Temporal Knowledge Graphs (TKGs) extend static knowledge representation by incorporating dynamic temporal attributes, offering novel insights for spatiotemporal big data mining and analysis. Despite their temporal capabilities, TKGs remain incomplete, necessitating effective interpolation approaches to infer missing facts. Existing methods treat all facts within TKGs as time-varying and make significant efforts on exploring geometric operations on various embedding space. Nevertheless, they overlook persistent, time-invariant relationships and fail to fully exploit the hierarchical richness of temporal information. To address these limitations, this paper proposes Dual², a simple yet effective TKG interpolation method that explicitly integrates static and dynamic knowledge with multi-granular time information. By jointly modeling calendar-derived semantics and position-aware periodicity, Dual² captures both absolute temporal contexts and relative temporal patterns to comprehensively learn the evolutionary patterns of TKGs. Experimental evaluations on three benchmark datasets demonstrate Dual²’s effectiveness, achieving the state-of-the-art performance with 8.26% Mean Reciprocal Rank (MRR) and 12.11% Hits@1 improvements over existing methods. These results validate the importance of decoupling static-dynamic modeling in TKG interpolation.