v2.0.0

Every seed's output changes from 1.x.

• Dry spells: a semi-Markov rain chain; the year's longest dry spell is now 3% short of the stations' (6% in 1.x).
• The wettest days: Weibull amounts tied to how deep in the rain the day is; the 95th percentile is no longer light.
• Fronts: a day's temperature follows the rain on it and around it, warm before the rain and cold after.
• New fields: snow_depth (snow lies, melts in a thaw and settles as it ages), gust, fog, solar_radiation, and the fog and drizzle conditions. Sunshine is fitted on CRU sunshine records.
• One world per seed: weather_request::spatial, on by default, makes nearby places share their weather, with correlations that fall with distance as between weather stations. spatial = false gives each cell its own weather, as in 1.x.
• JSON always has "units".

A single day costs about 4x 1.x at a random place, more in winter in cold places (the snow season so far is simulated); a year's series about 1.5x. See Data and Method for the fits, the validation and the known shortfalls.

v1.1.0

Three opt-in request fields; a plain request gives the same output as 1.0.

• units: unit_system::imperial gives °F, inches and mph.
• hourly: each day's 24 hours in weather::hours, in local mean solar time. The temperature curve runs into the nights on either side, rain falls in spells that add up to the day's total, and cloud, wind, humidity and sunshine stay consistent with the day.
• historical: each month of 1991-2020 as it was, from CRU TS 4.10 monthly anomalies (resources/history/history.tsv, loaded with load_history(), not compiled in).

v1.0.1

Uses dasmig-core 1.3.0's degree trigonometry and its gamma, beta and Weibull draws in place of its own copies, which moved into core unchanged. Output and known answers are byte-identical to 1.0.0; requires core 1.3.

weather-generator 1.0.0

First release. Daily weather for any place on land and any date: the low, mean and high, rain, sleet or snow and how much, thunder, frost, cloud, humidity, wind, sunrise, sunset, day length and hours of sunshine. A Richardson-type weather generator (a rain Markov chain, gamma amounts, autocorrelated daily mean and range, month anomalies) around CRU TS 1991-2020 normals, fitted on 162 GHCN-Daily stations and checked against 119 it never saw. A day depends only on the seed, the place and the date, so single days and series agree. Points up to 100 km off the coast take the nearest land's weather and report the distance. climate_at for the normals, deterministic seeding, batches, JSON and CSV, layers, lite (2-degree) and full (half-degree) tiers, compiled-in data. The climate tables are adapted from CRU TS 4.10 and CRU CL 2.0 under the ODbL; see LICENSE_DATA.txt.