Note: A version of this piece, which I penned, originally ran on Climate Capital’s substack as part of the Resilient Systems Capital launch (a new $4 million vehicle focused on resilience x climate tech businesses). See here.

It’s been another summer of heatwaves and wildfires across North America and Europe, as well as devastating floods and droughts worldwide, from the U.K. to Mongolia. Water levels in the Danube and the Rhein rivers recently got so low that it disrupted transportation and nuclear reactor operation (the same has happened to the Rhein in several recent years, and the river is actually still hitting new all-time lows as of this writing on Tuesday the 29th, even though I started this article draft at the beginning of the month). Global ocean temperature records are being set almost continuously (which is predominantly, in the short-term, a function of the developing and staggering El Niño). With said El Niño growing and increasingly posed to reach unprecedented levels of strength, the next year or so will likely occassion more disruption and disaster. Huzzah. The brightspot in that, if there is one, is that it will also offer us valuable insight into what a future with >2° C warming may look like, given that’s a future we’re almost certainly headed for in coming decades.

More recently and in acutely devastating fashion, a glacial collapse led to an astonishingly rapid flash flood that killed thousands in Nepal, with many still missing and untold destruction otherwise (the economic toll will exceed $2B and could rise towards ~$5B+). Either way, those are costs Nepal’s economy alone cannot shoulder, and loss and damage mechanisms by which rich countries are supposed to pay for the damages that less developed countries cannot themselves bear, exacerbated by climate change to which they contributed little, are almost entirely dsyfunctional (Nepal is not getting much help, certainly not anything near the level of help it should).

To be sure, this single event was not directly attributable to climate change, per se. Single event attribution is always perniciously hard and divisive, even if the science is improving; many climate practitioners and journalists in particular are too quick and/or imprecise in ascribing climate change as a “cause” of negative and damaging extreme weather events as they happen. It’s typically not that simple, and, to let my inner pedant run while, climate change, as an observation of changing averages, cannot technically “cause” anything. It can, over time, reflect the changing likelihoods thereof.

Still, we do know that climate change does include changes in conditions that make disasters like the one witnessed recently in Nepal more likely. A warming climate has, in many places, made mountainside less stable, particularly where there is permafrost melt, considering that permafrost helps bind together rock and ice. Climate change also tracks increased rates of water infiltration dynamics in mountainsides (and ice sheets), which destabilizes rocks (this pattern roughly holds for other glaciers, too; where ocean water ‘intrudes’ underneath glaciers, it can accelerate their melt and destabilization).

In perplexing and sad irony, some of the best research on all these fronts comes out of Nepal. The International Centre for Integrated Mountain Development (ICIMOD) released its latest glacier assessment on August 12th. The report specifically warns of “particularly rapid losses in the eastern and central HKH.” Central HKH is an area that covers the specific locations where this disaster occurred. On August 26th, exactly two weeks after the report was released. Clearly, governments should work more closely with scientists and pay attention to their work, and/or we need more intermediaries doing decision-relevant translation of science into information for policymakers (and other stakeholders, for that matter). These are field-building and coordination needs, which, when ignored, can prove as costly as other lags and snags in science and R&D, even if the only effective adaptation strategies at present amount to relocation at present.

OK. Pause. That was a lot already, and quite the sobering opening. I promise this won’t all be doom and gloom. That said, collectively, as a species, we must occasionally reckon, with full attention and awareness, with the full scope of the challenges climate change poses. This occasionally requires being present to devastation (emphasis on occasionally; I don’t recommend doing it all the time and letting it take over your life. Been there, and while my results are based on a sample size of 1, they weren’t positive).

