Enhancing uncertainty of regional climate models for climate change projection at Western Nile Delta

ABSTRACT

The objective of this research is to refine the precision of uncertainty estimates associated with Regional Climate Models (RCMs) for precipitation forecasts in the Western Nile Delta region of Egypt up to the year 2100. This refinement is predicated upon a comprehensive evaluation of the performance of selected models. The study employed two RCMs, specifically RCA4 and RACMO22, which integrated outputs from three distinct Global Circulation Models (GCMs): ICHEC-EC-EARTH, CCCma-CanESM2, and MPI-M-MPI-ESM-LR, each characterized by different resolutions. To assess the performance of these models, a suite of statistical metrics was utilized, including the Percentage of Bias, Root Mean Square Error (RMSE), and Nash-Sutcliffe Efficiency Coefficient. The evaluative process encompassed an analysis of precipitation patterns across three strategically selected locales within the study area: Alexandria, Borg-Al-Arab, and Wadi El-Natroun. The resultant performance indicators were within deemed acceptable limits, evidencing RMSE values up to 6.91, Percentage of Bias not exceeding 0.33, and Nash-Sutcliffe Efficiency Coefficients up to 0.78. The analysis elucidated that among the models scrutinized, the RCA4 model, when driven by output from the ICHEC-EC-EARTH-SMHI GCM, demonstrated superior performance, thereby underscoring its efficacy for detailed precipitation projection in the context of the Western Nile Delta.

KEYWORDS:

• Climate change
• regional climate model
• precipitation
• Egypt

Introduction

Addressing climate change is pivotal for the fulfillment of the United Nations’ seventeen Sustainable Development Goals (SDGs), given the significant threats and disasters posed by climate change to human populations, natural resources, and overall survival. The development of adaptation and mitigation strategies necessitates access to precise data and climate services capable of providing real-time and future climate projections. Global Climate Models (GCMs), which are instrumental in projecting future climatic conditions, often suffer from limitations related to their coarse spatial resolution, rendering them less effective for detailed assessment studies (Themeßl, Gobiet, & Leuprecht, 2011). This necessitates the refinement of these models to higher spatial resolutions through downscaling techniques, thereby enhancing the accuracy of climate change impact assessments and reducing associated uncertainties. Downscaling can be achieved through either dynamic or statistical methodologies, with each approach having distinct advantages. In regions characterized by complex topographies, such as coastal areas, downscaling is essential for generating precise local climate impact predictions (Flato et al., 2014).

Dynamic downscaling, driven by physical processes, is complex and computationally intensive, whereas statistical downscaling, being data-driven, is simpler and less demanding in terms of computational resources (Giorgi, 2019). Regional Climate Models (RCMs), predicated on dynamic downscaling, numerically simulate interactions between the atmosphere, oceans, land surfaces, and ice, solving the fundamental equations governing mass, energy transfer, and radiant exchange within the Earth’s climate system (IPCC, 2013; Pachauri et al., 2014). These models, integrating GCM outputs under identical boundary conditions, are designed to simulate atmospheric interactions more precisely. However, discrepancies often arise between historical observations and projections derived from RCMs, attributable to various sources of uncertainty. Such discrepancies highlight the challenges in modeling climate parameters accurately across different regions and seasons, and in the context of varying driving forces and scenarios for representative concentration pathways (RCPs) (Prein et al., 2016; Samuelsson et al., 2011).

The impact of climate change is particularly pronounced in developing nations, exacerbating vulnerabilities and socio-economic challenges (Asfaw, Bantider, Simane, & Hassen, 2021; Babaousmail et al., 2021; Demissie & Sime, 2021). Egypt, for instance, is significantly susceptible to climate change impacts, which manifest in the North Western part of the Nile Delta through alterations in precipitation patterns, extreme weather events, and the consequent reduction in freshwater availability due to saltwater intrusion, among other issues (IPCC, 2013; Pachauri et al., 2014; UNDP, 2017). Such challenges underscore the critical need for robust adaptation measures to address sea-level rise and its associated impacts (El-Nahry & Doluschitz, 2010).

Given the escalating frequency and intensity of climate-related disasters, such as droughts and floods, and the anticipated increase in temperature extremes across Egypt (Ayugi et al., 2020; Hamed, Salehie, Nashwan, & Shahid, 2023), there is an urgent requirement for precise climate data to facilitate local climate change impact assessments and the formulation of adaptation strategies at various administrative levels. This necessitates a comprehensive examination of historical climate patterns to inform future projections and impact assessments from diverse perspectives (Giorgi & Gutowski, 2015; IPCC, 2013). The escalating incidence of extreme weather events, inducing significant casualties and economic losses, further accentuates the need for such analyses, particularly in a country like Egypt where the Nile Delta plays a crucial role in the nation’s agriculture and economy, making it highly vulnerable to climate change impacts (EEAA, 1999, 2010; Kottek, Grieser, Beck, Rudolf, & Rubel, 2006; UNDP, 2017).

