Observers are important components of dynamic positioning (DP) systems for marine vessels, estimating unmeasured states, filtering noise, filtering out wave induced motions, and predicting states in the case of signal loss. In this paper, a sensor-based hybrid observer concept is investigated. The concept assumes that noisy position measurements are available only occasionally at a non-constant sampling rate. Predictions of position between the samples are provided by integrating acceleration measurements, which are available at a high rate (approximated to be continuous sampling). Estimates with smaller variance are computed by averaging multiple observer copies of position and velocity. The contributions of this paper include the design, stability analysis and simulation of a sensor-based hybrid observer combining noisy acceleration, velocity and position measurements of a vessel in DP using different sampling rates.