Identifying changes and critical drivers of future temperature and precipitation with a hybrid stepwise-cluster variance analysis method

Abstract

In this study, a hybrid stepwise-cluster variance analysis (HSVA) method is developed for generating future climate projections and identifying critical drivers of temperature and precipitation changes. The proposed HSVA method incorporates global climate models (GCMs), stepwise-cluster analysis (SCA), and analysis of variance (ANOVA) techniques within a general framework. It has advantages in (1) dealing with continuous/discrete variables as well as nonlinear relations between predictors and predictands and (2) quantifying the significant effects of atmospheric variables and interactions on climate change. The proposed methodology is then applied to the Kaidu watershed in northwest China for examining its applicability with consideration of different GCMs and representative concentration pathways (RCPs) during 2010–2099. Results demonstrate both increases in annual temperature (with a rate of 0.1–0.6 °C per 10 years) and precipitation (with a rate of 1.0–13.6 mm per 10 years). Surface upwelling longwave radiation (RLUS), highly related to land surface energy exchange, is the primary factor that has significant effect on climate change. The interactions between RLUS and the minimum near-surface temperature (TASmin) as well as RLUS and near-surface temperature (TAS) would affect future temperature and precipitation, implying that RLUS has a large relation with climate projections and needs to be paid more attention in downscaling practice. Identification of these factors and interactions could help better understand the process of climate change and improve downscaling performance.