Land use and land cover (LULC) dynamics are an integral component of global change. In this thesis, various approaches were developed to unravel the complexity of LULC classification and the subsequent application of the multi-temporal LULC data for land change modelling. This complexity is particularly relevant in this study, whereby the available multi-temporal remote Landsat images are noisy and of relatively low spatial resolution. First, a semi-automated object-based method using rulesets and supervised classification was developed. This method was applied to the multi-temporal Landsat images to produce LULC maps. As the outcomes of the classification were not sufficiently accurate for land change modelling, the LULC maps were subsequently augmented using expert knowledge and input from landowners. Second, since high-resolution aerial photos were available for portions of the study area for 1998 and 2004, a case study was done with image fusion. The case study compared LULC maps derived from the different levels of fusion to those from the non-fused images. The results indicated that the feature- and decision-level fusion produced LULC maps which could be used for land change modelling. Third, in order to develop a land change model, the augmented multi-temporal LULC maps were used for extracting transition probabilities for a Markov-chain land change model. However, the classical Markov-chain method does not consider the neighbourhood influence, whereas the cellular automata does. A flexible hybrid approach, combining the Markov-chain and cellular automata algorithms, was developed. This was done to model the LULC dynamic transition probabilities to drive the change. The model’s sensitivity was assessed and the hybrid approach was tested by simulations of contemporary and future LULC patterns in the lower Hunter Valley, NSW with transition probabilities derived from various methods.