Methodology

LL97 covers real property in New York City with more than 25,000 sqft aggregate floor area under DOB's aggregation rules. This product treats DOB's Covered Buildings List (CBL PDF, ingested annually) as the authoritative universe: every row here is a BBL on that list, resolved to LL84 emissions filings, LL33 letter grade, and enforcement history. Non-CBL BBLs receive a diagnostic “not on CBL” record when accessed directly and are excluded from all product-wide totals and charts.

A tiny subset of CBL rows lack pathway indicators; those are marked pathway_unknown and appear in filters and lists but never receive a cap, delta, or penalty. See pathway unknown.

Every load-bearing input, its dataset id, refresh cadence, staleness threshold, and last-known-good pointer. Rows with warn/errorstatus flow into the product-wide freshness pill in the header and drive the stale banner. This table is the authoritative answer to “when did this refresh?”

DOB publishes selected aggregates as press releases and PDFs. We store them verbatim with citation URL and use each as an independent check on our computed values. When the computed value drifts beyond the per-metric tolerance, we display an “outside tolerance” badge on the Overview page and do not overwrite the citation.

The Urban Green Council's LL97 progress report (July 2024) projects that approximately 57% of Article 320 buildings will exceed their 2030-2034 emissions cap at current filing trajectories, absent additional intervention. The cliff page shows this citation side-by-side with our computed 2030 over-cap percentage, applies a ±5 pp tolerance, and marks the delta “outside tolerance” when it exceeds that band. The citation is stored in nyc.ll97_dob_aggregate_citations with metric name urban_green_2030_benchmark and URL to the source PDF.

Urban Green is a widely-cited independent nonprofit; this is not a DOB-published figure. When DOB or another authoritative source publishes a competing 2030 exceedance projection, we will list both side-by-side rather than overwrite.

Cap_bbl,period = Σ (occupancy_coefficient_occ,period × gross_floor_area_bbl,occ). Coefficients are taken from the LL97 rules table (as loaded into nyc.ll97_cap_coefficients). GFA is taken from the most recent LL84 filing per BBL; PLUTO is used only when LL84 is absent. Delta_bbl= actual_ghg − cap; penalty_estimate = max(0, delta) × $268 / tCO2e (2024 rate).

Buildings with unresolved GFA, unresolved primary property type, or occupancy classifier confidence < 0.7 are marked unresolved_estimate and excluded from both cap and penalty computations.

Article 321 buildings comply by implementing the prescriptive measures package. There is no numerical cap. Cap, delta, and penalty columns are intentionally suppressed throughout the product (Buildings list, Building detail, CSV export). Article 321 buildings still appear on lists, they just do not compare against a cap.

A small subset of CBL rows do not clearly indicate Article 320 vs Article 321. We do not guess: we mark the pathway as pathway_unknown, suppress cap and penalty, and expose it as a filter on the Buildings list. When DOB publishes an updated CBL PDF, we pick up any resolutions on the next ingest without operator intervention.

“Cliff” refers to the 2030-2034 compliance period, where LL97 caps become substantially stricter than the 2024-2029 caps. On the cliff page, we compute an exceedance percentage as over_cap_2030 / (over_cap_2030 + under_cap_2030) × 100 from nyc.ll97_computed_exposure for period 2030-2034. Buildings with unresolved estimates are excluded from the denominator. The distribution histogram bins buildings by delta_tco2e (actual - cap) at −2000, −500, 0, +500, +2000, and +10000 tCO2e boundaries.

LL97 assigns two distinct penalty computations:

• Article 320 cap exceedance:$268 per metric ton of CO2e over the annual cap, per year of exceedance. This is the headline “penalty exposure” shown on the buildings list. It is a per-building, per-year estimate; DOB may negotiate mitigations, credits, or good-faith deferrals that we cannot see in public data.
• Article 320 no-filing: $0.50 per sqft per year for BBLs with LL97 filing obligations that failed to file. This product does not currently render a distinct no-filing penalty; it appears in aggregate DOB citations as non_filer_count.
• Article 321 non-compliance: $2,000-per-year violation for failure to complete the prescriptive measures package. Not currently computed here; DOB enforcement records are the source of truth.
• False reporting: $500,000. Not computed here.

A BBL is on the Covered Buildings List when the property meets the LL97 25,000 sqft threshold under DOB's aggregation rules. BBLs absent from the CBL show a “not on CBL” panel on the building detail page and are not counted in product-wide totals.

Enforcement events are linked to a covered BBL primarily via BIN; when BIN is missing (some OATH records), we fall back to a normalized address match against PLUTO. Events that still cannot be linked appear under “unlinked” on the Enforcement page for auditability. Events flagged is_ll97_specific=false are shown only when the user explicitly opts in to building-scoped context.

The EJ overlay attempts to load NYC LL64 Designated Communities (IRCM-RCJD). When LL64 is unavailable or restricted, the product falls back to the NYS PEJA dataset (state-level environmental justice designation). Every cliff view surfaces a persistent disclaimer identifying which source is active and warns that PEJA has a different legal basis and geographic scope than LL64.

