Tensor-CA: A high-performance cellular automata model for land use simulation based on vectorization and GPU

With the ability to understand linkages and feedbacks between land use dynamics and human-land relationships, cellular automata (CA) are extensively applied in land use/cover change (LUCC) simulation. However, with complex transition rules and a growing volume of spatial data, conventional serial CA models cannot meet the demands of efficient computation. In this article, a Tensor-CA model using vectorization and Graphics Processing Unit (GPU) technology based on a tensor computation framework for optimizing multiple LUCC simulations is presented. Complex transition rules of LUCC-CA models are vectorized and formalized to tensor operations which are effectively solved by GPU. The proposed Tensor-CA model was applied to LUCC simulations in the Pearl River Delta of China. The experimental results indicate that the proposed model effectively improved the performance compared to Serial-CA, Parallel-CA, and GPU-CA.