Territories are governed, administered and observed from partitions of space into territorial units. All these territorial partitions change over time, for political or administrative reasons. In this paper, we present two innovative ontologies - Territorial Statistical Nomenclature (TSN) and TSN-Change - for the modeling of territorial partitions over time, adopting a Linked Data (LD) approach understandable by both humans and machines. These ontologies are innovative as they are generic, enabling the publication of any territorial partition into the LD Web. Above all, they allow for rich descriptions of changes, from one partition version to another, through Multi-Levels Change Graphs that link together changes that happen at different territorial levels (e.g., major regions or districts levels). These RDF graphs constitute a knowledge base to extract patterns of change and simulate scenarios.