Address search resolves free-text input to a 10-digit BBL through two paths:

1. NYC GeoSearch (geosearch.planninglabs.nyc): the official Department of City Planning geocoder. Returns candidates with BBL, BIN, borough, and coordinates. Used first for “250 Broadway, Manhattan”-style queries.
2. PLUTO fallback: when GeoSearch cannot resolve an address, we pattern-match against nyc.ll97_pluto_parcels.address. This surfaces buildings that GeoSearch marks ambiguous but PLUTO has a canonical row for.

A direct 10-digit BBL entry skips both and short-circuits to the building record. If the BBL is not in the CBL, LL84, or PLUTO tables, the record page renders a “record not found” state with the three-source diagnostic and a fresh address search field.

LL97 caps are calculated per LL97 occupancy group (offices, hospitals, retail, etc.), which is a distinct taxonomy from the LL84 “primary property type” string. We use a rules-based mapper: LL84 property type + PLUTO building class produces a most-likely occupancy group, with a confidence score in [0, 1]. Rows with confidence < 0.7 receive the occupancy_low_confidence data-quality flag and their computed cap is suppressed from headline metrics. When DOB publishes canonical occupancy-group assignments (BEAM field ll97_occupancy_group), this approximation is replaced without operator intervention.

Each row in nyc.ll97_computed_exposure may carry any of the following flags:

• missing_gfa · LL84 filing lacked GFA.
• missing_property_type · LL84 lacked a primary property type.
• gfa_disagreement_gt_5pct · PLUTO vs LL84 GFA differ > 5%.
• occupancy_low_confidence · Classifier confidence < 0.7.
• ll84_zero_ghg · LL84 filed 0 tCO2e; treated as unresolved.
• historic_reporting_missing · No LL84 for the reference year.
• simulated_shadow · Row is a simulated best-effort placeholder for a not-yet-public BEAM field.
• cap_not_applicable_pathway · Cap suppressed because pathway is 321 or unknown.

LL84 filings for a calendar year are due May 1 of the following year. This product displays the most recent available LL84 report per BBL. CY2025 filings are in-flight as of this refresh; when the annual LL84 dataset lands, the product picks up new rows on the next ingest.

• Assessed penalty (vs theoretical)
  Actual penalty DOB assessed against a building, distinct from the theoretical over-cap exposure computed from LL84 emissions.
  Public partial shadow: ecb_summons_issued
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• Audit assignment
  Buildings selected by DOB for LL97 filing audit and the audit stage.
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• Audit outcome
  Result of an LL97 audit (accepted / correction ordered / penalty referred).
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• Extension flag
  Whether a per-building filing was accepted under a filing extension (June 30 grace or BEAM extension to August 29).
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• Filing received date
  Timestamp DOB received a per-building LL97 compliance filing.
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• Good-faith effort application
  Whether an owner has applied for a good-faith effort determination and its status.
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• Mediated resolution application
  Application for the LL97 mediated resolution process and its outcome.
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• Notice of deficiency status
  Whether a notice of deficiency has been issued for a building and its current stage.
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• Per-building filing status
  Filing status for each covered BBL/BIN combination (filed / late / non-filed / exempt) as tracked in DOB's BEAM system.
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• Pre-OATH correspondence
  DOB correspondence with owners before an OATH matter is opened (notice of deficiency, cure letters).
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• RDP attestation
  Registered Design Professional attestation attached to each per-building compliance filing.
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• RECs / offsets claimed
  Renewable Energy Credits and beneficial-use offsets claimed to reduce reported emissions.
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact
• 60-day cure window
  The 60-day cure window state for a building under LL97 enforcement (open / cured / expired).
  Release status: announced_not_yet_released · Owner: DOB Sustainability data contact

Not every LL97 input is publicly accessible. When a source is throttled (Socrata rate-limit, missing NYC_APP_TOKEN), restricted (LL64 IRCM/RCJD), or otherwise unavailable, the product does one of three things:

• Fallback datasets: when a documented fallback exists (LL64 → NYS PEJA), the ingest loads the fallback, records the substitution in the source inventory, and surfaces a persistent disclaimer wherever the affected metric appears.
• Last-known-good: when the current refresh fails but the previous successful load is still valid, we serve the last-known-good rows and surface an amber “last refresh warned” banner with the reason and the timestamp.
• Empty state: when a source has never landed and no fallback exists, the dependent view renders an explicit empty state naming the source, the escalation owner, and a link to the source's public URL.

Every load-bearing source in the inventory has an escalation owner and trigger. The trigger fires automatically when ingest_runs.status = 'error' for two consecutive attempts, or when last_success_at exceeds the source's stale_after_hours. See source inventory for per-source owners.

The product is designed to meet WCAG 2.1 AA on every route:

• Every form control has a bound label (via htmlFor).
• Every chart bar and interactive element has a screen-reader label with the numeric value and the meaning of that value.
• Focus order follows the visual reading order; every interactive element is keyboard-reachable and shows a visible focus outline.
• Color is never the sole channel: pathway badges combine color with text, over/under cap uses ↑/↓ + text, and every “within/outside tolerance” state has icon + text.
• A skip-to-main-content link is present on every route for keyboard users.

Known limitations: some Recharts default tooltips may not announce in every screen reader. If you encounter an accessibility issue, contact the escalation owner listed in source inventory.