Lila Soto: Iris, tell me honestly — did anything about this week's prep actually surprise you, or were you just confirming things you already suspected?
Iris Holm: One thing genuinely stopped me. The CMIP6 spread — 1.8°C to 5.6°C for climate sensitivity. I knew the range was wide. I didn't know it was the widest ever. Wider than any model generation since the 1990s.
Lila Soto: Mm, that's the part that feels almost philosophically uncomfortable. We built better tools and got less certain.
Iris Holm: Which is a very specific kind of failure. Not ignorance — more like revealed complexity. Anyway, that's what we're in today.
Lila Soto: Climate feedbacks. Why warming doesn't stop where the physics says it should. Because here's the place I want to start — the contrast that I think most people haven't sat with. Doubling CO₂ in the atmosphere, just from radiative forcing alone, predicts about 1°C of warming. One degree. Clean, physical, calculable.
Iris Holm: And we're not looking at one degree.
Lila Soto: Three to four times that, in the likely range. And the concept that bridges the two is Equilibrium Climate Sensitivity — ECS. What temperature does the planet actually settle at, over the long run, after that CO₂ doubling? And the reason CMIP6 produces a range from 1.8 to 5.6 instead of a single number is that the answer depends entirely on feedbacks. Positive feedback loops — mechanisms where the initial warming triggers more warming, independent of anything humans do next.
Iris Holm: The gap between 1°C and 3–4°C isn't measurement error. It's the feedbacks.
Lila Soto: Which is — I mean, that reframe matters. Because once you call it a feedback problem rather than an emissions problem, you're asking a different question entirely. Can you switch it off? Or once it's running, is it just running?
Iris Holm: That's the question I want to stress-test. Because the assumption underneath most climate policy is that cutting emissions cuts warming. And if feedbacks are already self-reinforcing — that assumption has a hole in it.
Lila Soto: A pretty significant one. Let's find out how big.
Iris Holm: Here's the simplest way I can put it. Radiative forcing is you turning a thermostat up one notch. Predictable. Physics. Done. A positive feedback is the thermostat — after you've walked out of the room — starting to call for more heat on its own.
Lila Soto: Oh. The house gets warmer without you touching anything.
Iris Holm: And it keeps calling for more because it's warmer. That loop — that's what the 3°C actually is. Not extra emissions. The system feeding itself.
Lila Soto: Okay but — and this is the question I'd want answered if I were just listening — how do we know that's real? That it's not just, I mean, a model artifact? Something the simulations produce but that doesn't actually happen in the atmosphere?
Iris Holm: Right. And the answer isn't the models. It's the ice. Paleoclimate records — ice cores, Cenozoic sediment — they show fast-feedback climate sensitivity pinned at roughly 3 plus or minus 1 degree Celsius per 4 watts per square meter of CO₂ forcing. That's not a simulation. That's Earth's own receipts from past climate states.
Lila Soto: James Hansen's work, yeah — the Cenozoic CO₂ and sea-level co-variations. The thing that struck me reading it is that the feedbacks show up every single time. Consistently amplifying. Not occasionally, not sometimes.
Lila Soto: So the models aren't inventing the feedback signal. They're — kind of trying to catch up to something the rock record already confirmed.
Iris Holm: Which reframes CMIP6 somewhat. The 1.8 to 5.6 spread isn't uncertainty about whether feedbacks exist. That part's settled. It's uncertainty about which feedbacks dominate — cloud behavior, ice-albedo timing, ocean stratification. The amplification is confirmed. The magnitude is the open question.
Lila Soto: And the near-term number — Transient Climate Response — that already bakes in some of this, right? Even before you get to the long-run equilibrium?
Iris Holm: TCR averages around 1.7°C across CMIP6 models. That's at the moment of CO₂ doubling — not the final settled temperature, which is ECS. But even 1.7 is already feedback-amplified. The raw forcing would be lower. So the feedback signal is in the near-term number too, not just the long-run one.
Lila Soto: So there's no version of this story where the feedbacks aren't already in the math. It's not a future risk layered on top. It's — mm — it's the number we're already living inside.
Iris Holm: And living inside it is the part that gets concrete fast. Because now we're talking about three specific mechanisms — not the math, the actual physical systems — that are already amplifying.
Lila Soto: I want to start with a scene. There's a permafrost researcher in Siberia — June morning, pulling core samples from ground that, in 1995, was frozen solid at this time of year. Now it's thawing six weeks earlier. She's logging depth, date, carbon content. No drama. Tuesday is Tuesday.
Iris Holm: What's in the log, though.
Lila Soto: That's the thing. What's in the log is the leading edge of 1,400 to 1,700 billion tonnes of carbon. Stored in permafrost — which covers about 22% of Northern Hemisphere land. And that number — I mean, say it again — that's two to three times what's currently in the atmosphere.
Iris Holm: Two to three times. That's not a reserve. That's a second atmosphere, frozen.
Lila Soto: And when it thaws, microbial decomposition kicks in — bacteria essentially eating the ancient organic matter — releasing CO₂ and methane. Both. The methane part is where the near-term picture gets — yeah, actually alarming. EPA puts methane's global warming potential at 28 to 36 times CO₂ over a hundred years. But over twenty years? Substantially higher.
Iris Holm: Which means the decade we're in is disproportionately exposed.
Lila Soto: Exactly what I mean — and that framing has consequences we'll get to, the net-zero timeline question and what the choice of methane's timescale is actually hiding about the near-term window.
Iris Holm: But first — what makes permafrost self-sustaining once it starts? Because that's the feedback claim.
