Ask a question small enough to answer
“What is my garden doing?” is too broad for one sensor. “Does this raised bed stay damp beneath its mulch overnight?” is a question you can investigate. Define the decision first, then choose a measurement that could change it. If you would water exactly the same way regardless of the number, the instrument may not yet have a useful job.
Separate local conditions from regional forecasts. A forecast describes an area and a future probability, while a rain gauge observes precipitation at one location. Neither tells you directly how much water reached the roots beneath a canopy. Use them as different pieces of evidence rather than competing claims to a single truth.
Make placement part of the reading
For air temperature, direct sunshine and reflected heat can bias an exposed sensor. A suitable radiation shield and ventilated location help, but the complete installation matters. A sensor beside a warm wall may be useful for understanding that wall’s microclimate; it should not be presented as representative of an open garden or official station.
Rain gauges need a stable, level installation away from obvious obstructions and splash sources, following the maker’s guidance. Leaves, bird activity and a tilted collector can all alter results. Wind measurements are particularly sensitive to height and nearby obstacles. A sheltered balcony anemometer can describe that balcony, but it cannot stand in for a standard open-site wind observation.
The World Meteorological Organization’s observation guidance and NOAA’s Cooperative Observer Program explain why exposure is part of a measurement. You do not need to create a professional station, but you should record the ways your site differs from one.
- Location
- Bed, wall, balcony or open plot
- Exposure
- Sun, shield, obstacles and shelter
- Timing
- Observation time and last update
- Meaning
- What decision this reading supports
Give soil probes a local reference
Soil is not a uniform material. Texture, salinity, compaction, temperature and installation can influence electrical measurements. A low-cost probe may report an arbitrary scale, a converted percentage or an estimated water-content value. Read its documentation before naming that output “soil moisture” in a log.
Compare the installed probe with repeated hand checks over a natural wetting and drying cycle. Note whether its trend matches what you find at the same depth. Do not deliberately stress valuable plants simply to find a dry endpoint. A comparison with a second inexpensive probe can reveal disagreement, but agreement alone does not prove that either instrument is accurate.
Maintain good soil contact, avoid air gaps and keep the probe away from an emitter unless that exact wetting pocket is your intended subject. If you relocate it, treat the new position as a new observation series. The old threshold may no longer describe the same conditions.
Make bad data visibly bad
A dashboard should display the time of the last successful observation, not just the last value. A plausible reading from yesterday can be more misleading than an obvious error. Mark missing, stale and out-of-range data explicitly. Never turn a failed sensor into a perfectly dry reading merely because zero is a convenient software default.
Store units with your notes. Distinguish instantaneous readings from averages and accumulated rainfall from rainfall rate. Preserve gaps rather than drawing a smooth invented line across them. When a battery change or firmware update changes behavior, annotate the event so a jump in the graph is not mistaken for a sudden change in the soil.
If data controls watering, require a recent plausible reading and apply a separate runtime limit. Keep weather-station advice distinct from safety-critical storm or flood warnings. Follow official local warning services for dangerous conditions; a hobby station and a home-built alert are not substitutes.
Budget for cleaning and communications
Inspect housings, cable entries and mounting hardware after severe weather. Clear a rain collector according to its instructions and check that a shield has not become a home for insects. Keep connectors off the ground and include drip loops where appropriate. Do not coat a sensing surface unless the manufacturer explicitly permits it.
Check radio reception at the final installed position, with doors closed and the garden in leaf. Wi-Fi, Bluetooth Low Energy and LoRaWAN solve different communication problems, and none guarantees coverage at your particular shed. LoRaWAN also requires suitable network or gateway access. Record battery replacement dates and test whether low-power warnings actually reach you.
Run the disagreement experiment
For one week of ordinary garden visits, write down the probe trend, a root-zone hand check, observed rain and the action you chose. Add a note whenever the instruments disagree with each other or with your senses. Investigate placement and maintenance before buying another sensor.
The useful outcome is not a beautiful graph. It is a clearer rule: this reading supports this decision under these conditions. If you cannot write that sentence honestly, keep the instrument in observation mode and leave the watering loop under manual control.