Performance Evaluation of CMIP6 Climate Model Projections for Precipitation and Temperature in the Upper Blue Nile Basin, Ethiopia

Abstract

The projection and identification of historical and future changes in climatic systems is crucial. This study aims to assess the performance of CMIP6 climate models and projections of precipitation and temperature variables over the Upper Blue Nile Basin (UBNB), Northwestern Ethiopia. The bias in the CMIP6 model data was adjusted using data from meteorological stations. Additionally, this study uses daily CMIP6 precipitation and temperature data under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios for the near (2015–2044), mid (2045–2074), and far (2075–2100) periods. Power transformation and distribution mapping bias correction techniques were used to adjust biases in precipitation and temperature data from seven CMIP6 models. To validate the model data against observed data, statistical evaluation techniques were employed. Mann–Kendall (MK) and Sen’s slope estimator were also performed to identify trends and magnitudes of variations in rainfall and temperature, respectively. The performance evaluation revealed that the INM-CM5-0 and INM-CM4-8 models performed best for precipitation and temperature, respectively. The precipitation projections in all agro-climatic zones under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios show a significant (p < 0.01) positive trend. The mean annual maximum temperature over UBNB is estimated to increase by 1.8 °C, 2.1 °C, and 2.8 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5 between 2015 and 2100, respectively. Similarly, the mean annually minimum temperature is estimated to increase by 1.5 °C, 2.1 °C, and 3.1 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5, respectively. These significant changes in climate variables are anticipated to alter the incidence and severity of extremes. Hence, communities should adopt various adaptation practices to mitigate the effects of rising temperatures.

1. Introduction

The changing climate is a worldwide concern which poses the threat of the century and has resulted in significant damages, including irreparable losses in ecosystems on both land and underwater [1]. It is caused by greenhouse gas emissions into the atmosphere, which warm the planet [2]. As per the United Nations Development Programme (UNDP) report, smallholder farmers in developing nations will be particularly affected by the negative effects of climate change because they mostly rely on natural systems to raise cattle and grow crops [3]. Africa is among the most susceptible continents to the consequences of climate change, including droughts and floods caused by warming air temperatures and altered precipitation patterns that endanger economic development, agriculture, and water supply [4,5]. The continent has limited adaptability because of restricted access to financial assets, technology, and knowledge to cope with the effects of climate hazards [6,7,8]. The lives and economies of people living in the Horn of Africa, notably Ethiopia, are significantly impacted by climate change and variability, since a large portion of the nation’s economy depends on climate-sensitive industries including forestry, agriculture, and fisheries [9]. Agriculture and food security in Ethiopia are predicted to suffer greatly due to the effects of climate change [10].The Upper Blue Nile Basin (UBNB) (Abay), which is the focus of this particular study, is one of the main agricultural productive areas of Ethiopia [11]. The community’s primary source of income is rain-fed agriculture, which is greatly influenced by changing climate due to significant seasonal variations in precipitation and temperature variables [12]. Similarly, variations in periodic and spatial patterns of temperature and precipitation would affect cropping patterns, productivity, wetness of the soil, groundwater levels, and the duration of droughts [13,14]. Hence, projections of climate change are essential for defining climate-related risks and feasible options for response [15]. Additionally, global climate models (GCMs) are the main instruments for projecting future climate patterns and trends. The recent generation of the Coupled Model Intercomparison Project phase six (CMIP6) includes improvements to current parameterizations, the inclusion of recently developed physical processes, and increased resolution compared with other CMIP generations [16]. In addition, CMIP6 global model was set up to assess the effectiveness of GCMs in simulating past, present, and future climate variables under various conditions. Recently, the CMIP6 has been released, which integrated the representative concentration pathways (RCPs) and shared socioeconomic pathways (SSPs) (population, technology, gross domestic product (GDP), and land-use scenarios) and made projections more authentic [16]. More importantly, bias correction is a statistical approach used to address data exhibiting systematic errors or biases, thereby adjusting it to its correct values [17]. It corrects for the propensity to underestimate/overestimate the mean value of downscaled variables by using the statistics of observed and historically simulated variables for similar periods [18]. As a result, outputs from GCMs often cannot be used directly for climate assessments [19]. Bias-correction methods include linear scaling (LS), power transformation (PT) (for rainfall), local intensity scaling, distribution mapping (DM), variance scaling (for temperature), delta change, and quantile mapping (QM) [20,21].

There have been studies conducted in the study area evaluating the quality of the CMIP6 climate model’s outputs using GCM simulations. Hence, this study addresses the issues by evaluating the performance of the CMIP6 climate model and projecting precipitation and temperature variables under multiple scenarios in different agro-ecological zones of the UBNB. This will help to formulate effective strategies for mitigation, as well as adaptations for climate-related impacts. In addition, the projection, as well as the identification, of past and future changes in climate extremes is important to analyze heatwave duration, heat-related illness, fire risk, livestock heat stress, agricultural productivity decline, and so on. Similarly, this study is critical for disaster-prone locations in order to design adaptation and mitigation actions. This is especially relevant when utilizing the latest-released global climate models (GCM) from CMIP Phase 6.

2. Methodology

2.1. Location of the Study Area

This study was undertaken in the Upper Blue Nile Basin (UBNB), located in the northwestern part of Ethiopia. Geographically, the basin is situated between 7°45′ and 12°45′ N latitude and 34°05′ and 39°45′ E longitude. The basin is drained by the Abay/Nile River and has an estimated area of 200,718 km², comprising 18 percent of Ethiopia’s total area (1,112,000 km²). It includes diverse topographic characteristics, with elevations ranging from 478 to 4260 m above mean sea level (masl) (Figure 1). The UBNB contributes approximately 70 percent (54.4 billion m³/year) to the total annual flow of the main Nile [22,23]. Additionally, the basin experiences an average annual rainfall of 1180.8 mm, with maximum and minimum temperatures of 26.7 °C and 12.1 °C, respectively. The highest rainfall occurs during the summer season, with about 77.7 percent of the total rainfall occurring between June and September. In addition, the study area is endowed with five agro-ecological zones based on its altitudinal classification, namely, desert from the lowest altitude (478 masl to 500 masl), lowland (Kolla) from (500 masl to 1500 masl), midland (Woina-dega) from 1500 to 2300 masl, highland (Dega) from 2300 to 3200 masl, and upper highland/cold (Wurch) from 3200 to 4260 masl.

