1. The shared label conceals different physical questions
A proposal promises to help cool the planet. Another promises to remove carbon dioxide. A third promises to prevent emissions that would otherwise occur. These statements can concern related climate goals while describing different physical changes. Putting them in one spreadsheet under climate benefit does not make their measurements interchangeable. Before comparing the proposals, a reader needs to know what quantity changes, where it changes, over what interval, and relative to which alternative.
The IPCC's 2022 mitigation assessment defines carbon dioxide removal around taking carbon dioxide from the atmosphere and storing the carbon durably. IPCC on carbon dioxide removal
Its 2021 physical-science assessment treats solar radiation modification separately, as intervention in Earth's radiation budget. The latter can affect the carbon cycle indirectly, but it is not the same operation as withdrawing and storing atmospheric carbon. IPCC on radiation modification and the carbon cycle
The relevant official excerpts were available through search; direct chapter pages returned access errors, so this article confines its attribution to those verified passages.
This distinction is not an argument that every proposal in one category is good and every proposal in another is bad. It is a demand for a claim that can be checked. A storage project needs evidence about carbon flows and retention. A radiative-intervention proposal needs evidence about radiative effects and their consequences. Each also raises other questions that a single success metric may fail to capture. An article should not treat a favorable number on one axis as a complete evaluation.
The examples below use fictional carbon batches, invented measurements and deliberately simplified boundaries. A local script actually checks their arithmetic, identifiers and unit errors. No real project was audited, no atmospheric intervention was carried out, and no climate model was run. The exercise is not a recommendation to buy credits, invest in a technology, or conduct a field experiment. It is a way to expose claims that appear compatible only because their units and assumptions have disappeared.
The central example follows two carbon batches from capture to storage while retaining their associated emissions as separate entries. Other examples examine a change in a land-carbon stock, a later release, a measurement interval, and a mistaken addition of capacity to output. They are intentionally small enough to reconstruct. The goal is not a universal calculator but an evidence trail whose boundaries remain visible when the results are translated into ordinary prose.
A useful final account should be able to say more than the intervention worked. It should identify the quantity measured, the relevant comparison, the observation period, the uncertainty, and the unmeasured consequences. That structure lets a reader criticize a specific assumption or request a specific missing record. It also prevents one promising result from being quietly promoted into proof of a broader climate outcome that was never measured.
2. Begin with the mass unit rather than the headline number
The Energy Information Administration explains the approximate 44-to-12 mass relationship between carbon dioxide and its carbon component: oxygen contributes the remaining mass of the molecule. Thus a tonne of carbon is not a tonne of carbon dioxide. The conversion concerns chemical composition, not a multiplier that creates additional removed carbon. It is essential to know which quantity a document reports before combining it with another. EIA's explanation of carbon and carbon-dioxide mass
In a simple calculation using that approximate ratio, ten metric tonnes of carbon correspond to about 36.67 metric tonnes of carbon dioxide. A second document reporting ten tonnes of carbon dioxide describes a different mass. If an analyst drops the unit suffix and adds the two as 20 tonnes of the same substance, the resulting total has no consistent meaning. The numerical values looked compatible only because the information identifying what they measured was removed.
The fictional ledger uses metric tonnes of carbon dioxide throughout its physical mass fields. Its associated emissions are also stipulated to be carbon dioxide only. It does not estimate a mixture of greenhouse gases or choose a conversion metric for them. That limitation is deliberate: introducing a carbon-dioxide-equivalent label without specifying the underlying gases and method would create a different accounting problem that the small fixture is not designed to solve.
A unit check should also preserve the distinction between a quantity and a rate. A statement that a facility can process 1,000 tonnes per year is not a statement that it processed 1,000 tonnes during a particular month. If an operations report covers three months, an annualized figure may be a projection or a capacity description rather than observed output. The article should identify which it is instead of silently treating the reporting interval as irrelevant.
Imagine two versions of a project summary. The first says 950 tonnes stored during the observation year. The second says a planned annual storage capacity of 10,000 tonnes. Adding them to obtain 10,950 tonnes stored would combine an observed quantity with a capability claim. In the toy comparison, 950 is 9.5 percent of 10,000, but even that ratio does not establish a utilization rate unless the capacity and observation periods refer to comparable operating conditions.
