Abstract
This study evaluated the performance of five Global Climate Model (GCM) outputs from the Coupled Model Intercomparison Project phase 6 (CMIP6) in reproducing the historical precipitation and temperature. Observational data from the National Meteorological Agency are used for model evaluation and bias correction. Then, the projections from representative GCMs are used to understand the future climate (2031–2060) of the Baro River Basin under two Shared Socioeconomic Pathways (SSP2-4.5 and SSP5-8.5) with respect to the historical datasets (1985–2014). Statistical metrics (percent of bias, root mean square error, and coefficient of determination) are used to assess the model's performance in reproducing precipitation and temperature and Compromise Programming (CP) was used in ranking GCMs. GFDL-CM4, INM-CM5-0, and INM-CM4-8 models for precipitation; CMCC-ESM2, MRI-ESM2-0, and INM-CM4-8 for maximum temperature; and GFDL-CM4, INM-CM4-8, and INM-CM5-0 for minimum temperature were selected based on their better simulation. The projected annual precipitation shows increases of 6% under SSP2-4.5 and 16.46% under SSP5-8.5. The mean annual maximum and minimum temperature show increases of 1.43 and 1.96 °C under SSP2-4.5, and 1.81 and 3.11 °C under SSP5-8.5, respectively. Overall, the ensemble of three models outperforms the ensemble of all models for the Baro River Basin when utilising the representative GCMs.
HIGHLIGHTS
Multi-model GCM outputs help in reducing model discrepancies.
Multi-criteria evaluation enables the development of proactive approach to facilitate evaluation.
Relying solely on the ensemble of all models is not suitable for climate impact studies.
Careful model selection in climate impact studies should be emphasized.
Comprehensive understanding of climate dynamics helps to explore the future outlook of regions.
INTRODUCTION
Rapid global warming has hastened global and regional climate change, changing the severity and frequency of extreme climatic events with far-reaching consequences on natural ecosystems, human development and significant economic losses each year (Chen et al. 2020; Srivastava et al. 2020; Ge et al. 2021; Wu et al. 2021). Understanding the intensity, frequency, and spatial extents through reliable future projections of climate changes is a pre-requisite to make reliable climate change adaptation planning (Kim et al. 2020). Due to its low adaptability and resilience to natural disasters, climate changes have serious implications for developing countries such as Africa. Moreover, weak adaptive capacity due to limited access to information, technology, finance, and capital assets, intensifies the impacts of climate change in developing countries (Sylla et al. 2016). According to IPCC (2013), in many parts of Africa, changes in rainfall and temperature are causing fluctuations in freshwater systems, affecting the quality and quantity of water accessibility.
Ethiopia relies primarily on available water resources from the highlands and rain-fed agriculture, but much of the south and east region is very dry and prone to drought and desertification (Roth et al. 2018). Over the past 55 years, it has been known that there was a warming trend in the annual minimum temperature ranges, increasing by about 0.37 °C every decade (NMA 2007). The climate changes studies in different parts of Ethiopia revealed that climate change-induced changes in seasonal and annual hydrological variables show increasing stress on water resources availability (Melke 2015; Shiferaw et al. 2018; Taye et al. 2018; Worku et al. 2018; Dibaba et al. 2020; Mengistu et al. 2020). However, different studies reported varying magnitudes of the impacts. For example, Mengistu et al. (2020) and Dibaba et al. (2020) reported increasing mean annual temperature and decreasing precipitation in the Upper Blue Nile river basin and Finchaa catchment, respectively. Temperature increases have also been reported in the Ilala watershed, but there has been no significant change in precipitation (Shiferaw et al. 2018). According to Melke (2015), the monthly and seasonal temperature and precipitation in the future time horizons did not show a systematic increase or decrease in Gumara Catchment, Lake Tana Basin.
Despite being the main water resource for Nile basin riparian countries, the Baro River Basin is facing immense pressure and different competitive uses. Moreover, the Baro River Basin is subject to high climate variability with frequent changes in hydrology as a result of climate change (Mengistu et al. 2021; Muleta 2021). Variations in the hydrological process due to climate change-induced extreme events and water scarcity limits agricultural production and the community livelihood in the basin and the country at large. For example, using the A1B emission scenario, Muleta (2021) reported an increase in temperature, but precipitation does not manifest a systematic increase or decrease. Mengistu et al. (2021) reported increasing temperature and decreasing annual precipitation trends for the same basin using bias-corrected CORDEX-RCMs (Regional Climate Models). Overall, climate projections indicated that current water stress challenges will be exacerbated under a warming climate. All scenarios projected warmer temperature, but the studies have shown discrepancies in precipitation predictions. This could be due to the nature of precipitation, type of climate models and characteristics of climate scenarios adopted. One mechanism for reducing climate model discrepancies is the use of multi-model GCMs (Dibaba et al. 2019). However, using multi-model GCMs requires a proper evaluation of the model's performance before identifying representative GCMs for a particular location (Kim et al. 2020). It is therefore critical to assess, identify and map the long-term climate projections using state-of-the-art climate modeling with the latest datasets to aid the development of water resources in the basin.
Projections of global climate variables have been carried out using General Circulation Models, which provide projections at large spatial scales. Jaagus & Mändla (2014) stated that greenhouse gas concentrations are used as input statistics for the GCM. Over the years, many of the GCMs have been developed for various IPCC assessment report scenarios, including phases 3 and 5 of the Coupled Model Intercomparison Project (CMIP) and the recently released phase 6. The main differences between the CMIP6 simulations and the previous CMIP phases (CMIP3 and CMIP5) are the future scenario start years and new sets of specifications for concentration, emission, and land-use scenarios (Gidden et al. 2019). Moreover, CMIP6 represents a major increase in terms of the number of modeling groups involved, the number of future scenarios considered, and the number of diverse experiments conducted (Chen et al. 2020). From GCMs to Earth System Models, CMIP6 models have a wider range of complexity due to improved physical processes and higher spatial resolution (Eyring et al. 2016). The Scenario Model Intercomparison Project (Scenario MIP) is the primary activity of CMIP6 and is based on alternative scenarios of future emissions and land-use change generated by integrated assessment models.
Climate change projections are critical for improving the understanding of the climate system as well as characterizing societal risks and response options (O'Neill et al. 2016). To explore the impacts of climate change, adaptation, and mitigation, CMIP6 utilizes a new scenario framework combining future radiative forcing and associated climate changes with alternative socioeconomic development pathways (SSPs) (O'Neill et al. 2014). In conclusion, CMIP6 is preferred over CMIP5 due to a number of characteristics that improve climate projections. A wide range of enhancements provided by CMIP6 includes increased resolution, broader parameterization, and updated emission scenarios. These advancements considerably bolster the precision and comprehensiveness of climate forecasting. Moreover, CMIP6 incorporated the most recent generation of climate models, integrating the latest insights into climate processes and featuring improved representations of physical, chemical, and biological interactions. Additionally, CMIP6 introduces a broader spectrum of Earth system components model processes, including land-use changes, biogeochemical cycles, and ice-sheet dynamics. The incorporation of these novel elements amplifies the models’ capacity to capture intricate interactions and feedback mechanisms within the Earth system. Consequently, the projections generated by CMIP6 encompass a more comprehensive understanding of the complex dynamics governing our planet.
