Spectral occupancy measurements and subsequent data modelling have been the subject of increasing attention since the early 1980s. With the congested nature of the HF spectrum, where users with differing requirements compete for diminishing available bandwidth within the International Telecommunications Union (ITU) frequency allocations, there is strong motivation for developing models for spectrum usage. Such models find application in both short-term forecasting of channel availability for HF radar and communications applications, and for long-term prediction of HF spectrum usage for spectrum management and evaluation. In this paper, an extension of the channel availability model of Gibson (1993) and Gibson and Arnett (IEE Sixth International HF Conf., p.150-4, 1994) is detailed, which accounts for diurnal variations in spectral usage and propagation conditions. Channel availability within statistically similar frequency bands in a given season and solar cycle epoch is modelled as a cyclostationary two-dimensional first-order Markov chain. Three Markov parameters are required in the model, with each parameter expressed as a function of time of day.