Hot, dry, and intermittently awake.
A year-ish local sensor trace shows a room that spent much of its measured life above normal comfort temperatures, with dry-air episodes, distinct light/sound daily rhythms, and a few sensor anomalies large enough to deserve their own quarantine.
1. Heat Was The Default State
Aggregation logic: convert UTC to America/Chicago, keep days with at least 100 readings, then chart daily median, daily 95th percentile, and daily maximum temperature in °F. Headline shares use all valid temperature readings.
Highest adequately covered daily median, on 2025-06-02; that same day reached 95.9°F. The coolest covered day still had a median of 72.9°F.
The striking pattern is not a rare heat spike; it is the baseline. More than half the valid readings sit above the common comfort range, and the upper tail keeps pushing past 90°F.
2. The Room Has A Daily Signature
Aggregation logic: group readings by local hour of day. For each hour compute median light, 90th percentile light, median sound, and the share of readings with sound at or above the global 95th percentile.
Brightest typical local hour by median light. The quietest median sound hour is 00:00, while the largest share of sound spikes occurs around 18:00.
Light and sound do not move as one simple occupancy proxy: light has broad daylight/desk-light structure, while sound is mostly quiet with narrow bursts.
3. Dryness Is A Regime, Not A Moment
Aggregation logic: use daily median and daily minimum relative humidity on adequately covered local days; monthly bands summarize the seasonal drift. Dry readings are valid humidity values below 20% RH.
Driest adequately covered daily median RH, on 2025-12-15; the minimum reading that day was 14.2% RH.
The lowest-humidity period is sustained enough to show up in medians, not just isolated minima. That makes it plausibly lived-room dryness rather than a single bad sample.
5. The Extremes Tell A Different Story
Aggregation logic: rank valid rows by raw light, raw sound, temperature, and humidity extremes. Pressure value 72.0 hPa was excluded as impossible before pressure summaries, but retained in provenance as a known-bad range issue.
The largest light value is far beyond the rest of the distribution and should be read as sensor saturation or a glitch, while sound spikes are brief event markers rather than a stable noise level.
7. Pressure Moves Are Visible, But Not Causal
Aggregation logic: drop impossible pressure, compute daily median pressure/temperature/humidity on adequately covered days, then compare same-day daily medians and day-to-day median changes. Correlations are descriptive only.
Correlation between daily median pressure and daily median humidity. Temperature correlation is -0.41; without outdoor weather or HVAC state, this is association only.
Pressure contains a real weather signal, but the indoor response is modest and entangled with season, heat, and collection gaps. It is useful context, not an explanation by itself.
8. Overnight Is Calmer, Workday Is Hotter
Aggregation logic: bucket local hours into overnight 00-05, transition 06-08, workday 09-17, and evening 18-23. Compare median temperature, interquartile range, median humidity, and median raw sound.
Workday median temperature minus overnight median temperature. Workday median raw sound is 1.0 units higher than overnight.
The office does settle overnight, but not into a cool profile; the day-night difference is smaller than the overall high baseline.