A bottom up approach to the modelling of costal and land use evolution throught GIS

Abstract

Coastal areas stand in the intersection of human and physical factors, some of them interacting in a non linear fashion and presenting feedbacks, which typically are characteristics of complex systems. This kind of systems are very difficult to model in a top-down approach, due to the high number of variables in question and the nature of relations between them. The spatial dimension itself, also introduces some complexity which is difficult to capture in a deterministic approach. In this study, we propose a bottom-up approach for modelling the coastline evolution, integrating land use cover. The system is based in a probabilistic Cellular Automata (CA) model, which divides the study area in a grid. The hermetic structure of CA is overcomed by using transition rules whose weights have been calibrated by a Artificial Neural Network (ANN). In this way, the knowledge of past events is incorporated into the model and projected into the future, by means of using ”intelligence”; this contrasts with techniques such as linear regression, whose efficacy in the case study was evaluated to be much inferior. Finally, is important to mention that the use of a Geographical information Systems (GIS) environment, enabled to assemble together different types of data and overlap them in an efficient way, which would be very hard or impossible to do otherwise; therefore, we believe that the use of GIS is crucial in spatial based simulations. The case study for this model was an area of the Municipality of Almada (Portugal) that has an extensive coastal line, both Atlantic and estuary. The modelling of spatially dynamic and naturally complex phenomena occurring in these coastal areas, is important for the definition of an innovative strategy for their physical planning and also their environmental management.