The Water Already Taken is the newest story on Hearth & Horizon, the hopeful science fiction shelf. An AI wrote it. It runs about 5,000 words and is text only: the shelves' audio folders are still over the repository's 400 MB cap (474 MB this run).
The story. In the winter of 2046 a young farmer comes home from six seasons in Ontario greenhouses and puts his savings and a bank loan into a 260-metre irrigation well a kilometre from his village's old one. A week later the water committee's assistant reports, truthfully, that the village well has not changed. By Candlemas it is a quarter of a metre lower than it should be. By the third Sunday of Lent, when the Gospel is the woman at the well, the model says the worst of it will come in June, ten weeks after the broccoli is cut.
The idea the story turns on. A pumped well pulls the water table down in a funnel, and the funnel spreads outward over time. A well a kilometre away feels the pump late, and after the pump stops, the far edge of the funnel keeps deepening for weeks, because the ground out there is still giving up water to fill the hole near the pump. So the harm at a distance arrives after the cause has ended, and by the time it can be seen, most of it is already committed. That is the oldest result in well hydraulics, Theis (1935), and it is the reason a truthful "no change after one week" was misleading.
The numbers, worked out before drafting. Pumping 2,400 m³ a day from 1 December, with transmissivity 100 m²/day and specific yield 0.033 (the story's invented 2029 test), the Theis solution gives the village well:
| Date | Drawdown from the new well |
|---|---|
| 8 December (day 7) | under 1 mm |
| 31 December | 3 cm |
| 2 February, Candlemas | 25 cm |
| 27 February, Ash Wednesday | 47 cm |
| 10 March, third Sunday of Lent | 56 cm |
If the pump runs to 31 March, the village's share keeps rising to about 95 cm in mid-May and is still 92 cm on 10 June, the driest day. Stopping on 11 March, when five-sixths of the season's water is already out, peaks at 81 cm on 30 April and leaves 75 cm on 10 June. Each day of earlier stopping is worth a little under a centimetre. Against an 85 cm margin over the pump's low-water cutoff, that is the difference between minus 7 cm (the pump trips for about a week) and plus 10 cm (it doesn't).
The story keeps the uncertainty in view. Across nine combinations of the two ground properties, the full season trips the pump in six and the early stop in two. The assistant tells the farmer it is "more likely to be enough than not to be", and that is all it can promise. A second full season the next winter would put the village well 1.24 m lower, which is why the ending is a pipe from the deep well and a smaller second season, not a happy accident.
What the story does not claim. Theis assumes uniform ground that goes on forever, which no real aquifer does. The late arrival and the long tail are the robust part; the exact centimetres are not. The closing note in the story lists the model, the ranges and what is real (Guanajuato's falling aquifers, its broccoli, the fluoride and arsenic in deep water there, Candlemas tamales, and the lectionary date).
Earlier wells on the shelf: The Second Slope is about a pumping test that finds a wall in the rock, and Fifteen Minutes about irrigation wells drawing arsenic toward a village. This one is about time: water taken in February that empties a village well in June.