Projection of rainfall variability in Egypt by regional climate model simulations

During the 20th century, a large shift occurred in the mean global surface temperature (1.0 °C), accompanied by drastic changes in precipitation distributions, resulting in increased intensity and frequency of floods and droughts (Kharin et al. 2007; IPCC 2014; Rogelj et al. 2018), and affecting the availability of freshwater resources. Some of the changes in weather and climate extremes observed in the late 20th century are expected to continue (O'Neill et al. 2016; Grubler et al. 2018), due to the predicted increase in global temperature (1.3–5.1 °C by 2100) (Rogelj et al. 2018). Accordingly, there is an urgent need to study potential changes in precipitation in different regions of the world using available climate models.

Climate scenarios have been adopted to provide reasonable descriptions of how the future will evolve, covering a wide range of uncertainties about human contributions to climate change and the response of the earth system to human activities (Hulme et al. 1999). Globally, by the end of the 21st century, an increase in the intensity of precipitation extremes of about 6, 10, and >20% was projected under representative concentration pathways, namely RCP2.6, RCP4.5, and RCP8.5, respectively (Kharin et al. 2013). However, in its Fifth Assessment Report (AR5), the Intergovernmental Panel on Climate Change (IPCC) indicated that major regional deviations from the global-scale pattern in temperature and precipitation are expected (Christensen et al. 2013). Zhao et al. (2014) concluded that rainfall changes would vary across areas, with more precipitation over wet areas and less precipitation over dry areas. In the Sixth Assessment Report (AR6), IPCC (2022) stated that the impacts of climate change previously disclosed are more widespread and have greater consequences than expected.

Global climate models (GCMs) are typically used to reproduce the current climate and to predict the possible future changes of climate indices. However, their spatial resolutions are too coarse to assess the catchment and site-specific impacts of climate change (Bordoy & Burlando 2013; Laflamme et al. 2015; Potter et al. 2019). Therefore, some post-processing is necessary to improve upon these global-scale models, such as downscaling processes that address the scale difference between global models and the local scale in which impact studies are generally conducted. Two widely used techniques of downscaling are available in the literature: dynamic downscaling (DD) and statistical downscaling (SD). DD is based on dynamical formulations using the boundary conditions of GCMs to achieve higher spatial resolution of the climate model (Kanamaru & Kanamitsu 2006), which are known as regional climate models (RCMs). On the other hand, SD involves linking some large-scale variables of GCM or RCM (predictors) to other small-scale variables of the catchment or site scale (predictands) (Wilby et al. 1998). The main strengths of SD over DD are the low computational cost and the relative ease of application. Though, it is assumed that the statistical relationship between large and local processes remains constant in the future climate, which may not be true given the nonstationary nature of the evolving climate.

Generally, RCMs have been proved to provide more reliable results for regional impact studies of climate change relative to GCMs (Schmidli et al. 2006; De Cáceres et al. 2018). Hence, they have become very important tools for describing and predicting different hydrological processes such as precipitation. However, their applications in climate change studies are challenging due to inherently systematic errors. Biases occur not only with variables' values but also with their seasonal fluctuations (Teutschbein & Seibert 2012). For instance, RCMs tend to underestimate (overestimate) the frequency and intensity of heavy (light) rainfall events (Fowler et al. 2007; Berg et al. 2012). Thus, it is necessary to apply adjustment techniques (e.g., bias correction) to minimize projection errors caused by GCM/RCM outputs. The idea behind bias-correction methods (BCMs) is to correct the simulated data in accordance with the observed data during a specific historical period (Chen et al. 2013); then, applying the resulted correction to the future observations. Several methods of bias correction have been developed, such as linear scaling (LS) (Jacob et al. 2007), local intensity scaling (LOCI) (Schmidli et al. 2006), and quantile mapping (QM) (Fowler et al. 2007).

