We have estimated the 3-D distribution of crack density parameter (ε) and saturation rate (ξ) in the 1995 Kobe earthquake region by applying a crack theory to values of seismic velocity and Poisson's ratio obtained from seismic tomography. Our results show that ε and ξ exhibit considerable heterogeneity. Higher ε and ξ exist in the Kobe fault zone, in particular, at the mainshock hypocenter. High ξ areas are generally consistent with high ε areas, but high ξ areas have a much wider distribution, indicating that microcracks exist in more localized areas, while fluids are permeating through wider areas in the crust. A significant discrepancy between ε and ξ exists beneath Osaka Bay where ε is low while ξ is high. ξ is generally high beneath seas but low beneath land areas. Osaka Bay and the present sea/land distribution in southwest Japan have existed for 2 millions years. This period is long enough for the sea water to permeate down to the deep crust through the active faults that would have been ruptured during many earthquakes cycles.