31-08-2002 03:45									#1
	Random map scenarios Support Apolyton, buy Civilization III: Complete

In accordance with Mark's suggestion, I am starting a thread dedicated to issues relating to random map scenarios.

Demo 7.1 does have random maps, but they are not yet scenarios, since they have no civilizations.

The maps are simple fractals using a midpoint algorithm, with a sea level added later to set the amount of land.

Since the only variable is altitude, only altitude related characteristics are used. This means that the available terrain types are ocean, coastal sea, flat land, rolling, broken and mountains, in increasing order of altitude.

As Laurent said:

quote: Then we need hygrometry (desert/humid squares) and vegetation (forests/plains).

Both of these models are dependent on climate patterns, which I will add when I figure out how to do it. If anyone has any references on coding such things I would much appreciate any help here.

I certainly do want this completed in time for D8.

Cheers

	31-08-2002 07:31									#2
	Put an end to popups!

Hey Gary. On climate...

Here's a link to a site with an interesting digram of wind patterns and how they emerge. Just scroll down a little bit

http://www.franksingleton.clara.net/tw2001.html

I presume a crude estimate of water put into the air, combined with where the winds tend to blow it would be a first step. When you think you're a few weeks from doing the first crude pass at the climate coding let me know, and I'll see if I can lure Paul Crocker into working on it a bit. Paul was responsible for the first shot at a map generation model way back, and seemed to know his stuff! An eight-zone model like that shown in the diagram on that web page is the sort of thing he wanted to use.

Last edited by Mark_Everson on 31-08-2002 at 07:40

	02-09-2002 15:35									#3
	Increase Your PM Length

Please take a look at the ecology model. I have designed it with map generation in mind and set up a way to link climate with disasters, vegetation etc.

http://apolyton.net/forums/showthre...&threadid=36592

Last edited by Simon Loverix on 02-09-2002 at 17:03

	03-09-2002 08:00									#4
	Support Apolyton or Terrorists Win

I did reread the ecology model, but it tends to concentrate on rather minor ecological details (like wetlands) and does not give any real help in a large scale map generation algorithm, apart from saying things like "wind patterns should be taken into consideration".

What wind patterns? Where exactly are they? How do mountains affect them? How, exactly, do they generate high or low rainfall areas? How do they vary seasonally? In the long term?

In order to produce good map generating code I need specific answer to these and similar questions, in a similar detail to that provided for, for example, the social model.

Cheers

	03-09-2002 11:21									#5
	Support Apolyton, buy Civilization 2

I thought there was an effect of altitude stated in the model, like moving climate zone 110km towards the pole for every 100feet altitude or something, and a repartition of zones per latitude.

	04-09-2002 04:19									#6
	Support Apolyton, buy GURPS/ Alpha Centauri

quote: I thought there was an effect of altitude stated in the model, like moving climate zone 110km towards the pole for every 100feet altitude or something, and a repartition of zones per latitude.

That may help for temperature (if we knew the temperature distribution by latitude, which we don't), but does not help with wind dominated moisture levels. Nor does it help simulating things like the Gulf Stream.

Cheers

	09-09-2002 17:35									#7
	Support Apolyton buy from Amazon

Gary, are you referring to the most recent version of the ecology model (3.0)? Follow the link I pasted into my previous post, you can download it from there.

	10-09-2002 14:22									#8
	Help yourself to an AD-FREE life

Wind directions, as can be seen in one of the pictures on the page Mark_Eversons linked to:
0-30: east to west, trades
30-60: west to east, westerlies
60-90: east to west, polar easterlies
A rough way to similute moisture would be to see how long stretch of water a wind has covered before reaching land, and let the air moisture be proportional to that. Then for each land square, the moisture in the air is decreased, and instead the moisture on the land for the square is increased. Mountains should be more wet than plains, since they generate more rainfall.

Water currents often go in the same direction as the winds, but follow circular patterns. A sea large enough to be swept by both westerlies and trades, will have a clockwise current (if it lies on the northern hemisphere) because the south of it will be swept to the west, and the north of it to the east. When the water moves north it will be warmed, and by the time it moves back it will have cooled. So on the northern hemisphere east coasts should be warmer than west coasts.

