 2014; Tapoglou et al. 2014; Ying et al. 2014; Juan et al. 2015; Khaki et al. 2015; Khalil et al. 2015; Mohanty et al. 2015; Nourani et al. 2015; Yang et al. 2015; Gong et al. 2016; Nourani & Mousavi 2016; Sun et al. 2016; Yoon et al. 2016; Barzegar et al. 2017; Ebrahimi & Rajaee 2017; Wen et al. 2017; Ghose et al. 2018; Kouziokas et al. 2018; Yu et al. 2018; Guzman et al. 2019; Lee et al. 2019; Tang et al. 2019; Al-Najjar et al. 2021a, 2021b). In the Gaza Strip of Palestine, due to excessive water pumping and limited precipitation as a result of rising temperatures, the Gaza coastal aquifer, which is the only available water resource for water supply in the Gaza Strip, is anticipated to be severely strained throughout the mid-future. Due to the ambiguity of climate variations, this study used a simulation-based approach to assess potential changes in climate conditions and future impacts of climate and anthropogenic practices on Gaza coastal aquifer groundwater table fluctuations, as well as the impact of alternative water resources, such as seawater desalination and wastewater treatment, on groundwater resource recovery.

The Gaza Strip, as shown in Figure 1, is a 365 km² coastal strip of land located on the Mediterranean Sea's southeast coast, stretching 42 km and varying in width from 6 to 12 km. With a population of more than two million people, the Gaza Strip is considered one of the world's most densely populated areas (Palestinian Central Bureau of Statistics (PCBS) 2020). The Gaza Strip's water supply system is semi-collapsed with the only source of water being severely contaminated and deteriorating as a result of unsustainable pumping of roughly 200 million cubic meters per year (PWA 2011, 2012, 2014, 2015; UN 2012). Drought has become more common in the Gaza Strip as a result of climate change's effects, which have harmed the coastal aquifer's vulnerability and the viability of agricultural activities where the drought investigation studies refer to an increase in the incidence of drought occurrence from about 20% in the 1970s to more than 80% in the last 10 years (Al-Najjar et al. 2020). The Palestinian Water Authority (PWA) has established water management strategies and mitigation measures, as shown in Figure 2, to deal with the Gaza Strip's water shortage in response to the effects of climate change and anthropogenic concerns. The proposed intervention management plans include the construction of three seawater desalination plants to provide 13 million cubic meters of water per year, as well as the construction of a large-scale seawater desalination plant to provide 110–120 million cubic meters per year, to reduce water abstraction from the coastal aquifer by about 35–60% over the next 20 years (PWA 2011, 2014; UfM 2011; Abualtayef et al. 2016).

In parallel to this, the PWA approved a long-term strategic plan to build three large-scale wastewater treatment plants to support the agricultural sector with about 75 million cubic meters of treated water for irrigation and to cope with the drought impact, as well as to reduce groundwater abstraction from the coastal aquifer by about half, from 180 to about 70 million cubic meters per year by 2032 (PWA 2011, 2013, 2014; Abuatayef et al. 2020).

Historical total monthly based raw data for the climate parameter of precipitation (P_o) were obtained from nine meteorological stations located inside the geographical extent of the Gaza Strip from January 1974 to December 2020. Furthermore, the available monthly time-series data for the climate parameters of minimum temperature (T_min), average temperature (T_avg), maximum temperature (T_max), sunshine (S_o), evaporation (E_o), and humidity (H_o) are regularly measured only at the Remal meteorological station and were gathered from January 1974 to December 2006. For groundwater condition analysis, the time series for the municipal monthly groundwater consumption (C_o) from January 1997 to December 2011 were retrieved from the metered records of the monthly groundwater pumping. In addition, available measurements for the monthly fluctuating level of the mean sea level (MSL) groundwater table response (R) were obtained from 10 piezometric groundwater wells for the period from January 1974 to December 2018. Each class of time-series data was statistically processed and temporally standardized over the period from January 1974 to December 2040 using ARIMA stochastic models. The ensemble climate model of CIMP5 was regionally downscaled for the parameters of monthly rainfall and average temperature under three greenhouse gas emission scenarios of RCPs of 2.60, 4.50, and 8.50. The artificial one hidden layer feed-forward backpropagation MLP network of 20 neurons was structured to combine the impact of the climate and human factors on the groundwater level of the Mediterranean Gaza coastal aquifer. The new quantities of the alternative water resources deriving from seawater desalination and wastewater treatment were simulated by reducing the groundwater consumption by the amount of the supplementary water supply.

The available historical records for the total monthly precipitation (P_o) are presented in the heat maps, as shown in Figure 3. According to the analysis, the Gaza Strip is a low-rainfall region, with the maximum total monthly rainfall of less than 330 mm throughout the rainy season, which spans from November to March.

The future projection of the precipitation in the Gaza Strip up to the year 2098 is presented in the boxplot of Figure 4. The regional trend indicates a decreasing manner in the total rainfall from the north to the south of the Gaza Strip as well as the future projection refers to a decreasing temporal trend over 2006–2036, 2037–2067, and 2068–2098.

The geographical distribution for the variation in the total rainfall, as shown in Figure 5, reveals a decreasing trend in the total rainfall over the near and far future. Hence, the variation in the rainfall throughout 2006–2036 compared to the base period of 1975–2005 is within 0 to −5% for the RCP 2.60, while the variation is assigned to −5 to −15% and −15 to −25% for the RCP of 4.50 and 8.50, respectively.

The available historical records for the minimum temperature (T_min), average temperature (T_avg), and maximum temperature (T_max) were determined in the heat maps, as shown in Figure 6. The temperature in the area of the Gaza Strip fluctuates between about 7 °C in cold seasons and 35 °C in hot seasons.

