The central question: what does the NO2 number actually subtract?
Many ambient nitrogen dioxide monitoring systems use chemiluminescence to measure nitric oxide, NO, then infer nitrogen dioxide, NO2, from a second measurement made after a converter transforms NO2 into NO. In the simplified ideal case, NOx = NO + NO2 and NO2 = measured NOx minus measured NO. The analyzer detects light from a chemical reaction in its chamber, not the ambient NO2 concentration by directly shining a light through NO2 itself.
To read an NO2 result correctly, verify which analyzer method was used, whether the converter is selective, whether conversion efficiency and channel timing were valid, and which units and intervals the data represent. A heated molybdenum converter may also convert other oxidized reactive nitrogen compounds, making apparent NO2 higher than true NO2 under some conditions. Photolytic conversion and direct NO2 methods have different interference patterns and should not be treated as identical.
This article uses illustrative, constructed datasets to explain subtraction, conversion efficiency, uncertainty and quality control. It aims to teach precise scientific translation: distinguish NO, NO2, NOx and the broader reactive-nitrogen family, then state what the instrument can establish about ambient air at a site. The calculations are not real observations or a replacement for regulatory operating instructions.
Four labels that look similar but are not equivalent
NO is nitric oxide, a nitrogen-oxygen gas emitted in several combustion and atmospheric processes. NO2 is nitrogen dioxide, a different molecule with its own environmental significance. NOx commonly denotes NO plus NO2 when used as a monitoring parameter. NOy refers to a broader family of reactive oxidized nitrogen compounds, often including species such as nitric acid and organic nitrates in addition to NOx. Whether a particular instrument channel reports one of these families must be established from the exact method.
If a dataset lists NO, NO2 and NOx separately, do not add all three together to describe total nitrogen oxides. In the ideal case NOx already contains NO and NO2; adding NO2 again double-counts it. A monitor may present reported concentrations for several channels, but these are related measurements, not independent stacks of molecules.
The same three letters can also appear in policy documents with operational definitions. Read the parameter code and method designation instead of relying only on the abbreviation in a chart title. Precision about species is the first defence against comparing unlike data.
Why the instrument produces light
In a chemiluminescence detector, nitric oxide reacts with ozone generated or supplied inside the instrument. One product pathway creates electronically excited nitrogen dioxide, which releases light as it relaxes. Under controlled instrument conditions, the detected emission is related to the nitric oxide entering the reaction cell. This is different from simply detecting the colour of NO2 in ambient air.
The analyzer needs control of reaction gases, sample flow, detector conditions and background response. Any fault in these components can affect the relationship between light and NO concentration. A calibrated response is therefore evidence only when the full system behaves as specified. The emitted light is a measurement signal from an induced chamber reaction, not a direct picture of the local atmosphere.
Chemiluminescence is attractive because it can be sensitive and automated. But the direct signal is associated with NO. To estimate NO2 in a conventional converter-based instrument, a separate step creates more NO from the sample. That is why the converter’s chemistry is as important to the final reported nitrogen dioxide as the photodetector itself.
The two channels and the key subtraction
In the direct path, untreated sample air enters the reaction system to give an NO measurement. In the converted path, the sample passes through a device designed to transform NO2 into NO, so the downstream detector responds to original NO plus converted NO2. The instrument or data processor subtracts the NO component to estimate NO2. This is the conceptual basis for NO2 = NOx − NO.
To make the interpretation valid, measurements must represent comparable air and periods. Some analyzers switch channels in sequence, so a rapid plume can alter NO and converted signals between switching steps. Instruments have sampling and averaging logic to handle this, but a reader should not assume two visually adjacent dashboard numbers were necessarily measured at exactly the same instant.
A useful teaching diagram has two arrows from one ambient sample: an untreated path labelled NO, and a converter path labelled converted NOx. Subtracting the first from the second yields apparent NO2. Label it ‘apparent’ until converter selectivity, efficiency and quality controls have been established.