I digress. Addressing climate change will also increasingly require:

The confluence of extreme weather and climate change-related damages this year (among other factors) has brought climate change back into mainstream discourse to a degree (pun number one, sorry). We’re still far from the heyday of, say, 2021, at least in terms of the enthusiasm that existed then for private sector climate tech, public policy for the energy transition in the U.S., and attention paid to climate conversations generally. The policy environment in the U.S. is certainly still, shall we say, ‘different’ today than it was five years ago. That’s, of course, my tongue-in-cheek understatement of the year. Even the suggestion, let alone legislation, that the U.S. should stop exports of liquefied natural gas, which were taken very seriously during the Biden admin, are at present unimaginable). Plus, most of the attention that does still flow to “climate”-related or adjacent things is monopolized by topics, startups focused on, or massive fundraising deals surrounding cleaner energy when it’s useful for AI and data centers. And that’s fine!

But the confluence of the catastrophic physical impacts and changes across the Earth systems, as well as their downstream impacts on human systems, has also started to attune more people to the idea that there are larger shifts afoot. People are starting to appreciate that the speed, scope, scale, severity, and “shape” of climate change isn’t the same as it was in, say, 2015. It’s getting gnarlier, realer, more visceral, more ‘now,’ not to mention more complicated and challenging. All this is to say that climate change itself is changing, and not in our favor, so to speak. There’s evidence for this, which we’ll explore here shortly. But first, again, an apology for exposing you to what is a stark fact to reckon with, especially in an atmosphere (pun number two, sorry) where there’s plenty of other stuff going on that’s worth worrying about).

We’ll now walk through a few examples and case studies of how and why Earth systems (read as the planet’s largest physical systems, like ocean circulations), are changing and why that matters as much if not more than greenhouse gas emissions alone do:

Global warming is accelerating. That doesn’t just mean the Earth is continuing to warm. It means the rate at which the Earth is warming is rising. Specifically, the world got ~0.35° C warmer between 2015 and 2025, almost double the average decadal warming rate between 1970 to 2015, which was closer to ~0.20° C.

This is a function of multiple drivers. Global warming and the climate change it causes isn’t just driven by greenhouse gasses (“GHGs”) and their greenhouse effect. Of course, GHGs are the prominent drivers of warming since preindustrial times, and the fact that global emissions of most of them remain at all-time highs is a contributing factor in accelerating warming. But factors beyond GHGs are also significant.

To break this down, let’s introduce the concept of Earth’s Energy Imbalance (“EEI”), a term that refers to the difference between how much radiation Earth, in aggregate, absorbs versus how much it reflects back to space. When EEI is positive, it means the Earth is absorbing more energy than it is reflecting back to space on balance, and the planet accumulates heat as a result. When it’s negative, as has been true many times in the geologic record, including as recently as within the past thousand years, the Earth loses heat on balance. If EEI is close to zero (exhibiting little imbalance one way or another), as has also often been true, the climate trends, more or less, towards stability.

Right now, EEI is at an all-time high for the time periods in which it’s been measurable with reasonable accuracy. And the reason why isn’t just GHGs.

Beyond GHGs, a major contributor to EEI is Earth’s reflectivity. A common example (and a relatively straightforward and appreciable one, hence why I’ll use it here) focuses on Arctic sea ice. Sea ice cover at the end of summer in the Arctic has shrunk, on average, by ~12% per decade relative between 1979 and 2024 (relative to a 1981–2010 average), largely due to global warming (and the fact that the Arctic is warming 4x faster on average than global averages, as I’ve often noted at this point I feel like).

Ice and snow are among the most reflective natural surfaces on Earth; they reflect a lot of energy from the sun back into space. This keeps the planet cooler, offsetting some of the warming impact of the GHGs accumulating in the atmosphere. But, as the Earth warms, Arctic sea ice (and ice elsehwere) melts, giving way to open ocean. The ocean is not particularly reflective, especially not relative to the sea ice it once was in this specific case.

Hence, as sea ice melts, the Earth loses reflectivity. This increases EEI, which maps pretty directly to more global warming. This example is also useful because it’s emblematic of a reinforcing feedback. The higher the positive EEI, the more the planet warms. The more the planet warms, the more, in general, sea ice, ice sheets, glaciers, and snow worldwide melt, reducing reflectivity, further increasing EEI. There are plenty of other dynamics involved; that causal chain I just drew was an oversimplification. But the trend definitely holds on the whole, as we’ll discuss next.

