The impact of global changes in near-term climate forcers on East Africa’s climate

Abstract

Climate change and air pollution are two interconnected daunting environmental challenges of the twenty-first century. Globally, stringent public health and environmental policies are set to mitigate the emissions of near-term climate forcers (NTCFs) because they double as air pollutants. While the global climate impact of NTCF mitigation has been investigated using coarse resolution climate models, the fine scale regional climate impacts over East Africa are not fully known. This study presents the first 2021–2055 downscaled model results of two future scenarios which both have increasing greenhouse gas emissions but with weak (SSP3-7.0) versus strong (SSP3-7.0_lowNTCF) levels of air quality control. NTCF mitigation is defined here as SSP3-7.0_lowNTCF–SSP3-7.0. The results reveal that NTCF mitigation could cause an increase in annual mean surface temperature ranging from 0.005 to 0.01 °C decade⁻¹ over parts of Kenya, Ethiopia and Somalia. It could also cause an increase in annual mean precipitation ranging from 0.1 to 1 mm month⁻¹ decade⁻¹ over parts of Uganda, Kenya, Tanzania, South Sudan and Ethiopia. Majority of the precipitation increase is projected to occur during the MAM season. On the other hand, Zambia, Malawi and southern Tanzania could also experience a decrease in annual mean precipitation by up to 0.5 mm month⁻¹ decade⁻¹. Majority of this decrease is projected to occur during the DJF season. These findings suggest that pursuing NTCF mitigation alone while ignoring greenhouse gas emissions will cause additional climate change over East Africa. Mitigating both of them concurrently would be a better policy option.

Introduction

Near-term (short-lived) climate forcers (NTCFs) are atmospheric chemistry species which exert a modifying effect on climate within two decades after they have been emitted or formed (Szopa et al. 2021). This is due to their short atmospheric residence times that range from hours to a few months for most NTCFs. Direct NTCFs such as aerosols, ozone (O₃) and methane (CH₄) exert a radiative forcing which influences the radiation budget of the atmosphere, while indirect NTCFs such as nitrogen oxides (NO and NO₂), sulphur dioxide (SO₂), carbon monoxide (CO) and non-methane volatile organic compounds (NMVOCs) are precursors to the direct NTCFs. Through complex chemical processes, the indirect NTCFs control the atmospheric abundance of the direct NTCFs. For example, ozone is formed through a photochemical reaction between NO_x and volatile organic compounds (Li et al. 2022; Lu et al. 2019), while sulphate and nitrate aerosols are formed through the oxidation of SO₂ and NO₂, respectively (Jacob 2021).

Besides their influence on climate, NTCFs are also air pollutants, and they trigger and/or exacerbate respiratory illnesses like lung cancer, acute lower respiratory infection and chronic obstructive pulmonary disease (World Health Organization 2018). They are also associated with cardiovascular illnesses such as coronary artery disease, cardiac arrest, and heart failure (Miller and Newby 2020). This pollution commonly takes the form of tropospheric ozone and particulate matter with a diameter of less than 2.5 µm (PM_2.5). A small contribution also comes from the precursor gases such as NO₂, SO₂ and CO (Fowler et al. 2020; Pozzer et al. 2023). Globally, ambient air pollution is considered the leading environmental risk factor, causing more than 50% of the deaths which is approximately 4.2 million deaths yearly (Cohen et al. 2017; Pozzer et al. 2023). Due to this, environmental and public health policies now target the reduction of these air pollutants to improve air quality.

Currently, NTCF emissions are concentrated in highly populated regions of the world. These include China, India, eastern United States, and Europe (Hoesly et al. 2018). Africa is also an important emission source for black carbon (BC) and organic (OC) aerosols which are associated with biomass burning. Major emissions of NO_x and SO₂ are also concentrated over the global oceans, along the international shipping routes (Hoesly et al. 2018). Since the year 2000, Eastern and Southern Asia have had the highest levels of NTCF emissions due to industrial development (Szopa et al. 2021). By contrast, over North America and Europe, the emissions of some NTCFs such as NO₂ and SO₂ have declined in the past decade, 2010 to 2019 (Aas et al. 2019; Jiang et al. 2018; Miyazaki et al. 2017). China also had major SO₂ declines within that decade (Zheng et al. 2018). CO abundance is also on a decline globally (Buchholz et al. 2021). These declines were all due to adoption of policies that restrict the emission of NTCFs.

