A novel statistical decomposition of the historical change in global mean surface temperature

Abstract

According to the characteristics of forced and unforced responses to climate change, sophisticated statistical models were used to fit and separate multiple scale variations in the Global Mean Surface Temperature (GMST) series. These include a combined model of the Multiple Linear Regression (MLR) and Autoregressive Integrated Moving Average (ARIMA) models to separate the contribution of both the anthropogenic forcing (including anthropogenic factors (GHGs, aerosol, land use, Ozone, etc.) and the natural forcing (volcanic eruption and solar activities)) from natural internal variability in the GMST change series since the last part of the 19th century (which explains about 91.6 % of the total variances). The multiple scale changes (Inter-Annual Variation (IAV), Inter-Decadal Variation (IDV), and Multi-Decadal Variation (MDV)) are then assessed for their periodic features in the remaining residuals of the combined model (internal variability explains the rest 8.4 % of the total variances) using the Ensemble Empirical Mode Decomposition (EEMD) method. Finally, the individual contributions of the anthropogenic factors are attributed using a Partial Least Squares Regression (PLSR) model.