Non-GHG factors are becoming more significant drivers of change. Here’s the real rub. Since 2001, observations from NASA’s CERES instruments document a sustained increase in absorbed radiation, indicative of declining reflectivity. Specifically, data from NASA’s CERES instruments show Earth’s average “reflectance” fell from ~29.3% to ~28.7% between 2003 and the end of 2025, possibly falling more rapidly in recent years.

That may seem like a diminutive reduction in reflectivity. But there’s a massive amount of inherent leverage to it. Reflectivity, while not discussed all that much, at the level of the entire Earth system, wields massive influence on EEI. The warming that Earth’s declining reflectivity has driven has, alongside all-time high GHG emissions, of course, led to a doubling (or even a tripling by some estimates) in EEI since the turn of the century (another scientific source can be found here), potentially becoming a more salient driver of increased EEI than GHGs over the time period (read more on that here).

The main reflectivity losses aren’t just in snow and sea ice, either. Those have certainly declined and contributed. But in total, ice, glacial, and snow losses account for ‘only’ ~27% of lost reflectivity since 2001. The other ~73% is roughly attributable to a combination of diminishing cloud cover and losses of atmospheric aerosols. These two categories also directly influence and interact with each other, which complicates disambiguation of which is driving more or less, how, and under what conditions.

Suffice to say, reflectivity matters, and some things that get very little attention in conventional climate contexts (like clouds!) actually have a lot of leverage over the Earth system. For another example of underappreciated drivers, I used to write extensively about methane and other GHGs beyond CO2. Those are also playing a role in the changing nature of climate change; warming induced emissions and feedback loops, i.e., potentially self-reinforcing changes in natural systems that release GHGs, are finally getting more press (and, more importantly, dedicated research programs).

The point of setting the stage, as we have thus far, with an introduction to EEI was to underscore how it’s come to pass that climate change is not only firmly in an “overshoot” era, i.e., one in which almost everyone, even the laggards, acknowledge that warming will exceed 1.5°C and even 2°C, and likely for a sustained period of time (central estimates are now tracking towards ~2.9°C under the new “medium” scenario [CMIP7]).

The point is that climate change is accelerating and entering a more “nonlinear” era due to increasing dislocations driven by the other components, like reflectivity, that impact EEI.

I’m far from the first to make these points. Kelly Erhardt, who works at Outlier, a stellar climate-focused non-profit that funds a lot of niche work to address underserved and overlooked yet high-leverage solutions (comparable to what my employer, ARC tries to do), said it succinctly in a recent interview. Plenty of other smart people and organizations are also sounding the alarm about ‘nonlinear’ climate change, including the likes of JP Morgan, which became the first major bank to address climate tipping point risks directly. Sarah Kapnick, JPMorgan’s global head of climate advisory, noted explicitly that the world may be moving out of a “linear” phase of climate change and…

Perhaps the most valuable addition I personally have to add is that the Earth system and climate, let alone climate change, aren’t particularly linear systems or system changes to begin with. There are more dynamics in the Earth system that aren’t linear than ones that are. Look no further than the relationship between CO₂ molecules in the atmosphere and their warming impact on the Earth. That relationship is not even linear (it’s logarithmic)!

Appreciating nonlinearity as a baseline of the Earth system also encourages us to think more comprehensively about what climate change is, how it can change (and surprise us and our modeling efforts, regardless of how computationally powerful), and what the implications thereof are for how to respond to it. Here’s part of the answer: We need to do a lot more of the conventional things we already do. Then, we also need to consider doing other, fundamentally different and new things as well. The current toolkit of mitigation, adaptation, and carbon removal alone may not suffice.

