SectionKnowledge BaseTopicData QualityDepthComplete reference
Quick Answers
  • Common sensor drift patterns for temperature and humidity
  • Outlier detection techniques for automated publishing
  • Cross-validation with nearby CWOP and professional stations
  • Practical recalibration methods for hobby stations
  • When to replace sensors vs when to apply correction factors

Publishing inaccurate data undermines trust in your station and, if you contribute to networks like CWOP, can lead to your data being flagged or excluded. Sensor drift is inevitable — capacitive humidity sensors degrade, temperature sensors develop offsets from solar radiation exposure, and rain gauges lose calibration as tipping mechanisms wear. This guide covers practical detection and correction techniques drawn from maintaining multiple stations over a decade, building on the data validation concepts in Publishing Fundamentals.

Common Drift Patterns

Humidity Sensors

Capacitive humidity sensors are the most drift-prone component in consumer weather stations. After 2–3 years, readings typically shift upward by 5–15%. At high humidity (above 85%), readings may peg at 99–100% when the actual value is 90%. This is caused by moisture absorption into the dielectric material — a known limitation of the technology.

Detection: compare your humidity readings against a nearby professional station during stable, overcast conditions when spatial variation is minimal. A consistent offset of more than 5% indicates drift.

Temperature Sensors

Thermistor-based temperature sensors are relatively stable, but the radiation shield housing them can degrade. A cracked or dirty shield allows solar radiation to artificially heat the sensor, showing temperatures 2–5°C above actual during sunny afternoons while being accurate at night. UV exposure also yellows white plastic shields, reducing their reflectivity.

Barometric Pressure

Barometer sensors are generally the most stable instruments in a weather station. However, they require correct sea-level correction based on your station altitude. If your altitude setting is wrong by even 10 metres, all pressure readings will be offset. Verify your station altitude with a GPS or topographic map, not by matching a nearby station (their correction may be wrong too).

Rain Gauges

Tipping-bucket rain gauges lose accuracy as debris accumulates in the funnel or the pivot mechanism wears. Under-reading is more common than over-reading. Clean the funnel monthly during seasons with falling leaves or pollen. Test calibration annually by pouring a known volume of water through the funnel and comparing the reported total.

Outlier Filtering

Occasional spurious readings — a temperature spike of 20°C in one sample, or wind speed jumping to 999 km/h — are typically caused by communication errors between the sensor and the base station. Implement these filters in your publishing pipeline:

Cross-Validation

Compare your station’s readings against nearby reference stations. CWOP data, nearby airport METAR reports, and other personal weather stations on networks like Weather Underground provide comparison points. Focus on:

Troubleshooting Matrix

SymptomLikely CauseFix
Humidity always reads 99–100%Saturated capacitive sensorReplace humidity sensor module; apply correction offset as interim measure
Temperature high in afternoon, fine at nightRadiation shield degradation or poor sitingReplace/clean radiation shield; relocate sensor away from heat sources
Pressure consistently 2–5 hPa offWrong altitude settingVerify station altitude with GPS; recalculate sea-level correction
Rain gauge under-reportingDebris in funnel or worn pivotClean funnel; test calibration with known water volume; replace mechanism if worn
Occasional wild spikes in any readingRF interference or sensor communication errorImplement range and rate-of-change filters; check for nearby RF interference sources
Wind speed always reads zeroJammed anemometer cups or broken reed switchPhysically inspect anemometer; clear ice/debris; replace if mechanism is broken

FAQ

How often should I recalibrate sensors?
Check against reference stations quarterly. Most consumer sensors do not support field recalibration — the practical approach is to apply correction offsets in GraphWeather’s sensor editor and replace sensors when offsets become too large (typically more than 10% for humidity, more than 1°C for temperature).
Can I automate outlier detection?
Yes. GraphWeather supports configurable data validation rules. Set range limits and rate-of-change thresholds in the sensor configuration. Readings that fail validation are logged but not published, preventing bad data from reaching your website.
My readings match nearby stations — is my station accurate?
Probably, but “matching” does not guarantee accuracy — you could both be wrong. Cross-validate against a professional station (airport METAR) for the most reliable reference. Temperature should agree within 1–2°C under calm, overcast conditions.
How do I report issues to CWOP if my data is flagged?
CWOP quality control is automated. If your data is flagged, correct the underlying issue (usually pressure correction or humidity drift) and the flags will clear automatically once the data falls back within quality bounds. See METAR and CWOP Basics for details.