This paper is concerned with adaptive Doppler processing in high frequency (HF) over-the-horizon (OTH) radar to improve the detection of slow-moving ships against clutter. A knowledge-based adaptive detection approach that exploits additional information to identify the surfaces that scatter clutter in each radar resolution cell (i.e. land, sea or land-plus-sea) is proposed to guide the selection of training data for the adaptive algorithm. The practical performance of the KB scheme is compared with conventional FFT-based Doppler processing and a traditional adaptive algorithm that does not exploit additional information.