The Gap

What the record does not contain

Environmental conditions

Noise: no readings in the night-limit hours

The Bronx jail site has a noise limit for the hours between six in the evening and seven in the morning. The contractor’s published monitoring reports record no measurement in those hours.

Each monthly report states its schedule: noise is recorded by five stationary meters during weekday working hours, seven in the morning to half past three in the afternoon, and by handheld spot checks. Every report from January to July 2026 says the same.1

Since 21 April 2026 a City rule has required more. Where a new building of 200,000 square feet or more works after hours under a noise mitigation plan lasting thirty days or longer, within fifty feet of a home, a monitor must record twenty-four hours a day and send its readings to the Department of Environmental Protection every hour. The jail’s floor area, about 1.07 million square feet, is more than five times that threshold.2

Whether the rule reaches this site cannot be determined from any public record. Nothing published says whether the site holds a noise mitigation plan or for how long, whether its work lies within fifty feet of a home in the rule’s terms, or whether an earlier plan predates the rule. And where the rule does apply, the readings go to DEP and are published nowhere.3

The limit is stated, and a rule now requires the measurement in some cases. The public record shows a weekday window, a requirement whose reach cannot be checked, and — where it does reach a site — a stream of readings with no public end.

The Department of Buildings authorized work at the jail site through the night on thirteen consecutive nights, 14 to 27 September 2026, under a single after-hours variance (X5515292), each window flagged as within 200 feet of a residence.4

The independent monitor’s weekly summaries print Lmax ranges reaching 90.4 dBA and 95.7 dBA, above the 85 dBA Lmax limit the same reports state for the weekday daytime hours, seven in the morning to six in the evening. The logs beneath them record Leq, not Lmax, so no published reading backs either peak, and the exceedances cannot be counted. The same summaries state that no exceedances were “determined to be caused from project-related activities.”5

Dust has the same shape. Four separate instruments impose a dust obligation on this site, and particulate monitoring runs on the identical weekday window, seven to half past three. Weekday construction is permitted to six in the evening without a variance.6

Vibration, on the same site, is monitored twenty-four hours a day. That is what the difference looks like when it is stated by the same contractor in the same report.

Dust: no upwind value in handheld readings

The dust limit is expressed as a concentration above upwind background — a hundred micrograms per cubic meter above background to trigger suppression, a hundred and fifty to stop work, each averaged over fifteen minutes.

The stationary meters log at fifteen-minute intervals, so their readings can be compared to it.

The handheld readings cannot be. Of 625 handheld dust readings taken at the site perimeter over four weeks, 572, better than nine in ten, are one-to-two-minute readings, recorded in absolute figures; none of the 625 carries an upwind value.7

They are measurements against a standard they are not capable of testing. The number is real, the limit is real, and the two cannot be put in the same sentence.

Dust: no source document for the stop-work limit

That stop-work limit governs whether work continues on a construction site beside occupied homes. No document the project holds names where it comes from.

It is not in the City’s construction dust rules, which impose work practices and carry no numeric limit. It is not stated in the environmental review. It appears in the monitoring reports as a given, and in the contractor’s own materials as a given.8

A threshold whose provenance nobody states is a threshold nobody can dispute.

Air, asthma and schools: what is measured, and where

Every figure the City publishes for the air over this ground is an area figure. The table gives them as the registers hold them, with the geography each one is measured over. None of them is a measurement at a school, a shelter or a home.

Sources: the Department of Health’s Environment and Health Data Portal for the air ranks and the two asthma rates, and the Department of Education’s school location file for the schools. The two asthma rates are not alternatives: they are measured over different areas, a neighborhood and a community district, and both are published.

The zoning amendment polygon: 76% on another block

The City publishes a dataset of zoning map amendments. Each amendment is drawn as a polygon, and each polygon is meant to describe what its resolution describes.

For one amendment in this study area, seventy-six percent of the lot area inside the published polygon lies on a block the resolution does not mention.9

The resolution is unambiguous, the zoning map edition that carried it is held, and no later edition reverses it. The drawing and the document are simply not the same shape, and nothing in either one says so.

This is the kind of error that cannot be found by reading a dataset. It is found by reading the dataset against the document it claims to represent — which is work almost nobody does, because the dataset exists precisely so that nobody has to.

311 complaint counts, and what they measure

The vacate order on this corridor does not appear in the record as an order. It appears as the disposition of a complaint.10

Baker describes one benefit of a tenant association as being that neighbors are taught to call when they have an issue, to take a picture, and to leave a name, so that the complaint is recorded.11

A complaint count therefore measures two things at once: what is wrong in a building, and whether anyone there has been organized to report it.

This project reads the same records. The archive holds 311 service-request extracts for the study area. Where a count of complaints is read as a measure of conditions, it is also a measure of organization — in the City’s hands and in this project’s.12

Violence on Rikers, and where the complaints go

The City’s jails have been under federal court supervision for violence since 2015, as SCHEDULED sets out.13 The City’s public record of 311 service requests contains no request routed to the Department of Correction; its twenty-nine requests at a Rikers Island address are street matters, illegal parking and street lighting the commonest.14 In March 2026 the news site THE CITY reported that calls from people held in the City’s jails go to the Department as “customer comments”, which are not published, and counted more than 40,000 such calls a year.15

This project’s own errors

A spatial join that contradicted its own source: three lots on the Concord Avenue row were tagged as reached by a 1987 amendment whose own resolution does not reach them. The tagging is corrected, and the error is recorded because a reader cannot see it — the dataset is published and only the document behind it shows the join was wrong. The boundary, the margin and the correction are set out on ZONED.

A column read as something it was not. An older export of building department filings carried one date column, headed “Date Filed.” It was read as a sign-off date. The wrong date then appeared in five scripts, six data files, a drawing and four documents, and was corrected on 14 September.

An amount read as another amount. A figure recorded as an agency’s award on a work order was published as what the work cost — in five passages, and in one finding covering tens of millions of dollars. The numbers were right. The word attached to them was not.

What the records show

Five kinds of gap, each with a different shape: an obligation with no public measurement, a measurement that cannot test its own limit, a threshold with no stated source, a dataset that contradicts its own document, and an absence in one archive mistaken for an absence in the world.

The project’s own errors appear in the same list, as instances rather than as a separate category, and they are the same class of thing as the City’s. A centroid falling inside the wrong polygon and a polygon drawn on the wrong block are the same error at two removes.

None of these is visible to a reader of the record. Every one required setting one document against another. That is the condition this project is describing, and it is the reason the method is stated at the point of use rather than in a footnote.