Worked subtraction with matching units
Suppose an analyzer reports 48 ppb from its converted NOx channel and 21 ppb from its NO channel for a matched, valid interval. The apparent NO2 concentration is 48 − 21 = 27 ppb. Reverse check: NO plus NO2 gives 21 + 27 = 48 ppb, consistent with the simplified channel definition. Both numbers must be in the same mixing-ratio units.
If the NO channel is reported in ppm, convert before subtraction. For example, 0.021 ppm equals 21 ppb. Subtracting a number 0.021 directly from 48, when one is ppm and the other ppb, is nonsense despite producing an ordinary-looking decimal. A spreadsheet needs explicit unit columns, and formulas should operate on harmonised values.
The result 27 ppb is a calculation from the constructed record. It has not yet proved that the converter was selective or that no other nitrogen compounds interfered. A full scientific statement includes the method and the appropriate acceptance of the underlying instrument data.
How heated molybdenum can create positive interference
Some conventional NOx analyzers reduce NO2 to NO using a heated molybdenum-based converter. The conversion process may also transform other oxidized nitrogen species, such as nitric acid and organic nitrates, into NO. The downstream signal then includes more than the original NO and NO2. If the entire converted-minus-direct signal is labelled NO2, the reported number may be biased upward relative to true ambient NO2.
The effect is particularly relevant in chemically aged or photochemically active air containing non-NOx reactive nitrogen, often called NOz in atmospheric research. It is not correct to say that every molybdenum-converter reading is wrong by a fixed percentage. Interference depends on mixture, converter design, conditions and calibration. Evidence from independent methods helps quantify the problem.
EPA’s method-code guidance distinguishes standard chemiluminescence, photolytic-chemiluminescence and direct NO2 analyzer types. That distinction is not cosmetic metadata. It tells a reader which chemistry and possible interference pathways need to be considered before interpreting a trend or comparing stations.
Numerical illustration of a false high NO2
Suppose the true ambient NO is 21 ppb and true NO2 is 27 ppb, so ideal NOx would be 48 ppb. Imagine an imperfectly selective converter adds the equivalent of 5 ppb from other reactive nitrogen compounds to its converted signal. It would then indicate 53 ppb while the untreated NO channel remains at 21 ppb. Subtraction produces apparent NO2 = 53 − 21 = 32 ppb, five ppb above the true assumed value.
The positive bias relative to true NO2 is 5/27 ≈ 18.5 percent. Relative to the apparent reported 32 it is 15.6 percent. The denominator must be declared; both percentages can be calculated, but they answer different questions. This exercise demonstrates a mechanism rather than a universal correction factor.
One cannot recover true NO2 by always subtracting five. The non-NOx interference varies with atmospheric chemistry, converter performance and measurement method. Without independent evidence, the apparent discrepancy should be discussed as a possibility, not numerically corrected by guesswork.
Photolytic converters address a different part of the problem
A photolytic converter uses selected light to dissociate NO2 to NO, generally seeking better chemical selectivity than a broad catalytic converter. That can reduce some positive interference from other oxidized nitrogen compounds. But conversion efficiency may be less than complete, and the photolytic response has its own dependence on wavelength, cell design, flow and possible spectral interactions.
Do not interpret ‘photolytic’ as ‘perfect’. The analyzer still needs calibration and verification, and the converted signal must be corrected in accordance with the approved method when necessary. It is possible for a better-selective system to produce lower apparent NO2 than a molybdenum system during an aged photochemical air mass without either simply having a faulty numeric display.
Direct optical methods, including some cavity-based techniques, measure NO2 without the same conventional NO-to-NOx subtraction. They bring different calibration and interference considerations. Before deciding that one city’s NO2 changed across a monitoring upgrade, inspect whether the analyzer methodology changed.
