Thank you, Adam. That's the connection I keep coming back to. The Holocene is the only climate state agriculture and cities have ever been tried in — everything we call history happened inside it. Which is why the duration is the part that matters to me as much, if not more, than the peak.
My goal has been to try and make climate change more legible and accessible. If it would be useful to your readers, you're welcome to crosspost it.
From the research/reading I’ve done it seems like 2 to 2.3celsius warmer than preindustrial is the “game over” range for organized agriculture both crop/livestock.
Although with the blinding speed of the fossil fuel warming we’re inflicting on the planet it could be lower because of the violence of the rapid heating.
Honestly, agriculture isn't my area, although I did grow up on farm. I plot CO₂ and temperature and try not to claim more than the data I've plotted, so take this as a pointer rather than an answer.
What I've read in AR6 doesn't support a single cliff. Losses concentrated in the low and mid latitudes, some high-latitude gains, and adaptation making a real difference — for Africa the median loss at 2 °C drops from roughly a third to roughly a tenth once adaptation is counted. Serious, regionally brutal, but not one number where it stops working.
Where I think your instinct about speed is right is that the threshold framing may be the wrong one. Agriculture was invented inside the Holocene, and every crop and every planting calendar we have was tuned to it. What my figure says isn't that we cross 2 °C — it's that we don't come back. On the pathway current policy delivers we stay outside that window for about sixteen thousand years. A bad decade is weather. A climate that never returns to the one your seeds were bred for is a different problem, and it's the one I'd worry about.
BTW, I see your restack and thanks for that. A crosspost is different. It actually goes up as a post on your Substack and goes out directly to all your subscribers. I get that that's a bigger ask so that's totally up to you.
Good question, and it's answerable from the same published data the figure uses — though these are medians of a large ensemble, so I'll give the spread as well.
Your hypothetical is closest to ScenarioMIP's Very Low pathway, which reaches net-zero greenhouse gases around mid-century and then holds modest net-negative emissions. On the same axes as the figure:
Peak around 2050, near 15.7 °C — about 1.8 above preindustrial. By 2100, about 15.5 °C (+1.6), with a 5–95% range of 1.0 to 2.4. By 2200, about 15.2 °C (+1.3), range 0.8 to 2.1.
CO₂ falls from roughly 445 ppm at mid-century to 405 by 2100 and 365 by 2200. Very Low keeps removing a little carbon past net zero, so concentrations decline rather than hold flat, and temperature follows them down.
The Medium pathway on my figure is about 16.9 °C in 2100 and 17.9 in 2200. What's worth noting is that the ranges don't overlap — the warmest Very Low outcome at 2200 is still cooler than the coolest Medium one.
Today is about 15.3 °C, so Very Low in 2100 sits a little above the present and by 2200 a little below.
I would not read anything into a second decimal place. These are FaIR medians with wide ensembles behind them, and the Earth system model runs that will eventually replace them aren't finished.
Since our starting point is below the equilibrium temperature, it seems like the temperature should continue rising until equilibrium is restored. The projections show that the temperature is expected to peak and then start to fall even when it is still well below the equilibrium temperature for the anticipated CO2 concentration.
Good question — the answer is that the CO₂ doesn't hold still. Temperature is chasing a target that is itself falling. After net zero the ocean, and then seafloor carbonate, draw carbon dioxide down, so the equilibrium the system is relaxing toward comes down faster than the system can climb toward it. The gap closes by the target descending, not by temperature rising.
Zeke Hausfather's explainer is the best short treatment: https://www.carbonbrief.org/explainer-will-global-warming-stop-as-soon-as-net-zero-emissions-are-reached . He names exactly this confusion — holding CO₂ fixed at today's level is a different world from reaching net zero and letting concentrations fall. One difference of scale: he's describing the first few centuries, where temperature stays roughly flat. The figure runs to tens of millennia, which is where the decline shows up.
