Start with what the filter actually measures
PM2.5 filter gravimetry measures the mass gained by a preweighed filter after a specified volume of ambient air has passed through a size-selective sampler. Its simplest equation is final conditioned filter mass minus initial conditioned filter mass, divided by accepted sampled air volume. Translating a PM2.5 result requires checking the sampler’s size-selection behaviour, filter handling, temperature and humidity conditioning, weighing precision, valid flow history, blank observations and sampling interval. The result is mass concentration, usually micrograms per cubic metre, not the count of individual particles or a direct chemical composition.
A daily result of 18 micrograms per cubic metre is meaningful only with its method and time window. A filter may have collected during a nominal 24-hour interval, but a shutdown or flow excursion can change the valid period. Moisture on the filter can mimic particle mass if pre- and post-weighing environments differ. Conversely, some semi-volatile particulate species can be lost from the collected deposit. Good interpretation does not hide these physical realities behind a single precise-looking decimal.
Before using a gravimetric PM2.5 number, write down the filter identifier, initial and final masses and units, actual approved sampling volume, dates and elapsed minutes, conditioning details, flow checks and relevant quality flags. Separate the legally defined reference-method calculation from research-specific blank corrections or exploratory mass adjustments. The purpose of this article is to teach that separation, calculate defensible concentrations and identify what the number can and cannot tell us.
What PM2.5 names and what it does not
PM2.5 denotes particulate matter selected by an inlet with a specified aerodynamic performance around a nominal 2.5-micrometre cutpoint. It is not a literal sieve that accepts every perfectly spherical particle at 2.49 micrometres and rejects every particle at 2.51. Aerodynamic diameter reflects how particles behave in air under specified conditions and depends on particle density and shape as well as geometric size. A sampler’s certified or validated performance defines the selected population.
The filter collects the airborne material that survives transport through the inlet and collection train during the sample. Inlet contamination, flow shifts, particle bounce, line losses and changing aerodynamic selection can alter what reaches the substrate. Such problems cannot be corrected later merely by measuring the filter more accurately. It is possible for a laboratory balance to deliver highly repeatable masses from a filter that does not represent the intended aerosol population.
PM2.5 concentration is also not a toxicity score. Particle composition, morphology, source and chemical reactivity can differ strongly between locations and days with the same mass. Likewise, two PM2.5 samples with identical chemistry may have different particle numbers and size distributions. The mass is a useful common measure because it can be determined through gravimetry, but interpretation becomes misleading if it is portrayed as a complete physical or health characterisation.
Why preweighing and postweighing must be comparable
A clean filter’s initial mass is sometimes called the tare. After sampling, its final mass includes collected material plus any storage or handling effects. The difference is useful only when weighing conditions are consistent. Filters can absorb or lose water, accumulate electrostatic charge, pick up laboratory contamination or change their own substrate mass. These effects are particularly important because collected particle masses may be only hundreds of micrograms on a substantially heavier filter.
Regulatory reference methods prescribe environmental conditioning and quality checks rather than leaving each technician to weigh whenever convenient. The United States PM2.5 federal reference method in 40 CFR Part 50 Appendix L specifies a 20–23 degrees Celsius mean conditioning temperature, generally 30–40 percent relative humidity, and not less than 24 hours of conditioning before the prescribed weighings, with method-specific controls and exceptions. These numbers are method-specific requirements, not universal settings for every sampler worldwide.
The practical lesson is comparability. If a preweighing was made on a dry, stabilised substrate but the final weighing occurred just after carrying the filter from a humid corridor, the apparent mass gain can include transient water. A well-operated gravimetric laboratory documents environmental logs, conditioning duration, balance performance and filter handling. Without that record, a reported change in air quality may actually be a change in the weighing room.
Construct the concentration with units
The gravimetric equation is C = (Wf − Wi) / V, where C is aerosol mass concentration, Wf is final conditioned filter mass, Wi is initial conditioned filter mass, and V is the accepted volume of ambient air sampled on the applicable basis. If Wf and Wi are micrograms and V is cubic metres, C is micrograms per cubic metre. Parentheses matter: the mass difference is computed first, then divided by the volume.
