East African population exposure to precipitation extremes under 1.5 °C and 2.0 °C warming levels based on CMIP6 models

The study utilizes historical and future precipitation and surface air temperature datasets from 26 CMIP6 models (Eyring et al 2016). The datasets cover two time slices: 1850–2014 and 2015–2100 for historical and future periods, respectively. The comparison analysis is conducted relative to the baseline period given in the AR6: 1995–2014. The SSP2-4.5 and SSP5-8.5 scenarios that represent modest mitigation and worst-case scenarios are utilized in the study. The SSP2-4.5 scenario is considered a more plausible outcome where modest mitigation implementation will curb global warming to ∼2.5 °C warming relative to the pre-industrial period by the end of the 21st century (O'Neill et al 2017). On the other hand, SSP5-8.5, also referred to as 'business as usual', represents a fossil-fuel intensive future, void of stringent climate mitigation, leading to nearly 5 °C of warming by the end of the century. The choice of the two radiative forcing scenarios from the available five possible frameworks is informed based on the assumption that differences in climate outcomes from the different scenarios for the same global pathways are likely small relative to varying regional climate features or/and inter-model uncertainties (O'Neill et al 2017). The first realization ensemble is considered in the study to allow equal comparable analysis, except for a few GCMs having the first variation member as r1i1p1f2 or r1i1p1f3. Table S1 (available online at stacks.iop.org/ERL/17/044051/mmedia) shows the first ensemble member of models employed in the study, their country of origin, and native resolutions.

The estimation of population changes is established using projections obtained from SSPs scenarios (Jones and O'Neill 2016). The socio-economic datasets are sourced from the Inter-Sectoral Impact Model Intercomparsion Project framework (Warszawski et al 2014). The SSPs for projected population are available at a global scale on 50 × 50 km grid-level. Given the varying grid resolution for models and observed datasets, re-gridding is performed at 1° × 1° using a bilinear interpolation technique. Jones (1999) recommends using bilinear interpolation when re-gridding from coarser to finer resolution, while a conservative technique can be employed when re-gridding from finer to coarser resolution. For consistency, this study combines the population under SSP2 and SSP5 with CMIP6 models under both scenarios to estimate population exposure correspondingly. Liu et al (2020) employed a similar approach.

2.2.1. Timings of the 1.5 °C and 2 °C warming targets

2.2.2. Extreme precipitation indices and estimates of avoided impacts

The study employs five precipitation indices defined by the expert team on climate change detection and indices. The summary of the indices used in this study is presented in table 1. Each index represents different categories of precipitation occurrences, such as precipitation intensity, duration, and frequency. They include the duration of dry days (CDD), the simple daily intensity (SDII), very heavy precipitation (>20 mm) days greater (R20mm), extremely wet days (R95p), and annual total precipitation (PRCPTOT). Zhang et al (2011) provide details on the climate indices used for examining the impacts of climate change. The reliability of the climate indices has been affirmed in numerous studies across various regions (Akinsanola et al 2020, 2021, Dike et al 2020, Zhu and Yang 2020, Ayugi et al 2021b). To assess how the precipitation indices differ between future and reference period, the climatological mean difference was computed and the Student t-test for unequal variances was performed to evaluate their statistical significance at the 95% confidence level. In order to ascertain the possible scenario of avoiding the impacts of higher warming at 2.0 °C above pre-industrial levels, the avoided impacts (AI) caused by additional 0.5 °C warming is computed from equation (1);

$$\begin{equation}{\text{AI}} = \left[ {\left( {\frac{{{\text{G}}{{\text{W}}_{2.0}} - {\text{G}}{{\text{W}}_{1.5}}}}{{{\text{G}}{{\text{W}}_{2.0}}}}} \right)} \right] \times 100\% \end{equation} \tag{ 1 }$$

Table 1. Names, abbreviations, definitions, and units of climate indices used in the study.

Category	Description	Acronym	Unit
Duration indices	Consecutive dry days	CDD	day
Percentile-based index	Extremely wet days	R95p	mm
Threshold-based index	Heavy precipitation days	R20mm	days
Intensity-based index	Wet-day intensity	SDII	mm d−1
Annual precipitation total	Wet-day precipitation amount	PRCPTOT	mm

where GW_2.0 and GW_1.5 denote the change in 1.5 °C and 2.0 °C warming, respectively, relative to the present baseline. A similar approach has been used in other recent studies (e.g. Chen et al 2020, Wang et al 2020).

