Extended Data Fig. 2: Individual expert judgments about the relationship between climate and conflict risk.

Raw numbers of the subjective probabilistic estimates of each expert that are documented in the elicitation. For each expert, the first 6 rows correspond to the 6 causal factors that the expert ranked as most influencing conflict risk to date, drawing from a list of 16 factors collectively generated by the full expert group. The last 3 rows correspond to past examples of organized armed conflict overall (labelled as ‘past’) and to conflict risk under approximately 2 °C and approximately 4 °C warming scenarios (labelled as 2 °C and 4 °C). Numbers within each row are estimated probabilities. For each causal factor (the first six rows), the probabilities reflect judgments of how frequently climate variability and change have led to substantial, moderate or negligible changes in conflict risk for violent conflicts to date involving that factor (probabilities ordered as: substantial decrease, moderate decrease, negligible change, moderate increase and substantial increase). For total risk of violent conflict to date (past), the probabilities reflect judgments across past examples of conflict overall. For the approximately 2 °C and approximately 4 °C warming scenarios, specified probabilities reflect judgments of potential changes in conflict risk compared to the current climate; these hypothetical scenarios consider effects for current societies, assuming current levels of (for example) socioeconomic development, population and government capacity. Shading categories visualize patterns. CN, conflict in neighbouring areas; CR, corruption; ES, economic shocks; EI, external intervention; ID, illiberal democracy; II, intergroup inequality; MG, mistrust of government; NR, natural resource dependency; PG, physical geography; PP, population pressure; PS, political shocks; RH, recent history of violent conflict; SC, low state capability; SD, low socioeconomic development; VC, climate variability and/or change; VI, vertical income inequality. Confidence levels³⁷ are indicated in the rightmost column: vl, very low; l, low; m, medium; h, high; vh, very high.