The effect of spatial resolution on the performance of cellular automata-based land use change spatial simulation

Studies of land use change have been increasingly carried out in recent years. Land use change is seen as one of the fundamental factor for the operation of environmental system from local to global scale. There are various methods to study land use change and cellular automata(CA)-based spatial simulation is a popular one. While this method (CA-based spatial simulation) is widely used in study of land use change, there are many aspects that need to be explored regarding its performance. Exploring the effect of spatial resolution on CA-based spatial simulation of land use change is the main objective of this research. Yogyakarta urban area was preferred as research area because of its interesting characteristic. Built up land is continuously increasing while agriculture land tend to decrease. Yogyakarta urban area consisted of the city of Yogyakarta and its suburban areas. Spatial simulation combined with experimental analysis were used as the main methods. CA-based spatial simulation were performed on different scenarios i.e. different spatial resolution of the data input. This study used three different spatial resolution that are 10 m, 50 m, and 75 m. Univariate statistical analysis against empirical data of land-use change was conducted to determine those spatial resolutions. Performance of CA-based spatial simulation was accessed using Kappa Index of Agreement (KIA) and two indices of spatial pattern, i.e. variance to mean ratio (VMR) and Moran's I. This study shows that higher spatial resolution of data input tend to generate a more clustered spatial pattern on the simulated map. The minimum and average value of actual land use change area could be utilized as consideration for determining appropriate spatial resolution. Medium spatial resolution particularly for extended spatial simulation produce more "visually realistic" spatial pattern. © 2019 SPIE.