The application of bias correction may increase uncertainty in climate projection (Ahmed et al. 2018); thus, the performance of different BCMs in different regions has been evaluated in several studies. For RCM-simulated daily precipitation over Europe, the gamma distribution mapping proved to be efficient in correcting not only the mean but also the higher moments of the distribution (Piani et al. 2010). Over the Alps region (Austria), Themeßl et al. (2011) compared seven methods of bias correction in daily precipitation, and both QM and LOCI proved significant advantages over the traditional multiple linear regression. In the United Kingdom, Lafon et al. (2013) concluded that the empirical QM method is the most accurate compared to three other methods of bias correction (linear, non-linear, and gamma-based QM). In China, several studies have evaluated different BCMs. For instance, Fang et al. (2015) compared the performance of five BCMs for precipitation in downscaling RCMs and concluded that the power transformation (PT), QM, and LOCI methods were the best performing. Luo et al. (2018) compared the performance of seven BCMs and showed that both QM and daily bias-correction (DBC) methods are the best. In North America, some studies have evaluated the performance of BCMs using different RCMs (e.g., Chen et al. 2013) or GCMs (Hassanzadeh et al. 2019). Similar studies have been conducted in different regions such as Senegal (Sarr et al. 2015), Malaysia (Noor et al. 2019), and Syria (Homsi et al. 2019). Table 1 summarizes some recent studies on BCMs for rainfall.

In Egypt, there are only a few studies available in the literature where attempts have been made to investigate rainfall projection in the context of CC, and in particular to assess the skill of different BCMs in improving rainfall information generated by climate models. For example, Elmenoufy et al. (2017) applied the LS method to the RCM (RegCM4) outputs, driven by the European Community-Hamburg Atmospheric Model (ECHAM5), to assess the rainfall projection over the country. They used the monthly gridded data of the period (1976–2005) and predicted an increase in annual rainfall of 3 and 20 mm under RCP4.5 and RCP8.5 scenarios, respectively. In contrast, using the gauged data at eight stations, Mostafa et al. (2019) expected a significant decrease in the annual precipitation from 2010 to 2100 under the following two scenarios: RCP4.5 (−0.48 to −0.9 mm/year) and RCP8.5 (−0.95 to −1.40 mm/year) over the northern part of the country. Nashwan & Shahid (2019) evaluated the performance of 31 GCMs in simulating the spatial climate pattern in Egypt and anticipated a decrease in the rainfall quantity (−62%) over the northern coast and an increase (20%) over the south of the country. In a recent study, Nashwan et al. (2020) anticipated changes in rainfall using five GCMs with 0.1° spatial resolution over the Central North region of Egypt. The results concluded that precipitation would increase in the near future (2020–2059), but it will significantly decrease in the distant future (2060–2099). Furthermore, they evaluated the performance of three BCMs (LS, PT, and Empirical Quantile Mapping (EQM)) for the period (1983–2005), revealing that the LS method was the best.

The potential alterations of precipitation indices over Egypt during the current century due to the impacts of climate change will be investigated via selected RCM outputs, in order to identify the most appropriate strategies to mitigate and adapt to these changes. First, the rainfall simulation of the RCM ‘RCA4, MPI-ESM-LR’, developed by SMHI (Swedish Meteorological and Hydrological Institute), will be evaluated with reference to the observed daily rainfall data of 15 stations for the control period (1995–2005). Second, the performance of three common BCMs (LS, LOCI, and EQM) will be evaluated to determine which is the most effective method in correcting RCM simulations. Third, the best correction method will be applied to the outputs of three scenarios (RCP2.6, RCP4.5, and RCP8.5) to accurately project the rainfall over three future periods: near (2020–2030), mid (2060–2070), and far (2090–2100).

A common period of record is favored for climate change studies; thus, the reference period (1995–2014) was chosen to include as many stations as possible (15 gauged stations) distributed across the country to cover the diversity in our case study (Figure 1). Since rainfall over Egypt is concentrated in the north and rapidly decreases to Middle Egypt and then becomes very rare in Upper Egypt and western dessert (Gado & El-Agha 2020), the distribution of stations in the current study is as follows: nine stations in northern Egypt, five stations in the middle of the country, and one station in the south. The database was taken from a previous study (Gado et al. 2019), in which the daily rainfall data were examined for independence and homogeneity, to represent the historical observed climate. After building the database, four precipitation indices were established on an annual basis for each station. These indices include annual maximum precipitation (AMP), annual total precipitation (ATP), annual number of rainy days (ANRD), and simple daily intensity index (SDII). The statistical characteristics of the annual precipitation of the selected stations are recognized by high variability, as shown in Figure 1 and Table 2. The average of AMP ranges from 1 to 43 mm/day, ATP (2–174 mm/year), ANRD (1–34 day/year), and SDII (1–10 mm/day).

