利用分位数回归模型和经验模态分解 (empirical mode decomposition, EMD) 方法对 气候及其变化进行分析和预测. 首先, 采用全球热力图对气温数据进行描述统计, 并采用经验 模态分解方法降噪获取趋势项来引入全球气温周期概念, 探究全球气候变暖趋势; 然后, 基于 多元线性回归模型及分位数回归模型寻找全球气温的影响因素, 并对气温进行建模及预测. 研 究结果可为全球气候分析提供统计学支撑.

The purpose of this study is to analyze climate’s variation trend and forecast the climate. The main methods are known as quantile regression and empirical mode de- composition (EMD). Firstly, a global heat map is utilized for the descriptive statistics of global temperature data. The EMD method is applied for data denoising to analyze global temperature’s variation trend, and the concept of global temperature cycle is introduced. These aim to study the trend of global warming. Secondly, the multivariate linear regres- sion model and the quantile regression model are applied to identify factors influencing global temperature. Then the temperature model is built to predict temperature changes. The findings can provide statistical support for global climate analysis.