276 | Gavin Schmidt on Measuring, Predicting, and Protecting Our Climate
What this covers
Gavin Schmidt, director of NASA's Goddard Institute for Space Studies, discusses the current state of climate measurement and prediction with Sean Carroll. The conversation centers on three interconnected claims: that human-caused climate change is now robustly measurable and well-explained by decades of modeling; that recent temperatures have unexpectedly exceeded empirical predictions; and that reaching net zero emissions remains humanity's primary lever for stabilizing future warming. Schmidt moves between the physics of greenhouse gases, the history of climate modeling success, and the puzzles posed by 2023–2024 record temperatures breaking records by what he calls "record-breaking margin."
The conversation spans multiple scales of inquiry. Schmidt explains how signal versus noise works in paleoclimate, why attractor in phase space matters more than year-to-year chaos, and how Milutin Milankovic's orbital cycles pace ice ages. He describes the success of 40-year-old climate models at predicting global mean temperature and rainfall patterns, yet contrasts this with the failure to predict the recent extreme warmth—raising the troubling possibility that empirical versus first-principles relationships may no longer hold in a warming regime. The discussion also covers adaptation and resilience alongside mitigation, the practical limits of solar radiation management and the termination shock problem, and how individual agency extends beyond consumer choices into community and political voice. Schmidt does not offer reassurance; his role is to describe what is actually happening.
Schmidt argues that human-caused climate change is now unambiguously measurable and well-understood through decades of robust modeling, yet the unexpectedly extreme warmth of 2023-2024 exposed the limits of empirical prediction, and that reaching net-zero emissions remains the only reliable lever to halt warming.
- Climate models built 40 years ago have successfully predicted global mean temperature change because the climate signal is far more robust to detail than expected.
- The 2023-2024 record temperatures broke records by a record-breaking margin, falling outside the empirical prediction envelope and remaining unexplained.
- Global warming will stabilize when emissions reach net zero, meaning humanity retains agency over future climate.
Human emissions have altered the climate system on a scale comparable to major geological events.
- Human emissions of greenhouse gases (carbon dioxide, methane, CFCs, ozone changes, nitrous oxide) plus deforestation have made humanity a geophysical force whose imprint on the climate system is comparable to the biggest events of the last 65 million years.
“We are now very clearly a geophysical force, and our imprint on the system is comparable to almost the biggest things that have happened in the last 65 million years.”
- The roughly 2.5 watts per square meter of forcing humans have added by changing greenhouse gas levels over the last century vastly exceeds both the geothermal flux (~0.05 W/m²) and societal waste heat (~15 terawatts spread over the planet's surface), so direct heat from cars, cities and volcanoes is negligible compared to the greenhouse-driven energy imbalance.
“there's an energy imbalance. And so you would calculate that and it turns out to be roughly two and a half watts per meter square. That's the forcing that we have put on the system in the last a hundred years or so.”
- Global warming will continue to climb as long as we keep adding carbon dioxide to the atmosphere, and the best estimate is that warming will stop only when we reach net zero emissions; if we reached net zero tomorrow, temperatures would effectively stabilize, meaning future temperature change is governed by future emissions and humanity therefore retains agency.
“our best estimate of when global warming will stop is when we get to net zero.”
- Greenhouse gases are natural and keep the planet much warmer than empty space because triatomic and more complex molecules (water vapor, CO2, methane) have vibrational modes that absorb in the infrared at the frequencies the Earth emits, whereas the dominant nitrogen, oxygen and argon are inert to infrared radiation.
“all of these slightly more complex atoms have vibrational modes that absorb in the infrared. And that's what makes them greenhouse gases, and that's why they are so useful at keeping the earth's planet, the earth's surface much warmer than it otherwise would be.”
- The warming out of the last ice age took about 10,000 years to warm five degrees Celsius, whereas humans have caused 1.5 degrees of warming in just the last hundred years, making the current rate of change dramatically faster than typical geological transitions.
“the warming out of the last ice age took 10,000 years to warm five degrees, and we've changed 1.5 degrees in the last a hundred. So that's quite a bit faster.”
