Evaluation and projection of mean surface temperature using CMIP6 models over East Africa

This study evaluates the historical mean surface temperature (hereafter T2m) and examines how T2m changes over East Africa (EA) in the 21 century using CMIP6 models. An evaluation was conducted based on mean state, trends, and statistical metrics (Bias, Correlation Coefficient, Root Mean Square Difference, and Taylor skill score). For future projections over EA, five best performing CMIP6 models (based on their performance ranking in historical mean temperature Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 29 January 2021 doi:10.20944/preprints202101.0611.v1 © 2021 by the author(s). Distributed under a Creative Commons CC BY license. simulations) under the shared socioeconomic pathways SSP2-4.5 and SSP5-8.5 scenarios were employed. The historical simulations reveal an overestimation of the mean annual T2m cycle over the study region with fewer models depicting underestimations. Further, CMIP6 models reproduce the spatial and temporal trends within the observed range proximity. Overall, the best performing models are as follows: FGOALS-g3, HadGEM-GC31-LL, MPI-ESM2-LR, CNRM-CM6-1, and IPSL-CM6A-LR. During the three-time slice under consideration, the Multi Model Ensemble (MME) project many changes during the late period (2080 – 2100) with expected mean changes at 2.4 °C for SSP2-4.5 and 4.4 °C for the SSP5-8.5 scenario. The magnitude of change based on Sen’s slope estimator and Mann-Kendall test reveal significant increasing tendencies with projections of 0.24°C decade-1 (0.65°C decade-1) under SSP2-4.5 (SSP5-8.5) scenarios. The findings from this study illustrate higher warming in the latest model outputs of CMIP6 relative to its predecessor, despite identical instantaneous radiative forcing.


Introduction
The global mean surface temperature (GMST) records continue to set new levels year after year from 2011 to present (IPCC, 2018;WMO, 2020). Despite the global impacts of coronavirus disease , the year 2020 stood to be the warmest year, with new records set to about 1.2 °C above the pre-industrial (1850 -1900) level (WMO, 2020). The situation is further exacerbated with the recent report highlighting an alarming tendency of GMST warming at 0.1 -0.3 per decade, which continue to have an adverse impact on human and ecosystem (IPCC, 2018).
Comparative analysis before and after the 1950s reveals an amplification of extreme events resulting from heightened warming of global land and ocean (IPCC, 2014). Efforts to mitigate the impacts of global warming levels (GWLs) to below 2 °C or even much more preferred targets of 1.5 °C under current greenhouse levels (GHGs) remains an elusive ambition despite the efforts by United Nations Framework Convention on Climate Change (UNFCCC) Paris Agreement (UNFCCC, 2015). In order to stabilize the GMST from further upward trajectories, concerted energies from all stakeholders will be required to devise appropriate and urgent actions to guide the earth system from crossing the threshold to the worst-case scenario of "Hothouse Earth" (Steffen et al., 2018).
The scientific community plays a critical role in providing timely and accurate information regarding the evolution of climate anomalies that currently define the present world and the future frequencies. The advent of the Coupled Model Intercomparison Project (CMIP) has significantly aided in understanding the projected patterns of the climate system. The resultant advancement is the General Circulation Models (GCMs), capable of enhancing our understanding of the climate system. The output of these models informs relevant stakeholders on the formulation of effective and sustainable policies.
