Weather realism #34515

Kodiologist · 2019-10-06T16:41:22Z

Summary

SUMMARY: Balance "More realistic temperature, precipitation, humidity, and pressure"

Purpose of change

Closes #34139 - More realistic precipitation and temperature.

Describe the solution

The generation of temperature, humidity, and pressure using time and simplex noise has been overhauled to more closely match real 2010s data from New England, particularly Concord, New Hampshire, since New Hampshire is roughly in the middle of New England. I also edited the humidity and pressure thresholds for the various precipitation types so that the rate at which the player sees light precipitation (Drizzle, Flurries) or moderate precipitation (Rain, Snowing, etc.) is similar to the real data. Finally, I've added tests for temperature and precipitation realism.

I've moved the default INITIAL_DAY to Spring 60, which has a typical low around 35 °F, since a typical low on Spring 30 is now around 25 °F (from 35 °F before this PR).

While I've changed the amount of rainfall produced by generic_wet and generic_very_wet, I haven't adjusted the other effects (field decay and player wetness) because I don't know how to relate them to real weather data.

Some downstream gameplay effects:

Funnels are much less useful, both because rain is much less frequent and because rainfall rates are lower. (But, they already weren't very useful in a game full of rivers, swamps, and swimming pools.)
Also because rain is much less frequent, rain gear is less important and fire is more useful.

Describe alternatives you've considered

It's clear that many more changes could be made, such as:

Properly linking temperature to sun position
Generating absolute humidity instead of relative humidity
Making precipitation stochastic instead of a deterministic function of pressure and humidity
Representing rainfall as continuous instead of discrete (not wet vs. wet vs. very wet)

But I think the PR I'm proposing improves things, without requiring lots of code changes.

