Sensor calibration

Sensor calibration is the process of comparing a sensor's readings against a known, trusted reference and adjusting its output, through an offset, a scaling factor, or a more complex correction curve, so its measurements reflect the true value rather than a systematically biased one. A sensor can be precise, giving very consistent readings, while still being inaccurate if it is not calibrated.

Calibration is typically performed by co-locating the sensor with a reference-grade instrument across a range of real-world conditions, fitting a correction model, and then applying that correction to the sensor's raw output going forward; the correction can drift over time as the sensor ages or its environment changes, which is why calibration is repeated periodically rather than treated as a one-time step. This differs from general data quality checks, which catch missing, duplicated or out-of-range values, whereas calibration corrects systematic bias in values that otherwise look perfectly plausible.

Calibration underpins trust in low-cost environmental sensor networks measuring particulate matter and water quality index indicators, and in any remote sensing instrument, where on-orbit calibration against ground reference sites keeps satellite measurements consistent over the instrument's lifetime. The main pitfalls are calibrating only under one set of conditions, a correction fit in dry weather can fail in humid conditions, and letting calibration lapse on a sensor network, silently degrading data quality while dashboards keep reporting numbers with no accuracy caveat.