Change pattern for intermittent water areas

[jeisenbe]

Expected behavior

Intermittent and seasonal areas of water should be rendered with a natural-looking pattern, as is the case for other areas of natural landcover. For example, a blue dot pattern, as seen in the two lakes on the right in this sample image:

Actual behavior

Intermittent lakes and rivers are rendered with strong horizontal blue and white stripes. This looks unnatural and does not work well with narrow lakes or rivers.

Intermittent river polygons look particularly strange:

The dashed line for intermittent waterway=river also clashes with the horizontal stripes:

Proposed fix

Use a random dot pattern, similar to those used for sand or beaches, with blue on a light background, for intermittent or seasonal natural=water. I believe this should also be done for waterway=riverbank polygons and natural=water with water=river.

Text example with natural=riverbank:

[jeisenbe]

The remaining question for river water polygons is whether the pattern should be transparent, so that we can see the underlying surface (eg sand, shingle, grass). But this would allow the line of the waterway=river to show as well.

I've updated the original post with some examples of intermittent rivers, and a test rendering I made.

[matkoniecz]

For example, a blue dot pattern, as seen in the two lakes on the right in this sample image:

It looks like a wetland for me. I would need to check discussions whatever it was intentional but current rendering of wetlands areas makes this mistake unlikely (at least for me) and (at least for me) works well to suggest that it may be either full of water or completely dry.

[kocio-pl]

I like their current look. We try to show something that can't be depicted in a simple way, something unusual - passage of time on static image, so it's the upside of "unnatural" rendering. We use more natural patterns (like what you've proposed) to show natural looking areas.

[jeisenbe]

What do you think about the intermittent river water areas? I find that the strong horizontal lines make it harder to visually follow the riverbanks, and the lines clash with the dashed line of an intermittent river waterway.

…

On Mon, Oct 22, 2018 at 3:49 AM kocio-pl ***@***.***> wrote: I like their current look. We try to show something that can't be depicted in a simple way, something unusual - passage of time on static image, so it's the upside of "unnatural" rendering. We use more natural patterns (like what you've proposed) to show natural looking areas. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#3468 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AoxshDEKhiJ1qtCnBi-JubMf40-tB6_Bks5unMGvgaJpZM4XycbR> .

[Adamant36]

1. What would the dot pattern look like with a river line underneath it? It would probably make it harder to distinguish.
2. It would probably make it impossible to tell that a water body is intermittent higher zoom levels, for instance the intermittent lakes in California that are currently visible as such at z6, due to the dots potentially being to condensed at that point.
3. It probably wouldn't look right with a surface underneath it. Unless it was semi transparent. Which would probably take extra, unnecessary at this point, code to pull off with surfaces as exceptions to normal rendering.

[jeisenbe]

Heres an example with a river aligned east-west. I think this doesn't work well with the current horizontal line pattern:

Priluk, Russia https://www.openstreetmap.org/#map=13/61.3360924/42.981087

Current rendering z13:

Dot pattern z13:

But vertically-oriented rivers and lakes also look a little strange now. You can also see intermittent water near wetlands (mangroves) in the image below:
Timika, Papua Indonesia https://www.openstreetmap.org/#map=10/-4.6411825/136.9440304

[jeisenbe]

1. Re: "What would the dot pattern look like with a river line underneath it?"
I believe waterways currently render over polygon areas, not under. This makes sense because you still want to be able to see the course of the river through an intermittent or seasonal waterbody.

Current rendering with waterway=river:
https://www.openstreetmap.org/#map=15/-4.4002763,136.9389841

Dot pattern for natural=riverbank, with waterway=river:

Current rendering with intermittent river:

Dot pattern for water area, with waterway=river intermittent=yes:

2. Re: intermittent lakes at z6:
The dot pattern doesn't change with the zoom level, so I believe legibility is improved compared to the current pattern.

