In this talk, I discuss techniques for estimating static parameters in chaotic dynamical models. Such problems arise, for instance, when sub-grid scale phenomena are included in climate and weather models using different parameterization schemes. In chaotic systems, classical parameter estimation approaches where model simulations are directly compared to observations are often infeasible, since errors in the initial conditions can lead to large, unpredictable deviations from the observations. One way forward is to compare summary statistics of model simulations and observations. Alternatively, one can explicitly deal with uncertainties in the initial conditions by formulating the system as a state space model. Here, I review our experiences with these techniques, and illustrate the approaches with various toy models and full scale climate and weather models.