Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Stream is barely visible on scrub #2098

Closed
sommerluk opened this issue Mar 27, 2016 · 60 comments
Closed

Stream is barely visible on scrub #2098

sommerluk opened this issue Mar 27, 2016 · 60 comments

Comments

@sommerluk
Copy link
Collaborator

Stream is barely visible on scrub.

The waterway=stream is barely visible on the natural=scrub, especially when combined with intermittent=yes. Screenshot from http://www.openstreetmap.org/#map=19/5.34678/-3.98195
screenshot

Not sure what would be a good solution for this, if any…

@SomeoneElseOSM
Copy link
Contributor

It's bug-for-bug compatible with real life - even on the ground in summer you sometimes can't tell where the streams would run after it's been raining :)

@moliha
Copy link

moliha commented Mar 28, 2016

But the map should help finding streams and other waterways. This could be easier, if the color used for waterways had better contrast to the various greens used for forest, scrub etc. I think a darker and more blue color for waterways would be better.

@kocio-pl
Copy link
Collaborator

Similar to #1781 and #1427.

@kocio-pl
Copy link
Collaborator

Unfortunately still the same with a new water color.

@HolgerJeromin
Copy link
Contributor

for reference
image

@Adamant36
Copy link
Contributor

@kocio-pl making scrub a tad darker seems to help if that's an option. I think its a tad to light.
scrub with stream alternate
Its just a mock up, but it looks better. Maybe it could be a little lighter if need be and still have the same effect.

@kocio-pl
Copy link
Collaborator

kocio-pl commented Sep 1, 2018

All the colors look different, did you use Chrome for a screenshot? It has known problems with color profile.

@Adamant36
Copy link
Contributor

I just did it in paint with the paint can. The only things I changed where were the scrub fill color and some of the plant icons. I'll have to test it in kosmtik though now that you mention it.

@Tomasz-W
Copy link

Tomasz-W commented Sep 3, 2018

@Adamant36

Here is a visualisation of different green shades used for certain plants in osm-carto.

zielen2

As you can see, scrub colour is quite not maching the rest. The fourth square in 50-200 cm column is an experimental scrub pattern on a #aedfa3 (orchard/ vineyard colour), Can you make some test renderings with streams with it?

@Tomasz-W
Copy link

Tomasz-W commented Sep 27, 2018

@Adamant36 Can you take care of this issue next? I'm planning a quite big proposal of coordinating non-landcover green shades (i.a. @park and @leisure-fill change), but of course it has to match another green shades to make complex system, so this issue and #3411 have to be done first.

Here are my Ps mock-ups with proposed above #AEDFA3. Visibility of a intermittent stream seems to be better. It's not perfect, but we propably won't find better shade without making mess with another green shades.

(click to view full size!)
scrub + intermittent stream before
1
scrub + intermittent stream with #AEDFA3
2
"camo" before
3
"camo" after
4

@Adamant36
Copy link
Contributor

@Tomasz-W, sure. I'll try to get to it in the next couple of days.

@Adamant36
Copy link
Contributor

Adamant36 commented Sep 30, 2018

Scrub before and after with intermittent stream. I'll do some more tests later. It seems like it needs to be a tad darker though maybe.
before
scrub before
after
scrub after

@Tomasz-W
Copy link

Tomasz-W commented Sep 30, 2018

@Adamant36 The idea is to just use "@orchard" colour for scrub in landcover code. If we start to play with this colour, we will have to test also orchard, vineyard and plant_nursery compared to grass, forest, etc. I think it's not worth it and IMO it would propably end as a waste of time. Examples above shows nice improvement with using orchard colour and I'm OK with it.

Please test also some places with orchard/ vineyard/ plant_nursery mix with scrub like here: https://www.openstreetmap.org//#map=18/50.05146/19.87933

@Adamant36
Copy link
Contributor

Adamant36 commented Nov 18, 2018

@Tomasz-W, is there a reason scrub couldn't be changed to more of a golden brown/greyish color since scrub land isn't really green most of the time anyway and there can be a lot of patches of dirt in between the woody plants? To me, a solid green seems to much like woods or that there is consistent ground cover when there isn't.

Id also suggest the same for grassland/meadow. More of golden brown/gray color. Especially if the color of parks/pitches/sports centers is going to be changed. We could create a scrub/grassland/meadow color category made up of gold/brown and free up green for parks/woods. Otherwise, I don't think any of them are doable because we don't have enough green shades, or that just adding a pattern to meadows would be adequate. Plus, it just looks more realistic.

What do you think about it @jeisenbe?