A number of studies have evaluated the performance of CMIP6 climate model outputs using GCM simulations (eg. Karim et al. 2020; Afolayan et al. 2021; Babaousmail et al. 2021; Ngoma et al. 2021; Shiru & Chung 2021; Zhang et al. 2021; Zhou et al. 2021). However, being recently released, CMIP6 is not known to have been applied in the Baro River Basin. On the other hand, there have only been a few studies on climate projection in the Baro River Basin and the analysis of previous studies was not based on recent climate model projections. Some of the climate change studies in the basin were projected based on climate scenarios A1B (balanced scenario group of storyline A1) and B1 (intermediate level of the storyline) (Kebede et al. 2013; Muleta 2021). Although increasing temperature is reported by the studies, the studies have shown discrepancies in precipitation predictions. In the previous climate scenarios, describing the developments of anthropogenic drivers for the possible future climate change with socioeconomic development is inattentive. The studies are also based on the application of a single climate model without any performance evaluation among the alternative climate models. It is true that most of the climate change studies in Ethiopia are based on either a single GCM–RCM model or do not involve model evaluations before using GCM–RCM for climate change and impact assessments as also used in the Baro Akobo River basin. We believe that one mechanism for reducing climate model discrepancies is the use of multi-model GCMs with the most recent state-of-the-art climate models available using proper evaluation of the individual and ensemble models. However, using multi-model GCMs requires a proper evaluation of the model's performance before identifying representative GCMs for a particular location. In this regard, comprehensive investigation is required to establish the improvements of CMIP6 models in simulating precipitation and temperature over various spatio-temporal scales and the selection of climate models for climate projections and impact studies requires assessment of multi-model performance.
The new approach used in this study, the use of multi-model output using multi-criteria evaluation, has a very important contribution for proper establishment of comprehensive assessments of multiple GCMs under various scenarios in Ethiopia where only limited information is available to develop proper strategic information in understanding current and future climate projections for a better adaptive capacity and mitigations. The contribution could also be related to evaluating the models’ dependability, understanding their strengths and limitations, or identifying the most appropriate model for a specific application. The new set of simulations will also help to evaluate key aspects of society with a comprehensive potential climates that would imply a range of challenges in mitigating and adapting to climate change and socioeconomic outcomes. Consequently, this research emphasized on developing a more comprehensive understanding of the regional climate governing the region through careful model selection and evaluation which also help to explore the future outlook of the basin.
Based on the five GCMs obtained from the CMIP6 model, the objectives of the current study were to: (a) evaluate the capability of the CMIP Phase 6 (CMIP6) based Global Climate Models (GCMs) to accurately simulate precipitation, maximum temperature, and minimum temperature in the Baro River Basin; (b) evaluate the effectiveness of bias-corrected GCM simulations in replicating observed precipitation, as well as maximum and minimum temperature patterns; (c) quantify the projected alterations in future precipitation, maximum temperature, and minimum temperature.
MATERIALS AND METHODS
Study area
The basin is bordered by Sudan to the west, the Upper Blue Nile basin to the north and northeast, and the Akobo River to the east and southeast. The Baro River is formed by the confluence of the Birbir and Gebba Rivers east of Metu in the Illuababor zone of the Oromia region. The study area is covered 23,500 km2 with the outlet at the Gambella station. Mean monthly precipitation in the basin varies from about 23 mm in February to more than 300 mm in August, with two distinct seasons. From April to October is the rainy season, and from November to March is the dry season. The average monthly maximum temperature ranges from 23 °C in July to 30 °C in March, while the minimum temperature in the basin ranges from 12 °C in June to 14 °C in January
Datasets
Observed data
Daily meteorological data were collected from National Meteorological Agency (NMA) and used to evaluate the GCMs model output data before using the GCMs simulations for future projection of climate change. For undertaking climate analysis, complete datasets with no gaps or missing information are typically required (Dibaba et al. 2019). In this respect, rainfall stations with a significant number of missing values are excluded from the evaluation process. Eleven stations were chosen for precipitation data, while seven stations were selected for maximum and minimum temperature, as indicated in Table 1. In both cases, all the selected stations have missing values of less than 10% for the records. In this research, the missing meteorological data were filled with the Markov Chain Monte Carlo (MCMC) imputation technique using XLSTAT. The quality and consistency of the meteorological data were checked using the homogeneity test and Double Mass Curve analysis.
Observed point data information used in this study
No. . | Name of stations . | Latitude . | Longitude . | Elevation . | Tmax . | Tmin . | PCP . |
---|---|---|---|---|---|---|---|
1 | Aemteferi | 8.9 | 35.23333 | 1,630 | ✓ | ✓ | ✓ |
2 | Bure | 8.28 | 35.11 | 1,704 | ✓ | ||
3 | Gimbi | 9.201389 | 35.83944 | 1,844 | ✓ | ||
4 | Gore | 8.157283 | 35.54983 | 2,024 | ✓ | ✓ | ✓ |
5 | Masha | 7.752683 | 35.37 | 2,235 | ✓ | ✓ | ✓ |
6 | Mizenteferi | 7.0026 | 35.58428 | 1,444 | ✓ | ||
7 | Matuhospital | 8.297067 | 35.57615 | 1,702 | ✓ | ✓ | ✓ |
8 | Tepi | 7.204883 | 35.43748 | 1,208 | ✓ | ✓ | ✓ |
9 | Uka | 8.18 | 35.35 | 1,667 | ✓ | ||
10 | Yubdo | 8.9525 | 35.44694 | 1,550 | ✓ | ✓ | ✓ |
11 | Dembidollo | 8.54 | 34.79 | 1,811 | ✓ | ✓ | ✓ |
No. . | Name of stations . | Latitude . | Longitude . | Elevation . | Tmax . | Tmin . | PCP . |
---|---|---|---|---|---|---|---|
1 | Aemteferi | 8.9 | 35.23333 | 1,630 | ✓ | ✓ | ✓ |
2 | Bure | 8.28 | 35.11 | 1,704 | ✓ | ||
3 | Gimbi | 9.201389 | 35.83944 | 1,844 | ✓ | ||
4 | Gore | 8.157283 | 35.54983 | 2,024 | ✓ | ✓ | ✓ |
5 | Masha | 7.752683 | 35.37 | 2,235 | ✓ | ✓ | ✓ |
6 | Mizenteferi | 7.0026 | 35.58428 | 1,444 | ✓ | ||
7 | Matuhospital | 8.297067 | 35.57615 | 1,702 | ✓ | ✓ | ✓ |
8 | Tepi | 7.204883 | 35.43748 | 1,208 | ✓ | ✓ | ✓ |
9 | Uka | 8.18 | 35.35 | 1,667 | ✓ | ||
10 | Yubdo | 8.9525 | 35.44694 | 1,550 | ✓ | ✓ | ✓ |
11 | Dembidollo | 8.54 | 34.79 | 1,811 | ✓ | ✓ | ✓ |
Note: ✓ represents the data are available at that station, PCP means precipitation, Tmax means maximum temperature and Tmin means the minimum temperature of the observed data.
CMIP6 Global Climate Model (GCM) data
In this study, the CMIP6 GCMs were used to simulate precipitation and temperature in the Baro River Basin. A set of criteria was used to select among the various GCMs from the CMIP6. They include: the nominal resolution (100 km), scenario/experiment and analysis period. Further, the availability of the CMIP6 models with a complete first ensemble member (r1i1p1f1) at the time of analysis was considered, i.e. only models having a similar variant label are used in order to have an unbiased comparison of all of them. In this case, models with different variant labels are not considered for comparisons as models with different variant labels can be associated to different realization, initialization and characteristics. Moreover, only models that have three simulation conditions, i.e. historical simulations, and SSP2-4.5 and SSP2-4.5 simulations for both precipitation and temperature are only considered. Considering these factors five GCMs are selected (Table 2). Historical data from 1985 to 2014 was used as a baseline for this study. Future climate datasets from CMIP6 GCMs are available from 2015 to 2100. For each model, precipitation, maximum, and minimum temperature data are obtained for the SSP2-4.5 and SSP5-8.5 scenarios. Further information on the GCMs data used is available online at (https://esgf-node.IInI.gov/search/cmip6/).
