[OSM-talk] Macromapping problems

Daniel Koć daniel at xn--ko-wla.pl
Mon Aug 21 23:42:35 UTC 2017

I guess most of you understand the problems of so called "micromapping" 
(very detailed geo data mapping). But when I started to look at the 
opposite end (macro level), it looks even worse.

Low zoom rendering on default map style lacks lot of important things 
(see https://github.com/gravitystorm/openstreetmap-carto/issues/2688 ). 
We have basically:

- land/water distinction (but some big lakes are missing for example)

- ice areas

- country (and district) borders

- capitals and big cities

- big roads

- nature reserves

- railway and ferry lines

- big military areas

What we don't have there:

- big landuses, like forests, farmlands, sands and grasslands

- ocean and sea names

- continent names

- big deserts

- big mountain chains

- big rivers

Of course we can have some of them, but while landuses can be added 
quite easily, it's different with the rest. Rivers have their own 
problems, which we've just started to solve with crafting some 
classification, but all the big area objects share at least one common 
issue: lack of precise borders.

That's probably why we have Asia mapped as a node in the middle of nowhere:


the same for Andes:


and not a trace of Sahara _desert_ in Nominatim:


You get the picture probably. We started as a city-scale project with 
routing as the most important goal, but now we're "The map" for every 
GIS data-related activity. I think this is the time to look how should 
we map the biggest objects in the world? Our attitude is high accuracy 
and ground truth, which are not easy to apply there.

There are also other problems, like no special tools for macro objects 
probably - for example Overpass service is choking with some data 
requests even at country level. Another problem is importing data for 
planet/continents, because there's too much of them already. There is a 
tool to filter some of them ( https://github.com/gmgeo/osm-carto-lowzoom 
), but it's not easy to customize - you need some programming skills.

How do you think we could improve macro level mapping, rendering and 
data analysis?

"Like a halo in reverse" [M. Gore]

More information about the talk mailing list