The recent model outputs of CMIP6 promise to exemplify the most accurate projections of future climate due to the massive improvements compared to previous outputs (Eyring et al., 2016). To illustrate this, higher spatial resolution (~ 70 km) in comparison to coarser resolution (~ 250 km) for CMIP5 characterizes the current model generation. Besides, improved physical processes and biogeochemical cycles, new features such as improved aerosols' effect or refined parameterization schemes, assorted range of socioeconomic pathways (SSPs; van Vuuren et al., 2014;O'Neill et al., 2017), and large ensemble members are among the developments that describe the latest model outputs. Subsequently, large volumes of research outputs based on CMIP6 have highlighted notable improvements in modelling various aspects of the climate system or comparative analysis of the added value in CMIP6 as compared to CMIP5 (Voldoire et al., 2019;Mauritsen et al., 2019;Hajima et al., 2020;Moseid et al., 2020). Studies focusing on simulations or projection of mean and extreme climate based on CMIP6 (e.g., Akinsanola et al., 2020;Almazroui et al., 2020a, b;Grose et al., 2020;Jiang et al., 2020) or comparative studies of CMIP6 against CMIP5 performance have also reported better and more reliable results (Gusain et al., 2020;Jiang et al., 2020;Luo et al., 2020;Nie et al., 2020;Senevirante and Hauser, 2020;Zamani et al., 2020;. Remarkably, these studies continue to enhance our understanding of the suitable models to be employed for accurate diagnosis and projection of impact analysis. Over East Africa ( Figure. 1), the livelihood of a population that is mainly dependent on climate variables is under threat posed by pronounced changes in the climate system, mainly due to an increase in T2m and reduced precipitation (Shongwe et al., 2011;Senevirante et al., 2012;Niang et al., 2014;Ongoma et al., 2018;Ayugi and Tan, 2019;Gebrechorkos et al., 2019). The damages witnessed over recent years is threatening the economy and ecosystem (IPCC, 2014). For instance, recurrent droughts (Nicholson, 2014;Haile et al., 2019;Tan et al., 2020) and floods (Kilavi et al., 2018;Juma et al., 2020) remain the signature features affecting millions of people and has a negative impact on the gross domestic product in the region that is mainly dependent on the agrarian economy (World Bank, 2012;FAO, 2019). Characterizing observed events has been based on various reanalysis or gauge-based estimates that have pointed clear historical patterns and vulnerable regions identified (Liebman et al., 2014;Lyon, 2014;Gebremeskel et al., 2019).
At present, few studies have been conducted using the latest model output to clearly delineate the local changes in climate features that keep evolving with new challenges due to the steady rise in greenhouse gas (GHGs) emissions. Existing studies based on previous CMIPs versions or regionally downscaled models under the CORDEX framework have documented the spatial and temporal variance, trends, extreme occurrences, and possible projections with respective attributes highlighted (IPCC, 2001;Collins et al., 2011;IPCC, 2014;Omondi et al., 2014;Camberlin, 2017;Ongoma and Chen, 2017;Ongoma et al., 2018a;Gebrechorkos et al., 2018;Osima et al., 2018;Ayugi and Tan, 2019). The aforementioned studies have pointed to an increase in T2m, trends, and increase (decrease) in observed maximum (minimum) temperatures.
The observed changes have a massive impact on the wellbeing of society and ecosystems services.
In order to provide the latest information regarding the possible future scenarios based on CMIP6 models, there is a need to establish suitable models capable of reproducing local climate and provide accurate simulations (Flato et al., 2013;Sillman et al., 2013). As a basis to identify the most suitable models for climate projections, this study evaluates the historical T2m using an observational gauge product from the climatic research unit (CRU TS4.03; Harris, 2020) and examines how T2m will change in the 21 st century. The suitability and reliability of the observed datasets have been proven in several studies over the study domain (Ogwang et al., 2015;Ayugi et al., 2016;Ongoma and Chen, 2017;Karendi et al., 2017). The rest of the paper is organized as follows: Section 2 presents the study area, data, and techniques employed, while section 3 gives the results and discussions. Finally, section 4 highlights the critical conclusion and future recommendations for further studies. Figure 1 shows the study area, demarcated within the geographical coordinate of longitude 28º E -40º E and latitude 12º S -5º N. The region is situated on the eastern belt of the African continent and mostly referred to as East Africa (EA), and is composed of five countries, namely Kenya, Uganda, Tanzania, Burundi, and Rwanda. Intricate landscape features such as Mt. Kilimanjaro, Mt. Elgon, Mt. Kenya, Mt. Rwenzori, Lake Victoria, a large expanse of arid and semi-arid lands (ASALs), and highland regions depict fluctuations of temperature from one region to another (Graffiths, 1972;Camberlin, 2018). However, the standard deviation (SD) of mean annual and seasonal temperatures is minimal, with a maximum variation of 2 ºC < SD < 5 ºC experienced. For instance, the lowest temperature range is noted in the region along the equator from western Kenya through southern Uganda, Burundi and Rwanda. Conversely, the highest temperature is experienced over the ASALs regions located over eastern Kenya, northwest/east Kenya and northern Uganda, with maximum temperature occurring throughout the year, mainly attributed to lack of cloud and vegetation cover (Pepin et al., 2014). Annual climatology of temperature range between 5 ºC to 35 ºC while seasonal temperatures occur during December -February (DJF) and

Historical Datasets and Analysis
The study employed thirteen historical models (Table 1) for assessment studies over East Africa.