Imagine an initial filter mass of 150.345 milligrams and a final mass of 150.780 milligrams. The gain is 0.435 milligrams or 435 micrograms. If accepted volume is 24.0 cubic metres, the reported illustrative concentration is 435 / 24.0 = 18.125 micrograms per cubic metre. The digits ultimately displayed should reflect the reporting and measurement rules; 18.1 may be a sensible rounded presentation, while 18.125 shows the transparent arithmetic.
Someone who divides 0.435 by 24 and labels the result micrograms per cubic metre has mixed milligrams with micrograms. Their value, 0.018125, is in milligrams per cubic metre, equivalent to 18.125 micrograms per cubic metre. The arithmetic is not wrong if the unit is stated correctly; the translated label is wrong. This simple example explains why retaining a units column is as important as retaining the numerical results.
Air volume must be accepted, not guessed
A sampler might operate near 16.7 litres per minute. Over exactly 1,440 minutes, constant flow at that rate gives 16.7 × 1,440 = 24,048 litres, equal to 24.048 cubic metres. This is a useful dimensional check, not proof that the instrument actually collected that volume. A flow controller may vary over the day, and a site’s accepted volume may depend on integrated readings and sampler-specific calibration corrections.
Consider a monitor that collected for 23.5 hours at an observed average flow of 16.5 litres per minute. Its illustrative volume is 16.5 × 1,410 / 1,000 = 23.265 cubic metres, not 24.048. If the net collected filter mass were 400 micrograms, the concentration using the measured volume would be about 17.19 micrograms per cubic metre. Dividing by the nominal 24.048 would give about 16.63, a difference driven by denominator choice alone.
Review the complete flow record, flow calibration and any instrument flags before selecting V. A sudden zero period, extreme temperature condition or leak may make parts of the record unacceptable under the method. Be careful with conversions between actual ambient volume and standardised volume: different methods can use different conventions. Never convert to standard conditions and then silently label the result as the actual-volume concentration required by another method.
The time window is part of the observation
A filter’s integrated mass is the inventory collected over the accepted sampling interval. It does not tell us when the particles arrived during that interval. A 24-hour filter could include a short wildfire plume, several traffic peaks, a clean overnight period and a windy afternoon. The final concentration is a period average on the sampling basis; it is not evidence that every minute had the same air quality.
Under the US PM2.5 reference method, a fully valid normal sample generally has a sampling duration between 1,380 and 1,500 minutes, or 23 to 25 hours, with additional provisions for exceptions and flags. These are regulatory-method conditions, not an instruction to reject every shorter research sample in another programme. Always identify the actual governed method before deciding validity. A sample’s number may be useful scientifically yet fail a particular compliance protocol.
When comparing an integrated filter with an hourly optical sensor, aggregate the time-resolved data over matching valid hours. An hourly sensor’s maximum should not be compared to a 24-hour filter mean as if the time supports were identical. Also inspect whether an instrument reports local time, UTC or a daylight-saving adjustment. An unrecognised date boundary can shift a paired comparison by many hours and create a false calibration dispute.
Filter conditioning is a physical control
Humidity changes the mass of hygroscopic material and sometimes of the filter substrate. The relationship can exhibit hysteresis: particles that have absorbed water may not lose it instantly when surrounding humidity decreases. Conditioning helps establish a reproducible weighing state, reducing unwanted mass differences caused by the room rather than the atmosphere. The laboratory’s specified humidity window and acclimatisation time therefore belong to the measurement method.
Temperature affects equilibration, convection around the balance and sometimes volatile sample constituents. Static charge can produce unstable balance readings, especially on low-mass filters. The US reference method calls for a suitably sensitive analytical balance, controlled conditioning, and techniques to neutralise static. It also directs comparable pre- and post-weighing conditions and, where feasible, use of the same balance and analyst. These details matter because the quantity of interest is a small mass difference.
A common misconception is that aggressive heating would give a purer ‘dry PM2.5’ weight. It can instead drive off legitimate semi-volatile particulate matter, changing the defined measurand. A valid method does not simply dry everything as hot as possible. It chooses a controlled condition and reports the resulting operational mass. If the objective is thermodynamic particle water content, that is a different experiment requiring additional information.
