SF Block Score

How to read the dots

MEASURED pulled from an authoritative source: DataSF open data, SFPD incident reports, SFMTA's Vision Zero network, and the USGS 1 m elevation model.

DERIVED computed from measured data. The hazard and air layers fall here — sea level rise, tsunami, flood and landslide all come off real elevation; air and noise off real road geometry.

ESTIMATED hand-encoded from documented public sources because no bulk feed was reachable. Transit times, liquefaction polygons and the fog model are here. Screening quality — good for narrowing a search, not for closing on a house.

What each factor actually measures

• Air & noise — near-road exposure using the decay curve from the literature: steep inside 150 m, levelling by 300–500 m. Fronting one big arterial is penalised separately from being ringed by several.
• Transit to downtown — modelled walk time to the nearest rail stop or rapid bus corridor plus encoded in-vehicle minutes to Powell St. No routing engine was available, so treat these as ±5 min.
• Walkability — distance-weighted density of 4,289 active businesses by NAICS code. Computed the way Walk Score works rather than copied from it.
• Crime & 311 — two years of SFPD reports and a year of 311 cases, normalised per unit area and smoothed from neighbourhood centroids.
• Hazards — scaled 0–100 linearly, so 100 means no modelled risk rather than “best in the city”.

Known limits

• A 400 m hex cannot tell a house fronting 19th Ave from one 200 m east of it. The gradient is real; the resolution is not parcel-level.
• No vehicle traffic-count data exists on DataSF, so arterial weights are judgment calls informed by the Vision Zero injury network.
• Liquefaction polygons are approximated from published hazard maps. A parcel-level geotechnical report is the only real answer.
• Supermarket locations are a curated list of full-size stores, snapped to measured grocery points where one sat within 350 m. 38 of 44 snapped.
• Grocery counts include small markets that self-report the supermarket NAICS code, so corner-store coverage is undercounted.
• School proximity is weighted low on purpose: SFUSD still runs citywide choice with an attendance-area tiebreaker, and the zone-based system is postponed to roughly 2028.