We describe a data-driven method to estimate the top-formation resistivity using marine controlled-source electromagnetic data. The procedure exploits the fact that the airwave contributions are traveling up or down with a Poynting vector close to the vertical axis. The proposed method is based on forming an impedance in the same way as it is done in magnetotelluric processing. The top-formation resistivity is used to perform up-down decomposition of the electric field below the seabed. This procedure suppresses the contribution from the airwave in the resulting upgoing electric field and increases the data sensitivity to the subsurface resistivity distribution. The upgoing electric field is used in the misfit kernel for an inversion scheme. This is a method that is intended for use in shallow water. Inversion using the upgoing electric field means that the observed electric and magnetic fields contribute to the data misfit. The optimization procedure seeks to reduce the misfit between the observed and predicted data for both fields. We compared this procedure with the conventional procedure of predicting electric fields only. The addition of magnetic data improved the resolution for the synthetic and the real data examples. We performed postinversion modeling on the final resistivity models. The weighted misfit of the electric data was added to the weighted misfit of the magnetic data to form the total error. We found that this total error was smallest for the case in which we used the upgoing electric field in the misfit kernel.