Jannis Hoch , Sophie Pieternel de Bruin, Halvard Buhaug , Nina von Uexkull , Rens van Beek, Niko Wanders 

In the past decade, several efforts have been made to project armed conflict risk into the future. One arising technique is the use of machine-learning (ML) models. In this study we explore its opportunities to project sub-national armed conflict risk for three shared socio-economic pathway (SSP) scenarios and three Representative Concentration Pathways (RCPs) by 2040-2050 in Africa, using the novel and open-source ML framework CoPro. Results are consistent with the underlying socio-economic storylines of the SSPs, and the resulting out-of-sample armed conflict projections obtained with RandomForest classifiers agree with comparable studies. In SSP1-RCP2.6, conflict risk is low or absent in most regions, although the Horn of Africa and parts of Kenya, Tanzania and Mozambique continue to be conflict-prone. Conflict risk intensifies in the more severe SSP3-RCP6.0 scenario, especially in central Africa and large parts of western Africa. We specifically assessed the role of hydro-climatic indicators as drivers of armed conflict. Overall, their importance is limited but can differ locally depending on the overall sign of climate change impact and the contextual (socio-economic) factors defining the overall magnitude of those impacts. With our study being at the forefront of ML applications for conflict risk projections, we have identified various challenges for this arising scientific field. A major concern is the inconsistent data availability of observed conflict events as well as of socio-economic indicators for the various SSPs. Nevertheless, ML models such as the one presented here are a viable way forward in the field of armed conflict risk projections, and can help to inform the policy-making process with respect to climate security.