Australia has many intermittent lakes, small to huge. Here's an example, "Lake Yamma Yamma (Way 220925120) (note that I added surfaces eg grass, shingle, mud, for testing in the small lake)

z6 dots (note this is only the coastline and lakes; I couldn't download all of Australia)

z6 stripes

z7 dots

z7 stripes

z8 dots (note that the test surfaces are visible in the smaller lake at this level)

z8 stripes

3. Re: Surfaces under water.
This is easy to implement, because the pattern is transparent. I believe this is already the case.
These are some examples I made in JOSM; top right to bottom left are grass, sand, shingle, bare_rock, scrub, mud and wetland=marsh. (The last two look strange in both of the styles; I believe there is an open issue about mud wetland rendering on water)

Current rendering z12 with landcover:

Dot pattern z12 with landcover:

z9 dots:

z9 stripes:

@Adamant36, if you have time to test this in California, just add this to /symbol:
, and change the code in water.mss from intermittent_water.png to water_intermittent.png

[Adamant36]

@jeisenbe, thanks for the tests. It actually looks a lot better then I thought it would. I'll have to give it a try in California when I have the time. What did you use to create the pattern?

[Tomasz-W]

I'm against this change, beacuse we need a way to show some property of area (so current rendering works good in this aspect), not it's surface.

[jeisenbe]

I've found some rivers are mapped with intermittent water plus natural=sand for sand islands that are covered seasonally. The horizontal stripes obscure too much of the sand pattern. A random dot pattern makes the underlying sand more apparent:

https://www.openstreetmap.org/#map=15/61.3362/72.0177

Regular stripes at z15:

Random dots at z15:

Regular stripes at z16:

Random dots at z16:

[jeisenbe]

@imagico made the pattern, I downloaded it from his fork. I believe it was made with his pattern generator tool. It might be nice to try a few more patterns, eg more closely spaced dots or slightly larger dots. If others are interested in seeing this

…

On Mon, Oct 22, 2018 at 5:59 PM Adamant36 ***@***.***> wrote: @jeisenbe <https://github.com/jeisenbe>, thanks for the tests. It actually looks a lot better then I thought it would. I'll have to give it a try in California when I have the time. What did you use to create the pattern? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3468 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AoxshGxKQu0VhNRHoq3hqXrFTasKdSIaks5unYjngaJpZM4XycbR> .

[jeisenbe]

Re: “We need a way to show some property of area, not it's surface.” What do you mean? I don’t understand this sentence. The suggested change is from a regular pattern of stripes to an irregular pattern of dots. Both are transparent, so both the current and proposed rendering show underlying sand/rock/grass etc.

…

On Mon, Oct 22, 2018 at 7:55 PM Tomasz Wójcik ***@***.***> wrote: I'm against this change, beacuse we need a way to show some property of area (so current rendering works good in this aspect), not it's surface. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3468 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AoxshAY-sbze4yPtehSvJK11nr4ehvymks5unaQegaJpZM4XycbR> .

[kocio-pl]

What do you mean? I don’t understand this sentence.

I guess the same as what we claimed before - that regular pattern does not suggest that it's kind of wet surface, it's clear that it's something more abstract.

This kind of pattern looks nicer in some places (with intermittent river, I admit), but it's misleading as a symbol for me.

[jragusa]

Please also consider #709 for which I also proposed to use the same pattern than intermittent water

[jeisenbe]

Good idea. I agree that retention basins should be rendered as intermittent water, because they are usually dry but occasionally fill with water. The pattern used can be the same as with natural=water plus intermittent=yes or seasonal

…

On Wed, Oct 24, 2018 at 7:02 PM Jérémy Ragusa ***@***.***> wrote: Please also consider #709 <#709> for which I also proposed to use the same pattern than intermittent water — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3468 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AoxshHQjxP7ZztIcg3LiOTf_aZ7fjSmKks5uoDq1gaJpZM4XycbR> .

[turnsole80]

I like the idea of change in this case. The lines are a bit too strong. Although the dots are very wetland like, that's not such a problem, since the idea should be to warn of areas that may be wet or hazardous

[jeisenbe]

In the USA, intermittent lakes and pools are considered wetlands, according to the government, and some wetlands are only seasonal. Personally I find the random dots to be not much more similar to wetland rendering than the horizontal stripes. The current generic wetland pattern is horizontal dashes.

…

On Wed, Oct 24, 2018 at 9:50 PM Chris ***@***.***> wrote: I like the idea of change in this case. The lines are a bit too strong. Although the dots are very wetland like, that's not such a problem, since the idea should be to warn of areas that *may* be wet or hazardous — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3468 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AoxshA_ds5qzNQ51dowVQJKYHkghiAnkks5uoGIUgaJpZM4XycbR> .

