Capstone · Columbia GSAPP · May 2026
Where the Zone Ends
An early warning framework for industrial displacement in the Greenpoint-Williamsburg IBZ
Where, block by block, are industrial uses under the most pressure? I scored 445 industrial-compatible lots on 16 indicators and tested where the scores cluster. The strongest concentration of hot-spot blocks sits inside the IBZ core, where the designation is meant to protect industry.
- Type
- M.S. Urban Planning capstone, individual
- Advisor
- Alanna Browdy
- Reader
- Matthew Bauer
- Client
- Evergreen Exchange
- Data
- MapPLUTO 24v4 joined by BBL, plus ACRIS, DOB, 311 and DOT
- Tools
- ArcGIS Pro, Getis-Ord Gi*, InDesign
The question
The IBZ is a policy designation without independent zoning force, and the city has no block-level tool showing where industrial uses inside and around the zone face the most pressure. I set out to build one Evergreen could use to set priorities.
The framework is a screening and prioritization tool. It flags conditions associated with displacement pressure. It isn't a predictive model and hasn't been validated against later closures.
My role
I developed the 16-indicator framework, assembled the parcel-level evidence, ran the spatial analysis, and produced the report. I also designed and administered my own business survey and facilitated a March 2026 focus group with industrial business operators.
Evergreen's 2023–2025 engagement surveys, used as a separate source, were designed and administered by Evergreen staff.
Methods
Study area. The Greenpoint-Williamsburg IBZ plus a half-mile buffer: 5,278 tax lots across 257 blocks. The analysis covers the 445 industrial-compatible lots within it (MapPLUTO LandUse 5 or 6, plus LandUse 4 where the primary zoning is pure manufacturing).
Indicators. Each lot gets a 0 or 1 on 16 binary indicators in four equally weighted categories:
For example, VP-2 flags a sale above $300 per square foot of building area (ACRIS, 2019–2024); UR-1 flags a change from industrial to non-industrial occupancy in DOB jobs; CS-4 flags flooding, sewer backup or utility failure in the prior 24 months; AP-3 flags an industrial lot within 300 feet of a manufacturing-zone boundary.
Scoring and clustering. A block's composite score is the sum of positive lot indicators divided by its industrial lots. I tested block scores with Getis-Ord Gi* using a fixed 500-foot distance band, mapping blocks at 90%, 95% (z > 1.96) and 99% (z > 2.58). Indicator thresholds were set before the hot-spot results were observed.
| Dataset | Records used |
|---|---|
| MapPLUTO | 24v4 parcel layer |
| ACRIS | Deed transactions, 2019–2024 |
| DOB Job Filings | 15,162 Brooklyn CB1 records, filtered to 2,228 in the study area |
| DOB Certificates of Occupancy | 307 certificates |
| 311 | 92,227 Brooklyn CB1 complaints, Jan 2024–Sept 2025, before filtering |
| DOT Parking Regulations | 2,868 Brooklyn records → 124 truck and commercial signs → 78 on study-area streets |
Four qualitative streams supported the indicators: Evergreen's surveys, my own survey, the focus group, and a December 2025 property tour of about 25 addresses.
Findings
Every lot showed at least one flag.
The framework assigned 2,010 flags across the 445 lots: 120 lots carry one to three, 276 carry four to six, and 49 carry seven or more.
Thirty-three blocks are hot spots, in three corridors.
The 33 hot-spot blocks span the 90%, 95% and 99% confidence levels; 17 blocks are cold spots and 207 are not significant. The cold spots, around Humboldt, Meeker and the residential interior, may partly reflect an absence of industrial lots and shouldn't be read as protected industrial places.
The most significant block and the highest-scoring block are different places.
The composite score measures how many flags a block carries; Gi* measures whether a block's score sits among similarly high neighbors. Block 2616 leads on significance (z = 3.24, p = .0012); Block 2617 has the highest composite score in the study area, 8.67.
Rent gaps are wide.
Illustrative rent comparisons on these corridors put non-industrial rents at 5 to 12 times industrial rents.
An earlier loss estimate, kept in its own scope.
Working from Evergreen's earlier consolidated record of about 388 IBZ-listed businesses, the capstone found about 60% had closed or left. That figure predates the reconciliation I completed for Twenty Years On and isn't comparable to that report's cleaned census, in which 74 of 251 businesses (29%) are confirmed losses.
Proposed recommendations
The capstone proposes three lanes of action. None has been adopted.
- Annual monitoring of the indicators.
- Corridor-specific agency action on a 12- to 24-month horizon.
- Industrial-first co-location finance, drawing on the GMDC Brownsville Industrial Center at 805 Rockaway Avenue as a precedent.
Limitations
- Screening, not prediction. Gi* tests the clustering of scores I constructed; significance doesn't indicate future closure or causal impact.
- AP-1. The residential and loading proximity indicator flagged all 445 lots, so it adds no discrimination between lots.
- Weighting. The four categories are weighted equally, and the weights weren't sensitivity-tested.
- Measurement. VP-1 is a single-year assessed-value cross-section, and binary flags record presence, not intensity.
- Complaints. 311 data reflect reporting behavior, not every operating conflict.
- My survey. Small and targeted; presented as qualitative evidence, not a representative sample.
Selected findings from the May 2026 capstone. The full document is available subject to permission.