在众多天气因素中，本文主要对北京市冬季平均气温进行分析和预测。 首先，再进行更深入的分析之前，我们先对北京市1977-2006年12月，1月，2月三个月的气温资料做了初步整理，并进行正态性 检验，这个是后面很多分析方法的基础和前提条件。 其次，我们关心在这30年内，北京市冬季的平均气温是否有很大的波动。在对这方面进行分析是，选取了传统的$QC$方法。通过 画出bar{X}和R控制图，来分析30年中气温的波动。从最后的图中可以看出，这三十年中，北京市冬季气温没有发生太大的 变化。 再次，在分析气温的过程中，很重要的一个方面是对影响气温的因子进行回归。我主要采用了四种方法：多元线性回归，逐步回归， 以及事件概率回归和logistic回归。针对每种不同的方法，可以得到回归方程。 接着，对三个月的气温进行了主分量分析，将其降维，得到两个主分量，并能够解释这三个变量。 最后，在对气温的预测中，采用了时间序列的方法。由于原始序列不平稳，因此对一节差分后的序列建立了$ARMA(2,1)$模型。 在这个模型的基础上得到五年的预测值。 除此之外，在本文最后还对北京市冬季降雨量给出了回归分析，找到影响降雨量的主要因素和回归方程。能够更好的帮助人们了解和认识影响气候的原因。

Among several factors which affect the climate, I made the analysis and forecast focusing on the winter average temperature of Beijing. In the first place, before the further analysis, I deal with the data simply and do the normal test for the data. Actually, the normal assumption is the fundamental to the further analysis. In the second place, I concerned about whether there was obvious fluctuation during these 30 years. I applied the traditional QC method to detect it. From the bar{X} and R Chart I drawn, I could tell that there was no obvious evident to show the out of control condition. In the third place, one of the most important parts in the process of analyzing is regression. Here, I made use of four different ways: multivariate regression analysis, step regression analysis,regression estimation of event probability(REEP) and logistic regression. Regards to each method, I got the regression equation. Furthermore, in order to reduce the dimension of factors, I took advantage of principal component analysis. And based on the result, the 3 factors have been reduced to 2 factors which could explain the original ones. Finally, the time series has been used to forecast the temperature. Since the original series is not a stationary series, I made one step difference. Based on the new stationary series, I constructed the ARMA(2,1) model and did the 5 years forecasting.