It looks like Hausfather’s explainer implies that we are much closer to equilibrium already than what your chart shows. That’s the disconnect. You in fact show the concentration dropping over the next 5-to-10k years, which is why the equilibrium temperature comes down. But it is always higher than the current and predicted temperature. From a simple controls-theory perspective, the temperature has to keep driving towards equilibrium the whole time. I think you and he just have different ideas of where the equilibrium is.
The band isn't my estimate of equilibrium — it's the fitted paleo record, plotted as it comes. Judd, EPICA, Clark. Whether that slope is the right thing to compare a transient against is a fair question, but it isn't a disagreement between Hausfather and me about where equilibrium sits.
Very interesting post Dean. I liked your comment about feedbacks as that was exactly what I was thinking reading it. Looking at the method post too, I think what you have created is a fully believable best case scenario. It would be interesting to see a version with some permafrost melt contribution built in. The ICCI has some basic data in their last state of the cryosphere report.
Maybe add up to a 5Gt annual contribution to the Medium emissions profile ramping up from 0.5 today to 5Gt by 2100, then stable with a slow decline afterwards. The mechanism would slow for permafrost as it melted away but be replaced by wetland inundation as sea levels rose through ice sheet decay. You could stabilise them out at say 10,000 years.
I'd be happy to run some scenarios if you send me the model code.
Hi Dean, another beautiful graph and essay. Your graph clearly shows how far we are and will be out of equilibrium and also suggests the path back to equilibrium. As you mentioned the modeled paths to 2500 do not include the slow climate feedbacks such as glacier and ice sheet melting, change of vegetation and other things. Whereas it is OK to ignore the slow feedbacks during this century they probably can't be ignored over the next 500 years and certainly not over the ten thousand years of your extension.
I think one can get an estimate for the magnitude of the time constant of the slow climate feedbacks from the temperature and CO2 record during the ice ages, particularly the end of the last ice age. The temperature and CO2 concentration tracked each other very well - a fact you used for your equilibrium curve. During the rapid changes at the end of the last ice age the lag between temperature and CO2 was about 500 years or less. I think this could be used as an estimate for the slow feedback time constant.
If true then the path to 2500 will already start to significantly bend towards the equilibrium band and probably reach equilibrium before the end of this millenium. Carbon sequestration from weathering will then mostly happen while the earth is on the equilibrium curve in the same way it always happened.
Unfortunately this implies significantly higher temperatures of 20 C or more. I think this is what Jim Hansen means when he talks about "warming in the pipeline". I understand that it is difficult to include the slow feedbacks in the climate models but that doesn't mean that they can be ingored.
homas — thank you, and this is the right thing to press on.
I think the lag you're pointing at measures something different from what we need. At the end of the last ice age, CO₂ follows temperature — the ocean gives up carbon as it warms. In the glacial cycles CO₂ was a feedback, not a forcing; the forcing was orbital. So that 500 years is the carbon cycle's response time to warming, not the ice sheets' response time to CO₂.
There's a second problem, and it's visible on the figure. The ice-age points sit essentially on the band — scatter about the fit is around 1 °C. That isn't because the slow feedbacks are fast. It's because orbital forcing is slow: precession, obliquity and eccentricity run over 20,000 to 100,000 years, far slower than any feedback in the system, so the ice ages never departed far from equilibrium. You can't read a response time out of a record where the system is always equilibrated. The tight tracking is evidence about the forcing, not the response.
The direct measure is the deglaciation itself. The ice sheets took roughly ten thousand years to go. That's the time constant — millennia, not centuries.
On Hansen, I don't want to wave him away; it's a serious argument. But notice where the premise leads: at Medium's 619 ppm the band sits at 23.5 °C, not 20. If the slow feedbacks equilibrated on 500 years, the answer would be worse than the number you quoted.
Where I come out is that the gap is real and it isn't warming owed to us. It closes from the other side — CO₂ falls before the slow feedbacks can act on it. That's the whole shape of the loop.
Thanks for the quick response. This discussion helps me to develop my thoughts, so here we go:
First a fundamental point that I think is not properly appreciated by many climate scientists: CO2, unlike other GHGs, determines a planet's temperature because the time constant of the carbon cycle is much longer than the time constants of the climate feedbacks, fast or slow. If this were not the case a planet would never warm up long term due to CO2 emissions. CO2 would have a similar effect as methane - only a short term warming.