A large number of future climate projections, which can be obtained from GCM–RCM combinations, are available for climate change studies. Their simulations are coordinated in different frameworks, such as the Coordinated Regional Climate Downscaling Experiment (CORDEX), which provide the input simulations for climate change studies under different families of future scenarios. The rainfall outputs of the RCM ‘MPI-ESM-LR/RCA4’, developed by SMHI (Swedish Meteorological and Hydrological Institute) and available through the CORDEX project (https://esg-dn1.nsc.liu.se/projects/cordex), were exploited here to project the rainfall pattern over Egypt. This RCM, which has a resolution of 50 km × 50 km (0.44° × 0.44° grid spacing), was recommended for the Middle East and North Africa (MENA) region by CORDEX and some previous studies (e.g., Shalby et al. 2020; Gado et al. 2021). All rainfall data of the considered RCPs (RCP2.6, RCP4.5, and RCP8.5) were obtained for three future periods: near (2020–2030), mid (2060–2070), and far (2090–2100).

The following steps were adopted in this study to predict the spatial and temporal changes in rainfall in Egypt:

• Bias correction for the RCM rainfall data for the historical period (1995–2005), using the observed rainfall for 15 gauged stations as a reference dataset.

• Evaluating the performance of three BCMs for rainfall (LS, LOCI, and EQM), and selecting the best method based on different statistical metrics.

• Utilizing the best BCM in the projection of future rainfall for three future periods (near future: 2020–2030, mid future: 2060–2070, and far future: 2090–2100) compared to the historical period (2004–2014) under three RCPs (RCP2.6, RCP4.5, and RCP8.5).

Bias-correction procedures employ a transformation algorithm to adjust the outputs of climate models. The underlying idea is to develop a relationship between observed and simulated climate variables for the reference period, which is the basis for correcting both the reference and future periods. Three widely used bias-correction techniques have been applied to the rainfall data extracted from the RCM for the study area: LS, LOCI, and EQM. All BCMs are performed daily and are assumed to be stationary, i.e., the parameters of the correction for the current climatic conditions are valid for the future.

While the LS method only accounts for the bias of the mean, the LOCI method takes it one step forward and adjusts the mean and both wet-day frequencies and intensities of precipitation in three steps as follows (Turco et al. 2017):

• A specific threshold level for the RCM raw data (P_th,raw) is calibrated so that the number of days exceeding this level matches the number of wet days of the observed data (Schmidli et al. 2006). Then, the number of precipitation events is corrected for both control and future periods so that all days with precipitation below P_th,sim are redefined to dry days with 0 mm precipitation (Equation (2)).

  

  (2)
• A scaling factor is estimated based on the average long-term monthly wet-day intensity (Fang et al. 2015). Taking into count only wet days, the intensity scaling factor (S) is calculated by Equation (3):

  

  (3)
• The corrected precipitation is calculated by multiplying the scaling factor with the raw data, as shown in Equation (4) (Turco et al. 2017).

  

  (4)

By definition, adjusted precipitation has the same mean, wet-day frequency, and intensity as the observed time series. This method is efficient in improving the raw data containing many days with light rain, known as the drizzle error (Fang et al. 2015; Choudhary & Dimri 2018), as in dry areas such as our case study.

This study projects rainfall in Egypt in the context of climate change for three future periods: (2020–2030), (2060–2070), and (2090–2100), under three different scenarios (RCP2.6, RCP4.5, and RCP8.5). The RCM ‘RCA4, MPI-ESM-LR’, developed by SMHI was evaluated in this study against the observed rainfall data from 15 stations in Egypt (Figure 1). Three BCMs (LS, LOCI, and EQM) were evaluated in correcting biases in the RCM simulation. The performance of each BCM is evaluated based on its ability to detect the observed precipitation pattern. The bias-corrected precipitation values are compared with observations for the period (1995–2005) using frequency-based metrics. The frequency-based indices include the mean, the STD, the 99th percentile, and the probability of wet days (the number of rainy days divided by all days).

The characteristics of observed and simulated rainfall were compared before bias correction to assess the performance of the selected RCM at all stations (Table 3). The RCM simulation data underestimated the average daily rainfall at 10 stations and overestimated it at five stations. The differences between simulation and observation data are substantial at some stations (Al-Arish, Alexandria, Baltim, Hurghada, and Minya). The values of STD and the 99th percentile of the RCM-simulated data are much lower than those of the observation data at most stations, and the difference between them is significant. In contrast, the probabilities of wet days of the RCM simulation are close to those of the observation data of all stations (Table 3). These biases in the mean, the STD, and the 99th percentile may have negative effects on the results of hydrological models based on such simulation data.