@Tomasz-W
Copy link

Tomasz-W commented Nov 18, 2018

@Adamant36 My assumpion is that we have definetely too much unnecessary green shades now, which makes map cluttered and not intuitive in some areas. Another problem is a threat of close similarity of some colours of different features.

I'm sure that meadows and grasslands (#3143) are rather green, so we should stay with planned 'grass + pattern' rendering for them. Heath (#780) has another colour, but very similar height (see the diagram #780 (comment)), so I would include it also in this rendering. As scrubs can be yellow/ brown as well as green, and they have different height, we can try some olive shade for them (maybe heath colour re-use after moving it to green?), but you have to remember about another similars colours in use (see https://wiki.openstreetmap.org/wiki/AreasTab) to avoid conflicts. I consider previously proposed solution (#2098 (comment)) as a good enough one, but it's always good to try other options if someone wants to.

@Adamant36
Copy link
Contributor

@Tomasz-W, I think grassland is only green for the two or three months a year that it gets rain in most places. Then its more of a high grass, dry golden brown hay color. Which is one of the reasons its used for grassing land and why I proposed it be colored similar to farmland. Meadows on the other hand might be green more often, but they are also usually a mix of grass and wild flowers during "normal" times. So, make sense to not them color meadows as straight green.

@Tomasz-W
Copy link

@Adamant36 That's why I propose to use pattern, which would suggest that is not a typically green area without making conflicts with other colours, see #3143 (comment)

@Adamant36
Copy link
Contributor

Adamant36 commented Nov 18, 2018

Id also add that through a lot of research I have done, at least in america there is a major amount of miss tagging when it comes to grassland/meadows/grass/heath/scrub. Heath seems to be a mainly European thing, although there is a lot of things here miss tagged as it that should probably be scrub (or maybe visa versa?). Whereas, know one seems to know the difference between a meadow, grassland, or grass. While I think the issue is partly due to them being landuse instead of landcover tags, the very similar similar colors also probably don't help (1). Giving grassland a pattern fill would help slightly, but I think giving them drastically, more realistic, colors would also.

  1. People in America also seem to miss tag whole parks as playgrounds a lot to. That could be further evidence for the fact that coloring different things very similar leads to tagging problems or it could just allude to the fact that Americans as a group are extremely bad at the subtleties involved in things like this. Who knows.

P.s. The pattern might work for meadows, but I don't think it does for grassland. As grassland isn't green most of the time in most places and its not a "patterned" looking thing in sky shots either. Also, it doesn't help free up green colors or deal with the issues mentioned above this paragraph of miss tagging. Which I think making grassland part of farmland (most of the time its managed for hay) would.

If nothing else, at least it would free up a potential better green color for scrub. There's no reason we can't change the color of grassland and also still have a pattern on it either. They aren't mutually exclusive.

@jeisenbe
Copy link
Collaborator

jeisenbe commented Nov 18, 2018 via email

@jeisenbe
Copy link
Collaborator

jeisenbe commented Nov 18, 2018 via email

@Adamant36
Copy link
Contributor

just like we render all developed land the same at low
zoom
There's actually an issue about that with quarries being rendered the same as residential areas when zoomed out #2905. So, that type of unification isn't always necessarily a good thing. Also, what about with more zoomed in levels? I agree with you that there should be a gradation, but I'm not sure how that would work with leisure stuff also being green/darker green.

There's a big need to free up green stuff in general. You might be correct about grassland. It makes sense. I've looked around California and done searchers on Google images for references. I still think though that making it more hay color or similar to farmland would be a good idea. Even if it might not as realistic. Especially since farmland is being rendered at low zooms levels now and it would be cool if it was chunked in with it on that since its a part of farming infrastructure.

Scrub, I don't know. It could be colored something else, but I guess green would work also. @Tomasz-W, can you come up with a color block thing with the plant colors you already did in the comments above and add in the different leisure colors, along with your new proposed park color? We might need a natural green color meta ticket or a natural/leisure green color meta ticket. I'm not sure we can change them in isolation of each other.

@jeisenbe, say green is over used and we need to free it up some. What are you willing to sacrifice if not scrub or grassland?

@jeisenbe
Copy link
Collaborator

jeisenbe commented Nov 18, 2018 via email

@Tomasz-W
Copy link

We might need a natural green color meta ticket or a natural/leisure green color meta ticket. I'm not sure we can change them in isolation of each other.

@Adamant36 They are listed at the end of #3298 (comment). I've opened #3517 because they are connected to each other by "lighten/ darken" and they all have simillar functions. I think that in physical green areas case meta-ticket won't help because anyway we have to test different rendering options for each one, one by one.

