In global-mean terms, 2026 has been hot, but 2027 will be even hotter. Not just because of the long-term upwards trend in global temperatures, but because a Super El Niño is developing (perhaps becoming stronger than any previously observed; also sometimes bizarrely referred to as a “Godzilla” El Niño). Barring some colossal tropical volcanic eruption, the global mean temperature in 2027 will be higher than 2026 (or any previous year, for that matter).
El Niño and its cool counterpart La Niña, which together form the oceanic component of the El Niño–Southern Oscillation (ENSO), dominate interannual (i.e., year-to-year) variability in global-mean temperatures. They are one of the key reasons why the global mean temperature does not always increase year-on-year, despite continuously increasing greenhouse gas forcing: El Niño years are relatively warmer than La Niña years. The overall warming trend does, however, remain evident when accounting for ENSO: globally, more recent El Niño years are hotter than previous El Niño years, and more recent La Niña years are as hot as previous El Niño years, despite being cooler relative to neighbouring years.
There is a positive trend in annual-mean near-surface temperatures in most regions of the globe. So, if we look at the correlation between the global-mean annual-mean temperature and the temperature at each grid point, we find a very strong positive correlation almost everywhere (Figure 1). This is just another way of relating global warming to local warming trends: as the globe warms, so does the temperature where you live. No surprises there. (Note that using historical temperature time series includes all climate forcing and internal variability, and so does not simply represent the response to increased CO2, hence some of the spatial inhomogeneity.)
What this map does not tell us is whether there is a relationship between the interannual variability in the temperatures at the local (grid-point) level and the global level. In other words, what is the relationship between the non-trending component of temperatures globally and locally? To try to answer this question, we can linearly detrend the temperature time series at each grid point, and then compute the correlation between the grid-point and global mean time series (Figure 2).
The tropics light up as the hotspots, while the correlations over Northern Hemisphere landmasses are modest at best. In particular, there is effectively no linear relationship for large parts of Europe: here, there is little reason to assume a globally relatively warm year (such as 2027) will be tied to a relatively warm year locally, because the non-trending components are uncorrelated.
We can also see this looking at the UK’s annual mean temperatures over the same period (Figure 3). The correlation with the global mean temperature is strong and significant in the absolute sense, but this is being driven by the linear trends: the correlation between the detrended series is small and nonsignificant.
This might not be too surprising to many, but it is important context when discussing interannual global temperature variability and its relationship with local/regional impacts. While the upward march of global temperatures means the local temperatures also march up, knowing whether a specific year is relatively warm or cool for the globe – something we can predict very well – might not actually be that useful for what you experience, and could well be confusing.
Notes: Temperature can be either be high or low, not hot or cold. Something with a high temperature can be referred to as being “hot”. Linear trend removal does not perfectly isolate interannual variability but I hope it is good enough for this blog.