We compare two methods of evaluating head covariance for two-dimensional steady-state flow in mildly heterogeneous bounded rectangular aquifers. The quasi-analytical approach, widely used in stochastic subsurface hydrology, is based on the Green s function representation, and involves numerical four-fold integration. We compare this approach with a numerical solution of the two-dimensional boundary-value problem for head covariance. We show that the finite differences integration of this problem is computationally less expensive than numerical four-fold integration of slowly-convergent infinite series.