If you or @jeisenbe or anybody else wants to work on this, at this stage I think that the best way to solve this mess would be to simply put test renderings of landcover-mixed areas in certain tickets. We would have a comparsion to another green areas then, and if it will be needed, we always can reference another tickets.

@Adamant36
Copy link
Contributor

@jragusa, at least "harmonizing all natural landcovers" would help make intermittent water paths on scrubs more visible. At least in theory. Also, its extremely hard to change one landuse color in isolation without considering the colors of other ones along with it.

@jragusa
Copy link
Contributor

jragusa commented Nov 19, 2018

I agree but in this case it would be better to see intermittent waters on the modified colour to avoid any regressions ;)

@Adamant36
Copy link
Contributor

Adamant36 commented Nov 19, 2018

@jragusa, first picture in #2098 (comment). I'm sure there will be more.

@Tomasz-W
Copy link

Tomasz-W commented Nov 19, 2018

@jeisenbe:

  • at this stage I wouldn't touch grass colour (it's not necessary and we have much other things to do, but we can try with moving similar campsite colour for it later, colour proposed by you is too yellowish imo)
  • using the same colour for heath and scrub (with different patterns) is ok for me.
  • I think that golf colour wouldn't fit any natural landcover, but we can consider moving it for pitches, as you wanted to change their colour

Can you try to make a PR for heath and scrub based on #d1e0b4 with different patterns?

@jeisenbe
Copy link
Collaborator

Before we go to a PR, we should consider if it is possible to still have a different background color for heath and scrub.

Back in 2015 there was a PR that proposed to unify the scrub and heath background color, and add a new pattern for heath, which was rejected: #1733

Also see the discussion about heath starting here: #780 (comment)

@imagico disagreed with combining heath and scrub. But he suggested that we might try having a continue of colors from heath through scrub to forest/wood, eg getting darker and greener as we go along, but grass / meadow could be a different sort of color because the type of vegetation and landuse are quite different.

However, I do like the pattern shown for heath on the PR #1733 by @matkoniecz; something like that could work, or perhaps something a little more rounded on top, but with a different background color than that used from scrub.

@jeisenbe
Copy link
Collaborator

Here's a new comparison of the current vegetation and green colors:
http://davidjohnstone.net/pages/lch-lab-colour-gradient-picker#eff0d5,cdebb0,d6d99f,b5e3b5,add19e,aedea3,c8facc,aae0cb,b5e3b5,def6c0

The list order is: farmland, grass/meadow, heath, scrub, forest/wood, orchard/vineyard, park, pitch, golf, campsite.

Here is just the scrub change, with heath, grass, and the rest left the same:
http://davidjohnstone.net/pages/lch-lab-colour-gradient-picker#eff0d5,cfecb1,d5d89f,d1e0b4,add19e,aedea3,c8facc,aae0cb,b6e3b5,def6c0

Rendering without the patterns, eg what we would see at lower zoom levels, only other change is d1e0b4 for scrub:
scrub-d1e0b4-test1
scrub-d1e0b4-test2

The new d1e0b4 scrub color has a delta (difference) of 8.4 with the current heath color, and 11.3 with the current wood/forest color. This looks pretty good; its different enough from heath to be distinct, but still somewhat similar. It has a difference of 9.0 with the current campsite color, which is a little close but ok, and 12.5 with farmland, according to the linked color picker.

So we could keep the current heath color and grass color for now, and just change scrub, though it would be nice to have a slightly less brown color for heath.

Re: test renderings with new patterns; I have not yet installed the pattern generator yet.

@Adamant36
Copy link
Contributor

@jeisenbe, normally I wouldn't mind waiting to do a PR until we get the other things besides scrub dealt with, but this issue was specific to scrub and the new release is coming out on the 23rd. So it would be good if scrub was included in it in order to see it in the wild and also to honor the intent of the original issue creator. Otherwise, we will have to wait another month until the next version comes out for it to be widely released. There's no reason we can't do that and then work on the other things in a meta ticket or here in the mean time.

@jeisenbe
Copy link
Collaborator

If we want to change heath, we may need to change campsite and grass as well.

I tried making some extensive changes, including the leisure and campsite and golf unification as @Tomasz-W suggested.