[matkoniecz]

Personally I find the random dots to be not much more similar to wetland rendering than the horizontal stripes. The current generic wetland pattern is horizontal dashes.

Note that it is preferable to have map that is as clear as possible to someone new, not only to ones who know it well.

according to the government

I am not convinced that legal classification is important - for various reasons things are legally classified in sometimes bizarre ways.

[turnsole80]

I would suggest following that which is used across multiple platforms, and/or best practice, which IMO would lean toward using dots rather than stripes

[matkoniecz]

I would suggest following that which is used across multiple platforms, and/or best practice, which IMO would lean toward using dots rather than stripes

Can you give examples of maps showing intermittent status of water areas? AFAIK this information is rarely displayed and usually in ways not suitable for this map style.

I am pretty sure there is no standard for that.

Such research takes long time, but would be useful (assuming that it looks for rendering overview, not only for map examples that match closely to the preferred rendering).

[Adamant36]

@matkoniecz, the USGS uses dots in the style suggested here to show intermittent water bodies on some of its topographic maps. Although I can't find the specific map where I saw it, here's a link to their map symbols list. Which shows it on the 3rd and 4th pages. Interestingly, they also show some intermittent lakes as lines also. So I'm not sure which is the preferred method, but at least it shows that intermittent status being used by a major map maker.

https://www.wvgs.wvnet.edu/www/maps/topomapsymbols_MapX1B.pdf

[jragusa]

@jeisenbe could you try with thinner blue stripes ? That would resolved the rendering with intermittent river

[turnsole80]

Not a problem. I was able to assemble a couple:

Australia:

Australia/New South Wales:

Australia/Queensland:

New Zealand:
Not rendered

South Africa (historical):

[jeisenbe]

Thank you Chris, that was helpful. Does anyone have links to examples of other possible renderings for intermittent lakes and river areas? Joseph

…

On Sat, Oct 27, 2018 at 8:10 PM Chris ***@***.***> wrote: Not a problem. I was able to assemble a couple: *Australia:* [image: image] <https://user-images.githubusercontent.com/26226665/47603219-bedb5580-da34-11e8-91b1-777426892b30.png> *Australia/New South Wales:* [image: image] <https://user-images.githubusercontent.com/26226665/47603235-06fa7800-da35-11e8-808c-fa776f49974f.png> *Australia/Queensland:* [image: image] <https://user-images.githubusercontent.com/26226665/47603128-445e0600-da33-11e8-8784-9225a552028c.png> *New Zealand:* *Not rendered* *South Africa (historical):* [image: image] <https://user-images.githubusercontent.com/26226665/47603184-41afe080-da34-11e8-95ff-d5ba1c2e37de.png> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3468 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AoxshK5WGUWAJz_f8aESVmjgACN0N1oEks5upD8LgaJpZM4XycbR> .

[jragusa]

I have published an example on french totopographic map in the previous discussion #996:

[jeisenbe]

Thank you for the link. This is the French topo map mentioned: https://user-images.githubusercontent.com/9897203/36991179-340db25a-20a7-11e8-9854-97ec997cf850.png But note “The horizontal hatching in French topographic maps indicates Zone inondable - which means area subject to flooding which is not the same as intermittent waterbodies in OSM” - see http://www.viewranger.com/help/FranceLegend.pdf “Zone inondable has a fairly specific meaning in French - see https://fr.wikipedia.org/wiki/Zone_inondable - that is strongly tied to French climate and the flooding patterns“

…

On Mon, Oct 29, 2018 at 5:57 PM Jérémy Ragusa ***@***.***> wrote: I have published an example on french totopographic map in the previous discussion #996 <#996>: [image: intermittent] <https://user-images.githubusercontent.com/9897203/36991179-340db25a-20a7-11e8-9854-97ec997cf850.png> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3468 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AoxshKpOnXY3am4ufuag0UnoeG2uc9kWks5upsL5gaJpZM4XycbR> .

[turnsole80]

I have to say, I'm really taken with the way the South African maps handle water (and sand too actually - I think there are some ideas there we could work with), so that would be my preferred route to pursue.

[kocio-pl]

Which example do you mean, all of them in general or some particular things?

[turnsole80]

Just the rendering of intermittent water for now. As far as the sand is concerned, that can wait until another time. The other water features, such as rice, can also wait, but are worth noting for when the time comes :)

[jeisenbe]

I managed to make version of the current intermittent water pattern, but with stripes 1/2 thickness. I believe it should be close to 1 pixel now.

