Model-based processing for towed arrays is based on a Kalman-type estimation scheme. In the narrow-band case the measurement equation is based on either a plane wave or curved wavefront monochromatic signal model. In the broadband case one could model the signal as a set of parallel narrow-band signals. However, this would be prohibitively computationally intensive. Here, it is shown how the signal can be modeled with very few parameters using an autoregressive moving average (ARMA) model for the signal, i.e., the measurement equation in the Kalman filter, where the parameters of interest, the bearing and range, are combined with the ARMA parameters in a single state vector, thus providing an adaptive scheme wherein these parameters are adapatively estimated in an iterative manner. Examples of broadband bearing and range estimation based on simulated data will be shown.