Consequences of 1.5 °C and 2 °C global warming levels for temperature and precipitation changes over Central Africa

We define Central Africa as the region located approximately between latitudes 15° S and 15° N, and longitudes 5° E and 35° E. The region's climate is complex and varies spatially. While coastal and central areas experience an equatorial climate (with two rainy seasons), the northern parts undergo a semi-arid climate regime (with a single rainy season). Precipitation is strongly variable in space and time due to its convective nature. The length of the rainy season varies with latitudes. Annual average temperature ranges from 26 °C–30 °C.

This study is based on daily temperature and precipitation output data from the CORDEX-Africa multi-RCMs ensemble of transient simulations described in Nikulin et al (2018, table 2S). The ensemble consists of 25 RCM simulations including 10 simulations from the Swedish agency for weather, climate, hydrology, and oceanography, all used to downscale large-scale driving fields from a set of 13 CMIP5 GCMs (Taylor et al 2012) over a Pan-African numerical domain, using a grid-spacing of 0.44° (approximately 50 km). The control simulations, forced by observed natural and anthropogenic atmospheric composition (CO₂ and non-CO₂ greenhouse gases, aerosols, and land cover), cover the period from 1950–2005, whereas climate projections (2006–2100) are forced with emissions from the Representative Concentration Pathways (Moss et al 2010, van Vuuren et al 2011). AR5 uses four RCPs to describe scenarios for future emissions, concentrations, and land-use, ending with radiative forcing levels of 2.6, 4.5, 6.0, and 8.5 W m⁻² by 2100. RCP 2.6 assumes that strong mitigation policies result in a low forcing level and a peak between 2010–2020 for global annual GHG emissions (measured in CO₂-equivalents), with emissions declining substantially thereafter. Emissions in RCP 4.5 peak around 2040, then decline. In RCP 6.0, emissions peak around 2080, then decline. In RCP 8.5, emissions continue to rise throughout the 21st century. RCPs are not fully integrated scenarios i.e. a complete package of socioeconomic, emissions and climate projections. RCPs are only consistent with components of radiative forcing that meant to serve as input for climate modelling, pattern scaling and atmospheric chemistry modelling.

RCP8.5 combines assumptions about high population and relatively slow income growth with modest rates of technological change and energy intensity improvements, leading in the long term to high energy demand and GHG emissions in absence of climate change policies (Riahi et al 2011). Compared to the total set of RCPs, RCP8.5 receives more attention in the climate modelling community and corresponds to the pathway with the highest greenhouse gas emissions. It is technically possible to limit radiative forcing from RCP8.5 to lower levels comparable to other RCPs (2.6–6 W m⁻²). CMIP5 RCPs were not designed to address GWL concerns, nor to analyze difference between the effects of 1.5 °C and 2 °C of global warming (James et al 2017). Of the existing RCP2.6 GCM simulations, which can be considered as the most appropriate proxy for holding GWL below 2 °C, only ten CORDEX-Africa simulations have been generated by only two RCMs (see for detail Nikulin et al 2018, table 2S). In this study, therefore, we utilise the CORDEX-Africa runs driven by the RCP8.5 scenario, as, first, it comprises the largest ensemble (25 runs) and, second, may be considered as the most realistic business-as-usual scenario given the current trajectory of greenhouse gases emissions.

Zoom In Zoom Out Reset image size

There are several approaches to simulate 1.5 °C and 2 °C global warming above preindustrial levels, with different pathways, as well as range of methods for identifying regional climate signals associated with global temperature limits (James et al 2017). Each of these methods shows strengths and weaknesses. In this paper, we use GCMs' transient climate simulations to define a 1.5 °C or 2 °C warming period. We take 1861–1890 to define the pre-industrial (PI) period as it is available across all CMIP5 historical simulations. Then for each GCM downscaled the timing of the relevant GWL is defined as the first time the 30 year moving average (centre year) of global temperature is above 1.5 °C or 2 °C compared to pre-industrial. The corresponding 30 year period is then extracted from the downscaling RCM for analysis using 1971–2000 as a control period. It is worth highlighting that the computed changes in regional climate do not therefore represent the total change induced by 1.5 °C or 2 °C of global warming, as significant global temperature increase had already been experienced by 1971–2000 Rather, the changes computed here represent the difference between the late 20th century climate and the climate projected in a 1.5 °C or 2 °C world. Further details about the methodology, definition of warming levels, models and experimental design are described in a companion paper (see for detail Nikulin et al 2018, table 1S).

