Research overview: model transferability and novel climates in felid projections

Projecting a species into the future requires transferring a model into climate conditions that may not exist anywhere today, raising the risk of unreliable extrapolation. This overview surveys the problem of novel and non-analog climates, methods to detect when models are extrapolating beyond their training data, and the resulting uncertainty in felid range projections. It reviews tools such as multivariate environmental similarity surfaces that flag where predictions should be trusted least. The synthesis frames transferability as a core limitation of climate-change distribution modeling, explaining why projected habitat maps for cats decades ahead must be read with attention to where the model is interpolating from data versus guessing into unfamiliar conditions.