Detroit / Air Studies

01 / 07

26:163 / 88101 / 2026

Reading the record

548 daily PM2.5 observations across seven Detroit monitors, January1–March31,2026. Eighty-two station-days have no matching record. The main series is New Mt. Herman Church,0097, in48217: all90 days are present. Southwest0015 remains in comparisons with43 readings.

ug/m³ is particulate mass per cubic metre of air. Every plotted concentration comes from an EPA row. Derived medians, quantiles, means, correlations and differences are described below; they are summaries, not additional measurements.

One shared playhead

Playback opens on March31, the latest matching Detroit date in the selected archive, then traverses January–March at two seconds per day. Swiping preserves the selected date across all seven works. The slider inspects a date and holds it. The pause button or Space resumes playback. A moving cursor indicates progress between dates; numeric values change only at the next actual record.

01 / Register

Seven monitor columns show the selected month’s actual readings. Desktop shows the month; portrait shows fifteen neighboring days at a readable scale. The thin7×90 strip records availability: blank means absent, ink means a reported sample. It is not a concentration heatmap.

02 / Field

Exact daily observations projected into seven categorical layers. Time runs left to right, concentration increases upward on one shared43.4µg/m³ scale. This is not terrain. The small matrix prints Pearson correlations between station series, using only overlapping valid dates for each pair; it is descriptive association, not evidence of a common emission source.

03 / Departure

Primary daily concentration minus its quarter median6.45µg/m³, above or below the baseline. The companion trace is the trailing seven-calendar-day arithmetic mean; at the start it uses the available1–6 days. The median is not a health limit.

04 / Frequency

Ninety numerical observations grouped into twelve equal concentration bins from0 to43.4µg/m³. Two observations fit each row; exact bin counts appear below. PrintedQ00/Q25/Q50/Q75/Q100 summarize the primary distribution using linear interpolation between sorted observations (Type7 quantiles). The maximum is included in the last bin.

05 / Comparison

Monitor rows sorted by the selected date’s concentration; missing readings sort last and show--. Bars share the quarter-wide0–43.4 scale, so widths remain comparable across dates. Small traces show each monitor’s quarter. Desktop’s rightmost difference is that monitor minus the selected primary reading. Row positions briefly settle when the actual rank changes; bar values are never interpolated.

06 / Rank

Each connector preserves one primary observation between chronological day position and descending concentration rank. Earlier dates break ties. Endpoint heights also encode concentration. The small lower strip is the signed change from the previous calendar day; day1 has no predecessor and is absent.

07 / Cycle

One circle, ninety daily spokes. Angle is chronological day across this quarter, not a full year. Length uses the shared concentration maximum. Numerical annotations are actual concentrations; portrait reduces label frequency to prevent overlap. The selected-day readout sits inside. Desktop adds nearby records, median,90th percentile and day-to-day change.

Shared registers

Bottom readouts pair each station’s current value with its quarter mean and coverage. Missing records remain-- and interrupt traces. No visual element claims a pollution source, citywide exposure, health forecast or live measurement.

Arrows / swipe: change work. Space: hold/resume. Date slider: inspect. F: fullscreen. S or download icon: export the current frame at3840×2160, or2160×2880 for portrait. Reduced-motion devices start held. The guide pauses playback.

Full mapping notes ↗ · EPA source selection ↗ · p5.js source ↓

03 / 31

SWIPE