This looks much better and it is more similar to the French map above, although I still prefer the dot pattern as seen in maps used in most English-speaking countries, especially for rivers and long horizontal lakes (see z8 and z10 below)

However, I now believe that the dot pattern really needs to have an outline, either a thin blue line or a dashed blue line. This would also improve the stripe pattern, as seen in the French map.

I've also tried examples with an approximately doubled density of dots. These are not entirely random; I just copied and rotated the original image by 90 degrees and overplayed the result. I would need to use @imagico's random pattern generator before making a real PR, but this gives and idea of what is possible. This denser pattern would be better for thin lakes and rivers.

Does anyone know how the HikeBikeMap style manages to combine water polygons before drawing outlines? Is there some preprocessing of water polygons involved? Outlines would be even better.

Examples in Northern Territory, Australia:

z5 Northern Territory

1. Current wide stripes
   

2. Narrower stripes
   

3. Dots
   

4. Double Dots
   

(@Adamant36 is this low zoom enough for you? 😄 )

z6 Northern Territory, Australia

1. Current wide stripes
   

2. Narrower stripes
   

3. Dots
   

4. Double Dots
   

z7 Alice Springs, NT, Australia

1. Wide stripes (current rendering)
   

2. Narrow stripes
   

3. Dots
   

4. Double Dots
   

z8 Alice Springs, Northern Territory

1. Wide stripes
   

2. Narrow stripes
   

3. Dots
   

4. Double dots
   

z9 NT Lakes

1. Wide stripes
   

2. Narrow stripes
   

3. Dots
   

4. Double dots
   

z10 Lake Amadeus, near Alice Springs

1. Wide stripes
   

2. Narrow stripes
   

3. Dots
   

4. Double dots
   

[kocio-pl]

Thanks for testing.

Narrow stripes are hardly visible for me and it looks like a paler shade of water. Effectively it is the same problem as with all the dots: it looks like new kind of natural water-related surface, which it isn't. Wide stripes look the best for me, since they clearly suggest that the water is still a water (the same color) and this is not depiction of a surface but some abstract idea (changes in time).

[sommerluk]

Agree with @kocio-pl . The stripes might have a not so pleasant look, but it represents quite well that there might be either a lot of water or no water at all. (The dots however look to me like something in the middle, like wetland instead of intermittend).

[Adamant36]

@sommerluk, less dots perhaps? In the example maps there's a lot more space between them. Maybe that would help.

[kocio-pl]

I don't see a point in more testing this idea, because problem is not related to a density of dots. It was nice to see another solid research by @jeisenbe instead of just guessing, but it proved this looks like a surface with every other pattern. It seems unlikely to me that further playing with this idea will change anything, so I will close this ticket now.

[jeisenbe]

Does anyone else think thinner stripes would be an improvement? It certainly makes it easier to follow the outline of the water body. Right now the 2 pixel tall gaps between blue stripes lead to significant visual artifacts when the edge of the water area is complex, as seen in the zoom 7 to 10 examples above. I was going to try testing adding a 0.5 to 1.0 pixel outline for intermittent water bodies to help with this problem; this looks really nice on the version of the HikeBikeMap that is deployed on http://MapOSMatic.osm-baustelle.de/new/

…

On Sun, Nov 4, 2018 at 1:51 AM kocio-pl ***@***.***> wrote: I don't see a point in more testing this idea, because problem is not related to a density of dots. It was nice to see another solid research by @jeisenbe <https://github.com/jeisenbe> instead of just guessing, but it proved this looks like a surface with every other pattern. It seems unlikely to me that further playing with this idea will change anything, so I will close this ticket now. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3468 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AoxshGrbzN-dN9bn5U8n4Bsj5bg5hyGzks5urcl9gaJpZM4XycbR> .

[Adamant36]

I don't know. I like the thinner stripes if its zoomed in more, but zoomed out they blend together to much and you lose the effect.

[sommerluk]

Does anyone else think thinner stripes would be an improvement? It certainly makes it easier to follow the outline of the water body.

Smaller stripes makes it is indeed easier to follow the outline – especially for water bodies having a bigger est-west distance than north-south.

