“…In contrast to the use of the numerical dispersion error, the approach in [21,22,19] was based on the minimization of the order of the local truncation error and did not require the existence of exact and numerical harmonic solutions. However, the use of uniform meshes in [21,22,19] reduces the application of the technique to rectangular domains and significantly restricts the value of the proposed approach. Moreover, in our paper [19] it was shown that the direct application of some techniques with improved accuracy on uniform meshes (e.g., the MIR techniques in [12,51,24]) to nonuniform meshes leads to a significant degradation in the order of accuracy (e.g., by three orders for the wave equation).…”