Conversion efficiency and the temptation to divide blindly
If a converter transforms only a known fraction of NO2, its response can be modelled conceptually as converted signal = NO + η × NO2, where η is the relevant conversion efficiency. In a simplified illustration with η = 0.90, original NO = 10 ppb and true NO2 = 20 ppb, the converted path would show 10 + 0.90×20 = 28 ppb. Direct NO is 10, so the raw difference is 18 ppb.
If a valid, method-appropriate efficiency correction is applied, 18/0.90 = 20 ppb. But division by a guessed efficiency can do more harm than good. The actual analyzer may internally compensate using its calibration, or its measurement scheme may not conform to that simple equation because of interferences and background. Never apply a second manual correction without knowing whether the exported NO2 result was already corrected.
Record whether reported values are raw channels or final validated results. The presence of a converter-efficiency number in a manual does not mean all users should recalculate every public NO2 value. Use the operator’s approved method and data definition.
Why subtracting two readings amplifies low-end uncertainty
At low concentrations, NOx and NO may be nearly equal. Their difference can be much smaller than either channel. Noise, offsets and time alignment then have greater relative importance. Suppose NOx is 3.0 ppb and NO is 2.8 ppb; apparent NO2 is 0.2 ppb. A small change of 0.3 ppb in either channel could change the sign of the difference. Reporting 0.2 with unjustified precision is misleading.
If independent channel uncertainties have standard deviations u1 and u2, the uncertainty of their difference is often approximated by the square root of u1² + u2² when the uncertainties are uncorrelated. For each channel at 0.5 ppb, that gives about 0.71 ppb. Real analyzer channels can share correlated errors, so an instrument-specific calculation may differ. The example simply illustrates why the difference may be less certain than either large parent number.
A slightly negative calculated NO2 value is not physical negative gas concentration. It may be a baseline or noise consequence of subtraction. Use the network’s prescribed data treatment instead of silently zeroing, deleting or claiming a negative gas was present.
NOx is not automatically NOy
NOx in common ambient air monitoring is used for the sum of NO and NO2, though method definitions should be checked. NOy is a broader family of reactive oxidized nitrogen species. An NOy analyzer may intentionally convert many species to NO for detection, which is different from an instrument whose intended parameter is NOx but inadvertently detects some non-NOx compounds.
This distinction matters because interference depends on purpose. A converter that responds broadly may be a problem for selective NO2 reporting and useful in a purposely designed broader reactive-nitrogen measurement, depending on its validated method. The same chemical conversion cannot be judged without asking which analyte family was intended.
An educational report should not combine a nearby NOy value with NOx or subtract one from the other unless their definitions, units, methods, intervals and uncertainty support the operation. Technical vocabulary is a map of different measured quantities, not a basket of interchangeable labels.
The relationship with ozone outside the analyzer
NO and ozone react in ambient air. Near fresh NO emissions, ozone may be depleted locally while NO2 forms. During daylight, NO2 participates in photolytic cycles that can contribute to ozone formation in the presence of suitable precursor chemistry. This explains why NO, NO2 and ozone patterns often vary by time and site.
But the ozone used inside an NO chemiluminescence detector is a controlled reagent for generating an optical signal. It should not be confused with ambient ozone that appears as a separate environmental measurement. A schematic with both chemical contexts must label them explicitly. Otherwise learners may conclude that the analyzer simply reads the city’s existing ozone glow.
A graph in which NO2 rises while ambient ozone falls can be consistent with local titration, but transport, emissions and sunlight also affect both. A single pair of curves cannot uniquely assign the cause. Combine site type, meteorology and suitable independent pollutant measurements before drawing attribution.
Time stamps and channel switching
Some analyzers switch between untreated and converted flow paths. If a sharp traffic plume passes during one path but not the other, an instantaneous subtraction could be misleading. Commercial instruments use internal cycling and averaging methods, yet externally exported data may have different resolutions or time stamps. Read the manufacturer and network definitions before aligning channels manually.