The practical record request is simple: preserve the original unit, interval, measurement location and status of the figure. If those fields are absent, the analyst should mark the quantity unresolved rather than choose a convenient interpretation. A precise-looking integer can be less informative than a modest range with an explicit unit and boundary. The first task of the ledger is to prevent precision in typography from being mistaken for precision in meaning.
3. Follow one batch without counting the same carbon twice
Batch A in the invented fixture begins with 1,000 tonnes of carbon dioxide captured from air. Of that amount, 950 tonnes reach the stipulated storage boundary and 50 tonnes are released before storage. The numbers satisfy a simple mass balance: captured carbon dioxide equals stored carbon dioxide plus the specified loss. This is an authored closed example, not a statement about the loss rate of any actual process or facility.
The Department of Energy's direct-air-capture life-cycle guidance emphasizes assessing the different phases of a system, estimating net removal, and considering uncertainty and environmental tradeoffs. A capture-device output alone is not a complete life-cycle result. The guidance is a framework for defining an assessment, not evidence that the fictional numbers used here describe a real technology. DOE's life-cycle assessment guidance for direct air capture with storage
For Batch A, the example separately assigns 150 tonnes of electricity-related emissions, 30 from heat, 20 from materials and ten from transport. Those terms total 210 tonnes. Subtracting that total from the 950 tonnes stored gives a net of 740 tonnes under the stipulated boundary. Starting instead with the captured 1,000, subtracting the 50 lost before storage and then the 210 in associated emissions gives the same result.
Two common spreadsheet mistakes produce different answers. Subtracting the 50-tonne loss again from the already reduced storage figure gives 690 tonnes, counting the same loss twice. Adding captured and stored quantities produces 1,950 tonnes, counting the same carbon at two stages as though it were two separate batches. The script checks the intended mass relationship, but the meaning of each field still has to come from the measurement design and the documents that define it.
A document-level example makes the duplication easier to see. One certificate records the amount leaving the capture unit; another records the amount entering storage. They can be two pieces of evidence about one batch, not two independent removals. Their identifiers, timestamps and measurement boundaries determine whether they can be reconciled. Matching only their similar titles would not establish that they concern the same material, while adding them without a match would be equally unjustified.
A useful public statement could therefore say that the example's net result is 740 tonnes after the specified losses and emissions, while listing what the boundary excludes. It should not simply repeat the largest available gross number. Nor should it claim that successful arithmetic validates the input measurements. The calculation establishes what follows from the authored entries; it does not establish the truth of an unsupplied meter reading or the completeness of an actual project's inventory.
4. Test which assumption changes the result
The electricity term for Batch A comes from 500 megawatt-hours multiplied by an invented emissions factor of 0.3 tonnes of carbon dioxide per megawatt-hour. The product is 150 tonnes. This factor is a stipulated scenario input, not a current grid average or a recommendation about an energy supplier. Its value is kept in the fixture so that a reader can see which assumption produced the electricity contribution.
In a second scenario, hold every physical batch quantity and every other emissions term constant but replace the electricity factor with 0.05. Electricity emissions become 25 tonnes; total associated emissions become 85 tonnes; and the net result becomes 865 tonnes. The 125-tonne change in the net result comes entirely from the altered factor. It is not evidence that the capture equipment processed more carbon or that additional storage was measured.
That comparison points to a concrete evidence question. Which factor belongs to the actual period and energy supply under investigation, and what accounting convention does it use? A proposed future energy contract, an annual average and a measurement tied to a particular operating interval are different records. This article does not select a universal convention among them. It shows why the convention and its evidentiary basis must be visible before the result can be compared with another estimate.
Sensitivity analysis can also identify a misleading disagreement. Suppose two analysts agree on the 950 stored tonnes and on 60 tonnes of non-electricity emissions, but publish net results of 740 and 865. Their spreadsheets need not disagree about a single observed mass. They may differ only in the electricity assumption. A correction should address that assumption rather than accuse one analyst of miscounting capture output or imply that two physical experiments produced different storage results.
Conversely, agreement on a final total does not prove agreement on the underlying inventory. One analyst could omit a 20-tonne materials term while overstating another term by 20 tonnes. The errors cancel numerically, leaving the same net figure. Comparing only the final total would miss the discrepancy. A more informative reconciliation compares component values, boundaries and source documents, then explains how each difference affects the result.
