Carbon Capture and Removal

Core Concepts

Two different problems, two different tools

When people talk about "carbon capture," they often conflate two distinct strategies that address different parts of the climate problem.

Point-source capture intercepts CO2 at the exhaust of a power plant or industrial facility — before it enters the atmosphere. It prevents new emissions. Carbon dioxide removal (CDR) extracts CO2 that is already in the atmosphere and stores it somewhere else. It creates negative emissions.

As Carbon Brief explains, the climate system does not treat these symmetrically: the carbon cycle's response to CO2 emissions is not equal and opposite to its response to removals of the same magnitude. Balancing a tonne of emission with a tonne of removal does not simply zero out the climate impact.

Both are necessary. Avoidance and point-source capture reduce future emissions; removal addresses accumulated past emissions. For net-zero targets — where residual hard-to-abate emissions must be offset — IPCC frameworks treat removal as essential.

What "negative emissions" actually requires

For a CDR activity to count as genuinely removing carbon, the entire process must remove more CO2 from the atmosphere than it generates. This life-cycle framing matters enormously: it is why a BECCS plant that burns wood from recently cleared forest may not actually produce negative emissions, and why DAC powered by a coal grid could be releasing more CO2 than it captures.

Permanence: how long does the carbon stay gone?

Not all carbon storage is equal. The key question is permanence: how long will the captured carbon remain out of the atmosphere?

• Geological storage (injecting CO2 into rock formations) offers the longest timescales. Frontiers research shows natural reservoir leakage rates on the order of 10⁻⁷ per year — extraordinarily slow. Depleted gas wells are less reliable (10⁻⁴ to 10⁻⁶ per year), and defining absolute permanence at geological timescales remains scientifically challenging.
• Forest and soil carbon is biologically active and reversible. Wildfire, drought, pest outbreaks, or land-use change can release stored carbon in a season. The carbon is gone from the tree; it does not stay gone from the climate.
• Biochar occupies a middle ground: its aromatic ring structure resists microbial breakdown, so carbon can persist for centuries to thousands of years — but it is not as permanent as geological storage.

This permanence gradient is also a cost gradient. A comparative analysis in Nature Communications frames it plainly: nature-based solutions cost €40–200/tCO2 but are vulnerable to reversal; engineered solutions with geological storage cost $150–700/tCO2 but offer durability across geological timescales. The portfolio question is how to combine near-term affordable options with the long-term durable ones.

What IPCC scenarios actually require

The headline finding from IPCC's Special Report on 1.5°C is stark: every pathway limiting warming to 1.5°C with limited or no overshoot requires CDR at unprecedented scale. The IPCC projects cumulative CDR use of up to 1,000 gigatonnes of CO2 over the 21st century — with BECCS contributing a median 334 Gt and DAC a median 30 Gt.

To put that in perspective: 1,000 Gt is roughly 25 years of current global annual emissions. Conservative estimates require 4.2 Gt of novel CDR per year by 2050; more demanding analyses put it above 10 Gt per year. Current global deployment is measured in megatonnes — three orders of magnitude smaller.

The IEA's Net Zero Emissions scenario calls for DAC alone to capture 85+ million tonnes per year by 2030 and ~980 million tonnes per year by 2050. If every planned project were built at full capacity tomorrow, they would capture around 3 million tonnes by 2030 — less than 5% of what the NZE pathway requires.

Compare & Contrast

A taxonomy of CDR approaches

It helps to see the landscape as a whole before going deeper into any one approach. The table below organizes the major pathways by cost, permanence, scalability, and readiness — using evidence from the claims in this module.

The critical distinction that cuts across the whole table is between reversible biological storage (forests, soils) and durable geological or mineral storage (DAC with CCS, enhanced weathering, OAE). The former is cheaper and available now; the latter is permanent but expensive and not yet at scale. A credible climate strategy needs both.

Worked Example

What it takes for DAC to actually remove carbon: the energy problem

Direct air capture is often presented as a silver bullet — a machine that can just suck CO2 out of the sky. The physics makes it harder than that.