To compare NOx and NO in a spreadsheet, use matched accepted intervals and consistent units. A row labelled 12:00 may mean the hour ending or starting at noon. Combining an 11:00–12:00 NOx mean with a 12:00–13:00 NO mean can manufacture a negative NO2 during a changing plume.
Cloud upload time is another variable. If a remote monitor transmits buffered data later, a data-receipt timestamp should not be used as physical observation time. Preserve interval boundaries and time-zone conversions, especially when comparing with traffic, weather or ozone measurements.
Zero, span and converter verification
NO and NOx channels need appropriate zero and span checks, flow and pressure stability, reaction cell and detector diagnostics, and converter-efficiency verification as specified for the instrument. A low-noise detector cannot compensate for a converter whose selectivity has changed. A converter test does not establish that the sampling inlet was leak-free. These are separate parts of measurement validity.
Suppose a span challenge of known 40 ppb NO is reported as 42 ppb. The absolute difference is +2 and the relative difference is +5 percent. These numbers do not by themselves determine whether the instrument passed; consult the applicable acceptance limit and challenge method. A challenge targeting NO2 conversion must be evaluated differently from one that simply tests direct NO response.
Keep as-found and as-left service readings. A maintenance record saying ‘calibration fixed’ is less useful than a record identifying the initial error, adjustment, post-service confirmation and possibly affected prior intervals. If a public dataset is later revalidated, the reason should be traceable.
Direct NO2 methods do not invalidate all older observations
Some direct optical methods, including cavity attenuated phase shift approaches, provide a different way to measure ambient NO2 without relying on the same conventional difference between two chemiluminescence channels. That can be advantageous for evaluating converter-related interference. It does not mean every conventional historic NO2 dataset must be discarded as worthless.
Use the designated method and data-quality context. In a comparison between a direct method and a converter-based method, matched time, site, flow and validation are prerequisites. A systematic difference during aged photochemical conditions might indicate converter interference; it could also require examination of the direct method’s own calibration and potential artefacts.
An instrument-method change within a long time series can shift the reported quantity even if emissions do not change. Document method codes and transitions before interpreting decade-scale trends, and avoid mixing raw and finalised values under one vague heading.
A conversion-efficiency challenge is not a converter-selectivity test
Conversion efficiency asks how much of an introduced NO2 challenge becomes detectable NO. Selectivity asks whether other substances also become detectable NO. An instrument could convert virtually all challenged NO2 and still respond to atmospheric nitric acid or organic nitrates. Both characteristics matter, but they cannot be verified with one number. An operator reporting only excellent conversion efficiency has not established that the resulting ambient NO2 measurement is chemically specific. A reader should inspect which gases the check used and what other species are plausible at the sampling location.
Construct a controlled exercise with a zero-air baseline, a directly calibrated NO response, and an independently established NO2 challenge of 40 ppb. If the converted-minus-direct response attributable to that challenge is 36 ppb under the stated protocol, the observed response ratio is 36/40 = 0.90, or 90 percent. Whether that passes depends on the governing method. If the analyzer already incorporates an efficiency adjustment, taking the output and dividing by 0.90 again would incorrectly duplicate that adjustment. Check exactly whether the logged signal is raw, calibrated, compensated or final.
A separate selectivity experiment would introduce a known potential interferent while holding true NO and NO2 fixed, ideally comparing against an independent method. If the inferred NO2 rises without any addition of actual NO2, the response is evidence of interference. Interference in a real atmosphere is more complicated because multiple nitrogen compounds, relative humidity and inlet conditions vary together. One successful challenge cannot supply an all-weather correction. Keep the distinction between analytical recovery and chemical identification as carefully as one distinguishes correct arithmetic from the correct scientific premise.
How converter loss and chemical interference can occur together
Positive converter interference and incomplete NO2 conversion are not mutually exclusive. Suppose a constructed example contains 25 ppb true NO2, a converter transforms 90 percent of that component, and other oxidized nitrogen adds an equivalent of 4 ppb NO to the converted path. Ignoring calibration adjustments, the converted-minus-direct signal attributable to these terms is 0.90 × 25 + 4 = 26.5 ppb. Although conversion is incomplete, the apparent NO2 exceeds 25 ppb. The two errors happen to partly offset. Agreement with truth would not prove that either mechanism is absent.