The elucidation of knowledge gaps, stemming from a paucity of data, necessitates a comprehensive evaluation of Regional Climate Models (RCMs) to ascertain the most suitable models for specific regions. This process requires the assessment of multiple RCMs to discern their performance characteristics and capabilities (Demissie & Sime, 2021; Dibaba, Miegel, & Demissie, 2019; IPCC, 2013). Accordingly, rigorous analyses are imperative to refine these datasets and mitigate uncertainties inherent within them. Accurate climatic data are essential for assessing the regional and local impacts of climate change, conducting impact analyses, and formulating adaptation strategies at both regional and national levels (Giorgi & Gutowski, 2015; IPCC, 2018). Such endeavors involve bias correction and the evaluation of various RCMs to identify a model of reliability and precision. Employing a multitude of performance evaluation methodologies enhances the comprehensiveness of understanding regarding the efficacy of models and bias correction techniques (Daniel, 2023; Deser, Phillips, Bourdette, & Teng, 2012). The Intergovernmental Panel on Climate Change (IPCC) advocates for the reduction of uncertainty through the exploration of both pessimistic and optimistic climate scenarios, and the extension of historical data series to spans no less than 30 years, thereby encompassing the full spectrum of climatic change and variability (IPCC, 2018).

Evaluating the competencies of historical RCMs is pivotal for selecting the model that demonstrates superior performance, which, in turn, is crucial for future climate projections and impact studies. The fidelity of future climate change predictions hinges on the precision with which climate models replicate historical climate variability (Pachauri et al., 2014). Hence, appraising the performance of RCMs at local and regional scales, such as the North Western part of the Nile Delta (NWDE), is vital to understanding the efficacy of each RCM in climate change impact assessment studies, acknowledging the presence of uncertainties within these models (Demissie & Sime, 2021).

In the present study, the performance of four RCMs was scrutinized, focusing on two specific models (RCA4 and RACMO22T) driven by three General Circulation Models (GCMs) of varying resolutions. The evaluation of these RCMs employed statistical metrics, including the percentage of bias (PBIAS), root mean square error (RMSE), Nash – Sutcliffe efficiency coefficient (NS), and mean absolute error (MAE). Following this assessment, the output from one regional climate model (ICHEC-EC-EARTH- SMHI- RCA4) was selected for bias correction, aiming to project future precipitation scenarios inclusive of extreme events under both optimistic and pessimistic Representative Concentration Pathways (RCPs). The derived information will inform the development of optimal adaptation and management strategies to mitigate the risks posed by climate change-induced disasters.

Material and methods

Study area description

The geographical focus of this investigation encompasses the North Western part of the Nile Delta in Egypt (NWDE), with a detailed simulation presented in Figure 1. For the purposes of detecting climate change and its variabilities, meteorological stations located in Alexandria, Borg El-Arab, and Wadi El-Natroun were strategically selected. Geographically, the study area is delineated by coordinates ranging from 30.1°N to 31.4°N in latitude and from 29.25°E to 30.1°E in longitude, as indicated in Table 1, which documents the precise locations of these selected stations. This region spans across three administrative governorates – Alexandria, El Behira, and Giza – and includes two significant coastal lakes: Idko and Maryuot. The area’s water resources are predominantly sourced from a network of irrigation canals, underground aquifers, and the seasonal rainfall, which averages approximately 120 mm/year along the coastal zones (Ashour, El Attar, Rafaat, & Mohamed, 2009). Notably, the region has experienced several extreme weather events, such as the notable occurrences in 2015, which have had adverse impacts on various human activities including agriculture, industry, tourism, and the monastic communities within the Wadi El-Natroun area (Salem, 2017). Historically characterized by its agricultural utility, the NWDE has recently been earmarked for land reclamation initiatives, with numerous ongoing projects aimed at bolstering the Gross Domestic Product (GDP) within the economic sector.

Figure 1. Study area.

Table 1. Location of selected stations.

Station name	Latitude (degree)	Longitude (degree)
Alexandria	31.2°N	29.95°E
Borg El-Arab	30.91°N	29.68°E
Wadi El-Natroun	30.4°N	30.2°E

Climate of the study area

In the designated region of investigation, encompassing the North Western part of the Nile Delta of Egypt (NWDE), the annual precipitation exhibits considerable spatial variability, with recorded averages ranging from 85 mm in Wadi El-Natroun to 214 mm in Alexandria. This precipitation distribution is largely seasonal, with the majority of rainfall occurring during the winter months, specifically from December through February. The geographical positioning of the delta, adjacent to the Mediterranean Sea, occasionally induces coastal showers and augments cloud cover. Such meteorological phenomena play a contributory role in enriching precipitation amount in this area, relative to the arid regions elsewhere in Egypt.

Climate data

For the quantitative assessment of climate change within this study, monthly average precipitation data were sourced from the Climatic Research Unit (CRU) at the University of East Anglia. The specific dataset utilized was the CRU Time Series (CRU-TS) version 4.04 (Harris, Jones, Osborn, & Lister, 2014), spanning from 1901 to the current period and characterized by a spatial resolution of 0.5° × 0.5°, thereby offering comprehensive global coverage. In addition, data for daily precipitation were acquired from the Africa Coordinated Regional Climate Downscaling Experiment (CORDEX) repository. The historical dataset encompasses a 35-year duration, extending from 1971 to 2005. This dataset was bifurcated into a calibration period, comprising the initial 20 years (1971–1990), and a validation period, consisting of the subsequent 15 years. Furthermore, datasets for two Representative Concentration Pathway scenarios, RCP 8.5 and RCP 4.5, were obtained for the projection timeframe from 2005 to 2100.

