Summary
We build a model of discretization errors, known as directional aliasing, to theoretically evaluate how biases in the microtremor spatial autocorrelation (SPAC) coefficient, or the real part of the SPAC spectrum of microtremor analysis, are related to the magnitudes of the imaginary part when a seismic array of only two sensors is used. By using this model, we investigate the potential utility of the imaginary spectrum component as an indicator of applicability of the two-sensor SPAC method to the field of microtremors generated at an observation site. Field data of microtremors from compact seismic arrays (1–15 m) are used to test the model. It is found that, when the imaginary components are very large in magnitude (where the threshold depends on the rk, the array radius times the wavenumber), the field of microtremors is dominated by waves arriving from a single direction parallel to the array axis and the SPAC coefficients tend to be underestimated in small rk ranges (i.e. rk < 3.8; the range considered throughout this study). In the present study, which is based on the observations of 400 microtremor arrays, the underestimates seldom exceeded 30 per cent. The SPAC coefficient estimates could be corrected in that case by using information on the imaginary part. When the imaginary components are very modest in magnitude, by contrast, there are two possible scenarios: either (i) the waves are arriving predominantly from a single direction perpendicular to the array axis and the SPAC coefficients are wildly overestimated (i.e. there was a small percentage of low-quality data, with relative errors exceeding +50 per cent, based on the observed data analyses), or (ii) the wavefield is close to isotropic and the SPAC coefficients are unbiased (i.e. 70–90 per cent of all observed data fell within the relative error range of ±20 per cent). It is difficult in that case to have SPAC coefficient estimates corrected by using information on the imaginary part alone.