Here I have used the current golf color for the new park color and used the new leisure (new park lightened 5%) for campsites and other areas. This frees up the current campsite color to use for grass (it is quite similar to the current grass color), and allows the use of a lighter, greener heath color. I've also had to adjust the scrub color to match better. It's quite difficult to get it right.

http://davidjohnstone.net/pages/lch-lab-colour-gradient-picker#eef0d5,def6c0,dde8ad,c9deb3,add19e,aedea3,c8facc,aae0cb,b6e3b5,def6c0

Farmland eef0d5, Grass/Meadow def6c0, heath dde8ad, scrub d1e0b4, forest/wood add19e, orchard/vineyard aedfa3, *park b5e3b5, pitch b8dabd, leisure lighten park 5%; also campsite and golf = leisure now. I've had to adjust the outlines and text color, still WIP:

Text area 1 at z15:
greenchanges-lessline-test1
z16:
green-lessline-test1-z16

Test 2 at z15:
greenchanges-lessline-test2
z16:
green-lessline-test2-z16

Mountain area test:
z13-mountainstest-newgreens
Next to bog and marsh areas (which match grass and heath color)
bog and marsh test

@jeisenbe
Copy link
Collaborator

Comparisons between current master, scrub changed to d1e0b4 (without pattern just for testing) or extensive color changes (without patterns; would need to be added for heath and scrub)

Northern Ireland, NE of Belfast:
(notice the patch of scrub next to bog in the upper left corner, also parks, pitches and golf areas)
http://www.openstreetmap.org/#map=14/54.6300/-5.8296
Current z14
ne-belfast-before
Scrub d1e0b4
ne-belfast-scrubonly
Extensive changes to grass, heath, scrub, park, leisure
greens-northeast-belfast-after-z13

Dunmurry, Northern Ireland:
Park, stadium, scrubland, orchards etc
http://www.openstreetmap.org/#map=15/54.5422/-5.9806
Current z15
dunmurry-z15-before
Scrub d1e0b4 only
greens-dunmurry-z15-scrubonly
Extensive changes to greens as above
greens-dunmurry-z15-after

@Adamant36
Copy link
Contributor

@jeisenbe, can you just do some test renderings with only the scrub color changed that involves intermittent streams please? Its probably hard for people to keep up with the 6 different color changes at once and not get off the topic of how intermittent streams look in the process. Plus, its what the issue is about and its kind of pointless to change the scrub color if it doesn't look good with streams still. Also, the other stuff can be changed later in a meta issue if need be.

Its better to tackle one thing at a time if its possible. We have already established that the new scrub color looks good, not that it works good with intermittent streams though. I think you should just focus on that, give other people a chance to provide feedback on it, and then move on to the other stuff once we know its good.

@jeisenbe
Copy link
Collaborator

I've already shown this image with intermittent streams on scrub with the new color suggested by ATomasz-W: d1e0b4

And this image, with the old png still used for the scrub pattern (this needs to be updated):

Here are some real-world examples from Hawaii, with a golf course next door, by chance:
http://www.openstreetmap.org/#map=15/20.7490/-156.8960
(Should I admit that I just realized that the current scrub background color is the same as the current golf course fill color, b5e3b5? This suggests it would not be perfect as a park color; intermittent stream are rare in developed parks, but not unheard-of)
z15 current
manele-before-z15
z16 current
manele-z16-before
z17 current
manele-z17-before

With d1e0b4 for scrub (no pattern yet)
z15 scrub d1e0b4
manele-z15-scrubonly
z16 d1e0b4
manale-z16-scrubonly
z17 d1e0b4
manele-z17-scrubonly

@Tomasz-W
Copy link

@jeisenbe Please add some test renderings with #d1e0b4 with scrub pattern.

@jeisenbe
Copy link
Collaborator

jeisenbe commented Nov 21, 2018

I've made a new scrub pattern file with #a3b886 for the pattern to match the #d1e0b4 background best:

#a3b886
scrub-screenshot-a3b886

It looks like this pattern is a little denser and more random than the current scrub pattern. I can try making some others if necessary.

Hawaii:
z15
manele-z15-scrub1

z16
manale-z16-scrub1

You can see the file (scrub2.png) on https://github.com/jeisenbe/openstreetmap-carto/blob/scrub-only/symbols/scrub2.png

@jeisenbe
Copy link
Collaborator

@jeisenbe
Copy link
Collaborator

Maybe the pattern needs to be lighter; the forest/wood pattern is much closer to the background color, and it looks like the current scrub pattern is also closer to the background color than what I've just shown.

Scrub pattern 4: b5c498 pattern on d1e0b4 background
scrub3-b5c498-d1e0b4-screenshot

Test area with scrub 4:
scrub4-b5c498-d1e0b4-test

Hawaii z15 with scrub 4:
scrub4-hawaii-z16

PNG for testing:
scrub3-b5c498-d1e0b4

@jeisenbe
Copy link
Collaborator

We've been discussing the scrub color change on another issue, #780 "Improve Natural=Heath Color", and there is a suggestion to use a 3% darker color than the #d1e0b4 that has been shown above, eg. #c8d7ab.