Basic information on the CMIP6 used in this study
Model name . | Institution and country . | Resolution (Lon × Lat) . | Variant label . |
---|---|---|---|
CMCC-ESM2 | Fondazione Centro Euro-Mediterraneo sui Cambiamenti Climatici, Italy | 1.25° × 0.9424° | r1i1p1f1 |
GFDL-CM4 | Geophysical Fluid Dynamics Laboratory, Princeton, USA | 1.25° × 1° | r1i1p1f1 |
INM-CM4-8 | Institute for Numerical Mathematics, Russia | 2° × 1.5° | r1i1p1f1 |
INM-CM5-0 | Institute for Numerical Mathematics, Russia | 2 × 1.5 | r1i1p1f1 |
MRI-ESM2-0 | Meteorological Research Institute, Japan | 1.125° × 0.9424° | r1i1p1f1 |
Model name . | Institution and country . | Resolution (Lon × Lat) . | Variant label . |
---|---|---|---|
CMCC-ESM2 | Fondazione Centro Euro-Mediterraneo sui Cambiamenti Climatici, Italy | 1.25° × 0.9424° | r1i1p1f1 |
GFDL-CM4 | Geophysical Fluid Dynamics Laboratory, Princeton, USA | 1.25° × 1° | r1i1p1f1 |
INM-CM4-8 | Institute for Numerical Mathematics, Russia | 2° × 1.5° | r1i1p1f1 |
INM-CM5-0 | Institute for Numerical Mathematics, Russia | 2 × 1.5 | r1i1p1f1 |
MRI-ESM2-0 | Meteorological Research Institute, Japan | 1.125° × 0.9424° | r1i1p1f1 |
Methods
Global climate models performance evaluation


Compromise programming





Downscaling
Downscaling is a method for producing high-resolution climate projections from coarse-resolution global climate models. While downscaling can provide more detailed information on climate variability and change, it also introduces additional uncertainties that must be taken into account. Some of the uncertainties associated with downscaling include model uncertainty, spatial and temporal uncertainty, and statistical uncertainty. The uncertainties associated with downscaling arise from the inherent uncertainty in the structure of downscaling models and the uncertainty in the input data (Khan et al. 2006). Downscaling often involves the use of statistical models to relate large-scale climate variables to local-scale variables. The accuracy of these statistical models depends on the availability and quality of data, assumptions made in the model, and other factors that can lead to statistical uncertainties. Despite these uncertainties, downscaling can be a useful tool for studying the regional or local effects of climate change. Downscaling can provide useful information for decision-making and planning, such as identifying areas that are particularly vulnerable to the effects of climate change or assessing the effectiveness of various adaptation strategies. There are different techniques of downscaling the GCMs for climate change studies. For example, a recent study by Kadkhodazadeh et al. (2022) used a delta change factor for downscaling the CMIP6 GCMs and Kumar et al. (2023) used Empirical Quantile Mapping (EQM) to get the statistically downscaled and bias-corrected precipitation and temperature data. Both studies reported that downscaled GCMs products are more representative than the coarser GCMs. However, it is critical to recognize the limitations and uncertainties of downscaling and use the results appropriately. Overall, statistical and dynamical downscaling are the two main methods of interpolating GCM data (Flint & Flint 2012). The GCM results are utilized as boundary conditions for the regional-scale GCM models at finer-resolution when dynamically downscaling. This method requires sophisticated computers, and finer-scale, complete physical weather and ocean models, which are computationally costly. Statistical downscaling approaches use a statistical relationship between coarse and fine resolution historical climate records to interpolate from coarse model outputs to finer-resolution (Flint & Flint 2012). This requires less computational effort, but it frequently requires drastic simplifications of physical relationships. Statistical downscaling uses the application of different statistics-based approaches to discover links between large-scale climate patterns determined by global climate models and observable local climatic responses (Salihu 2018). These connections are used with GCM data to convert climate model outputs into statistically refined products that are frequently more suitable for use as input to regional or local climate impact studies. Owing to the above-mentioned facts, a statistical downscaling method was used in this study.
Bias correction
Many climate models and bias reduction methods recommend adopting an ensemble approach using bias-corrected data. The main idea is to parameterize the strain correction algorithms used to modify the simulated historical climate data with the observed historical climate variables and the simulated historical climate variables. The same correction algorithm applies to future climate data. Bias correction methods are used to minimize deviations between observed and simulated climate variables at daily time intervals, so hydrological simulations driven based on corrected simulated climate data are reasonably consistent with simulations using the observed climate data (Rathjens et al. 2016). There are different techniques of bias correction for precipitation and temperature. A recent study by Tumsa (2022) evaluated six bias correction techniques in the upper Awash basin of Ethiopia based on frequency and time series metrics. The result revealed Linear Scaling (LS) performed best in removing the model biases for precipitation. Dibaba et al. (2020) on the other hand reported Distribution Mapping (DM) as the best method in removing temperature biases. Furthermore, a thorough analysis of the bias correction techniques published by Teutschbein & Seibert (2012) revealed that each method has improved the simulation of precipitation and temperature biases. The percentiles and standard deviations of the daily precipitation series, however, fluctuate depending on the correction techniques used. Luo et al. (2018) also used five and seven bias correction methods for temperature and precipitation, respectively. The outcome showed that DM is able to minimize temperature biases while LS do for precipitation.
The LS method adjusts GCM/RCM simulation rainfall and temperature using multiplicative and additive factors, respectively (Worku et al. 2019). The factors are created by comparing observed data to historical GCM/RCM simulations. Using monthly adjustment values based on the variations between observed and raw data, it functions (raw GCMs simulated data in this case). Every month, precipitation is often multiplied to adjust for errors. The distribution function (CDF) of GCM-simulated rainfall and temperature values is adjusted with the CDF of observed rainfall and temperature values in the DM method. The DM method modifies the mean, standard deviation, extremes, and distribution of GCM/RCM output rainfall and temperature events. The Gamma and Gaussian distributions are used in DM to fit GCM/RCM rainfall and temperature distributions to observational data. For temperature, the Gaussian distribution (normal distribution) with a mean and a standard deviation is
is usually assumed to fit temperature best.
This study used LS and DM methods to process the precipitation and temperature bias corrections, respectively. These bias correction techniques were selected owing to their effectiveness demonstrated in previous studies (Fang et al. 2015; Mengistu et al. 2021). Mengistu et al. (2021) in the Baro Akobo basin reported that LS performs slightly better than DM in correcting the bias from the individual raw models and ensemble for precipitation and DM outperforms LS for temperature bias correction.
Quantification of changes in precipitation and temperature






RESULTS AND DISCUSSIONS
Performance of GCM for precipitation
The calculated precipitation performance metrics for precipitation for all GCMs and the ranking of the GCMs derived using CP are shown in Table 3. The ideal values vary for each criterion, but the overall ranking indicates that the GCM that may have the most ideal values for more metrics does not necessarily rank best. For example, the ideal values of the percent of bias are 0.6 but the INM-CM5-0 is on the second rank. For precipitation, the top three GCMs are GFDL-CM4, INM-CM5-0, and INM-CM4-8, respectively. The lowest performing GCM for precipitation is CMCC-ESM2. For precipitation, the entropy method assigns weights of 0.27, 0.66, and 0.064 to the individual criteria RMSE, PBIAS, and R2, respectively.