2.2. Datasets

Daily rainfall and temperature datasets were obtained from 27-gauge stations for the period 1995–2020 from the National Meteorology Agency of Ethiopia (NMAE). The data were primarily used to assess the performance of CMIP6 climate model data. In addition, baseline (1995–2014) and future (2015–2100) CMIP6 projections, based on SSP scenarios [16], were extracted from the World Climate Research Program’s (WCRP) CMIP6 database https://esgf-metagrid.cloud.dkrz.de/search/cmip6-dkrz/ (accessed on 7 August 2024). These projections include lowest emission (SSP1-2.6), medium (SSP2-4.5), and strong emission scenarios (SSP5-8.5) for near-term (2015–2044), mid (2045–2074), and far (2075–2100) time-scales. The division of time periods is commonly used in climate studies and other long-term projections. For instance, the near-term period (2015–2044) reveals immediate responses to the current levels of greenhouse gasses already present in the atmosphere. Changes in temperature and precipitation during this period are largely influenced by short-term climate variability, such as El Niño-Southern Oscillation (ENSO), as well as existing human actions. This timeframe allows for the assessment of short-term adaptation needs and immediate mitigation efforts.

The mid-term period (2045–2074) reflects the increasing influence of cumulative emissions and the socio-economic development choices made over the preceding decades. It helps to evaluate the effectiveness of current mitigation policies and guides medium-term adaptation strategies in sectors such as agriculture, water management, and infrastructure. The far-term period is critical for assessing high-risk scenarios, long-term impacts on ecosystems, and extreme events, such as intensified droughts or floods. Therefore, understanding these distinctions assists local policymakers in selecting adaptation strategies based on various future scenarios. Hence, seven CMIP6 GCMs were selected (

Table 1. List of seven selected CMIP6 climate models for climate projection in UBNB.

CMIP6 GCM	Institution	Resolution	References (Author, Year)
BCC-CSM2-MR	Beijing Climate Center, China	1.1° × 1.1°	Wu et al. [28]
CMCC-ESM2	Euro-Mediterranean Centre on Climate Change (CMCC), Italy	1.1° × 0.9°	Lovato et al. [29]
MPI-ESM1-2-HR	Max Planck Institute for Meteorology (MPI), Germany	0.9° × 0.9°	Kamruzzaman et al. [30]
MRI-ESM2-0	Meteorological Research Institute, Japan	1.1° × 1.1°	Shiru and Chung [31]
INM-CM5-0	Institute for Numerical Mathematics (INM), Russia	1° × 1°	Volodin et al. [32]
INM-CM4-8	Institute for Numerical Mathematics (INM), Russia	1° × 1°	Volodin et al. [32]
NORESM2-MM	Norwegian Climate Center, Norway	0.9° × 1°	Seland et al. [33]

) based on their performance to convey rainfall and temperature details for historical and future SSPs, their nominal resolution, frequency (daily), and analysis period. Furthermore, recent studies [12,24,25,26,27] have used these models to project future trends in temperature and rainfall across various regions of the country.

2.3. Data Processing and Evaluation

2.3.1. Bias Correction Techniques

This study applied power transformation (PT) for precipitation bias correction, and distribution mapping (DM) was used for temperature bias correction, utilizing the CMhyd tool. PT and DM are especially recommended for correcting GCM for precipitation and temperature, respectively [18,21,34]. The CMhyd tool [35] implements these bias-correction techniques by comparing raw GCM products with baseline data, calculating the variation between observed and simulated data, and applying various bias-correction methods to adjust both historical and future model outputs. CMhyd was designed to provide climate data simulations that are extracted, bias-corrected, and potentially representative for each gauge location used in the study area [35]. The frameworks used in this study for bias correction method was shown in Figure 2.

2.3.2. Performance Evaluation of CMIP6 Climate Model

It is critical to evaluate the effectiveness of climate model products before using them for any projections [36]. To evaluate the performance of CMIP6’s output against observed data, the evaluation techniques applied include the Pearson correlation coefficient (r), mean error (ME), mean absolute error (MAE), root mean square error (RMSE), and percentage of bias (PBias). These methods are important for analyzing how effectively the model performs.

The Pearson correlation coefficient (r) measures the strength and direction of the relationship between two variables, with values ranging between −1 and 1. In this study, it is used to determine how closely the model’s output correlates with the observed data.

$$\mathrm{r}={\displaystyle \frac{\sum (\mathrm{O}-\mathrm{Ō}\left)\right(\mathrm{M}-\mathrm{ḿ})}{\sqrt{{\sum \left(\mathrm{O}-\mathrm{Ō}\right)}^2}\sqrt{{\sum \left(\mathrm{M}-\mathrm{ḿ}\right)}^2}}}$$

(1)

where r = the correlation coefficient, O = the value of observed rainfall, Ō = average observed rainfall measurement, M = the value of model rainfall estimate, and ḿ = average model rainfall.

The mean error (ME) refers to the average of all the errors in a set of measurements, a positive ME value indicates an overestimation of the model, and a negative value shows an underestimation.

$$\mathrm{M}\mathrm{E}={\displaystyle \frac{\sum \left(\mathrm{M}-\mathrm{O}\right)}{\mathrm{n}}}$$

(2)

where ME = the mean error in mm, M = model rainfall estimate, O = observed rainfall measurement, and n = number of data points.

The root mean square error (RMSE) measures the average magnitude of the estimated errors between the model and the station data. A lower RMSE value means greater central tendencies and smaller extreme errors. An RMSE value of zero shows the perfect score.

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{{\displaystyle \frac{\sum _{\mathrm{i}=1}^{\mathrm{n}}{\left(\mathrm{M}\mathrm{i}-\mathrm{O}\mathrm{i}\right)}^2}{\mathrm{n}}}}$$

(3)

where RMSE is the root mean square error in mm, Oi = observed rainfall measurements, Mi = model estimates, and n = number of data points.

The normalized root mean square error (NRMSE) is a better indication for evaluating model performance, since normalizing the RMSE (the NRMSE) can help to make the RMSE scale-free. The NRMSE is a percentage that connects the RMSE to the observed range of the variable. This gives a number between 0 and 1, with values closer to 0 indicating better-fitting models.

$$\mathrm{N}\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}={\displaystyle \frac{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}{\stackrel{-}{\mathrm{X}}}}$$

(4)

where RMSE is the root means square error and X is the mean.

The Bias, on the other hand, represents how closely the mean of model rainfall correlates to the mean of actual rainfall. It expressed as:

$$\mathrm{P}\mathrm{B}\mathrm{i}\mathrm{a}\mathrm{s}={\displaystyle \frac{\sum \left(\mathrm{ḿ}-\mathrm{Ō}\right)}{\sum \mathrm{Ō}}}\times 100\mathrm{\%}$$

(5)

where ḿ and Ō are the mean of the model and observed rainfall estimates, respectively.

2.3.3. Performance Assessment Using Taylor Diagram

In addition, the Taylor diagram (TD) [37] provides a comprehensive statistical summary of the correlation and standard deviation between the model and observed data. The performance is assessed based on a selected reference using the Taylor diagram. This method is recommended to identify how closely a model’s output fits with observations in terms of correlation. Specifically, the TD describes how two fields are statistically related: a “test” field (often representing a model simulation) and a “reference” field (usually representing the “truth,” based on observations). A correlation coefficient close to 1 indicates a high similarity between the CMIP6 datasets and the observed data, whereas a coefficient of −1 implies an opposite relationship.

