Data & Methods
Data
I selected 23 General Circlulation Models (GCMs) from CMIP6 from ten countries and one continentally from Europe (Table 1) with criteria that the models had monthly average minimum temperature (°C), monthly average maximum temperature (°C) and monthly total precipitation (mm) as well as complete scenarios for SSP2-4.5. The scenario “middle of the road” is characterized by medium challenges to mitigation and adaptation with global warming capped to 4.5°C (GLISA 2021). I considered the normal period 1970-2000 for the historical climate data and future climate data for the period 2041-2060. All the data were obtained from WorldClim (Fick and Hijmans 2017) a high resolution climate database. A spatial resolution of 10 minutes was selected for computational purposes.
Table 1: Selected GCMS in the CMIP6 with their institution and countries of origin. Source: IPCC, 2021: Annex II: Models [Gutiérrez, J M., A.-M. Tréguier (eds.)]. In Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change
The global data was cropped to the spatial grid range of Africa; 17°W to 51°E and 37°N to 35°S (Figure 2). I used IPCC’s regional classifications for Africa resulting in ten regions; Mediterranean (MED), Sahara (SAH), Western-Africa(WAF), Central-Africa(CAF), N.Eastern-Africa (NEAF), S.Eastern-Africa (SEAF), W.Southern-Africa (WSAF), E.Southern Africa (ESAF), Madagascar (MDG) and the Arabian-Peninsula (ARP) (Iturbide, Gutiérrez et al. 2020).
Methods
Bioclimatic variables and anomalies
I used the monthly average minimum temperature (tmin), monthly average maximum temperature (tmax) and monthly total precipitation (prec) for the future climate and normal climate to calculate their bioclimatic variables. I used the dismo package in R (Hijmans, Phillips et al. 2022) with the syntax: biovars(prec, tmin, tmax)
I calculated bioclimatic variables’ anomalies between future and normal climate by using differences between values for normal climate and the future climate. The temperature variables’ anomalies (MAT1, MDR2, ISO3, TSeas4, MTWarmM5, MTCold6, TAR7, MTWetQ8, MTDryQ9, MTWarmQ10 and MTColdQ11) were calculated in degrees celcius (°C) and the precipitation variables (MAP12, PWetM13, PDryM13, PSeas15, PWetQ16, PDryQ17, PWarmQ18 and PColdQ19) were expressed as a percentage change (%).
Table 2: A sample datable showing the aggregated regional model averages of each of the 19 bioclimatic variables. The full table contains 23 GCMs with 10 rows each for the Africa regions.
Cluster analysis and Principal Component Analysis
To visualize the similarity among models, I performed a cluster analysis on the regional anomaly values of Annual Mean Temperature (MAT1) and Mean Annual Precipitation (MAP12) using the pvclust package and function (Suzuki, Terada et al. 2019). I used Ward.D2 hierarchical clustering algorithm with a Euclidean distance of scaled MAT1 and MAP12 values. To visualize the model estimates of MAT1 and MAP12 and correlation with GCMs, I used principal component analysis.
Model Selection for "best", "middle" or "worst" scenario
Using regional model averages, I color condition the data to visualize magnitudes of change. To visualize the interaction of increasing MAT and change in MAP (increase or decrease) I obtain a difference of their standardized value. This is also coded to show the "dry and hot" or "wet and cool".