GCM performance metrics and ranking for precipitation
. | Performance matrix . | Normalized matrix . | Normalized weighted matrix . | . | . | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|
GCM . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | Sum . | Rank . |
GFDL-CM4 | 15.71 | −5.53 | 0.99 | 0.06 | 0.09 | 0.26 | 0.00 | 0.03 | 0.00 | 0.03 | 1 |
INM-CM4-8 | 39.25 | 6.58 | 0.86 | 0.16 | 0.11 | 0.22 | 0.03 | 0.07 | 0.00 | 0.09 | 3 |
INM-CM5-0 | 26.35 | 0.60 | 0.94 | 0.11 | 0.01 | 0.25 | 0.04 | 0.00 | 0.00 | 0.04 | 2 |
MRI-ESM2-0 | 79.22 | 14.21 | 0.44 | 0.32 | 0.24 | 0.11 | 0.07 | 0.15 | 0.01 | 0.23 | 4 |
CMCC-ESM2 | 89.83 | 32.64 | 0.60 | 0.36 | 0.55 | 0.16 | 0.08 | 0.36 | 0.01 | 0.44 | 5 |
Ideal values | 15.71 | 0.6 | 0.99 | 0.06 | 0.01 | 0.26 |
. | Performance matrix . | Normalized matrix . | Normalized weighted matrix . | . | . | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|
GCM . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | Sum . | Rank . |
GFDL-CM4 | 15.71 | −5.53 | 0.99 | 0.06 | 0.09 | 0.26 | 0.00 | 0.03 | 0.00 | 0.03 | 1 |
INM-CM4-8 | 39.25 | 6.58 | 0.86 | 0.16 | 0.11 | 0.22 | 0.03 | 0.07 | 0.00 | 0.09 | 3 |
INM-CM5-0 | 26.35 | 0.60 | 0.94 | 0.11 | 0.01 | 0.25 | 0.04 | 0.00 | 0.00 | 0.04 | 2 |
MRI-ESM2-0 | 79.22 | 14.21 | 0.44 | 0.32 | 0.24 | 0.11 | 0.07 | 0.15 | 0.01 | 0.23 | 4 |
CMCC-ESM2 | 89.83 | 32.64 | 0.60 | 0.36 | 0.55 | 0.16 | 0.08 | 0.36 | 0.01 | 0.44 | 5 |
Ideal values | 15.71 | 0.6 | 0.99 | 0.06 | 0.01 | 0.26 |
The difference between the matrix and the ideal value represents the statistical value, and the sum of all values represents the rank of the model, with lower values representing better performance and higher values representing the least performance. The percentage bias of the top three models is less than 7% and the R2 value is greater than 0.855. Additionally, the three models follow unimodal rainfall. Therefore, based on observational data and comparing with historical GCM runs, the three climate models, i.e. GFDL-CM4, INM-CM5-0, and INM-CM4-8, are selected for precipitation.
Spatial distribution of the GCM and observed average annual precipitation over the Baro River Basin of historical simulation (1985–2014).
Spatial distribution of the GCM and observed average annual precipitation over the Baro River Basin of historical simulation (1985–2014).
Comparison of average monthly precipitation observed with the GCMs and their ensemble for historical simulation (1985–2014).
Comparison of average monthly precipitation observed with the GCMs and their ensemble for historical simulation (1985–2014).
The percentage bias for the ensembled three models is 0.55 and the percentage bias of all model ensemble and first ranked models are 9.701 and −5.53, respectively. This shows the performance of the three-model ensemble is superior to the ensemble of all models and the first ranked model GFDL-CM4.
The findings of GCMs performance evaluation in the Baro Akobo River basin show that CMIP6 GCMs are suitable to simulate the precipitation adequately. However, the selection of specific climate models for a specific region is subjective based on the factors considered and techniques used for evaluation. This is consistent with the study report by Almazroui et al. (2020) who reported CMIP6 GCMs are able to simulate the East African major climate variables. Similar findings of best performing models are reported in different parts of Africa. For example, a study by Ongoma et al. (2017) found GFDL-CM4 to be the best performing and first ranked model for annual precipitation simulation from the 15 GCMs evaluated in Uganda. On the other hand, Shiru & Chung (2021) reported INM-CM5-0 and INM-CM4-8 were ranked 11th and last from the 13 GCMs evaluated in Nigeria. The over/under estimation of mean monthly precipitation by some models reported in this study is also reported by Babaousmail et al. (2021) over North Africa. According to a study conducted by Babaousmail et al. (2021), the CMIP6 models effectively replicate the average annual climatic conditions during both dry and wet months. The findings of this study demonstrate that the CMIP6 models successfully replicate the mean monthly precipitation, with the top three ranked models performing particularly well. Nevertheless, a few models indicate a minor overestimation or underestimation for certain months. The study also reported adequate simulation of the mean ensemble than the individual models.
The variations in the spatial distribution of mean annual precipitation reported by the GCMs in Baro Akobo revealed that all models are not equally worth for specific locations and topographies. For example, the study by Dibaba et al. (2019) reported a varying spatial simulation of the climate models in the Finchaa and Didessa catchments. Especially, climate uncertainties in precipitation simulation increase in mountain regions. However, the mean ensemble of three models adequately estimated precipitation across the entire domain. However, the overall linear distribution over the basin shows the ensemble of the three models fits the observed precipitation better than the individual and ensemble of all models.
Performance of GCM for maximum temperature
The performance metrics for maximum temperature calculated for all GCMs and the ranking of GCMs derived using CP are presented in Table 4. The highest model ranking is CMCC-ESM2, MRI-ESM2-0, and INM-CM5-0, respectively. The least ranked GCM for maximum temperature using the CP method was GFDL-CM4. For maximum temperature, the entropy method assigns weights of 0.202, 0.79, and 0.006 to the individual criteria RMSE, PBIAS, and R2, respectively.
GCM performances metrics and ranking of maximum temperature
. | Performance matrix . | Normalized matrix . | Normalized weighted matrix . | . | . | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|
GCM . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | Sum . | Rank . |
CMCC-ESM2 | 1.19 | 0.50 | 0.75 | 0.12 | 0.04 | 0.17 | 0.001 | 0.00 | 0.00 | 0.001 | 1 |
GFDL-CM4 | 2.30 | 8.11 | 0.96 | 0.22 | 0.59 | 0.22 | 0.023 | 0.44 | 0.00 | 0.461 | 5 |
INM-CM4-8 | 1.13 | 3.26 | 0.91 | 0.11 | 0.24 | 0.20 | 0.000 | 0.16 | 0.00 | 0.160 | 4 |
INM-CM5-0 | 2.51 | 0.94 | 0.92 | 0.25 | 0.07 | 0.21 | 0.027 | 0.03 | 0.00 | 0.054 | 3 |
MRI-ESM2-0 | 1.97 | 0.99 | 0.90 | 0.19 | 0.07 | 0.20 | 0.017 | 0.03 | 0.00 | 0.046 | 2 |
Ideal values | 1.13 | 0.5 | 0.96 | 0.11 | 0.04 | 0.22 |
. | Performance matrix . | Normalized matrix . | Normalized weighted matrix . | . | . | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|
GCM . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | Sum . | Rank . |
CMCC-ESM2 | 1.19 | 0.50 | 0.75 | 0.12 | 0.04 | 0.17 | 0.001 | 0.00 | 0.00 | 0.001 | 1 |
GFDL-CM4 | 2.30 | 8.11 | 0.96 | 0.22 | 0.59 | 0.22 | 0.023 | 0.44 | 0.00 | 0.461 | 5 |
INM-CM4-8 | 1.13 | 3.26 | 0.91 | 0.11 | 0.24 | 0.20 | 0.000 | 0.16 | 0.00 | 0.160 | 4 |
INM-CM5-0 | 2.51 | 0.94 | 0.92 | 0.25 | 0.07 | 0.21 | 0.027 | 0.03 | 0.00 | 0.054 | 3 |
MRI-ESM2-0 | 1.97 | 0.99 | 0.90 | 0.19 | 0.07 | 0.20 | 0.017 | 0.03 | 0.00 | 0.046 | 2 |
Ideal values | 1.13 | 0.5 | 0.96 | 0.11 | 0.04 | 0.22 |
Comparison of average monthly maximum temperature observed with five GCMs and ensemble model of historical simulation (1985–2014).