Your paths on your graph make this explicit: because you (and the climate models) assume that the carbon sequestration is faster than the slow climate feedbacks, you quickly (10k years) end up at the starting temperature. With carbon sequestration faster than climate feedbacks the earth would never have warmed up much. Your beautiful band of climate equilibrium would never have existed.
Then the next question is: what is the slow climate feedback time constant? It is difficult to determine from observations today, which is what the climate models mainly rely on. We have to look at the earth's climate history on the scale of at least tens of thousands of years. The record of temperature and CO2 during the ice ages is a great place to look.
To determine a time constant of a system we need to observe the response to an excitation or forcing as the climate scientists like to call it. The type of forcing doesn't matter, it only has to bring the system out of equilibrium. If the forcing is fast then the system relaxes with its time constant. If the forcing is of similar duration as the time constant then you get a lag. I think this happened at the end of the last ice age. This is why I think the slow climate time constant is about 500 years. Note that this doesn't identify what the time constant is due to. It is just the response of the earth system. Also, the time constant can and likely does depend on the earth's temperature, but we are probably close enough to the ice ages that it should not be very different.
BTW, I do think that it could be much worse than 20 C.
Your first point I think is right and underappreciated — CO₂ is the control knob precisely because its residence time is long compared with the response times of what it drives. Methane is the counterexample that proves it.
On the time constant, I'd point you at Clark et al. 2016 in Nature Climate Change, "Consequences of twenty-first-century policy for multi-millennial climate and sea-level change." Twenty-two authors including Pierrehumbert, and Peter Clark is one of the sources behind my equilibrium band. They put the ice-sheet response squarely in the millennia, with sea level still rising ten thousand years out. Their argument is that the 2100 framing obscures the problem, which is close to what I'm trying to say with the loop.
On where my return starts and what it relaxes toward — you're pushing on the right seam, and it deserves better than a comment box. I'm writing a companion piece on how the model is built: what's imposed, what's calibrated, what falls out, and where it's weakest. The slow-feedback question is going into it. I'll come back to you when it's up.
Thanks for your reply and I am looking forward to your next essay. I am making a few more comments but don't feel you need to reply.
Your graph stimulated me to look into the slow feedback time constant, something I wanted to do for a while. I think CO2 and temperature are a coupled system: when CO2 goes up then the temperature goes up, when the temperature goes up CO2 goes up and also the inverse. So, from a system perspective there is only equilibrium conditions and time constant that define the system. No need to know any details of the system. This makes the record of CO2 and temperature during the ice ages so valuable. The close tracking of the two means strong coupling and short time constant compared to the scale of the ice age duration. From the lag during fast changes the slow climate change time constant is about 500 years.
I looked at the Clark et al paper. It is mostly about sea level rise, which is driven by ice sheet melting. As they are pointing out in the paper sea level rise continues long after the temperature has stabilized during the ice ages. This means the temperature/CO2 time constant is much shorter than the ice sheet melting time. I also noticed that they use an ECS of 3.5 C for their future projection but the ECS of the temperature/CO2 data for the past in figure 1 is 6.2 C even though both past and future are of similar duration. This inconsistency is what you get if you only use models and ignore the information from the climate history, something Jim Hansen first pointed out.
Excellent piece.
When it comes to understanding the climate crisis the Holocene is key.
Before 12 thousand years ago the highest estimates put the total global human population at 300 thousand.
After the beginning of the Holocene and in the geological blink of an eye, we’re now at 8 billion.
Thank you, Adam. That's the connection I keep coming back to. The Holocene is the only climate state agriculture and cities have ever been tried in — everything we call history happened inside it. Which is why the duration is the part that matters to me as much, if not more, than the peak.
My goal has been to try and make climate change more legible and accessible. If it would be useful to your readers, you're welcome to crosspost it.
I already did.
From the research/reading I’ve done it seems like 2 to 2.3celsius warmer than preindustrial is the “game over” range for organized agriculture both crop/livestock.
