Forecasting land-use changes with the use of neural networks and GIS

Abstract

In the present study a spatial model, which combines GIS with artificial neural networks, has been developed for forecasting changes in land use. The model has been parameterized for the island of Lesvos (NE Greece) for the time period between 1975 and 1999 and employs an artificial neural network for predicting the patterns of development of the island's urban areas and olive groves, based on a series of input parameters such as population density, transportation network, location of urban areas, proximity to the coastline and elevation. In this context, data from 1975 and 1990 have been used as input and the model has been run to project (i) urban land development and (ii) patterns of olive grove cultivations, for the year 1999. Results demonstrate that the model can predict reasonably well the patterns of change of the island's urban areas, however its predictive ability regarding the changes in the extent of coverage of olive cultivations is considerably lower. The overall performance of the model and its advantages and limitations are critically assessed and future improvements are suggested.