吉林省夏季降水量的混合预报模型研究

发布时间：2018-06-10 06:45

本文选题：大气环流因子 + 线性回归模型　；参考：《东北师范大学》2017年硕士论文

【摘要】：本文是一篇注重探索性和过程性的文章,立足于多元统计分析、Lasso变量选择、均生函数、人工神经网络等多种统计学研究思想和建模方法,结合地理学、气候学、大气物理学等多学科理论知识,着眼于探索和建立吉林省夏季降水量的短期气候预报模型。经过不断的尝试比较和大量的实验分析,文章给出了一种新的建模方案,即利用Lasso变量选择技术,筛选出影响夏季降水量的主要大气环流特征量等物理因子和代表降水量主要周期振荡特性的均生函数函数延拓序列,并采用人工神经网络方法建立了双重因子的非线性混合预报模型。分别以吉林省西北部地区、中南部地区和东部地区三个气候区夏季6～8月降水总量作为预报对象进行建模预报试验。结果表明:基于双重因子的神经网络混合预报模型,比大气环流、ENSO等气象指数作为预报因子的回归预报模型和均生函数回归预报模型两种单因子线性预报模型,以及基于环流因子和均生函数延拓序列的混合因子双重线性预报模型,要具有更好的物理基础和预报能力。
[Abstract]:This paper is an article focusing on exploration and process, based on the multivariate statistical analysis of Lasso variable selection, average generating function, artificial neural network and other statistical research ideas and modeling methods, combined with geography, climatology, etc. This paper focuses on the exploration and establishment of short-term climate forecast model of summer precipitation in Jilin Province. After continuous trial and comparison and a lot of experimental analysis, this paper presents a new modeling scheme, which is to use Lasso variable selection technology. The physical factors, such as the main atmospheric circulation characteristics, which affect the summer precipitation, and the equal-generating function continuation series, which represent the main oscillation characteristics of the precipitation, are screened out. The nonlinear mixed prediction model with double factors is established by artificial neural network method. The total precipitation in three climatic regions of northwestern Jilin Province, central southern region and eastern part of Jilin Province in June to August was used as the forecast object for modeling and forecasting. The results show that there are two kinds of single factor linear forecasting models based on the neural network mixed forecasting model based on double factors, the regression forecasting model based on the weather index such as ENSO and the average generating function regression forecast model. The double linear prediction model of mixed factors based on circulation factor and extension sequence of mean generating function should have better physical basis and prediction ability.
【学位授予单位】：东北师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：C815

【参考文献】