Ambient air contains about 400 parts per million of CO2. Industrial flue gas contains 5–15%. The low concentration in air means DAC requires significantly higher energy inputs and costs compared to point-source capture. You are fishing for CO2 in an extremely dilute solution.

The two main DAC technology families handle the regeneration step differently. Solid sorbent systems release captured CO2 by heating the sorbent to 80–120°C — temperatures compatible with waste heat or low-cost renewables. Liquid solvent systems based on strong bases require 300–900°C — temperatures that often require burning fossil fuels, which partially cancels the benefit. National Academies and WRI both document this difference.

The energy source is not a detail — it is the mechanism. Life cycle assessment data shows that wind-powered DAC achieves 78% net carbon removal efficiency, while grid-powered DAC with fossil electricity achieves only 3.6%. A facility drawing from a coal grid is not removing carbon; it is rearranging it.

Now consider what scaling requires. The IEA's NZE scenario targets ~980 MtCO2/yr of DAC by 2050. Scaling barriers research puts DAC's energy demand at 300–500 MW per million tonnes of CO2 removed. At 980 Mt/yr, that is 294,000–490,000 MW of dedicated clean power — roughly the entire current electricity generation capacity of the United States, used solely to run air capture machines.

None of this means DAC should not be built. Learning curves are real: first-of-a-kind (FOAK) projects cost $350–700/tonne; nth-of-a-kind (NOAK) at megaton scale can reach $150–230/tonne. But learning requires building. Disclosed order prices for durable CDR fell from ~$490/tonne in 2023 to ~$320/tonne in 2024 as market volume increased — a real signal that deployment drives cost reduction. The question is whether the world will build enough, fast enough.

Common Misconceptions

"Planting trees is a reliable carbon sink"

Forests are real carbon stores, but forest carbon offsets have a permanence problem that the market has significantly underpriced.

Research on U.S. forest offsets found that 26% of existing forest carbon offset credits face wildfire hazard, and annual forest fire acreage is projected to quadruple by end of century under moderate emissions scenarios. The same study found that total carbon stock exposed to reversal liability is approximately three times greater than the total credited carbon — meaning the buffer pools (the financial insurance against reversals) are severely undercapitalized relative to realistic risk.

This is compounded by the additionality problem. Research on U.S. forest offset markets suggests that the majority of credits — particularly those based on unrealistic baselines — provide no real offset to greenhouse gas emissions. The fundamental question — "would this carbon have been sequestered anyway without the offset payment?" — is often answered optimistically by project developers.

The point is not that trees are useless for the climate — they are not. Protecting young secondary forests (20–40 years old) can provide up to 8x more carbon removal per hectare than planting new trees, and forest-based CDR accounts for roughly 10% of the near-term reductions in 1.5°C scenarios. The point is that a forest offset credit is not equivalent to a tonne of CO2 permanently stored underground.

"BECCS is a clean bridge to net zero"

BECCS — burning biomass for energy and capturing the resulting CO2 — appears in nearly every 1.5°C scenario model. The IPCC projects a median BECCS contribution of 334 Gt over the century, with scenario requirements ranging from 1 GtCO2/yr in 2030 to 16 GtCO2/yr in 2100.

The gap between model and reality is severe. Current global BECCS deployment captures roughly 2 million metric tonnes of biogenic CO2 per year, with less than 1 million tonnes permanently stored — against IPCC scenario requirements of 500–5,000 million tonnes per year by 2050.

Beyond the deployment gap, BECCS has structural constraints the models often abstract away:

• Land competition: IPCC assessments project 56–482 million hectares of additional cropland would be required for BECCS biomass at 1.5°C scale — land that competes directly with food production and biodiversity.
• Carbon debt from land conversion: Converting existing land to biomass can take decades to repay the initial carbon debt, and in high-carbon-stock conversions (forests, peatlands), the debt may never be repaid.
• Energy penalty: Real-world data from the Drax plant shows that post-combustion CCS reduces facility efficiency from 36.2% to 20.9% — a 170 MW parasitic load on a 630 MW turbine.

BECCS may yet play a role in climate mitigation, but it is not a ready-made solution. It is a set of contested bets about future land availability, energy systems, and carbon accounting.