Change only the assumed interferent contribution to 1 ppb. Then the same signal is 0.90 × 25 + 1 = 23.5 ppb, a value 1.5 ppb below the true input. The direction of total error changes without a change in converter efficiency. This is why a single statement such as ‘molybdenum always raises NO2 by twenty percent’ is invalid. Two physical mechanisms with different magnitudes determine the observed error, and their relative importance changes with the sampled air.
If a method explicitly permits and defines an efficiency correction, apply only that documented operation to the appropriate component of its signal; do not mechanically divide a channel contaminated by unknown non-NO2 nitrogen and announce a true NO2 value. In the first example, 26.5/0.90 is about 29.4 ppb, not 25, because the 4 ppb interference also gets amplified. Better specificity evidence or an independent instrument is needed before claiming a resolved measurement. The practical lesson is to separate calibration, selectivity and chemical assumptions in the report.
Quantify uncertainty in a difference, rather than in either channel alone
Because inferred NO2 is the difference between a converted NOx response and an untreated NO response, the uncertainty of the difference matters. For two measurements X and Y with uncertainties uX and uY, an elementary independent-error model gives u(X−Y) = square root of (uX squared plus uY squared). If the measurements share calibration drift or timing artefacts, their covariance must also be considered: variance of X−Y equals variance of X plus variance of Y minus twice their covariance. This is a mathematical framework, not a substitute for an approved measurement uncertainty budget.
Take a hypothetical converted-channel result of 60 ppb with standard uncertainty 2 ppb and an NO result of 45 ppb with standard uncertainty 1.5 ppb. If the uncertainties are independent, calculated NO2 is 15 ppb with combined standard uncertainty square root of (4 + 2.25) = 2.5 ppb. The relative uncertainty is 2.5/15 = 16.7 percent even though the two underlying measurements appear relatively stable. This illustrates why low NO2 obtained from two larger numbers deserves special caution.
Now imagine both channels undergo the same positive zero drift. Their difference might partly cancel that common change; conversely, different channel offsets need not cancel. It is not defensible to claim complete cancellation without studying the design, switching sequence and statistical correlations. In a public-facing explanation, present the reported value, its method, applicable uncertainty or data-quality flags, and any limitations in the assumptions. When those inputs are unavailable, say that a formal uncertainty calculation cannot be completed instead of supplying a decorative percentage.
From mixing ratio to mass concentration without losing the conditions
Ambient NO2 can be displayed as ppb by volume or as micrograms per cubic metre. These are not interchangeable labels. For an ideal gas, a useful relation is mass concentration in micrograms per cubic metre equals the mixing ratio in ppb multiplied by the compound molar mass in grams per mole and divided by molar volume in litres per mole. That molar volume depends on temperature and pressure. A regulatory or monitoring network may prescribe reference conditions; the reader must use those, not an arbitrary room-temperature shortcut.
For a teaching example at 25 degrees Celsius and 1 atmosphere, ideal-gas molar volume is approximately 24.45 litres per mole. NO2 molar mass is about 46.0 grams per mole, giving roughly 46.0/24.45 = 1.88 micrograms per cubic metre for each ppb. Thus 27 ppb would correspond to approximately 50.8 micrograms per cubic metre under these explicitly stated conditions. If the reference conditions change, so does the factor. The same concentration number should not be transferred to another gas by retaining the NO2 molecular mass.
The conversion adds another layer to the original problem: a correctly converted apparent NO2 value still retains any underlying converter interference. Changing units does not improve selectivity or validate a previously flagged hour. Also distinguish a concentration reported under reference conditions from a value physically measured at ambient temperature and pressure. When merging datasets, preserve raw units, reference assumptions and conversion provenance. A tidy numerical column without those facts can make an international comparison look exact while quietly mixing different quantities.
