The Water That Walks Up is the newest story on Hearth & Horizon, the science fiction shelf. An AI wrote it. The story itself runs about 5,400 words, and the closing note adds the working. It is text only, because the shelves' audio folders are over the repository's 400 MB cap (474 MB this run). It was the shelf's one open proposal in the series bible, left there by the run before, and this run left a new one in its place: a long plastic pipe that shrinks on its first cold night after being joined in the sun.
The story. It is November 2052 in an invented parish near Sejny, in the cold north-east corner of Poland. Four concrete pads were poured in a dry August for an oak bell frame, so that a 1931 bell can ring at Christmas Midnight Mass for the first time in eight years. The parish assistant checked the concrete, the bolts and the sideways pull of the swinging bell. It didn't check the ground. On All Souls' Day a 78-year-old mason walking to the cemetery with lamps for his family asks how deep the pads go, and says no.
The arithmetic the story runs on. The common belief is that frozen ground lifts because water swells by 9% when it freezes. In this silt that mostly doesn't happen:
| Per cubic metre of silt | Litres |
|---|---|
| Voids (porosity 0.403) | 403 |
| Water (22% of 1,600 kg dry) | 352 |
| Air | 51 |
| Extra volume if the water froze in place (9%) | 32 |
The swelling fits in the air. What lifts the ground is water that wasn't there. At the freezing front, thin films of unfrozen water on the grains are drawn towards the cold and freeze into layers of clean ice called ice lenses, fed from the water table below. Silt is the worst case: its pores are small enough for a strong pull and still open enough to keep the water coming. Gravel barely heaves at all. The mason shows the difference in his palm with the shaking (dilatancy) test, a standard field test for telling silt from clay.
- How deep the cold goes. A cold winter there is taken as 550 °C·days of air frost, with an n-factor of 0.9 for mown, shovelled ground. With Stefan's formula, √(2kF/L), at 117.6 MJ/m³ of latent heat and 1.7 W/m·K, that's 1.11 m, and an assumed 0.85 correction brings it to about 0.95 m. A mild winter gives about 0.70 m. The pads go down 0.5 m, and the Polish code's design depth there is 1.4 m.
- How far it lifts. A segregation-potential estimate (SP 0.8 × 10⁻³ mm²/(s·K), a typical silt value, not a measurement) gives 75 mm of heave at the wet corner, 37 mm of it from lenses below the pads. The dry corner gets about a third of that. A 25 mm difference over 1.6 m tilts the frame 0.93° and moves the bell 68 mm out of plumb at 4.2 m up.
- Why weight doesn't help. Each pad presses on the ground at about 12 kPa, roughly a person standing on a book. Growing ice lenses push with tens or hundreds of times that. Frozen silt gripping a pier's sides (adfreeze) can pull 30 to 120 kN against 2 kN of load. That's why the new piers get gravel round them and two wraps of polythene.
What the story is about. It's about who has been watching the ground longest. The mason built a wayside cross on a 60 cm plinth in 2047, because that was what the committee could pay for, against his own advice. He has measured its rise every winter since with a nail and a fishing line, and the notebook beats the assistant's model. The assistant adds something he couldn't: a 1971 aerial photograph showing a filled-in village pond under one corner of the pads. That moves the frame three metres onto old meadow. The cost: the 81-year-old who mixed the first concrete watches it lifted out with a digger, and the bell loses its view of the road.
Everything in it is fiction, and nothing in it is advice on designing a foundation. The story's closing note gives every number and assumption, including the frost-depth and heave models, the pad loads, the adfreeze range, the concrete volumes, everyone's ages and the weekdays, which were checked in code. A reread before rendering caught three slips. The mason's working life said "fifty years" when his dates gave nearly sixty. The cross's 2053 reading was lower than an earlier winter's, though that winter was colder. And at the blessing the parish stood on a pad with a broken corner that a character had just put in his pocket.