A frost alert system is three simple parts working together: a temperature sensor sitting where the crop is, a threshold that defines danger, and a message that reaches you in time to act. On a small farm that message can be the difference between a healthy block and a ruined one. The good news is that a cheap, slightly noisy alert beats an expensive one you never installed, and it beats none at all.

What radiative frost actually is

Not all frost is the same. Advective frost arrives with a cold front and wind, and there is little a smallholder can do about it. The kind you can defend against is radiative frost, and it forms on clear, calm winter nights. With no cloud to hold it in, heat radiates from the ground up to the open sky, the air just above the soil cools, and that cold, dense air slides downhill and pools in the low spots.

This is why one corner of a field can freeze while the slope above it stays clear. It is also why the weather report can mislead you. Official air temperature is read in a screen about chest height, but your seedlings sit lower, where it is colder. Frost can bite the crop while the nearest station still reads a few degrees above zero.

Siting the sensor is where most systems fail

Which sensor you buy matters far less than where you put it. Place it at plant height, in the low pocket where cold air collects, not up on a pole or against a north-facing wall that soaks up the day's sun and stays warm after dark. Shield it from rain and direct sun with a small louvred screen, but keep it ventilated so it reads the air and not sunlight or its own warm housing.

One reading at the shed tells you almost nothing about the frost hollow two hundred metres away. If your land has more than one cold pocket, it is worth a sensor in each rather than trusting a single point to speak for the whole farm.

Lead time versus accuracy

Every frost alert trades warning time against certainty, and you cannot escape that trade. A bare threshold alarm, along the lines of message me when it hits 2 degrees, is dead simple and reliable, but it only buys you the time between that trigger and freezing, perhaps an hour on a fast-cooling night. Add the rate of fall, or a dew-point and forecast layer, and you gain lead time at the cost of more false alarms.

For most small farms, start with the simple threshold. A false alarm costs you a broken night's sleep; a missed frost costs the crop. Cheap and a little noisy wins.

What a good alert should do

An alert is only useful if it wakes the right person and tells them enough to act. Before you trust a system through a cold snap, check that it does all of this:

• Reaches a real person at 3 a.m. on a channel they will notice, such as an SMS or a phone call, not just an app nobody opens overnight.
• Runs on battery or solar backup, because frost nights and load shedding both happen in the dark.
• Sends the actual number, not just cold, because you respond differently to 2 degrees falling fast than to half a degree holding steady.
• Leaves you time to deploy a defence: overhead sprinklers, row covers or fleece, a wind machine, or the call to harvest early.
• Keeps a log, so next winter you know which block froze first and where the sensor really belongs.

Build your own, and start simple

None of this needs a costly commercial rig. At heart a frost alert is a temperature sensor, a rule and a message, which makes it a genuinely useful build rather than a toy. A microcontroller such as the sheenbot∞ board can read a sensor and fire off an alert, and the parts to do it are in the store. It is the kind of project our holiday workshops lean on, because it solves a real problem while teaching sensors, thresholds and logic.

Wherever you start, keep it simple. Site the sensor low and at plant height, set an honest threshold, and make sure the message gets through when the power is out. The best frost alert is not the cleverest one; it is the one that wakes you in time. For more from our farming series, see the newsroom.