The authors isolate the trend of human-driven warming in the long-term global temperature record, removing the influence of natural factors, such as El Ni\u00f1o, volcanic eruptions and solar variation.<\/p>\n
They find that the world had been warming at a rate of around 0.2C per decade since the 1970s, but has \u201caccelerated\u201d since 2015 to a rate of 0.35C per decade.<\/p>\n
The study warns that if the current rate of warming persists, the 1.5C Paris threshold will be breached in the next few years.<\/p>\n
\u201cThe essential result of this paper isn\u2019t how fast we\u2019re warming, but that warming is now happening faster than before and that the difference isn\u2019t negligible,\u201d an author on the study tells Carbon Brief.<\/p>\n
Warming signal<\/p>\n
The year 2024 was the hottest on record<\/a>, with global average temperatures at the surface exceeding 1.5C above pre-industrial levels for the first time.\u00a0<\/p>\n Crossing the 1.5C threshold in a single year is not equivalent to a breach of the Paris Agreement<\/a>, which refers to long-term warming \u2013 typically interpreted as over a 20-year period<\/a>.<\/p>\n However, rapidly rising global temperatures are prompting scientists to ask when this internationally recognised threshold might be broken.<\/p>\n Human activity has been the primary driver<\/a> of rising global temperature in the long term, through greenhouse gas emissions and land-use change. However, natural factors also have warming and cooling effects from year to year.<\/p>\n The study authors identified three main sources of this natural variability.<\/p>\n El Ni\u00f1o and La Ni\u00f1a \u2013 collectively referred to as the El Ni\u00f1o-Southern Oscillation (ENSO)<\/a> \u2013 are generally the largest<\/a> drivers of year-to-year fluctuations in global temperatures. The study authors identify volcanic activity<\/a> and changes in solar variation<\/a> as the other two main natural influences on global temperature trends.\u00a0<\/p>\n Study author Dr Grant Foster<\/a>, formerly from the consulting firm Tempo Analytics and now retired, describes these sources of natural variability as \u201crandom noise\u201d that sits on top of the long-term warming signal. He explains that \u201cthe larger the noise, the harder it is to see the real trend\u201d.<\/p>\n To isolate the warming trend, the authors used a statistical technique that they first employed in a 2011 paper<\/a> to remove the contributions of ENSO, volcanic activity and solar variation from the global temperature record.<\/p>\n The authors carried out this analysis on five separate datasets of global average surface temperature \u2013 NASA<\/a>, NOAA<\/a>, the Met Office Hadley Centre<\/a> and University of East Anglia<\/a>\u2019s HadCRUT5<\/a>, Berkeley Earth<\/a> and Copernicus ERA5<\/a>.\u00a0<\/p>\n The plots below show the global temperature between 1880 and 2024, relative to pre-industrial temperatures<\/a>, from the five datasets.<\/p>\n Each plot shows the original warming record (light blue), in which all drivers of warming are included, as well as the adjusted record (dark blue) which excludes the effects of ENSO, volcanoes and solar activity.\u00a0<\/p>\n Removing the effects of natural variability makes the years 2023 and 2024 slightly cooler, the study notes, but they remain the two warmest years since the beginning of instrumental record.<\/p>\n Acceleration<\/p>\n Record-high temperatures in recent years have led scientists to ask whether global warming is accelerating.<\/p>\n The authors of the new study decided to use two different statistical approaches to test whether they can identify a \u201cstatistically significant\u201d acceleration in global warming from the long-term temperature record.<\/p>\n The \u201cnoise\u201d from natural drivers of temperature change, such as ENSO, can make it tricky to spot underlying trends. However, Foster tells Carbon Brief that after removing the influence of natural variability, \u201cacceleration is easy to prove statistically \u2013 some might even say it becomes obvious\u201d.<\/p>\n Both tests find that warming is accelerating with more than 98% confidence<\/a> for each of the five datasets. When the same tests were run on the unadjusted data, they failed to reach even 95% confidence, showing the importance of removing natural variability from the warming signal, according to the study authors.<\/p>\n Under the first statistical approach, called a quadratic<\/a> analysis, the authors applied a single curved trend line to the warming signal.<\/p>\n For the second approach, the authors used a technique to identify the month when the rate of global warming changed noticeably. The different datasets estimated this date to range from February 2013 to February 2014. They then calculated the speed of global warming both before and after these dates.<\/p>\n Global temperatures increased at an average rate of around 0.2C per decade over 1970-2015, according to the study.<\/p>\n In contrast, the authors find that warming rates have increased to 0.34-0.42C per decade, across the five different datasets, since the February 2013-February 2014 period.