African continental free trade area (AfCFTA): projected economic impact assessment under future warming in CMIP6

The climate projection data used in this study are the Climate Model Intercomparison Project (CMIP) Phase 6 (CMIP6) for the GFDL-ESM4 model from National Oceanic and Atmospheric Administration (NOAA) Geophysical Fluid Dynamics Laboratory (GFDL). These data were downloaded from the Potsdam Institute for Climate Impact Research (PIK) cloud database and can also be found online at Coupled Model Intercomparison Project Phase 6. The GFDL-ESM4 datasets run from 2015 to 2100 and are taken at a resolution of about 1^∘ with 49 levels of comprehensive, interactive chemistry and aerosols from precursor emissions Earth System (ESM4). Three different CMIP6(GFDL-ESM4) datasets were analyzed for three different Shared Socioeconomic Pathways (SSPs) (ssp126, ssp370, and ssp585). The SSPs are different greenhouse gas emission scenarios that have been integrated into the CMIP6 climate models. They help to better understand how current levels of greenhouse gas emissions may affect future warming. The ssp126 is the emissions scenario with climate policy. ssp370 and ssp585 are emissions scenarios without climate policy to achieve climate targets.

The historical data are from the Climate Forecast System Reanalysis (CFSR) from the National Centers for Environmental Prediction (NCEP). This dataset ranges from 1979 to 2014 and can be accessed online at Global Weather Data for SWAT for any location on Earth with the coordinates provided. The data includes daily maximum and minimum temperatures. Because the CMIP6 data use daily mean surface temperatures, it was difficult to reconcile the daily mean maximum and minimum temperatures for high latitude countries with the mean daily temperatures from the CMIP6 data. This is because these high latitude regions have large seasonal temperature differences compared to low latitude countries. To address this issue, the maximum data from SWAT are considered and rescaled to fall below current temperatures. Therefore, the historical data used here are for reference only and have not been used to model projected GRP growth rates per capita.

The gross national product and national population data used to create figure 3 were obtained from the World Bank [43, 44]. The CMIP6 data (GFDL-ESM4) are in netCDF4 format, while the CFSR/NCEP data can be downloaded as a CSV file. For each dataset, surface temperature data (measured at 2 m above ground level) are extracted over some of the major economic centers in Africa (Lagos, Nairobi, Cape Town and Algiers) with specific longitude, latitude and altitude data. These cities were selected based on their significant contribution to their country's GDP and regional economy. The data is then fitted to mathematical models to conduct trend studies and economic impact.

The method of moving averages has widely been used by scientists in all fields to analyze time-series data [45]. A moving average is commonly used with time series data to smooth out short-term fluctuations and highlight longer-term trends or cycles. Mathematically, let us consider a sequence of surface temperature data points $\lbrace T_i\rbrace_{i = 1}^{N}$, an n-moving average is a new sequence $\lbrace \tilde{T}_i \rbrace_{i = 1}^{N - n + 1}$ defined from $\lbrace T_i\rbrace$ by taking the arithmetic mean of subsequences of n terms [45],

$$\begin{align} \tilde{T}_i = \dfrac{1}{n} \sum_{j = 1}^{i + n - 1} T_j. \end{align} \tag{ 1 }$$

To capture trends in the data sets, the CFSR/NCEP data and the CMIP6 data are aggregated to get an idea of the historical and future trends in surface temperature taken at 2 m above ground level.