Comparison of average monthly maximum temperature observed with five GCMs and ensemble model of historical simulation (1985–2014).
Spatial distribution of GCMs and observed average maximum temperature over the Baro River Basin of historical simulation (1985–2014).
Spatial distribution of GCMs and observed average maximum temperature over the Baro River Basin of historical simulation (1985–2014).
Performance of GCM for minimum temperature
The performance metrics for minimum temperature calculated for all GCMs and the ranking of GCMs derived using CP are presented in Table 5. For minimum temperature, the entropy method assigns weights of 0.52, 0.41, and 0.063 to the individual criteria RMSE, PBIAS, and R2, respectively. The top three ranking models are GFDL-CM4, INM-CM4-8, and INM-CM5-0, respectively.
Performances metrics and ranking of minimum temperature GCMs
. | Performance matrix . | Normalized matrix . | Normalized weighted matrix . | . | . | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|
GCM . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | Sum . | Rank . |
CMCC-ESM2 | 5.58 | 42.1 | 0.82 | 0.25 | 0.31 | 0.21 | 0.08 | 0.09 | 0.001 | 0.171 | 5 |
GFDL-CM4 | 2.03 | 13.91 | 0.88 | 0.09 | 0.10 | 0.23 | 0.00 | 0.00 | 0.001 | 0.001 | 1 |
INM-CM4-8 | 3.5 | 19.34 | 0.56 | 0.16 | 0.14 | 0.14 | 0.04 | 0.02 | 0.006 | 0.057 | 2 |
INM-CM5-0 | 3.45 | 21.2 | 0.73 | 0.16 | 0.15 | 0.19 | 0.03 | 0.02 | 0.003 | 0.059 | 3 |
MRI-ESM2-0 | 5.34 | 40.45 | 0.91 | 0.24 | 0.30 | 0.23 | 0.08 | 0.08 | 0.000 | 0.159 | 4 |
Ideal values | 2.03 | 13.91 | 0.91 | 0.09 | 0.10 | 0.23 |
. | Performance matrix . | Normalized matrix . | Normalized weighted matrix . | . | . | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|
GCM . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . | Sum . | Rank . |
CMCC-ESM2 | 5.58 | 42.1 | 0.82 | 0.25 | 0.31 | 0.21 | 0.08 | 0.09 | 0.001 | 0.171 | 5 |
GFDL-CM4 | 2.03 | 13.91 | 0.88 | 0.09 | 0.10 | 0.23 | 0.00 | 0.00 | 0.001 | 0.001 | 1 |
INM-CM4-8 | 3.5 | 19.34 | 0.56 | 0.16 | 0.14 | 0.14 | 0.04 | 0.02 | 0.006 | 0.057 | 2 |
INM-CM5-0 | 3.45 | 21.2 | 0.73 | 0.16 | 0.15 | 0.19 | 0.03 | 0.02 | 0.003 | 0.059 | 3 |
MRI-ESM2-0 | 5.34 | 40.45 | 0.91 | 0.24 | 0.30 | 0.23 | 0.08 | 0.08 | 0.000 | 0.159 | 4 |
Ideal values | 2.03 | 13.91 | 0.91 | 0.09 | 0.10 | 0.23 |
Comparison of average monthly minimum temperature observed with five GCMs and ensemble model of the historical simulation (1985–2014).
Comparison of average monthly minimum temperature observed with five GCMs and ensemble model of the historical simulation (1985–2014).
Spatial distribution of the GCM and observed average minimum temperature over the Baro River Basin of historical simulation (1985–2014).
Spatial distribution of the GCM and observed average minimum temperature over the Baro River Basin of historical simulation (1985–2014).
Overall, the study results show the importance of evaluating different GCMs into representative models and determining the most realistic ones for model aggregation into ensembles for climate projection to lessen the individual uncertainties linked to various GCMs. However, this choice does not necessarily guarantee that GCMs will be equally important for different climate variables. For example, GFDL-CM4, which performed well at precipitation, performed poorly for maximum temperature. Similarly, CMCC-ESM2 which is ranked 1st for maximum temperature showed a lower rank for precipitation and minimum temperature. Moreover, the best performing GCMs in one region or study area did not rank equally in other regions. To conduct a thorough assessment of climate models, it is essential to have a performance evaluation. According to Thorarinsdottir et al. (2020), climate model simulations must undergo an evaluation process by comparing the distributions of their outputs to the corresponding distributions of observational or reanalysis data products. Evaluating how well CMIP6 models simulate extreme climate events in the past can serve as a basis for generating dependable projections of future climate (Fan et al. 2020). According to Karim et al. (2020) efficient predictions of future climate changes depend on a thorough and accurate observation of past changes. Our study has examined how well each CMIP6 GCM performs in simulating the climate of the Baro Akobo river basin, particularly in relation to the distributions of daily observed climate variables.
Evaluation of bias-corrected precipitation and temperature
Raw GCMs outputs underestimate and overestimate the mean monthly precipitation as compared to the observed precipitation data. The statistical metrics such as PBIAS, RMSE, and R2 showed a substantial difference between raw and bias-corrected GCMs output as compared to the observed long-term monthly precipitation data. The raw ensemble GCM simulations are heavily biased from observed data. For example, comparing the observed precipitation to the GCM output, the highest PBIAS (overestimation) was 30.06 and 26.83% for the Dembidollo and Bure stations, respectively. On the other hand, a large PBIAS (underestimation) was −20.85% for Masha and there is a small present bias in most stations (Table 6). The biases in the raw precipitation data were substantially reduced after applying a LS bias correction.
Statistical performance for raw and bias-corrected GCM precipitation
. | . | . | Before . | . | . | Bias corrected . | . |
---|---|---|---|---|---|---|---|
Station . | Variable . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . |
Alem teferi | PCP | 33.55 | − 5.34 | 0.95 | 0.02 | 0.00 | 1.00 |
Bure | PCP | 42.11 | 26.83 | 0.93 | 0.02 | − 0.01 | 1.00 |
Dembidollo | PCP | 51.01 | 30.06 | 0.86 | 0.03 | − 0.01 | 1.00 |
Gimbi | PCP | 45.05 | − 7.44 | 0.96 | 0.01 | 0.00 | 1.00 |
Gore | PCP | 23.39 | − 2.26 | 0.95 | 0.00 | 0.00 | 1.00 |
Masha | PCP | 47.16 | − 20.85 | 0.92 | 0.01 | 0.00 | 1.00 |
Metu hospital | PCP | 31.86 | 16.68 | 0.95 | 0.01 | 0.00 | 1.00 |
Mizen teferi | PCP | 31.95 | − 13.13 | 0.91 | 0.02 | 0.00 | 1.00 |
Tepi | PCP | 31.13 | 8.93 | 0.89 | 0.01 | 0.00 | 1.00 |
Uka | PCP | 23.64 | − 0.35 | 0.95 | 0.02 | 0.00 | 1.00 |
Yubdo | PCP | 10.26 | − 2.93 | 0.96 | 0.01 | − 0.01 | 1.00 |
. | . | . | Before . | . | . | Bias corrected . | . |
---|---|---|---|---|---|---|---|
Station . | Variable . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . |
Alem teferi | PCP | 33.55 | − 5.34 | 0.95 | 0.02 | 0.00 | 1.00 |
Bure | PCP | 42.11 | 26.83 | 0.93 | 0.02 | − 0.01 | 1.00 |
Dembidollo | PCP | 51.01 | 30.06 | 0.86 | 0.03 | − 0.01 | 1.00 |
Gimbi | PCP | 45.05 | − 7.44 | 0.96 | 0.01 | 0.00 | 1.00 |
Gore | PCP | 23.39 | − 2.26 | 0.95 | 0.00 | 0.00 | 1.00 |
Masha | PCP | 47.16 | − 20.85 | 0.92 | 0.01 | 0.00 | 1.00 |
Metu hospital | PCP | 31.86 | 16.68 | 0.95 | 0.01 | 0.00 | 1.00 |
Mizen teferi | PCP | 31.95 | − 13.13 | 0.91 | 0.02 | 0.00 | 1.00 |
Tepi | PCP | 31.13 | 8.93 | 0.89 | 0.01 | 0.00 | 1.00 |
Uka | PCP | 23.64 | − 0.35 | 0.95 | 0.02 | 0.00 | 1.00 |
Yubdo | PCP | 10.26 | − 2.93 | 0.96 | 0.01 | − 0.01 | 1.00 |
Table 7 compares the raw and bias-corrected monthly maximum and minimum temperature GCMs using PBIAS, RMSE, and R2. The result showed that the raw output of the ensemble GCMs showed substantial biases compared to the observed data. For example, a large source of bias PBIAS (19.64%) and RMSE (4.46) was simulated for the Masha station. The application of bias correction reduced the biases substantially at most of the stations. For example, the PBIAS of 19.64% and RMSE of 4.46 at the Masha station were removed after applying a bias correction, showing excellent agreement with the observed maximum temperature.
