Because of the nonuniform distribution of earthquake epicenters, a uniform distribution of seismological stations is not necessarily optimal for sampling the Earth in order to determine global Earth structure. Taking into account the already existing digital long-period seismological stations (GDSN, IDA, and GEOSCOPE), we propose a strategy for the siting of additional stations. Here, we shall develop a criterion designed to optimize global resolution of upper mantle structre using mantle wave observations; other kinds of seismological studies will lead to different criteria, which should, however, be amenable to a similar treatment.
For an Earth model developed in spherical harmonics up to a high degree M, and for a given distribution of epicenters and stations, the variance of the solution expanded to some lower degree L ≦ M is obtained at each point of the earth surface by use of averaging kernels, the width of which depends on L. Two measures of the efficiency of an additional station are the reduction of the mean variance and the reduction of the variance oscillations over the Earth's surface; these measures of “new station efficiency” are mapped as a function of the location of the new station (Figures 8 and 9) for the averaging kernels corresponding to L = 8.
Considering the reduction of variance oscillations as the main aim, we find that a new station added in south east Africa, in the eastern Pacific, or in Pakistan is up to 6 times more efficient than a station added randomly, and many times more effective than a station added in the Central United States, the Philippines, or Japan.