The analysis of climate change in Central Africa focuses on seasonal mean precipitation and temperature, together with drought and flood indices: consecutive dry days (CDD), for quantifying drought risk, and three indices for flood risk including the consecutive wet days (CWD), total wet day precipitation (PRCPTOT) and maximum consecutive five day precipitation (rx5day). We examine the range of modelled responses, and then present the multi-model response. There are also many methodologies used to determine the robustness of a climate change signal (see Collins et al 2013). We consider a climate change signal robust if the following two conditions are fulfilled (i) more than 80% of model simulations agree on the sign of the change, (ii) the signal to noise ratio, i.e. the ratio of the mean to the standard deviation of the ensemble of climate change signals, is equal to or larger than one. The first criterion considers model agreement and the second is a measure of the strength of the climate change signal (with respect to the inter-model variability in that signal). We use both in defining robustness because the first criterion may be fulfilled even in the case of a very small, close to zero change.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We use Taylor diagrams to assess RCMs' performance in simulating rainfall variability at a monthly time scale (Taylor 2001). Figure 1 summarizes the root-mean-square difference (RMSD), pattern correlation (r), and the standard deviation (STD) of seasonal mean rainfall of each models' ensemble mean, with respect to the reference field (GPCC) for SON and MAM rainy seasons in 1971–2000 CRU data (Harris et al 2014) is also compared to GPCC (Schneider et al 2011), to provide a measure of observational uncertainty. The performance of RCMs' control simulations shows a wide spread and varies across seasons. In MAM, figure 1(a) indicates r values between 0.6 (RACMO/EC-EARTH) and 0.9 (ALADIN/CNRM-CM5), RMSD ranging from 0.5 (CRCM5/MPI-ESM-LR) to about 1 (e.g. RCA4/EC-EARTH), and STD from 0.75 (RACMO/EC-EARTH) to 1.5 (RegCM4/MPI-ESM-MR). In SON, models' performances are more clustered and indicate slight improvement relative to MAM.

We start by translating the regional scale response of global warming for the two GWLs, using global and regional climate models. Figure 2 shows projected changes in annual mean temperature and precipitation over Central Africa for 1.5 °C and 2 °C warmer worlds. Since CORDEX simulations are integrated from 1950 through 2100, a direct assessment of projected climate changes relative to the preindustrial period (PI) is only appropriate for the CMIP5 GCMs ensemble simulations referred to as 'GCM r1 PI'. In addition, we assess the regional changes relative to the control period 1971–2000 for CMIP5 (GCMs r1 CTL), for the CMIP5 GCMS used to force the CORDEX RCM ensemble (GCM CORDEX CTL), and the CORDEX RCM ensemble (RCMs CORDEX CTL). As would be expected, projected changes are larger when computed relative to the preindustrial (in GCM r1 PI), than when computed relative to the late 20th century (in GCM r1 CTL). This provides important context for interpreting the CORDEX results: these are calculated relative to the late 20th century, and if regional changes were calculated relative to the preindustrial, they would very likely be larger.

Zoom In Zoom Out Reset image size

Figure 2 also shows a comparison between 1.5 °C (and 2 °C) changes calculated from RCP8.5, RCP 4.5, and RCP2.6. Since each RCP has a different pathway towards 1.5 °C (and 2 °C), with differences in both the rate of warming and the mix of anthropogenic forcings, there could be differences between RCPs in the regional change associated with 1.5 °C and 2 °C. In terms of annual mean temperature and precipitation, figure 2 suggests that the uncertainty ranges are reasonably consistent between RCPs, although this may be different if calculated for seasonal means or extremes (Pendergrass et al 2015, and Mitchell et al 2016). All datasets show warming of Central Africa associated with 1.5 °C and 2 °C worlds (figure 2(a)). When calculated relative to the preindustrial, the regional warming is larger than the global mean in the majority of models (GCM r1 PI). Whilst there is variation between models in the magnitude of the warming, the range of responses is markedly warmer at 2 °C than 1.5 °C. With respect to the period 1971–2000 the range of mean regional warming of CORDEX RCM CTL resembles that of GCM r1 PI and GCMs r1 CTL, although the number of simulation varies across the different ensembles. However, the interquartile range of CORDEX RCM CTL projected warming is narrower than that of the driving GCM across RCPs, except for RCP 2.6. This illustrates that the uncertainty in regional change is different depending on which dataset is employed, this should be considered when assessing projected responses from only one dataset.

This need for careful assessment of uncertainty is perhaps even more relevant for precipitation. Figure 2(b) shows projected changes in annual mean precipitation for the 1.5 °C and 2 °C GWLs, revealing uncertainty in the sign as well as the magnitude of precipitation change. Relative to 1971–2000, GCMs r1 CTL and GCMs CORDEX CTL responses remain similar across RCPs and indicate a clear increase in precipitation changes. In contrast, CORDEX RCM CTL projections tend to be more uncertain showing equally positive and negative changes. Differences in the mean response between 1.5 °C and 2 °C appear to be small, however, the uncertainty ranges at 2 °C are larger, with some of the models projecting changes in precipitation of a magnitude not seen at 1.5 °C. This could indicate an increase in risk associated with 2 °C.

