分段回归在剔除精河、库尔勒水平摆气象因素影响的探索

Subsection Regression Method for Removing Meteorological Factors in the Jinghe, Korla Horizontal Pendulum

摘要:

以精河、库尔勒水平摆观测数据为因变量,以地温、气温、气压为自变量进行分段回归分析,结果如下:(1)地温和气温是影响精河、库尔勒水平摆观测数据年频段信息的主要因素,它们之间具有准线性关系;(2)去掉趋势后,观测数据与地温、气温、气压之间的线性相关性明显增强,说明观测数据的趋势转折变化不是气象因素造成的,可能是观测区域的构造应力变化所致;(3)研究表明分段回归是全样本回归分析方法的一种改进,是降低时间序列数据处理难度的有效方法。

Abstract:

Crustal deformation observation stations built near the surface cannot avoid distractions that include information from the Earth's interior and exterior.Identifying earthquake precursors and other distraction factors is currently difficult.The regression method is a mathematical model for quantitatively deducting distraction factors.However,a mathematical model cannot describe the change in tendency during different stages when building a regression model.Subsection regression was used to divide data into sections to assist in the use of a mathematical model to conduct regression analysis.The dividing point of the regression model was the turning point for the change in tendency of deformation data.The selection of this dividing point can influence model quality.This study used ground tilt observation data from horizontal pendulums at the Jinghe and Korla stations as the dependent variable and meteorological factors as the independent variable.The tendency change of deformation data was divided into several parts and correlation coefficients were calculated.Scatter diagrams were created to test the results with the model.The results indicated that air and ground temperature were the main distractions for the fixed-point deformation of the horizontal pendulums at the Jinghe and Korla stations.The relationship between meteorological factors and deformation data strengthened existing knowledge regarding deformation observations.The linear correlation of meteorological factors and deformation data was strengthened after deducting the change in tendency.Therefore,the change in tendency may have been caused by regional stress changes according to the study on deformation data from the Sichuan-Yunnan rhombic block and the anomalies observed before the Xinyuan,Hejing earthquake of M_S6.6 in Xinjiang by Sun Yi and Wang Zai-hua.Subsection regression may be the most efficient method to reduce data processing difficulty because it improves performance of the regression method.