2.4. Data Analysis Methods

The Methodological framework illustrating the steps used in this paper was shown in Figure 3.

2.4.1. Rainfall and Temperature Trend Analysis

Mann–Kendall (MK) trend test, which is non-parametric, is well suited for handling not-routinely distributed data, outliers, and missing values that are commonly encountered in hydrological time series [38]. In this study, the MK trend test was used to investigate historical and future trends in rainfall, temperature variables, and extreme indices. Each data value is compared to all previous data values. If a later-period data value is greater than an earlier-period data value, the statistic (S) is increased by one. Conversely, (S) decreases by one if the later-period data value is less than the earlier-period data value. The final value of (S) is obtained by averaging all such changes [39,40]. It is computed as follows:

$$\mathrm{S}=\sum _{\mathrm{i}=1}^{\mathrm{n}-1}\sum _{\mathrm{j}=\mathrm{i}+1}^{\mathrm{n}}\mathrm{s}\mathrm{g}\mathrm{n}\left(\mathrm{x}\mathrm{j}-\mathrm{x}\mathrm{i}\right)$$

(6)

A trend test is applied to a pair of time series: Xi, which is ranked from i = 1, 2…n − 1 and Xj, which is ranked from j = i + 1, 2…n. Each of the data points of Xi are taken as a reference point, which is compared with the rest of the data points Xj and sgn (xj − xi), which are the sign function computed as:

$$\mathrm{s}\mathrm{g}\mathrm{n}\left(\mathrm{x}\mathrm{j}-\mathrm{x}\mathrm{i}\right)=\left\{\begin{array}{c}+1,\hspace{1em}\mathrm{if} \left(\mathrm{x}\mathrm{j}-\mathrm{x}\mathrm{i}\right)>0\\ 0,\hspace{1em}\mathrm{if} \left(\mathrm{x}\mathrm{j}-\mathrm{x}\mathrm{i}\right)=0\\ -1,\hspace{1em}\mathrm{if} \left(\mathrm{x}\mathrm{j}-\mathrm{x}\mathrm{i}\right)<0\end{array}\right\}$$

(7)

where Xi and Xj, respectively, are the yearly values in years j and i (j > i).

$$\mathrm{Z}=\left\{\begin{array}{c}{\displaystyle \frac{5-1}{\sqrt{\mathrm{v}\mathrm{a}\mathrm{r}\left(\mathrm{S}\right)}}} , \mathrm{S}>0\\ 0, \mathrm{S}=0\\ {\displaystyle \frac{5+1}{\sqrt{\mathrm{v}\mathrm{a}\mathrm{r}\left(\mathrm{S}\right)}}} , \mathrm{S}<0\end{array}\right.$$

(8)

where Z’s positive and negative values, respectively, indicate increasing and decreasing trends.

In addition to identifying whether the trend exists, the magnitude of the trend was also estimated by a slope estimator (mij), and it is considered as the median for all datasets for various combinations. The positive value of (mix) indicates an upward trend, and the negative value indicates a downward trend. The magnitude of the trend was predicted by Sen’s slope estimator with the slope (mi) of all data pairs computed as follows [41]:

$$\mathrm{m}\mathrm{i}={\displaystyle \frac{\mathrm{X}\mathrm{j}-\mathrm{X}\mathrm{i}}{\mathrm{j}-\mathrm{i}}}$$

(9)

where the data values at time j and i (j > i) are denoted by the variables Xj and Xi, respectively. Sen’s estimate of slope is the median of these N values of Ti. It is computed as follows:

$$\mathrm{M}\mathrm{m}\mathrm{e}\mathrm{d}=\mathrm{m}\left[\begin{array}{c}{\displaystyle \frac{\mathrm{N}+1}{2}},\hspace{1em} \mathrm{If} \mathrm{N} \mathrm{is} \mathrm{Odd}\\ {\displaystyle \frac{\frac{\mathrm{N}}{2}+\frac{\mathrm{N}+2}{2}}{2}}, \mathrm{If} \mathrm{N} \mathrm{is} \mathrm{Even}\end{array}\right]$$

(10)

Ultimately, a non-parametric model has been used to compute the slope median (Med), which yields the trend and slope magnitude. In the time series analysis, a positive value of mi indicates an upward trend, whereas a negative value indicates a falling or downward trend. In the same way, a zero number denotes the absence of any data trend.

2.4.2. Variability Assessment

Coefficient of variation (CV)—It was employed to evaluate both the seasonal and annual precipitation data variability. The rainfall time series data exhibit greater variability when the CV value is higher. It expressed as:

$$\mathrm{C}\mathrm{V}={\displaystyle \frac{\mathsf{\sigma }}{\stackrel{-}{\mathrm{X}}}}\times 100$$

(11)

where X represents the average rainfall data and σ stands for the standard deviation. Typically, the CV values were categorized as follows: low (CV < 20%), moderate (20% < CV < 30%), high (CV > 30%), and extremely high (CV > 40%) [38,42].

3. Results and Discussions

3.1. Performance Evaluation of CMIP6 Climate Model Data

The model’s performance was examined after bias correction of the seven daily time series CMIP6 models with respect to measured climate data for replicating precipitation, maximum, and minimum temperatures over UBNB for climate projection. For this study, distribution mapping and power transformation were the bias adjustment methods used to correct the historical simulations of rainfall and temperature from GCMs for the period 1995–2014. The performance of CMIP6 models under different Shared Socioeconomic Pathways (SSPs) was assessed in estimating future climate variables (precipitation and temperature) in the UBNB based on “hydrogof” under R software package version 4.4.1. The statistical methods used to evaluate the performances included the Pearson correlation coefficient (r), mean error (ME), the root mean square error (RMSE), and percent of bias (PBias). The table below (

Table 2. Performance evaluation of selected CMIP6 models for monthly precipitation (1995–2014). The bold values indicate the best-performing model for future projection and statistical analysis.