Adoption of these policies will cause future reductions in the emissions and abundances of NTCFs, which will have a small but important impact on climate (Allen et al. 2020). This scientific domain is being explored using model simulations from coupled chemistry-climate models such as version 2 of the Community Earth System Model (Emmons et al. 2020; Gettelman et al. 2019; Tilmes et al. 2019) and the Max Planck Institute Earth System Model (Tegen et al. 2019). First, idealized experiments have been used to investigate the climate impact that will arise from the total removal of anthropogenic emissions of selected NTCFs (Baker et al. 2015; Kasoar et al. 2016; Lelieveld et al. 2019; Samset et al. 2018). Secondly, coordinated modeling experiments have been carried out under the Aerosol Chemistry Model Intercomparison Project (AerChemMIP) as part of phase 6 of the Coupled Model Intercomparison Project (CMIP6). AerChemMIP is designed to quantify the climate and air quality impacts of aerosols and chemically reactive gases, specifically, NTCFs (Collins et al. 2017). The latter is more realistic as it offers a variety of model simulations which are important for quantifying model uncertainty.

To provide context of the full impact that reductions in NTCFs will have on climate, AerChemMIP experiments are compared against experiments from the Scenario Model Intercomparison Project (ScenarioMIP) (O’Neill et al. 2016). Specifically, two future scenarios are compared, these are; SSP3-7.0 from ScenarioMIP and SSP3-7.0_lowNTCF from AerChemMIP. SSP3-7.0 has weak air quality control and high NTCF emissions while SSP3-7.0_lowNTCF has strong air quality control and low NTCF emissions. The mathematical difference between these two experiments is then used to reveal the climate impact (Allen et al. 2020; Collins et al. 2017; Hassan et al. 2022). It is important to note that these experiments have the same levels of well-mixed greenhouse gases and CH₄. Therefore, the climate impact revealed here is due to changes in only non-methane NTCFs.

The first multi-model global assessment of this climate impact was done by Allen et al. (2020) and it revealed that the mitigation of non-methane NTCF emissions would cause a future increase in surface temperature and precipitation due to the net warming effect that the reduction in aerosols would induce. This would also be accompanied by an increase in the number of hottest and wettest days. Over Africa, Allen et al. (2020) noted that the bulk of the precipitation increase would occur in East Africa (Fig. 1). Though useful, their results were based on coarse resolution models which could only offer limited insight. Exploring this problem using model simulations at finer spatial scales might be reveal new insights. Therefore, this study sought to further investigate these projected changes in temperature and precipitation over East Africa by dynamically downscaling one of the CMIP6 global climate model (GCM) experiments analyzed by Allen et al. (2020). This is the first study to downscale these model experiments over East Africa.

Fig. 1

Map of East Africa showing its complex topography and countries. These are; Uganda (UG), Kenya (KE), Tanzania (TZ), Rwanda (R), and Burundi (B) along with the neighboring regions including South Sudan (SS), Democratic Republic of Congo (DRC), Zambia (ZA), Malawi (M), Mozambique (MZ), Somalia (SM) and Ethiopia (ETH). This is also the simulation domain used for the downscaling process in WRF

