In 2026 the city's Law Department published roughly 64,000 scanned documents on the municipal response to the attacks: air-quality lab sheets, DEP building files, the 7 World Trade Center emergency bunker, EPA correspondence. The official portal is a search box. This is the archive map that goes with it: every box and folder, every building with a file, and the school air-testing campaign laid out in time.
DEP's folder labels are street addresses with block, lot and building identification numbers: one folder per building inspected, referred or cleaned after the collapse. Geocoded with NYC Planning's GeoSearch. Circle size is the number of documents. Click a circle, then a folder name, to open it in the navigator. Yellow pins are the schools whose lab results are on the readings tab; click one to jump to its samples.
More than half the archive is one-page transmission electron microscopy count sheets from asbestos air samples taken in Lower Manhattan schools, filed by the Board of Education's contractor ATC Associates. Each sheet is a tally form: the typed fields (analysis date, air volume, sample number, analyst) read cleanly in the city's text layer, while the structure counts are handwritten tick marks that no text recognition captures. The lab's calculated concentrations appear on separate summary reports, which the Lab readings tab covers. This is the testing campaign week by week, plus the school projects named in ATC's summary reports.
| Project named in ATC report | Docs | First date | Last date | Open |
|---|
Every number ATC's lab reports gave the Board of Education: asbestos in air, dioxin on surfaces, lead, metals, and the indoor-air sweeps (CO₂, CO, particulates, temperature, humidity) from the typed results tables of 754 reports, plus settled-dust results and room locations read off 628 handwritten chain-of-custody and microscopy sheets. Hollow dots are below detection. The yellow dashed line is a reference, not a safe/unsafe border: the caption under each chart says whether it is a health limit, a legal clearance level, or a comfort guideline, and what crossing it means. Rows read from handwriting say so, with a legibility grade. Click a dot or a row to open the page it came from.
| School | Date | Location | Sample | Value | Notes | Source page |
|---|
Everything here is derived from the City of New York's 9/11 Document Portal, published by the Law Department in 2026. Documents are linked, never copied: every title opens the PDF on the city's own server.
The portal runs on a Mindbreeze search appliance with an open JSON interface. A script paged through the index agency by agency and saved the metadata the city attached to each PDF: title, agency, production volume, box name, folder name, page count, plus the first two thousand characters of the city's own OCR text. No PDFs were downloaded. The index also contains a Markdown twin of every PDF; those are the OCR layer and are excluded from the counts.
Building locations come from parsing DEP folder names of the form 68 WARREN STREET Block: 137 Lot: 1 and geocoding the address with NYC GeoSearch. Folders that are not addresses (email binders, lab batches, EPA working-group files) are in the navigator but not on the map.
The school testing timeline reads the analysis date printed on each count sheet. Sheets whose date did not survive OCR are omitted; the numbers are a floor, not a census.
Lab readings come from two passes over the 1,911 multi-page DOE reports. Typed results tables were rebuilt from the city's own text layer by sorting words on page coordinates; each value was checked against the report's narrative range, and dropped-decimal OCR slips were corrected and flagged. Handwritten chain-of-custody forms and bulk-asbestos microscopy sheets (628 pages) were transcribed by a vision model page by page, with a legibility grade and a note on every uncertain cell; those rows are marked "handwritten" in the table. Room locations for asbestos air samples were matched from custody forms by school, sample number, pump volume and date. The portal holds several scanned copies of many reports; duplicates were collapsed to one reading.
Box and folder labels were typed by whoever scanned the box, so spelling varies and some folders are mislabeled. The "Date" field in the city's index is the date of indexing, not of the document, and is ignored here. Production volumes (NYC-WTC0001 through 0010) are the batches in which the Law Department released records, not an archival series.
Lab readings are machine-extracted and unverified by any lab or agency. Typed values pass through the city's text recognition and a coordinate re-sort; handwritten values pass through a vision model with a legibility grade. Error modes seen in testing: dropped decimal points, 1/7 and 0/8 confusion, row boundaries slipping on dense tables, and years missing from fax headers. Values were checked against each report's narrative range where one exists; metals-panel and TVOC values were not and include obvious misreads. Treat every number as a pointer to the linked page, not as the record itself.
python harvest.py then python build_site.py, from the repository. The harvest is resumable and free.