Used regional climate models and scenarios

The refinement of uncertainty in Regional Climate Models (RCMs) constitutes a critical endeavor for identifying and utilizing the most efficacious model to analyze the ramifications of climate change. This study’s precipitation projections were derived from four climate models, employing the methodology established by the Coordinated Regional Climate Downscaling Experiment (CORDEX), under the auspices of the World Climate Research Programme (WCRP). The selection of CORDEX as the foundational framework for this simulation was informed by its global recognition, proven efficacy in analogous projections across various locales, user-friendly interface, and capability to furnish high-resolution climate data across multiple grid cell scales. CORDEX integrates an array of RCMs and Empirical Statistical Downscaling (ESD) methods, facilitating the downscaling of local climate phenomena for a specified domain and deriving inputs from a collection of General Circulation Models (GCMs) as part of the Coupled Model Intercomparison Project Phase 5 (CMIP5) (IPCC, 2013), with the term “domain” referring to a geographically delineated area designated for local downscaling analysis.

For the purpose of this investigation, three distinct horizontal resolutions were employed to assess climate change impacts within the study region, utilizing three specific domains to encompass the requisite resolution parameters. These domains include the Africa domain (AFF44) with a grid resolution of 0.44 degrees, covering the entirety of Africa; the Mediterranean and North Africa domain (MENA22) with a grid resolution of 0.22 degrees; and the European domain (EUR11) with a grid resolution of 0.11 degrees, spanning Europe and the Mediterranean, with the study area situated to its south. The RCM outputs were meticulously analyzed to discern the model yielding the most credible projections (Panitz, Dosio, Büchner, Lüthi, & Keuler, 2014).

Notably, two RCMs, RCA4 and RACMO22T, were driven by three distinct GCMs – namely, ICHEC-EC-EARTH, CCCma-CanESM2, and MPI-M-MPI-ESM-LR – each characterized by unique boundary conditions, resolutions, and scenarios, as detailed in Table 2. The datasets generated by these models are accessible via the Earth System Grid Federation node at https://esg-dn1.nsc.liu.se/search/cordex/, with sponsorship and partnership from Earth System COG. The selection of models that yielded plausible outcomes facilitated the projection of future precipitation patterns within the study area.

Table 2. Used regional climate models (RCMs).

No	Climate model	Driving Model (GCM)	RCM	Developer
RCM1	ICHEC-EC-EARTH-SMHI-RCA4	ICHEC-EC-EARTH (Hazeleger et al., 2010)	RCA4	“Rossby Center Regional Climate model” (RCA4) (Samuelsson et al., 2011) and developed by “Swedish Meteorological and Hydrological Institute” (SMHI)
RCM2	ICHEC-EC-EARTH-KNMI-RACMO22	ICHEC-EC-EARTH (Hazeleger et al., 2010)	RACMO22	KNMI regional climate model RACMO22T (Meijgaard et al., 2012) and developed in “Royal Netherlands Meteorological Institute” (KNMI)
RCM3	CCCma-CanESM2-SMHI-RCA4	CCCma-CanESM2	RCA4	“Rossby Center Regional Climate model” (RCA4) (Samuelsson et al., 2011) and developed in Swedish Meteorological and Hydrological Institute (SMHI)
RCM4	MPI-M-MPI-ESM-LR-SMHI-RCA4	MPI-M-MPI-ESM-LR	RCA4	“Rossby Center Regional Climate model” (RCA4) (Samuelsson et al., 2011) and developed in “Swedish Meteorological and Hydrological Institute” (SMHI)

Models assessments

Climate models serve as sophisticated instruments for simulating Earth’s climate change (CC) dynamics. Critical to leveraging these models for climate projections is the comprehensive evaluation of their performance, particularly their capacity to accurately simulate precipitation. Such evaluations hinge on the comparison of model-generated precipitation outputs against observed data, facilitating the identification of models that yield veracious results for specific climate scenarios. This study focuses on the rectification of Regional Climate Models (RCMs) outputs for the period 1991–2005, assessed against observed precipitation data to ensure reliability and minimize uncertainties in climate projections.

The assessment encompassed the performance analysis of four climate model outputs over an extensive dataset of observed precipitation, covering a 35-year span. The models under evaluation were ICHEC-EC-EARTH-SMHI-RCA4, ICHEC-EC-EARTH-KNMI-RACMO22, CCCma-CanESM2-SMHI-RCA4, and MPI-M-MPI-ESM-LR-SMHI-RCA4. These models were subjected to a calibration process against long-term observed data from 1971 to 1990, with a subsequent evaluation period from 1991 to 2005 employing bias-corrected data. This process was conducted at three strategic locations: Alexandria, Borg El-Arab, and Wadi El-Natroun, utilizing a spatial resolution of 50 km × 50 km for both daily and monthly precipitation data.

To quantitatively assess the models’ performance, several statistical metrics were employed: the percentage of bias (PBIAS), root mean square error (RMSE), Nash – Sutcliffe efficiency coefficient (NS), and mean absolute error (MAE), as delineated in equations Eq 1, Eq 2, Eq 3, and Eq 4, respectively (Fang, Yang, Chen, & Zammit, 2015). Additionally, measures of central tendency and dispersion, such as the mean, median, and standard deviation, were computed to provide a comprehensive evaluation of the RCMs’ outputs. This rigorous analytical approach aims to refine the selection of RCMs that demonstrate the most accurate precipitation simulations, thereby enhancing the precision of future climate projections and reducing the associated uncertainties.

