Motivated by a computer experiment for the design of a rocket booster, this article explores nonstationary modeling methodologies that couple stationary Gaussian processes with treed partitioning.