The surface temperature (tas) datasets were retrieved from the CMIP6 repository (https://esgfnode.llnl.gov/search/cmip6). The first realization (r1i1p1f1) for each model members were considered for all timescales. The models at each timescale were first standardized for unit and calendar date formats. The standardized datasets were then re-gridded to a common grid of lowest model resolution by utilizing the nearest neighbor interpolation technique. The aforementioned technique interpolation follows the better classification of diverse geography by triangulating nearest points and sub-regionalization of grid points by the nearest cell center of input grids (Mallika et al., 2015;Vermeulen et al., 2017).
All re-gridded models were sliced to the 1970 -2014 time period for plotting and further analysis. The annual scale multi-model ensemble (MME) for the historical timeline  was developed from the thirteen models. The MMEs reduce biases by canceling them out partly (Pincus et al., 2008); and distinguish the forced signal and natural variability by averaging the spread in natural variability (Eyring et al., 2016). The CRU TS4.03 (Harris et al., 2020) mean temperature dataset for 1970 -2014 was utilized for benchmarking the models' and MME's performance in simulating annual mean temperature over the study area.
At first, the simulated spatio-temporal climatology for annual mean temperature, MME and CRU datasets were plotted for historical timescale. The temporal anomalies for all datasets were also plotted to observe variability in mean temperature. The performance of models and MME in simulating annual mean temperature against the benchmark CRU was assessed using bias, root mean square difference (RMSD), and correlation coefficient (CC). Mathematical equations for the mentioned metrics are given in Babaousmail et al., 2019;Ongoma et al., 2019;: where M identifies the model simulation, CRU refers to observed values, i refer to observed and simulated pairs, and n stands for a total number of pairs in a time series. Trend analysis provides the spatio-temporal progressions of the changes and variability in a variable or phenomenon. The modified Mann-Kendall (m-MK) trend test (Kendall, 1975;Mann, 1945;Hamed and Rao, 1998) was utilized to detect the significance of spatio-temporal trends over East Africa. The magnitude of trends was measured by the Theil and Sen's Slope estimation technique (Sen, 1968). The mentioned method exempts datasets from normal distribution requirements, non-missing values, and no outliers in a time series. Several studies (Ayugi et al., 2018;Ayugi and Tan, 2019;Mumo et al., 2019;Karim et al., 2020;Tadeyo et al., 2020;Ngoma et al., 2021a, b) have followed the trend test and estimation technique in trend analysis.
Further, models were ranked as per their performance using the Taylor diagram. The Taylor diagram coherently compares and represents the variation of the correlation coefficient (CC), and root mean square difference (RMSD) terms in a single diagram (Taylor, 2001). The Taylor skill score (TSS) summarized from the Taylor diagram was used to rank models: where Rm identify spatial correlation coefficient between the model and observation, Ro is the maximum correlation coefficient (0.999), σm is standard deviation of models' spatial patterns extracted and σo identify the observed spatial patterns for annual mean temperature. A dataset with TSS value near to 1 is considered as the best performing.