Since most GCM output is at coarse spatial scales of between 100 and 500 km, they are unable to accurately resolve local features such complex topography, mesoscale circulations, land use/land cover and coastlines, all of which are important in determining local climate patterns (Giorgi and Gutowski 2015; Walton et al. 2015; Xu et al. 2019). Due to this, GCM output only offers limited insight into local climate dynamics. To improve their utility, downscaling is carried out to provide finer detail in the spatial and temporal patterns, which offers several advantages. For example, the finer resolution and improved representation of topography allows for better representation of near-surface temperature gradients and the rain shadow effect that is associated with steep mountain ranges (Di Luca et al. 2012, 2015). Typically, for the GCM output to be useful for local applications, it has to be downscaled to a resolution of at least 25 km or something finer than that. This is the based on recent downscaling studies (Fernández-Alvarez et al. 2023; Rahimi et al. 2020; Xu 2021; Yang et al. 2023; Ye et al. 2022; Yu et al. 2023). Over East Africa, the impact that global mitigation of NTCFs will have on temperature and precipitation at such fine spatial scales is not fully known. This study is the first attempt to investigate this impact.

Data and methods

Study area

East Africa’s climate requires keen scientific study because of the complex mix of factors that influence the region’s climate. For example, the biannual crossing of the Intertropical Convergence Zone (ITCZ) creates a bimodal precipitation cycle, consisting of the long precipitation season that occurs from March to May (MAM) and the short precipitation season that occurs from September to November (SON) (Nicholson 2019). The El Niño Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD) enhance precipitation receipts over East Africa during SON, thus creating more interannual precipitation variability during SON compared to MAM (Nicholson 2019; Wenhaji Ndomeni et al. 2018). The other important factors include the Madden–Julian Oscillation, tropical cyclones, and the existence of mountains and large inland lakes (Finney et al. 2020; Nicholson 2017; Walker et al. 2020). There is also the East African climate paradox that has further confounded the study of the region’s climate (Mölg and Pickler 2022; Wainwright et al. 2019).

The historical climate trends show that East Africa’s precipitation has remained the same in the majority of the region and only changed in a few areas. For example, during the period 1981 to 2017, a precipitation increase of between 3 and 15 mm year⁻¹ occurred in the region between Uganda and Kenya that surrounds Lake Victoria. A precipitation decrease of 20 mm year⁻¹ occurred near Mt. Kilimanjaro in northern Tanzania and other declines in precipitation of ~ 4 to 10 mm year⁻¹ occurred over small regions in central Kenya, central Uganda, southwest Tanzania, and western Rwanda (Gebrechorkos et al. 2019a; Muthoni et al. 2019). Over Uganda alone, a small increasing trend in precipitation was identified starting in 2010 (Ngoma et al. 2021). On the other hand, East Africa’s temperature has shown an increasing trend in the majority of the region. For the 1979 to 2010 period, this increase reached 1.9 °C (Gebrechorkos et al. 2019a, b).

Concerning the future climate projections, the majority of the CMIP6-based climate modeling studies over the region have focused on the role of well-mixed greenhouse gases (Akinsanola et al. 2021; Ayugi et al. 2021, 2022a; b; Makula and Zhou 2022; Vashisht et al. 2021). By contrast, this study has investigated the role of NTCFs since they also pose a threat to East Africa’s climate.

Data

Much as statistical (Thrasher et al. 2022) and machine learning-based (Baño-Medina et al. 2020) downscaling is possible, this study preferred a physics-based dynamical downscaling because its results are physically consistent and scientifically reliable. Therefore, this study downscaled the Max Planck Institute Earth System Model (MPI-ESM1-2-HAM) using a physics-based regional climate model, that is, the Weather Research and Forecasting (WRF) model. The downscaling was from 250 to 23 km (Fig. 2). The two models and their data are described below.

Fig. 2

Near surface air temperature on 1st July 2030 as simulated by MPI-ESM1-2-HAM (250 km) and WRF (23 km) for SSP3-7.0_lowNTCF

MPI-ESM1-2-HAM model

The MPI-ESM1-2-HAM model is a fully coupled ocean-atmosphere model which was run for CMIP6 by the HAMMOZ Consortium (Mauritsen et al. 2019; Neubauer et al. 2019). The model’s atmosphere component was set up with a horizontal resolution of 250 km and 47 vertical levels up to 0.01 hPa (Neubauer et al. 2019). This model used identical ozone levels in both SSP3-7.0 and SSP3-7.0_lowNTCF experiments and only aerosol levels were allowed to change. Therefore, the differences in climate generated here were only due to changes in aerosol amounts. Further, the experiments were only run from 2015 to 2055 because this is the time period during which changes in aerosol and precursor emissions are expected to be significant (Collins et al. 2017).

