InSAR time-series (InSAR-TS) analysis enables us to obtain the displacement time-series by using a number of SAR images repeatedly acquired on the area. Among the factors affecting the accuracy of the InSAR-TS analysis, this study focuses on three factors that may severely affect the signal detection limit: 1) the selection of the reference point (determining the offset in each interferogram), 2) ramp-type artifact that originate from inaccuracy in the orbit data or ionospheric disturbance, and 3) altitude-correlated tropospheric noise. Fukushima et al. (2019, Earth, Planets and Space) proposed an InSAR-TS analysis method to simultaneously solve for the displacement time-series and the error terms mentioned above as well as the error in the digital elevation model. In the proposed method, the unwrapped phase in interferograms is assumed to be composed of a linear combination of the LOS displacement, offset, planar ramp, altitude-correlated phase, and error in the used digital elevation model. A set of unwrapped small-baseline interferograms is then inverted to simultaneously obtain the displacement time-series and the parameters describing the error terms under the minimum norm condition on the displacement time-series. In this study, I applied the above-mentioned method after some updates such as introduction of the temporal constraint adopted by the NSBAS algorithm (Doin et al., 2011) and data masking, on the ALOS-2 data acquired around the Arima-Takatsuki fault zone in western Honshu, Japan. Data of four different Paths (20 and 21 from descending orbit, 127 and 128 from ascending orbit) obtained between August 2014 and March 2021 were analyzed. Some of the original interferograms contained severe noise such as a phase ramp equivalent to approximately 25 cm of LOS displacements. The average velocity field obtained by applying the method captured a relative range decrease of a few mm/year on the southern side of the fault, consistent with the results obtained from Sentinel-1 data analysis. Given the fact that the Sentinel-1 dataset had much favorable conditions (much larger number of data and much smaller ionospheric noise), the consistency in the average velocity field suggests the effectiveness of the proposed approach.