Laboratory blanks and field blanks answer different questions
A laboratory blank generally remains in controlled laboratory handling rather than going through normal field exposure. It tests whether storage, preparation, weighing and related procedures are introducing mass changes. A field blank travels to the site and is installed and recovered according to the quality plan without collecting a normal air sample. It can reveal contamination or mishandling associated with the field process. Both are valuable checks, but they expose different stages.
Some research programmes apply a defined blank correction to sample mass. In the US PM2.5 federal reference gravimetric calculation, however, field and laboratory blanks are quality-control tools and the prescribed concentration formula is based on the difference between the sample filter’s final and initial masses divided by sampled volume. Do not replace that regulatory equation with an ad hoc average-blank subtraction. If a separate research correction is used, label it and explain the quality plan and measurement objective.
For illustration only, a research dataset has sample gain 435 micrograms, mean field blank gain 15 micrograms and V = 24 cubic metres. Its policy-defined blank-adjusted result would be (435 − 15) / 24 = 17.5 micrograms per cubic metre, compared with the unadjusted sample equation giving 18.125. Neither number should be silently substituted for the other. Which one is appropriate depends on whether the governing method permits and requires that adjustment.
Why blank variability matters even when the mean is small
Suppose five field blank changes are +2, +7, −3, +5 and +4 micrograms. Their mean is +3 micrograms, but their scatter matters. A genuine sample gain of 300 micrograms is less sensitive to this blank behaviour than a sample gain of 12 micrograms. Quality control should therefore consider both the typical blank drift and its dispersion relative to real sample loadings. A tiny average blank does not prove that every individual handling event is harmless.
A field blank can also detect localised unusual contamination: a technician may accidentally touch one filter or open its cassette near a source of dust. Such an outlier should be investigated, not automatically averaged away. Look for handling logs, container integrity, instrument records and matching laboratory blank results. A cleaning or retraining response may be appropriate, but retrospective sample exclusion must follow objective programme rules rather than a wish to improve a plotted trend.
If a sample’s apparent mass gain is negative, this is not a physical negative PM2.5 atmosphere. It may reflect substrate loss, humidity difference, balance behaviour or a small change dominated by noise. Keep the actual mass observations for the audit and apply the programme’s data-validation or qualification scheme. Simply changing the concentration to zero without a record can conceal recurring problems in filter handling and bias statistics at clean sites.
Handling a filter without adding a particle
A filter must be tracked from supply lot through conditioning, initial weighing, sampling, transport, final conditioning and final weighing. Each transfer is an opportunity for contamination or material loss. Operators should use validated clean handling procedures, identifiable containers and a chain-of-custody record. A filter placed in the wrong cassette or swapped in a stack can produce a mathematically plausible but meaningless mass difference.
Physical damage is especially troublesome. A folded filter may shed fibres or disrupt uniform deposition. An incorrectly seated filter may allow particles to bypass the medium. A tear may admit material to the wrong surface or allow loss in transport. In the laboratory, visual inspection should flag defects before analysts choose whether a sample meets method requirements. The balance cannot detect which of two visually similar filters actually saw the air.
Transport temperature and storage time deserve consideration too. Semi-volatile nitrate and organic constituents can change after collection if storage is warm or prolonged. Regulatory methods have specific time constraints and conditional extensions for cooled samples. A delayed weighing may remain permitted under the stated method, or it may require a qualification; the right answer depends on the record. Never invent a storage history because the observed mass looks reasonable.
An uncertainty budget begins at the balance
If final and initial masses each have uncertainty, their difference can be less certain in relative terms than either individual mass. The same balance and environmental state can introduce correlated components; simplistic independent-error assumptions may overestimate some terms and miss common systematic biases. The uncertainty of the sample mass difference must also be distinguished from uncertainty of the airflow integration and losses in sample collection.
As a simple independent-error exercise, suppose the standard uncertainty of the mass difference is 8 micrograms and the relative standard uncertainty of accepted volume is 2 percent. For a 400-microgram gain and 24-cubic-metre volume, mass-relative uncertainty is 8/400 = 2 percent. If the terms are independent, combined relative uncertainty is square root of 2 squared plus 2 squared = 2.83 percent, and the concentration 16.67 micrograms per cubic metre carries an illustrative standard uncertainty of about 0.47 micrograms per cubic metre.