The exercise's second batch, B, stores 390 tonnes and has 60 tonnes of associated emissions, giving a net of 330. Combined with A's baseline, the two batches store 1,340 tonnes, incur 270 tonnes of emissions and yield a net of 1,070. With A's lower electricity-factor scenario, the combined net is 1,195. These are actual script outputs for the fictional inputs, not a prediction of what a real portfolio will deliver.
5. Distinguish an atmospheric withdrawal from a prevented addition
The starting location of the carbon matters. In one fictional process, 1,000 tonnes of fossil carbon dioxide would reach the atmosphere without capture. Capturing and retaining 900 leaves 100 released. Relative to that stipulated alternative, the process prevents 900 tonnes from being added. It does not, through that operation alone, withdraw 900 tonnes that were already in the atmosphere. The distinction follows from the direction of the specified flow, not from whether the equipment is impressive.
Consider a transport pipeline that receives both kinds of material. At the storage end, a meter may report the combined mass accurately while providing no information about how much originated in ambient air and how much came from a fossil source. The storage measurement alone cannot recover the missing provenance. The analyst needs upstream batch records and a method for connecting them to the downstream quantity without double-counting transfers.
For example, suppose an invented delivery contains 700 tonnes identified as air-derived and 300 identified as fossil-source carbon dioxide. Reporting 1,000 tonnes stored is consistent with that input. Reporting 1,000 tonnes removed from air is not. If 100 tonnes lack a supported origin assignment instead, the unresolved fraction should remain visible. The analyst should not classify it as atmospheric solely because the destination is described as a removal facility.
There is also a difference between a physical flow and an institutional claim about that flow. A seller's invoice, an operator's measurement and a buyer's sustainability report can all refer to the same batch. Their existence does not multiply the physical withdrawal. An account that adds the 740-tonne figure from two organizations' reports would produce 1,480 even if both documents describe Batch A. Shared identifiers and claim boundaries are needed to recognize that overlap.
6. A storage permit answers a particular regulatory question
EPA describes Class VI wells as part of its Underground Injection Control framework for geologic sequestration. Its overview includes site characterization, monitoring, financial responsibility and a post-injection phase involving well plugging, plume and pressure monitoring, and a demonstration that underground sources of drinking water are not endangered before closure. That is a defined regulatory purpose and process, not a certificate that every claimed climate-accounting quantity has been verified. EPA's Class VI project overview
Suppose an invented project packet contains a permit authorizing specified operations and a separate annual mass report. The permit establishes what the issuing authority authorized under its terms. The mass report supplies a quantity claimed for a particular interval. The analyst should not use the existence of the first document as proof of the exact number in the second. Nor should a disagreement about the mass report be casually described as proof that the permit is invalid.
The same scope discipline applies to capacity. A modeled storage capacity describes a proposed or assessed ability to hold material under specified assumptions. An injection record concerns what actually entered during a period. A monitoring record concerns observations at particular locations and times. The documents may support one another, but they are not interchangeable. A public summary should identify which record supports each part of the conclusion instead of listing them together as a general seal of approval.
Now introduce an anomaly into the fictional packet. One monitoring instrument reports an unexpected value, while the annual storage total remains unchanged. The analyst cannot determine from that pairing alone whether carbon leaked, an instrument malfunctioned, or another explanation applies. The useful next documents would be the instrument's checks, the investigation of the anomaly, and any resulting revision to the relevant inventory. The anomaly deserves attention without being prematurely assigned a cause.
A record of maintenance or a corrective action can also be misunderstood. Its existence may show that an issue was identified and addressed; it does not automatically show that the issue was harmless or that the entire project failed. The account should identify the specific condition, the evidence for its consequence and the stated response. If those links are absent from the public packet, the uncertainty should be attached to that question rather than converted into a sweeping verdict.
This is why duration belongs beside the storage claim. An observation through a stated date is evidence through that date, with the limits of the monitoring arrangement. A projection of longer retention is a different kind of statement. The article should preserve the distinction between measured history, modeled future behavior and obligations that continue after an operating phase. A one-word label such as permanent does not remove the need to identify which of those statements the record actually supports.