(1) $\mathit{NS}=1-\frac{{\sum }_{i=1}^n{\left(Y_i^{obs}-Y_i^{sim}\right)}^2}{{\sum }_{i=1}^n{\left(Y_i^{obs}-Y_i^{m\mathrm{ean}}\right)}^2}$(1)

(2) $P_{\mathit{BIAS}}=\frac{{\sum }_{i=1}^n\left(Y_i^{obs}-Y_i^{sim}\right)}{{\sum }_{i=1}^n\left(Y_i^{obs}\right)}$(2)

(3) $\mathit{MAE}=\frac{{\sum }_{i=1}^n\left|Y_i^{obs}-Y_i^{sim}\right|}{\mathit{n}}$(3)

(4) $\mathit{RMSE}=\sqrt{\frac{{\sum }_{i=1}^n{\left(Y_i^{obs}-Y_i^{sim}\right)}^2}{\mathit{n}}}$(4)

Where:

Y _i^obs and Y_i^sim : observed and simulated variable i $Y_i^{\mathit{mean}}$ : mean observed variables

n: observation total number.

Bias corrections and observed data

Reflecting the necessity for enhanced accuracy in climate projections at finer scales, research has identified persistent systematic biases within regional climate models (RCMs) and general circulation models (GCMs). Despite their efficacy in capturing broad climatic patterns, these models exhibit limitations when applied to localized studies (Shrestha, Acharya, & Shrestha, 2017). The employment of the linear scaling (LS) method emerges as a pivotal strategy for the bias correction of precipitation data. This method, aligning the monthly means of corrected outputs with observed datasets, operates on the premise of adjusting monthly values through a ratio derived from observed data and uncorrected RCM outputs (Lenderink, Buishand, & Van Deursen, 2007). Leveraging the Grid Analysis and Display System (GrADS), a tailored script was developed to ascertain the optimal RCM for historical data analysis, facilitating the derivation of bias correction coefficients. These coefficients are instrumental in refining future precipitation projections and in quantifying the frequency of extreme events, thereby offering invaluable insights for sectors prone to climate vulnerabilities. The applicability of these methodologies was evaluated through case studies in Alexandria, Borg El-Arab, and Wadi El-Natroun, underscoring their potential in enhancing climate resilience.

(5) $P_{\mathit{cor},\mathit{\hspace{0.17em}}m,d}=\mathit{p},\mathit{\hspace{0.17em}}{\mathit{\hspace{0.17em}}}_{\mathit{raw},\mathit{\hspace{0.17em}}m,d}\ast \left(\frac{\mathit{\mu },\mathit{\hspace{0.17em}}\left({\mathit{P}}_{\mathit{obs}.,\mathit{\hspace{0.17em}}m}\right)}{\mathit{\mu }\left(P_{\mathit{raw},\mathit{\hspace{0.17em}}m}\right)}\right)$(5)

Where$P_{\mathit{cor},\mathit{m},\mathit{d}}$corrected value of precipitation monthly.

$P_{\mathit{raw},\mathit{m},\mathit{d}}$raw value of precipitation monthly.

$P_{\mathit{obs}.,\mathit{m}}$Observed value of precipitation monthly.

µ Represents the expectation operator (e.g. µ P_obs) represents the mean value of observed precipitation at given month m.

The methodology for selecting an appropriate Regional Climate Model (RCM) projection for the study area incorporates a hybrid approach, amalgamating considerations of RCM performance and resolution through a delineated six-step process. The conceptual framework delineating the study’s phases is illustrated in Figure 2. This procedure commences with the preliminary selection of climate models, grounded in the calibration and evaluation of precipitation outputs from four distinct RCMs, each characterized by a spatial resolution of 0.44 degrees. Subsequent to this initial phase, the final model selection is predicated upon a rigorous validation process. This entails a comparative analysis between the outputs of the optimally simulated RCM – selected based on its superior performance – and the Climatic Research Unit Time-Series (CRU-TS) version 4.04 gridded data (Harris, Jones, Osborn, & Lister, 2014), across three specified horizontal resolutions. Following this comparative validation, the projected precipitation data pertinent to the study region are then elucidated, providing a foundation for subsequent analysis and interpretation.

Figure 2. Conceptual framework of study processes.

For the delineation of the regional climate model (RCM) projections within the specified study area, an integrative methodology combining RCM performance and resolution characteristics was employed through a structured six-step process. This methodology is depicted in Figure 2, which outlines the conceptual framework governing the study’s phases. The preliminary selection of climate models was predicated on a comprehensive calibration and performance evaluation of four distinct RCM precipitation outputs, each characterized by a spatial resolution of 0.44 degrees. Subsequent to this initial assessment, the final model selection was informed by a rigorous validation procedure, which entailed a comparative analysis of the outputs from the optimally performing RCM across three distinct horizontal resolutions against the Climate Research Unit gridded Time Series (CRU-TS) version 4.04 data (Harris, Jones, Osborn, & Lister, 2014). This facilitated the illustration of projected precipitation data for the study region, underscoring the methodological robustness underpinning the selection process.

