The lawn that turned out to be two

A second private estate here on the Island came into our care this summer. The grounds went live in early August: a gateway, a controller, sprinkler heads around a single lawn fed by two lines, and probes in the soil.

At the first estate we put one moisture probe in each zone. Here we did something we had not done before and put two in the same lawn, at the same depth, a few dozen paces apart. It seemed a reasonable thing to try. Within a day it had told us something we would not otherwise have learned for a season, and within a fortnight it had made us change how the system thinks.

The disagreement

The two probes did not agree. Not by a little, and not occasionally: one corner of that lawn ran roughly twelve percentage points drier than the main sweep, every day, steadily, for weeks.

Twelve points is not noise. It is the difference between ground that is comfortable and ground that is beginning to suffer. Had there been a single probe in that lawn, we would have had one number and complete confidence in it, and which number we got would have come down to where the spade went in on install day.

The cause turned out to be plumbing rather than soil. That corner sits at the far end of the second line, behind a boundary hedge, at the edge of what the heads there can throw. It has probably always been dry. Nobody had ever had a way to know, because a lawn that is watered on a timer produces no evidence of its own condition — only a bill.

Where our own rule went wrong

Finding the dry corner was useful. What happened next was more useful, and less comfortable.

We had a rule, and we believed in it: within a zone, the driest probe governs. If part of the ground is short of water, that shortage sets the watering, not the average. Averaging is how a dry half of a lawn gets quietly abandoned — the wet half pulls the mean up, the number clears the threshold, and nothing is ever done about the part that needed help. We had written the rule specifically to prevent that.

It is a good rule. It was also, here, producing a bad outcome.

The dry corner is about a tenth of that lawn. Under driest-drives, that tenth was sizing the water for all of it. Night after night the system ordered the longest run it is permitted to order, and the other nine tenths — already comfortably at target — got wetter still. It was watering ninety per cent of a lawn it had no reason to water, in order to chase a corner it could not reach anyway.

Both facts were true at once: the corner really was dry, and the response really was wrong.

The change

So the rule changed. Each probe now carries the share of ground it actually stands for, and the zone's condition is worked out from those shares. A probe watching a tenth of a lawn now counts for a tenth.

Two things guard it, and both matter more than the arithmetic.

The first is that a share has to be a real measurement of real ground, not a convenient number. Weighting a probe down is a deliberate decision to let it stop governing the water, and that is only honest if the share is true. It is written down, along with the reasoning, because it will change: cut that boundary hedge back and the heads will throw further, and the number will be wrong the next day.

The second is that if the probes still reporting cover less than about half a zone, the system does not decide at all. It holds. A weighted system with a sensor missing can quietly hand a whole lawn back to the judgement of one small corner, and we would rather it stopped and said so than carried on sounding confident. That is the same instinct as the frost gate and the stale-data gate: where the evidence runs out, so does the authority to act.

What it does not fix

Here is the part worth being plain about. Weighting stopped us over-watering that lawn. It did nothing whatsoever for the dry corner.

No amount of clever arithmetic solves a watering problem that is really a coverage problem. If the water does not physically reach a piece of ground, the answer is not a better algorithm — it is a sprinkler head. That work is now on the list, along with a proper measurement of that corner rather than the provisional figure we are running on today. The system's contribution was not to fix it. It was to know, with evidence, that it needed fixing, and to stop drowning the rest of the lawn in the meantime.

Two probes, not one

We went in thinking a second probe was a nice-to-have — a bit of redundancy, a slightly better picture. It was neither. It was the only reason any of this came to light. A single probe would have given us a confident number and a lawn with a permanent dry corner nobody could name.

Jersey ground is rarely uniform, and a lawn is not one thing simply because it looks like one thing from the terrace. Grounds teach you their habits slowly, as we said when the first estate came into our care. This one taught us in a fortnight, and it taught us that we had a rule that needed changing. That is the whole point of measuring instead of assuming: the ground gets to argue back, and sometimes it wins.