However, this obvious benefit has to be compared with the disadvantage that smaller stripes are more difficult to recognize as stripes (as @kocio-pl said, “it looks like a paler shade of water”).

For me, in this case the disadvantages weight more than the benefits.

[jeisenbe]

Here are some examples with a 0.4 to 1.2 pixel outline. This really helps with smaller water bodies and at low zoom levels. For the dotted pattern I've tested a dashed outline at z7 and up, but for the stripes I have used a solid line because dashes would interfere with the stripe pattern. I set the size of the outline based on the size of the waterbody, but it would look similar to set it by zoom level

On my 1280 x 800 monitor the thin stripes still look good, but I imagine if you are using a retina display (eg recent smartphone or new laptop) they would be too thin. This is a problem for many of our patterns, which were clearly designed for standard desktop resolution displays.

Examples in Northern Territory, Australia:

z5 Northern Territory
Current

2 pixel stripes + outline

1 pixel stripes + outline

Dots + outline

z6 Northern Territory, Australia
Current

2 pixel stripes + outline

1 pixel stripes + outline

Dots + outline

z7 Alice Springs, NT, Australia
Current

2 pixel stripes + outline

1 pixel stripes + outline

Dots + dashed outline

z8 Alice Springs, Northern Territory
Current

2 pixel stripes + outline

1 pixel stripes + outline

Dots + dashed outline

z9 Lakes
Current

2 pixel stripes + outline

1 pixel stripes + outline

Dots + dashed outline

z10 Lake Amadeus, SW of Alice Springs
Current

2 pixel stripes + outline

1 pixel stripes + outline

Dots + dashed outline

[StyXman]

@jeisenbe do you have this tests in a branch somewhere? I definitely like the dots + dashed outline.

[jeisenbe]

I've rebased the branch "intermittent" and pushed it to https://github.com/jeisenbe/openstreetmap-carto/tree/intermittent This has the new pattern (borrowed from @imagico - in the osm-carto-alternative-colors style) and also a dashed outline for intermittent water areas.

[jeisenbe]

The current pattern still works poorly for intermittent river areas.

And as mentioned above, the dashed line for intermittent waterway=river features clashes with the horizontal stripes, and strong horizontal lines make it harder to visually follow the riverbanks.

[jragusa]

The dots + dashed outline you proposed previously looks good to me.

Can you test this pattern on your example above ?

[jeisenbe]

Rivers are the main problem with the current pattern (though intermittent reservoirs often have the same issue at mid and low-mid zoom levels , if they follow a narrow river valley):

z15 Aroyelvi river, Norway

It took me a minute (and zooming in) to figure out the meaning of this crazy zig-zag pattern

z16

dots + dashed outline you proposed previously looks good to me

For intermittent rivers the dashed outline isn't possible without significant pre-processing, since most rivers are made of more than one area, and the outline would show in between.

I think the dot pattern works well for natural=water (lakes) but not so well for waterway=riverbank features. There we would need a more solid pattern so that the shape of the river is more visible, and so the dashed waterway=river line is not showing.

[jeisenbe]

Here's the test images as requested. The outline works in this case, but it won't work in all situations:

z17 Montpellier - le Verdanson - dot pattern and dashed outline

z18

No outline - this would be feasible

z17 le verdanson dots no line

z18

Futaspranget, Norway

https://www.openstreetmap.org/#map=17/61.27694/7.16017

z17 current

z17 dots

z16 dots

z15 dots

With a waterway=river + intermittent=yes line, the strongly dashed pattern is quite prominent with both the current pattern and with the dots.

[jeisenbe]

I didn't originally understand why @imagico used a different pattern for intermittent river areas in the alt-colors style, but I now see that this is necessary to deal properly with the intermittent river lines showing through the pattern when it is less solid.

Here are examples with the new river color and the river_intermittent.png pattern from alt-colors:

z17 le Verdanson, Montpellier

z18

z17 Futaspranget, Norway

z16

z15

We will have to think about this while picking a pattern for water with salt=yes

[jeisenbe]

Todd River, in Alice Springs, Australia.

• I believe this river is almost always dry, but the water runs after rare, heavy rainfall.

z15 Todd River current

z16 current

z15 dots - Todd River

z16 dots

z15 river_intermittent.png more solid pattern

z16 river_intermittent.png

(It would also look better if the dashes were longer for intermittent river lines)