Now the focus is placed on the spatial distribution of the multi-model climate change response from the CORDEX RCMs in the SON and MAM seasons, which correspond to the peaks in climatological precipitation. Several studies indicate that SON is wetter than MAM in most observational datasets (Dezfuli 2011, Washington et al 2013, Dezfuli and Nicholson 2013, Badr et al 2016).

Zoom In Zoom Out Reset image size

The mean MAM temperature for the control period (1971–2000) as simulated by RCMs exhibits a meridional gradient with high values in the northern part (greater than 25 °C) and low values in the southern flank (figure 3). Projected changes in temperature show differences in warming rates between 2 °C and 1.5 °C GWL. The 0.5 °C increase in global temperature (from 1.5 °C–2 °C) is associated with larger increases in regional temperature, on average 0.6 °C for Central Africa, but even larger for parts of central DRC and northern Central Africa (up to 0.7 °C). It is also worth noting that this is an average across the models and some models show larger change. Temperature change in SON exhibits a reduced intrusion of warming along coastal areas in the west (figure 3). Higher warming is projected in the northern and southern parts of the study area with temperature increase greater than 1 °C between the 1.5 °C and 2 °C levels, whereas future changes of temperature will be less than 0.5 °C over Atlantic ocean (relative to the late 20th century).

There is a decrease in precipitation over much of inland Central Africa during MAM for the 1.5 °C GWL (figure 3), while coastal areas experience a slight increase in precipitation. Precipitation increase over the equatorial coastal region is projected to strengthen under 2 °C GWL. However, there is a slightly decrease in precipitation over southern coastal region from 1.5 °C–2 °C level of warming. Broadly, reinforcement of dry condition inland is expected under 2 °C level of warming. However, there is little agreement between regional models in these multi-model mean responses: this season is characterized by high uncertainty in projected rainfall change.

Zoom In Zoom Out Reset image size

The pattern of projected rainfall changes in SON is associated with a marked dipole distribution (figure 4). Northern Central Africa is expected to moisten while southern regions are projected to dry. Broadly, this pattern is reinforced from 1.5 °C–2 °C GWLs, although over eastern Central Africa the decrease in precipitation is less pronounced under 2 °C GWL. The increase of precipitation is larger in coastal areas in the two rainy seasons (figures 3 and 4). Model consensus is large for the 2 °C global warming level over southern Central Africa, with some areas indicating more than 80% of agreement across models with respect to the sign of the change. Model results show consensus in projected rainfall decrease only over a small region, encompassing parts of Angola and Zambia. This supports previous results from Dosio and Panitz (2016).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

This section focuses on the change of rainfall extremes under 1.5 °C and 2 °C global warming levels. We use a subset of widely used rainfall indices coordinated internationally, which allows the results of studies from different parts of the world to fit together seamlessly. These include the maximum length of dry spell (CDD) and maximum length of wet spell (CWD).

The situation is more contrasted for SON (figure 6), with clear delineation of CDD change in the south and north parts. Broadly, there is a clear increase in CDD over southern Central Africa in SON, with good agreement between models (figure 6). CDD is projected to increase faster in a 2 °C warmer world than for the 1.5 °C global warming level in the southern part of Central Africa in SON. CDD change tend be smaller in the central parts of the Congo basin.

Under either the 1.5 °C or 2 °C global warming scenario, there is significant decrease of CWD relative to today's climate despite localised hotspots of increase depicted in SON (figure 6). There are strong local changes. The area influenced by changes grows somewhat with increasing warming level. The decrease in CWD would be weakened by more than 2 days under the 2 °C threshold compared to the 1.5° level of global warming over most of the Central African region, particularly over Congo basin in both MAM and SON seasons.

Figure 7 shows the spatial distribution of precipitation indices (PRCPTOT and rx5day) and their projected changes during MAM. There is weakening of PRCPTOT is prominent over the continent, with a reduction lesser than 50 mm; in contrast, PRCPTOT tends to intensify in oceanic region exceeding 40 mm and have more consistent geographical patterns of the change precipitation. Differences between the warming levels are easy to discern, PRCPTOT at 2 °C warming is clearly higher and changes more robust than at 1.5 °C in most of continental Central Africa. For the two levels of warming, the changes are strongest in SON (figure 8). At 2 °C warming, southern part of the study area is subject to robust changes this season. Examination of rx5day reveals a contrast in pattern of change for 1.5 °C and 2 °C global warming levels (figure 8). A clear increasing pattern emerges over much Central Africa with more than 15 mm, with some localised hotspots in northern region showing decrease (figure 8). This increase in rx5day is an important element with respect to flood hazards, thus the increase in rx5day obtained under these global warning levels is likely to amplify the probability for these risks. Difference in change of rx5day between 2 °C and 1.5 °C GWLs is uncertain and shows a complex spatial structure. Additionally, models have better agreement in the sign of projected changes in PRCPTOT than in rx5day during SON in the south.