Soil probes
What the dirt is actually doing
Eight Ecowitt WH52 stakes, each reporting moisture, soil temperature and electrical conductivity — the last of which measures dissolved fertiliser salt. Colour follows moisture alone.
Waiting for the first reading…
Those cut-offs are a starting point, not gospel. Moisture percentage from an FDR probe depends on the mix it is buried in, so spend the first fortnight comparing the number against what the pot actually feels like, then move the thresholds to match your finger.
The tower
Top, middle and bottom of the column
Nothing has ever established in the bottom pockets of the GreenStalk, in any season. Two probes sat in the same column of soil, one high and one low, because the difference between them is the diagnosis. A third went into the middle tier on 14 September. As of 25 August the answer is in, the column has been flushed twice, and the whole tower has been sown with bush beans — the first crop deliberately planted into the tier that has never grown anything.
Settled: it was salt, and two flushes took 71% of it out
Two theories were bought to be told apart. A dry pocket should have shown the bottom tier drier than the top. A leaching zone should have shown it lower in conductivity. What the probes found instead was an enormous conductivity gap between the two tiers — the bottom carries multiples of the salt the top does.
Even watering does not mean even accumulation. Water leaves through the bottom carrying every dissolved solute from every tier above it, and the bottom pocket is the last stop. High salt is what stops roots drawing water even when there is water to draw — which is exactly what "nothing establishes down there" looks like, and it explains why the failure never cared what season it was.
The plants mostly agree. Of everything that went into this tower, the basil and the dianthus are thriving and the surviving parsley is genuinely healthy — while the thyme is now dead outright, the oregano is down to one stunted plant, one of two parsley was lost, and the marigolds and cilantro are gone entirely.
A second finding arrived alongside the salt one and is arguably larger. This tower had never actually been watered. The standing habit was two gallons through the top; roughly five cubic feet of mix needs about seven and a half gallons absorbed to go from its usual 35% to field capacity, so two gallons moved it about five points. The archive shows the bottom tier sitting at 33–36% every single day for the week before the flush. The new rule is about five gallons weekly, half through the top and half delivered pocket by pocket. One probe reads one pocket, so the vertical gradient is now measured and the horizontal spread is still a finger test.
11 August — the flush happened. Eight gallons went through: five poured into the top in half-gallon doses, three more delivered pocket by pocket. The reading three days later looked like a failure. Conductivity in the bottom tier had not fallen; it read 660 µS/cm at 19% moisture on 8 August and 1,850 µS/cm at 38% on 11 August. That was never a tripling of the salt — one of these probes measures bulk conductivity, which needs water to carry current, so a dry pocket always under-reports what is sitting in it. But the conclusion drawn at the time, that the flush had moved nothing, was wrong.
15 August — the flush worked, and it was read too early. The bottom tier now reads 830 µS/cm at 37% moisture. Set that against 1,850 µS/cm at 38% on 11 August: the moisture is within one point, so this is a like-for-like comparison, and conductivity has fallen by 55%. Salt genuinely left the tier. What the 11 August spike measured was salt freshly dissolved into solution by eight gallons of water and not yet carried out — a flush reads its worst partway through, before the water has finished leaving. The gap between the tiers is down from 7.4× to 4.9×, so the bottom is still far the saltier of the two and one more flush is now justified. The lesson for next time is not to take a conductivity reading three days after a flush and believe it.
25 August — the second flush, and a five-month clock on the salt. About ten gallons went through in one session, half into the top and half pocket by pocket, in doses roughly ten minutes apart. The minute-level probe record makes the mechanics visible for the first time: the upper tier started rising at 08:40 and the lower tier at 08:49, so water takes about nine minutes to cross the column — which sets the floor on how far apart the doses can usefully be. Conductivity in the bottom tier spiked from 900 to 1,720 µS/cm and crashed back to 430 within forty minutes. That spike is the flush working, not failing; it is salt dissolving into solution on its way out. The same day brought a correction that made the numbers worse rather than better. The tower was completely rebuilt in early spring — taken apart, drip trays washed, restacked with fresh mix — which had never been recorded anywhere. So the 870 µS/cm sitting in the bottom tier did not accumulate over years. It accumulated in about five months. It is a rate, not a legacy, and the planned soil refresh was retired rather than skipped, because there was never anything old down there to refresh.
