A dashboard can be wrong while every sensor is online
An urban heat dashboard shows a sudden rise at one location. The change might reflect genuine conditions, sunlight striking a poorly shielded instrument or a faulty device. Data quality assurance determines which interpretation is justified before planners act.
Define fitness for purpose
A sensor network designed to compare shaded pedestrian routes needs different spatial coverage and quality requirements from one intended to observe regional weather patterns.
Preserve installation metadata
Record instrument type, location, mounting height, shielding, maintenance history and relevant surroundings. Without this context, two readings may not be comparable.
Separate calibration from quality assurance
Calibration assesses instrument performance against suitable references. Ongoing quality assurance also checks completeness, timing, plausibility, consistency and processing errors.
Flag rather than silently overwrite
Suspicious spikes, flatlines, missing periods and clock errors should be identified and investigated. Preserve appropriate raw observations and document corrections.
Compare neighbouring observations carefully
Nearby instruments can legitimately differ because of shade, airflow and surfaces. An anomaly rule must not automatically erase real microclimate variation.
Assign ownership for faults
The system needs a clear route from anomaly detection to inspection, repair, data qualification and communication with downstream users.
Break-test: a sensor is moved
After relocation, its time series changes. Can analysts distinguish environmental change from changed placement and flag the discontinuity before publishing a trend?
Connected knowledge
See Environmental Sensor Placement, Microclimate Sensor Calibration and NEA.