There is an animated picture that illustrates the principle on http://mbgnet.mobot.org/salt/motion/current.htm

	11-09-2002 05:48									#9
	Support Apolyton, pre-order Civilization IV

Hi RM, thanks for the writeup and link! I think the solar flux hitting the water is another important factor in how much water vapor gets into the air. FE much more water should go into the air near the equator than at the poles. I'm hoping someone will turn up that can characterize a relatively simple model that still gets the important factors. I think we have most of the building blocks, just specifying them in terms of the map details is what's needed.

Of course, a relatively simple model might not work, but I think it stands a good chance of giving ok, or even intriguing maps.

	11-09-2002 14:04									#10
	Suffering from ads?

I did include a similar way to handle wind currents in the ecology model. But the page on the web site was never upgraded, though I sent a copy to Dom. Mark, can you put it on it's place? (link in the above post)
Thanks.

	12-09-2002 05:01									#11
	Support Apolyton or Terrorists Win

Hey Simon, thanks for the reminder. I've asked che to put it up. As Gary said though the v3 model doesn't have enough details to give him what he'd need for coding. Interested in making it more specific, per his comments?

	12-09-2002 05:38									#12
	Suffering from ads?

Well for wind patterns and percepitiation, most major wind currents generally flow west to east and this in combination with mountain ranges causes wetter areas to the west and dryer to the east.

Also like water, wind takes the path of least resistance, so a small grouping of mountains may not get much wind as the current flows around it.

Major post-industial Cities affect climate by creating local areas of high pressure zones above them that will bounce small fronts around them.

I'm not quite sure how major watter currents are formed, except they generally help regulate the ocean tempatures, but can be broke down easily with heavy rainfalls.

	13-09-2002 01:45									#13
	Avatar Enlargement: We've got the solution

I have taken a look at ecology model 3.0 and have some suggestions on wind and moisture distribution. It departs much from model 3.0 in these two areas, but I think it will lead to generation of smoother worlds and less special cases in the code. It is a rather long description, be warned.

It seems that in the model winds will be represented by vectors, and they will be rare compared to the number of squares. Why not let each square have its own (smaller) wind vector? It seems to me that if you want to use the winds for things like sailing movement you will still need the wind at each square, and then you will have to interpolate, if you use model 3.0. I would suggest generating wind direction for each square when generating the world and then be done with it.

I do not think mountains should turn the winds away completely, as suggested by model 3.0. Perhaps slowing windspeed (If you use 1 wind/square, then the vector length can be used as windspeed rather than actual stretch) is reasonable though.

Winds move mostly in west-east direction or east-west direction in 7 bands. The direction can be approximated for each square by something like f(y) = sin(5*pi*y/total_height_of_map) ), where positive direction is west. But you probably want to modify the strength of the winds by some latitude dependent amplification, and add some random noise, both in north-south and east-west directions, to make it less uniform and predictable. And change the winds with respect to mountains, of course.

Moisture can be generated by
1. First approximating general air humidity in each square using solar flux as suggested by Mark_Everson. Air humity near the equator should be very high, near the poles low. Air humidity in land squares should initially be zero.
2. Air humidity can then be "smudged" by the winds by an iterated process moving some of the air humidity in the wind direction for each square.
3. Moisture on land can then be calculated by some nice formula using air humidity, altitude and distance to poles and equator. You probably want to test empirically which formula makes the nicest moisture distributions before deciding exactly which. When you are done, you can release the air humidity data from memory, if you do not wish to use it any more. You probably only want it when generating the world.

Later when rivers and lakes are implemented, that can also be taken into account for in the moisture calculation process.

The main benefits of my suggestions:
1. One way to calculate moisture and winds for all squares regardless of climate zone ( temperate, tropical etc. ), no special cases in the code needed and most likely worlds that seem less "discrete".
2. Winds calculated for all squares ONCE at world generation. No interpolation between longer windstretches needed. (Local wind changes near cities, like Lord God Jinnai suggested, can of course still be implemented.)