27 August — settled, and better than predicted. Comparing raw conductivity across different moisture levels is meaningless, so the measure that matters is conductivity divided by moisture. That ratio went 24.9 before the flush, 8.5 that evening, 7.1 the next morning, and 7.1 again a day later at identical moisture — flat for twenty-four hours, which is what settled looks like. That is a 71% reduction. Two predictions were put on the record beforehand, 290–300 and then 250 µS/cm; the actual is 340 at 48% moisture. Stated against itself: the tier never dried back to the 35% of the baseline, so this is a normalised comparison rather than a like-for-like one, and no clean comparison is available ever again this season — the tier is planted now, and beans transpire, so the moisture curve stops meaning what it meant on bare soil. The number to watch from here is not the flush but the rebuild. After the first flush the ratio climbed from 22.4 to 24.9 in nine days, about 11%. Repeat that from 7.1 and the tier is back where it started by late October. Feeding practice, not flushing, decides how this ends — which is why fertiliser never goes through the top of this tower, only plain water.
27 August — rain does not water this tower, and now there is a number on it. Four-tenths of an inch fell over a couple of hours with five probes reporting every minute, which is as clean a penetration test as this garden will ever hand over. The two open pots took 31 and 18 points of moisture. The tower's top tier took five. The bottom tier took nothing at all — flat, and its conductivity did not move off 340 for the duration, so this is not water arriving and draining away, it is water never arriving. The tier was at 48%, well short of field capacity, with room to absorb. Brian has said this for months, that an inch at the top of the tower takes forever or never really reaches the bottom, and the mechanism is visible in the shape of the thing: every pocket below the top sits under the lip of the tier above it. The rule that follows is blunt — for this tower, a rain forecast is not a small factor, it is zero. Four-tenths of an inch and a completely dry day produce the same reading in the bottom tier, so rainfall probability comes out of tower watering decisions altogether.
A probe was lying, and the tell was the conductivity, not the moisture. The cucumber probe recorded nothing from that same rain, which looked at first like a sheltered pot. It was not. It had been sitting in an air void, and firming it back into the soil on 27 August moved it from 28% to 45% with conductivity coming off a floor of zero in the same reading. Two published claims died with it: that the cucumber took nothing from the rain, and that it was the thirstiest pot in the garden. The warning sign was there to be read all along, and it was not the moisture — it was that this probe reported exactly zero conductivity in 58% of all its samples, against four and five per cent for the two pepper and tomato stakes and none at all for either tower probe, never once exceeding 140. Bulk conductivity needs soil pressed against the blades to conduct at all. A probe living on the conductivity floor while its neighbours are not is a probe to go and firm.
The mushrooms are evidence, not decoration. The flowerpot parasols (Leucocoprinus birnbaumii, harmless to the plants, poisonous to eat) came up again after the flush — and they came up in a lower tier. Where they fruit marks where the water and the organic matter actually are, which makes them an independent physical confirmation, from a completely different mechanism, that the flush reached the bottom of the column. Two unrelated signals now say the same thing. Their presence also says the mix still holds undecomposed organic matter, which argues for the dump-fluff-and-amend plan over a full replacement.
One caveat on the same day's numbers, stated because it cuts against the earlier conclusion: the upper probe was reseated on 11 August after the soil around it was found loose and airy. An FDR stake in fluffy soil reads low. The top tier's older moisture figures are therefore suspect, and the specific claim that the bottom tier was the wetter of the two has been withdrawn — it rested on a 15% reading that may have been measuring air.
What's next
The care schedule
Standing rules: feed onto damp, drained soil — never soaked, never bone dry. Spray only after sunset. Harvest before spraying. Every pot propped off the deck boards.