"CDR gives us more time to cut emissions"

This is the moral hazard question, and it is genuinely contested. The concern is that the expectation of future large-scale CDR gives political and economic cover to delay emissions cuts today. Some oil and gas companies have explicitly described CDR as justification for maintaining fossil fuel production. Climate models that assume large future CDR allow present-day budgets to be spent more freely.

But the empirical evidence on how this plays out at the individual and policy level is mixed. Studies of public perception find the moral hazard effect is very small or null depending on study design. And Nature Communications research on CDR policy risks frames the matter simply: net zero first requires deep emissions cuts, with CDR addressing the hard-to-eliminate residual.

The structural risk is real even if the behavioral evidence is ambiguous. Scenarios that assume large future CDR deployment are making bets on technologies that do not yet exist at scale. If those bets fail to materialize, the overshoot of carbon budgets cannot be undone.

Boundary Conditions

Where the CDR portfolio breaks down

Each approach in the CDR toolkit has conditions under which it fails to deliver as advertised.

DAC requires clean electricity to produce genuine negative emissions. Without low-carbon power, net carbon removal efficiency drops from 78% to 3.6%. It also lacks a formal accounting methodology in IPCC National Greenhouse Gas Inventory guidelines, which means DAC removals cannot currently be counted toward UNFCCC mitigation commitments. And as noted above, subsidy requirements for gigatonne-scale deployment run into the trillions of dollars.

Enhanced weathering depends critically on the energy source for rock grinding. Life cycle assessments show that grinding and transportation can re-release 10–30% of the captured CO2, making decarbonization of the supply chain essential to viability. Additionally, field trial evidence shows that actual CO2 removal is low in soils where pH is below 6.4, and three years of trials show limited and dispersed removal estimates — suggesting that theoretical potential may not transfer reliably to field conditions.

Ocean alkalinity enhancement shows a promising mechanism — adding alkalinity shifts ocean chemistry to draw down atmospheric CO2, with carbon storage lasting ~10,000 years, and can partially reverse ocean acidification. But a meta-analysis of 68 studies across 84 species found highly variable responses: 35% positive, 26% negative, 39% neutral. The first EPA-permitted field trial (LOC-NESS) showed no measurable harm, but trials remain limited in scale and duration. Larger, longer deployments could produce ecosystem effects not yet observed.

Nature-based solutions broadly — forests, soils, blue carbon — face an irreducible tension between their near-term cost advantages and their structural vulnerability to reversal. Reforestation alone cannot recover the carbon released during wood harvesting within typical rotation periods. Soil carbon faces inconsistent measurement methodologies and limited policy frameworks that make crediting at scale difficult. Even blue carbon — mangroves and salt marshes that sequester carbon at rates 2–4x faster than tropical forests — depends on protecting existing ecosystems rather than creating new ones, and their degradation already releases 0.15–1.02 billion tonnes of CO2 annually.

Thought Experiment

The locked door

Imagine that in 2030, a technically credible report is published showing that CDR technologies have advanced far enough — and deployment has ramped sufficiently — that reaching 1.5°C now appears achievable without reducing fossil fuel consumption by more than 30% from today's levels, rather than the 80%+ reductions previously required.

The report is peer-reviewed, the modeling is transparent, and the assumptions are disclosed. The conclusion follows from the evidence. Several major oil-producing countries cite the report as justification for slowing their phase-out commitments. Several large emitters adjust their national targets accordingly.

Now consider:

• The report assumes CDR technologies perform according to their median projections. Historically, large-scale energy technology deployments achieve the median roughly half the time. What happens if deployment falls short by 20%? By 50%?
• The permanence of CDR in the report is geological storage. But a significant portion of deployed capacity uses biological storage. What is the plan if the carbon is re-released by a decade of drought and fire?
• Who bears the consequences if the bets do not pay off? Emissions reduction is distributed cost; if CDR underperforms, the consequences are concentrated in places and populations least responsible for the accumulated CO2.

This is not a hypothetical designed to argue that CDR is bad. It is an invitation to be precise about risk. The moral hazard problem is not that CDR is wrong — it is that deploying CDR assumptions before CDR exists at scale shifts risk onto those with no say in the decision.