Among the nine coupled chemistry-climate models (Table 1) that performed both SSP3-7.0 and SSP3-7.0_lowNTCF experiments, only MPI-ESM1-2-HAM was selected for the downscaling in this study. This model was the only one that provided outputs for both experiments at a 6-hourly temporal resolution. Data at such a high temporal resolution was necessary for generating realistic boundary conditions for driving WRF. The variables used were the eastward and northward near surface wind, eastward and northward wind, near surface specific humidity, specific humidity, soil temperature, moisture in the upper portion of the soil column, surface air pressure, sea level pressure, surface temperature, near surface air temperature, air temperature and geopotential height. All these data were downloaded from the Earth System Grid Federation (ESFG) database hosted at the Lawrence Livermore National Laboratory.¹ To be suitable input data for WRF, the MPI-ESM1-2-HAM variables were preprocessed into 6-hourly intermediate files, with each file containing all the variables. This was done in Python 3.9.7. Thereafter, the data were interpolated onto the WRF model grid using the metgrid function within the WRF preprocessing system.

Table 1 The nine models that performed both SSP3-7.0 and SSP3-7.0_lowNTCF experiments under CMIP6

In addition to the meteorological variables, the model output for PM_2.5 mass mixing ratio in lowest model layer was also obtained and analyzed to show the difference in PM_2.5 aerosol burden between SSP3-7.0 and SSP3-7.0_lowNTCF. The data were converted from mass mixing ratio (kg kg⁻¹) to concentration (μg m⁻³) following the procedure described by Gomez et al. (2023). The quantity was multiplied by air density which was calculated as in Eq. (1). The unit of pressure was Pascals, while that of temperature was Kelvins. The dry gas constant was 287 JK⁻¹ kg⁻¹.

$$Air\, density= \left(\frac{Surface\, pressure }{Surface\, temperature\times Dry\, gas\, constant}\right)\times {1e}^{9}$$

(1)

Weather Research and Forecasting (WRF) model

Version 4.3 of the WRF model, Advanced Research WRF dynamics solver (Skamarock et al. 2019) was used for the downscaling experiments. It is a regional atmospheric modeling system based on the physical equations. WRF is designed with Arakawa C-grid staggering, 2nd and 3rd order Runge–Kutta time integration, and a terrain-following hybrid sigma-pressure vertical coordinate system (Skamarock et al. 2019). For this study, the model was set up with one domain (Fig. 1) at a horizontal resolution of 23 km (110 × 100 grids) and a vertical resolution of 31 eta levels ending at 50 hPa at the top.

The model was set up with the Lin microphysics (Lin et al. 1983), Grell 3-D cumulus parameterization (Grell and Dévényi 2002), the Noah land surface model (Chen and Dudhia 2001), Yonsei University Scheme for the boundary layer parameterization (Hong et al. 2006) and the Rapid Radiative Transfer Model for GCMs (Iacono et al. 2008) for both shortwave and longwave radiation parameterization. These parameterization choices were based on previous studies focused on East Africa (Nooni et al. 2022; Otieno et al. 2019). The model was run from 2021 to 2055 at intervals of 1 month while using 1 extra day for model spin-up.

WRF generates about 255 different output variables, but the ones that are associated with precipitation are the accumulated total cumulus precipitation and the accumulated total grid scale precipitation. The two components were summed up to obtain the total accumulated precipitation value that was analyzed here. Furthermore, the variable directly associated with surface temperature is the 2-m temperature variable. The other variables include the eastward and northward surface wind, eastward and northward wind at model levels, surface pressure, total pressure, surface mixing ratio and geopotential height. The full list of variables can be found in Skamarock et al. (2019).

