I wrote a new paper on predicting conflict zones in civil wars. In a nutshell, the basic assumption here is that the determinants of violence in civil wars are often local (as the literature tells us). I use conflict events from the UCDP GED dataset and geographic covariates of violence to fit Point Process Models to predict the spatial distribution of violence in ten civil conflicts. In a cross-validation test, the models fitted to nine countries predict the remaining country rather well. Check out the extrapolations for Africa and the Greater Middle East if you have Google Earth installed here or take a quick glance at this poster for the general idea.