Data-driven methods for approximating complex systems with stochastic models. Techniques are fairly general, but motivated by fluid mechanics.