A soil moisture sensor does not measure the “true moisture” of an entire pot. It describes a small area around its probe. Used with that limitation in mind, it can reveal drying, confirm that a dripper worked, and prevent an unnecessary watering cycle. Used as an absolute judge, it turns irrigation into a lottery: the rule reacts to one point while roots occupy a volume.
Use a new sensor for observation only during the first several days. Water manually, note how fast its value rises, where it peaks, and how long it takes to return to the usual range. This shows the behavior of the zone rather than one number: a light substrate dries quickly, a dense mix holds water longer, and a large pot changes more slowly than a small one.
Where to place the sensor
Do not place the sensor directly under a dripper. It will get wet first and tell the controller that watering is complete while the rest of the pot may still be dry. Avoid the wall as well, where temperature and evaporation differ from the central root zone. A useful position is usually between the dripper and the roots, deep enough to describe the substrate rather than its dry surface.
One sensor rarely describes a long bed. Treat it as a reference point and continue to inspect individual irrigation lines: one blocked dripper may not affect the shared sensor. In a greenhouse, combine moisture readings with a watering log and periodic visual checks.
Why readings drift or jump
Readings depend on substrate, salts, temperature, compaction, and contact around the probe. Repotting can introduce air gaps. Feeding can shift a reading because conductivity changed rather than water content. Inexpensive resistive probes also degrade in soil faster than capacitive designs.
Do not copy a threshold from somebody else's setup. A displayed 35% can mean a comfortable root zone in one substrate and a dry pot in another. Find your own range by recording the value before manual watering and again after the water has distributed. The guide to overwatering and underwatering patterns explains how to read the shape of the trend.
Choosing a sensor type
| Type | Strength | Limitation |
|---|---|---|
| resistive probe | inexpensive way to test an idea | exposed electrodes corrode and respond strongly to salts |
| capacitive analog sensor | no exposed galvanic electrode pair | needs ADC calibration, protected electronics, and stable power |
| ready-made Zigbee soil sensor | battery operation and automatic Zigbee2MQTT discovery | exact exposes depend on model ID and hardware revision |
| pot load cell | tracks the change in water stored across the whole pot | more mechanical work and compensation for plant and equipment mass |
A supported Zigbee model is convenient for monitoring. A capacitive sensor and ESP32 suit DIY experiments. In either case, the number becomes useful only after verification in the actual substrate.
Calibrate without pretending one percentage is universal
Collect at least three repeatable reference cycles:
- install the probe in a permanent position and let it stabilize;
- record the value when your normal checks indicate that watering is due;
- add a measured volume of water;
- record the value after distribution, for example 30–60 minutes later;
- repeat at least three manual watering cycles;
- choose a threshold inside the stable range and add hysteresis.
Hysteresis uses different boundaries for entering and leaving a “dry” state. It prevents a noisy value near one threshold from rapidly toggling a rule. Recalibrate after repotting, changing substrate, feeding, or moving the probe.
Build a safer watering rule
The most dangerous rule is “if moisture is below the threshold, start the pump.” A real system needs a minimum interval between runs, a maximum duration, and a daily water limit. If the reading is stale, the controller must block watering instead of acting on the last known value.
A sensible starting rule requires several consecutive low readings, enough time since the previous watering, no leak alarm, a non-empty tank, and maintenance mode disabled. Run the pump briefly and then wait for water to distribute. Without that soak time, the controller may start again before the probe can respond.
Signs of a sensor fault, not dry soil
| Trend or state | Possible cause | Automation response |
|---|---|---|
| the value instantly falls to its minimum | wiring, power, or a changed ADC scale | block watering and report a sensor fault |
| a perfectly flat line for hours | updates stopped or the probe lost contact | check last_seen and availability |
| sharp jumps without watering | contact, power, radio, or analog-input noise | reject an isolated spike as a trigger |
| no expected rise after watering | water missed the probe, the placement is wrong, or the sensor failed | block repeat cycles and inspect the line |
| the baseline drifts over weeks | salts, corrosion, compaction, or root-zone change | run a manual reference cycle and recalibrate |
Maintenance and replacement
Inspect the cable, enclosure, probe position, and trend stability periodically. A line that no longer responds to watering may indicate lost contact or failure; a suddenly noisy trace points to power or connectivity. Enable maintenance mode before replacement so that a missing sensor cannot be interpreted as dry soil. See service mode and sensor replacement for the sequence.
Conclusion
A soil moisture sensor is an observation tool, not an absolute judge. It supports irrigation only when installed in a meaningful location, verified in the actual substrate, and combined with safe limits. In GrowerHub, use it for charts and alerts first; enable automatic action after the zone's behavior is understood.
See the Zigbee sensor examples and the self-service connection path for a first setup. The model choice remains open, but checking the exact model ID and exposed fields in Zigbee2MQTT before purchase avoids surprises.