Using geographically weighted regression to predict the spatial distribution of frozen ground temperature: a case in the Qinghai–Tibet Plateau

Abstract

This paper combines the use of principal component analysis (PCA) and the geographically weighted regression (GWR) model to predict the spatial distribution of frozen ground temperature. PCA is used to reduce the multicollinearity among covariates, while the GWR model is used to address the spatially non-stationary relationship between frozen ground temperature and its predictors, such as air temperature, latitude, longitude, and vegetation cover. Our approach is applied in a typical permafrost area in the Qinghai–Tibet Plateau, Western China. The result demonstrates the applicability of our approach in the spatial distribution of frozen ground temperature and shows that the approach can be used for analysis and prediction. This study offers insight into temperature monitoring of frozen ground, which helps prevent regional geological disasters.

Keywords

• geographically weighted regression
• principal component analysis
• frozen ground
• ground temperature
• permafrost
• spatial heterogeneity