基于元胞自动机的海上溢油扩散模拟

Simulation of marine oil spill diffusion based on cellular automata

利用逻辑回归算法和决策树C5.0算法分别获取溢油扩散的转换规则,并构建了基于逻辑回归的CA模型和决策树CA模型。这两个模型仅需要设置起始影像、影响因子和权重等少数的变量,便可以方便地模拟出溢油的动态变化情况。把逻辑回归CA模型和决策树CA模型应用到Deep Spill项目的海上溢油模拟实验,结果表明逻辑回归CA模型的模拟总精度达到96.4%,Kappa系数达0.893,而决策树CA模型的模拟结果更为理想,其精度和kappa系数分别提高了0.2%和0.006。利用元胞自动机能够很好地模拟并预测出海上溢油的动态变化,可以满足对溢油快速响应的要求。

Cellular automata （CA） is an effective tool for simulating geographical process. In this paper, logistic regression and decision tree algorithm （C5.0） are introduced to obtain transition rules, which are used to build logistic regression CA model and decision-tree CA model. These two models are very convenient because they only need a few variables, such as starting image, impact factors and weights. And the simulation results of oil spill can be obtained. The logistic regression CA model and decision-tree CA model are applied to simulate the movement of oil spill in Deep Spill projects. Experiment re-sults showed that the overall accuracy and Kappa coefficient of simulation results in logistic regression CA were 96.4%and 0.893.Better results could be obtained using decision-tree CA model. Its overall accuracy and kappa coefficients increased by 0.2%and 0.006. Our experiment results showed that the CA models could simulate the dynamic changes of the oil spill and meet the requirements for rapid response of governments.