Although with the blinding speed of the fossil fuel warming we’re inflicting on the planet it could be lower because of the violence of the rapid heating.
What have you come across?
https://finance.yahoo.com/markets/commodities/articles/worst-us-wheat-crop-since-120105002.html?fr=sycsrp_catchall&guccounter=1&guce_referrer=aHR0cHM6Ly9kdWNrZHVja2dvLmNvbS8&guce_referrer_sig=AQAAAFLesnRcz5NMix-lE9-6ye-vJhsXTqMl6xBgWM0S2HwULS9H1rlP0WdsH5_WnXQVmVkiLpKomTSpovHOccohLqWjaMiSsqTIZMv3pqoRe4Wn5ugBTWYuISgBN59TnB6yuQOReKV5J_JY6hnf9LEMuBqprrf0tK1sjyFbm-n0a_E2
Honestly, agriculture isn't my area, although I did grow up on farm. I plot CO₂ and temperature and try not to claim more than the data I've plotted, so take this as a pointer rather than an answer.
What I've read in AR6 doesn't support a single cliff. Losses concentrated in the low and mid latitudes, some high-latitude gains, and adaptation making a real difference — for Africa the median loss at 2 °C drops from roughly a third to roughly a tenth once adaptation is counted. Serious, regionally brutal, but not one number where it stops working.
Where I think your instinct about speed is right is that the threshold framing may be the wrong one. Agriculture was invented inside the Holocene, and every crop and every planting calendar we have was tuned to it. What my figure says isn't that we cross 2 °C — it's that we don't come back. On the pathway current policy delivers we stay outside that window for about sixteen thousand years. A bad decade is weather. A climate that never returns to the one your seeds were bred for is a different problem, and it's the one I'd worry about.
BTW, I see your restack and thanks for that. A crosspost is different. It actually goes up as a post on your Substack and goes out directly to all your subscribers. I get that that's a bigger ask so that's totally up to you.
Let's say that, by some miracle, we reach net-zero in 2050. What will global temperatures be in 2100 and 2200?
Good question, and it's answerable from the same published data the figure uses — though these are medians of a large ensemble, so I'll give the spread as well.
Your hypothetical is closest to ScenarioMIP's Very Low pathway, which reaches net-zero greenhouse gases around mid-century and then holds modest net-negative emissions. On the same axes as the figure:
Peak around 2050, near 15.7 °C — about 1.8 above preindustrial. By 2100, about 15.5 °C (+1.6), with a 5–95% range of 1.0 to 2.4. By 2200, about 15.2 °C (+1.3), range 0.8 to 2.1.
CO₂ falls from roughly 445 ppm at mid-century to 405 by 2100 and 365 by 2200. Very Low keeps removing a little carbon past net zero, so concentrations decline rather than hold flat, and temperature follows them down.
The Medium pathway on my figure is about 16.9 °C in 2100 and 17.9 in 2200. What's worth noting is that the ranges don't overlap — the warmest Very Low outcome at 2200 is still cooler than the coolest Medium one.
Today is about 15.3 °C, so Very Low in 2100 sits a little above the present and by 2200 a little below.
I would not read anything into a second decimal place. These are FaIR medians with wide ensembles behind them, and the Earth system model runs that will eventually replace them aren't finished.
That is comforting. We could survive +1.8. We haven't destroyed ourselves yet.
Since our starting point is below the equilibrium temperature, it seems like the temperature should continue rising until equilibrium is restored. The projections show that the temperature is expected to peak and then start to fall even when it is still well below the equilibrium temperature for the anticipated CO2 concentration.
Good question — the answer is that the CO₂ doesn't hold still. Temperature is chasing a target that is itself falling. After net zero the ocean, and then seafloor carbonate, draw carbon dioxide down, so the equilibrium the system is relaxing toward comes down faster than the system can climb toward it. The gap closes by the target descending, not by temperature rising.
