We're perhaps between 50 and 60 years too late to prevent catastrophic climate change simply by ceasing emissions, scientists say.
Temperatures will continue to rise for a long time after we reach carbon-zero, new research has found, and the only way to avoid the worst will be to start actively taking carbon out of the atmosphere.
But the research has been criticised by Kiwi climate scientists, saying the findings are unrealistic and based on low-complex modelling, rather than state-of-the-art systems.
Two Norwegian researchers ran simulations of the global climate between 1850 and 2500, looking at how different levels of greenhouse gas emissions affected temperatures and sea level rises.
Under the most likely optimistic scenario - emissions peaking in the 2030s before gradually declining to zero by 2100, by 2500 the seas would be 3m higher and temperatures on average 3C warmer than in 1850. 1850 is when the concentration of carbon dioxide in the atmosphere began to skyrocket, thanks to industrialisation.
Even if we stopped emissions today, "global temperatures will still be around 3C warmer and sea levels will rise by around 2.5 metres by 2500", the modelling showed.
"We have identified a point-of-no-return in our climate model ESCIMO - and that it is already behind us," climate scientists Jorgen Randers and Ulrich Goluke wrote in journal Scientific Reports.
"In ESCIMO the global temperature keeps rising to 2500 and beyond, irrespective of how fast humanity cuts the emissions of man-made greenhouse gas emissions."
There are three main drivers of this continued growth in temperatures and sea level rises. The first is a "cycle of self-sustained melting of the permafrost". It's already melting, releasing methane - a gas with an even stronger greenhouse effect than carbon dioxide - which causes more temperature rises, and more melting.
The second is "lower surface albedo" - with less of the Earth's surface covered by ice and snow, less of the sun's energy is reflected back into space. This means it absorbs more, contributing to warming.
And the third is water vapour. A warmer atmosphere is a more humid one, and water vapour is also a greenhouse gas.
Under the first scenario - cutting emissions to zero by 2100 - the model found temperatures would peak at 2.3C above 1850 around 2075, before declining over the next 75 years.
But then something unexpected happens - temperatures start going back up again.
"The surprising fact is that this rise takes place 50 years after man-made emissions have ceased, and after the concentration of CO2 in the atmosphere has been significantly reduced through absorption in oceans and biomass."
The culprit is the melting permafrost and ice.
"The effect of surface albedo continues on its smooth upward path throughout this period... it has enough momentum to push the climate system back onto a path of rising temperatures, with its secondary effects of raising humidity and permafrost melting, which then in turn help the system become warmer and warmer, even if man-made greenhouse gas emissions are zero.
"A cycle of self-reinforcing processes is established... The process is self-sustaining, at least until all carbon is released from permafrost and all ice is melted."
Even if emissions are cut to zero right now, the same thing happens - albeit about 75 years later. In other words, we're past the point of no return.
The scientists ran further tests to see when we had to act by, and the news wasn't good.
"The answer is that all man-made emissions would have had to be cut to zero sometime between 1960 and 1970 - when global warming was still below 0.5C."
The temperature now is about 1C higher than it was in 1850, and increasing about 0.2C a decade - at an accelerating rate.
To avoid the self-sustaining melting, they calculated 33 gigatonnes of carbon dioxide would need to be removed from the atmosphere from 2020.
"In other words, building 33,000 big carbon capture and storage plants and keep them running forever."
Worldwide, there are presently only 51 large-scale carbon capture and storage plants either in operation or under construction, according to a May report from US-based non-profit Resources for the Future.
The authors of the latest study admit building 33,000 plants is "technically feasible but would be hugely expensive".
"Cheaper opportunities exist to stop self-sustained global warming (through various forms of geo-engineering), but these will have unintended and undesired side effects beyond lowering the temperature."
While other climate scientists described the study as "interesting", the modelling used has been criticised as simplistic.
"Earth system models typically contain hundreds of thousands of lines of code, and require a supercomputer to run on. Even then, a 100-year simulation may still take several months to run," said Laura Revell, environmental physicist at the University of Canterbury.
Instead they used a low-complexity model that can be run in a matter of seconds.
"Reduced-complexity climate models contain highly simplistic representations of the climate system and how different processes interact with one another. The authors do acknowledge this by inviting other researchers to investigate their findings with complex models."
James Renwick of the Victoria University School of Geography, Environment and Earth Sciences said the findings were at odds with other research, which suggest no further warming once emissions are stopped.
"The results presented in this paper are very implausible and should not be seen as cause for alarm."