Identify the data stage before comparing hours or sites
A modern monitor may produce instantaneous detector signals, short-cycle channel readings, instrument-generated minute values, telemetry records and later quality-assured public observations. A number visible on a maintenance dashboard is not automatically the certified value used in a regulatory report. Look for the data product definition, time-zone convention, averaging interval, completeness criteria and quality-code vocabulary. For example, an hour labelled 14:00 may represent the hour beginning at 14:00 or the hour ending then, depending on the system. The label alone does not identify the window.
Consider an illustrative sequence of four valid 15-minute apparent NO2 averages: 18, 22, 26 and 30 ppb. If each covers an equal duration and the network permits this aggregation, the hour average is (18+22+26+30)/4 = 24 ppb. If one quarter-hour is invalid, simply taking the remaining three would yield 23.3 ppb, but that number is not necessarily an authorised hourly value. The governing completeness rule may require additional valid coverage or a flag. A calculation and a publishable observation are separate decisions.
A related trap is averaging already rounded hourly figures and reporting excessive decimals. If four inputs were stored only to the nearest integer, 23.25 ppb is an arithmetic outcome but may imply more precision than the inputs justify. Retain higher-resolution archived values where available, apply documented rounding at the appropriate reporting stage and mark preliminary estimates clearly. Where an official dataset offers both provisional and validated records, prefer the latter for definitive claims. The date on a chart is not proof of data-finalisation status.
Separate source inference from the chemistry of NO, NO2 and ozone
Fresh combustion plumes can contain substantial nitric oxide, while reactions involving ozone convert some NO toward NO2. Sunlight also drives photolysis of NO2 and a network of reactions involving volatile organic compounds and radicals. The resulting local NO-to-NO2 ratio is not a permanent fingerprint of one emitter. Wind, mixing, sunlight, background ozone and travel time help determine the chemical balance. A station measuring a high NO fraction near a road and another measuring more oxidised nitrogen downwind are not automatically measuring contradictory emission inventories.
A simple example illustrates that transformation is not the same as new nitrogen. Imagine an air parcel whose NO falls from 35 to 20 ppb while NO2 rises from 15 to 30 ppb, with no loss, gain or conversion to other reactive nitrogen in this intentionally simplified model. Total NOx stays 50 ppb throughout, but the reported NO2 doubles. The unchanged sum shows that a chemical redistribution can alter the molecule of interest without increasing this simplified total. Real parcels also mix and react further, so do not generalise the conservation assumption to an actual site.
For any observed hourly spike, the responsible question is what additional data could distinguish a local emission event from transported or chemically transformed air. Useful independent evidence may include meteorological measurements, traffic conditions, co-located ozone, other nitrogen channels, time-of-day patterns and an appropriate dispersion analysis. A single gas concentration cannot identify a specific vehicle, business or neighbourhood source. Scientific translation means keeping a measured concentration, a supported association and a causal attribution in separate sentences.
Sampling lines, residence time and validation are part of the result
Air must travel from an outdoor inlet through tubing and conditioning components to the detector. Residence time, leaks, material surfaces, moisture and filter maintenance can change how well a sampled quantity represents the air outside. NO2 and other reactive nitrogen species do not necessarily behave identically in a poorly maintained sampling path. A clean instrument zero check conducted only downstream of the inlet may miss an upstream problem. The entire sampling system, not just the optical detector, must be included in meaningful site-quality review.
Suppose two nominally collocated analyzers differ by 6 ppb during moist mornings but agree after lunch. Before declaring one converter biased, check whether their inlet heights, tube lengths, dryers, operating temperatures, time alignment and maintenance records are comparable. Agreement later in the day does not establish the reason for morning divergence. A useful diagnostic plan changes one potentially important condition at a time, records it and repeats checks with independent evidence. This avoids attributing every discrepancy to a popular interference mechanism.