7. A certification method is not an observed project outcome
The European Commission's current certification-methodologies page identifies Delegated Regulation (EU) 2026/285 as establishing methodologies for direct air capture with carbon storage, biogenic emissions capture with carbon storage, and biochar carbon removal. The page separately identifies other pathways being explored. The existence of a methodology is a statement about a framework; it does not, by itself, establish that a particular operator has delivered a particular quantity. The Commission's current carbon-removal methodology overview
Imagine a fictional operator's brochure citing that regulation beside a claimed annual removal total. The citation may correctly identify a relevant method, but the numerical claim still needs its own supporting records. The analyst should ask which method version was applied, what project boundary was assessed, what period was covered and what determination was actually issued. A general reference to a recognized framework cannot fill those fields on its own.
A simple land-stock example shows why the comparison must be explicit. A measured stock rises from 100 to 112 units during a stated interval. The observed increase is 12. In the authored scenario, a specified comparison case would rise from 100 to 106 over the same interval. The difference between those changes is six units. Reporting the full 12 as the incremental effect of the intervention would disregard the comparison the example itself supplied.
That does not establish that six is the correct answer for a real project. The comparison case is a hypothetical assumption, not an observation of the same land in an alternative history. Its credibility would need evidence and a method appropriate to the setting. The example distinguishes the arithmetic from the argument about the baseline. A spreadsheet can calculate the difference perfectly while the comparison remains weak, disputed or outside the scope of the available evidence.
The measurement design matters as well. Suppose the first stock measurement samples one depth and the second samples a different depth. A change in the reported total could reflect the changed measurement boundary. Or suppose the area included in the second survey expands. The analyst needs consistent area, depth, timing and units before interpreting the difference as a change in the same stock. This is an original record-comparison exercise, not a claim that a particular certification method permits such mismatches.
A careful account would therefore report the observed change and the incremental estimate separately, explaining the assumed comparison. It would identify which quantity a certificate or public statement actually claims. If the documents do not reveal the comparison, the article should not invent one. The most useful question is not whether the brochure contains an official logo or a regulation number, but whether the claimed quantity can be traced through a specified method to evidence with a compatible boundary.
8. Keep later releases separate from earlier delivery claims
Return to Batch A's baseline net of 740 tonnes under the original assumptions. Now stipulate that 100 tonnes of its stored carbon dioxide are released in a later period, with all other entries unchanged. The retained quantity becomes 850 tonnes, and subtracting the original 210 tonnes of associated emissions leaves 640. The later event changes the ongoing balance. It does not mean the original arithmetic was wrong; it means the history now contains another flow.
This example is intentionally physical rather than contractual. It does not specify who bears liability, what a registry would cancel, or what replacement obligation applies. Those questions would require the relevant rules and agreements. An analyst should not assume that correcting a registry entry physically recaptures a released tonne, or that a physical release automatically establishes a particular financial remedy. The two accounts can be connected, but they need distinct evidence.
Consider a separate three-year sequence of net removals: 100 tonnes in the first year, 150 in the second and 50 in the third. The cumulative amount is 300. A later release of 80 reduces the stipulated cumulative balance to 220. A chart displaying only the original positive annual deliveries would omit that later change. A chart that erases all earlier entries would instead destroy the history of when removal and release occurred.
The appropriate presentation depends on the question. A delivery-history chart can preserve each dated event. A retained-balance chart can show the cumulative sum after later changes. A claim about a particular vintage needs to identify which earlier quantity the later event affects. These are not three interchangeable ways to decorate the same number. They are different views of a time-dependent record, and a reader needs to know which view is being presented.
Now suppose a release estimate is revised from 80 to 60 after better measurements become available. The balance would change from 220 to 240 under the same simplified sequence. The analyst should retain the initial estimate, the revision, its basis and the date the revision became known. Calling the 20-tonne difference a new removal would misidentify the event. It is a correction to a prior estimate, not a newly observed transfer from the atmosphere.
This distinction also makes uncertainty more manageable. The account can state which quantities are measured, which are estimated, and which remain subject to later adjustment. It need not imply that every estimate is worthless or that every revision signals misconduct. The credibility of a long-running record depends partly on whether revisions can be reconstructed, not on a promise that the first number will never change.
