Weather: fronts and forecastsWE

WE1: the model and the forecast numbers

The weather replaces the one roll per day of Calendar.WeatherOn(day) (three types, a hash of the day number, so day 12 had the same weather in every save). It is a pure function of the save seed and the hour, so every sibling computes the same sky and nothing is synced.

Fronts, not dice

• A hidden hourly curve per save, a pure function of (State.seed, hour). Fair spells sit between fronts. A front arrives every 3-6 days, builds cloud for a few hours, rains for 6-24 h, then clears with a wind shift. Warm fair afternoons can pop a shower. A clear, calm night after rain leaves fog the next morning (it burns off by ~10:00 and leaves the lawn soaked).
• Cloud cover, rain chance, rain amount, fog, wind and temperature are all read from the curve, so changes are gradual and streaks are realistic (a wet week, a dry week).
• Periods: simulated every hour, shown as morning (7-11), afternoon (11-15), evening (15-19) and night (19-7). Night isn't played but is shown small: rain overnight means a soaked lawn at 7:00.
• The first 3 days are dry (grandma's garden, Mrs. Higgins are learned on dry grass): fair weather forced, 0% shown.

An honest forecast

• The curve gives each period a rain chance; whether it actually rains is then rolled from that chance (SimRandom.Value(seed, period index, salt), a named salt, SD2). "30%" rains about 3 times in 10. A rare shower under a "clear, 2%" forecast is fair and funny. The forecast can be wrong; the percentage is always true.
• The percentage comes first and the icon is drawn from it. If the percentage only followed the icon, it would add nothing. A player who mows anyway under 30% wins 7 times in 10, and knew the risk.
• Today and tomorrow are shown per period. Days 3-7 show one icon, the day's rain chance (1 − the product of the morning, afternoon and evening "no rain" chances), high/low and wind.
• The forecast never changes as the day gets closer (revisions are WE15).

Sky types

Stored as values (cloud cover, rain chance and amount, fog, wind, temperature); the icon is derived, so the sky changes gradually instead of switching between presets.

Thunderstorms (WE7) and snow (WE13) come later.

Season, temperature, wind

• Season: the code is season-ready (a season table, 14-day seasons, a 56-day year) but runs one endless late spring: growth ×1.2, lows ~50 °F, highs ~70 °F. Calendar.SeasonsEnabled = false. Filling the table is WE9.
• Temperature: for now it only sets the dew (clear calm nights = heavy dew). Shown in °F (WE4).
• Wind: mph, a direction (8 points) and gusts. Ambiance (trees, rain slant) plus the two effects in WE3.

Climate targets (late spring)

What the generator is tuned and tested against, over a 56-day run:

Hooks built now, empty

So later weather is data, not a rewrite:

Replacing the old calendar

• WeatherModel (Sim, pure) gives the hourly state; the forecast per period and the truth rolled from it; the daily chance for days 3-7.
• It replaces Calendar.WeatherOn / HumidityAt for every caller (Company, ClientTexts, HubComputer, GarageUI, HudController, MissionController, GarageController, EgoReactions). A WeatherType-style summary stays for the old callers until each one is moved.
• ForceWeatherCommand (host-only, debug) sets a day's weather.
• Tests: determinism (same seed → same week); calibration (each 10% bucket rains within ±3 points over 10,000 periods); the climate targets over a 56-day run; days 1-3 dry.

Needs: SD2 · Done when: the tests pass, and the old UI still shows sensible weather from the new model.

WE2: the sky follows the game clock

• Weather.cs takes continuous values (overcast, rain, fog, wind) interpolated between hours instead of three presets. Its visuals stay (smooth blending, showers, wet ground, rain slant), fed by the model.
• Clouds build 1-2 h before rain: the sky is the first warning in the field.
• Morning fog (URP fog density) is new.
• The HUD clock line shows the current period's icon and the next one ("rain from 15:00"). A client may say it ("Looks like rain, dear", through VoiceLines with WE4's lines).
• The same continuous values feed SoundWeather (SN1): rain and wind are heard from them, and the rain edge's position (rain sweeps in down the street) comes from the same seed and clock.
• DY1's running clock comes first; DY3 and this idea both edit the sky, so they're built back to back.

Needs: WE1, DY1 · Done when: on two instances a "60%" afternoon turns grey and rains (or not) at the same minute.

WE3: the weather in the work

Wet grass

• Lawn wetness = dew + recent rain − drying (sun, wind). It drives the existing humidity (heavier load, more clogs). Above ~50% wet clippings stick and drop as clumps (counted as clippings left) and the bag fills faster. These effects are read through grass condition: WE3 writes wetness into LawnWeather (GR2), whose wet row gives bag fill ×1.5 and clumps above 0.5, and MO6 drops the clumps.
• Dew dries by the sun (day.md): shaded cells stay wet longer.
• Shapes of a day, for the same lawn: a dry, clear day loses its dew by 11:00; after rain overnight the lawn stays above 50% until about 10:00; a 15:00 shower soaks it again for an hour or two. So the best plan differs: mow any time after 11:00 on the dry day, wait until the afternoon after a wet night, start early when a 15:00 shower is forecast.

Rain grows grass

• Soil moisture (derived from the last days' rain, never stored) multiplies daily growth by 0.8 + 0.5 × moisture in Company.GrassHeightFor (dry week ×0.85, average week ×1.05, wet week ×1.25). A wet week puts on ~2 cm more than a dry one, so clients call back sooner a few days later. Example: a lawn that grows 0.6 cm a day in late spring (×1.2) puts on 4.3 cm in a dry week (moisture 0.1), 5.3 cm in an average one (0.5) and 6.3 cm in a wet one (0.9).
• Soil moisture also goes into LawnWeather (GR2: lush grass, scorch).

Wind

• The leaf blower pushes further with the wind and barely works against it. Target: with the same blower in a 12 mph wind, a pile lands ~6 m away blown with the wind, ~2 m against it, and some leaves drift back.
• Mower and trimmer clippings drift downwind (onto patios, into pools).
• Leaf blows are replayed on every machine, so the wind must be the same everywhere: it is, because it comes from the seed.

Rain starts mid-job

• You can always keep working; it just gets wetter (heavier load, clumps, the bag fills fast).
• In heavy rain you may also stop: the client pays for the share done, no abandon review, no reputation loss. The rest can be booked again. Leaving a dry job early still gets the usual "left halfway through" review (AbandonReview).

Tests: the growth multiplier, the stop-in-rain settlement.

Needs: WE1, GR2 · Done when: a wet week makes clients call back sooner, and blowing against the wind is visibly worse on both instances.

WE4: where players see it, and bookings

Where it shows

Bookings

You pick the day when you book a newspaper client, and you can move a visit when rain is likely. The forecast the evening before decides:

• Anyone can make the call (MoveBookingCommand, actor recorded), from the phone.
• The washout call and "Looks like rain" lines go through VoiceLines.
• Tests: the washout rule; moving a booking costs no reputation.

Needs: WE1 · Done when: a guest moves a rainy booking and the host's calendar shows it, with no reputation lost.