Utilizing observed gridded data from the Climate Research Unit (CRU TS3.21) for the period 1971–2006 (Harris, Jones, Osborn, & Lister, 2014), a meticulous comparison and bias correction were conducted on historical data derived from the highest-performing RCM output. This procedure involved the application of bias factors, meticulously calculated to refine future precipitation projections. The efficacy of these adjusted projections was systematically investigated across three distinct future intervals: short-term (2006–2030), mid-term (2031–2060), and long-term (2061–2100), leveraging the superior resolution of the selected RCM. The projections were further scrutinized under two divergent climate scenarios, encompassing both pessimistic (RCP8.5) and optimistic (RCP4.5) pathways. A comprehensive numerical analysis, supported by RCM performance indicators, facilitated the estimation of climate change uncertainty ranges, detailed in Figure 2 through a sequential process: (1) comparison of RCMs to identify the highest-performing model, (2) bias adjustment between simulated and historically observed data to rectify common errors, (3) statistical evaluation of the corrected data, (4) selection of the ICHEC-EC-EARTH-SMHI-RCA4 as the model with the best performance, (5) employment of three distinct resolutions (AFF 0.44, MENA 0.22, and EU0.11 degrees) to enhance data precision, and (6) projection of two representative concentration pathways (RCP4.5 and RCP8.5) over three 30-year intervals to aid decision-makers in developing adaptation and mitigation strategies in response to climate change.

Results and discussion

In the present study, we undertook a comprehensive evaluation of four regional climate models (RCMs) to ascertain the most suitable model for predicting precipitation patterns within our designated study area. This evaluation was grounded in a comparative analysis of model outputs against historically observed data spanning from 1971 to 2005, employing a rigorous calibration and validation process. The culmination of this analysis facilitated the identification of the optimal RCM, which was subsequently employed for probabilistic precipitation forecasting. The selected model underwent further refinement to address bias discrepancies when contrasted with historical observations, enhancing the reliability of future precipitation projections.

Employing the best-fit model, we executed probabilistic projections for precipitation under the Representative Concentration Pathways (RCP) 4.5 and 8.5 scenarios, extending up to the year 2100 across three distinct future intervals: 2006–2030, 2031–2060, and 2061–2100. This approach enabled a nuanced understanding of potential precipitation trends and their implications under varying climate scenarios, thus providing a robust foundation for developing targeted adaptation and mitigation strategies in response to climate change.

RCMs performance

The evaluation of four Regional Climate Models (RCMs) – namely ICHEC-EC-EARTH-SMHI-RCA4, ICHEC-EC-EARTH-KNMI-RACMO22, CCCma-CanESM2-SMHI-RCA4, and MPI-M-MPI-ESM-LR-SMHI-RCA4 — was meticulously conducted against long-term historical observed data. This critical assessment aimed to mitigate uncertainties inherent in climate modeling by juxtaposing the simulations of daily and monthly historical precipitation data, at a resolution of 0.44 degrees, with long-term observed datasets spanning from 1971 to 1990. Figure 3 illustrates a minor bias in the average monthly precipitation across the four RCM outputs compared with Climate Research Unit (CRU) data for the same period, across three strategically selected stations: Alexandria, Borg El-Arab, and Wadi El-Natroun. Subsequently, bias corrections were applied to hindcast data for the period 1991 to 2005, culminating in a comparative analysis of the average monthly precipitation for these years against CRU data for an equivalent timeframe, as depicted in Figure 3. This meticulous process underscores the commitment to enhancing the reliability of regional climate projections, serving as a cornerstone for future climate resilience strategies.

Figure 3. Historical average precipitation (1971–1990) and (1991– 2005) by RCMs at Alexandria, Wadi El-Natroun, Borg El-Arab.

The efficacy of climate model simulations was rigorously assessed through four statistical metrics: standard deviation (STDV), root mean square error (RMSE), Nash-Sutcliffe efficiency (NSE), and percent bias (PBIAS). These metrics facilitated a comprehensive evaluation of the models’ performance by quantitatively comparing their bias-corrected outputs against observed climatic variables, detailed in Tables 3, 4 , and 5. NS is a measure of the agreement between the simulation and the observation, which varies from −∞ to 1, where 1 indicates an exact match. The model’s reliability relative to the mean increases with this value. PBIAS is the mean deviation of the simulated data from the observed data. Positive PBIAS values imply that the simulation overestimates the observation, while negative values imply that it underestimates it. PBIAS values close to 0.0 reflect more accurate model simulations. MAE is the mean absolute error of the model prediction, which is less affected by large errors than other error metrics (Fang, Yang, Chen, & Zammit, 2015). Among the models evaluated, the RCA4 model, driven by ICHEC-EC-EARTH dynamics, emerged as the most accurate in simulating precipitation patterns. This model’s outputs displayed remarkable alignment with observed data, characterized by minimal bias across all evaluated stations. Specifically, PBIAS values of 0.182, 0.015, and 0.11; RMSE scores of 6.52, 5.16, and 4.08; and NSE coefficients of 0.82, 0.76, and 0.83 were recorded for Alexandria, Borg-El-Arab, and Wadi El-Natroun, respectively. These indicators substantiate the RCA4 model’s superior capability in replicating observed precipitation trends, affirming its utility in regional climate modeling and the projection of hydrological variables.