	13-09-2002 02:33									#14
	Inflate your Upload Space

Interesting post, RM. And it wasn't that long, either.
I am a bit afraid that per-square (or -polygon) winds would still need some smoothing. Winds in a square depend on the neighbouring squares, so some interpolation is inevitable if you don't want to have winds like WWEWW in neighbouring squares, with the Eastward anomaly explained by nothing. I am not sure whether it would change the generation of terrain enough to bother with, though.

	13-09-2002 08:33									#15
	Full PM-box? Change here!

RM:
I still think intereplation would be the easier course to go given LD's response.

Also mountains should affect both direction and speed, depending on the type and area.

On the moisture thing:
When generating oceans, 1 point should be chosen ouside a continental shelf. Then fill all squares to X height like a bucket or whatever. This allows for below-sea level lands which otherwise would require more work later on.

The next step seems fine, assuming everything else works the way its suppose to.

Thus areas with high mostiure concentration and are fairly close to sea level would become swamps while the those far below sea level would become lakes. This also applies to high level areas too, but those are much rarer. Basically right now the differeance in determining a swamp vs. a lake is if the water can escape to the ocean.

Next would come rivers and underground water sources, both of which, would probably be randomly generated as i don't think we will be having a complex geology infrastructure.

Then Erosion needs to go last as it will affect lakes...perhaps making them larger in area and also adding land to deltas...

Finally vegitation.

	13-09-2002 13:14									#16
	Support Apolyton

LDiCesare:

How smooth the winds will be depends on how smooth the wind generating function will be. I think a sine function with a small random disturbance should be smooth enough. And if it is not, and interpolation is necessary, at least the interpolation is done only once.

I am thinking of the possible use of winds in the actual game as well as world generation. A ship moving from Spain to the Caribbean by sail should go faster than it will go in the opposite direction, for example. That would require interpolation during gameplay unless all squares already has a wind of their own, and unnecessary computations during gameplay will slow the game down. Perhaps this is a bit off topic for world generation, but I believe it is better to consider this problem/possibility before it arises.

As a side note about WWEWW winds, it is possible with a high pressure very close to a low pressure creating such patterns. But I guess winds involved have to be weak in order for it to be realistic.


Lord God Jinnai:

The ocean filling idea could work, depending on how common land below sea level that is not sea is. It is very rare, except in Netherlands.

Another idea for creating lakes, is to create them at the same time as rivers. Select areas with high altitude and moisture as starting points for a river, and let the river always move to the lowest adjacent square. If there are no adjacent square that is lower than the current square, then create a lake on the current square. Let the river continue from the lake to the lowest adjacent square.

	14-09-2002 01:53									#17
	Tired of ads?

Lands below sea level exist in every continent except antartica, so i think they are common enough to put in, several are quite large, such as the rift valley in Africa where a new plate is forming.

	14-09-2002 21:04									#18
	Support Apolyton or Terrorists Win

Thanks, RM, for the more detailed proposal. I suspect Gary would like yet more details before coding, like what exactly is the "nice formula using air humidity, altitude and distance to poles and equator". But every day its getting more in the direction of a useable model. Thanks for the comments everyone!

	16-09-2002 15:46									#19
	Help yourself to an AD-FREE life

LGJ:
Wetlands and lakes are generated separately, but only if a lake is adjacent to sea it becomes a wetland.
Erosion should be used when generating the height, terrain, etc. Especially if we are going to use tectonics. The river starting points used there can be used for the final rivers also.
A major city area is rather similar to a desert area considering vegetation, but the temperature rise is important for generating disasters.


RM:
The moisture takes into account latitude, height, climate zone, nearby rivers, ocean or sea and wind and water currents. It is in a step-by-step format and not in a formula because all those influences are themselves only generated in steps, they are not all available at once.

Wind and water currents are fairly complicated to generate, since they form one big system adapted to itself, so in fact you can't place them one by one if you want to be correct. As is said above, it is best to depart from certain latitudes (the roaring forties fe) that have a uniform direction for the entire planet and then make deviations. In fact, it is not that bad if we generate them rather randomly if we can't easily find a simple pattern.

