Most of the current global land cover datasets and global scale land-use models use a classification of land cover based on the dominant land cover type within a distinct region or pixel. Such a classification disregards the diversity and intensity of human influence on land systems. In this paper we propose a novel way of classification and modeling land-use using a classification based on land-use systems (LUSs) that represent specific combinations of human-environment interactions. A cluster analysis was used to identify and map these LUSs. The analysis accounted for population density, accessibility to market places, land-use/cover types and livestock densities. A conceptual framework was developed to model dynamics in LUSs accounting for both land cover and land management changes. LUSs changes were simulated based on changes in both local socio-economic and biophysical conditions and regional-scale changes in demand for agricultural products. The new land-use systems change model was used in the context of the integrated assessment model IMAGE. © 2012 Elsevier Ltd.