Table 3. Indicators for assessing RCMs at Alexandria for the period (1991–2005).

	Statistics	ICHEC-EC-EARTH-KNMI-RACMO22T	ICHEC-EC-EARTH-SMHI-RCA4	CCCma-CanESM2-SMHI-RCA4	ICHEC-EC-EARTH-SMHI-MPI-M-MPI-ESM-LR	CRU
Alexandria	Mean (mm)	16.62	20.63	20.15	14.02	20.34
	Median (mm)	10.56	8.29	12.32	7.41	7.4
	STDV (mm)	19.83	21.13	22.66	17.20	23.71
	NS	0.732	0.82	0.791	0.71
	PBias %	−0.014	0.182	0.09	0.31
	RMSE	10.43	6.52	8.44	11.85

Table 4. Indicators for assessing RCMs at Wadi El-Natroun for the period (1991–2005).

	Statistics	ICHEC-EC-EARTH-KNMI-RACMO22T	ICHEC-EC-EARTH-SMHI-RCA4	CCCma-CanESM2-SMHI-RCA4	ICHEC-EC-EARTH-SMHI-MPI-M-MPI-ESM-LR	CRU
Wadi El-Natroun	Mean(mm)	8.23	7.11	1.43	0.38	7.11
	Median(mm)	3.33	5.18	0.19	0.14	5.18
	STDV (mm)	12.14	7.03	2.18	0.58	7.03
	NS	0.71	0.83	−1.92	−2.254
	PBias %	−0.31	0.11	0.77	0.81
	RMSE	7.79	4.08	7.40	8.90

Table 5. Indicators for assessing RCMs at borg-El-arab for the period (1991–2005).

	Statistics	ICHEC-EC-EARTH-KNMI-RACMO22T	ICHEC-EC-EARTH-SMHI-RCA4	CCCma-CanESM2-SMHI-RCA4	ICHEC-EC-EARTH-SMHI-MPI-M-MPI-ESM-LR	CRU
Borg-El-Arab	Mean(mm)	11.00	11.31	1.71	0.66	11.1
	Median(mm)	5.51	5.41	0.16	0.12	3.84
	STDV (mm)	11.65	14.4	9.47	4.66	13.4
	NS	0.72	0.76	−1.23	−1.87
	PBias %	0.012	0.015	0.84	0.94
	RMSE	7.59	5.16	14.59	16.55

Furthermore, a comparative analysis of the mean monthly simulation accuracy across all stations, both pre- and post-bias correction, underscored the superior performance of the RCA4 model relative to the RACMO22T and MPI-ESM-LR models in daily precipitation reproduction. This outcome corroborates findings by Ismail, Zahran, ElmouSTAFA, Attia, and Hassan (2021), which posited that the RCA4 model’s simulation of rainfall was markedly more accurate than other CORDEX RCMs across the Western Coast Zone of Egypt).

RCMs assessment based on resolutions

The selection process for the Regional Climate Model (RCM) commenced with a rigorous evaluation of its precipitation simulation capabilities over two distinct periods: the calibration period spanning 1971–1990 and the validation period from 1991 to 2005. This assessment underscored the ICHEC-EC-EARTH-SMHI-RCA4 model as the optimal fit for the study at hand. Notably, model uncertainty was observed to fluctuate with adjustments in horizontal resolution. Consequently, the study delved into the analysis of precipitation data within the designated area employing the ICHEC-EC-EARTH-SMHI-RCA4 model, which was scrutinized at varying horizontal resolutions (0.44°, 0.22°, and 0.11°). This approach aimed to refine the precision of climate parameter predictions. Figure 4 elucidates the comparative analysis of RCM data across the three resolutions against CRU data for the period 1971 to 1990, revealing a marked enhancement in simulation accuracy at finer resolutions

Figure 4. Average monthly precipitation for 1971 to 1990, for Alexandria, Borg El-Arab and Wadi El-Natroun with different resolutions by (AFF0.44, MINA 0.22, and EUR 0.11).

Furthermore, a detailed examination of the RCM’s output statistics – encompassing standard deviation, median, mean, percent bias (PBIAS), Nash – Sutcliffe efficiency (NSE), and mean absolute error (MAE) – was conducted against the observed precipitation data’s horizontal resolutions for all stations. As delineated in Tables 6, 7 , and 8, a resolution of 0.11° yielded the lowest root mean square error (RMSE) values at Alexandria. Conversely, for Borg El-Arab, minor discrepancies were observed among the resolutions, reflective of the models’ differential impacts. During the rainy season (DJF), the MENA 0.22° resolution emerged as the closest to actual precipitation values.

Table 6. Indicators for ICHEC-EC-EARTH- SMHI- RCA4 based horizontal resolution at Alexandria for the period (1991–2005).

Location		AFF-44	MNA-22	EUR-11	CRU
Alexandria	Mean(mm)	16.62	15.27	17.10	20.34
	Median(mm)	36.2	8.72	9.17	7.4
	STDV (mm)	19.83	16.49	18.66	23.7
	NS	0.733	0.616	0.776
	Pbias %	0.224	0.332	0.19
	RMSE	6.91	5.55	5.37

Table 7. Indicators for ICHEC-EC-EARTH- SMHI- RCA4 based horizontal resolution at borg-El-arab for the period (1991–2005).

