中文摘要：

建立城市大气污染预测模型是治理城市大气污染的重要工作。在简述时间序列方法基本原理的基础上，分析了系数为变量的自回归滑动平均（ARMA）模型、截断ARMA模型，和残差为自回归综合滑动平均（ARIMA）的半参数方法等城市大气污染预测模型。以法国某城市为例，分别采用AR模型和系数为变量的AR模型对大气污染进行了预测。通过比较预测结果可知，基于非线性时间序列方法的城市大气污染预测模型可以提高预测精度，降低预测误差。

英文摘要：

It is important to propose reasonable models for foretelling the atmosphere pollution. After reviewing the fundamental theory of time serial methods, variable parameter autoregressive moving average（ARMA） model, truncation ARMA model and remnant ARIMA semi-parameter model were analyzed. A comparison was made between experimental and theoretical results obtained by autoregressive（AR） and variable parameter AR model respectively. It seems that the nonlinear time serial model can be used to forecast the atmosphere pollution.