The evaluation was performed by comparing the raw/bias-corrected output of the historical simulation of the RCM versus the reference observation datasets at each station. All of the studied BCMs have improved the RCM-simulated data, but there are substantial differences in their abilities to reproduce precipitation data with characteristics comparable to those of the observed data. The performance of these methods depends on the evaluation indices and the characteristics of the precipitation data (i.e., the station). For the frequency-based statistics, the EQM method achieved the best performance of all indices at most stations, the LOCI method came in second, and the LS method was the last in most cases, as shown in Table 3.

The AMP in Egypt is characterized by high variability (Table 2), with an average ranging from 1 mm/day in the south to 43 mm/day in the north. The AMP over Egypt was projected under the three RCPs for the three periods mentioned, as shown in Figures 4 and 5. The projected rainfall varies greatly depending on both the scenario applied and the future period. Overall, most stations are expected to experience significant increases in AMP compared to the observation period. Four stations (Marsa Matrooh, Wadi El-Natroon, Aswan, and Minya) could experience a decrease in AMP over the three future periods under the three scenarios. Four other stations (Alexandria, Al-Arish, Cairo, and Port Said El-Gamil) may suffer from a decrease in AMP during one or two future periods under some scenarios. For RCP2.6, the maximum increase in AMP is expected in Baharia to be around 745% during the first future period. On the other hand, the maximum decrease for this scenario could be about −70% during the first future period in Wadi El-Natroon and the second period in Minya. For RCP4.5, it is expected that the maximum increase will occur in Baharia (497%) during the first future period. The maximum decrease is projected to occur in Minya at −82%, during the last future period. For the worst-case scenario (RCP8.5), Siwa is expected to experience the maximum increase in AMP during the first future period by 604%. In contrast, Minya could experience the most decrease (−85%) during the last future period. For this scenario (RCP8.5), nationwide, the rate of increase in AMP is expected to decline by the end of the century (Figure 5). The western part of the country is the most affected by climate change, anticipating a significant increase for the three scenarios during the three periods, contrary to the regions of the Nile Valley, Delta, and the North Coast, which may witness a slight decrease in AMP (Figure 5).

The ATP in Egypt ranges from 2 mm/year in the south to 174 mm/year in the north (Figure 1 and Table 2). For each station, the ATP was analyzed to identify the changes with the observation period, and the results are presented in Figures 7 and 8. Like AMP, most stations are expected to experience significant increases in ATP. Four stations (Alexandria and Marsa Matrooh, Wadi El-Natroon, and Minya) could experience a decrease in ATP over all future periods in all scenarios. Additionally, five other stations (Baltim, Cairo, Ismailia, Port Said El-Gamil, and Salloum) could experience a decrease in ATP over one or two future periods under some scenarios. For RCP2.6, the maximum increase in ATP is expected in Baharia of about 610% during the first future period. On the other hand, the maximum decrease for this scenario may reach around −44% in Minya during the last future period. For RCP4.5, the maximum increase in ATP is expected to occur in Hurghada (419%) during the first future period. The maximum decrease is projected to occur in Minya by −61%, during the last future period. For the worst-case scenario (RCP8.5), Siwa is expected to experience the maximum increase in ATP of 860% during the first future period. On the other hand, Minya could experience the most decrease (−75%) throughout the third future period. For this scenario (RCP8.5), nationwide, the rate of increase in ATP is expected to decline by the end of the century (Figure 8). The western part of the country is the most affected by climate change, anticipating a large increase in ATP for all cases (Figure 7).

The ANRD in Egypt ranges from 1 day/year in the south to 34 day/year in the north (Table 2). The relative changes of the ANRD are presented in Figures 8 and 9. In general, most stations are expected to experience significant increases in ANRD. For RCP2.6, only one station (Port Said El-Gamil in the north) could witness a slight decrease of −5% in the far future. The maximum increase for this scenario is anticipated in Siwa (983%) in the far future. For RCP4.5, a decrease in ANRD is expected at only two stations (Baltim and El-Tor) during one or two future periods, with a maximum of −29% in El-Tor during the third period. On the other hand, it is expected that the maximum increase will occur in Hurghada by 565, 377, and 278%, during the three future periods, in ascending order. For RCP8.5, a decrease in ANRD is expected in five stations (Alexandria, Baltim, El-Tor, Marsa Matrooh, and Port Said El-Gamil), during the furthest period only, with a maximum of −50% in El-Tor. The Hurghada station is the most altered under this scenario, expecting an increase of 507, 480, and 234%, over the three future periods, in ascending order. Under this scenario, nationwide, the rate of increase in ANRD is anticipated to decline through the end of the century (Figure 9).