That calculation deliberately omits biases such as volatilisation, sampler size selection, substrate changes and systematic calibration error. A more complete uncertainty budget may be necessary for interlaboratory comparison or regulatory decisions. Do not claim an uncertainty of ±0.47 as a universal method property. It belongs only to the hypothetical inputs and independence assumption. Present accuracy limitations in words alongside the numerical propagation, especially when the study has not measured every contribution.
Micrograms per cubic metre is not a microscopic particle count
An equal total mass can consist of very many tiny particles or fewer larger ones. Particle number concentration is often expressed as particles per cubic centimetre, while gravimetric mass is usually micrograms per cubic metre. Converting between them demands assumptions about size distribution, effective density and morphology. Dividing a mass concentration by a single representative particle volume without evidence can lead to wildly misleading particle counts.
For example, imagine one set of particles has twice the diameter of another, both spherical and of equal density. One particle’s volume and mass are eight times as large. Equal total mass therefore requires about eight times as many of the smaller particles. Real atmospheric aerosols are more complicated, with multiple size modes and nonspherical shapes, but the illustration shows why mass and number cannot be treated as interchangeable columns in a dataset.
Related eduKateSG explanations about mobility sizing, aerodynamic particle sizing and cloud condensation nuclei address different properties. A filter mass can complement those observations but does not secretly contain their number distributions. If a public dashboard presents PM2.5 mass beside a particle count, check the labels and methods before inferring which direction human exposure has changed. The underlying measurements answer distinct questions.
Physical losses and chemical changes are not the same error
Particle transport losses occur when airborne material deposits on inlet walls, bends or tubing before reaching the filter. Chemical loss can occur when a semi-volatile constituent evaporates from deposited particles or transforms during sampling or storage. Both can lower a measured filter mass relative to an intended atmospheric measurand, but their remedies differ. Replacing a bend may reduce transport losses; it will not automatically stop ammonium nitrate from partitioning between particle and gas phases.
Positive artefacts also occur. Gas-phase compounds can adsorb to filters or deposits, increasing apparent particulate mass under certain conditions. Water retention can inflate weight if conditioning is inadequate. Laboratory dust or filter handling can add material that was never present in the intended air stream. Analysts should distinguish a physical aerosol episode from these artefacts using method-specific controls and, where available, paired measurements.
It is poor practice to apply a universal ‘PM2.5 filter loss correction’ taken from another season and sampler. A correction should be linked to demonstrated mechanism, operating conditions and validation. For education, label hypothetical effects separately: mass collected, mass lost upstream, semi-volatile mass lost after deposition, and spurious mass added by handling. This causal ledger is more informative than a single unexplained adjustment factor.
Compare filter gravimetry with a continuous sensor properly
Continuous PM monitors may infer mass from light scattering, beta attenuation, oscillation frequency or another physical proxy. Some instruments incorporate heating or humidity control that affects volatile and water-associated material. A reference filter method uses direct mass difference under its specified procedure. Agreement can vary with aerosol type, particle optical properties, temperature, relative humidity and temporal averaging.
Suppose a continuous instrument reports hourly values over the same 24-hour window as a filter. Compute the average of valid hourly values using the appropriate time weights, not the maximum or a selectively chosen subset. Then compare this aggregate to the filter concentration and inspect missing intervals and instrument calibrations. If the continuous mean is 22 while the filter is 18 micrograms per cubic metre, the four-unit gap needs investigation; it is not automatically evidence that one monitor is broken.
A helpful comparison records which quantity each monitor operationally measures, the matched sampling interval, its humidity handling and its quality controls. Differences can be scientifically revealing, for instance when volatile loss affects a warmed system or optical response changes with aerosol refractive index. A broad category such as ‘PM2.5 sensor’ does not guarantee method equivalence. Translation requires keeping the physical principle attached to every figure.
