Carrero, R., Navas, F., Malvárez, G., Guisado-Pintado, E. 2014. Artificial intelligence-based models to simulate land-use changes around an estuary. In: Green, A.N. and Cooper, J.A.G. (eds.), Proceedings 13th International Coastal Symposium (Durban, South Africa), Journal of Coastal Research, Special Issue No. 70, pp. 414–419, ISSN 0749-0208.
Understanding human-driven land-use changes is a key element for successful coastal management and planning. Land-use change modelling helps to increase our comprehension of the patterns of these changes and the interacting factors affecting them. In previous years, a wide variety of land-use modelling approaches have arisen, however not all of them fit with the complex and especially dynamic behaviour of land use changes in coastal areas. In this sense, the so called ‘fifth generation’ of models represent an opportunity, as they integrate computer modelling with artificial intelligence technology allowing a more flexible and non-linear approaches that better emulate the real and complex behaviour of land use changes in rapidly evolving environments such as coastal areas. In this paper the applicability of artificial intelligence based models to simulate land-use changes is explored and applied to a specific case in a coastal stretch. First the potential of a Cellular Automata (CA) is compared with other modelling approaches. In the study CA proved to be a powerful approach due to its capacity of performing dynamic and complex spatial modelling, its affinity to work with geographic and remote sensing data, and its compatibility with other models. Second, a specific CA-Stochastic model is tested in an estuary located in southwestern Spain. The model is applied to simulate land-use changes from 1999–2007, using a spatial resolution of 10×10 m-cell, considering eight representative land-use classes, 30 different land-use transitions and fifteen variables affecting each transition. Validation based on a fuzzy similarity method is performed to compare the real and simulated maps. Results show high analytic capacity and good performance of the model. Four major strengths are identified: i) the integration of physical variables with human-controlled variables; ii) the successful simulation of multiple, simultaneous land-use changes; iii) the management of high-resolution spatial data; and iv) the flexibility of the model. In terms of its application to coastal management and planning, CA-based models not only help to understand past land use changes and the natural and human factors behind them, but also represent a great potential to forecast land use changes and create future spatial-explicit scenarios to assist decision-making in the mid-long term.