Curious if you ever found a solution to this issue, @sfrav? The raster package says “Any type of model (e.g. glm, gam, randomForest) for which a predict method has been implemented (or can be implemented) can be used”; however, that does not appear to be the case for brms’ predict function. In my case, I fit a simple brm multilevel model with a quadratic effect of elevation and latitude on animal population density, and am trying to predict the density values for existing elevation and latitude values across the entire study area as a rasterized map (e.g., by adapting the lm/glm methods: HKU Movement Eco in R Workshop - The homepage for the Movement Ecology in R Workshop hosted at Hong Kong University, January 3-12, 2018).