2.2.3. Exposure to extreme precipitation

In the present study, population exposure (PE) is defined as the number of people exposed to R95p or CDD (Jones et al 2015, Liu et al 2017, Coffel et al 2018, Zhao et al 2021). The R95p (CDD) represents the prevalence of flood (drought) extremes that can affect the region (Kilavi et al 2018, Ongoma et al 2018, Haile et al 2020, Tan et al 2020, Wainwright et al 2020). The PE for R95p is measured in person.mm while CDD is computed in person.day. The study adopts 20 year averages for precipitation extremes and population to reduce the inter-annual variations (Chen et al 2020). Population exposure to EPIs is estimated by multiplying the annual frequency of extreme precipitation events (i.e. R95p and CDD) by the number of people (Jones et al 2015, Chen et al 2020). Considering the two SSPs employed in CMIP6 models (SSP2-4.5 and SSP5-8.5) and the two SSPs populations (SSP2 and SSP5), the study adopts a two by two matrix of climate and population scenarios. This framework presents the basis for computing the expected EPIs related exposure and are identified as SSP2-4.5|SSP2 and SSP5-8.5|SSP5. The multi-model ensemble mean of 26 CMIP6 GCMs is used to compute the spatial PE over the study region. The roles of climate and population changes on the exposure were also investigated as employed in previous studies (e.g. Liu et al 2017, Chen et al 2020, Iyakaremye et al 2021, Ma and Yuan 2021). It is noteworthy to mention that the EA population experiences diverse risks, given that the magnitude of exposures is due to weak adaptive structures and other multi-faceted challenges (Ahmadalipour et al 2019). Nevertheless, this study focuses on climate risks and not estimations of the changes in other vulnerability and risk evaluations. The study employs a technique proposed by Jones et al (2015) that examines the influence of climate and population on exposure through the decomposition of the change and exposure into three effects: climate effect, population effect, and nonlinear interaction effect. The exposure change $\left( {\Delta {\text{PE}}} \right)$ can be expressed mathematically using equation (2);

$$\begin{equation}\Delta {\text{PE}} = {Y_1} \times \,\Delta X + {X_1} \times \Delta Y + \Delta X \times \Delta Y\,\,\,\,\end{equation} \tag{ 2 }$$

where $\Delta PE\,$ is the total change in population exposure, ${Y_1}$ and ${X_1}$ are population and precipitation extremes, respectively, in the baseline period. $\Delta Y$ and $\Delta X$ are the changes in population and precipitation extremes at 1.5 °C and 2.0 °C warming levels relative to the baseline period. The term ${Y_1} \times \,\Delta X$ is the climate effect. The study ascertains the effect of climate on exposure by controlling the population in the baseline period while allowing the climate to change according to the projected precipitation extremes. The ${X_1} \times \Delta Y$ signifies population effect. In this segment, the population is allowed to vary according to the predicted population, leaving climate unchanged in the baseline period. ΔX × ΔY specifies the interaction effect (i.e. the change in exposure that results from simultaneous change in both climatic and population effects) and is computed under all the selected SSPs × SSPs combinations as the difference between total exposure and the sum of the climate and population effects. Detailed explanations and mathematical equations can be obtained in related literature that examined changes in exposure to heat extremes or precipitation extremes at a global level or continental level (e.g. Liu et al 2017, Chen et al 2020, Ma and Yuan 2021).

Figure 1(a) presents the projected area-averaged population distribution under SSP2 and SSP5 scenarios, respectively, for 2015–2100, while figure 1(b) shows the spatial area-averaged population density over EA during 1995–2014. Currently, the population density is mainly concentrated over the western parts of EA, with countries like Burundi, Rwanda, Uganda, and western Kenya having a greater percentage of the population as compared to other regions (figure 1(b)). The eastern side of Kenya and parts of Tanzania have a sparse population density distribution, except for the coastal belts.

