Optimum strong-motion array geometry for source inversion-II