Landscape (i.e., land cover, land use or vegetation maps) is a very complex mosaic of thousands of patches, and this makes its interpretation very challenging. Class areas and shared perimeters between classes are two pivotal properties of its structure. In addition, landscape structure changes over time as a consequence of many interacting processes. Hence, there's an urgent need for a synthetic and intuitive representation of its structural attributes. I advocate here network graphs as an aid to interpreting and checking temporal and spatial properties of landscapes. I also suggest several hints to fitter use network graphs in landscape representation. As a case study, I apply network graphs to the Ceno valley (Parma, Italy), but the proposed approach is suitable for any landscape maps.