Abstract: Aiming to investigate the dynamics of arable land, specify the underlying driving forces and project future spatiotemporal dynamics, categorical maps of arable land in the Yellow River Delta from 1992 to 2005 were obtained by via remote sensing and geographic information system techniques. Cellular automata models were calibrated with arable land maps in 1992, 1996 and 2001, and were conducted to predict the arable land maps in 2001 and 2005. The overall accuracy of simulation in 2001 and 2005 were 82.90% and 84.48%, while the values of Cohen’s kappa index amounted to 0.658 and 0.689. The results showed that the cellular automata model was able to simulate arable land dynamics effectively and could be a useful tool to project future patterns of arable land in the Yellow River Delta. Cellular automata model was conducted to predicted arable land patterns in 2010 under four different scenarios. Scenario analysis showed, due to the high level of soil salinity, arable land around the abandoned Yellow River of Diaokou was the most region to degenerate.