High-resolution model data are used to estimate the statistically typical mixing ratio variabilities of trace species as a function of distance and time separation. These estimates can be used to explain the fact that some of the differences between observations made with different observing systems are due to the less-than-perfect co-location of the measurements. The variability function is approximated by a two-parameter regression function, and lookup tables of the natural variability values as a function of distance separation and time separation are provided. In addition, a reparametrization of the variability values as a function of latitudinal gradients is proposed, and the seasonal independence of the linear approximation of such a function is demonstrated.