The future downscaling of ensemble GCM of CIMP5 for the average temperature, as shown in Figure 7, indicates an increasing trend by about +1 °C within 2006–2036 above the average temperature of 20 °C recorded throughout 1975–2005. However, throughout 2037–2067, the incremental increase in temperature is expected to be +1 °C above the average temperature recorded during 1975–2005. However, by 2068–2098, the temperature increase is probable to reach +4 °C above the average temperature recorded during 1975–2005.

The stochastic ARIMA models for the climate parameters are depicted in Figure 8 in terms of calibration and validation and Table 1 for models' terms of seasonal and nonseasonal parameters. The ARIMA models show accurate simulation for the climate parameters with accuracy ranges between 74 and 99%.

To provide cross-validation for the results of ARIMA models, a comparison was performed throughout 2006–2036 for the parameters of precipitation and average temperature, as shown in Figure 9. For the parameter of precipitation, the deviation of the ARIMA outputs to the outputs of the RCP scenarios is about +7%. However, in the parameter of temperature, the deviation is assigned to +8.5%.

The extraction of groundwater from drilling wells throughout the Gaza Strip has a substantial influence on the water budget of the Gaza coastal aquifer and is the main cause of groundwater storage instability. This research recommends using the stochastic ARIMA model of (2,1,5) (2,1,1)₁₂ to forecast and back-cast the time series of groundwater extraction from the Gaza coastal aquifer at each Gaza Strip governorate to lengthen the time series and give a deeper understanding of groundwater hydraulic behavior and actions. As shown in Figure 10, the stochastic model predicts that groundwater utilization began in the Gaza Strip's northern governorates in 1986. The extraction of groundwater began in the Gaza governorate in the mid-1970s, while the southern governorates began using it around the end of the 1980s and the beginning of the 1990s. Recently, groundwater extraction has been observed in a rapid and widespread way, particularly in the southern governorates of the Gaza Strip.

The stochastic model offers an ideal simulation for the time series of groundwater extraction quantities, with a correlation coefficient (r) of above 90%. Groundwater extraction practice reveals an annual increasing tendency that varies by location by roughly 12–59%, with an overall average of 24%. The agricultural methods in the Gaza Strip's northern governorates encourage rapid groundwater extraction, which puts pressure on the Gaza coastal aquifer.

The dominant feature that determines the interplay between the various physical factors in nature is a nonlinear function. Due to the obvious complexity of the association between meteorological and hydrological indicators, numerous scientific approaches have been proposed and coupled to interpret the intrinsic relationship between these factors. Therefore, the artificial MLP approach, as shown in Figure 11, was exploited to model the response of the Gaza coastal aquifer to the climate and human-induced changes.

The network was developed by training the network on 540 combinations of the data for each time series of the 10 groundwater wells. In terms of performance testing, it shows a valid presentation of the observed groundwater level. In an overall manner, as shown in Figures 12 and 13, the stochastic models describe the relationship between the observed and the simulated data by a correlation coefficient (r) of 95–99%, RMSE of 0.1–0.22, and NSE of 0.93–0.99

The parameters of performance testing for the calibration and validation are illustrated in Table 2.

The Gaza Strip's southern region seems to be the most densely populated region of the area. The local wells produce groundwater at a volume of far more than 100 m³ per hour, which negatively affects the availability of the Gaza coastal aquifer in this area. The groundwater depletion cone started to form in the Gaza Strip in 1992, and it has been spreading northward ever since. The cone's diameter was less than 1 km in 1992, and it is anticipated to grow to 4–5 km by 2040. As a result, seawater intrusion is the most common phenomenon that has an impact on groundwater quality by lowering the chloride concentration to an acceptable level. The model predicts that obtained groundwater from the Gaza coastal aquifer will rise from 124 million cubic meters in 2020 to around 191 million cubic meters in 2040, with an annual increase rate of about 3%. The groundwater simulation findings, as shown in Figure 14, show that the groundwater-level drop is between −0.38 and −18.49 m below the MSL in 2020 and between −1.13 and −27.77 m below the MSL in 2040.

The simulation of the groundwater recovery as a result of the new water resource interventions is shown in Figure 15 for the southern region of the Gaza Strip, which exhibits critical deterioration in the coastal aquifer.

The simulation results show that using desalinated saltwater and treated wastewater instead of groundwater over-pumping operations results in a considerable 82–91% increase in the groundwater level.

The groundwater contained in the coastal aquifers is the prime resource for water supply in the Mediterranean region. The water supply of the Gaza Strip is in jeopardy due to the potential of extreme climate, as well as the increase in demand for water services as a result of exponential population growth. The future climate projection refers to an increase by +1 °C by the end of 2040 and by about +4 °C by the end of 2099 when compared to the historical temperature of 1975–2005. In addition, variation in the total monthly precipitation is expected to vary between −5% within 2006–2036 and −25% within 2068–2098. As a result, the Gaza coastal aquifer is projected to be under the critical conditions of overuse and imbalance where the simulation refers to a drop in the groundwater table reach to −28 m by the year 2040. The necessity for urgent intervention to sustain the groundwater resource in the Gaza Strip has directed enterprises toward using the nontraditional water resources produced from seawater desalination and wastewater treatment, especially in the southern governorates of the Gaza Strip where the groundwater resource is overexploited. The simulation outputs for the process of compensation for the deficit in the water supply through the non-conventional water resources showed an optimized recovery in the groundwater level where the drop in the groundwater level is expected to be less than −5 m below the MSL.

The authors declare no conflict of interest.

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