Statistical performance measure of raw and bias-corrected GCM Tmax and Tmin
. | . | . | Before . | . | Bias corrected . | . | |
---|---|---|---|---|---|---|---|
Station . | Variables . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . |
Alem teferi | Tmax | 0.71 | − 7.79 | 0.84 | 0.00 | 0.00 | 1.00 |
Tmin | 3.42 | 21.08 | 0.24 | 0.01 | − 0.03 | 1.00 | |
Dembidollo | Tmax | 2.51 | 9.73 | 0.93 | 0.00 | 0.00 | 1.00 |
Tmin | 3.49 | 25.52 | 0.86 | 0.00 | 0.00 | 1.00 | |
Gore | Tmax | 1.79 | 7.08 | 0.91 | 0.00 | 0.00 | 1.00 |
Tmin | 2.71 | 8.43 | 0.09 | 0.02 | − 0.03 | 1.00 | |
Masha | Tmax | 4.46 | 19.64 | 0.79 | 0.00 | 0.00 | 1.00 |
Tmin | 5.09 | 44.23 | 0.70 | 0.01 | − 0.04 | 1.00 | |
Metu | Tmax | 2.09 | − 7.10 | 0.92 | 0.00 | 0.00 | 1.00 |
Tmin | 3.87 | 32.25 | 0.91 | 0.02 | − 0.06 | 1.00 | |
Tepi | Tmax | 2.88 | − 9.56 | 0.94 | 0.00 | 0.00 | 1.00 |
Tmin | 1.30 | 6.13 | 0.89 | 0.00 | 0.00 | 1.00 | |
Yubdo | Tmax | 2.30 | − 7.21 | 0.91 | 0.00 | 0.00 | 1.00 |
Tmin | 1.95 | 0.87 | 0.37 | 0.01 | − 0.03 | 1.00 |
. | . | . | Before . | . | Bias corrected . | . | |
---|---|---|---|---|---|---|---|
Station . | Variables . | RMSE . | PBIAS . | R2 . | RMSE . | PBIAS . | R2 . |
Alem teferi | Tmax | 0.71 | − 7.79 | 0.84 | 0.00 | 0.00 | 1.00 |
Tmin | 3.42 | 21.08 | 0.24 | 0.01 | − 0.03 | 1.00 | |
Dembidollo | Tmax | 2.51 | 9.73 | 0.93 | 0.00 | 0.00 | 1.00 |
Tmin | 3.49 | 25.52 | 0.86 | 0.00 | 0.00 | 1.00 | |
Gore | Tmax | 1.79 | 7.08 | 0.91 | 0.00 | 0.00 | 1.00 |
Tmin | 2.71 | 8.43 | 0.09 | 0.02 | − 0.03 | 1.00 | |
Masha | Tmax | 4.46 | 19.64 | 0.79 | 0.00 | 0.00 | 1.00 |
Tmin | 5.09 | 44.23 | 0.70 | 0.01 | − 0.04 | 1.00 | |
Metu | Tmax | 2.09 | − 7.10 | 0.92 | 0.00 | 0.00 | 1.00 |
Tmin | 3.87 | 32.25 | 0.91 | 0.02 | − 0.06 | 1.00 | |
Tepi | Tmax | 2.88 | − 9.56 | 0.94 | 0.00 | 0.00 | 1.00 |
Tmin | 1.30 | 6.13 | 0.89 | 0.00 | 0.00 | 1.00 | |
Yubdo | Tmax | 2.30 | − 7.21 | 0.91 | 0.00 | 0.00 | 1.00 |
Tmin | 1.95 | 0.87 | 0.37 | 0.01 | − 0.03 | 1.00 |
Note: Tmax and Tmin means maximum temperature and minimum temperature, respectively.
The results showed that the raw outputs from GCMs exhibited large biases for maximum and minimum temperatures. However, the biases were substantially reduced after applying DM bias correction. The improvement after bias correction particularly for PBIAS and RMSE was very good.
Quantifying the future climate projection
Precipitation projection
Bias-corrected mean annual and seasonal precipitation over the entire steady area for the future periods from 2031 to 2060 are presented in Table 8. The result shows that wet season precipitation is projected to increase by 9 mm (4.2%) and 27.2 mm (12.6%) under SSP2-4.5 and SSP5-8.5 scenarios, respectively. On the other hand, the projected annual precipitation was predicted to increase by 6.8 mm (6%) under the SSP2-4.5 and 23.63 mm (16.46%) under the SSP5-8.5 scenarios. The projected seasonal precipitation during the dry season shows increasing precipitation by 8 mm (19.15%) and 18 mm (44.4%) under the SSP2-4.5 and SSP5-8.5, respectively. Similar studies have reported an increasing trend of seasonal and annual precipitation in the future period (Ongoma et al. 2017; NAP 2019; Alaminie et al. 2021). Mekonnen & Disse's (2018) reported an increasing trend of precipitation, and temperature over the Upper Blue Nile River Basin using the ensemble mean of five GCMs. Similarly, a recent study by Alaminie et al. (2021) in the Upper Blue Nile River Basin reported an increasing precipitation trend under four SSP scenarios.
