Empirical evidence for a double step climate change in twentieth century

TL;DR

This study provides empirical evidence for two major climate regime shifts in the 20th century, characterized by sharp temperature increases, and develops simple models to explain temperature anomalies using ENSO, PDO, and these shifts.

Contribution

It identifies two significant climate regime shifts in 1925/1926 and 1987/1988 and demonstrates that simple linear models using ENSO, PDO, and regime shifts can effectively reproduce historical temperature anomalies.

Findings

Two major climate regime shifts identified in 1925/1926 and 1987/1988.

Sharp temperature increases of about 0.28/0.36°C during shifts.

Linear regression models using limited data and two factors effectively reproduce temperature anomalies.

Abstract

In this study we used the sea surface temperature (SST), El-Nino southern oscillation (ENSO) and Pacific decadal oscillation (PDO) time-series for the time period 1900-2012 in order to investigate plausible manifestation of sharp increases in temperature. It was found that the widely observed warming in the past century did not occur smoothly but sharply. This fact is more pronounced at the latitude zone 30S - 60N during the years 1925/1926 and 1987/1988. We hypothesise that there were two major climate regime shifts in 1925/1926 and 1987/1988 years. During these shifts the mean value of temperature rises, over which natural variability associated with ENSO, PDO and other factors occurs. During each sharp increase mean SST in tropics/north middle latitudes increased by about 0.28/0.36 {\deg}C. Most of other temperature anomalies are explained by ENSO and PDO. The existence of these shifts tends to be masked by natural variability. This hypothesis has allowed us to develop very simple linear regression models which explain the main features of temperature anomalies from 30oS to 60oN observed in the past century. Additionally, two remarkable outcomes revealed from this analysis. The first one is that linear regression coefficients can be calculated by employing a limited length of the temperature time-series (e.g., from 1910 till 1940) and reproduce quite well the whole time-series from 1900 up to now. The second one is that a good quality of reproduction is achieved by using only two factors (ENSO/PDO and sharp changes for tropics/northern middle altitudes).