SSP3-7.0 and SSP3-7.0_lowNTCF emissions

The SSP3-7.0 scenario assumes a future with high inequality between and within countries, the so-called ‘regional rivalry’. It is envisaged that there will be weak and non-uniform air pollution legislation and no greenhouse gas mitigation (Fujimori et al. 2017). Consequently, the world under this scenario will have the highest emission levels of NTCFs and substantially high levels of greenhouse gases. The latter is the reason why some literature refer to this scenario as ‘lacking climate policy’ (Fujimori et al. 2017; Gidden et al. 2019). By contrast, the SSP3-7.0_lowNTCF scenario uses the same assumptions as SSP3-7.0 except that it assumes a world where stringent policies are enacted to mitigate NTCFs but while completely ignoring greenhouse gas emissions. Therefore, NTCF emissions decrease while greenhouse gas emissions continue to increase (Gidden et al. 2019). SSP3-7.0_lowNTCF assumptions for NTCFs are similar to those under the sustainability scenario, SSP1. For example, it uses the same CH₄ emission reduction rates and the same emissions factors for the other air pollutants, that is, NO_x, CO, BC, OC, NH₃, VOC and sulfur (Gidden et al. 2019). For this study, the NTCF emissions used for SSP3-7.0 and SSP3-7.0_lowNTCF were obtained from the input forcing data for CMIP6 (Gidden et al. 2018a, b). These data were also downloaded from the ESFG database hosted at the Lawrence Livermore National Laboratory.² The data for 2015 to 2055 were plotted to show the difference in NTCF emissions between the two experiments.

Methods

Difference variable

The difference variable (Willmott et al. 1985) was applied to the model output from both SSP3-7.0 and SSP3-7.0_lowNTCF as demonstrated in earlier studies (Allen et al. 2020; Collins et al. 2017; Hassan et al. 2022). The effect of NTCF mitigation was calculated as shown in Eq. 2.

$$NTCF\, mitigation=SSP3\_7.0\_lowNTCF -SSP3\_7.0$$

(2)

Mann–Kendall trend test and Sen’s slope estimator

Two non-parametric tests, that is, Mann–Kendall (Kendall 1975; Mann 1945) and Sen’s slope estimator (Sen 1968) were applied in both spatial and temporal context to test for the existence of a trend at a significance level of 95%. This implies that the standard normal statistic, Z would have limits as + 1.96 and − 1.96. These methods have previously been applied in a similar way to meteorological time series data and were found adequate in revealing the trend (Alemu et al. 2015; Gebrechorkos et al. 2019a; Gocic and Trajkovic 2013; Muthoni et al. 2019; Ngoma et al. 2021; Onyutha 2016). In this study, the methods were applied to PM_2.5, temperature and precipitation data. All the data covered 35 years (2021 to 2055). For temperature and precipitation, the temporal trends were only done for countries whose geographical extent was fully contained within the study area. These were; Uganda, Kenya, Tanzania, Rwanda and Burundi. For the neighboring countries; DRC, South Sudan, Ethiopia, Somalia, Mozambique, Malawi and Zambia, the temporal trend statistic would not make complete scientific sense as only small portions of the geographical areas of these countries were contained within the study region.

Results and discussion

SSP3-7.0 and SSP3-7.0_lowNTCF emission differences

Figure 3 shows the global average emission estimates of major NTCFs from 2015 to 2050. Under SSP3-7.0, the emissions of CO and BC are projected to increase until 2040 after which they will start to decrease. OC emissions will also increase until 2040 and remain fairly constant until 2050. Emissions of VOC will increase from 2015 to 2050. NO_x emissions will increase up to 2020 and remain fairly constant until 2030, after which they will decrease. SO₂ emissions are projected to decrease. The decrease from 2015 to 2020 will be stronger than from 2020 to 2050. The increase in the majority of the NTCF emissions under SSP3-7.0 are estimated to mainly come from central Africa and southeast Asia and are associated with the continued reliance on fossil fuel sources for energy, transport and cooking needs (Gidden et al. 2019).

