Random Coastlines, Rivers, and Roads

On page 173 of the DMG, Appendix B: Random Wilderness Terrain, Gary Gygax wrote, "In any event, the DM must draw in rivers, large lakes, seas, oceans, and islands as these features cannot easily be generated by a random method."

I would add roads to the list of features his random method does not account for and challenge Gygax's conclusion that such features are difficult to generate randomly. Moreover, I am working on a method to generate such features to supplement Gygax's tables and am eager to find out if other DMs know of any such methods or have come up with any themselves.

As a starting point, I've calculated the likelihood of encountering the coast during an overland journey as roughly 0.6% per square mile traversed. This is arrived at by dividing the total number of miles of coastline on Earth by the total number of square miles of the Earth's land surface. A chance of about 0.3% per square mile traveled of encountering the coast when travelling across the ocean is arrived at by dividing miles of coastline by the total surface area of Earth's oceans.

A similar figure for the chance of encountering a river could be found by totaling the length of all rivers on Earth and dividing by land surface area. For some reason this information isn't readily available online.

Dividing the length of all Roman roads by the area of the empire in the 4th century gives a road-encounter chance of 2%.

I'm modeling these distributions from the real world in an effort to produce a geography that is somewhat familiar and, in the case of roads, mirrors something close to a medieval level of infrastructure.

So for overland travel, I'm proposing an additional d100 roll for each space entered with the following tentative results:

1 coastline
2-3 river
4-5 road

Once encountered, these features can be explored along their lengths if the party so wishes. The initial course of each feature will lie perpendicular to the path of travel that the party was taking. With the exception of roads (which will generally run straight), their courses can be determined by a roll on the following table as each new space is entered:

d6
1-2 turns clockwise
3-4 continues straight
5-6 turns counterclockwise

When roads encounter settlements (and perhaps rivers where there will be bridges, no doubt), a check for number of branches should occur, determined by rolling a d4, perhaps.

Thoughts?

I totaled up the world's 180 longest rivers and it seems there is less river than I imagined. The revised table would be something like:

d1000
1-4 river
5-10 sea coast
11-30 road
31-000 no result

Hriston said:

Thoughts?

Click to expand...

Hriston said:

I totaled up the world's 180 longest rivers and it seems there is less river than I imagined.

Click to expand...

I think your discovery about rivers is a good reflection of my thoughts - I think this is very hard to do at anything but a rather abstract, almost arbitrary, level.

When it comes to rivers, for instance, there is the question of "what counts"? I live in Australia, a rather dry country by European and even North American standards (more than one of the Australian entries in that Wikipeida "world's longest rivers" page are dry for much of their length most of the time, and the "lake" into which they flow - Lake Eyre - is a dry salt lake bed most of the time).

But there are still small streams, creeks that come up when it rains, many small coastal rivers with tidal estuaries, etc. I think this is fairly hard to capture with any fidelity via random methods (eg one is more likely to come across a river if near the coast than inland).

So my focus would be more on results that give a tenable feel, and perhaps err on the side of interesting features. (Just as the density of strongholds and patrols in Appendix C seems rather high.)

Hriston said:

I totaled up the world's 180 longest rivers and it seems there is less river than I imagined. The revised table would be something like:

d1000
1-4 river
5-10 sea coast
11-30 road
31-000 no result

Click to expand...

I think then technically and for clarity, you better change the first line to "1-4 Major River". If you don't, someone that doesn't understand your methodology might thing that every hex without a "river" is dry, when in fact many hexes that are not "river" will have tributaries to those rivers, or smaller rivers of various sorts. You'll need some way of determining when the river ends.

Even so, I think you'll end up with many of the problems you see in say Minecraft with its procedural generation. For example, you'll have a river that discharges on one end to the ocean, and yet at the same time has its source after you meander to its other end... in the ocean.

pemerton said:

I think your discovery about rivers is a good reflection of my thoughts - I think this is very hard to do at anything but a rather abstract, almost arbitrary, level.

When it comes to rivers, for instance, there is the question of "what counts"?

Click to expand...

For my purposes, "what counts" is "what's quantifiable". And that turned out to be the list of rivers from Wikipedia, which means, as @Celebrim has pointed out, that a result of 1-4 indicates a major river, one of the world's 180 longest. This subset of rivers includes those in both wet and dry climates. I believe the details need to be left rather abstract for the system to be adaptable to various established situations. Including a table, for example, that gives results of dry river bed, varying widths of river, waterfalls, rapids, or even underground watercourses, might be interesting, but may also produce some nonsensical results. I'll have to think about it. I'm comfortable for now with the idea that we're talking about one of the world's major rivers, maybe as wide as a mile or more when it is encountered, which may or may not be dry as seems suitable.

As for smaller rivers and streams, I'm planning on generating those on a smaller scale than the one mile or larger spaces of DMG, p 173. There's an article on page 7 of The Dragon #10, (Vol. 2, No. 4, October, 1977), that presents a system for generating a square encounter area about 1/13th of a square mile in area, assuming a scale of 10 yards per inch. It gives a probability of encountering a stream of 4% in clear, rough, and wooded terrain types, 5% in mountainous, and 10% in marshy.

