Open, reproducible datasets on the Arbor Hills Landfill in Salem Township, Washtenaw County, Michigan. Each one is built from public regulatory records and published with its source files and the scripts that produced it, so any figure can be checked back to where it came from. All free to download.
Every EGLE-filed wellhead reading — temperature, methane, oxygen, carbon monoxide, and gas ratios — well by well, with source PDFs and build scripts.
2021–2026 · CSV · operator EGLE filings (SRN N2688)
The operator's own per-well wellhead hydrogen-sulfide readings from its 2026 annual survey, all 506 wells, with the source PDF and a reproducible parser.
2026 · CSV · operator annual survey (SRN N2688)
Fence-line methane and hydrogen-sulfide readings from the six-station perimeter network, with a raw snapshot bundled for offline reproducibility.
May 2022–Sept 2026 · CSV · GFL public perimeter feed (Barr Engineering)
Data-quality note (updated October 9, 2026; corrects the September 25 note): for long stretches, several fence-line stations' hydrogen-sulfide (H2S) channels have reported a single constant value while the station kept reporting and its methane channel kept varying. A sensor in clean air can legitimately read 0.0 for weeks or even months, but a constant non-zero value for months, or a constant 0.0 for more than a year while neighboring stations vary, is a standard data-quality flag, not evidence of clean air. The feed marks every reading below 7 ppb as "BDL" (below detection limit), apart from a small number with no label, and that includes the stuck 3.0–3.2 and 5.4 ppb values described here. So the BDL label cannot tell a stuck sensor from a working one, and a "BDL" reading is not on its own evidence that the air was clean. GFL's public dashboard shows only these labels (for both hydrogen sulfide and methane), not the numbers, so a stuck sensor and a working one look the same there; the numeric values are available in the feed's underlying data service and in this project's dataset. Now: MS-3 and MS-4 are reporting varying readings. MS-5 and MS-2 have resumed varying readings only recently, so their records are short: MS-5 since September 26, 2026 (up to 11.0 ppb), after reading 0.0 from May 28, 2025 (apart from one 0.1 ppb reading) and sending no data from September 16 to 26; MS-2 since September 24, 2026 (up to 2.3 ppb), after reading 0.0 from June 30, 2025; for its last week of that stretch, from 9 p.m. on September 16 to 7 p.m. on September 24, MS-2's methane channel also read a constant 0.0 ppm. Since 11 a.m. on October 6, MS-5 and MS-2 have each read 0.0 ppb in all but one hour. MS-6 read 3.0–3.2 ppb from May 28, 2025 (apart from two readings on June 11, 2026) until 11 a.m. on September 29, 2026, when it dropped to 0.0 ppb; it then sent no data after 5 p.m. that day until it resumed at 11 a.m. on October 5, read 0.0 ppb H2S and 0–2 ppm methane every hour through 5 p.m. on October 6, and has sent no data since. MS-1 has sent no data since noon on September 17, 2026, after reading exactly 0.0 for both gases from 9 p.m. on September 16. Earlier: MS-1 read a constant 3.0 or 3.1 ppb from May 28, 2025 to May 28, 2026, starting the same hour MS-5 and MS-6 went flat; MS-3 read a constant 5.4 ppb from July 15 to October 16, 2025, and 3.1 ppb from November 24 to December 1, 2025, the same value MS-1 and MS-6 read in those hours. A quiet record at a station during these periods is not evidence of clean air. The stations then reporting varying readings recorded hourly values above 30 ppb (the 2022 Consent Judgment's fence-line action level, which is defined on a 15-minute rolling average; an hourly value from the public feed is a screening signal, not a confirmed crossing), most recently on September 14, 2026, up to 153.4 ppb at MS-1 (which has sent no data since September 17). None of the stations reporting varying readings has recorded an hourly value at or above 30 ppb since. These are observations from GFL's public perimeter feed, hosted by Barr Engineering; the cause has not been publicly confirmed by GFL, EGLE or Barr Engineering. Status as of October 9, 2026, 12:00 p.m. ET.
A smaller, curated subset of the wellfield dataset — the elevated-temperature reading history for the flagged wells.
Curated subset · CSV · operator EGLE filings (SRN N2688)