Water tiers two, three and four — lightly, into the pockets
Four-tenths of an inch of rain fell on 27 August, and the tower got none of it where it matters. Measured across the probes as it fell: the Better Boy pot took 31 points of moisture, the Serrano 18, the tower's top tier five, and the bottom tier nothing at all — flat, with its conductivity unmoved at 340 the whole time. It sat at 48%, well short of field capacity, so it had room to take water. None arrived. Every pocket below the top sits under the lip of the tier above it, and that is the whole mechanism. The standing rule that comes out of this: for this tower, the rain forecast is not a small factor, it is zero. Take it out of the decision entirely — four-tenths of an inch and no rain at all produce the same reading down there. So: about a third of a cup into each sown pocket on tiers two, three and four, enough to damp the top two inches and no more; a half-gallon jug covers all three. Check the top tier rather than assuming — it caught the five points and its surface is exposed. Leave the bottom tier alone at 48%. Never pour through the top, which would carry salt straight back into the tier that was just cleared twice. The seed sits an inch to an inch and a half down and the probe reads deeper than that, so the seed zone is drier than the number on this page, and a bean that takes up water then dries out part-way through germinating dies rather than pausing. No emergence yet on 27 August, which is right for day two. Due 29–31 August.
Basil cuttings — the reason changed, the job did not
This was on the list because the soil refresh would disturb the roots. That refresh has been retired — the tower was rebuilt from scratch in early spring, so there was never any tired mix to replace, and the original justification is void. Take the cuttings anyway, for a better reason: this is the best plant in the garden, it is an annual, first frost is early-to-mid November, and tips root in water in a fortnight. That is a winter windowsill crop and a free head start on 2027 instead of losing the plant outright to the first cold night.
Flashlight the mint
The tap test came back empty: no mites, no thrips. But the leaves have holes, and holes mean something chews. Look at leaf undersides, new growth, the soil line and the pot rim, after dark, when night feeders are actually out.
Re-read the cucumber probe, late afternoon
That probe was found sitting in an air void and was firmed back into the soil on 27 August at 11:11. Moisture jumped from 28% to 45% and conductivity came off a floor of zero in the same reading. Two claims died with it: that the pot took nothing from that morning's rain, and that it was the thirstiest pot in the garden — the steep daytime dry-down behind that second one was most likely a void drying fast rather than a vine drinking. The test now runs for free. Same pot, same plant, same weather, only the seating changed, so if the threefold dry-down gap against the Better Boy collapses over the next two days, the probe was the whole story. Nothing before 11:11 on 27 August compares with anything after it, and the first day of readings is equilibration rather than signal.
Tie off the tomato's split branch
An old healed crack at a Y-crotch on the Better Boy. It will snap under fruit load. Tie both heavy arms to the cage above and below the split, and leave the wound unsealed.
Hardware cloth under the wheat straw
A feeder is going on the deck rail, and a seed supply raises the pot-digging risk rather than lowering it. Cut cloth to pot diameter, notch a slit to the centre around the stem, straw back on top. The look survives; the digging stops.
Bush beans into the tower
Sown after sunset on 25 August into a column flushed that morning, two seeds a pocket, an inch to an inch and a half down — the whole bottom tier, the whole second, and whatever fitted on the third, fourth and fifth. Three varieties, kept one to a tier. Provider went into the bottom tier and the top one, the two tiers that carry probes, with the white-seeded Blue Lake and Tenderette filling the middle three. That was done by instinct rather than design and it is luckier than the plan was: the two measured tiers now hold the same variety, the same seed lot, the same day and the same density, so the salt comparison between top and bottom has nothing else differing inside it. Soil was 84°F at sowing — above the 65–75°F the packets ask for, but well under the mid-nineties where beans start failing, so germination runs fast at four to six days. Emergence due 29–31 August, first picking around 14–19 October against a first frost in early-to-mid November. Tight, but real. Beans come off the feed schedule entirely — legumes fix their own nitrogen, and feeding the bottom tier now would throw away two flushes.
Probe shakedown ends
The return window on the five stakes closes. Everything currently reports between −62 and −72 dBm against a −105 floor, so nothing is going back — but the wet-weather signal check should happen before the window shuts, not after.
Catalogue
Every plant, by where it lives
Five zones, and what is actually still growing in them. Plants that have been pulled come off the list; the lessons they taught are in the standing rules above.
Album
Photo passes
Newest first. Catalogue labels are read off these frames — say the word on any matched to the wrong plant.