August 25, 2026

Soil Moisture Sensor for Your Irrigation: What It Actually Decides (and What It Can't)

Search "soil moisture sensor for irrigation system" and you'll find two kinds of pages: Arduino builds where you wire a probe to a relay, and vendor pages promising your garden will "water itself." Both skip the question that actually matters — what does the sensor decide, and what still needs a human or a valve? Get that split wrong and you either overwater or trust a number that was never controlling anything.

What a soil sensor decides — and what it doesn't

A capacitive soil probe measures one thing: how much water is in the ground at its tip. That's genuinely useful, but it's a reading, not an action. A bare probe doesn't open a valve, doesn't tell your hose timer to run, and doesn't override your irrigation controller. Any system that "waters automatically" from a sensor has a second layer bolted on — a relay, a smart valve, or a controller wired to obey the probe. The sensor decides whether the soil is dry; something else decides to move water.

That's the honest framing Siltwell sticks to: it reads the probe and tells you which bed is dry — "water this bed, skip that one." It alerts; it does not run your irrigation. If you want hands-off watering, you still pair that alert with a timer or a drip zone you control. What you gain is knowing you're not watering a bed that's already damp.

How Siltwell decides a bed is actually dry

The trap with cheap sensor-to-valve setups is that a single noisy reading can trigger watering — or a stuck-high reading can starve a bed. Siltwell avoids that by requiring agreement over time.

• Two readings, not one. A "this bed is low" warning and push fire only when the two most recent readings are both below the bed's threshold. One weird blip doesn't count.
• A threshold you tune per bed. The default for a capacitive probe is 25%, but that's a starting point — you set a per-bed cutoff once you've watched how that bed dries. Sandy beds and clay beds cross the line at different real-world moisture.
• Tension sensors too. If you run Watermark-style tension probes, the dry trigger is around 60 cb, again needing both recent readings past the line.

So the decision Siltwell makes is narrow and honest: this specific bed has been dry for two readings running, here's a nudge to water it. What it never claims is a moisture-percentage table handed down as fact — the right threshold is the one you learn for your own soil.

Rain-skip is a garden-level call

The other half of "should I water" is whether the sky already did it. This is where people expect a soil sensor to handle rain — and it can, indirectly, but it's a different mechanism and a different scope. Siltwell's rain-skip guidance is garden-level, not per-bed, because rain falls on the whole garden at once. It combines your station's own rain gauge with the forecast to land on one of three calls:

• Hold — 5mm or more of rain is forecast in the next day or two, so skip the can.
• Covered — 5mm or more has already fallen in the last three days.
• Water — no real rain coming or recent, so it's on you.

The soil probe answers "which bed," the rain-skip answers "is watering pointless right now for the whole garden." Together they cover the two ways you waste water — irrigating a bed that's still damp, and irrigating right before a downpour. The which-zone guide goes deeper on running only the zones that need it.

The hardware that feeds it

Siltwell reads probes through supported weather stations: Ecowitt, Ambient, Tempest, and Davis. Two honest notes worth having before you buy:

• Tempest has no soil sensor. If you run Tempest, add an Ecowitt WH51 capacitive probe alongside it — it's the best-value probe and pairs to any Ecowitt gateway or Wittboy console.
• AcuRite and La Crosse aren't supported — no soil sensor and no open API to read from. There's an open-path workaround on the AcuRite page if that's what you own.