Fig. 3

Global average anthropogenic emission estimates (kg m⁻² s⁻¹) of major NTCFs from 2015 to 2050. These were the emission trajectories used for SSP3-7.0 and SSP3-7.0_lowNTCF experiments. The horizontal black line shows the 2015 value

Under SSP3-7.0_lowNTCF, the emissions of CO, BC and OC remain constant until 2020 after which they start to decrease until 2050. The NO_x emissions follow a similar trajectory, although they show a small decrease from 2015 to 2020. Emissions of VOC increase from 2015 to 2020 after which they decrease until 2050. Lastly, the SO₂ emissions will decrease sharply from 2015 to 2050. These NTCF emission reductions will be due to adoption of air pollution reduction policies.

Changes in PM_2.5

Figure 4 shows the projected changes in annual global average PM_2.5 from 2015 to 2055. Within the model framework, PM_2.5 is generated by the direct contribution of all fine aerosol varieties including nitrate, sulphate, carbonaceous, ammonium, dust and sea salt. The difference in PM_2.5 between SSP3-7.0 and SSP3-7.0_lowNTCF will be most apparent starting in 2045. That is when the mitigation efforts are estimated to start showing large benefits. Furthermore, as expected, NTCF mitigation will cause a significant decrease in PM_2.5 at a rate of 0.0045 μg m⁻³ decade⁻¹ over the 2015 to 2055 period and this expected to generate large benefits for air quality.

Fig. 4

a Projected changes in annual global average PM_2.5 concentration under SSP3-7.0 and SSP3-7.0_lowNTCF. b Projected changes due to NTCF mitigation. The plot includes the Mann–Kendall (MK) statistic and Sen’s slope at a 95% confidence level

Downscaled temperature

Figure 5 shows the spatial trend of the projected mean surface temperature over the 2021 to 2055 period under annual, December–January–February (DJF), March–April–May (MAM), June–July–August (JJA) and September–October–November (SON) aggregations. Both SSP3-7.0 and SSP3-7.0_lowNTCF show significant warming across most of the region. This projected warming generally ranges from 0.001 to more than 0.015 °C decade⁻¹ and is associated with increasing CO₂ and CH₄ levels in both scenarios (Allen et al. 2020). A small but insignificant cooling is also projected during SON over Lake Malawi under SSP3-7.0.

Fig. 5

Projected annual and seasonal mean surface temperature trend over the 2021 to 2055 period. Stippling denotes Mann–Kendall trend significance at the 95% confidence level

The resultant NTCF mitigation signal shows warming in the majority of the region. The annual mean warming ranges between 0.005 and 0.01 °C decade⁻¹ and its mostly significant over parts of Kenya, Ethiopia, Somalia, and the Indian Ocean. DJF, MAM, JJA and SON have areas of both warming and cooling, although only the warming was significant. During MAM the warming intensifies to over 0.015 °C decade⁻¹ over southern Ethiopia and northern Tanzania. A similar warming trend happens during JJA over Ethiopia and Lake Victoria. During the same season, warming ranging between 0.005 and 0.01 °C decade⁻¹ occurs over the ocean. This projected warming due to NTCF mitigation results from global increase in the effective radiative forcing when aerosol loadings are reduced. Consequently, the cooling that comes from aerosol–radiation and aerosol–cloud interactions is reduced, and this causes the warming (Allen et al. 2020; Smith et al. 2020; Westervelt et al. 2015).

Surface temperature changes during DJF were not significant. Further, surface temperature was also averaged over selected countries to obtain time series of annual and seasonal aggregations for the 2021 to 2055 period (Fig. 6). When the Mann–Kendall test and Sen’s slope estimator were applied in this context, no trends were found (Table 2).