The SDII in Egypt ranges from 1 mm/day in the south to 10 mm/day in the north (Table 2). The SDII was projected under the three RCPs for the three periods mentioned, as shown in Figures 10 and 11. Overall, most stations are expected to experience a decrease in SDII relative to the observation period. For RCP2.6, three stations (Baharia, Cairo, and El-Tor) will witness increases in SDII during all periods, with a maximum of 268% in El-Tor during the middle period. On the other hand, the maximum decrease for this scenario is −77%, which is expected in Minya during the middle period. For RCP4.5, an increase in SDII is projected at five stations (Baharia, Cairo, El-Tor, Baltim, and Salloum) during the three future periods, with a maximum of 560% in El-Tor for the far future. The maximum decrease for this scenario is expected in Minya (−81%) for the far future. Regarding RCP8.5, five stations (Al-Arish, Baharia, Baltim, Cairo, and El-Tor) are expected to experience increases in SDII over the three future periods, with a maximum of 230% in El-Tor for the furthest period. In contrast, the maximum decrease for this scenario is expected to be in Minya (−81%) during the last period. El-Tor and Minya respectively recorded the most increases and decreases in SDII for all cases.

The climate in Egypt is expected to be highly volatile in the future due to global warming, which may severely affect water resources and increase residents' vulnerability to natural disasters. Therefore, it is important to study the future changes in the country's precipitation in the context of climate change. This study predicted the future rainfall in Egypt using the high-resolution RCM (RCA4, MPI-ESM-LR), which is believed to be a reliable mechanism for the study area; however, the simulations are not free from bias. The results revealed that raw RCM simulation data overestimated low rainfall amounts and underestimated extreme rainfalls. This matches the results of previous studies regarding the so-called drizzle effect of climate models (e.g., Fowler et al. 2007; Sun et al. 2011); hence, the hydrological models based on this raw data can lead to inaccurate results.

The applicability of BCMs has not been sufficiently investigated in semi-arid regions such as Egypt; therefore, it is imperative to find an appropriate BCM to assess the impact of climate change on water resources. In this study, three BCMs (LS, LOCI, and EQM) have been applied to adjust the RCM outputs as a requirement for analyzing the impact of climate change on the rainfall pattern in Egypt. Although the bias-corrected methods improved the raw RCM outputs, the residual biases are still large and highly non-linear. EQM performed better than both LOCI and LS for most of the evaluation indices, as it can produce statistical characteristics roughly comparable to those of the observed data in such a semi-arid region. These results are consistent with those of previous studies (e.g., Choudhary & Dimri 2018; Luo et al. 2018), but differ from a recent study by Nashwan et al. (2020), who found that the performance of LS is better than EQM for North Central Egypt. This discord may be due to different precipitation regimes in different regions.

The projected rainfall was corrected by the EQM method for three future periods: I (2020–2030), II (2060–2070), and III (2090–2100) under three different scenarios (RCP2.6, RCP4.5, and RCP8.5). Temporal and spatial changes in the projected rainfall were analyzed for four annual precipitation indices (AMP, ATP, ANRD, and SDII). Most stations are expected to see significant increases in AMP, ATP, and ANRD in most cases. For example, Hurghada will witness a remarkable increase in ATP in the range of (374–500%), (259–353%), and (147–453%) for the three future periods in ascending order, depending on the scenario applied (Table 4 and Figure 6). Furthermore, Hurghada, which already has a low ANRD (4 days, Table 2), will witness a significant increase in ANRD, depending on the scenario applied, in the range of (9–19 days), (13–15 days), and (16–21 days) for the three future periods in ascending order. Other stations will witness the same, including Aswan (5–9 days), Salloum (25–42 days), Siwa (5–26 days), and Wadi El-Natroon (13–23 days), depending on the applicable scenario and the future period (Table 4). In contrast, some stations will witness a significant decrease in the ATP and become highly vulnerable to climate change. For instance, Minya already has a very low ATP (13 mm, Table 2), which will become even smaller (I: 6–8 mm, II: 5–9 mm, and III: 3–7 mm, depending on the applicable scenario), as shown in Table 4. Marsa Matrooh is also vulnerable to climate change, as it will suffer from a significant decrease in ATP (I: 71–88 mm, II: 70–95 mm, and III: 52–82 mm), with its current average ATP being 128 mm. For SDII, El-Tor recorded the largest increase for all future periods in all scenarios (I: 83–182%, II: 107–268%, and III: 150–560%). This significant increase of SDII in El-Tor, which is located in southern Sinai, explains the frequent intense rainfall events that occurred recently in the area, as reported in some recent studies (e.g., El Afandi & Morsy 2020; Omran 2020).