Climate Zones	GCMs	ME	MAE	RMSE	r	R2	PBias
Lowland	BCC-CSM2-MR	−2.4	29.9	60.1	0.87	0.72	−2.5
	CMCC-ESM2	0.00	31.8	55.2	0.89	0.77	0.00
	INM-CM4-8	0.00	28.9	48.5	0.91	0.82	0.00
	INM-CM5-0	0.00	28.8	47.3	0.91	0.83	0.00
	MPI-ESM1-2-HR	0.00	36.7	61.7	0.86	0.71	0.00
	MRI-ESM2-0	0.00	32.7	55.3	0.89	0.77	0.00
	NORESM2-MM	0.00	39.7	68.0	0.84	0.65	0.00
Midland	BCC-CSM2-MR	−2.5	26.9	50.3	0.91	0.80	−2.3
	CMCC-ESM2	0.00	30.5	43.4	0.93	0.85	0.00
	INM-CM4-8	0.00	26.8	39.1	0.94	0.88	0.00
	INM-CM5-0	0.00	26.7	38.3	0.94	0.88	0.00
	MPI-ESM1-2-HR	0.00	30.6	45.8	0.92	0.83	0.00
	MRI-ESM2-0	0.00	29.5	43.5	0.93	0.85	0.00
	NORESM2-MM	0.00	38.5	60.3	0.88	0.72	0.00
Highland	BCC-CSM2-MR	−2.5	35.1	58.3	0.89	0.75	−2.3
	CMCC-ESM2	0.00	38.9	55.6	0.89	0.77	0.00
	INM-CM4-8	0.00	35.8	52.8	0.90	0.79	0.00
	INM-CM5-0	0.00	33.4	48.6	0.91	0.82	0.00
	MPI-ESM1-2-HR	0.00	36.6	54.9	0.89	0.78	0.00
	MRI-ESM2-0	0.00	37.7	55.8	0.89	0.77	0.00
	NORESM2-MM	−0.01	54.1	88.2	0.79	0.44	0.00

Bold values indicate best performed CMIP6 model for future projection and statistical analysis.

) shows that for precipitation, the highest-ranking GCM was INM-CM5-0, while INM-CM4-8 ranked highest for both maximum and minimum temperatures across all agro-ecological zones. Consistent with this finding, Gebisa et al. [43] in the Baro River Basin, Ethiopia, also selected INM-CM5-0 for precipitation projections based on observed data and comparisons with historical GCMs. Conversely, the lowest-ranking GCM for precipitation across all agro-ecological zones was BCC-CSM2-MR. In contrast, Alaminie et al. [12] found BCC-CSM2-MR to be the best-performing model for precipitation projection among the 12 GCMs evaluated for the entire UBNB. This discrepancy may be due to the current study’s use of agro-climatic zone-based classification.

For precipitation, the INM-CM5-0 model shows better performance with statistical values of r = 0.91 and R² = 0.83 in the lowland part of the study area. Similarly, in the midland part, the INM-CM5-0 model recorded the highest statistical values of r = 0.94 and R² = 0.88, along with the lowest mean error values. In the highland part of UBNB, the model achieved the highest values of r = 0.91 and R² = 0.82 (Table 2). The values are overestimated in the model when the PBIAS value is positive, while they are under-estimated when it is negative. Consequently, the BCC-CSM2-MR model shows the highest bias, with values ranging from −2.3% to −2.5%, indicating that it underestimates monthly precipitation and has a larger percentage bias compared to other GCMs in the study area. Additionally, smaller absolute values of RMSE and Pbias indicate better model performance. RMSE measures the deviations between measured and modeled climate variables, with a value of zero representing the best performance. All agro-climatic zones exhibited relatively small extreme errors and greater central tendency, with RMSE values of 1.7 mm, 0.99 mm, and 1.1 mm for monthly time steps, respectively. For maximum temperature, the INM-CM4-8 model performed with r = 0.85 and R² = 0.7 in the lowland, r = 0.89 and R² = 0.77 in the midland, and r = 0.86 and R² = 0.72 in the highland parts of the study area (

Table 3. Performance evaluation of selected CMIP6 models for maximum temperature (1995–2014). The bold character indicates well-performed model for future projection and statistical analysis.

Climate Zones	GCMs	ME	MAE	RMSE	r	R2	PBias
Lowland	BCC-CSM2-MR	−0.11	1.4	2.3	0.75	0.44	−0.3
	CMCC-ESM2	0.00	1.4	1.8	0.83	0.67	0.00
	INM-CM4-8	0.00	1.3	1.7	0.85	0.70	0.00
	INM-CM5-0	0.00	1.4	1.7	0.84	0.70	0.00
	MPI-ESM1-2-HR	0.00	1.6	2.0	0.80	0.58	0.00
	MRI-ESM2-0	0.00	1.4	1.8	0.83	0.67	0.00
	NORESM2-MM	0.00	1.4	1.8	0.83	0.66	0.00
Midland	BCC-CSM2-MR	−0.1	0.78	1.85	0.71	0.19	−0.40
	CMCC-ESM2	0.00	0.82	1.0	0.88	0.75	0.00
	INM-CM4-8	0.00	0.77	0.99	0.89	0.77	0.00
	INM-CM5-0	0.00	0.81	1.1	0.88	0.76	0.00
	MPI-ESM1-2-HR	0.00	0.92	1.2	0.84	0.66	0.00
	MRI-ESM2-0	0.00	0.71	0.93	0.90	0.80	0.00
	NORESM2-MM	0.00	0.86	1.1	0.87	0.73	0.00
Highland	BCC-CSM2-MR	−0.1	0.86	1.85	0.70	0.17	−0.40
	CMCC-ESM2	0.00	0.90	1.1	0.85	0.69	0.00
	INM-CM4-8	0.00	0.84	1.1	0.86	0.72	0.00
	INM-CM5-0	0.00	0.92	1.2	0.85	0.69	0.00
	MPI-ESM1-2-HR	0.00	0.92	1.2	0.84	0.66	0.00
	MRI-ESM2-0	0.00	0.85	1.1	0.86	0.71	0.00
	NORESM2-MM	0.00	0.88	1.2	0.85	0.70	0.00

Bold values indicate best performed CMIP6 model for future projection and statistical analysis.

). BCC-CSM2-MR underestimates the monthly maximum temperature in the study area. In contrast, Alaminie et al. [12] reported that MRI-ESM2-0 showed the highest agreement with baseline period observational datasets over the Upper Blue Nile Basin.

In addition, the performance metrics for minimum temperature were calculated for all GCMs, and the rankings of the GCMs were derived using different statistical parameters. All models exhibited a positive correlation with observed values, except for BCC-CSM2-MR, which revealed a negative correlation. For minimum temperature, the INM-CM4-8 model showed the best performance with r = 0.62 and R² = 0.31 in the lowland, r = 0.72 and R² = 0.41 in the midland, and r = 0.76 and R² = 0.56 in the highland parts of the study area (

Table 4. Performance evaluation of selected CMIP6 models for minimum temperature (1995–2014).