Fig. 6

Time series of area-averaged, projected surface temperature due to NTCF mitigation. The trend values are shown in Table 2

Table 2 Mann–Kendall test, standard normal statistic (Z) and Sen’s slope (°C decade⁻¹) for surface temperature

Downscaled precipitation

Figure 7 shows the trend of projected accumulated precipitation over the 2021 to 2055 period. Under SSP3-7.0, the region is dominated by a decrease in annual mean precipitation ranging between 0.1 and 0.5 mm month⁻¹ decade⁻¹. This was most significant over the southern Kenya–northern Tanzania area and over the upper region of the Indian Ocean. This projected decrease is emphasized during JJA and SON. A significant precipitation decrease of about 0.5 mm month⁻¹ decade⁻¹ is seen over South Sudan and the Indian Ocean during JJA and over large parts of Zambia, Tanzania, Kenya and Somalia during SON. During both seasons, a stronger precipitation decrease ranging from 0.7 to over 1.5 mm month⁻¹ decade⁻¹ is seen over Lake Victoria and the coastal areas of Kenya and Tanzania. The DJF season differs from the other seasons, as it is dominated by precipitation increase ranging from 0.1 to over 1.5 mm month⁻¹ decade⁻¹. This was significant over central Uganda and the Lake Victoria shores, northeastern Kenya, Somalia, parts of South Sudan and Ethiopia and parts of western and southern Tanzania.

Fig. 7

Projected annual and seasonal trends in accumulated precipitation over the 2021 to 2055 period. Stippling denotes Mann–Kendall trend significance at the 95% confidence level

Under SSP3-7.0_lowNTCF, the region is dominated by increase in annual mean precipitation. This is between 0.1 and 0.5 mm month⁻¹ decade⁻¹ over Rwanda, northern Tanzania, southern and northern Uganda, west and central Kenya, parts of Ethiopia and South Sudan. The largest increase ranging from 1 to over 1.5 mm month⁻¹ decade⁻¹ is over Lake Victoria and the southern shore of Tanzania that borders the Indian Ocean. A small decrease in annual mean precipitation is seen over Zambia, Malawi and southern Tanzania. This decrease ranges from 0.1 to 0.5 mm month⁻¹ decade⁻¹. These precipitation changes are emphasized during different seasons. Precipitation increase over the southern coast of Tanzania is emphasized during DJF and MAM while the increase over Lake Victoria and the land areas is emphasized during MAM and SON. On the contrary, the precipitation decrease over the land areas is emphasized during SON, where large parts of Zambia, Malawi, southern Tanzania and northern Mozambique show a precipitation decrease of between 0.5 and 1 mm month⁻¹ decade⁻¹. Precipitation decrease over the Indian Ocean is emphasized during JJA. There is also a precipitation increase over the ocean during MAM although not significant.

The resultant NTCF mitigation signal shows that the region is dominated by an increase in annual mean precipitation. This projected increase ranges from 0.1 to 1 mm month⁻¹ decade⁻¹ over northeastern Uganda, west and central Kenya, northern Tanzania and parts of South Sudan and Ethiopia. However, there is also a decrease in annual mean precipitation of up to 0.5 mm month⁻¹ decade⁻¹ in parts of Zambia, Malawi and southern Tanzania. This projected decrease in precipitation is emphasized during DJF, where it ranges from 0.5 to over 1.5 mm month⁻¹ decade⁻¹. In the same season, other areas are seen to experience a precipitation decrease. These are central Uganda at about 0.5 mm month⁻¹ decade⁻¹, the Lake Victoria shore at about 1.5 mm month⁻¹ decade⁻¹ and northeastern Kenya at about 0.5 mm month⁻¹ decade⁻¹. There is also a small area in southern Kenya which has an increase in precipitation of up to 0.5 mm month⁻¹ decade⁻¹.

The bulk of the precipitation increase is projected to happen during MAM and JJA. During MAM, it ranges from 0.1 to 1 mm month⁻¹ decade⁻¹ over Rwanda, Burundi, western Tanzania, northeastern Uganda, western Kenya and Somalia. In the same season, larger increases of over 1.5 mm month⁻¹ decade⁻¹ are seen over Lake Victoria, the upper coastal region of Kenya and Somalia, the southern coast of Tanzania and over the ocean. During JJA, the precipitation increase ranges from 0.1 to just over 1 mm month⁻¹ decade⁻¹ and is mostly over western Kenya, Ethiopia and South Sudan. A stronger increase of about 1.5 mm month⁻¹ decade⁻¹ is also seen over Lake Victoria. The precipitation increase during SON is largely insignificant. Furthermore, the spatially averaged precipitation time series for selected countries were also obtained and tested for a trend (Fig. 8 and Table 3). An increasing precipitation trend occurs over Uganda, Kenya and Rwanda at rates of 0.196, 0.116 and 0.187 mm month⁻¹ decade⁻¹ respectively. Significant increments also occur during the MAM season over Kenya, Rwanda and Burundi at rates of 0.278, 0.285 and 0.265 mm month⁻¹ decade⁻¹ respectively.

