Artificial neural networks (ANNs) have recently attracted the attention of applied geographers. They can be employed to examine relationships in complex non-linear datasets in the same way as conventional statistical techniques, but without many of the parametric restrictions of these techniques and prior assumptions about the nature of the data relationships (such as linearity). One of their uses is in forecasting. In this study, summer surface ozone concentrations are estimated using surface meteorological variable as predictors by a multi-layer perception neural network for five locations in the UK. The relationship between weather and ozone is highly complex and non-linear. The performance of the ANNs is evaluated by comparison with results from a regression-based model. It is found that although this ANN model improves the accuracy of prediction it is not considered to be a dramatic improvement.