Climate Zones	GCMs	ME	MAE	RMSE	r	R2	PBias
Lowland	BCC-CSM2-MR	−0.1	1.3	2.2	0.46	0.3	−0.5
	CMCC-ESM2	0.00	1.5	1.9	0.48	0.34	0.00
	INM-CM4-8	0.00	1.2	1.5	0.62	0.31	0.00
	INM-CM5-0	0.00	1.2	1.6	0.58	0.23	0.00
	MPI-ESM1-2-HR	0.00	1.6	2.1	0.39	0.27	0.00
	MRI-ESM2-0	0.00	1.2	1.7	0.55	0.17	0.00
	NORESM2-MM	0.00	1.3	1.7	0.53	0.00	0.00
Midland	BCC-CSM2-MR	−0.1	0.67	1.56	0.49	0.23	−0.70
	CMCC-ESM2	0.00	0.79	0.98	0.6	0.37	0.00
	INM-CM4-8	0.00	0.59	0.75	0.72	0.41	0.00
	INM-CM5-0	0.00	0.59	0.77	0.72	0.37	0.00
	MPI-ESM1-2-HR	0.00	0.79	1.04	0.58	0.14	0.00
	MRI-ESM2-0	0.00	0.64	0.81	0.68	0.30	0.00
	NORESM2-MM	0.00	0.83	1.01	0.59	0.32	0.00
Highland	BCC-CSM2-MR	−0.1	0.87	1.6	0.61	0.34	−0.9
	CMCC-ESM2	0.00	0.81	1.0	0.73	0.40	0.00
	INM-CM4-8	0.00	0.71	0.88	0.79	0.56	0.00
	INM-CM5-0	0.00	0.73	0.88	0.79	0.56	0.00
	MPI-ESM1-2-HR	0.00	0.83	1.1	0.71	0.31	0.00
	MRI-ESM2-0	0.00	0.74	0.93	0.77	0.51	0.00
	NORESM2-MM	0.00	0.89	1.1	0.68	0.27	0.00

Bold values indicate best performed CMIP6 model for future projection and statistical analysis.

). Consistent with this finding, Gebisa et al. [43] reported that INM-CM4-8 performed well for minimum temperature projections. In contrast, a study by Shiru and Chung [31] found that INM-CM4-8 was ranked 5th in simulating minimum temperature among 13 GCMs evaluated. In terms of PBias, the BCC-CSM2-MR model underestimates (negatively biases) the monthly minimum temperature across all agro-climatic zones. Furthermore, the study showed that the total percentage of biases was slightly higher for minimum temperature compared to maximum temperature. This finding agrees with Alemu et al. [24] and Gebisa et al. [43], who found that the overall bias was higher for minimum temperatures compared to maximum temperatures. Generally, the two best-performing and top-ranked models for temperature and rainfall variables recorded the lowest mean error (ME), mean absolute error (MAE), root mean square error (RMSE), and a percentage bias of zero across all agro-climatic zones.

3.2. Assessment of Model Performance Using Taylor Diagram

In addition, (Figure S1) shows the abilities of GCMs for precipitation, maximum temperature, and minimum temperature in lowland (a1–a3), midland (b1–b3), and highland (c1–c3) areas, respectively, relative to observed data. Although the standard deviations differ, the correlation between the modeled and observed values ranges from 0.4 to 0.95 (Figure S1). The INM-CM5-0 model exhibited the smallest distance, indicating the highest correlation (>0.9) and the lowest standard deviation compared to observed data for precipitation. Meanwhile, the INM-CM4-8 model demonstrated the highest correlation (>0.7) for both maximum and minimum temperatures across all agro-ecological zones. In contrast, BCC-CSM2-MR models displayed either larger standard deviations or lower correlation coefficients, indicating poor performance in capturing periodical change in temperature and precipitation over the study area (Sub.1). The highest-ranking GCMs were found to have higher correlations. Consequently, these well-performing models were used in this study for temporal and spatial projections of precipitation and temperature analysis.

Furthermore, the top-performing GCMs in a certain research area or region did not always rank similarly in other areas. Performance evaluation is crucial for a comprehensive examination of climate models. In conclusion, the INM-CM5-0 and INM-CM4-8 models produced the lowest RMSE, with smaller biases and good correlations (>0.7) for precipitation and temperature variables, respectively. Additionally, after selecting the well-performing models, climate projections, extreme indices, and the starting and ending of the rainy season were analyzed from 1995 to 2100 for precipitation and temperature at all grid points under the socioeconomic pathways SSP1-2.6, SSP2-4.5, and SSP5-8.5 over the UBNB.

3.3. Temporal Distribution of Observed and Future GCMs

3.3.1. Temporal Distribution of Precipitation

The temporal distribution of observed (1995–2014) and CMIP6 model (2015–2100) precipitation in the lowland, midland, and highland areas of the study region is depicted in Figure 4. The observed maximum rainfall in the study area reached 2877 mm in 1998 in the midland region (Figure 4). Conversely, a minimum observed rainfall of 891 mm was recorded in the lowland region in 2001. Under the SSP1-2.6 scenario, the maximum and minimum future precipitation are projected to occur in 2098, with values of 2391.5 mm and 782.8 mm in the midland and lowland areas, respectively. Under the SSP2-4.5 scenario, the maximum and minimum future precipitation are projected for 2021 and 2017, with values of 2557.7 mm and 814 mm in the highland and lowland areas, respectively.

Similarly, in the SSP5-8.5 future scenario, a maximum precipitation of 2492 mm was observed in the midland parts of the study area in 2094, while a minimum precipitation of 1018.7 mm was recorded in the lowland parts in 2024. In general, the maximum precipitation was observed in the midland and highland regions, whereas the lowland areas exhibited lower precipitation levels across all observed and future scenarios. In this regard, Cheung et al. [44] concluded that the spatial patterns of precipitation and temperature in Ethiopia are influenced by altitude and diverse landscapes, which affects occurrences of drought.

3.3.2. Temporal Distribution of Temperature

Bias-corrected mean annual maximum temperatures for the observed and future periods over the study area are presented in Figure 5 under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios. The figure shows that the mean annual maximum temperature was highest in the lowland parts of the study area from 1995 to 2100 under all scenarios. In contrast, the highland parts of the study area consistently recorded the lowest mean maximum temperatures throughout the study period under all scenarios. Specifically, during the observed period, the maximum and minimum temperatures were recorded in 2008 and 1996, respectively, at 35.5 °C and 22.5 °C in the lowland and highland parts of the study area. Under the SSP1-2.6 scenario, the mean annual maximum temperature reached 36.25 °C in the lowland area in 2048. Similarly, under the SSP2-4.5 scenario, the highest and lowest mean annual maximum temperatures were recorded in 2089 and 2018, at 37.5 °C and 22.9 °C, respectively. Under the SSP5-8.5 scenario, the highest and lowest mean annual maximum temperatures were observed in 2096 and 2034 at 39.6 °C and 23.3 °C, respectively (Figure 5).