These are urgent lines of inquiry. We’re increasingly seeing the impacts of nonlinearity all over the planet, which ties back into the extremes we started this piece exploring. It also applies overall, at the level of global climate models. Some models, including prominent ones upon which many important decisions are based (and trillions in capital and policymaking decisions are made), did not predict the aforementioned declines in reflectivity well. This likely stems from the fact that there’s a lot we don’t know about cloud microphysics and aerosol-cloud interactions. There’s a lot we don’t know about aerosols in general (the error bars on their aggregate warming impact in IPCC reports are massive, as visualized below).

Setting aside those questions and uncertainties that are, for now, unanswerable and irreducible, such as why exactly many climate models have been very “off” when it comes things like reflectivity declines, the presence and magnitude of the divergence in them is problematic in itself as it suggests many models may underestimate climate sensitivity, a measure of how much Earth’s surface will warm in response to a doubling of atmospheric carbon dioxide concentrations. If climate sensitivity is higher than we thought and think, we may be underestimating future warming. Potentially by a lot!

The unknowns go on. Here’s another rabbit hole. Take the risks facing specific Earth systems. As defined by Global Tipping Points, Earth System Tipping Points (ESTPs) occur and/or are defined as:

The definition continues with a lot more detail in the 2025 Global Tipping Points report. While tombs could (and have and will) be penned on these risks, two more things I’ll emphasize are that ESTPs often exhibit hysteresis, i.e., the risk of irreversibility, and can set off “cascade[s] of accelerating and compounding damage…” (Wunderling et al., 2024).

ESTP risks are, similarly to extreme climate and weather events this year, also all over the place. Boreal forests. Coral reefs. The Amazon. Many ice sheets, both in the Arctic and Antarctica. Oceanic and atmospheric circulations (most notably, at least in the media of late, AMOC).

Let’s take a non-AMOC example. The Greenland Ice Sheet (GrIS) almost certainly faces tipping point risk; the 2023 and 2025 Global Tipping Points reports attribute high-confidence to this, drawing on evidence from ice-sheet models and paleorecords. This means the ice sheet, which is massive, could enter an irreversible phase of destabilization if certain critical global thresholds are passed. In the GrIS’s case, the threshold is likely thermal (i.e., related to warming). And the best estimate (1.5°C) of the potential tipping threshold is already close at hand; it’s one the world is likely to cross in the next decade (per UNEP, and on a trailing average basis, not just for one year [that will happen next year more likely than not]). Further, as if that all weren’t enough, the GrIS has all kinds of its own nonlinearities (which is, again, a common theme for ESTPs as well as an overarching dynamic of which they’re themselves emblematic).

So, the GrIS’s irreversible transition could start within decades. Or years. It could even already be underway. For reference, the GrIS holds 7+ meters of potential sea level rise. How long it would take and to what extent it would be ‘complete’ destabilization (a scenario where the GrIS collapses entirely), is all highly uncertain (another common theme of a lot of this science). Modeling efforts here are limited for all kinds of reasons, starting with the absence of observational data from historical precedents within human societal timelines. Machine learning and AI will help, as is true in many cases. But they won’t solve anything comprehensively. One saving grace is that a collapse would likely take a while (centuries or millenia). That’s not all that comforting, unless you only care about yourself, though. Plus, GrIS mass loss, once underway, may be (again) highly nonlinear, frontloaded, and the melt timelines are highly uncertain as well (they could be shorter).

My point is not to harp on the challenges ahead or to veer off into despair.

It’s to underscore that as climate change evolves, so to must our already insufficient efforts to address it. First, we need to reframe leftover, erroneous mental modes that position anything related to climate change, whether its drivers, its mechanics, or its consequences, as linear. In some select cases, maybe linearity is directionally accurate-ish. That’ll be the exception, not the rule. Then, we need to reconsider what is and may be required to respond to this evolving, and highly uncertain, climate change.

There’s also, as noted many times above, tremendous uncertainty about a lot of the specifics and the overall frame of how the next few decades, let alone centuries, will unfold. Hence, all this reimagining of climate action also requires some comfort operating within deep, and sometimes entirely irreducible, uncertainty. Easier said than done.