Celebrim said:

I think then technically and for clarity, you better change the first line to "1-4 Major River". If you don't, someone that doesn't understand your methodology might thing that every hex without a "river" is dry, when in fact many hexes that are not "river" will have tributaries to those rivers, or smaller rivers of various sorts. You'll need some way of determining when the river ends.

Even so, I think you'll end up with many of the problems you see in say Minecraft with its procedural generation. For example, you'll have a river that discharges on one end to the ocean, and yet at the same time has its source after you meander to its other end... in the ocean.

Click to expand...

This seems like the simplest solution. Upon rolling the 'major river' result, roll immediately to determine the river's direction of flow, whether to the right or left as approached by the party. When determining the river's downstream course, it continues through all terrain types until reaching the ocean. When determining its course upstream, if the river continues into a space that contains the seacoast, it is assumed that the headwaters of the river are separated from the coast by some high ground, perhaps a row of bluffs or some low hills. If, on the other hand, the river's upstream course continues into a space containing hills or mountains, it is assumed that the river has its source in that space.

Hriston said:

For my purposes, "what counts" is "what's quantifiable". And that turned out to be the list of rivers from Wikipedia, which means, as @Celebrim has pointed out, that a result of 1-4 indicates a major river, one of the world's 180 longest.

Click to expand...

As I posted, though, some of those "rivers" - eg most of the Australian ones on the list - are dry most of the time.

This is because the Lake Eyre basin, while perhaps a large enough one by world standards, is mostly desert and has water flows only when rain (which is not all that common) occurs.

So some of those major river hexes are really hexes with dry riverbeds. (I'll leave it to you, or someone else, to work out the proportions.)

I agree that generating rivers on the fly is probably not very effective. A river is really a feature of existing terrain, not the terrain itself, and as such it seems like you'd want to go back and fill in rivers only after the terrain has been generated. You could do this using a random percentage per hex that you've found a river, but once you find a river you would want to extend it upstream and downstream using some other method. You can certainly have rivers that dry up, or vanish into cracks in the earth, or some other termination, and headwaters can be practically anywhere, so random termination of the river both upstream and downstream would probably work, but you would want it heavily weighted toward having a source in highlands or lakes, and a terminus in a large body of water.

Coastlines are a little bit different in that, by definition, they must extend in two directions. You can't just have a single coastline hex in the middle of nowhere. If you are generating terrain procedurally and come across a coastline I'd recommend immediately extending the coastline a fair distance so that you can determine the extent of the land mass. Similar to rivers, coastline can be thought of as not really the terrain itself, but the limit of the terrain. You can certainly use random methods to extend the coastline, but you'll probably need to limit the random results so that it doesn't cut across terrain you've already generated.

pemerton said:

As I posted, though, some of those "rivers" - eg most of the Australian ones on the list - are dry most of the time.

Click to expand...

I'm a little confused by this response. Is it that, in your estimation, having a dry bed makes them not rivers, or at least not major rivers? I'm pretty sure I addressed this issue in the post you quoted from, i.e. the 'major river' result indicates an encounter with the bed of a large drainage system. Whether the riverbed is dry or full of water, I think, needs to be left open to be adaptable to the current situation established in the fiction.

This is because the Lake Eyre basin, while perhaps a large enough one by world standards, is mostly desert and has water flows only when rain (which is not all that common) occurs.

Click to expand...

The way I'm conceiving of the river-generation subsystem is that it operates independently of whether the party happens to be in the Lake Eyre basin or the Amazon basin. The larger situation is meant to inform how the indicated feature manifests in the fiction.

So some of those major river hexes are really hexes with dry riverbeds. (I'll leave it to you, or someone else, to work out the proportions.)

Click to expand...

Yes, but I'm not sure if the proportions need to or even can be worked out. It seems to me that such details depend too much on season and local geography to do without to some extent overhauling the system presented in the DMG, whereas what I'm attempting is a subsystem that can be added on to it.

Schmoe said:

I agree that generating rivers on the fly is probably not very effective. A river is really a feature of existing terrain, not the terrain itself, and as such it seems like you'd want to go back and fill in rivers only after the terrain has been generated.

Click to expand...

Yes, well the procedure as I've outlined it so far is to determine terrain type for the space being moved into as per the DMG, and then make a separate roll to determine the existence of a coastline, road, or major river which feature would lie across the path of the adventurers in roughly the middle of the space in question.

You could do this using a random percentage per hex that you've found a river, but once you find a river you would want to extend it upstream and downstream using some other method. You can certainly have rivers that dry up, or vanish into cracks in the earth, or some other termination, and headwaters can be practically anywhere, so random termination of the river both upstream and downstream would probably work, but you would want it heavily weighted toward having a source in highlands or lakes, and a terminus in a large body of water.

Click to expand...

