Weather Matrices for Real Locations [Simulationism]

Hi folks, I’ve been working on a project generating weather in RPGs and I wanted to share it. The concept is—given some environment, can we realistically generate its weather in a way that is easy to use at the table? I'll start out with my answer, which looks something like this. The idea is, each day is reduced to a weather code, giving temperature, cloud cover, and rain. Then you get a d100 table for that type of day, which determines how you transition to the next day. (The percentages on the first column give the average probability of encountering that type of day).



Now the background: I wanted to make a system that (1) relies on real data; (2) is algorithmic; and (3) is easy to apply to new environments. The idea being, then I can make it work anywhere we want.

To do so, I took data from the ERA5 dataset (accessed via Open-Meteo)—this is a big global weather database, that has weather for any location in the world since 1940. People generated this by taking weather records from different stations, combined with modeling efforts, to get global coverage. That means there is a record we can use for the weather of any place we want, with decades of data.

Then, I reduced each data to a code, for temperature, precipitation, and cloud cover. (I missed out on wind—which is complicated—and fog—where the data is poor). I then computed the day-by-day transition probabilities: e.g., if it is hot and overcast today, how likely is it to be hot, overcast, and rainy, tomorrow?

Here is a breakdown of the thresholds I used:

Temperature

Cloud Cover

Precipitation

Doing this, and ignoring low probability (<1%) days, gives you a ‘transition matrix’ for a particular locale. I then cast these into d100 tables. The idea here is, you find the type of day you are on, roll on the corresponding column, and use it to pick the new kind of day.

With a d100 transition matrix, you end up matching the transition probabilities pretty exactly. Here, for example, are the transition probabilities you get by rolling vs the data for Chicago winter (NB: transposed from the above).

Match to Real Data

And some other examples that I've generated.

Cairo Summer:

Seattle Autumn

Singapore Spring

I’m planning to build these for a wider variety of locales (e.g., 1 for each of the major climate classifications). But before I do I wanted to share these and get feedback about parts that are missing or ways they could be improved. I’ve written more about the process, mostly for my own edification, if anyone is interested in more depth (part 1, part 2, part 3, part 4, part 5). I’d love to hear any feedback or comments, positive or negative.

Some outstanding questions I have:
- Seasons will differ for different locales. How do I move past Spring/Summer/Autumn/Winter? I suppose location specific?

-Are the temperature, precipitation, and cloud cover cutoffs reasonable? Does anyone have clever ideas to get fog or wind?

-Is the presentation intuitive or confusing? Are there ways to improve it?

-Is 1% a reasonable cutoff, or would including more extreme (or less extreme) weather be preferable?

-Are there any locations you’re interested in seeing a table for?

You first give the premise as "can we realistically generate its weather in a way that is easy to use at the table?"

then present a complex table of random numbers - so I assume the answer to your question is no

It could be that you're using Fahrenheit rather than the much more sensible Celsius scale
Also your % arent useful as presented for gaming purposes

Its a great idea though, just being able to do it so its both realistic and simple is going to be a challenge, especially once you account for seasons and extreme events

Tonguez said:

You first give the premise as "can we realistically generate its weather in a way that is easy to use at the table?"

then present a complex table of random numbers - so I assume the answer to your question is no

It could be that you're using Fahrenheit rather than the much more sensible Celsius scale
Also your % arent useful as presented for gaming purposes

Its a great idea though, just being able to do it so its both realistic and simple is going to be a challenge, especially once you account for seasons and extreme events

Click to expand...

This is I think as realistic as you can get without using a computer. I can condense it to a smaller table at the expense of accuracy. (For example, just change the daily probabilities on the left column to a single d100 table, and ignore transitions). It's about the complexity of an old attack matrix, which ime doesn't take too long. But, I think the format is intimidating compared to single roll tables.

The seasons, fortunately, are not hard to get. I have versions for all four seasons of each location and it is easy to generate more.