Zoom In Zoom Out Reset image size

The changes in precipitation indices, except for CDD and heavy precipitation days (R20mm), are expressed as percentage changes relative to the reference period. Indices such as PRCPTOT, SDII, CDD, and R20mm are presented in the supplementary material as figures S3–S6, respectively. For demonstration, the results of R95p are shown in figure 2. The region experiences a pronounced increase in R95p by approximately 12% under SSP2-4.5 based on 1.5 °C global warming. However, the increase in R95p under SSP5-8.5 considering 2.0 °C global warming rises to about 18%. The increase in R95p under SSP2-4.5 scenario in 1.5 °C warming climates is approximately 8% smaller than in 2.0 °C, and even 13% smaller in 0.5 °C less warming between 1.5 °C and 2.0 °C warming climates. The areas with insignificant R95p change differences are mainly observed in southern Kenya and much of central and eastern Tanzania (figure 2(c)). The maximum increase occurs in eastern Kenya, consistent with the findings of Osima et al (2018) that was based on the CMIP5 dataset. The spatial distribution of changes for SDII is similar to that of R95p, while the change values in SDII are smaller than that of R95p (figure S4). The area-mean increase in SDII across EA is also lower under SSP5-8.5 than under SSP2-4.5 scenario in 1.5 °C warming climate (figure S4).

Zoom In Zoom Out Reset image size

Spatial mean average of extreme precipitation indices over EA, and the inter-model spread for 1.5 °C and 2.0 °C of global warming are presented in figure 3, while spatial changes for CDD and R20mm are shown in figures S5 and S6, respectively. Unlike other extreme precipitation indices, CDD show a lower change for SSP5-8.5 compared to SSP2-4.5 at 2.0 °C warming climates (figures 3(b) and S5). Besides, large uncertainties can be seen from the models for CDD across EA under the SSP5-8.5 scenario, with model spreads ranging from −2 to 3 d under the 1.5 °C warming target and −2 to 5 d under the 2 °C warming target (figure 3). Spatial analysis shows significant changes along southern Tanzania, where a notable increase in dry days is projected to occur at 6–8 d for SSP2-4.5 and SSP5-8.5 scenarios, respectively (figures S5(a), (b), (d), (e)). There is a remarkable difference in 0.5 °C less warming, with most parts of the region showing fewer changes in CDD in both scenarios (figures S5(c) and (f)). The magnitude of precipitation changes in R20mm shows significant changes along western sides and is less pronounced under SSP5-8.5/1.5 °C as compared to SSP2-4.5/1.5 °C (figures S6(a) and (d)). Nonetheless, under SSP2-4.5/2.0 °C and SSP5-8.5/2.0 °C scenarios, the region will experience significant changes in R20mm across most parts (figures S6(b) and (d)). Most parts of the study area will encounter a remarkable increase in R20mm under SSP2-4.5/0.5 °C as compared to SSP5-8.5/0.5 °C (figures S6(c) and (f)). Persistent, significant occurrence of R20mm is projected to occur over the western sides of the study area under the two scenarios for 0.5 °C less warming (figures S6(c) and (f)).

Zoom In Zoom Out Reset image size

In general, for 1.5 °C and 2.0 °C of global warming, the extreme precipitation indices averaged over EA are projected to increase under both scenarios, except for CDD that demonstrate a decrease in the northern-most parts of Kenya and Uganda. The largest increases occurred in eastern and northern Kenya for PRCPTOT, R95p, and SDII. On the other hand, a notable increase for R20mm occurs over the western sides of the study area under SSP5-8.5/1.5 °C. Moreover, the projected increase will intensify under higher degrees of global warming, despite the comparable inter-model spread between different warming targets (not shown). Under the SSP2-4.5 scenario, PRCPTOT will increase by approximately 2% and 4% over EA for 1.5 °C and 2.0 °C of global warming, whereas R95p will significantly increase by 12% and 18%, respectively. Projected changes under SSP5-8.5 for PRCPTOT and R95p are 2% (10%) and 4% (20%) under 1.5 °C and 2.0 °C of GWLs. Future changes in PRCPTOT is projected to increase, which leads to a potential higher increase in occurrences of R95p and the possibility of flooding. The increase in SDII is slightly smaller than that of PRCPTOT (figure 3(a)). The absolute increase in R20mm is small under both scenarios (figure 3(b)). Meanwhile, limiting warming target to below 1.5 °C but not 2.0 °C will avoid 37% (44.2%) and 92% (4%) impact occurrence of EPIs of R95p (CDD) under SSP2-4.5 (SSP5-8.5) scenarios, respectively (figure S7). These results indicate that a warming climate drives the increasing tendency of extreme precipitation indices with the exception of CDD, implying an intensification of precipitation processes in a warmer climate.