Additional context

An example weather log is attached. Here are some results from my analysis script:

``` TEMPERATURE (°F)

Mean absolute running difference - lows
CONCORD, NH 2014 - 6.385
CONCORD, NH 2015 - 6.432
CONCORD, NH 2016 - 6.134
BOSTON, MA 2014 - 4.261
BOSTON, MA 2015 - 4.514
BOSTON, MA 2016 - 4.414
CARIBOU, ME 2014 - 6.423
CARIBOU, ME 2015 - 6.016
CARIBOU, ME 2016 - 5.942
Cata - 5.866

Mean absolute running difference - highs
CONCORD, NH 2014 - 6.319
CONCORD, NH 2015 - 6.457
CONCORD, NH 2016 - 6.841
BOSTON, MA 2014 - 6.772
BOSTON, MA 2015 - 6.72
BOSTON, MA 2016 - 7.066
CARIBOU, ME 2014 - 6.162
CARIBOU, ME 2015 - 6.637
CARIBOU, ME 2016 - 6.419
Cata - 6.149

Mean within-day range
CONCORD, NH 2014 - 22.501
CONCORD, NH 2015 - 24.353
CONCORD, NH 2016 - 24.098
BOSTON, MA 2014 - 14.866
BOSTON, MA 2015 - 15.773
BOSTON, MA 2016 - 15.366
CARIBOU, ME 2014 - 18.781
CARIBOU, ME 2015 - 19.214
CARIBOU, ME 2016 - 19.421
Cata - 19.637

Extreme window center - lows
CONCORD, NH 2014 - 2
CONCORD, NH 2015 - 52
CONCORD, NH 2016 - 43
BOSTON, MA 2014 - 22
BOSTON, MA 2015 - 45
BOSTON, MA 2016 - 43
CARIBOU, ME 2014 - 2
CARIBOU, ME 2015 - 34
CARIBOU, ME 2016 - 352
Cata - 40

Extreme window center - highs
CONCORD, NH 2014 - 180
CONCORD, NH 2015 - 210
CONCORD, NH 2016 - 205
BOSTON, MA 2014 - 181
BOSTON, MA 2015 - 211
BOSTON, MA 2016 - 205
CARIBOU, ME 2014 - 180
CARIBOU, ME 2015 - 228
CARIBOU, ME 2016 - 194
Cata - 214

PRECIPITATION

Proportion of days with ≥1 mm of precipitation
CONCORD, NH 2014 - 0.293
CONCORD, NH 2015 - 0.266
CONCORD, NH 2016 - 0.238
BOSTON, MA 2014 - 0.26
BOSTON, MA 2015 - 0.258
BOSTON, MA 2016 - 0.235
CARIBOU, ME 2014 - 0.342
CARIBOU, ME 2015 - 0.342
CARIBOU, ME 2016 - 0.361
Cata - 0.212

Proportion of hours with light precip or more
CONCORD, NH 2014 - 0.039
CONCORD, NH 2015 - 0.031
CONCORD, NH 2016 - 0.028
BOSTON, MA 2014 - 0.039
BOSTON, MA 2015 - 0.032
BOSTON, MA 2016 - 0.029
CARIBOU, ME 2014 - 0.046
CARIBOU, ME 2015 - 0.035
CARIBOU, ME 2016 - 0.041
Cata - 0.039

Proportion of hours with heavy precip
CONCORD, NH 2014 - 0.014
CONCORD, NH 2015 - 0.012
CONCORD, NH 2016 - 0.01
BOSTON, MA 2014 - 0.016
BOSTON, MA 2015 - 0.01
BOSTON, MA 2016 - 0.012
CARIBOU, ME 2014 - 0.011
CARIBOU, ME 2015 - 0.006
CARIBOU, ME 2016 - 0.013
Cata - 0.015

Total precip per year (cm)
CONCORD, NH 2014 - 116.796
CONCORD, NH 2015 - 96.603
CONCORD, NH 2016 - 84.102
BOSTON, MA 2014 - 115.09
BOSTON, MA 2015 - 88.416
BOSTON, MA 2016 - 84.082
CARIBOU, ME 2014 - 114.754
CARIBOU, ME 2015 - 89.606
CARIBOU, ME 2016 - 111.868
Cata - 71.05

A plot of relative humidity (orange), vs. Concord in 2014 (blue):

A plot of pressure (hPa, orange), vs. Concord in 2014 (blue):

ghost · 2019-10-08T19:24:07Z

Looks good to me!

ghost · 2019-10-08T19:50:02Z

src/weather_gen.cpp

+   // Greatest absolute change from a day's average temperature, in kelvins
+constexpr double seasonality_magnitude_K = 15;
+   // Greatest absolute change from the year's average temperature, in kelvins
+constexpr double noise_magnitude_K = 8;  // 2;


// 2 commented-out previous code?

Yeah, that's the kind of code cleanup I'll do.

kevingranade · 2019-10-08T21:30:04Z

Will need some decent play testing, but in broad strokes this is IMO a drastic improvement that I'd love to have in the game.

Kodiologist · 2019-10-09T12:48:25Z

Awesome, I'll finish this up, then. @kevingranade, what is your opinion on the issue of moving the game start date?

ghost · 2019-10-09T12:53:53Z

src/weather_gen.cpp

+    ret.base_humidity = jo.get_float( "base_humidity", 50.0 );
+    ret.base_pressure = jo.get_float( "base_pressure", 0.0 );
+    ret.base_acid = jo.get_float( "base_acid", 0.0 );
+    ret.base_wind = jo.get_float( "base_wind", 0.0 );


I looked over the data output, and I saw that windspeed average was significantly lower than usual, and had far less variability over the year.

Infact the highest windspeed for the entire year was 13mph

Would this be due to this change? and also above to peaks and variation, and also the change to W = algorithm in get_weather() ? you didn't specify any rationale for change in windpower, although the changes to humidity pressure and precipitation look perfect.

Yeah, I simply haven't looked at wind data or adjusted the wind code in this PR, so it relates to pressure the same way it used to. Should I overhaul wind, too?

You set base_wind, base_wind_distrib_peaks, and base_wind_season_variation to 0, in this block of code. and changed the algorithm for W = in get_weather().

Wind , as far as I can tell, dosnt need overhauling because it was set to new england data in the same way as you are doing for pressure and humidity. it could be that your changes to pressure are resulting in a flatter and lower wind distribution, due to the relationship that it previously had with the pressure value.

Could you output the data , but with no changes to those wind variables?

Hmm, actually those are default loaded values if regional settings dont specify any. so it should still be the same, assuming you are using the standard regional settings.

It must be the pressure to wind relationship is different now.... may require some tweaking to get the previous wind behaviour with the new pressure

That's right: the default region settings are just less redundantly specified now. And the change on the line with the assignment to W is just a variable rename. The pressure-to-wind relationship is the same, but pressure is different, so wind is different.

