Weather Engine

Way back when I first started coding up this project (several years ago now), one of the first things I implemented was Alexis' weather generator.

It works pretty well. Working with real-world data as the seed input is easy. I've extended the system a little bit to make it more versatile throughout the year.

The only bits of data I need are: minimum and maximum monthly temperature in Fahrenheit (high and low), and the months in which the high occurs. I also collect high and lows for the precipitation (in mm) and the month in which that high occurs.

"climate": {
	"temperature": {
		"minimum": {
			"high": 77,
			"low": 47,
			"highMonth": 7
		},
		"maximum": {
			"high": 89,
			"low": 62,
			"highMonth": 7
		}
	},
	"precipitation": {
		"high": 126,
		"low": 70,
		"highMonth": 7
	}
}

With these numbers, it's a simple enough matter to fit a sinusoidal curve to give me the monthly average high and low for any given month. Admittedly, very few places on earth have a perfectly balanced summer/winter cycle, but this can be forgiven, and the players are unlikely to notice in any case (particularly across such a relatively small area).

A whole year of temperatures

I've incorporated a few extra bits into my version of the weather engine. The first is apparent temperature. At higher temperatures (80F+), high humidity makes the air seem hotter. This is usually reported as Heat Index. Conversely, when the air is cold, wind can make it feel even colder (Wind Chill).

Next, I used a IDW algorithm (my go-to) to spoof data for all hexes based on just a few inputs. I grabbed a few stations from around the State, but with this method, I don't need to worry about researching all 170+ points.

On one of the original weather posts, Vlad made a comment about how much rain it generates. These are some good thoughts for possible mods. I don't see that problem cropping up at the moment, but it's something I'll keep an eye on.

As another potential update, I could use a Gaussian distribution to generate the temperature for each day, or calculate the daily drift differently. But for now this works well.