Yes, once established as existing, the river is understood to extend in both directions, unless the party is currently following a river-course, in which case the newly encountered river is a tributary to the one being followed and extends upstream in the opposite direction to that which is indicated by its direction of flow. Direction of flow is determined with a 50% chance of each direction (right or left with reference to the adventurers' path) and indicates what is upstream and what is downstream. Downstream, the river extends (course determined randomly as given upthread) until terminating at the coast (or an interior drainage which is simply a coast line that turns back on itself), although the possibility of a river disappearing underground, perhaps into a cave system, is an interesting complication that it seems desirable to add in. Upstream, the river's course continues through all terrain types except mountains and perhaps hills, in which case it is considered to have its source in that space. Also, if the river's upstream course runs into a space containing a coastline, then it is considered to have its source in some high ground that lies between it and the sea itself.

Coastlines are a little bit different in that, by definition, they must extend in two directions. You can't just have a single coastline hex in the middle of nowhere. If you are generating terrain procedurally and come across a coastline I'd recommend immediately extending the coastline a fair distance so that you can determine the extent of the land mass. Similar to rivers, coastline can be thought of as not really the terrain itself, but the limit of the terrain. You can certainly use random methods to extend the coastline, but you'll probably need to limit the random results so that it doesn't cut across terrain you've already generated.

Click to expand...

Yes, the underlying methodology calls for established features to override any results that would contradict them and re-rolls to be made. The shape of the coast is determined by the same random method that determines the course of rivers, i.e. roll d6. Results 1-2 indicate that the feature turns clockwise, 3-4 indicates that it continues straight ahead, and 5-6 indicates that it turns counter-clockwise, but any result that has the feature doubling back into a space where it has already been established not to exist must be re-rolled.

Thank you to everyone who has given me feedback. Here's what I have so far. The main change to my approach is to resolve the presence of major features of the landscape (i.e. seacoasts, major rivers, and the shores of major lakes) at what Gygax refers to on page 47 of the DMG as "the scale of such world maps", for which he gives a value of around 20 to 40 miles per hex. This, of course, reminds me of the World of Greyhawk folio map, so I've gauged the following procedure to a 30-mile hex "world map" scale and a "smaller scale" of 1 mile per hex for the resolution of cultural features such as roads and settlements, as well as the actual shape that major features take on the landscape. This "world map" scale allows for the occurrence of major features at a greater frequency, letting me replace the rather unwieldy d1000 roll with something perhaps more intuitive. These scales can be tinkered without too much adjustment, I think.

One thing I also find lacking in Gygax's system is the concept of a starting hex. This is, of course, because his system was conceived as a supplement to a more intentionally deliberate form of world building, the idea being that random generation would only be needed if the players ventured into an unprepared area. My interest here is to create a system whereby world geography can be generated entirely at random, so I've included a table for the generation of a starting hex, or any random hex, which gives the proportions that result from carrying out Gygax's table to its mathematical conclusion.


Terrain Type
To determine the terrain type of a random/starting hex at any scale, roll on the following table:

Major Landforms
For the purpose of land exploration, there is a 1 in 4 chance that a world map scale hex will contain a major feature. If a feature is indicated, check each smaller scale hex for the presence of the feature, a roll of 1 in 20 indicating that the feature lies across the direction of travel. Roll on the following table to determine the type of feature.

When exploring the ocean at the world map scale, there is a 1 in 20 chance that the hex will contain a land coastline. If the presence of a coastline is indicated, check each smaller scale hex for the coastline by rolling a 1 in 20.

Lake exploration is conducted at the smaller scale. Check each hex for the presence of lakeshore by rolling a result of 1-4 on a d10 if the lake is fresh, or 1 on a d12 if the lake is salty.

The run, or course, of a major feature, when entering a new hex, is determined by consulting the following table:

In any case, the runs of major features must lead into unmapped hexes, and any other result must be discarded and rerolled.

When a river is encountered, there is a 50% chance that it will flow in one direction or the other.

When exploring a river upstream, the river’s source will be found in any encountered hex that contains high ground (hills and mountains), or a major freshwater lake. If a seacoast or shore of a major saltwater lake is encountered, high ground is presumed to exist between the river’s source and the body of water. A river will continue upstream through all other terrain types.

Downstream, a river will continue through all terrain types, reaching its terminus only upon encountering a seacoast or a major lakeshore. If high ground is encountered, it is assumed that the river continues on through a canyon, or possibly through an underground cavern at the DM’s discretion.

A river running through desert terrain tends to be dry unless it has rained recently, or there is reason to believe it is fed by meltwater or a perpetual spring.


Major Roads/Trackways
During land exploration, there is a 1 in 10 chance that each smaller scale hex contains a major road lying across the path of travel, running roughly in a straight line going in either direction. If such a road is indicated, there is a 2 in 10 chance it is a paved road.

If a road intersects with a river, an appropriate bridge, ford, or ferry will be located there, and the road will continue on the opposite side.

If a town or city is encountered, roll a die to determine the number of roads that branch out from the settlement, including any on which the party has arrived. For towns, roll a d4. For cities (including ruined cities), roll a d6. Observe the same chance of pavement as noted above.

As with major landforms, paths, tracks, and streams, roads must lead into unmapped hexes.


I hope this is of some small interest to others who are inclined towards random world building.