It's up to you, wind is not vastly broken after this or anything, but it would be nice if you could find a relationship between wind and the new pressure that preserves as much of the old wind data distribution as possible - as that was matched to real-world data as close as possible in a previous PR.

If you want to do that in a separate PR, that would be fine, imo.

Okay, let's focus on the other variables for now and maybe I'll do wind in a new PR.

ghost · 2019-10-09T12:57:45Z

This is wind-speed per-hour in winter , before this change :-

After your changes, this is wind-speed per-20 minute interval in winter ( so similarish intervals. )

Kodiologist · 2019-10-19T18:12:24Z

Okay, this PR should be good to go now.

AlexMooney · 2019-11-01T22:26:09Z

@Kodiologist do you happen to have the data on rain fall handy for another chart? I'd love to see the pre- and post-PR CATA precipitation versus the real world data.

Kodiologist · 2019-11-01T22:27:45Z

Yes, in the form of the weather log linked to above.

kevingranade · 2019-11-10T17:10:03Z

This pull request has been mentioned on Cataclysm: Dark Days Ahead. There might be relevant details there:

https://discourse.cataclysmdda.org/t/never-any-rain/21901/5

kevingranade · 2020-04-27T13:35:13Z

This pull request has been mentioned on Cataclysm: Dark Days Ahead. There might be relevant details there:

https://discourse.cataclysmdda.org/t/missing-weather/23427/6

Kodiologist changed the title ~~Weather realism~~ [WIP] Weather realism Oct 6, 2019

ZhilkinSerg added [C++] Changes (can be) made in C++. Previously named `Code` [JSON] Changes (can be) made in JSON Game: Balance Balancing of (existing) in-game features. Mechanics: Weather Rain, snow, portal storms and non-temperature environment labels Oct 7, 2019

ghost reviewed Oct 8, 2019

View reviewed changes

ghost reviewed Oct 9, 2019

View reviewed changes

Kodiologist force-pushed the weather-realism branch from 8c45d94 to 531ab31 Compare October 9, 2019 13:39

Kodiologist changed the title ~~[WIP] Weather realism~~ Weather realism Oct 19, 2019

Kodiologist force-pushed the weather-realism branch 4 times, most recently from 7c66f0d to 44a7b81 Compare October 25, 2019 14:49

Kodiologist mentioned this pull request Oct 26, 2019

More realistic precipitation and temperature #34139

Closed

Kodiologist force-pushed the weather-realism branch 2 times, most recently from 595bcf7 to c24bde2 Compare October 26, 2019 21:11

Kodiologist added 9 commits October 28, 2019 16:08

For debug weather logs, use the current weather generator

1e06ca2

Remove redundant uses of the default weather values

1e48b4d

Simplify: year_fraction, ctn

a0deeca

Remove the extra variable seasonal_variation

d9d95a0

Make midwinter the coldest part of the year

b21d0f1

Temperature overhaul

3cf7633

Allow seeding test_weather

c7fa3f8

Pressure overhaul

59df742

Humidity overhaul

37bb0ce

Kodiologist added 5 commits October 28, 2019 16:08

Precipitation overhaul

5dda869

Refactor the handling of rain data

ab4c258

Add tests of weather realism

fa03e06

Advance the default INITIAL_DAY to 60

1e43bf6

Astyle

d9b5949

Kodiologist force-pushed the weather-realism branch from c24bde2 to d9b5949 Compare October 28, 2019 20:13

kevingranade merged commit 25a44aa into CleverRaven:master Oct 31, 2019

Zireael07 mentioned this pull request Nov 4, 2019

Survivor has unnatural heat and cold resistance #35306

Closed

ghost mentioned this pull request Nov 15, 2019

Weather/Wind doesn't change. #35535

Closed

sdp0et mentioned this pull request Nov 20, 2019

Vehicle Funnels can't fill tanks with water in rain #35599

Closed

ghost mentioned this pull request Nov 28, 2019

Weather Tweaks - new light drizzle category and fix wind #35759

Merged

Kodiologist mentioned this pull request Dec 27, 2019

Update tips for 0.E. #36464

Closed

Holli-Git mentioned this pull request Sep 25, 2021

Rain and Weather seems to work well for the first few days, before breaking into just sunny, cloudy, or clear #51861

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Weather realism #34515

Weather realism #34515

Kodiologist commented Oct 6, 2019 •

edited

Loading

ghost commented Oct 8, 2019

ghost Oct 8, 2019

Kodiologist Oct 8, 2019

kevingranade commented Oct 8, 2019

Kodiologist commented Oct 9, 2019

ghost Oct 9, 2019

Kodiologist Oct 9, 2019

ghost Oct 9, 2019

ghost Oct 9, 2019

Kodiologist Oct 9, 2019

ghost Oct 9, 2019

Kodiologist Oct 9, 2019

ghost commented Oct 9, 2019 •

edited by ghost

Loading

Kodiologist commented Oct 19, 2019

AlexMooney commented Nov 1, 2019 •

edited

Loading

Kodiologist commented Nov 1, 2019

kevingranade commented Nov 10, 2019

kevingranade commented Apr 27, 2020

Weather realism #34515

Weather realism #34515

Conversation

Kodiologist commented Oct 6, 2019 • edited Loading

Summary

Purpose of change

Describe the solution

Describe alternatives you've considered

Additional context

ghost commented Oct 8, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kevingranade commented Oct 8, 2019

Kodiologist commented Oct 9, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ghost commented Oct 9, 2019 • edited by ghost Loading

Kodiologist commented Oct 19, 2019

AlexMooney commented Nov 1, 2019 • edited Loading

Kodiologist commented Nov 1, 2019

kevingranade commented Nov 10, 2019

kevingranade commented Apr 27, 2020

Kodiologist commented Oct 6, 2019 •

edited

Loading

ghost commented Oct 9, 2019 •

edited by ghost

Loading

AlexMooney commented Nov 1, 2019 •

edited

Loading