Location		AFF-44	MNA-22	EUR-11	CRU
Borg-El-Arab	Mean(mm)	11.00	9.68	10.75	11.14
	Median(mm)	28.29	4.81	5.34	3.84
	STDV (mm)	14.41	11.71	13.05	13.41
	NS	0.722	0.753	0.740
	Pbias %	0.012	0.131	0.035
	RMSE	4.82	3.36	4.11

Table 8. Indicators for ICHEC-EC-EARTH- SMHI- RCA4 based horizontal resolution at Wadi El-Natroun for the period (1991–2005).

Location		AFF-44	MNA-22	EUR-11	CRU
Wadi El-Natroun	Mean (mm)	8.23	7.43	5.98	6.00
	Median(mm)	18.1	3.48	2.63	1.77
	STDV (mm)	12.145	9.53	6.36	7.11
	NS	0.602	0.671	0.781
	Pbias %	−0.27	−0.19	0.003
	RMSE	3.38	2.77	2.25

Precipitations projection

In an effort to evaluate precipitation patterns in the designated study area, two scenarios were scrutinized: one with high radiative forcing (RCP 8.5) and the other with intermediate radiative forcing (RCP 4.5). The temporal scope of this investigation was divided into three distinct periods: near future (2006–2030), intermediate future (2031–2060), and far future (2061–2100), with all stations under consideration.

Upon examination of the Probability Density Function (PDF) for monthly, annual, and extreme values, it was discerned that the ICHEC-EC-EARTH-SMHI-RCA4 model, with a resolution of 0.11°, offered the most accurate estimate for future precipitation. This was evidenced by a slight upward trend in the PDF for precipitation projections relative to the observed projections across all three stations, as depicted in Figure 5.

Figure 5. Probability density function (PDF) for average monthly precipitation for RCP4.5 and RCP 8.5 at Alexandria, Borg El-Arab and Wadi El-Natroun (three time slices) EUR 0.11) for ICHEC-EC-EARTH- SMHI- RCA4.

Despite the observed increase in precipitation density values, a clear pattern emerged. Both the near future and intermediate future periods showed an increase in precipitation under both RCP4.5 and RCP8.5 scenarios. However, a decrease was observed in the far future under the same scenarios. Furthermore, the total annual precipitation is projected to decrease under both RCP 4.5 and RCP 8.5 scenarios across all time periods, as illustrated in Figures 6 and 7.

Figure 6. Precipitation average monthly and annually for near, intermediate and far future for RCP4.5 so as RCP 8.5 at Alexandria, Borg El-Arab and Wadi El-Natroun byICHEC-EC-EARTH- SMHI- RCA4 (EUR 0.11).

Figure 7. Precipitation average monthly for near, intermediate and far future for RCP4.5 so as RCP 8.5 at Alexandria, Borg El-Arab and Wadi El-Natroun byICHEC-EC-EARTH- SMHI- RCA4 (EUR 0.11).

In Alexandria, a thorough examination of precipitation projections for the near future (2006–2030) under different radiative forcing scenarios revealed noteworthy insights. Under the RCP 8.5 scenario, the projected precipitation exhibited distinct probabilities of occurrence and corresponding values. Specifically, when the probability of occurrence was at 4%, the projected precipitation ranged from 30 to 40 mm/year. At a probability of 10–20%, the precipitation was estimated to be between 5 and 10 mm/year. A 50% probability corresponded to a projected precipitation of 5 mm/year, while a 1% probability yielded a considerably higher precipitation estimate of 70 mm/year. The mean monthly precipitation for this scenario was calculated at 19.65 mm, with an annual total of 238.32 mm. The maximum monthly precipitation occurred in December, reaching 70 mm.

In the context of the RCP 4.5 scenario for the near future, the projected precipitation values varied based on probability of occurrence. At a 5% probability, precipitation ranged from 40 to 50 mm/year. A probability of 8% corresponded to an estimate of 10 mm/year, while a 50% probability resulted in 5 mm/year of precipitation. For probabilities of occurrence less than 1%, the projected precipitation increased notably to 70 mm/year. The mean monthly precipitation in this scenario averaged 18.5 mm, with an annual total of 221.6 mm. The highest monthly precipitation occurred in January, reaching 45.4 mm.

Transitioning to the medium future (2031–2060), under the RCP 4.5 scenario, the projected precipitation values were as follows: at a 5% probability, precipitation ranged from 40 to 50 mm/year. Within the range of 10% to 30% probability, precipitation was estimated between 5 and 10 mm/year. A 50% probability corresponded to 5 mm/year of precipitation, while a 1% probability yielded an average precipitation of 18 mm/year. The maximum monthly precipitation, observed in December, reached 63 mm, with an annual total precipitation of 216.7 mm.

Under the RCP 8.5 scenario for the same medium future period, precipitation projections exhibited variations based on probability of occurrence. Probabilities between 8–10% were associated with precipitation exceeding 10 mm/year, while probabilities between 7–30% indicated precipitation between 5–10 mm/year. A 45% probability was linked to a consistent 5 mm/year of precipitation. The mean monthly precipitation averaged 19.44 mm, with a total annual precipitation of 233.33 mm. The months of January and February registered the maximum precipitation, reaching 45.4 mm. These detailed projections offer valuable insights into the anticipated precipitation patterns for Alexandria under different radiative forcing scenarios and time frames.