Temperature has shifted detectably across the planet, and paleoclimate data can be reliably calibrated to constrain past climates.
- There is enough data to confirm that temperature distributions have shifted (warming detectable over ~98% of the planet since the 1970s, with seasonal cycles in cities shifting by about a month), but not yet enough data to confidently establish whether the variance of the climate distribution has increased.
“We don't quite have enough data to be able to say that the variance has increased. We have enough data to say that the distributions have shifted.”
- The further back in time you measure, the more uncertain the temperature estimates become, but the signal also gets larger, so the signal-to-noise ratio for past climate events is actually comparable to what we can detect over the last hundred years (e.g., ~10% noise on the 1.5°C modern warming versus ~20% on the 5°C Last Glacial Maximum cooling).
“the further back you go, the more uncertain it gets balance that though, the further back you go, the bigger the signal gets. And so the signal to noise ratio for a lot of stuff that happened in the past is actually pretty comparable to what we can detect even the last a hundred years”
- The ratio of Oxygen-18 to Oxygen-16 in water changes through evaporation and condensation, and because heavier isotopes preferentially rain out as air cools toward the poles, the isotope ratio preserved in ice cores and carbonate shells encodes a temperature-dependent signal that can be calibrated into a paleothermometer.
“if you measure the Oxygen-18 to Oxygen-16 ratio in that snow, that it becomes the ice that goes down through the ice core, when there's more Oxygen-18, it was warmer, and when there was less Oxygen-18, it was cooler.”
- The glacial-interglacial cycles of at least the last 3 million years are paced by Milankovitch orbital cycles (roughly 100,000-year, 40,000-year, and 19,000-year periods) in the Earth's orbit, which Milutin Milanković calculated by hand, and ocean sediment cores from around the world confirmed this once the dating was corrected.
“There's a hundred thousand year cycle, there's a 40 year thousand year cycle, there's a 19,000 year cycle, and put together, those of pace, have been the pacemaker of the glacial periods and the glacial-interglacial variations for at least the last 3 million years.”
- During the Cretaceous, when global temperatures were about eight degrees Celsius warmer than pre-industrial, sea level was roughly a hundred meters higher than today.
“for our listeners, it was about a hundred meters higher than it is today.”
Climate models successfully predict mean temperature and warming patterns but struggle with sea level and tipping points.
- Climate models successfully predicted global mean temperature, the pattern of warming, the rate, and rainfall impacts, but still struggle with sea level projections (which require dynamic ice sheet modeling) and cannot yet predict when rainforest tipping points occur under deforestation and drying.
“We still struggle with getting the sea level projections right, because you need to include the ice sheets... We haven't been able to successfully predict when the tipping points of rainforest systems are”
- Although the climate is enormously complex with many feedbacks, much of the apparent complexity is noise that cancels out, so the underlying climate signal can be extracted relatively easily in models, and that signal turned out to be far more robust to the details than scientists anticipated 40 years ago.
“that signal is in fact, far more robust to the details than we really anticipated 40 years ago.”
- Although temperature is a non-linear three-dimensional field, a global mean temperature is well-defined and meaningful because the heuristic energy-balance model (sun brings energy in, infrared goes out, mediated by surface temperature) works well, and the non-linearity objection turns out to be a small effect.
“The issue is not, can I define a global mean temperature? The issue is, is that meaningful?... that's a valid point, but it turns out that it's a pretty small point.”
- Climate modeling is somewhat ahead of stellar/solar modeling because, although stars require magnetohydrodynamics, it is far harder to get direct information from the middle of the sun than from the middle of the atmosphere, giving Earth scientists an observational advantage for useful predictions.
“it's a lot harder to get direct information from the middle of the sun, than it is to get information from the middle of the atmosphere. So I think we have some advantages on earth”
Empirical relationships derived from past data are breaking down and cannot reliably predict behavior in the emerging new climate regime.
