The autocovariance function (ACF) is central to analyzing time series, spatial, and spatio-temporal data, but classical parametric models are often too restrictive. Bochner's theorem guarantees that any positive spectral measure yields a valid ACF via the inverse Fourier transform, yet existing non-parametric spectral approaches rarely return closed-form expressions for the ACF itself. This paper addresses that gap by developing a closed-form class of non-parametric ACFs.
The method derives the inverse Fourier transform of B-spline spectral bases with arbitrary degree and knot placement. This yields a family of closed-form stationary covariance models with three key properties: it is uniformly dense among ACFs of weakly stationary processes that possess a spectral density; it accommodates univariate, multivariate, and multidimensional processes; and it is non-separable in general, with separability as a special case.
Jackson-type approximation bounds establish convergence rates, and simulation studies show accurate recovery of known univariate and bivariate ACFs. Applied to output from a high-resolution ocean simulation, the fitted space-time ACF is non-separable and demonstrates accurate process recovery.