Determination of impedance and attenuation coefficients for upper sediment layers from chirp sonar data is performed by using simulated annealing algorithms. Normal incidence reflection seismograms are calculated for layered marine sediments using the Biot theory of acoustic wave propagation in porous media to produce synthetic data. A time-domain inversion algorithm progressively identifies each reflection and applies random perturbations to corresponding model parameters to minimize the quadratic deviation between the data and synthetic seismograms. Numerical simulations indicate that such an inversion scheme can be efficiently used to survey large areas for characterization of sediment properties and their spatial distributions. In addition, impedance and attenuation estimates are used to predict the sediment porosity, density, permeability, and sound speed using the Biot theory. Finally, chirp sonar data (3 to 7 kHz) collected on the New Jersey Shelf as a part of SWARM-95 (Shallow Water Acoustic Random Media) experiment are analyzed. Inversion results confirm previous coring and in situ observations that alternating sand and clay layers are present at the Hudson Apron site.