Many regression models have been applied in the literature [33, 38–40, 46] to analyze the impacts of temperature variability on economic growth. In this work, to analyze long term economic impacts of elevated annual mean temperatures, a simplified form of the log-polynomial regression model is considered, which is similar to the one used in the paper by Burke et al [39] and discussed in Newell et al [47]. The model is given by,

$$\begin{equation} \log Y_{i,t} = f(T_{i,t}) + \alpha_i + \epsilon_i \end{equation} \tag{ 2 }$$

where $f(T_{i,t})$ is a 2nd degree polynomial function in the mean surface temperature , $Y_{i,t}$ are the GRP per-capita growth rates (in percent point), α_i and _i are some constant and error terms respectively. They are usually referred to as the regional dummy variables which account for institutional and cultural differences between regions [38]. The full model is given by,

$$\begin{equation} Y_{i,t} = \exp(a T_{i,t} + b T_{i,t}^2 + \alpha_i + \epsilon_i), \end{equation} \tag{ 3 }$$

where a and b are the coefficients of the linear and quadratic terms respectively, i stands for the data location and t for the time (day, year).

The impact of different emission pathways on future warming in the CMIP6 data is examined by calculating the number of days when daily maximum temperatures exceed 25 ^∘C. The years 2020, 2050 and 2100 are considered. Figure 1: The top row refers to the ssp126 emissions scenario, the middle row refers to the ssp370 scenario, and the bottom row refers to the ssp585 scenario. This figure shows that in the equatorial region of sub-Saharan Africa, the number of days when daily maximum temperatures exceed 25 ^∘C is much higher compared to high latitude regions, and this number is expected to increase further under emission scenarios ssp370 and ssp585. Therefore, this region is most vulnerable to climate change and is likely to suffer economic losses in the future compared to high latitude regions.

Zoom In Zoom Out Reset image size

Four megacities in different sub-climate regions (equatorial and semi-equatorial (Mediterranean and temperate)) were selected to study the trend of daily mean surface temperatures. Daily mean surface temperatures are then extracted from gridded CMIP6 data (GFDL-ESM4) over these cities with different latitudes and altitudes to investigate the effects of latitude and altitude differences on daily mean surface temperature trends. African countries generally have some form of quasi-centralised economies, with most coastal megacities contributing a significant share to the country's GDP. Based on this argument, Lagos, Nigeria, West Africa, a powerhouse of the Nigerian economy and the Economic Community of West African States (ECOWAS) with longitude = 3.406894, latitude = 6.446522 and an elevation of 26 m above sea level was selected. Other major cities in ECOWAS such as Accra and Abidjan have very similar mean surface temperatures. Similarly, Nairobi, Kenya, another major economic centre in East Africa, has longitude = 36.811673, latitude = −1.284469 and an elevation of 1668.0 m above sea level. The two megacities of Algiers (longitude = 3.070831, latitude = 36.691327 and elevation = 49 m above sea level) and Cape Town (longitude = 18.492373, latitude = −33.907693 and elevation = 16 m above sea level) in the extreme north and south of the continent with a completely different climate from the equatorial regions were also selected. Figure 2 shows the geographical location of the cities studied.

Zoom In Zoom Out Reset image size

To analyze trends in surface temperature data over Lagos, Nairobi, Cape Town and Algiers, their annual rolling mean surface temperatures in the aggregated historical CFSR/NCEP and CMIP6 (GFDL-ESM4) climate projection data over the selected cities are calculated from equation (1) and presented in figure 2. These regions have different threshold temperatures or summer room temperatures. The rates of increase in their mean surface temperatures relative to summer room temperatures are compared. Figure 2 shows a steep increase in annual moving average surface temperature for low-latitude, low-elevation equatorial regions relative to high-latitude, high-elevation regions. The blue dashed lines represent the average summer room temperature for the region (300 K for Lagos, 291 K for Nairobi, 290 K for Cape Town, and 291 K for Algiers), while the black dashed lines separate the historical CFSR/NCEP data from the CMIP6 projection data. It is also noteworthy that the projected increase in surface temperature in Nairobi (much closer to the equator but at a much higher altitude) shows a steady increase, while the temperature in Lagos (at sea level) shows an almost exponential increase.