Zeke Hausfather's explainer is the best short treatment: https://www.carbonbrief.org/explainer-will-global-warming-stop-as-soon-as-net-zero-emissions-are-reached . He names exactly this confusion — holding CO₂ fixed at today's level is a different world from reaching net zero and letting concentrations fall. One difference of scale: he's describing the first few centuries, where temperature stays roughly flat. The figure runs to tens of millennia, which is where the decline shows up.
It looks like Hausfather’s explainer implies that we are much closer to equilibrium already than what your chart shows. That’s the disconnect. You in fact show the concentration dropping over the next 5-to-10k years, which is why the equilibrium temperature comes down. But it is always higher than the current and predicted temperature. From a simple controls-theory perspective, the temperature has to keep driving towards equilibrium the whole time. I think you and he just have different ideas of where the equilibrium is.
The band isn't my estimate of equilibrium — it's the fitted paleo record, plotted as it comes. Judd, EPICA, Clark. Whether that slope is the right thing to compare a transient against is a fair question, but it isn't a disagreement between Hausfather and me about where equilibrium sits.
Fair enough. I’ll just say that if the paleo record does not track the equilibrium curve, that would bear some explaining.
The paleo record is what I’m using to define the natural equilibrium curve.
Very interesting post Dean. I liked your comment about feedbacks as that was exactly what I was thinking reading it. Looking at the method post too, I think what you have created is a fully believable best case scenario. It would be interesting to see a version with some permafrost melt contribution built in. The ICCI has some basic data in their last state of the cryosphere report.
Maybe add up to a 5Gt annual contribution to the Medium emissions profile ramping up from 0.5 today to 5Gt by 2100, then stable with a slow decline afterwards. The mechanism would slow for permafrost as it melted away but be replaced by wetland inundation as sea levels rose through ice sheet decay. You could stabilise them out at say 10,000 years.
I'd be happy to run some scenarios if you send me the model code.
Thanks again, very thought provoking.
Thanks, Tom. I just sent you an email with the code, data and an introduction.
Hi Dean, another beautiful graph and essay. Your graph clearly shows how far we are and will be out of equilibrium and also suggests the path back to equilibrium. As you mentioned the modeled paths to 2500 do not include the slow climate feedbacks such as glacier and ice sheet melting, change of vegetation and other things. Whereas it is OK to ignore the slow feedbacks during this century they probably can't be ignored over the next 500 years and certainly not over the ten thousand years of your extension.
I think one can get an estimate for the magnitude of the time constant of the slow climate feedbacks from the temperature and CO2 record during the ice ages, particularly the end of the last ice age. The temperature and CO2 concentration tracked each other very well - a fact you used for your equilibrium curve. During the rapid changes at the end of the last ice age the lag between temperature and CO2 was about 500 years or less. I think this could be used as an estimate for the slow feedback time constant.
If true then the path to 2500 will already start to significantly bend towards the equilibrium band and probably reach equilibrium before the end of this millenium. Carbon sequestration from weathering will then mostly happen while the earth is on the equilibrium curve in the same way it always happened.
Unfortunately this implies significantly higher temperatures of 20 C or more. I think this is what Jim Hansen means when he talks about "warming in the pipeline". I understand that it is difficult to include the slow feedbacks in the climate models but that doesn't mean that they can be ingored.
homas — thank you, and this is the right thing to press on.
I think the lag you're pointing at measures something different from what we need. At the end of the last ice age, CO₂ follows temperature — the ocean gives up carbon as it warms. In the glacial cycles CO₂ was a feedback, not a forcing; the forcing was orbital. So that 500 years is the carbon cycle's response time to warming, not the ice sheets' response time to CO₂.
There's a second problem, and it's visible on the figure. The ice-age points sit essentially on the band — scatter about the fit is around 1 °C. That isn't because the slow feedbacks are fast. It's because orbital forcing is slow: precession, obliquity and eccentricity run over 20,000 to 100,000 years, far slower than any feedback in the system, so the ice ages never departed far from equilibrium. You can't read a response time out of a record where the system is always equilibrated. The tight tracking is evidence about the forcing, not the response.
The direct measure is the deglaciation itself. The ice sheets took roughly ten thousand years to go. That's the time constant — millennia, not centuries.
