← Gallery

Where a sector's harm lands

PROTOTYPE
San Francisco Bay Area · 9 counties
⚠ Hub sector.
Deaths/yr, this sector (count)
no data / ~0 deaths from this sector in this cell
Each square is one model cell. The grid is not uniform — cells run from about 3.4 km across in the dense core to 55 km in the sparsest corners, so a big square covers far more ground, and more people, than a small one. That matters most in the deaths view, where a cell's count rises with the area it covers; the death rate view divides that back out. Of the 257 cells, every sector reaches 255 — the two with no modelled residents are never shaded. Cells are shaded against this sector's own range at equal-count (quantile) breaks, so the colours split this sector's cells into equal-sized groups rather than tracking one extreme cell. Between sectors compare the pattern here and the totals in the card — a colour means a different number for a different sector.

Honesty ledger — assumptions & limits

Counterfactual estimate: each shaded cell shows the deaths avoided if that one sector's modeled PM2.5 alone were removed and the reduction sustained.
One sector at a time; totals overlap: each sector is estimated on its own, removed against the same baseline while the others stay in place. Because the risk curve bends, those estimates overlap and add to more than the Bay-wide total (). Read one sector at a time; never add two.
Baseline death rate = per person per year — about deaths per 100,000 people per year. IHME Global Burden of Disease: California, non-communicable disease plus lower-respiratory infection, ages 25 and over, applied uniformly to every Bay cell.
Recorded in the data as: “
Risk curve (how much extra death risk a given PM2.5 level implies): the GEMM function. This is the high end of the published range; the alternative function the same literature recommends gives 30–70% lower numbers for the same air-quality change.
Recorded in the data as: “
Ports, airports and road traffic () have a whole fleet's emissions pinned to a single dock or terminal cell, so everything downstream inherits that distortion — a directional picture, not a reading for any one neighborhood.
Low-confidence sectors () look like a whole-sector total spread thinly over many small stand-in sites rather than measured facilities — flagged, and not solid enough to rank against the others. The sector this page opens on is chosen to keep that estimate out of the headline.
Dollar figures are illustrative. They price a small reduction in mortality risk spread across a large population, using a standard regulatory figure called the value of a statistical life. Prefer thinking in a range (~$7M–$13M) around the central value.
Recorded in the data as: “
• 2005 annual-average meteorology; steady-state annual, not an episodic or wildfire model. Fractional per-cell figures are statistical expectations — most sector-and-cell pairs carry a small fraction of one modeled death per year.