The main results of the different precipitation indices for the three scenarios are as follows:

• For RCP2.6, the maximum increase in AMP and ATP will occur in Baharia of about 745 and 610%, respectively, during the first future period. On the other hand, the maximum decrease could be about −70% (−44%) for AMP (ATP) in Minya during the second (far) future period. The maximum increase in ANRD is expected in Siwa (983%) in the far future, while ANRD could decrease slightly about −5% in only one station (Port Said El-Gamil in the north) in the distant future. For SDII, the maximum increase (decrease) will occur in El-Tor (Minya) of about 268% (−77%) during the middle period.

• For RCP4.5, the maximum increase in AMP (ATP) will occur in Baharia (Hurghada) of about 497% (419%) during the first future period, and the maximum decrease will occur in Minya by −82% (−61%), during the far period. It is expected that the maximum increase in ANRD will occur in Hurghada (565%) in the first future period, while ANRD could decrease in two stations (Baltim and El-Tor) with a maximum of −29% in El-Tor during the last period. For SDII, the maximum increase (decrease) will happen in El-Tor (Minya) of about 560% (−81%) for the far period.

• For the worst-case scenario (RCP8.5), the maximum increase in AMP and ATP will occur in Siwa by 604 and 860%, respectively, during the first future period; and the maximum decrease will occur in Minya by −85% for AMP and −75% for ATP during the far period. It is expected that the maximum increase in ANRD will occur in Hurghada (507%) in the first future period, while ANRD could decrease in five stations with a maximum of −50% in El-Tor during the last period. Nationwide, the rate of increase in AMP, ATP, and ANRD is projected to decline through the end of the century. For SDII, the maximum increase (decrease) will happen in El-Tor (Minya) of about 230% (−81%) for the far period.

Only few studies have been conducted to assess the rainfall projection in Egypt in the context of climate change (e.g., Elmenoufy et al. 2017; Mostafa et al. 2019; Nashwan et al. 2020). In these studies, some defects related to projection uncertainty can be identified, such as the use of coarse gridded GCMs (Nashwan & Shahid 2019; Nashwan et al. 2020), limited gauged stations (Mostafa et al. 2019), or the application of the simple LS BCM (Elmenoufy et al. 2017). Furthermore, these studies relied on gridded observation data to obtain the correction factor applied to the rainfall patterns projected under two scenarios (RCP4.5 and RCP8.5). In this study, however, three BCMs were evaluated in order to find the best method to correct the rainfall indices in Egypt. Then, the best method (EQM) was employed to project precipitation under three RCPs. Additionally, gauged data were used to estimate the adjustment coefficients applied to the most appropriate RCM for the study area. Thus, a thorough comparison between the spatial pattern in the prediction resulting from this study and the previous studies is not feasible. Yet, the current results, regarding annual precipitation, are in agreement with those obtained from Elmenoufy et al. (2017), who predicted an increase in rainfall under RCP4.5 and RCP8.5 scenarios. Moreover, the precipitation pattern obtained spatially corresponds to that produced by Nashwan & Shahid (2019), revealing an increase in annual rainfall over Egypt, except for the northern coast which may experience a significant decrease (−62%). In the MENA region, some recent studies have concluded that precipitation is likely to decrease with disastrous consequences for ecosystems (e.g., Lelieveld et al. 2016; Bucchignani et al. 2018). For instance, Terink et al. (2013) concluded that the annual precipitation will decrease in most countries in MENA, with a maximum decrease of 20% in southern Egypt for the period (2020–2030).