On Hansen, I don't want to wave him away; it's a serious argument. But notice where the premise leads: at Medium's 619 ppm the band sits at 23.5 °C, not 20. If the slow feedbacks equilibrated on 500 years, the answer would be worse than the number you quoted.
Where I come out is that the gap is real and it isn't warming owed to us. It closes from the other side — CO₂ falls before the slow feedbacks can act on it. That's the whole shape of the loop.
Thanks for the quick response. This discussion helps me to develop my thoughts, so here we go:
First a fundamental point that I think is not properly appreciated by many climate scientists: CO2, unlike other GHGs, determines a planet's temperature because the time constant of the carbon cycle is much longer than the time constants of the climate feedbacks, fast or slow. If this were not the case a planet would never warm up long term due to CO2 emissions. CO2 would have a similar effect as methane - only a short term warming.
Your paths on your graph make this explicit: because you (and the climate models) assume that the carbon sequestration is faster than the slow climate feedbacks, you quickly (10k years) end up at the starting temperature. With carbon sequestration faster than climate feedbacks the earth would never have warmed up much. Your beautiful band of climate equilibrium would never have existed.
Then the next question is: what is the slow climate feedback time constant? It is difficult to determine from observations today, which is what the climate models mainly rely on. We have to look at the earth's climate history on the scale of at least tens of thousands of years. The record of temperature and CO2 during the ice ages is a great place to look.
To determine a time constant of a system we need to observe the response to an excitation or forcing as the climate scientists like to call it. The type of forcing doesn't matter, it only has to bring the system out of equilibrium. If the forcing is fast then the system relaxes with its time constant. If the forcing is of similar duration as the time constant then you get a lag. I think this happened at the end of the last ice age. This is why I think the slow climate time constant is about 500 years. Note that this doesn't identify what the time constant is due to. It is just the response of the earth system. Also, the time constant can and likely does depend on the earth's temperature, but we are probably close enough to the ice ages that it should not be very different.
BTW, I do think that it could be much worse than 20 C.
Your first point I think is right and underappreciated — CO₂ is the control knob precisely because its residence time is long compared with the response times of what it drives. Methane is the counterexample that proves it.
On the time constant, I'd point you at Clark et al. 2016 in Nature Climate Change, "Consequences of twenty-first-century policy for multi-millennial climate and sea-level change." Twenty-two authors including Pierrehumbert, and Peter Clark is one of the sources behind my equilibrium band. They put the ice-sheet response squarely in the millennia, with sea level still rising ten thousand years out. Their argument is that the 2100 framing obscures the problem, which is close to what I'm trying to say with the loop.
On where my return starts and what it relaxes toward — you're pushing on the right seam, and it deserves better than a comment box. I'm writing a companion piece on how the model is built: what's imposed, what's calibrated, what falls out, and where it's weakest. The slow-feedback question is going into it. I'll come back to you when it's up.
Thanks for your reply and I am looking forward to your next essay. I am making a few more comments but don't feel you need to reply.
Your graph stimulated me to look into the slow feedback time constant, something I wanted to do for a while. I think CO2 and temperature are a coupled system: when CO2 goes up then the temperature goes up, when the temperature goes up CO2 goes up and also the inverse. So, from a system perspective there is only equilibrium conditions and time constant that define the system. No need to know any details of the system. This makes the record of CO2 and temperature during the ice ages so valuable. The close tracking of the two means strong coupling and short time constant compared to the scale of the ice age duration. From the lag during fast changes the slow climate change time constant is about 500 years.
I looked at the Clark et al paper. It is mostly about sea level rise, which is driven by ice sheet melting. As they are pointing out in the paper sea level rise continues long after the temperature has stabilized during the ice ages. This means the temperature/CO2 time constant is much shorter than the ice sheet melting time. I also noticed that they use an ECS of 3.5 C for their future projection but the ECS of the temperature/CO2 data for the past in figure 1 is 6.2 C even though both past and future are of similar duration. This inconsistency is what you get if you only use models and ignore the information from the climate history, something Jim Hansen first pointed out.