Worked example: an incorrect unit and a plausible chart
A file records initial weight as 0.150345 grams and final weight as 0.150780 grams, with accepted V = 24.0 cubic metres. Difference is 0.000435 grams. Multiplying by one million converts grams to micrograms, giving 435 micrograms. Concentration is 18.125 micrograms per cubic metre. Someone who multiplies by one thousand instead obtains 0.435 and mistakenly labels the chart 0.018125 micrograms per cubic metre.
The resulting plot may still have convincing day-to-day peaks and troughs because every value suffered the same factor-of-one-thousand error. Visual smoothness is not a validation check. Compare a raw filter mass difference to the concentration equation and retain dimensional units at each step. Once the conversion is corrected, all affected numbers change consistently while their relative day-to-day pattern remains intact.
The opposite error is also possible: reporting 18.125 milligrams per cubic metre when the calculation produced micrograms. That is a thousand-fold overstatement of particulate mass. Anyone translating an air-quality table for non-specialists should verify these prefixes before writing about severity, comparison to guidelines or regional differences. The smallest line in the header may control the most consequential claim in the article.
Worked example: a denominator with missing intervals
Assume an instrument collected particles at 16 litres per minute for 600 valid minutes, experienced a 60-minute outage, then ran at 15 litres per minute for 780 valid minutes. The accepted sampled volume is (16 × 600 + 15 × 780) / 1,000 = (9,600 + 11,700)/1,000 = 21.3 cubic metres. An analyst who uses the entire 1,440-minute clock period at 16 litres per minute assumes 23.04 cubic metres, which includes air that was never sampled.
If a filter gained 360 micrograms, the volume-aware concentration is 360/21.3 = about 16.90 micrograms per cubic metre. The incorrect nominal-volume calculation yields 15.625. A roughly 7.5-percent understatement here arises without any laboratory balance mistake. This example is illustrative; actual sample acceptance also requires applying the governing method’s duration, flow and outage validity criteria.
The repair is to preserve the time-stamped flow log and reconstruct the accepted volume from validated readings. Do not simply subtract one hour at the nominal flow if the other intervals were also operating differently. In a real monitoring project, instrument-provided volume may already incorporate the necessary integration. Review its calculation and traceability before reproducing it manually in a spreadsheet.
Worked example: blank influence near a clean-air day
Imagine a filter gain of 27 micrograms for 23 cubic metres and field blanks with changes of +8, −2 and +5 micrograms. The unadjusted sample concentration is 27/23 = about 1.17 micrograms per cubic metre. The blank changes are not negligible relative to that small gain. A research protocol might describe a mean blank of (8−2+5)/3 = 3.67 micrograms and separately estimate the net, but the uncertainty and validity of such a correction depend on the sampling programme.
If an unofficial user subtracts the largest field blank, 8 micrograms, without any method justification, the result becomes 19/23 = 0.826. That is a substantial change in the apparent clean-air concentration produced by a discretionary choice. Replacing the blank with the smallest observation could change the result in the other direction. A careful report states the raw gain, blank distribution, adopted policy and whether this is a regulatory or research result.
When the blank scatter approaches the sample mass, a strong numerical claim may not be supported even if the balance reads to micrograms. Resolution is not the same as analytical confidence. A small mass is not inherently invalid, but it demands more attention to handling, chamber control and repeatability. Readers should carry qualifiers through subsequent summaries rather than pretending that every numerical row has equal evidential strength.
Worked example: separate a real PM episode from humidity
Two sampling days produce filter mass gains of 480 and 620 micrograms on near-identical air volumes. A headline calls the second a 29-percent pollution surge because 620/480 − 1 = 0.2917. Before drawing that conclusion, inspect whether both filters underwent comparable conditioning and the balances were stable. If the second postweighing occurred after only a brief exposure to unusually humid air, the evidence for atmospheric change is compromised.
This does not mean humidity automatically explains the entire 140-microgram difference. It means the weighing procedure was not demonstrably equivalent. Repeat conditioning and weighing under the governing method, review chamber logs and inspect any damaged or contaminated filter. If the corrected observation still increases and sample quality checks pass, then a genuine difference in collected mass becomes more defensible. A method check establishes the measurement before the environmental explanation is offered.
