Revere Housing Gap Explorer

How the model works

The question

By 2070, climate-driven coastal flooding will displace thousands of households across the North Shore — modeled here for Revere, Winthrop, Chelsea, Lynn, and Saugus. Can Revere, with its transit access and unused zoning capacity, absorb that displaced demand? This tool lets you test that question by dialing in your assumptions about how much gets built, where, and how many households each community sends.

How supply is computed

The supply model starts from a cached set of redevelopment candidates: Revere parcels with measurable unused zoning capacity before hitting the current zoning ceiling. The default Broad rule matches today's dataset, about 7,700 parcels with at least 200 sf of zoning headroom. Site eligibility lets you tighten that definition before exclusions are applied:

• Broad keeps the current broad candidate universe: any parcel with 200+ sf of zoning headroom.
• Selective requires 5,000+ sf of headroom plus a classic, near-soft, or 1.50x underbuild signal.
• Classic uses the L1_04 soft-site definition only: vacant, parking, or underbuilt parcels.
• Custom exposes the minimum zoning headroom, underbuild ratio, and included site-type labels.

For each candidate that passes the selected eligibility rule, the tool knows:

• Its zoning envelope (maximum buildable square footage under Title 17), which gives a potential unit count when you pick an average unit size.
• Its suitability grade (A–F), computed live from the four factor-weight sliders in the sidebar (zoning capacity, transit access, underused land, parcel assemblage). A parcel at a subway stop with an empty lot scores high; a fully-built single-family on a dead-end street scores low.

The five buildout sliders control what share of each grade's paper zoning capacity actually gets delivered. At 50% for Grade A, a Grade-A parcel with 10 potential units contributes 5 delivered units to the pipeline. Parcels absorb in rank order of their grade/score, capped by an annual absorption rate (100–2,500 units/yr). A small vacancy buffer (5% by default) discounts the delivered total.

The difference between zoning capacity and delivered units is intentional: zoning capacity is the full paper envelope, while delivered units subtract buildout likelihood, how quickly projects can come online, and normal vacancy.

A single flood risk tolerance dial then removes flood-vulnerable parcels from supply: binary, "we don't build where the selected tolerance says not to build." Now keeps all parcels flood-eligible; moving toward the strict end progressively tightens the filter against a static composite of FEMA current SFHA and MC-FRM 2030/2050/2070, with the strictest setting equivalent to the old 10/100 composite-risk cutoff. Year markers on the slider are reference points, not year-specific inundation forecasts — MC-FRM does not publish a 2100 scenario, so the strict end is best read as a precautionary planning buffer beyond MC-FRM 2070. This dial is shared with Parametric Zoning — one slider, one meaning. Four optional categorical cutoffs also exist (21E contamination, institutional use, infrastructure/ROW, small single-family parcels).

How the tabs connect

Supply/Demand is the source of truth. It shows the zoning capacity, delivered units, and the gap given current zoning, buildout %, capture %, and absorption. Parametric Zoning reads the currently selected Supply/Demand year, takes that gap as the target, and allocates FAR uplift across the same site-eligible, non-excluded receiver pool used for supply, then discounts gross capacity by the same buildout and vacancy assumptions used in the Supply/Demand tab. It respects the selected-year added-unit delivery headroom and each parcel’s site ceiling. Inside that shared receiver pool, parcel differences become an upzoning score: assemblage, transit, actual built density, large-site redevelopment, business districts (Broadway / Shirley Ave), and industrial opportunity south of Rumney Marsh. Parcels are ranked by that score, then given a separate site ceiling that depends on site form rather than corridor priority alone. The result is a ranked low-to-high FAR gradient rather than a fake full zoning code.

How demand is computed

Five modeled sending places have a cumulative count of flood-displaced households for 2026, 2030, 2040, 2050, 2060, and 2070. 2026 uses MC-FRM present-day 5% AEP exposure. 2030, 2050, and 2070 use MC-FRM 5% AEP probability-raster exposure, with the intervening 2040 and 2060 values interpolated.

The capture sliders set what share of each place's displaced households Revere would receive. The Base preset sets Revere-internal capture at 85%; Winthrop, Chelsea, Lynn, and Saugus have active sliders. Total demand at any year is the weighted sum across active modeled places: sum(capture rate × displaced HH).

The gap

At every year between 2026 and 2070, the tool computes gap = demand − supply. Positive means an unmet housing shortfall; negative means surplus capacity. The four preset scenarios span a spectrum from "status-quo Inertia" (low buildout, minimal regional capture, 200 units/yr) through "Base" (REACH-ROX-pace) to "Subregional Receiver" (Revere as the formal regional receiver, 90% top-grade buildout, 2,000 units/yr). Moving any slider drifts you off the preset — the active preset shows a "modified" tag and a reset button.

Data sources and vintages

The live app loads the static browser assets for parcel geometry/properties, soft_sites.json, demand_timeseries.json, scenarios.json, sending_munis.geojson, and transit GeoJSON files. The file name soft_sites.json is historical; in the live model it functions as the broad redevelopment-candidate cache, while Classic preserves the narrower L1_04 soft-site definition.

All spatial preprocessing is in Massachusetts State Plane (EPSG:26986, meters). Browser display is reprojected to WGS84 (EPSG:4326).

Confirmed scenario capture rates

Limitations

• Only Revere is modeled as a receiving site. The regional alternative (siting some of the demand in Chelsea, Lynn, or Saugus) would reduce the burden on Revere — the tool doesn't capture that.
• Demand is flood-displacement only. No organic population growth, no in-migration, no speculative movement.
• No unit-mix / affordability layer. The tool counts units, not affordability tiers or household income distribution.
• The default site-eligibility rule is intentionally broad. Use Selective, Classic, or Custom to test stricter assumptions about what counts as plausible supply.
• The Supply/Demand baseline uses current zoning limits. The Parametric Zoning tab is a planning response layer, not a formally adopted zoning code or parcel entitlement.
• Deterministic. There's no Monte Carlo / probabilistic uncertainty quantification; the static tornado PDF shows one-at-a-time sensitivity.
• Running-max smoothing is applied to each muni's cumulative displaced households to reconcile the methodology shift from FEMA current-condition flood zones (2026) to MC-FRM projected future flood zones (2030+).
• The flood risk tolerance dial is a threshold over a static composite of FEMA-current and MC-FRM 2030/2050/2070 scenarios. MC-FRM does not publish a 2100 footprint; settings beyond 2070 on the slider are a precautionary planning buffer, not a modeled 2100 inundation. Demand also tops out at MC-FRM 2070 for the same reason.

Keyboard & sharing

• Time slider: ← / → = 2 years, Shift+← / → = 10 years. Play button advances automatically.
• All sliders: ← / → = 1 step, Shift+← / → = 10 steps. Announced via ARIA for screen readers.
• Share: the current scenario state (preset + any drift + year) is encoded in the URL; the Share button copies the link.
• Print: Cmd+P renders hero numbers, chart, and narrative card only — sliders and maps hidden.

The first parcel you open will show the current grade, the inputs behind that grade, the source logic for each input, and the expected housing contribution under the active dials.

• Current grade and score
• Zoning capacity and site eligibility
• Flood and exclusion status
• Expected delivered units
• Source cues for the scoring rows

Toggle any model input in the left panel and the map redraws.