The effects of daily temperature variation on regional growth rates have been analyzed for all regions of the world and published online by Maximilian et al [38]. Here the effects of an additional degree of daily temperature variability on regional growth rates for Africa is presented in figure 3. From this figure, the reduction in regional growth rates per additional degree of daily temperature variability is larger in equatorial countries with smaller seasonal temperature differences than in high latitude regions with larger seasonal temperature differences. The size of the scatter plot indicates the country's population, using 2017 World Bank values, and the color indicates the country's average seasonal temperature difference. The list of African countries and their alpha-3 codes used in this figure (figure 3) can be found in supplementary table 1 (available online at stacks.iop.org/ERL/16/094046/mmedia). The 2020 GDP data from International Monetary Fund (IMF) (supplementary figure 1) show that most of the African countries with the highest GDP are located in the equatorial zone of Sub-Saharan Africa, which forms the backbone of the African continental free trade area. Rising annual mean temperatures and daily temperature fluctuations can significantly affect the AfCFTA economy when countries such as Nigeria, Kenya, Ghana, Ivory Coast and the Democratic Republic of Congo are severely affected. The daily mean surface temperature data for 2015 over the selected cities with different latitudes and altitudes are extracted and studied. A comparative study of the daily temperature variability and seasonal temperature differences across the different locations is equally done. Figure 4 shows that the daily temperature variability and seasonal temperature difference are higher in high latitude regions than in low latitude regions. We also note that the daily temperature variability is higher in Nairobi, which is at a higher altitude, than in Lagos at sea level.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To assess the likely economic impact of this increase in mean surface temperature, daily mean surface temperatures from 2015 to 2100 over Lagos, Nairobi, Cape Town and Algiers are fitted to a log-polynomial regression model shown in equation (3). A comparative study is carried out to examine the contribution of latitude and altitude on the growth rates of GRP per capita. In figure 5 and supplementary figures 2–4, the daily and annual mean temperatures are fitted to the log-polynomial regression model with regression parameters from Burke et al [39] for the different emission scenarios. Figure 5 (top row) shows log(GRP) growth rates per capita plotted against annual mean temperatures. The middle row shows the density plots of observed data and the bottom row shows the density plots of log(GRP per capita growth rates). This figure shows that regions at higher latitudes and elevations with greater daily temperature variability and seasonal temperature differences exhibit a nonlinear relationship between log(GRP) per capita growth rates and temperature, with an increase in economic growth followed by a decrease at temperatures above 290 K, while regions at low latitudes and elevations exhibit a quasi-linear relationship between log(GRP) per capita growth rates and temperature, with a large decrease in economic growth. CMIP6 data were used under the ssp370 emission forcing. The percentage point change in log(GRP) and the change in marginal growth rates of these cities compared to the baseline (ssp126) are shown in figures 6(a)–(e). Here, Lagos shows a much larger (more negative) percentage change in log(GRP), followed by Kenya, while Algiers shows no significant change with increasing annual mean temperatures. While figure 6(e) shows that Lagos at low latitude and altitude will experience a very small (negative) marginal change in growth rate compared to Nairobi, Algiers and Cape Town at higher latitude and altitude under future warming. The calculated values of median change in log(GRP) are given in table 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The results indicate that the economic impact of the increase in regional average temperature will be felt most by countries in the equatorial region of sub-Saharan Africa, with countries such as Nigeria, Kenya, Ghana, Ivory Coast, Cameroon, the Democratic Republic of Congo and Somalia being the most vulnerable compared to high latitude countries such as Algeria, South Africa and other high latitude countries such as Morocco, Egypt, Tunisia and Botswana. Supplementary figure 1 shows that these low latitude countries form the foundation of the AfCFTA when their total GDP is added together, implying that the future increase in global mean surface temperature will significantly affect the AfCFTA economy. The AfCFTA agreement initially requires members to eliminate tariffs on 90% of goods. The United Nations Economic Commission for Africa estimates that this agreement will increase intra-African trade by $52\%$ by 2022. This means that effective regional trade policies can help mitigate future declines in regional economic growth due to climate change.