Fig. 8

Time series of area-averaged, projected accumulated precipitation due to NTCF mitigation. The trend values are shown in Table 3

Table 3 Mann–Kendall test, standard normal statistic (Z) and Sen’s slope (mm month⁻¹ decade⁻¹) for precipitation

The changes in aerosol–radiation and aerosol–cloud interactions also explain the increase and decrease in precipitation, although the stronger signal is the increase. This study aligns well with prior studies that show that aerosols have had and continue to have an important influence on East African precipitation (de Graaf et al. 2010; Mmame et al. 2023; Scannell et al. 2019). Further, this study has also shown that in addition to causing an increase and decrease in the precipitation over East Africa, NTCF mitigation will specifically cause the bulk of the precipitation increase during MAM. These results contradict Allen et al. (2020) who highlighted that NTCF mitigation only causes an increase in precipitation and mainly during DJF. This contradiction probably exists because the present study only downscaled one CMIP6 model and yet the results of Allen et al. (2020) are based on a multi-model ensemble of nine CMIP6 models. Furthermore, downscaling with WRF introduces additional biases and uncertainty to those already existing in the raw CMIP6 model output. Quantifying this model bias could help refine the results but it has not been addressed in the current scope of the study. Despite this, both studies agree on the increase in surface temperature.

Summary and conclusions

This study downscaled the MPI-ESM1-2-HAM global climate model output using the WRF regional climate model. Model experiments for two future scenarios, that is, one with weak air quality control (SSP3-7.0) and the other with strong air quality control (SSP3-7.0_lowNTCF) were downscaled from 250 to 23 km in order to study the local-scale projected climate change over East Africa due to global mitigation of NTCFs. This study concludes that global efforts to mitigate NTCFs could indeed improve air quality but it could also cause significant climate change in East Africa. Specifically, it could cause an increase in surface temperature in large parts of Kenya and some parts of Ethiopia and Somalia. It could also cause an increase in precipitation in several parts of the region, including Uganda, Kenya and Rwanda. Majority of the increase is projected to occur during the MAM season. On the other hand, parts of Zambia, Malawi and southern Tanzania could also experience a decrease in precipitation especially during the DJF season. Therefore, to avoid such a future in which air quality is improved but climate change worsened, it is recommended that both NTCFs and greenhouse gases be mitigated concurrently, both locally and on a global scale. If such policies are used, it will help achieve a double benefit of improving air quality and combating climate change.

For further understanding of the possible climate change due to NTCF mitigation, future studies should be done using daily values of the meteorological variables. This will make the calculation of the climate change indices (Karl et al. 1999) possible. This was not possible in the present study because the assessments were based only on monthly values. If possible, future studies could also downscale model experiments that have all NTCFs including CH₄. Such results can be used to reveal the climate impact of CH₄ mitigation alone and/or the mitigation of all NTCFs plus CH₄. Since CH₄ has a positive radiative forcing, its mitigation could cause cooling which might offset the warming caused due to the mitigation of the other NTCFs. This kind of investigation was not possible in the present study because the MPI-ESM1-2-HAM model did not perform the experiment that included CH₄ mitigation. In addition, the results presented here are thought to have a level of bias but this was not quantified within the current scope of the study. Future studies could cover this as well. Lastly, in future, NTCF emission estimates will be improved and updated using new methodologies. This will help improve the model simulations and advance our understanding of the climate impact of NTCF mitigation.