So here are some tests with scrub background color #c8d7ab and a new scrub pattern in #b0be93 (jsdotgenerator was used with distance 24, relaxed 3 times with relaxation radius 32 metric 2. Rendered without pixel alignment to get a more naturally fuzzy look.

twmbarlwm-scrub-c8d7ab

caldiera-scrub-c8d7ab

ribeira-scrub-c8d7ab

With intermittent streams:
estrada-scrub-c8d7ab

test-scrub-c8d7ab

This color is a little darker (3%) than the previous, but the streams are still visible

@Adamant36
Copy link
Contributor

@jeisenbe, I like the new pattern. Whats the advantage over the original though? Maybe the same thing could be done for the forest pattern also at some point.

@jeisenbe
Copy link
Collaborator

jeisenbe commented Nov 23, 2018 via email

@jeisenbe
Copy link
Collaborator

jeisenbe commented Nov 24, 2018

This was the command used to generate the new pattern with jsdotpattern:

http://www.imagico.de/map/jsdotpattern.php#x,512,jdp97762;g5,45,64,64;rx,250,2,64,64;rx,250,2,64,64;rx,250,2,64,64;rd,1,0,0,scrub2,1,5,12,0,jdp3379,8ece8f,b5e3b5;

Compare to the wood/forest pattern:

http://www.imagico.de/map/jsdotpattern.php#x,256,jdp6894;g,30,32,32;s,jdp33742;s,jdp81637;rx,250,2,32,32;s,jdp28824;s,jdp59702;s,jdp91550;s,jdp27774;rx,250,2,64,64;rd,1,0,0,tree%20pair,1,5,5,0,jdp52898,6b8d5e,add19e;

Actually it looks like the old pattern was only semi-random; this new pattern is more random, and perhaps the icons are a few pixels closer together on average.

I'm thinking of submitting a PR for just this change, and leaving some more time to work on heath and grass later, because this change looks good and resolves this issue.

@Adamant36
Copy link
Contributor

Adamant36 commented Nov 24, 2018

I was thinking the new scrub pattern was random the old forest pattern. It might be worth randomizing the forest pattern more at some point.

As far as the PR goes, you should do it for just this change. Heath is a different issue anyway. Your going to do it with scrub as %3 darker then heath (at #d1e0b4) right?

@jeisenbe
Copy link
Collaborator

Area in Hawaii with streams, intermittent streams on scrub near golf course:

Scrub background color #c8d7ab with scrub pattern in #b0be93:
z15
new-scrub-z15-hawaii
z16
new-scrub-hawaii-z16

@Tomasz-W
Copy link

As new scrub colour is based on a new heath colour, it's obvious for me that we should finish heath first, then a change for scrub would be easier (propably something like darken(@heath, 3%)

@jeisenbe
Copy link
Collaborator

jeisenbe commented Nov 24, 2018 via email

@Adamant36
Copy link
Contributor

I'm not exactly sure what the pros and cons of using a variable to represent a color is, but I know they are used a lot in the landcover.mss file. So I don't why it would be an issue. Unless its particular to this case for some reason. That being said, if you can figure out the exact hex value for darken(@heath, 3%) then that's fine.

Also, it then has the disadvantage of being harder to tweak in small amounts if need be. While most people would know what the difference between darken(@heath, 3%) and darken(@heath, 2.5) might be its much harder and takes more work to interpret and tweak a hex value with by using the lch whatever thing or a color slider in a photo editing app (where you might not be able to do it with that fine of detail). Which is probably why the darken thing is used.

As I said though, as long its the exact same color, it doesn't really matter how its represented in there. Although, thinking about it, I remember a couple of issues where the hex value was used instead of the darken variable and the maintainers said to use the darken variable instead because it shows how the color was obtained. So....I think there's definitely more of a slant toward using the variable when it can be. Plus, there's plenty in the contributing file about how code should be consistent and use already existing standards. Which the variable thing clearly is. Even if you might personally prefer it the other way.

@Tomasz-W
Copy link

#d0e3b6+ 10% vertical dash pattern for heath + #c8d6ab for scrub

scrub1
scrub2
scrub3
scrub4
scrub5

@jeisenbe
Copy link
Collaborator

jeisenbe commented Nov 25, 2018 via email

@Adamant36
Copy link
Contributor

Sounds good to me.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

10 participants