In this paper we discuss a low-fkquency computational pmcedurc based on the Finite ~~f f e r c n~w~~i r n e Domain (FDTD) algorilhm, for numeric$ modeling of electmmngnetic scattering by buried objects in sediment layers under s w water. The FDTD algorithm is found to be accurate for modeling buried objects in sediment layers tens of meters away from a constant current dipole source of IA in sea water. For validation of the lowfrequency FDTD modeling, the computed FDTD results are compared with those calculated by using analytic expressions and integra1,equation techniques. In this paper we also present a technique for detecting conductivity anomalies in sediments, e.g.. a buried object in sedimentary layers under sea water, by using the neural network approach.The electric field values are used as the inputs to the neural network and the associated conductivities are used as the targets. The neural network is then trained to arsociate these conductivities and field values. It is shown in this paper that a trained neural network can be used to estimate the conductivity of new objects that have not k n employed to train the network.