“…This effect is countered by employing a device given in [3] for sharpening the error estimate. It is, of course, possible to use two completely independent cubature rules (e.g., Haegemans [1] uses conic product Gauss formulas (see Stroud [8], Section 2.5) with 36 and 49 points, respectively), but it seems sensible to extend Kronrod's [2] idea, that Q2 should reuse the points of Q1, to the two-dimensional case.…”