In addition, Figure 6 reveals the distribution of mean annual minimum temperatures across different agro-climatic zones of the study area under each scenario. For the observed period, the highest and lowest mean annual minimum temperatures were recorded in the lowland and highland parts of the study area in 2008 and 1999 at 19.5 °C and 8.6 °C, respectively. Under the SSP1-2.6 scenario, the highest mean annual minimum temperature was observed in the lowland area in 2073, while the lowest was recorded in the highland part of UBNB in 2026. Similarly, under the SSP2-4.5 scenario, the highest and lowest mean annual minimum temperatures were recorded in 2089 and 2033 in the lowland and highland parts of UBNB at 21.3 °C and 9.2 °C, respectively. Finally, under the SSP5-8.5 scenario, the highest and lowest mean annual minimum temperatures were registered in the lowland and highland parts in 2099 and 2016, respectively.

3.4. Spatial Distribution of Observed and Future Climate Variables

Precipitation and temperature are crucial climatic factors with significant impacts on the environment and socioeconomic activities globally, exhibiting varied geographical and temporal distributions [45,46]. Figure 7 illustrates the spatial distribution of observed (1995–2014) and future (2015–2100) projections of precipitation (INM-CM5-0) and temperature (INM-CM4-8) under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios. The spatial distribution of these climate variables was generated using the inverse distance weighted (IDW) interpolation technique under the ‘geo-statistics wizard’ in ArcGIS 10.8.

The highest increase in annual baseline precipitation was observed in the midland/central region of the basin. Similarly, the midland and southwestern parts of the study area recorded the maximum precipitation across all scenarios, while the lowest precipitation was projected for the northern and northeastern parts of the basin. Future precipitation projections indicate a slight increase in precipitation across all scenarios. Alaminie et al. [12] also concluded that future precipitation simulations confirms the availability of moist conditions across the basin in each future scenarios. The spatial distribution of precipitation reflects variations with topography: lowland areas receive less precipitation, whereas midland and highland areas receive more. This finding is consistent with NMA [47], which notes that rainfall patterns in Ethiopia are influenced by meteorological systems such as the Somali Jet, Red Sea Convergence Zone (RSCZ), Subtropical Jet (STJ), Tropical Easterly Jet (TEJ), and Intertropical Convergence Zone (ITCZ). These systems collectively influence Ethiopia’s precipitation patterns, with their rates, locations, directions, and variability explaining variations in rainfall amount and distribution.

Figure 7 also shows the spatial distribution of maximum temperatures for observed and future projections under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 climate scenarios. Under all scenarios, the mean annual maximum temperature is expected to rise relative to the historical baseline. This finding aligns with Alaminie et al. [12], who noted enhanced warming over the UBNB under all future scenarios, with the warming being more pronounced in the northeastern parts of the basin relative to the southwestern parts. The highest annual average maximum temperatures were observed in the lowland areas under each scenario, while the lowest temperatures were projected for the midland and highland parts. Lower elevations, characterized by warmer temperatures, contrast with the cooler temperatures found in the higher elevations of the midland and highland areas.

Similarly, the spatial distribution of observed and projected minimum temperatures shows an increase in annual average temperatures under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios. The highest increases in annual average minimum temperature are projected for the lowland areas, whereas the lowest minimum temperatures are anticipated in the highland regions. The lowest elevated parts of the study area record warmer temperatures compared to the higher elevations around the highlands in each scenario. This suggests that geography significantly affects temperature. There is an inverse relationship between altitude and temperature, according to Gurara et al. [48], with high temperatures common in low-altitude regions whereas low temperature is common in high altitude areas. The hydrological cycle will be impacted by the anticipated rise in temperatures and variations in precipitation, which will have an impact on basin-wide planning and development aspects.

3.5. Future Trends and Variability Analysis of Precipitation and Temperature

According to Yue and Wang [40], trend analysis was a valuable technique for effective water resource management and planning. After bias adjustment of the CMIP6 models, a trend analysis of mean annual precipitation and temperature was conducted using the Mann–Kendall (MK) trend test (to assess the significance of the trend) and Sen’s Slope estimator (to determine the magnitude of the trend). This analysis was performed in R Studio using the ‘modifiedmk’ package across the study area. A positive value from the test indicates an increasing trend, whereas a negative value indicates a falling trend. The assessments were conducted for three time slices: near (2015–2044), mid (2045–2074), and far (2075–2100) future periods, using the scenarios SSP1-2.6, SSP2-4.5, and SSP5-8.5 for each agro-climatic zone in the UBNB.

3.5.1. Future Precipitation Trends and Variability Analysis

Table S2 presents the results of the Mann–Kendall (MK) and Sen’s slope trend tests for precipitation outputs from the INM-CM5-0 model across each climatic zone under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios. For the near future period (2015–2044), both MK and Sen’s slope trend tests indicate a significant positive trend (p < 0.01) in precipitation across all agro-climatic zones under each scenario (Table S2). The results also show an increasing positive trend in all SSPs, as indicated by the Z statistics value, suggesting that precipitation will rise in the near future under all SSPs in the study area.

In the mid-future period (2045–2074), there is an insignificant positive trend in the lowland parts of the basin under the SSP1-2.6 scenario, while a significant positive trend (p < 0.05) is observed in the midland and highland parts. Under the SSP2-4.5 scenario, an insignificant decreasing trend is recorded in the lowland and midland areas, while a significant (p < 0.01) decreasing trend is noted in the highland parts of the study area. In the far future period (2075–2100), an insignificant decreasing trend is observed in the lowland and highland parts of the basin under the SSP1-2.6 scenario, while an insignificant increasing trend is noted in the midland part. This finding aligns with Alemu et al. [24], who found a decreasing trend in the lower part of the Awash basin for the far future period.

Under the SSP2-4.5 scenario, there is an insignificant positive trend across all agro-climatic zones. In contrast, the SSP5-8.5 scenario shows a significant increase in precipitation (p < 0.01) in the midland and highland parts of the study area, while the lowland parts exhibit an insignificant positive trend. Overall, from 2015 to 2100, the precipitation trend shows a significant decrease in the lowland and highland parts under the SSP1-2.6 and SSP2-4.5 scenarios, with an insignificant negative trend in the midland part. Conversely, under the SSP5-8.5 scenario, precipitation increases significantly across all agro-climatic zones at a p < 0.01 significance level. This finding is consistent with Alaminie et al. [12], who observed an increasing trend in precipitation under all SSPs across the entire UBNB. Similarly, Gebrechorkos et al. [49] found that precipitation in Ethiopia is expected to rise in the 2050s and 2080s, with the western part of the country becoming much wetter compared to the baseline period. The IPCC’s fifth assessment report (IPCC, 2014) and Stocker et al. [50] also project that global-scale precipitation will gradually increase, with some regions experiencing increases, while others may see decreases or minimal changes.

