ChronosAIR · lab panel

generated 2026-10-05 14:02 UTC · regenerated hourly

running 26 of 24 runs in 24h.

Last run 05 Oct 14:02 UTC · 1365 runs logged · 13650 rows

Gaps in the last 48 hours

from	to	gap
No gaps over 90 minutes in 48h.

Validation cells

KeurboomPark_CPT

coastal_metro

n: 1365
mean residual: -2.8
range: -155.5 to +49.8
band match: 15%
hard misses: 190

sign flip

PTA_North

inland_metro

n: 1365
mean residual: -6.6
range: -56.5 to +30.1
band match: 37%
hard misses: 143

sign flip

RichardsBay_Harbour

industrial_coastal

n: 1365
mean residual: -3.7
range: -88.8 to +37.1
band match: 11%
hard misses: 158

sign flip

Robertson_WC

rural

n: 1365
mean residual: +1.7
range: -124.8 to +29.5
band match: 22%
hard misses: 75

sign flip

Vanderbijlpark_NWU

vaal_priority_area

n: 1365
mean residual: -9.1
range: -106.9 to +27.6
band match: 37%
hard misses: 137

sign flip

Residual against humidity

Coverage cells · why they are null

cell	outcomes
CapeTown_CBD	reading ×1354
Johannesburg_CBD	all_stations_stale ×51, reading ×1303
Nieuwoudtville_NC	all_stations_stale ×816, no_station_in_radius ×23, reading ×515
Steelpoort_LP	all_stations_stale ×376, no_station_in_radius ×23, reading ×955
eMbalenhle_Secunda	all_stations_stale ×214, reading ×1140

Excluded from the table above, because they are our errors and not missing readings: locations_request_failed on 11 runs, last 2026-08-19. They stay in the log.

Residuals are the difference between the modelled estimate and a low-cost optical reference sensor. Neither is a reference-grade instrument. A sign flip at a single cell is worth more than a run of consistent readings, because a systematic model bias cannot change sign. Do not derive a correction factor from this panel.