9. Do not translate a radiative result into removed tonnes without a model
The IPCC's physical-science assessment explains that solar radiation modification can influence land and ocean carbon sinks through climate feedbacks, while not stopping atmospheric carbon dioxide from rising or resolving the associated ocean acidification under continued emissions. The distinction is therefore not that radiation changes have no connection to carbon. It is that the connection is not the same as a measured atmospheric withdrawal and durable storage. IPCC's assessment of biogeochemical responses to radiation modification
An invented observation reporting a change in watts per square metre measures something different from a report of tonnes of carbon dioxide. Adding those values would be as meaningless as adding a distance to a mass. A model might relate specified radiative and carbon scenarios under explicit assumptions, but the relationship is not a universal unit conversion like kilograms to tonnes. It depends on the question, the time horizon and the physical system being represented.
Suppose a project summary reports a short-period radiative change over a particular region and describes it as equivalent to removing a large amount of carbon dioxide. A reader should ask for the model, comparison scenario, geographic scope and interval supporting that equivalence. Without those details, the numerical translation cannot be evaluated. A local result does not automatically establish a global climate outcome merely because the same word cooling appears in both descriptions.
A separate two-region arithmetic example illustrates another compression problem. Imagine an index that changes by plus 0.2 in one equally weighted region and minus 0.2 in another. The average change is zero. That does not mean that neither region changed. Nor does the zero average establish that the positive and negative changes are equally important to people, ecosystems or infrastructure. The example is not a climate simulation; it shows information lost by aggregation.
If the regions have different weights, the average changes again. Give the first region three times the weight of the second, and the weighted average is plus 0.1. The underlying regional values have not changed; the summarization rule has. A public account should identify the weighting and explain why it is appropriate for the claimed result. A global number without its aggregation method can hide a substantial interpretive choice.
The comparison between interventions should therefore preserve multiple dimensions when the question requires them. Carbon retained, radiative change, regional responses and duration are not automatically reducible to one score. A well-defined model can help compare scenarios, but its output should remain labeled as a modeled result with stated assumptions. The absence of a simple common unit is not a failure of communication. It is a property of the different physical questions being asked.
10. A temperature statement does not substitute for ocean chemistry
NOAA's ocean-acidification indicator explanation distinguishes dissolved-carbon-dioxide conditions, pH and aragonite saturation. It notes that pH is logarithmic and that lower pH indicates increasing acidity, even while typical seawater remains above neutral pH. These are chemical indicators, not alternative names for temperature. A claim about one does not supply a measurement of all the others. NOAA's explanation of ocean-acidification indicators
For a simplified numerical example, consider pH changing from 8.2 to 8.1 under a consistent measurement convention. The logarithmic relationship gives a hydrogen-ion activity ratio of ten raised to 0.1, approximately 1.259: an increase of about 25.9 percent. Dividing the 0.1 change by 8.2 gives about 1.2 percent, but that is a percentage change in the scale value, not the corresponding change in hydrogen-ion activity. The two calculations answer different questions.
This example does not estimate a real ocean trend, predict a biological response, or remove the need to specify measurement conditions and pH conventions. It demonstrates why a small-looking change on a logarithmic scale can be misdescribed. The source's explanation of the scale and the example's arithmetic are distinct from an actual survey. An article should not turn the illustrative numbers into observations merely because they resemble values encountered in public discussion.
Imagine a fictional monitoring packet with temperature observations from one station and chemistry samples from another, taken in different seasons. Combining them into a single before-and-after claim about an intervention would require a defensible connection. The differences in place, time and sampling could matter. The analyst should identify those mismatches instead of presenting the two records as synchronized measurements of the same water mass.
A more useful packet would preserve the sampling location, depth, time, measured quantity and method, then state what comparison is supported. If one chemical indicator was not measured, the article should say so. It should not infer that the indicator stayed constant because another variable changed in the desired direction. Nor should it infer that every chemical change has the same biological consequence without the relevant evidence about the setting and organisms.
The larger lesson is about the shape of a claim. A proposal intended to affect warming may still leave other consequences of elevated carbon dioxide unresolved. Conversely, a measured carbon-related improvement in one place does not establish that all environmental effects are beneficial. Keeping chemistry and temperature in separate fields is not a demand for endless caveats. It is a practical way to prevent a statement about one variable from being used as evidence about a different variable that was never observed.
