Editorial Standards

Ownership and sponsorship

IAQng is built, funded, and maintained by SpaceBot, which sells facilities AI to the same organizations this site addresses. IAQng is a SpaceBot publication, not an independent institute, and it is not presented as one anywhere on the site.

Editorial responsibility sits with the SpaceBot team that maintains IAQng. That team decides what gets published, what gets corrected, and what is held back. There is no separate editorial board, and claiming one would misrepresent how the site works.

The conflict this creates is direct: SpaceBot benefits when readers take indoor air seriously. The controls against it are that inclusion is never sold, that source material is linked so a reader can go around IAQng entirely, and that the tools state their assumptions and boundaries. Those controls are disclosed rather than presented as independence.

Sources and citations

Sources are ranked. When two sources conflict, the higher tier wins unless the page says why it does not.

Primary material. Public agency data and statements, published standards, original research, and official incident records. This is the default basis for any factual claim about air quality.
Institutional and expert material. Research institutions, standards bodies, and named domain experts writing in their field.
Reporting. Used for context, timelines, and discovery of primary material. Reporting can establish that an event was covered and how; it is not used as the sole basis for a physical or health claim.
Secondary summaries. Used only where labeled, and not to support a claim the primary source does not already support.

If a page names a person, standard, organization, tool, paper, or practice, a reader should be able to tell why it is there and follow it back to the original. Citations link to the source itself rather than to a summary of it wherever the source is public.

Data status

A source fact, an IAQng calculation, and an IAQng conclusion are not interchangeable, and pages label which one a reader is looking at. Uncertainty and missing data stay visible rather than being smoothed over.

Label	What it means
Certified	The publishing agency has finalized the value. It is not expected to change.
Provisional	Published quickly and subject to revision. Most current air data is provisional. Rankings built on it update when the agency finalizes the underlying data.
Modeled	Output of a model rather than an observation, including forecasts.
Derived	An IAQng calculation applied to a source value, such as converting PM2.5 mass to a cigarette-equivalent illustration.
Editorial grouping	IAQng grouped days, monitors, or metros into one episode. The grouping is a judgment, and the underlying observations are shown separately.
Automated attribution	A cause or source region assigned by the event pipeline from retained evidence, published without per-episode editorial sign-off. Every one links the evidence it rests on and carries an evidence-strength label, from agency determination down to unconfirmed. See event attribution below.
Unattributed	A significant episode the evidence does not resolve to a cause. It is listed as unattributed rather than given a likely-sounding label.
Research preview	Assembled but not surfaced as a finding. Research artifacts behind the published catalog are not cited as IAQng conclusions.

Air Records rank observed PM2.5 and nothing else. A first-place ranking means the particle levels were unusual for that metro; the ranking itself never assigns a cause. Where a ranked day carries an air-event label, that label comes from the catalog and is an automated attribution, not part of the measurement.

Analysis and interpretation

Where IAQng calculates, models, or concludes, it says so in the same place as the result rather than in a note elsewhere. Three things stay distinguishable on the page: what the source reported, what IAQng computed from it, and what IAQng concluded.

Scores, bands, and drivers are versioned, and the action rules a facility might follow are kept outside the score rather than hidden inside it, so a reader can disagree with the recommendation without discarding the measurement. Methodology changes that alter a published result are versioned and dated.

The cigarette-equivalent comparison used on the maps and records is a public-awareness illustration of fine-particle mass, not a medical equivalence. It is labeled as derived wherever it appears and is not presented as a diagnosis, an exposure assessment, or a regulatory measurement.

Air-event attribution

A named air event is a strong claim: it asserts that specific days in specific metros share a cause. IAQng publishes its air-event catalog, and the honest description of how those names are produced is this: episodes are grouped from the monitor record by an automated pipeline and given a cause where retained evidence supports one. They are not reviewed and signed off one by one before publication.

What the pipeline is required to do before it will name a cause:

Derive a stable name, slug, date range, and geographic boundary.
Resolve the affected metros against the monitor record rather than assuming them.
Cite cause and source-region evidence by ID, weighting official assessments above corroboration, hypothesis, and context.
Retain the supporting official and journalistic records rather than summarizing them away, and show them on the episode page.
Abstain where the evidence does not support a cause. Unattributed episodes stay unattributed; unknown means unknown, not "no smoke".
Mark which underlying data is provisional.

What that does not include is a person signing off on each episode. Because every cause links the evidence it rests on, a reader can check any one of them, and a correction is the fastest way to change one. Episode boundaries, causes, and names can change as evidence accumulates or as EPA certifies provisional days. An event that is later contradicted is marked superseded rather than quietly deleted.

Evidence strength

Every named episode carries a label wherever its cause appears, including in the record rankings. The label describes what kind of evidence backs the cause, never how much of it there is: volume can never promote an episode, so one agency record outranks any number of headlines. The distinction that matters most is between a government body describing the event and a government body flagging the measurement, which are both agency evidence but are not the same claim.