During the far future period (2061–2100) in Alexandria, under the RCP 4.5 scenario, the projected precipitation exhibited a probability of occurrence of 6% for 10 mm/yr., 30% for 4 mm/yr., and approximately 2% for 45 mm/yr. The maximum precipitation observed was 66.9 mm in December, with a mean monthly precipitation of 16.44 mm and a total annual precipitation of 197.27 mm.

Under the RCP 8.5 scenario, the projected precipitation demonstrated a probability of occurrence of 5% for approximately 10 mm/yr., 32% for 5 mm/yr., and around 1% for 55 mm/yr. The maximum precipitation recorded was 70.79 mm in December, with a mean monthly precipitation of 14.75 mm and a total annual precipitation of 177 mm. Comparable results were obtained for Borg El-Arab, albeit with slight variations observed for Wadi El-Natroun.

It is noteworthy that a subtle distinction emerged in the case of the MENA 0.22° and EUR 0.11° resolutions at the Borg El-Arab and Wadi El-Natroun regions, both of which are situated in lowlands with depressions. The probability density function (PDF) trend for precipitation projection consistently surpassed observed values across all examined time frames and stations.

Notably, the maximum precipitation values exhibited a marginal increase during near and intermediate future projections, transitioning from RCP 4.5 to RCP 8.5. However, a significant decline in maximum precipitation was observed in the far future projections under both RCP 4.5 and RCP 8.5 scenarios. This decline was consistent across all time frames. Concurrently, the total annual precipitation displayed a decreasing trend for both RCP 4.5 and RCP 8.5, spanning all time frames.

Specifically, the probability of extreme precipitation events, as indicated by occurrences with maximum precipitation values, demonstrated a decreasing trend. For RCP 4.5, the probability of occurrence was projected to be 1.6%, 2%, and 0.8% during the near, intermediate, and far future projections, respectively, at Alexandria. Similarly, the maximum precipitation probability of occurrence during extreme events remained below 1% throughout all time frames at Borg El-Arab and Wadi El-Natroun.

Given that the study area is characterized by a high population and numerous socioeconomic activities, it is imperative to consider the potential impacts of these projections when formulating adaptation strategies for climate change. This encompasses aspects such as irrigation, drainage systems, water availability, and infrastructure rehabilitation, with a particular focus on low-lying regions like Wadi El-Natroun. Furthermore, the establishment of early warning systems assumes significant importance, providing essential incentives for vulnerable populations in the event of extreme weather events and contributing to their resilience in the face of potential disasters.

Conclusion

The study focuses on enhancing the accuracy of climate change projections in the Western Nile Delta through the application of methods aimed at reducing uncertainty in regional climate models (RCMs). The performance evaluation involved the assessment of four RCM outputs, with a particular focus on two RCMs, namely RCA4 and RACMO22T, driven by three distinct global circulation models (GCMs).

Following a meticulous selection process to determine the most suitable RCM for the study area, climate data was further scrutinized at different resolutions. The findings unequivocally identify ICHEC-EC-EARTH-SMHI-RCA4 as the optimal model for climate change projections in the study area. Particularly, reliable resolution was observed in the MENA-0.22° and EU-0.11° regions, outperforming the AFF-0.44° resolution. This distinction is particularly pronounced in coastal zones such as Alexandria, characterized by frequent morphological changes and a propensity for rainfall.

The results obtained under the RCP 8.5 and RCP 4.5 scenarios demonstrated commendable agreement with the data from the calibrated and validated periods. It is worth noting that additional methods of bias correction, as suggested by Luo et al. (2018), may further enhance the confidence in these outcomes.

The implications of these findings are substantial. They underscore the importance of considering the impact of future projections when formulating adaptation strategies to address climate change across various development sectors. This encompasses critical aspects such as irrigation, drainage systems, water availability, and infrastructure rehabilitation, with a specific emphasis on low-lying areas, exemplified by Wadi El-Natroun.

Moreover, the regional climate experiments conducted herein can potentially serve as valuable references for other regions globally, provided that appropriate bias corrections are applied using accurate observed climate data.

Recommendation

The findings of this research underscore the paramount importance of addressing data uncertainties to enhance the reliability of climate change projections. One critical step toward achieving this objective is the establishment of robust early warning systems. These systems play a pivotal role in not only providing vital data and information but also offering incentives for individuals with limited incomes in the event of extreme weather events. Such proactive measures can significantly bolster their resilience in the face of impending disasters.

Furthermore, the integration of ensembles of regional climate models (RCMs) and scenarios, particularly in the context of assessment report six, warrants comprehensive elaboration. This approach holds the promise of yielding more accurate results with heightened confidence, thereby mitigating uncertainties in future predictions.

Simultaneously, it is imperative to redouble efforts in the realm of in-situ measurements. These on-site measurements hold the key to producing precipitation projections that mirror reality more closely. By investing in enhanced in-situ data collection, we can further refine the accuracy and realism of our precipitation projections, contributing to more effective climate change mitigation and adaptation strategies.

Disclosure statement

No potential conflict of interest was reported by the author(s).