Changes in mean annual precipitation for the future (2031–2060) relative to the baseline period (1985–2014)
Mean GCMs precipitation . | . | Annual . | . | AMJJASO . | . | NDJFM . |
---|---|---|---|---|---|---|
Scenarios | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 |
Baseline | 143.48 | 143.48 | 216.03 | 216.03 | 41.91 | 41.91 |
2031–2060 | 152.08 | 167.10 | 225.04 | 243.21 | 49.95 | 60.54 |
Change (mm) | 8.6 | 23.62 | 9.0 | 27.2 | 8.03 | 18 |
Change (%) | 6 | 16.46 | 4.2 | 12.6 | 19.15 | 44.4 |
Mean GCMs precipitation . | . | Annual . | . | AMJJASO . | . | NDJFM . |
---|---|---|---|---|---|---|
Scenarios | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 |
Baseline | 143.48 | 143.48 | 216.03 | 216.03 | 41.91 | 41.91 |
2031–2060 | 152.08 | 167.10 | 225.04 | 243.21 | 49.95 | 60.54 |
Change (mm) | 8.6 | 23.62 | 9.0 | 27.2 | 8.03 | 18 |
Change (%) | 6 | 16.46 | 4.2 | 12.6 | 19.15 | 44.4 |
The projected monthly precipitation shows consistent directions and magnitude of change between SSP2-4.5 and SSP5-8.5 scenarios except for the month of July. The average monthly projected precipitation was increased under SSP2-4.5 scenarios except for July. The projected precipitation in SSP5-8.5 is greater than in the SSP2-4.5 scenarios. Ongoma et al. (2017) reported similar trends of precipitation projection in East Africa, stating the increase in precipitation under SSP5-8.5 are greater than under SSP2-4.5 scenarios. A similar study by Alaminie et al. (2021) found precipitation projected under SSP5-8.5 are greater than SSP2-4.5 in all months except July over the Upper Blue Nile basin. Almazroui et al. (2020) over North East Africa also reported increasing precipitation in SSP5-8.5, but SSP2-4.5 have no uniform trend of precipitation. Alternatively, the study conducted by Gebresellase et al. (2022) on the Awash basin projected that future precipitation and temperature changes will result in increased precipitation intensities, more wet days, and longer dry spells, primarily due to a significant rise in temperature. Likewise, this study demonstrated that the projected precipitation will increase in the future under both SSPs (SSP2-4.5 and SSP5-8.5). In general, the study indicates that there will be heightened future changes in the basin's climate, specifically with regard to increased precipitation and temperature. Precipitation increases have the potential to increase water availability, which will benefit water resources. However, this is not true always as reported by Dibaba et al. (2020). According to Dibaba et al. (2020), regardless of the increasing or decreasing precipitation, the availability of water is highly affected by the temperature change and the associated evapotranspiration demands. On the other hand, if the predicted precipitation increases, it could have a number of consequences for the environment and human society. Increased precipitation can lead to more frequent and intense flooding events, particularly in areas that are already prone to flooding. This could have negative consequences such as infrastructure damage, property loss, crop damage, and even death. While increased precipitation can be beneficial to agriculture and water resources, it can also cause soil erosion and water pollution. Heavy rain has the potential to wash away topsoil and transport pollutants from agricultural and urban areas, affecting waterways and jeopardizing water quality.
Maximum temperature (Tmax) projection
The projected maximum temperature shows an increase in mean seasonal and annual temperature under SSP2-4.5 and SSP5-8.5 scenarios. The mean annual maximum temperature is projected to increase by 1.43 and 1.81 °C under SSP2-4.5 and SSP5-8.5 scenarios, respectively, compared with baseline (historical). The projected average maximum temperature predicted for the wet season was slightly higher than for the dry season. The projected maximum temperature during the wet season from April to October increases by 1.67 and 2.17 °C under SSP2-4.5 and SSP5-8.5 scenarios, respectively. On the other hand, the projected average maximum temperature during the dry season from November to March increases by 1.09 and 1.3 °C under SSP2-4.5 and SSP5-8.5 scenarios, respectively (Table 9).
Mean average seasonal and annual change of Tmax (2031–2060)
Mean Tmax (°C) . | . | Annual . | . | AMJJASO . | NDJFM . | |
---|---|---|---|---|---|---|
Scenarios | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 |
Baseline | 26.69 | 26.69 | 25.5 | 25.5 | 28.36 | 28.36 |
2031–2060 | 28.12 | 28.5 | 27.17 | 27.67 | 29.45 | 29.66 |
Change (°C) | 1.43 | 1.81 | 1.67 | 2.17 | 1.09 | 1.3 |
Mean Tmax (°C) . | . | Annual . | . | AMJJASO . | NDJFM . | |
---|---|---|---|---|---|---|
Scenarios | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 |
Baseline | 26.69 | 26.69 | 25.5 | 25.5 | 28.36 | 28.36 |
2031–2060 | 28.12 | 28.5 | 27.17 | 27.67 | 29.45 | 29.66 |
Change (°C) | 1.43 | 1.81 | 1.67 | 2.17 | 1.09 | 1.3 |
Note: Tmax denotes maximum temperature.
The simulated maximum temperature shows a consistent increase in all months under both SSP2-4.5 and SSP5-8.5 scenarios. The monthly maximum temperature increases in SSP2-4.5 range from 0.71 to 2.06 °C with a maximum increase in June, and the maximum temperature increase in SSP5-8.5 range from 0.92 to 2.65 °C, with a maximum increase in June. Overall, the increase in temperature reported in this study is consistent with the continuous increase in warming observed during the 21st century in the North Eastern Africa (Almazroui et al. 2020; Fan et al. 2020). According to Almazroui et al. (2020), temperature is projected to increase by 1.3 and 1.7 °C under SSP2-4.5 and SSP5-8.5, respectively. The temperature projections are also in line with the previous studies in different parts of Ethiopia (NAP 2019; Getahun et al. 2020; Alaminie et al. 2021). For example, Alaminie et al. (2021) reported that the mean annual maximum temperature is projected to increase in the Upper Blue Nile basin by 1.3 and 1.5 °C under SSP2-4.5 and SSP5-8.5 scenarios, respectively, for near-term (2031–2060) period. Getachew et al. (2021) reported an increase in maximum temperatures by 1.38–3.59 °C by the 2080s in the Lake Tana sub-basin of Ethiopia under the RCP4.5 radiative forcing scenario. Rising temperatures have several significant implications. For example, the overall rise in temperature is attributed to these changes in precipitation patterns (Getachew et al. 2021). According to the study by Gurara et al. (2022), projected changes in future precipitation and temperature lead to increased precipitation intensities, more wet days, and longer dry spells in the Awash basin. The significant rise in temperature is attributed to these changes. As temperatures rise, the atmosphere may change, changing rainfall patterns and causing more intense rainfall events to occur. It should be noted that the specific consequences of rising temperatures can vary depending on regional and local factors. The rise in temperature could also result in increased evapotranspiration demand which reduces the availability of water in the soil according to Dibaba et al. (2020).
Minimum temperature (Tmin) projection
Like maximum temperature, a projected minimum temperature show increases in seasonal and annual average temperature under SSP2-4.5 and SSP5-8.5 scenarios. The annual average minimum temperature is projected to increase by 1.96 and 3.11 °C under SSP2-4.5 and SSP5-8.5 scenarios, respectively. The projected average maximum temperature of the wet season was slightly greater than the dry season. The projected minimum temperature during the wet season from April to October increases by 2.32 and 3.62 °C under SSP2-4.5 and SSP5-8.5 scenarios, respectively. On the other hand, the projected average minimum temperature during the dry season from November to March increases by 1.45 and 2.4 °C under SSP2-4.5 and SSP5-8.5 scenarios, respectively (Table 10).
Mean average seasonal and annual change of Tmin (2031–2060)
Average Tmin (°C) . | . | Annual . | . | AMJJASO . | NDJFM . | |
---|---|---|---|---|---|---|
Scenarios | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 |
Baseline | 13.18 | 13.18 | 13.52 | 13.52 | 12.71 | 12.71 |
2031–2060 | 15.14 | 16.29 | 15.84 | 17.14 | 14.17 | 15.11 |
Change (°C) | 1.96 | 3.11 | 2.32 | 3.62 | 1.45 | 2.4 |
Average Tmin (°C) . | . | Annual . | . | AMJJASO . | NDJFM . | |
---|---|---|---|---|---|---|
Scenarios | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 | SSP2-4.5 | SSP5-8.5 |
Baseline | 13.18 | 13.18 | 13.52 | 13.52 | 12.71 | 12.71 |
2031–2060 | 15.14 | 16.29 | 15.84 | 17.14 | 14.17 | 15.11 |
Change (°C) | 1.96 | 3.11 | 2.32 | 3.62 | 1.45 | 2.4 |
Note: Tmin denotes minimum temperature.