In addition to analyzing precipitation trends, it is important to assess the variability of mean annual precipitation under different scenarios in each agro-climatic zone of UBNB. The coefficient of variation (CV) measures the proportion of the standard deviation to the mean. A higher CV indicates greater dispersion around the mean, represented as a percentage.

Commonly, CV values are classified as follows: low (CV < 10%), moderate (10% < CV < 20%), high (20% < CV < 30%), and very high (CV > 30%) [38,42]. Hence,

Table 5. Coefficient of variation (CV) for model INM-CM5-0 precipitation output in each climate zone under different scenarios from 2015 to 2100.

Coefficient of Variation (CV (%)) from 2015 to 2100
	SSP1-2.6	SSP2-4.5	SSP5-8.5
Lowland	14.2	14.5	16.2
Midland	15.8	11.2	15.0
Highland	15.6	17.2	18.4

presents the coefficient of variation (CV) values for future projections of mean annual rainfall under different scenarios for each agro-climatic zone of UBNB from 2015 to 2100. The result shows that precipitation in the study area varies moderately (CV < 20%) under all scenarios. Additionally, CV values increase from the SSP1-2.6 to the SSP5-8.5 scenarios in all zones of UBNB, except in the midland parts (Table 5).

3.5.2. Trend Analysis for Maximum Temperature

The trends in maximum temperatures over the UBNB are shown in (Sub.2). The projections indicate an increasing trend in maximum temperatures under both SSPs, except for the SSP1-2.6 scenario during the mid- and far future periods. However, temperatures projected under SSP5-8.5 are higher than those projected under SSP1-2.6 and SSP2-4.5. The maximum temperature for the entire basin under the SSP1-2.6 scenario shows an insignificant decreasing trend during the mid- and far future periods, except for a significant decrease in the highland part of the study area during the mid period. As reported by the IPCC [1] and aligned with Pradhan et al. [51], most geographical regions have seen increases in the frequency, intensity, and length of heat-related events, such as heatwaves, as a result of global warming. Many parts of Africa have also seen increases in the frequency and severity of droughts. Additionally, there has been a worldwide upsurge in the intensity of heavy precipitation occurrences.

3.5.3. Trend Analysis for Minimum Temperature

In addition, the minimum temperature trend shown in (Sub.3) reveals a positive trend with a significant rise in scenario SSP2-4.5 and SSP5-8.5 under all future periods. Similarly, a significant (at p < 0.01) increasing trend was observed in the near and far future periods in the SSP1-2.6 scenario, while in the far future period, lowland and midland parts of UBNB show a significant (at p < 0.05) decreasing trend. On the other hand, in the entire study period (2015–2100), there is an insignificant positive trend in minimum temperature across all agro-climatic zones. As per Gabisa et al. [43] and Moges and Bhat [52], we point out that the difficulties to agricultural output and food security posed by highest minimum temperatures include crop heat stress, poorer yields, disrupted pollination trends, and intensified pest and disease pressure. In conclusion, in line with [1,51], many parts of the world have seen changes in their climatic zones due to rising temperatures, including the extension of dry climate zones and the reduction in polar climate zones. Consequently, this has led to modifications in the categories, distributions, and seasonal movements of numerous plant and animal species. Because of global warming, a shift in precipitation and an increase in the frequency of some extreme events, food security has already been impacted by climate change. In many lower-latitude regions, observed modifications to the climate have had a detrimental effect on crop yields of some crops (such as wheat and maize), whereas across higher-latitude zones, these changes have a positive impact on agricultural yields of some crops in recent decades.

In a similar vein, the projected mean annual precipitation in the study area was predicted to increase by 5.6 mm (0.44%) under the SSP1-2.6 scenario, 8.7 mm (0.68%) under the SSP2-4.5 scenario, and 146.5 mm (11.5%) under the SSP5-8.5 scenario for the entire future study period (2015–2100) compared with the baseline (historical) (

Table 6. Projected changes in average precipitation (mm& %), Tmax and Tmin (°C) in baseline, near, mid, far, and long periods under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios (1995–2100).

INM-CM5-0 Mean Precipitation (mm/Year)
Time	Scenarios
	SSP1-2.6	SSP2-4.5	SSP5-8.5
Baseline (1995–2014)	1274.8	1274.9	1274.8
2015–2044 (near)	1342.4	1298.6	1375.0
2045–2074 (mid)	1285.3	1279.6	1453.0
2075–2100 (far)	1157.4	1277.4	1556.1
2015–2100 (long)	1269.2	1283.6	1421.3
Change in near future (mm/%)	67.6/5.3	23.7/1.86	100.2/7.86
Change in mid-future (mm/%)	10.5/0.8	4.7/0.37	178.2/13.98
Change in far future (mm/%)	−117.4/9.2	2.5/0.17	281.3/22.1
Change in long-term future (2015–2100) (mm/%)	5.6/0.44	8.7/0.68	146.5/11.5
INM-CM4-8 Mean Max. Temperature (°C)
Baseline	26.4	26.5	26.5
2015–2044 (near)	28.5	28.5	28.8
2045–2074 (mid)	28.9	29.1	30.1
2075–2100 (far)	28.6	29.8	31.3
2015–2100 (long)	28.2	28.6	29.3
Change in near future (°C)	2.1	2.0	2.3
Change in mid-future (°C)	2.5	2.6	3.6
Change in far future (°C)	2.2	3.3	4.8
Change in long-term future (2015–2100) (°C)	1.8	2.1	2.8
INM-CM4-8 Mean Min. Temperature (°C)
Baseline	12.1	12.1	12.1
2015–2044 (near)	13.8	13.9	14.2
2045–2074 (mid)	14.1	14.9	16.2
2075–2100 (far)	13.8	15.3	17.9
2015–2100 (long)	13.6	14.2	15.2
Change in near future (°C)	1.7	1.8	2.1
Change in mid-future (°C)	2.0	2.8	4.1
Change in far future (°C)	1.7	3.2	5.8
Change in long-term future (2015–2100) (°C)	1.5	2.1	3.1

). Similarly, in the near, mid-, and far future time scales, the mean annual precipitation shows an increasing trend under each scenario. However, the precipitation projection trend is not uniform under SSP2-4.5. For instance, the computed results show a declining trend in mean annual precipitation of up to −117.4 mm (−9.2%) under SSP1-2.6 by the end of the century (Table 6). In line with these findings, Mengistu et al. [53] concluded that by the end of the century, an overall decrease in annual average rainfall to −10.8% at RCP4.5 and −19.0% under RCP8.5. The results also indicate that projected precipitation under SSP5-8.5 is higher than under SSP1-2.6 and SSP2-4.5 scenarios. In this regard, Almazroui et al. [54] identify similar trends in Northeast Africa, noting increased precipitation under SSP5-8.5, while SSP2-4.5 showed no uniform trend. Similar studies, such as Alaminie et al. [12], Gebisa et al. [43], and Getachew [55], have reported an upward trend in annual precipitation for future periods. Furthermore, Mekonnen [56] concluded that there is a positive trend in precipitation and temperature in UBNB based on five ensemble GCMs. Additionally, Moges and Bhat [52] noted that rainfall projections in the Rib watershed are estimated to rise for both RCPs. Hence, as concluded by Moges and Bhat [52], any alteration in the value, distribution, and patterns of rainfall will directly affect agricultural productivity and have a major effect on the lives of small-scale farmers in rural areas. This might trigger an increase in the frequency of flooding and soil erosion, which could be dangerous for the production of crops.