This study assessed the spatial and temporal changes in future rainfall in Egypt in the context of climate change for three future periods: I (2020–2030), II (2060–2070), and III (2090–2100) under three different scenarios (RCP2.6, RCP4.5, and RCP8.5). The RCM ‘RCA4, MPI-ESM-LR’, developed by SMHI, was evaluated against observed rainfall data from 15 sites in Egypt. Three commonly used BCMs (LS, LOCI, and EQM) were evaluated to determine the most efficient method to use in correcting the daily precipitation simulated by the RCM.

The statistical characteristics of the raw RCM data are significantly different from those of the observed data. The results demonstrated improvements to some extent in the corrected estimates through the three methods, but at different levels. Overall, EQM shows the best results in terms of a general representation of the observed rainfall over Egypt, followed by LOCI and LS. Thus, the EQM method was selected to adjust future RCM simulations over the three future periods under the three scenarios. The temporal and spatial changes in the projected rainfall were analyzed in terms of four annual indices (AMP, ATP, ANRD, and SDII) in order to better understand future rainfall variations in Egypt. The alteration in rainfall indices varies greatly according to the scenario applied and the future period. Overall, AMP, ATP, and ANRD are expected to experience significant increases during the three future periods under the three considered RCPs at most sites, compared to the observation period. Countrywide, the rate of increase in these indices is expected to decline during the end of the century under the worst-case scenario (RCP8.5). In contrast, SDII was expected to recede significantly in most cases, as the expected increase in annual rainfall quantity would be accompanied by an extension of the rainfall period throughout the year (i.e., the number of rainy days).

The changes will not be distributed uniformly throughout the century, and their magnitudes will be regionally dependent. The western part of the country is the most affected by climate change, anticipating a significant increase in precipitation indices for all scenarios during all future periods; in contrast to the regions of the Nile Valley, Delta, and the North Coast, which may witness a decrease in precipitation. These results are consistent with those obtained from a recent paper (Gado et al. 2021) that analyzed future changes in temperature in Egypt during the late of this century and found that the Western Desert and Upper Egypt are the regions most affected by climate change, while the northern region of Egypt is the least affected. The potential increase in precipitation in the western part of the country can be vital to recharge the western aquifer that can be used for new agriculture projects. Moreover, this provides an opportunity to expand rainwater harvesting projects and increase rainfed agricultural lands. In contrast, the possible reduction on the Nile delta and the valley would increase the aridity in the region, which shall increase irrigation demand and limit groundwater recharge to the Nile aquifer. The spatial distribution also showed a decrease in rainfall in northern Egypt, which is currently dominated by most of the rainfall; and hence, this may lead to increased dependence on groundwater for irrigation in this area and increased seawater intrusion. Thus, the findings of the study can be helpful for adaptation and mitigation planning of the adverse impacts of climate change in the country.

The method proposed in this study will give some useful information regarding the appropriate bias-correction approach for RCMs in Egypt and provide more reliable precipitation projections, which have always been a great challenge due to the erratic behavior of precipitation in arid regions. The BCMs used were able to improve the precipitation quantity derived from RCM-simulated time series to mimic the observed rainfall patterns in Egypt, but did not significantly improve the coherence between the modeled and observed precipitation. Therefore, to simulate precipitation variability, a more advanced approach is recommended, such as the improved BCM of daily rainfall data using a sliding window technique developed by Smitha et al. (2018). Keeping in mind that the historical rainfall records used in the study consist of 15 stations; thus, it is recommended to consider more meteorological stations with a recent period of record, in order to forecast various rainfall indices in the country with high reliability. This study uses climate data from RCPs; however, new emission scenarios called the Shared Socioeconomic Pathways (SSPs) have been developed to provide a socioeconomic dimension to the integrative work initiated by the RCPs (Van Vuuren et al. 2017). These scenarios were not used in this study because they were not available at that time, and this might introduce a possible bias in our results. Accordingly, a systematic comparison between RCPs and SSPs would be advisable for a future investigation.

No funding was received for conducting this study.

T.A.G. and I.M.H.R. conceptualized the study; T.A.G. and R.M.E.-H. prepared the methodology; R.M.E.-H. prepared and wrote the original draft; T.A.G. and R.M.E.-H. wrote, reviewed, and edited the article. All authors have read and agreed to the published version of the manuscript.

On behalf of all authors, the corresponding author states that all data and materials support their published claims and comply with field standards.

The authors declare there is no conflict.