Lila Soto: Right. And there's research that pins this linearly — permafrost carbon release at 11 to 21 petagrams of carbon per 100 degree-years of warming overshoot. Linear. It doesn't plateau. And here's the counterintuitive piece: AMOC — the Atlantic Meridional Overturning Circulation — its slowdown actually cools the Northern Hemisphere regionally. You'd think that buys time for permafrost. It doesn't. That same linear relationship holds even under AMOC-induced cooling.
Iris Holm: Hold on. So two systems — AMOC slowing, Arctic warming — compound each other's carbon effect even when they're pulling temperature in opposite directions?
Lila Soto: Yeah. Cooling somewhere does not buy time. They're not canceling — they're stacking. And then there are the other two mechanisms. Ocean carbon sink weakening — warmer water holds less dissolved CO₂, reduced solubility — ScienceBrief synthesized over 130 peer-reviewed studies and landed on 10 to 70 additional parts per million of atmospheric CO₂ per degree of warming. From oceans alone just absorbing less. And then ice-albedo, which is actually the simplest — no chemistry required. Ice melts, darker ocean surface underneath absorbs more solar radiation, warms faster, melts more ice.
Iris Holm: That one doesn't even need a gas. It's just physics eating itself.
Lila Soto: And that researcher in Siberia — she's logging a data point inside all three simultaneously. The carbon coming up in her core sample. The AMOC running in the background. The pond forming next to where she parked her equipment, sitting on what used to be permafrost. All of it already in motion.
Iris Holm: All three already in motion — and that's actually where the clean version of the net-zero story starts to crack. Because the assumption underneath cutting emissions is that you're cutting warming. But these feedbacks aren't triggered by what we're currently emitting. They're triggered by temperature. Temperature that already exists.
Lila Soto: Oh. So the tap metaphor breaks down.
Iris Holm: Net zero closes the tap. It does not reach into the permafrost and refreeze it.
Lila Soto: I mean — okay, but doesn't ECS give us a ceiling? Like, if sensitivity lands at the lower end of that CMIP6 range, isn't there a level where things stabilize?
Iris Holm: ECS is — look, ECS is complicated by something called state-dependence of CO₂ forcing. The instantaneous radiative forcing from a CO₂ doubling isn't fixed. It increases roughly 25% for every additional doubling. And it's already risen about 10% since pre-industrial times, from stratospheric cooling. So the sensitivity number itself is shifting as we go.
Lila Soto: Wait — the ceiling is moving?
Iris Holm: The ceiling is not a fixed address. And part of why CMIP6 spans 1.8 to 5.6 is that cloud feedbacks and that state-dependence are doing as much work as the carbon cycle feedbacks. It's not just permafrost methane widening the range. It's clouds behaving differently at different base states. Whether carbon cycle feedbacks are even fully captured inside that spread — that's still debated.
Lila Soto: So some models are already embedding them, some aren't, and we don't fully know which.
Iris Holm: Right. Which brings me to the methane timescale — because this is where policy is making a choice it isn't advertising. The EPA's 28 to 36 times figure is over a hundred years. But methane decays faster than CO₂. Its near-term punch — the twenty-year window — is substantially higher than that number. And policies built around 2100 targets are using the hundred-year figure.
Lila Soto: Mm. So we're systematically underweighting the decade we're in.
Iris Holm: It's not a neutral communication choice. The timescale you pick determines how permafrost methane gets weighted in carbon budgets, in overshoot policy, in emissions inventories. Pick the century frame, the near-term signal shrinks. Pick the twenty-year frame, this decade looks categorically different.
Lila Soto: And overshoot — I mean, the 1.5°C overshoot scenario. That's now considered unavoidable. So if we blow past 1.5 and then pull back down — does the permafrost just come along for the ride back?
Iris Holm: That's the question that doesn't have a reassuring answer. Permafrost carbon loss after overshoot may be irreversible. Temperatures nominally return to target — the carbon doesn't go back into the ground. The process that released it is done. And no runaway feedback is expected inside this century, to be clear — bounded, not benign — but the permafrost piece specifically looks like a one-way door.
Lila Soto: So that researcher in Siberia — she's not just logging the warming. She's logging something that doesn't unlog. The date gets earlier every June, and there's no policy path that moves it back.
Iris Holm: Which shifts the question. Not whether feedbacks are running — they are. But what 'success' even means inside that. Because it used to mean: cut emissions, stop warming. Now I think it means something more like — watch the systems. Intervene if you can. The permafrost, the ocean chemistry, the ice dynamics. Monitor them in real time, because they're running on their own clock now regardless of what happens at a policy level.
Lila Soto: Yeah — and that's a different posture than what most people think climate action is. It's not just emissions accounting anymore. It's — I mean, what are we actually doing to watch those systems? That's maybe the most consequential climate question right now. Not the next COP target. What instrumentation exists for permafrost carbon flux, for ocean sink capacity, in real time.
Iris Holm: The bounded piece matters though. No runaway feedback anticipated this century — that's not nothing. The ScienceBrief synthesis, 130-plus studies, puts the carbon cycle contribution at substantial but capped. Not a cliff. But self-sustaining across any overshoot pathway.
Lila Soto: Bounded, not benign. That's the phrase that keeps coming back to me. And — actually, I want to come back to where we started. You said the CMIP6 spread genuinely stopped you. 1.8 to 5.6. Widest ever. At the top of this, that felt like a measurement problem. Now it feels like — I don't know. Like the spread is describing something real about the systems themselves.
Iris Holm: It's not measurement error dressed up as uncertainty. It's cloud feedbacks, state-dependence of forcing, feedbacks stacking in ways we can't fully isolate. The range is the honest answer. That's — yeah, that landed differently by the end.
Lila Soto: And that researcher is still logging. June comes, the depth changes, the date is earlier again. She's not waiting for the range to narrow. I think that's where I'll actually stop — not on a big frame, just on that. She's doing the most important measurement work on the planet on a Tuesday morning.