Future projection datasets and data analysis
For future projections over EA, five best performing CMIP6 models (based on their performance ranking in historical mean temperature simulations) under the shared socioeconomic pathways SSP2-4.5 and SSP5-8.5 scenarios were employed. The models are FGOALS-g3, HadGEM-GC31-LL, MPI-ESM2-LR, CNRM-CM6-1, and IPSL-CM6A-LR. The first realization members (r1i1p1f1) of models for both SSP scenarios were chosen for projections. At first, all models for each SSP scenario were standardized for unit scale and calendar date formats and then re-gridded to a common grid resolution following the previously mentioned approaches. In order to follow variations in temperature projections, the whole period was divided into the near-term (2020 -2049), mid-term (2050 -2079), and long-term (2080 -2100) periods of the 21 st century, relative to the historical period of 1985 -2014. Consequently, the multimodel ensembles (MME) for both SSP scenarios in the near-, mid-, and long-term future were developed for the analysis. Numerous studies agree on the use of MME in providing robust projections of future climate due to the reduction of inherent biases in individual models (Palmer et al., 2005;Miao et al., 2014;Kim et al., 2015;Ahmed et al., 2019). Future spatio-temporal annual T2m changes and uncertainties over EA were calculated by subtracting the near-term, mid-term, and long-term future MME annual mean temperature values from the historical (1985 -2014) MME annual mean value over the whole region. Besides, the spatio-temporal trends and the significance were calculated using previously mentioned statistical techniques. A number of studies have employed a similar approach to verify the robustness of the moddels in characterizing climatic trends over diverse regions Tadeyo et al., 2020;Karim et al., 2020). This study also focused on uncertainties in the projections of mean temperature and quantify them in the form of probability density function (PDF). The PDF has an integral function, applied to continuous random variables of multiple intervals, and in return gives the chance of occurrences for values in the intervals (Harris et al., 2006).  (Ogallo, 1993;Anyah et al., 2006;Ongoma et al., 2017;Ayugi and Tan, 2019). The lowest temperatures during JJA are a result of cold airmass advected from the south Indian Ocean, featured by Mascarenes High, which enhances the observed low temperature (Ogwang et al., 2015). Consequently, the model Ensemble, MPI -ESM1-2LR, CanESM5, Other studies points to the changes in the forcing data employed that play an important role, particularly the described concentrations of GHGs utilized to force the models (Wyser et al., 2020;Forster et al., 2020;Zelinka et al., 2020;Parsons et al., 2020). In order to assess the models' capability in reproducing the spatial variance, the mean T2m bias is computed as shown in Figure 3. The models show better performance if the simulated bias is close or equal to zero. Models that show higher (lower) values tend to over (under) estimate the temperature patterns. Overall, the majority of the models are able to depict the climatological temperature patterns of the study region. Nearly all models demonstrate an underestimation in regions with a complex topography and over most regions of Uganda. CanESM5 shows an underestimation by more than -6.5 ºC over Mount Kenya region. Contrarywise, warm bias is noted in MIROC6 model, which is more pronounced over eastern Kenya. The mentioned region is characterized by vast ASALs, thereby experiencing high radiation due to less cloud cover. The model MPI-ESM1-2-HR, on the other hand, show pronounced cold bias over the entire domain.

