Propagation delay in the atmosphere is the most important artifact that has to be mitigated in the interferometric SAR applications for displacement estimates. In the case of 3D spatial heterogeneous refractivity distribution, propagation delay in the atmosphere known as atmospheric phase screen (APS) is difficult to be corrected without a priori knowledge of displacement location. Especially for ground-based radar interferometric (GRI) measurement in the mountainous area suffers from a significant effect of APS which is difficult to fully be compensated by conventional compensation methods. This paper presents a novel approach to compensate for the APS, especially for the GRI application. In this framework, the time-series clustering approach is proposed to identify the local APS pattern due to the heterogeneity of refractivity. The correlation of each couple of temporal profiles is employed to cluster phase profiles by the rational of the k-means clustering approach. The proposed method further compensates the residual APS by estimating the temporal behavior of each cluster. Synthetic displacements are generated to assess the performances of the method in disentangling terrain displacements and propagation delay in the atmosphere.