Likewise, this study demonstrates that the projected maximum temperature shows an increase in annual mean maximum temperature under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios across all future periods. The mean annual maximum temperature is expected to increase by 1.8 °C, 2.1 °C, and 2.8 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios, respectively, compared with the baseline (historical) from 2015 to 2100 in the UBNB. More specifically, by the end of the 21st century, the mean annual maximum temperature is projected to increase by 2.2 °C, 3.3 °C, and 4.8 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios, respectively. Alaminie et al. [12] reported similar findings, with the increasing trends in temperature over the UBNB (Abay) projected to be 1.3 °C, 1.7 °C, 2.0 °C, and 2.6 °C under SSP1-2.6, SSP2-4.5, SSP3-3.7, and SSP5-8.5 scenarios, respectively. For the near (long)-term period, projected warming under SSP1-2.6, SSP2-4.5, SSP3-3.7, and SSP5-8.5 scenarios is 1.5 °C, 2.2 °C, 2.8 °C, and 3.8 °C, respectively. In line with this finding, Tarekegn et al. [10], in north-western Ethiopia, concluded that the average maximum temperature is expected to rise from 26.5 °C in the past (1981–2010) to 27.8 °C in the future (2041–2070), indicating an increase of 1.3 °C when contrasting the past and future temperature. Similarly, Gebisa et al. [43] concluded that, in comparison to the baseline for the Baro River Basin, the average annual maximum temperature is predicted to rise by 1.4 °C and 1.8 °C under the SSP2-4.5 and SSP5-8.5 scenarios, respectively.

Additionally, under the RCP4.5 radiative forcing scenario, Getachew [55] suggested that maximum temperatures in Ethiopia’s Lake Tana sub-basin would rise by 1.38–3.59 °C by the 2080s. Similarly, Moges and Bhat [52] concluded that a highest warming trend across the Rib watershed was observed for both current and future scenarios. Overall, as reported by the IPCC Sixth Assessment Report and in line with Lee et al. [57], over the last 2000 years, the rate of increase in the global average surface temperature has accelerated since 1970 more than it has during any previous 50-year span. Increased temperatures have a number of important effects. For instance, variations in precipitation patterns are attributed the general increase in temperature [55]. As per the study by Gurara et al. [48] in the Awash basin, anticipated changes in future temperature and precipitation are expected to cause higher precipitation intensities, more rainy days, and longer dry spells. Similarly, Gebisa et al. [43] noted that as temperatures increases, changes in the atmosphere might alter rainfall patterns and lead to more intense rainfall events.

On the other hand, like maximum temperatures, projected minimum temperatures are expected to show increases in annual average minimum temperature under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios. In the long term (2015–2100), the annual average minimum temperature is estimated to increase by 1.5 °C, 2.1 °C, and 3.1 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios, respectively. Similarly, in the near future (2015–2044), the mean annual minimum temperature is projected to rise by 1.7 °C, 1.8 °C, and 2.1 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios, respectively. In the mid-future period (2045–2074), the minimum temperature is estimated to increase by 2.0 °C, 2.8 °C, and 4.1 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios, respectively. Similarly, in the far future period (2075–2100), the minimum temperature is expected to increase by 1.7 °C, 3.2 °C, and 5.8 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios, respectively. Additionally, in the SSP2-4.5 and SSP5-8.5 scenarios, the rise in the minimum temperature is expected to be greater than the increase in maximum temperature. In this regard, Gebisa et al. [43] and Getahun et al. [58] concluded that the increase in minimum temperature under SSP2-4.5 and SSP5-8.5 scenarios was generally higher than the maximum temperature increase in the Awash and Baro river basins. Similarly, the recent IPCC Sixth Assessment Report IPCC [1], in line with Pradhan et al. [51] and Almazroui et al. [54], concluded that overall increasing temperature trends were observed in Northeast Africa. In addition, Stocker et al. [50] reported that global mean temperatures will continue to rise over the 21st century under all RCPs. Generally, as concluded by Gebisa et al. [43], IPCC [1,59,60], the future rise in temperatures and anticipated variations in precipitation will affect the hydro-climatic conditions, hence affecting the future development and planning of the basin. Additionally, variations in the patterns and trends in rainfall, along with rising temperatures, could have direct implications in rain-fed agriculture activities and the livelihoods of the farming communities [43,52,61].

4. Conclusions

Global Climate Models (GCMs) have contributed to the recent phase of the Coupled Model Inter-comparison Project (CMIP6). This study aims to assess the performance of CMIP6 climate models and their projections of precipitation and temperature over the Upper Blue Nile Basin in Northwestern Ethiopia. This study evaluates the effectiveness of seven CMIP6 climate model products using statistical parameters. Mann–Kendall and Sen’s Slope tests were computed to analyze the past and future trends in rainfall and temperature variables.

The results reveal that the INM-CM5-0 model (for precipitation) and the INM-CM4-8 model (for maximum and minimum temperatures) performed well across all assessment parameters in the UBNB. This study also shows that precipitation projections in all agro-climatic zones under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios exhibit a significant (p < 0.01) positive trend. The overall percentage change in the projected mean annual precipitation in the study area is predicted to increase by 5.6 mm (0.44%) under SSP1-2.6, 8.7 mm (0.68%) under SSP2-4.5, and 146.5 mm (11.5%) under SSP5-8.5 scenarios over the entire future study period (2015–2100) compared to the baseline (historical) period.

Additionally, the mean annual maximum temperature over UBNB is projected to increase by 1.8 °C, 2.1 °C, and 2.8 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios, respectively, from 2015 to 2100. Similarly, the mean annual minimum temperature is estimated to increase by 1.5 °C, 2.1 °C, and 3.1 °C under SSP1-2.6, SSP2-4.5, and SSP5-8.5, respectively. The projected changes in maximum and minimum temperatures will be warmer than the baseline period, particularly under SSP5-8.5 compared to SSP1-2.6 and SSP2-4.5 in the study area. These significant changes in climate variables contribute to altering the occurrences and severity of extreme events. Hence, stakeholders and farming communities should adopt various adaptation practices such as afforestation and reforestation campaigns and resource conservation mechanisms to address the impacts of warming temperatures.