Evaluation of model performance
The models FGOALS-g3, CNRM-CM6-1, GFDL-CM4, IPSL-CM6a-LR, and ensemble demonstrate an agreeable pattern, similar to the observed variance. The listed models have a minimal bias of 1.5 < bias < -1.5. The pronounced cold bias over the western sides of the region could be attributed to the vast vegetation cover and presence of most mountains and highlands, which has an influence on the local temperature variation. The models' ability to reproduce the linear spatial trends varies from one model to another, as presented in Figure 4 and Table 2 The trends were further evaluated and tested for their significance and magnitude. Table 2 shows the mean, slope, Z-score, and significance of linear trend of T2m for CRU and the 13 CMIP6 models. The observed T2m over the region exhibits significant increasing trends during the study duration. The observed mean (Z-score) values are 23.3 ºC (5.93) during 1970 -2014. These results agree with past studies over the study area (Camberlin, 2017;Ongoma et al., 2017;Camberlin, 2018). Remarkably, all models agree on the linear trajectory of T2m over the study region, with CMIP6 models depicting statistically increasing trends. Some models showed higher (lower) mean    , 1983, 1985, 1987, 1987, and 2003). This shows the inability of the models to capture the interannual variability of temperature trends. This feature has been noted in other regions, with models demonstrating incapability to capture major climatic anomalies Gbobaniyi et al., 2013;Ayugi et al., 2020a). The summary of model performance relative to observation over the study region is presented in Figure 6. The majority of GCMs depicted a larger standard deviation than observation with only six models (i.e., HadGEM-GC31-LL, FGOALS-g3, MPI-ESM-2-LR, MRI-ESM2-0, CESM2, and CESM2-WACCM) reproducing similar or smaller standard deviation. Conversely, most CMIP6 models (i.e., 7/13) simulated higher amplitudes depicting a large spatial variation of warming. The mean ensemble demonstrates better performance relative to observation, while MIROC6 shows the largest warm bias associated with overestimations. Spatial correlation analysis of the models highlights acceptable simulations by most CMIP6 models, with MIROC6 depicting the worst performance (CC; 0.29). The CMIP6 models ensemble shows robust simulations compared with individual models with > 0.90. Generally, temperature GCMs demonstrate good performance in reproducing observed spatial patterns over EA domain in the new model generation due to the improved parametrization schemes, enhanced spatial resolution, and physical processes, including the biogeochemical cycles (Eyring et al., 2016). Comparative studies across other domains equally show varying performances, with some studies depicting a slight improved performance of CMIP6 relative to CMIP5 in simulating climate variables (Xin et al., 2020;Zamani et al., 2020;Zhu and Saini, 2020).
The best performing models for this analysis are: CESM2, CESM2-WACCM, HadGEM-GC31-LL, FGOALS-g3, and MPI-ESM2-LR. The consistently poor performance in MIROC6 is noteworthy with attributions to the unskillful simulation of regional climate due to the model's ocean biogeochemical component that has been primarily updated to simulate the biogeochemical cycles of carbon, nitrogen, phosphorous, iron, and oxygen (Hamija et al., 2020). Duan et al. (2013) suggested that the indirect effect of sulfate aerosol could enhance the ''model's capability to simulate the annual and seasonal climate patterns. Conversely, the mentioned chemical (sulfate) was primarily missed in the recent updated biogeochemical cycle in the MIROC6 model. This shows that the enhancement of the mechanism for the observed bias remains a challenge.

Temperature projections for the 21 st century
Examination of possible changes in T2m over EA is conducted using two main scenarios representing plausible trajectories: modest mitigation scenarios (i.e., SSP2-4.5) or worst-case no policy possible pathways (and SSP5-8.5) . Figure 8 shows the annual projected changes in T2m for the 2015 -2100 timeline based on the best performing models and their respective MME under two scenarios. The projected changes over the study domain exhibit warming tendency with highest (lowest) values under SSP2-4.5 projected at 1.8 °C (1.2 °C) and at 3.2 °C (1.5 °C) under the SSP5-8.5 scenario. The MME projects 1.6 °C (2.4 °C) warming over the study region for the SSP2-4.5 (5-8.5) scenario (Figure 8a, b). During the three time slices under consideration, the MME projects many changes during the late period (2080 -2100) with expected mean changes at 2.4 °C for SSP2-4.5 and 4.4 °C for the SSP5-8.5 scenario. On the contrary, the near future (2020 -2049) is expected to experience 1.0 °C warming under the modest mitigation pathways, whereas 1.1 °C mean changes in T2m are expected if no mitigation measures to curb the GHG emissions are put into place. The findings of this study agree with previous studies conducted over varying regions in Africa using CMIP5 models (Ongoma et al., 2018b;Libanda and Ngonga, 2018). The study noted a projected increase in T2m by about 2.4 °C and 4.1 °C under the RCP4.5 and RCP8.5 scenarios. Further, the result of the present study is in harmony with a recent survey conducted over the entire Africa with a focus on various sub-regions such as Central East Africa (CEAF) and southeast Africa (SEAF) (Almazroui et al., 2020a). To illustrate, the study demonstrated a 66 % range of projected warming over CEAF and SEAF during the last period of the century with expected warming at 2.7 °C -4.9 °C under the SSP5-8.5 scenario.
