- 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:
- Range check — reject readings outside physically plausible bounds (e.g., temperature below -50°C or above 60°C, humidity below 0% or above 100%)
- Rate-of-change check — reject readings that change more than a threshold per sample interval (e.g., temperature changing more than 5°C in 5 minutes is almost certainly a sensor error)
- Persistence check — flag readings that remain exactly the same for too many consecutive samples (sensor may be stuck or disconnected)
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:
- Temperature during overcast, calm conditions (minimises microclimate effects)
- Pressure readings (should match closely after sea-level correction)
- Rainfall totals over multi-day periods (smooths out timing differences between tipping events)
Troubleshooting Matrix
| Symptom | Likely Cause | Fix |
|---|---|---|
| Humidity always reads 99–100% | Saturated capacitive sensor | Replace humidity sensor module; apply correction offset as interim measure |
| Temperature high in afternoon, fine at night | Radiation shield degradation or poor siting | Replace/clean radiation shield; relocate sensor away from heat sources |
| Pressure consistently 2–5 hPa off | Wrong altitude setting | Verify station altitude with GPS; recalculate sea-level correction |
| Rain gauge under-reporting | Debris in funnel or worn pivot | Clean funnel; test calibration with known water volume; replace mechanism if worn |
| Occasional wild spikes in any reading | RF interference or sensor communication error | Implement range and rate-of-change filters; check for nearby RF interference sources |
| Wind speed always reads zero | Jammed anemometer cups or broken reed switch | Physically inspect anemometer; clear ice/debris; replace if mechanism is broken |