The example also highlights the danger of changing measurement rules after seeing an inconvenient result. A laboratory should apply predetermined conditioning and reweighing rules consistently to all affected samples. Selectively reweighing only the high day creates a biased comparison. The aim is not to make two numbers agree; it is to ensure both numbers arise from the same validated process.
Mass closure: chemical components are additional measurements
Filter gravimetry provides total collected PM2.5 mass under defined conditions. Chemical speciation can then estimate how much is associated with sulfate, nitrate, ammonium, organic material, elemental carbon, mineral elements and other fractions. The sum of independently measured or reconstructed component masses may not match gravimetry exactly because of unmeasured species, conversion assumptions, water, artefacts and uncertainties.
Consider gravimetric PM2.5 mass 20 micrograms per cubic metre. A reconstructed chemical total of 18 does not prove that two micrograms of a named compound were present. It may reflect water, unmeasured constituents, differences in sampling channels or conversion factors. Conversely, a reconstructed total of 22 is not physically an atmospheric particle with negative remaining mass; it flags possible measurement differences, duplicate accounting or bias. Mass closure is a diagnostic comparison rather than an equation with one guaranteed answer.
The verified eduKateSG aerosol-filter XRF guide explains why elemental amounts need their own blanks and air-volume basis. The thermal–optical carbon guide explains why OC and EC are method-defined fractions. Combining these with a gravimetric measurement is valuable only after harmonising sample period, collection size fraction, units and reconstruction assumptions.
What a long-term trend can and cannot establish
A ten-year PM2.5 gravimetric trend may be a valuable record of environmental change. But the series should be audited for sampler replacements, network changes, filter medium, calibration, laboratory procedures, inlet maintenance and data-validation rules. A site relocation, a new building nearby or a modified sampling schedule may change the observed population even if air emissions are constant. Trend interpretation should include documented station history.
Weather can also influence daily mass. Stable atmospheric conditions may accumulate emissions near the surface, while rain or wind may reduce concentrations or change the source mix. A lower annual mean does not prove that every emitting activity became cleaner, and a high day does not prove one named source was responsible. These causal questions require supporting emissions, meteorological or chemical evidence.
If the purpose is public communication, a clear statement names the measure, interval and uncertainty: ‘This station’s validated 24-hour PM2.5 gravimetric concentration increased between the two sampling days.’ A second sentence can discuss possible explanations, explicitly conditional on supporting observations. Avoid replacing ‘measured concentration increased’ with an unqualified cause, just as one should avoid making the filter number itself sound more toxic or less toxic than the evidence supports.
Why QA rejection and data qualification differ
A failed quality requirement can invalidate a sample for a particular use. A flagged but analysable sample may remain useful for a research question with appropriate qualification. Examples include significant flow deviations, questionable conditioning, damaged filters, unexplained contamination or missing chain-of-custody information. The outcome depends on the method and predefined quality plan, not on whether the result supports an expected trend.
Do not erase a bad sample and leave no trace. Preserve its identifier, raw weights, collection period, logs and validation decision. A dataset might carry separate fields for numeric concentration, status, qualification reason and use recommendation. This design prevents a plot or machine-learning workflow from unknowingly treating a provisional figure as equivalent to a validated reference measurement.
At the same time, excessive exclusion can bias a dataset. If high-loading episodes more often produce sampler flow flags, dropping all flagged days without examining the missingness pattern could underrepresent pollution episodes. Follow the defined acceptance criteria, but describe the completeness and potential selection bias in research reporting. Measurement quality and population representativeness are related questions, not a single checkbox.
Transfer exercise: calculate without losing significant meaning
A practice record lists initial mass 0.212564 grams, final mass 0.212941 grams and valid sampled volume 22.5 cubic metres. The mass gain is 0.000377 grams or 377 micrograms. Dividing by 22.5 gives 16.7556 micrograms per cubic metre, conventionally about 16.8 after suitable rounding. If a student writes 0.0000168 with a microgram unit, they have carried grams numerically into the wrong label.
Now suppose the same record shows a flow-calibration uncertainty of 2 percent and a repeat-weighing uncertainty of 6 micrograms for the mass difference. The relative weighing uncertainty is 6/377 = 1.59 percent. Under a simplifying independent-components assumption, combined relative standard uncertainty is square root of 1.59 squared plus 2 squared percent, about 2.56 percent. For 16.76 micrograms per cubic metre that is about 0.43. This does not include artefacts or filter representativeness.