Winds were indeed intended to stay the same for the whole game, except in the case of a major climatic disturbance, where they would be regenerated from scratch.
Winds per square could be possible if you can find a way to backtrack their effects. Since winds influence moisture, there should be a change when they are deleted and regenerated anew.
The effects of the currents as I described them are quite significant, so better don't put in too much of them. It was my aim to be able to use them for custom scenario's, where you want a certain area to be desert. If there is no reason for the ecology model that it should be desert, it will be fertile grassland soon enough. So you'd better place a dry wind.

	16-09-2002 20:46									#20
	Suffering from ads?

Mark_Everson:

OK, if you want a formula, here is one:
Moisture = altitude*a + air_humidity * b + distance_from_the_equator * c,
where a,b and c are constants.
But I think the only way to get a function that get the wanted result, is to try different functions and see which works best. Too many details may distract from what is important.

Simon_Loverix:

The way I see it is that there are two ways to generate wind patterns.

One way is to take a brief look at a wind map for Earth, and try to estimate a rough formula ( i.e. a sine function with a small random disturbance ) for the winds, and then use that formula.

The second way is to try to construct a physical model for winds, generate a few random(?) winds, and then try to make it fit together, perhaps by using an iterated procedure, like a numerical solution to a differential equation.

It is clear that it is more difficult to take the second way, and not even certain to get better results. That is why I would recommend the first method. If you can formulate a procedure using the the second way, it is worth a try, but noone has suggested any such algorithm yet.

And now to answer some of the more detailed objections to my proposal:

"Wind and water currents are fairly complicated to generate"

The generation will not be more complicated than we make it. I think a goal should be to make as simple algorithm as possible, while still getting good results.

"As is said above, it is best to depart from certain latitudes (the roaring forties fe) that have a uniform direction for the entire planet and then make deviations"

I think you will find that a sine pattern scaled correctly will give the correct direction of wind for those latitudes that are fairly uniform. And for latitudes where the winds are often unpredictable, it will give winds that are close to zero (which would be the average wind). But I agree that some randomness should be taken into account. (And mountains, if we can find an easy enough way to do it).

"It was my aim to be able to use them for custom scenario's, where you want a certain area to be desert. If there is no reason for the ecology model that it should be desert, it will be fertile grassland soon enough. So you'd better place a dry wind."

With my suggestions, you can still make sure winds face the directions you want, you just have to do it square by square. At least until there is a good map editor. Besides, I think reevalution of moisture should not be done very often, since things like that do not usually change over night, and it will probably involve costly calculations to evaluate it, regardless of which of our methods are chosen.

	17-09-2002 00:24									#21
	Support Apolyton, buy Civilization: The Boardgame

quote: OK, if you want a formula, here is one: Moisture = altitude*a + air_humidity * b + distance_from_the_equator * c, where a,b and c are constants.

That is precisely the values of a,b,c, or at least a relative value that we need. As long as this is a mathematical model with no randomness, it can be tested on paper. Having no idea what the values should be like, the coder may come up with silly default values and never be able to get them right. We need rough starting values. You don't have any idea how much tuning values is long and difficult. With every change in the military code, I am afraid the cost values I carefeully balanced get totally changed because one unit will become too string for its cost, for instance. I iterated thousands of automated fights and could get short results for 1000 fights when I checked the costs of units. Checking 1000 maps is not that fast, but that would be needed if we start from scratch.
Even if we delivered a program that generates moisture correctly, there are many other factors, like wind, size of the world, etc., which makes lots of variables to tune. If you could provide some starting figures, it would really be very useful.

	12-12-2002 14:39									#22
	Bump Increase the size of your Attachments

(bump)
See also that thread: http://apolyton.net/forums/showthre...&threadid=30811

	10-01-2004 23:31									#23
	Support Apolyton, buy Civilization III: Complete

Bump

This is another project dear to my heart, which I had partly implemented.

Cheers and PF

Last edited by Gary Thomas on 11-01-2004 at 00:07