It is interesting to note that while the region is expected to warm rapidly, projections on precipitation show an increase in rainfall over the study domain (Shongwe et al., 2011;Ongoma et al., 2018;Osima et al., 2018;Ayugi et al., 2021). The expected recovery in precipitation is likely to offset the impacts of warming, which could adversely impact agriculture due to the sharp increase in evapotranspiration and subsequent increase in drought and aridity. The historical patterns over the EA region have been characterized by many occurrences of drought/flood incidences, which are mainly a result of anthropogenic influence and changes associated with internal variability; e.g., by El Nino Southern Oscillation(ENSO) and Interdecadal Pacific Oscillation (IPO) (Gu et al., 2013;Lyon, 2014;Hua et al., 2016;Dai, 2016). Overall, the best performing CMIP6 models show the capability to reasonably project possible future temperature changes over the study region (Figure 8a, b). Higher warming is expected during the last century under the SSP5-8.5 scenario. The temperature changes are likely to impact various sectors such as health, energy, agriculture and societal infrastructure during the mid and late century. This will call for robust policy changes to be instituted to avoid loss of livelihoods and property. The spatial distribution of projected changes in T2m over EA is described in Figure 9. The analysis shows warming under varying time scale, relative to the baseline period. The significance

Temperature projections for the 21 st century.
of the changes is further tested based on linear trends and m-MK analysis ( Figure 10 and Table 4).
Notably, the projections based on the SSP2-4.5 scenario during the near future show minimal warming with net change not exceeding 0 °C from the current temperature experienced ( Figure   9a). However, towards the end of the century, pronounced patches of warming of 0.8 °C -1.4 °C are expected over eastern parts of Kenya and western Kenya along with Lake Victoria (Figure 9c).
Under the SSP5-8.5 scenario, significant homogeneous warming is expected over the whole region exceeding 7 °C warming from current levels with more pronounced change over Tanzania and Uganda (Figure 9f).
To illustrate, Iyakaremye et al. (2020)     Finally, the changes in the variability and skewness of T2m over the EA region are presented using the PDFs (Figure 11). Figures 11a and 11b display the predictable PDFs from CMIP6 mean under different scenarios, with a baseline period of 1985 -2014 and projected timescales subdivided into three periods, namely, near-term (2020 -2049), mid-term (2050 -2079), and long-term (2080 -2100). Relative to baseline curves, the SSP5-8.5 scenario shows a clear pattern of a condensed peak, increased spread and a mean shift to the right, inferring large increases in the frequencies of warm extremes over the study region. The changes during SSP2-4.5 depict stabilization patterns towards the end of the century with the high peaks, higher than that of the observed period, and a noticeable increase in mean values of T2m at 25 °C -26 °C ( Figure 11a). Comparative analysis for SSP2-4.5 and SSP5-8.5 shows an increase in mean values from one time slice to another ( Table 4). The results show more robust global warming changes than previous studies based on CMIP5 (Ongoma et al., 2018b). For instance, the PDFs for projected change during the 2071 -2100 period relative to the baseline period (1961 -1990) over EA shows that the mean temperature will be 25.2 and 26.7 under RCP4.5 and RCP8.5 scenarios, while the present study demonstrates that the projected change during the same period will be 25.6 and 27.7 under SSP2-4.5 and SSP5-8.5 scenarios, respectively ( Figure 11 and Table 4). The results of the present study are in harmony with projected changes in T2m using CMIP6 over most regions (Grose et al., 2020;Fan et al., 2020;Tokarska et al., 2020;Almazroui et al., 2020a, b, c;Karim et al., 2021). The findings from this study illustrate higher warming in the latest model outputs of CMIP6 relative to its predecessor, despite identical instantaneous radiative forcing (Wyser et al., 2020). Conclusively, future population pressure, fossil fuel-based industrialization, transportation, commercialization, and land-use changes may intensify the GHG emissions and temperature rise over the EA region. It is necessary to study the evolution and variability in drivers of GHG emission and temperature increase over the study domain to understand future temperature warming-induced impacts in EA. Moreover, it will also help to develop comprehensive, national, and regional environmental, economic, disaster management, and climate change adaptation and mitigation policies.