EPA’s ambient monitoring quality-assurance overview explicitly recognises independent assessment, technical audits, siting criteria and validated data. Those elements explain why an accredited measurement programme needs more than a working analyzer display. A method has an instrument, a sampling environment, a recorded chain of checks and an authorised data product. Readers should ask whether all four match the conclusion being drawn. Missing maintenance context justifies more limited wording, not an invented correction.
Two independent instruments can disagree without either being useless
An informative comparison gives exact instrument types, clock alignment, sampling location, averaging interval and quality flags before calculating a difference. A direct optical NO2 instrument and a converter-based chemiluminescence instrument may not share the same interference pathways. A disagreement can reveal selective chemistry, but it can also reveal calibration or siting differences. The best comparison begins with controlled co-location and acceptance of both datasets, then asks whether residual differences track oxidized nitrogen chemistry, humidity or a method change.
Imagine ten paired, valid hourly values with a mean direct-method NO2 of 24 ppb and a mean converter-method apparent NO2 of 29 ppb. The absolute mean difference is 5 ppb. Relative to the direct-method mean, it is about 20.8 percent; relative to the converter-method mean, about 17.2 percent. Neither percentage is automatically the converter’s universal bias. Those are summary statistics of one constructed sample. Distribution, temporal correlation, instrumental uncertainty and chemical context determine how far the result can be transferred.
The same lesson applies to a trend break when a monitoring station changes analyzer model. Plot the method history alongside the concentration series and, where feasible, collect overlapping comparison data. If a clear discontinuity follows an equipment change, test the equipment explanation before announcing a policy success or failure. Do not silently splice methods while implying uninterrupted comparability. The EPA method-code memorandum provides a concrete example of why the method identity belongs with the observation.
Three further transfer problems, with the assumptions visible
Problem one: a valid paired interval reports NOx 84 ppb and NO 63 ppb. The apparent NO2 is 21 ppb. If a station later supplies a separately assessed potential positive interference of 3 ppb for this specific interval, the simplified estimated true NO2 would be 18 ppb only under the explicitly stated assumption that other biases are negligible and that such a correction is authorised. The percentage overestimate relative to the hypothetical true value is 3/18 = 16.7 percent. Relative to the originally reported value it is 3/21 = 14.3 percent. Both must be labelled.
Problem two: a raw converted-minus-direct calculation gives 2.2 ppb and independently estimated standard uncertainties of 1.8 ppb for the converted reading and 1.2 ppb for the direct reading. Assuming independent uncertainties, the standard uncertainty of the difference is approximately square root of (1.8 squared plus 1.2 squared) = 2.16 ppb. A difference of 2.2 ppb is therefore comparable to its estimated uncertainty. It would be misleading to claim a sharply determined low concentration without the network’s complete uncertainty and data-validation context.
Problem three: someone compares a reported 30 ppb hour in one city with a reported 45 micrograms per cubic metre hour elsewhere and concludes the latter is 50 percent higher. That conclusion fails immediately because the units differ. At the illustrative 25-degree-Celsius, one-atmosphere conversion described above, 30 ppb corresponds to approximately 56.4 micrograms per cubic metre, which is numerically higher than 45. Whether the two observations are scientifically comparable is another question involving reference conditions, methods, sites and time intervals. Correct unit translation is necessary, not sufficient.
Problem four: a monitor records NOx 50 ppb and NO 18 ppb at noon, and NOx 50 ppb and NO 30 ppb at night. Ideal inferred NO2 changes from 32 to 20 ppb while inferred NOx stays constant. The data demonstrate a difference in component proportions, not an emission change by themselves. Independent wind and photochemical information would be required to make a more ambitious causal claim. This transfer problem encourages readers to explain what the available channels prove before drafting a story about pollution sources.