Returning to the ESTPs examples (and starting to bring things full circle), ESTPs are emblematic of this challenge (and many of the others at hand). The risks they entail, both across the Earth system and the human systems that depend on its stability, are very much a product of a more general phenomenon. Emergent behavior—i.e., stuff there’s little to no observational data for, that surprises us, and that deviates from expectations—often creates chaos, especially when it enters into connected and complex systems. ESTPs, and other similar risks deviate from what the existing edifice of climate response efforts is set up to contend with. That’s why there’s misalignment between what we’re comfortable doing today re: climate change and what we may need to build comfort doing soon.

To be sure, I’m not trying to pull a fast one by alluding to climate interventions here at the end, by platforming them alongside the established pillars of climate action. Mitigation, adaptation, and carbon removal all warrant and really still necessitate orders of magnitude more investment, regardless what happens with interventions. But, as I’ll keep shouting until my face turns blue or the first ice sheets float, we’re gonna need to at least do the scientific and R&D leg-work for interventions. Because, unfortunately, if you ask me, it’s more likely than not that the scenarios will come to pass in which we will want to at least have them as back-up options. And it’s much preferred we can then evaluate and make decisions about their deployment if they’ve been (i) de-risked to a degree, (ii) made somewhat governable, and (iii) sustained by consistent legitimizing processes with relevant rights holders and communities. Otherwise, they may get deployed illegitimetly / irresponsibly, which is its whole own problematic scenario analysis we can skip for now.

Perpetuating a scientific vacuum (i.e., unilaterally opposing even small-scale field trials for things like solar radiation management [“SRM”] because of moral hazard concerns) creates more risk than it reduces. For one, as is often the case, the private sector may ‘do’ it anyway. To be sure, that someone else will do something anyway isn’t a cure all for moral challenges. I don’t accept that we should develop LLMs further, for example, if they pose even a 1% chance of extincting our species, just because China will do it even if the U.S. doesn’t.

A better moral argument (because we need not always revert to utilitarian ones) runs through assessing what the costs of failing to, at a minimum, assess additional intervention options, like SRM, might entail. One useful measure we can lean on here are estimates from the the UNFCCC for what’s called “losses and damages,” i.e., harms beyond what people and systems can adapt to. In other’s words:

One estimate of losses and damages from 2019 already ran up to around $400B a year between then and 2030 for these types of ‘unavoidable’ losses. Updated numbers, incorporating inflation alone, are probably closer to trillions of dollars annually, especially if one were to start looking out to the 2040s, 50s, and beyond. Here’s another problem to add to the list; not many, if any, people are really even running such analyses, at least not publicly. Moreover, the numbers we do have are purely economic; they ignore things like, say, human deaths. Hm. Pause. Accepting millions of deaths and billions of damages, annually, as unavoidable, even if they may indeed be avoidable—provided we allow scientists and researchers to assess various potential interventions that may be credibly capable of reducing some or many of these damages thought otherwise to be unavoidable, all while assuming they must clear a very high bar in terms of responsibility, governance, oversight, license, and legitimacy, etc, strikes me as a larger moral hazard than the risk that interventions are developed and then irresponsibly deployed.

What do the interventions I speak of look like? For this, I’m tapping in my good friend Paul Gambill, who laid out examples (not exhaustively) in a great newsletter of his own on Tuesday (P.S., definitely also subscribe to Paul’s Substack if this stuff is your jam):

So, what’s the TL;DR? Well, give my employer, ARC, money! In all seriousness, the point of this is to begin re-imagining what responding to climate change will require. If you’ve read this newsletter long enough, you already know it’s gonna require far more than EVs and solar panels. It can’t be solved by just doing SRM or SAI, either, as some people in Silicon Valley increasingly seem to think for some reason (the age old silver bullet does have its allure, I get it). Probably, it will require far more than society has collectively yet to imagine. For now, I’ll leave it here, and will punt the task of more imagining to my future, less tired self, as well as to you all, and to all of us in communion with one another. If anything is for certain at all, it’s that we won’t figure any of this stuff out on our lonesome.