11. Distinguish a research recommendation from an implementation decision
The National Academies' March 2021 report on reflecting sunlight recommended research and research governance aimed at improving policy-relevant understanding, rather than advancing a path to deployment. It is a dated scientific advisory report, not an authorization for a particular present-day intervention. Its research recommendation should not be described as proof that a technique has been demonstrated safe and effective for operational use. The National Academies' 2021 report highlights
An original document exercise can separate three records that are often collapsed: a research proposal, an experimental result and an implementation authorization. The proposal describes work someone wants to perform. The result describes what a specified investigation found. The authorization describes what a competent body permitted under stated conditions. The presence of one does not establish the existence or content of the other two, even when all concern the same broad technology.
Suppose a fictional proposal seeks to test an instrument over a small area. A later slide reports that the instrument detected the intended local signal. A promotional summary then says that the intervention has been approved to address global warming. The intermediate documents do not support that final statement. The instrument result may be valid and useful, but the claimed authorization and global outcome require records the packet has not supplied.
The reverse mistake would be to dismiss the instrument result because it does not answer the global policy question. A small experiment can answer a bounded question without resolving a larger one. A precise account should identify what the experiment measured and which inferences remain unsupported. It should neither inflate the finding into deployment readiness nor erase it for failing to establish something outside its design.
This same three-record approach works for carbon removal. A method paper, a project measurement and a certification decision have different roles. A model predicting future capture capacity is not a record of delivered storage. A permit is not a substitute for the measurements an accounting claim requires. The analyst should match each sentence to the record capable of supporting it rather than build confidence by accumulating documents with related titles.
For a reader deciding what further scrutiny is needed, that distinction produces concrete questions. Is the missing evidence a measurement, an explanation of how a measurement scales, an assessment of consequences, or an authorization? Different missing links require different work. Treating all uncertainty as one undifferentiated research gap makes it harder to know what progress would look like. A well-scoped account can identify the next decisive piece of evidence without claiming that it already exists.
12. Publish the balance together with the test's limits
The local fixture checked two authored batches and their stated mass units. It returned 740 tonnes for A and 330 for B under the baseline assumptions, totaling 1,070. It rejected three deliberately altered inputs: a duplicate batch identifier, a carbon-mass unit substituted for the required carbon-dioxide unit, and a mismatch between captured mass, storage and loss. Those are actual bookkeeping checks. They are not measurements of a facility, certification of a removal, or validation of a climate model.
The unit control is especially limited. It rejects an incompatible label; it does not infer the correct conversion or decide what an ambiguous source document meant. The balance control identifies inconsistent entries; it does not determine which meter or report is wrong. The duplicate control detects a repeated identifier; it does not prove that every real-world duplicate would share an identifier. These limits should remain attached to any claim that the data passed a check.
A separate uncertainty calculation uses 950 plus or minus 20 tonnes for stored mass and 210 plus or minus 30 for associated emissions. Treating these as bounded intervals, the lowest net is 930 minus 240, or 690; the highest is 970 minus 180, or 790. This is a worst-case interval under the stated bounds, not a confidence interval. It does not assume a probability distribution or independence between the measurement errors.
The coincidence that 690 also appeared in the earlier double-subtraction mistake is instructive. An identical number can arise from a legitimate lower-bound calculation or an invalid central estimate. The value alone cannot tell the reader which happened. The operations, assumptions and intended interpretation matter. A chart should not present an uncertainty bound as though it were the central result, nor defend a counting error merely because its answer falls inside a plausible range.
A strong final summary of the fictional project would report the baseline net, the alternate electricity assumption, the bounded uncertainty example and the later-release scenario as separate statements. It would not add them together or choose whichever makes the strongest headline. It would identify the observation period and preserve the distinction between actual fixture arithmetic and untested physical inputs. A reader could then reconstruct each result and see exactly what would change it.
That is the standard worth applying to public climate-intervention claims. Keep the physical quantity, source, destination, interval, comparison and uncertainty visible. Do not make a radiative observation impersonate a carbon withdrawal, a gross flow impersonate a net result, or an official framework impersonate a completed project. Different interventions may address connected problems, but their evidence must retain the differences that make their claims testable. The honest comparison starts by refusing to merge what the records have not shown to be equivalent.