\u00a0<\/p>\n The study reveals that the rate of warming observed over the past decade has been higher than any previous decade in the instrumental record.\u00a0<\/p>\n Foster tells Carbon Brief that \u201cthe essential result of this paper isn\u2019t how fast we\u2019re warming, but that warming is now happening faster than before and that the difference isn\u2019t negligible\u201d.<\/p>\n If this warming rate remains constant, the Paris Agreement <\/a>1.5C threshold would be breached between 2026 and 2029, the authors find.\u00a0<\/p>\n (Their approach estimates the 20-year period where the average exceeds 1.5C of warming, and the breach of the limit is taken as the halfway point in this period.)<\/p>\n The table below shows key results for the five different datasets, including estimates for the date that warming started accelerating, the rate of warming and the year that the Paris Agreement will be breached in each.<\/p>\n Results for the five different datasets, including estimates for the date that warming started accelerating, the rate of warming and the year that the Paris Agreement will be breached in each. Source: Foster and Rahmstorf (2026).<\/p>\n \u2018Statistical significance\u2019<\/p>\n There are \u201cmany opinions\u201d among climate scientists about how fast the planet is currently warming, Foster tells Carbon Brief.\u00a0<\/p>\n For example, a study<\/a> from Dr James Hansen<\/a> calculates a warming rate of 0.27C per decade after 2010. Similarly, the latest Indicators of Global Climate Change report<\/a> estimates warming of 0.27C per decade over 2015-24.<\/p>\n Foster continues:<\/p>\n \u201cBut we all agree it\u2019s higher than before. [The] thing is, we couldn\u2019t prove that statistically.\u201d<\/p>\n<\/blockquote>\n Foster tells Carbon Brief that in 2024, Dr Claudie Beaulieu<\/a> \u2013 an assistant professor at the University of California<\/a> \u2013 led a study<\/a> which concluded that \u201ca recent surge in global warming is not detectable yet\u201d.<\/p>\n Beaulieu used the same statistical method as Foster to investigate whether global temperature data shows an acceleration in warming. However, she did not first remove the natural drivers of temperature change, such as ENSO.<\/p>\n (Carbon Brief wrote about Beaulieu\u2019s work in more detail<\/a> when it was published.)<\/p>\n Foster tells Carbon Brief that the study was \u201cexcellent\u201d, adding:<\/p>\n \u201cThey found that confirming acceleration was a close call \u2013 the data are very suggestive \u2013 but not quite \u2018statistically significant.\u2019\u201d<\/p>\n<\/blockquote>\n Foster explains that after removing the natural influence, the warming trend is clearer, making it easier to find statistically significant warming levels.\u00a0<\/p>\n Beaulieu praises the new study, explaining that \u201cthe fact that the acceleration signal appears consistently across all five independent datasets is reassuring\u201d.\u00a0<\/p>\n However, she stresses that \u201cthe acceleration may prove temporary\u201d.<\/p>\n She says that \u201ccontinued monitoring over the next several years will be essential to determine whether the accelerated warming rate identified here represents a lasting shift\u201d.\u00a0<\/p>\n The study authors say that the main limitation of their work is that the method of removing natural variability is \u201cempirically based, but approximate and imperfect\u201d.\u00a0<\/p>\n Foster says:<\/p>\n \u201cWe estimate the impact of things like El Ni\u00f1o by comparing past values of the El Ni\u00f1o index to past temperature changes, hence we don\u2019t need to know the physics behind it, just the numbers. Statistical results like this are only approximate.\u201d\u00a0<\/p>\n<\/blockquote>\n Meanwhile, an acceleration in warming is supported by many other observations of the Earth\u2019s climate.<\/p>\n For example, ocean heat content \u2013 the measure of the amount of energy stored in the ocean \u2013 is rising year on year. There is also evidence of acceleration<\/a> in recent years, with the period from 2020 onward seeing the largest year-to-year increases in ocean heat content on record.<\/p>\n In addition, the Earth\u2019s energy imbalance<\/a>, which measures the difference between incoming solar radiation and outgoing radiation, has also increased<\/a> in recent years.<\/p>\n![\"Five](\"https:\/\/www.europesays.com\/ie\/wp-content\/uploads\/2026\/03\/warming-acceleration-compressed.png\")
Global temperature trends from five datasets, including (light blue) and excluding (dark blue) the effects of El Nino, volcanic activity and solar activity. Source: Foster and Rahmstorf (2026<\/a>).<\/p>\n\n\tData sourceDate of accelerationWarming rate (C per decade)Year to cross 1.5C
\n<\/tr>\n\n NASA<\/td>\n April 2013<\/td>\n 0.36<\/td>\n 2028<\/td>\n<\/tr>\n \n NOAA<\/td>\n February 2013<\/td>\n 0.36<\/td>\n 2028<\/td>\n<\/tr>\n \n HadCRUT<\/td>\n January 2014<\/td>\n 0.34<\/td>\n 2029<\/td>\n<\/tr>\n \n Berkley<\/td>\n February 2014<\/td>\n 0.36<\/td>\n 2028<\/td>\n<\/tr>\n \n ERA5<\/td>\n February 2014<\/td>\n 0.42<\/td>\n 2026<\/td>\n<\/tr>\n \n
\n
\n