Supposedly people can estimate the smoothness for the Matern covariance. We should try to get that to work.