Modern climatology relies on the 'forcing–response' paradigm: rising greenhouse gas concentrations should lead to continuous warming, possibly nearly linear over the interval considered. However, the observed temperature dynamics after removing short-period natural variability look different. This work fits within the broader research on climate variability, where neighboring studies examine the influence of solar activity, orbital cycles, and internal oscillations, but here the focus is on isolating the step-like structure of warming.
Monthly temperature anomalies from HadCRUT4 for 1950–2013 were used. For each grid point, the linear influence of the El Niño index, derived as the first principal component (EOF) of HadSST3 data, was subtracted. This procedure was performed using the Climate Explorer web tool, allowing the result to be reproduced in less than ten minutes. Then, linear trend maps were constructed for the three quasi-stable periods, along with maps of mean temperature differences between periods.
Global mean temperature after removing the linear ENSO influence shows a step-like pattern: trends for 1950–1987, 1988–1997, and 1998–2013 are near zero, and almost all warming occurred during the shifts of 1987/1988 and 1997/1998. Trend maps for stable periods show no global changes, only local variations. Meanwhile, difference maps between periods indicate that most regions warmed rapidly precisely during the shifts. A simple linear regression with two factors (ENSO and climate regime) explains 78% of monthly and 87% of annual variability in global temperature.
This work differs from neighboring studies by focusing on discrete transitions between quasi-stable states rather than seeking a continuous response to external forcing. It provides the first quantitative test of the hypothesis that warming is concentrated in short shifts rather than distributed evenly. Such a result calls into question the adequacy of a purely forcing-based paradigm and points to the possible existence of a regulatory mechanism not accounted for in current climate models.
The shift hypothesis can be tested on longer records, including paleoclimate reconstructions, and on data from other models. It is important to determine whether the shifts of 1987/1988 and 1997/1998 are unique or whether similar transitions occurred earlier. Also, the physical mechanism that could hold temperature near a setpoint and then rapidly switch it remains to be identified.
The results may influence the interpretation of climate trends in models and the assessment of climate sensitivity to greenhouse gases. They are also useful for statistical description and forecasting of global temperature.
Test the robustness of the step-like structure on independent datasets and with different methods of removing natural variability. Build and test models with an explicit thermostat mechanism.
This work is directly related to an unresolved problem in climatology: why observed warming is uneven in time and what role internal variability of the climate system plays. It also touches on the question of whether negative feedbacks exist that could stabilize global temperature.
The idea of a climate thermostat was discussed as early as 1991, when the New York Times wrote about possible self-regulation of climate, and a cloud-based mechanism was proposed by Ramanathan and Collins.