Label	What it means
Agency determination	A government body issued a record naming this cause for this event: an NWS air-quality alert, a NOAA Storm Events record, or a state or local official archive.
EPA monitor flag	EPA attached an exceptional-event qualifier to the affected samples. The agency flagged the measurement itself; no separate agency record describes the event.
Documented	An agency or publisher page assesses this cause, without an agency record naming the event or an EPA qualifier on the data.
Corroborated	Independent reporting agrees on the cause, with no agency determination, monitor flag, or published assessment behind it.
Inferred	No evidence names a cause, but the calendar does the work: the episode falls on July 4th or New Year, when fireworks dominate fine-particle readings. IAQng says so rather than pretending not to know, and labels it a guess. This is the only place the site names a cause without evidence, and it is the weakest claim it makes.
Reported	The cause rests on contextual or hypothesis-level material only. Treat it as a lead rather than a finding.
Unconfirmed	The cause comes from grouping the monitor record alone, with no retained supporting evidence. Weakest tier on the site.

Weak episodes are published and labeled rather than withheld, and every significant episode the pipeline could not resolve is listed as unattributed rather than given a plausible-sounding cause. Matching a named fire to an official incident record establishes which fire, not that it caused a given episode, so those matches never raise an episode's evidence strength.

Correlation in timing is not attribution. A smoke episode that overlaps a known fire is not attributed to that fire unless retained evidence connects them. Severity alone, proximity alone, the season, and article counts can never establish a cause, and no PM2.5 threshold is treated as proof of one.

Inference from the calendar

One exception is drawn deliberately. Two dates in the year produce a particle signature that is not seriously in doubt: July 4th and New Year. Where an episode falls on those days and nothing else explains it, IAQng names fireworks and marks the claim Inferred, which is a weaker label than any evidence-backed tier.

The reasoning is that "cause not established" over the 4th of July tells a reader less than a labelled guess does. The rule is narrow on purpose: it applies to those dates only, only to episodes short enough for a date to mean something, and it never overrides an attribution that has evidence behind it. It is the one case where the site names a cause it cannot cite, and it says so on every affected event.

Automation and human review

IAQng uses automation, including AI-assisted research, to collect data, normalize it, discover candidate events, group and compare records, generate metadata, run validation, and prepare research for review. That work is substantial and it is not hidden.

Automated output is not evidence. It is a starting point that has to be checked against a source. The following require human editorial review before publication, without exception:

Any material public claim, including event names, causes, and source regions.
Any rights or licensing decision.
Any interpretation of what a source says.
Any change in publication status, including moving research to published.

Citations are never generated without a retrieved source. A failed lookup is recorded as a failed lookup; it is not converted into a description of what the source probably said.

Media and camera rights

Public access is not permission to reuse. Every camera source, image, and video on IAQng carries its own terms, and those terms are reviewed per provider and per media mode before anything is republished.

Rights are recorded per source, not assumed from the set. Some camera imagery is a government work in the public domain; much of it is not. No blanket license is applied across the camera program.
Provider-native media is preferred over re-encoded or re-hosted copies, and identity, coordinates, timestamps, ownership, and attribution are preserved exactly.
Only assets with documented reuse rights enter press downloads. Media that is viewable on the site is not automatically media that IAQng offers for reuse. The reuse posture of every registered source, and the terms on every offered asset, are published in the asset index, which is generated from the rights registry and validated by an automated check that fails if the two disagree. See press and reuse.
Camera imagery is visual context, not measurement. A camera view is never used as an air-quality reading and is not treated as evidence that smoke is present.
Camera failures are isolated. A camera going dark does not affect environmental data, forecasts, health language, or facility guidance.

Inclusion and endorsement

Being named on IAQng is not an endorsement, and it cannot be bought. Placement in source lists, record pages, and the field pages is an editorial decision, and IAQng does not sell it, trade it, or condition it on a commercial relationship with SpaceBot.

The reverse also holds. Appearing on IAQng does not mean a person or organization supports SpaceBot, IAQng, or any product, and IAQng does not describe anyone as a partner or supporter on the basis of being cited.

Where the site recognizes contributions to the field, it is crediting work, not ranking people or certifying vendors. Where a page names a tool or a company, it is pointing at something a reader can evaluate, not recommending a purchase.

Corrections and material updates

A correction is reviewed against the source it disputes. If it holds, the page changes. What happens next depends on how much the change matters:

Routine fixes to wording, links, or formatting are made without a note.
Material corrections that change a fact, a ranking, a named cause, or a conclusion carry a dated note on the page saying what changed.
Data updates that follow from an agency finalizing provisional values are not corrections. Pages state when a ranking is provisional and that it updates on finalization.
Methodology changes that alter a published result are versioned and dated, and the page says which version produced the number shown.
Retired or superseded material is marked as such rather than deleted, so a citation made in good faith does not become a dead link.

A site-wide correction log is not published yet. It is planned alongside the asset and versioning work described in press and reuse. Until it exists, material corrections are recorded on the affected page.

Submit a correction

The most useful correction names the page, identifies the weak claim or missing source, and points to a better paper, standard, public agency page, archive, project page, or practical example. General disagreement is welcome but harder to act on.