Comparison of spatial patterns between cellular automata model simulations and remote sensing observations

Cellular automata (CA) have been increasingly applied to ecohydraulics modelling owing to its advantages in simulating local behaviours and spatial patterns. However, methods for evaluating CA model performance, mostly comparing with RS observations, have yet to be established. A conventional way is to compute the overall error between model outputs and observations. This aggregated approach without spatial information makes CA lose its fundamental brilliance in ecosystem modelling. Therefore, methods that can not only indicate numerical difference, but also retain spatial information, have to be developed. In this paper, fractal analysis and multiple resolution "goodness-of-fit" have been introduced for the purpose. The methods developed are illustrated through a comparison between two synthesized images that came from different simulation scenarios of a CA based ecological model. K e y w o r d s ce l lu la r a u t o m a t a ; fractal ana lys i s ; mode l ca l ib ra t ion ; r e m o t e s ens ing ; spatial pa t te rn c o m p a r i s o n