The correct answer to the exercise is not merely 16.8. It is a measured average under a stated sampling interval, with an approximate simplified uncertainty and additional possible systematic limitations. A calculator can perform division, but it cannot infer an unrecorded sampling failure. Technical literacy requires both the numerical work and the ability to explain what the input record can justify.
Transfer exercise: compare two monitoring systems fairly
A filter gives 12.0 micrograms per cubic metre over 24 hours. A continuous sensor has six valid hours at 8, six at 10, six at 14 and six at 16 micrograms per cubic metre. Its matched 24-hour mean is (8 + 10 + 14 + 16) / 4 = 12.0, because each value represents the same six-hour duration. The two methods agree on the average in this example, even though hourly conditions varied considerably.
If only the sensor’s highest interval, 16, is chosen for comparison, a false disagreement appears. If the sensor’s last six hours were missing, simply averaging the remaining three six-hour values and calling that a complete daily mean would introduce an observation-coverage problem. A fair comparison requires aligned periods, valid sampling and method-specific uncertainty. Agreement of two means is encouraging but is not proof of uniform performance under every aerosol condition.
The same lesson applies to educational assessments or business dashboards: a whole-period average and a short-window peak are different quantities. Translation means preserving the denominator, duration and coverage, not choosing whichever number makes the conclusion more memorable.
What to ask a laboratory before relying on the table
Which exact gravimetric sampling method and size-selective inlet were used? What filter medium and balance resolution were involved? Were temperature and humidity conditioning records within the applicable method’s requirements? Were pre- and postweighings traceable to the same filter identity? What sampling volume basis was used? Which sampler flow flags, field blanks and laboratory blanks were reviewed? Was the result calculated from accepted measurements rather than a nominal run time?
Ask whether the reported figure is a regulatory reference-method result, a research-adjusted concentration or an unvalidated preliminary calculation. If blanks were subtracted, ask for the written rationale and supporting quality controls. If a number was censored, reweighed or excluded, ask for the reason and whether the same decision rule applied to all other samples. The answers make the result reproducible to an informed reader.
A good report can be concise while still containing essential provenance. Tables need method names, units, averaging periods, uncertainty or qualifiers and explanatory notes. The reader should be able to reconstruct at least one example from raw masses and volume. If a report relies only on a clean chart and a dramatic caption, the route from measurement to claim is too fragile.
Questions readers often ask
Can PM2.5 mass be negative? Physically not as a true concentration; a negative mass change or blank-corrected value is an analytical outcome requiring review or qualification. Is water part of the measurement? The operational mass depends on how a method conditions filters; uncontrolled water is an artefact, but ambient water and semi-volatile content require careful method-specific interpretation. Are two filters with identical mass the same pollution? No, their chemistry, size distributions and sources can differ.
Does the filter identify traffic or wildfire smoke? Not by mass alone; chemical tracers, meteorology and other evidence are needed. Why not use a smartphone particle sensor as a direct replacement? Different devices may infer mass from light and require calibration and humidity considerations. What does a field blank prove? It can reveal aspects of handling and contamination, but it does not guarantee the sample’s inlet and sampling physics were correct.
The durable habit is to translate every number into a sentence that includes how it was acquired, which population it represents, which corrections were applied and what uncertainty remains. In the context of PM2.5 gravimetry, that sentence begins with a preweighed filter, ends with an accepted air volume and avoids attaching causes that the weighing method never measured.
Primary method references and related reading
For the formal US reference-method requirements, consult 40 CFR Part 50 Appendix L: PM2.5 reference gravimetry. The text describes filter conditioning, balance requirements, blank quality checks, sample validity and the defining mass-over-volume equation. Regulatory applicability and later amendments should be checked against the currently effective official rules for the relevant jurisdiction; these are not automatically transferable to every research programme.
For context and comparison, see eduKateSG’s explanations of filter XRF elemental data and thermal–optical carbon fractions. These measurements can inform the chemistry of PM2.5 but should not be substituted for gravimetric total mass without an explicit and tested measurement model.