The simulated minimum temperature shows a consistent increase in all months under both SSP2-4.5 and SSP5-8.5 scenarios compared with the baseline (historical period). The monthly minimum temperature increases under SSP2-4.5 is between 0.67 and 2.81 °C with the maximum increase in September, while minimum temperature increases under SSP5-8.5 range from 1.04 to 4.4 °C with the maximum increase in September. The simulated minimum temperature shows an increasing direction of change between SSP2-4.5 and SSP 5-8.5 scenarios in all months. An increase in minimum temperature is larger than the increase in maximum temperature projected under both SSP2-4.5 and SSP5-8.5 scenarios.
The results are consistent with previous studies. For example, a study by Getahun et al. (2020) indicated that the increase in minimum temperature for all GCMs was generally higher than the maximum temperature increase for both the 2021–2051 and 2071–2100 periods in the Melka Kunture sub-basin of the Awash basin, Ethiopia. Overall, increasing trends in temperature were reported by recent studies in North East Africa (Almazroui et al. 2020). This is in line with the recent IPCC's sixth assessment.
Implication of climate change
The projected rise in temperatures and the anticipated fluctuations in precipitation will impact the hydrological cycle, thereby influencing future planning and development of the watershed. Agriculture plays a vital role as the backbone of the country's economy, and any fluctuations in productivity directly influence the growth of the Gross Domestic Product (GDP) (Addisu et al. 2015). Consequently, any alterations in the quantity and distribution of rainfall and temperature would pose a severe threat to agricultural productivity. These changes would have immediate implications for food production and security at a national level. Warmer minimum temperatures can extend growing seasons in these areas, allowing crops to grow for longer periods of time and potentially increasing agricultural productivity (Moges & Bhat 2021). While longer growing seasons can be beneficial, rising minimum temperatures can cause crop heat stress, lower yields, altered pollination patterns, and increased pest and disease pressure, posing challenges to agricultural production and food security. The increase in temperature not only impacts crop production but also has a profound effect on the growth and production patterns of livestock. According to Garnett (2009), as temperatures rise, livestock production faces detrimental consequences including intensified competition for natural resources, diminished feed quality and quantity, loss of biodiversity, and the additional challenge of increasing demands for livestock products.
The dynamic alterations in future rainfall, as stated in the IPCC's fifth assessment report (IPCC 2013), are largely attributable to the rise in global surface temperature induced by human activities. It is highly certain that in a considerably warmer world, changes in average precipitation will not occur uniformly. Certain regions will witness increases, while others may experience decreases or relatively minor alterations in precipitation patterns. According to Moges & Bhat (2021), even slight changes in the amount, distribution, and trends of rainfall can have a direct and substantial impact on agricultural production. Consequently, these variations significantly affect the lives of rural smallholder farmers who heavily rely on agriculture as their primary source of income. As stated by Fiwa et al. (2014), when excessive rainfall occurs without proper soil conservation structures in place, it results in a heightened rate of surface runoff and soil erosion. This, in turn, leads to the depletion of fertile topsoil in high slope areas and the accumulation of sedimentation in low slope areas. Flooding causes the inundation of agricultural fields, leading to the destruction of crops. Additionally, it can have a significant impact on the quality of grazing lands and irrigation facilities in the downstream watershed (Maharjan & Joshi 2013). The results of this study provide valuable insights into the spatial variations and temporal trends of rainfall and temperature. This information is crucial for effective water resource management and making informed decisions regarding farming practices.
CONCLUSIONS
This paper has evaluated the performance of five GCMs, which are part of CMIP6, to quantify the future precipitation and temperature changes under different scenarios. The performance evaluation of the individual GCMs and ensemble of the GCMs was held based on multi-criteria like mean annual spatial simulation, monthly analysis, CP and the combination of statistical metrics (PBIAS, RMSE, and R2). The method of Entropy weighting was selected for the allocation of weights to evaluate the dispersion of values. Three GCMs (GFDL-CM4, INM-CM5-0, and INM-CM4-8) outperformed precipitation simulations and two models (CMCC-ESM2 and MRI-ESM2) simulated precipitation poorly compared to the observed data (1985–2014). CMCC-ESM2, MRI-ESM2, and INM-CM5-0 are the top three models for maximum temperature simulation, whereas INM-CM5-0, and GFL-CM4, INM-CM5, and INM-CM4-8 are the top three models for minimum temperature simulation. The statistical performance evaluations for both precipitation and temperature show that the ensemble of three models outperforms both the individual and ensemble of all GCMs. The combination of statistical metrics, spatial and seasonal and annual climatology in the evaluation of CMIP6 over the Baro Akobo River basin showed that CMIP6 GCMs can be used for climate projections.
GCMs bias correction was processed using LS for precipitation and DM for temperature using the CMhyhd tool. For the climate change projections, the ensemble of the three models was bias-corrected to reduce the uncertainties associated with the individual GCMs. Annual and seasonal projected rainfall show an increasing trend under both SSP2-4.5 and SSP5-8.5 in the coming period (2031–2060) with reference to a baseline period (1985–2014). The projection of both maximum and minimum temperature shows an increasing trend under both SSPs, but the projected temperature under SSP5-8.5 is slightly greater than the projected temperature under SSP2-4.5. The increasing temperature may cause an increase in evapotranspiration and potential evapotranspiration, which in turn causes a reduction in soil water availability which could be a problem for subsistence agriculture. Overall, the changes in precipitation and increase in temperature will intensify under different climate change scenarios in the Baro Akobo River basin. In this regard, it is important to evaluate the climate change impacts which could help to understand the current and future change impacts. This could help to respond to what should happen that help to evaluate management/adaptations to compensate for the change impacts.
For future investigations in the basin, it is crucial to consider the evaluation of CMIP6 in simulating the precipitation extremes in order to evaluate the potential impacts of climate change on flooding and drought. In this study, the scenarios used for the projection of climate change are limited to two (SSP2-4.5 and SSP5-8.5). Therefore, similar studies should encompass a broad range of updated future forcing pathways, including both the middle to upper range and the lower end, to provide a comprehensive understanding for future research. With the set of criteria developed based on a nominal resolution of 100 km and GCMs with the first member (r1i1plf1), only 5GCMs are used. If the criteria change, there is a possibility to get many more GCMs. Moreover, this study used only five GCMs and with a set of criteria developed based on a nominal resolution of 100 km.
In this study, the ensemble of three models outperforms the ensemble of all models in the Baro River Basin. Overall, the best model for one climate variable (precipitation) may not be representative of the other variable (temperature), so it is important to evaluate the performance of both precipitation and temperature. The anticipated rise in precipitation levels has the potential to create a flood risk. Therefore, we suggest utilizing the average output from the highest-ranked models to examine the hydrological response to climate change. The result highlights the need for regional development and cooperation to promote strong climate-resilient management strategies and to counteract the rapid climate changes projected in the study area. In order to support decision makers, planners and stakeholders to devise appropriate basin management strategies, evaluating the potential impacts of climate change on the hydrology and water resources of the Baro Akobo basin will be the continuation of this study.
ACKNOWLEDGEMENTS
The authors thank Bule Hora University and Jimma University for hosting and supporting the study. Thanks to the Ethiopian National Meteorological Agency for providing the observed climate data. We also thank the Working Groups of the World Climate Research Program for making CMIP6 available in the Earth System Federation (ESGF) Archive.
FUNDING
The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.