Summary and Conclusion
This study uses CMIP6 experiments to present an analysis of the ability of the models to simulate current T2m over East Africa and their future changes under three time slices, namely near-term (2020 -2049), mid-term (2050 -2079), and long-term (2080 -2100) periods of the 21 st century, relative to the historical period of 1985 -2014. We examine future changes under modest mitigation and high-emission pathways (i.e., SSP2-4.5 and SSP5-8.5). The performance of models and MME in simulating annual T2m against the benchmark CRU is assessed using mean state, trends, bias, CC, RMSE, and Taylor skill score. The main findings are summarized as follows: The CMIP6-simulated T2m indicate that most of the models can reproduce the climatological temperature patterns of the study region with an aspect of overestimation by the majority of model outputs. Only few models portrayed an underestimation of the annual cycle of T2m. The spatial analysis demonstrates an underestimation in regions with a complex topography and over most areas of Uganda. Remarkably, CanESM5 shows an underestimation by more than -6.5 ºC over the Mount Kenya region.
The ability of the models to reproduce the linear spatial trends varies from one model to another. Most of the models simulated the trends within the observed ' 'range's proximity except CESM2 and CNRM-ESM2-1, which depict a higher positive annual temperature trend. The majority of models recorded higher spatial trends over the western sides of the study area around Uganda, Rwanda, Burundi, and parts of Tanzania. The "models' overall ranking from all the analyses ranging from mean cycle simulation, trend analysis, interannual variability, spatial patterns variability based on RMSE, bias, and CC are as follows: FGOALS-g3, HadGEM-GC31-LL, MPI-ESM1-2-LR, CNRM-CM6-1, and IPSL-CM6A-LR.
The projected changes over the study domain exhibit warming tendency with MME derived from best performing models showing warming over the study region by 1.6 °C (2.4 °C) for the SSP2-4.5 (5-8.5) scenarios. During the three time slices under consideration, the MME project many changes during the late period (2080 -2100) with expected mean changes at 2.4 °C for SSP2-4.5 and 4.4 °C for the SSP5-8.5 scenario. The magnitude of change based on Sen's slope estimator and m-MK reveal significant increasing tendencies during the late century at 0.24 °C decade -1 (0.65°C decade -1 ) under the SSP2-4.5 (SSP5-8.5) scenarios.
This ' 'study's findings illustrate higher warming in the latest model outputs of CMIP6 relative to its predecessor, despite identical instantaneous radiative forcing. Despite the robust results, some limitations feature the present study. Future studies may examine the response to different global warming threshold with more "models' outputs added to limit any uncertainty in the climate projections. Moreover, subsequent studies could also focus on extreme climate factoring the population exposure and local vulnerability following a novel study by Chen and Sun Compliance with ethical standards: Authors agreed to a unanimous correspondence for the publication of the manuscript.