This paper focuses on a spectrum usage model in the time domain in the context of dynamic spectrum access (DSA). To achieve a sophisticated dynamic spectrum access, understanding the spectrum usage is an important task. We focus on duty cycle (DC) as a feature quantity of spectrum usage in the time domain and observed DC (O-DC) obtained from long-term spectrum measurement results is used for the modeling. In fact, O-DC has stochastic and deterministic behaviors and we have been investigated modeling for both behaviors. O-DC is stochastic behavior and a mixture distribution based modeling has been considered for the model of stochastic behavior. We employ nonparametric Bayesian model (NPBM) in which the number of distributions is also an adjustable parameter. Statistics of O-DC, such as mean of O-DC, has a deterministic behavior in time domain. Specifically, the deterministic behavior is determined by the common daily habits, such as mean of O-DC during is night is low, but it is high during daytime. We show the validity of the stochastic and deterministic model for O-DC based on long-term spectrum measurement results.