In the past 30 years, the notion of landscape has emerged in ecology as a result of both theoretical considerations and practical aspects of land use and land cover. This has generated a variety of numerical models addressing both methodological and thematic objectives. Scientists model landscapes for at least two reasons: to better understand the landscape dynamics themselves (called intrinsic needs) and to offer a realistic frame to support other ecological processes (extrinsic needs). This paper mainly concerns the intrinsic needs; it reviews and discusses the way the socioeconomic and/or ecological mechanisms of various landscapes have been explored through modelling approaches in the past. Our objective is to identify the possible lack of understanding in landscape dynamics and to propose a unified view of this complex object. We outline the links between the concepts of landscape and of models using a double-entry matrix, focusing on one hand on the four main terrestrial landscapes (agricultural, forested, arid and urban) and on the other hand on the main landscape model characteristics (explicit or neutral, patchy or continuous, and multi- or mono-scale). The patterns and processes of each of the four landscape types, in particular, are analysed within a coherent framework. The heterogeneity of this yet coherent analytical matrix implies the need for unifying concepts and formalisms. The complexity theory and related concepts such as self-organization or formal grammar applied to landscape mosaics could help to further develop the mathematical formalisms necessary to assemble the various inner landscape processes. The discipline can now offer a theoretical dimension to dynamic landscape modelling aiming at understanding the mechanism unity underlying this complex object.