The Paris Agreement’s of 2015 aim is to strengthen the global response to the threat of climate change by keeping global temperature well below 2 ̊C above pre-industrial levels and to pursue efforts to limit the temperature rise even further to 1.5 ̊C. In 2015 global temperature had just reached 1°C above pre industrial levels (see globalwarmingindex.org). Five years later in July 2020, the World Meteorological Organisation (WMO) announced that there is a 20% chance that one of the upcoming 5 years would reach the 1.5°C limit showing that we are closer than we think to what the Paris agreement is trying to prevent.
A few weeks ago, the Copernicus Climate Change Service launched the Global temperature trend monitor, an app to allow anyone to put the current changes in global mean temperature trends in the context of the 1.5°C target limit. The app is based on the ERA5 Reanalysis, the most recent an extensive global dataset of weather variables, and allows to extrapolate the observed 30 year surface temperature trend to infer the date the limit is reached. A slider allows to choose a starting point in time between December 2000 until the current month. One can see the acceleration of the warming trend showing the 1.5°C passing date decreasing from March 2045 in 2000 to February 2034 with the latest observations (February 2021).
Last December Carbonbrief.org published estimates based on the latest IPCC climate model simulations showing global temperature passing 1.5°C between 2026 and 2039 with a median at 2030 under the worst case emission scenario (SSP5-8.5). An almost identical results compared to the same scenario (RCP8.5) of last simulations IPCC simulations (between 2027 and 2036, with a median at 2031).
In an attempt to contribute to these estimates we looked at the ability of CMIP5 models to reproduce the recent observed trend. Indeed, projections in CMIP5 started in 2006 giving an outlook of the 2006-2020 period for which we now have observations. Instead of the original projections, we use our High-resolution CMIP5 dataset for the same scenario (RCP8.5). A dataset based on the CMIP5 projections available in the Copernicus Climate Change Service Climate Data Store that we “calibrated” (i.e. statistically downscaled) with ERA5 reanalysis data. As a result, the dataset is a coherent extension of the ERA5 reanalysis and allows direct comparison with ERA5.
The graphic below shows the annual time series of global temperature between 1980 and 2080. Looking at the last 20 years, it is obvious that several models overshoot the observations, a well known feature that is widely commented on (see for example Realclimate.org or Climate-lab-book.ac.uk). We arbitrarily kept half the simulations by selecting those that had the better agreement ‘with the observed average over the 2021-2020 period. We use the 10 simulations subset to evaluate the year the +1,5°C is reached.
The 1,5°C limit is reached between 2024 and 2035 with a median at 2028, a somewhat earlier estimation than the previous ones that we have also reported on the graphic. A couple of models reach the limit before 2025 which can be seen as coherent with the forecast made by the WMO last June.
Hitting the 1°5 limit within the next ten years is certainly the most alarming of signals about the urgency to reduce greenhouse gas emissions.
Figure. Annual time series of global temperature between 1980 and 2080. Difference relative to preindustrial levels is inferred by setting the 1981-2010 average to 0,67°C according to data from Globalwarmingindex.org. The ERA5 data is the black line and the models values are in grey. We arbitrarily separated the available 20 models in two groups according to their agreement with the average over the 2001-2020 period: the 10 closest simulations to observations are in dark gray, and the 10 farthest in light grey. To smooth-out the interannual variability, as in in the other evaluations mentioned above, we use a 30 year centered running mean (red line). This running mean combines observations and model values generating an envelope (in red) beyond 2005 since it starts including model values from 2006 onwards. Similarly, the running mean contains observations until 2034.