A reader-ready interpretation checklist and a final limitation
Before using an ambient NO2 figure, write down the measured or derived species, the instrument principle, the converter type, units, averaging window, data-validation status and applicable flags. Then record whether a method change, chemical selectivity issue or inlet problem could materially affect the inference. If the reported NO2 came from converted NOx minus direct NO, say so. If the instrument directly measured NO2, describe that principle instead. Both are more informative than a naked number copied from a coloured dashboard.
A sound short conclusion distinguishes data from hypothesis. For instance: ‘The validated instrument reported an apparent 27 ppb NO2 for this interval from NOx-minus-NO chemiluminescence. Because the method used a converter, interference and efficiency are relevant quality questions; this record alone cannot apportion emissions among nearby sources.’ That wording is not timid. It describes exactly the chain of evidence, how far the measurement can be used and which further observation would sharpen the answer.
People sometimes treat a long technical explanation as proof of measurement quality. It is not. Scientific quality comes from appropriate measurement design, traceable controls, transparent uncertainty and independent checks. A lesson can train the reader to spot where those are missing, but it cannot retrospectively validate a faulty record. That is the governing idea behind every worked case in this guide: preserve the distinction between what was sensed, what was calculated, what was assumed and what an independent test actually supports.
Transfer exercise: read a multi-channel record
A matched valid record gives NOx = 62 ppb and NO = 37 ppb. Under the ideal selective-converter interpretation, apparent NO2 = 62−37 = 25 ppb. Reverse check: NO + apparent NO2 = 37+25 = 62 ppb. If a later assessment finds a positive converted-path interference equivalent to 4 ppb, a simplified teaching adjustment would reduce the inferred true NO2 to 21 ppb, but only if that interference is independently established and the method allows the adjustment.
The percentage of apparent NO2 attributable to that hypothetical 4 ppb is 4/25 = 16 percent. Relative to the corrected hypothetical value 21 ppb, the bias would be about 19 percent. Stating only ’16-percent error’ without denominator and model would be incomplete.
The challenge question is whether a public user can apply that adjustment to unrelated station data. Answer: no. The interference was invented for the exercise and is not a universal property of converter instruments.
Transfer exercise: a nearly zero difference
An hourly NOx result is 5.1 ppb and NO is 5.3 ppb. Direct subtraction yields −0.2 ppb apparent NO2. It would be wrong to interpret this as a negative amount of NO2 molecules in the air. At low signal levels, offsets, analytical uncertainty, channel timing and validation rules require review. The correct treatment depends on the network’s data policy.
Suppose someone replaces all negative calculated NO2 readings with zero before calculating a long-term mean. That changes the distribution and can bias the result upward. Unless a governing rule explicitly prescribes that transformation for the intended statistic, preserve original values and flags and document any separate presentation choice.
The lesson transfers to every inferred quantity obtained by subtracting two similar measurements. Difference signals require special care because small errors in large parent terms can dominate the result.
The practical conclusion from a monitoring record
A strong report might say: ‘For the stated valid hour, the converter-based chemiluminescence analyzer reported 48 ppb NOx and 21 ppb NO, giving an apparent NO2 value of 27 ppb. Converter selectivity and quality checks govern whether this should be interpreted as ambient NO2 without material positive interference.’ That statement names the method and the logical limitation.
Add site context and avoid assigning source ownership from one gas channel. A traffic-influenced site, an urban background monitor and a regional photochemical station may record different NOx patterns even with perfectly functioning instruments. Data should answer a defined question, not create a false story of universal neighbourhood exposure.
For an official regulatory or public-health conclusion, use the appropriate local authority’s validated records and current criteria. This article explains how to read evidence; it does not set new thresholds.
Primary technical sources and eduKate learning links
EPA’s method-code guidance for nitrogen oxides analyzers distinguishes chemiluminescence, photolytic and direct NO2 methods. The EPA ambient monitoring QA programme, PAMS monitoring description and gaseous data-quality assessments provide additional first-party context.
Related eduKateSG reading includes the ozone UV photometry and averaging guide, bridging inference, epistemic stance and research data management.