- Empirical correlations derived from the data that is available rather than the data you actually need can lead you astray, as happened with the early ice-core temperature-isotope relationships derived from spatial variations that did not match temporal changes; the 2023 prediction failure raises the concerning possibility that past empirical climate relationships no longer hold in a new climate regime.
“empirical relationships that are derived from, not from data that's available rather than the data that you need can indeed lead you astray.”
- In 2023-2024 global temperatures ran far hotter than predicted, with August through November anomalies breaking records by a record-breaking margin ('record breaking squared'), falling outside the uncertainty bounds of empirical predictions that had worked for a decade, and the cause remains unexplained.
“they were breaking the records where they were breaking the records by a record breaking amount as well. So that's the kind of double... That's record breaking squared”
Climate systems respond to forcings with slow propagation; renewable deployment outpaces expectations but fossil displacement remains incomplete.
- El Nino and La Nina events, a sloshing of warm water along the equatorial Pacific, drive teleconnections that affect weather thousands of miles away (wet periods in the American Southwest, drought and fires in Australia and Indonesia during El Nino, with roughly opposite patterns during La Nina), and these effects propagate with a delay rather than instantly.
“something happening in the Tropical Pacific is gonna affect the weather a long, long way, many thousands of miles away from the Tropical Pacific.”
- Cities are more livable now than 20 years ago, with lower pollution, more walkability, and more pedestrian space, and electric cargo bikes offer a win-win by reducing road pressure and pollution.
“Cities are more livable now than they were 20 years ago. Pollution levels are down, they're more walkable, there's a lot more space for pedestrians”
- The climate problem is not about working out where you are on the chaotic attractor (the year-to-year wiggles), but about where that attractor sits in phase space; almost all of the climate change we have observed is due to the attractor shifting, not uncertainty about our position on it.
“the climate problem is not working out where you are on the attractor. It's where that attractor is in face space.”
- Climate scientists are not motivational speakers obligated to give people hope; their job is to tell people what is happening in the real world, and the satisfaction now comes not from seeing accurate predictions come true but from communicating the science so people act to avoid those predictions being correct.
“I don't think that as scientists we are obligated to give people hope. Like that's not really our job. We're not motivational speakers.”
- Although renewable energy is rolling out faster than expectations (exponentially), coal use and per-capita emissions are falling in places like the UK and US, this growth is not actually displacing much fossil-fuel energy because retired fossil plants are being bought by bitcoin miners and overall energy demand keeps rising.
“The rollout of renewable energy is going very, very well, exponential, much half faster than expectations... But it isn't really displacing very much of the fossil fuel sourced energy.”
- Plant hardiness zones are moving poleward on roughly a decadal basis, such that New York City is now subtropical when it previously was not, affecting what can be planted and which pests appear.
“New York City where I live is now subtropical, it turns out. And it didn't used to be.”
Individual and collective agency operates through multiple channels beyond personal consumption, and mitigation must occur alongside simultaneous adaptation and resilience-building.
- Individual agency on climate is not limited to consumer or commuter choices; people wear many hats (faith communities, PTAs, voters, letter-writers, podcasters) and can multiply their voice far beyond personal consumption, as shown by Greta Thunberg, a schoolgirl with no disposable income who had a massive global effect.
“a Swedish schoolgirl with no disposable income has had a massive global effect.”
- Addressing climate change requires doing all three of building climate resilience, adapting, and mitigating simultaneously, because you cannot adapt to an ever-worsening situation (it must stabilize) and you cannot only mitigate since impacts are already being felt by places not even adapted to today's climate.
“we're gonna have to be building client resilience. We're gonna have to be adapting, we're gonna have to be mitigating, and you have to do all three. You can't adapt to an ever getting worse situation, it has to at some point stabilize.”
- Solar radiation management (injecting sulfates into the stratosphere) would scientifically cool the surface, but it fails as a solution because it must be sustained for hundreds of years while remaining robust to wars, economic downturns and geopolitical crises; if you start and then stop, all the suppressed warming arrives at once, making the consequences catastrophically worse.
“the problem with that as a solution, is that you need to keep doing it... And the answer is hundreds of years.”