The findings of the present study are in agreement with the recent studies that employed large ensemble members from large ensembles of RCMs outputs to project changes in extremes precipitation over EA region (Cattani et al 2018, Osima et al 2018, Ogega et al 2020, Dosio et al 2021). The aforementioned studies noted an increase in precipitation extremes over the study domain, which is mainly associated with the alteration in the Hadley circulations and thermodynamic impact linked to the Indian Ocean dipole (IOD) (Hastenrath et al 2011, Endris et al 2016, 2019).

We estimate the projected PE to precipitation extremes based on projections obtained from SSP scenarios (Jones and O'Neill 2016). Figure 4 show the spatial variation distribution of projected changes in population exposure to R95p relative to the baseline period while figures 5 show the responses under two global mean warming levels. Changes in PE to CDD are presented in figures S8 and S9, respectively. With the expected future warming, the PE under different warming scenarios demonstrates the most prominent net intense occurrence over Burundi, Rwanda, and some parts of Uganda (figure 4). In contrast, less change is noted to occur over vast parts of Kenya and Tanzania. Comparable changes are well delineated for different SSP scenarios, even though a substantial increase is expected under SSP2-4.5 scenarios and relatively lower exposure is projected under 1.5 °C as compared to 2.0 °C (figures 4(a), (b), (d), (e)). The exposure to extremely wet precipitation events under the SSP2-4.5 scenario at 1.5 °C warming levels is projected to occur less in most regions, except for Burundi, Rwanda and around east Lake Victoria basin (figures 4(a) and (b)). For example, under the SSP2-4.5 scenario, the PE under 2.0 °C is about 46 × 10⁵ billion person mm, which declines to 40 × 10⁵ billion person mm under SSP5-8.5|SSP5 and 2.0 °C warming levels (figure 5). In comparison, under 1.5 °C warming, exposure is reduced to 26 × 10⁵ billion person mm (23 × 10⁵ billion person mm) under SSP2-4.5 (SSP5-8.5), respectively. In contrast to 2.0 °C warming level, 0.5 °C less warming reveals noteworthy net changes for the exposure to precipitation extremes (figures 4(c) and (f)). The exposure is reduced to 19 × 10⁵ billion person mm (18 × 10⁵ billion person mm) under SSP2-4.5 (SSP5-8.5) scenarios (figure 5). The large population exposed to EPIs under SSP2-4.5 could be attributed to the projected population growth, urbanization, and spatial patterns of development under the SSP2 (middle of the road) scenario. The R95p causes impacts such as water-borne disease outbreaks, stressed sewage networks, landslides, wrecked homes and buildings, damaged crops and affected agricultural production, affected traffic conditions, and most importantly, heavy and deadly flooding. In contrast, the SSP5-8.5, which represents fossil-fueled development, is characterized by a low fertility rate, high income, and sprawl pattern patterns, thereby reducing the impact of PE to EPI. The findings show that 0.5 °C less warming could lead to a 42.1% (43.1%) reduction in population exposed to EPIs under SSP2-4.5 (SSP5-8.5) scenarios.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Risks related to climate change have been detected to significantly increase over recent decades and are projected to worsen under future warmer conditions (IPCC 2018, 2021). Estimating population exposure to extreme precipitation events is key to assessing the risk induced by extreme precipitation and flooding. The evolution of exposure to extreme precipitation with different warming levels signifies the speed at which this hazard affects society. Over the study region, the exposure of the population to climate events remains a significant feature (Niang et al 2014). Recent decades have witnessed an amplification of wet extreme event incidences (Kilavi et al 2018, Tramblay et al 2020, Wainwright et al 2020), mainly as a result of the positive phase of IOD (Cai et al 2018, Endris et al 2019, Ngoma et al 2021b). Such changes will directly impact the population that remains vulnerable due to its low adaptive capacity (Ahmadalipour et al 2019). The projected exposure increase over the study domain has been found in other regions across the globe (Chen et al 2020, Wang et al 2020, Ma and Yuan 2021, Zhao et al 2021), and will likely intensify due to the increase in global warming and the proliferation of urban populations. Lowering the warming by 0.5 °C in the future will have a ripple effect in minimizing the adverse impact of climate change, not only over the study region but likewise at the global level (Rogelj et al 2018, Chen et al 2020, Wang et al 2020). Overall, the exposure to precipitation extreme events is expected to increase substantially under a 2.0 °C warmer future as compared to 1.5 °C. The present study agrees with existing studies that project notable intensification of pluvial occurrences over EA during the near and middle 21st century (Nguvava et al 2019, Haile et al 2020, Tan et al 2020, Ayugi et al 2021a).

