Terahertz (THz, 0.3-3 THz) communications are envisioned as one of the enablers at the air interface for sixth-generation (6G) cellular systems. However, by utilizing large antenna arrays to overcome severe path losses, this system will suffer from micromobility phenomenon manifesting itself in occasional rotations and displacements of user equipment (UE) in the hand of a user. In this paper, based on the measurements of micromobility patterns we propose several models characterized by various degrees of details to capture micromobility patterns of different applications. By utilizing the time to the outage as a metric we compare their accuracy. Our results show that drift to the origin is a critical property that has to be captured by the model. While the two-dimensional Markov model is shown to provide the most accurate approximation, the decomposed Brownian motion model is characterized by the worst match of data. The decomposed one-dimensional Markov model provides the trade-off between simplicity and approximation accuracy. © 2021 ACM.