By using the method of prediction analysis, the accuracy of source inversion of strong-motion data is investigated for various arrays, and the effects of array configuration are evaluated. In the case of a circular array in which Ns array stations are uniformly distributed within radius R and when the strike-slip fault is located at the center of the array, an empirical relationship between the array parameters and the standard deviation of the seismic moment of each subfault, normalized by the seismic moment, is obtained as
where c is a constant, and Ne is the number of subfaults. The constant c depends upon the fault mechanism (dip slip or strike slip), aspect ratio of the fault, focal depth, the array configuration, rupture mode (unilateral or bilateral), crustal structure, etc. However, it would be useful for both the source inversion study and a future installation plan of strong-motion arrays.
An application of the method to several specific cases leads us to conclude that highly dense local array is not advantagenous for source inversion, but an array covering the earthquake fault with equally spaced stations is of benefit to the source inversion.
Two cases of array configuration examined are a straight-line array, which is perpendicular to the fault and another straight-line array, which is parallel to the fault. The perpendicular array is unsuitable for source inversion, but the parallel array is advantageous for source inversion. Furthermore, the inversion study for the 1979 Imperial Valley earthquake is simulated for three cases of strong-motion station distributions (case A: 20 United States stations; case B: El Centro strong-motion array; and case C: all 26 stations of the United States and Mexico). Our simulation shows that case B gives a very poor result, and that the best of three is case C.