Finally, the study appraised the relative importance of climate and population change under SSP2-4.5 and SSP5-8.5 at 1.5 °C and 2.0 °C GWLs (figure 6). Under the two scenarios explored in this study (figures 6(a) and (b)), we found future exposure to be predominantly driven by changes in population rather than climate and interaction factors. Under SSP2-4.5, the projected change due to climate (constant population) is nearly 2.3 (∼5.8) × 10⁵ billion person mm at 1.5 °C (2.0 °C) warming levels. However, population exposure (constant climate) is ∼21 (33.8 × 10⁵ billion person mm) at 1.5 °C (2.0 °C) warming levels (figure 6(a)). In other words, climate effects are anticipated to contribute by ∼10.6% (12.6%) of the total change in population exposure under 1.5 °C (2.0 °C) warming levels, while population and interaction effects are expected to contribute ∼77.4% (71.9%) and 12% (15.5%), respectively, under 1.5 °C (2.0 °C) scenarios. For SSP5-8.5 scenario, the expected changes in exposure resulting from climate effect is ∼3.8 (6.1 × 10⁵ billion person mm), i.e. 15.7% (17.5%) of the total change at 1.5 °C (2.0 °C) warming levels, whereas population account for 16.1 × 10⁵ billion person mm (75%) at 1.5 °C and 26 × 10⁵ billion person mm (66.2%) in 2.0 °C of the total change, respectively (figure 6(b)). Interestingly, the projected changes in regional exposure due to the interaction effects under SSP2-4.5|SSP2 are bigger than the climate effect, while the reverse pattern is observed under SSP5-8.5|SSP5. For instance, under SSP2-4.5, the interaction effect contributes about ∼3.1 × 10⁵ billion person mm (12%) and ∼7.0 × 10⁵ billion person mm (15.5%) of the total change in 1.5 °C and 2.0 °C GWLs (figure 6(a)). Under the SSP5-8.5 scenario, climate effects are higher at ∼3.8 × 105 billion person mm (15.7%) of the total change in 1.5 °C and further rises to ∼6.4 × 105 billion person mm (17.5%) in 2.0 °C (figure 6(b)).

Zoom In Zoom Out Reset image size

The projected population exposure over EA shows similar varying patterns to those observed in other regions or at a continental level. The contributing factors affecting most regions and at the continental level vary from either population effects, GDP, interactive effects, or climate effects (Winsemius et al 2016, Liu et al 2020, Chen et al 2020). For instance, over Europe, the climate effect has a strong influence due to low population growth, while regions of Asia, North/South America, and Oceania have a dominant climate effect as compared to the interaction effect (Liu et al 2020, Chen et al 2020). The present study agrees with recent research that noted the chronology of contributing factors in Africa, with population accounting for >75%, followed by interactive change and climate change having the least contribution (Liu et al 2020). The aforementioned study listed 10 countries with the highest population exposure, accounting for >53% of global exposure. Interestingly, among the listed nations, three countries (including Rwanda, Burundi, and Uganda) are situated in East Africa. Other countries include: Nigeria, the Philippines, Bangladesh, Haiti, the Netherlands, Luxemburg, and Belgium. Similar findings have been observed in other studies that established the impact of climate extremes on society's well-being (Cook et al 2015, Harrington and Otto 2018).

Given the importance of the interaction and population effects, clearly, any policy response designed to reduce population exposure to these extreme events needs to focus at the country level by considering their socioeconomic developments (population growth), and climate mitigation efforts. As a way forward, the high exposure noted over EA calls for a shift in policies with an adaptative measure to be put in place to cushion the effect on the exposed population. Actions such as relocations to high elevation ground, climate smart agriculture, and water harvesting will prove necessary as a first practical step to minimize population and socio-economic losses due to projected increases in precipitation extremes and high population exposure noted (Schlenker et al 2013, Nsubuga and Rautenbach 2017).