My worry is not that I am unable to map based on ‘high resolution’ imagery but that the available imagery is so old that in areas where there has been a great deal of development, including totally new road layouts that have been mapped using the more up to date Maxar imagery these new developments will now be removed by remote mappers. Africa is already plagued by poor quality mapping anyway. This makes it worse. I just go off and map rivers and other natural features that do not change so readily. Although new mining projects in remote areas do not appear in the Bing imagery. By my estimation some of the Bing imagery is over 10 years out of date in some places. Perhaps we just tag it ‘Here be dragons.’
3 Likes
Just to put some numbers against this, I had a look at changesets mentioning “Bing” or “Maxar” in an “imagery_used”** tag.
Worldwide in March:
changesets=> select count(*) from osm_changeset where created_at > '2023-03-01 00:00:00' and created_at < '2023-04-01 00:00:00' and tags -> 'imagery_used' like '%Bing%';
count
--------
269446
(1 row)
changesets=> select count(*) from osm_changeset where created_at > '2023-03-01 00:00:00' and created_at < '2023-04-01 00:00:00' and tags -> 'imagery_used' like '%Maxar%';
count
--------
131998
(1 row)
In Africa*** in March:
SELECT count(id)
FROM osm_changeset c, (SELECT ST_SetSRID(ST_MakeEnvelope(-30.0,-48.0,55.0,34.0),4326) AS geom) s
WHERE ST_CoveredBy(c.geom, s.geom) and created_at > '2023-03-01 00:00:00' and created_at < '2023-04-01 00:00:00' and tags -> 'imagery_used' LIKE '%Bing%';
count
-------
17489
(1 row)
SELECT count(id)
FROM osm_changeset c, (SELECT ST_SetSRID(ST_MakeEnvelope(-30.0,-48.0,55.0,34.0),4326) AS geom) s
WHERE ST_CoveredBy(c.geom, s.geom) and created_at > '2023-03-01 00:00:00' and created_at < '2023-04-01 00:00:00' and tags -> 'imagery_used' LIKE '%Maxar%';
count
-------
41871
(1 row)
** obviously this’ll select mostly iD rather than other editors, and “amount edited” isn’t proportional to “number of changesets”.
*** actually quite a big square box that includes Africa
3 Likes
Total changesets, Africa:
FROM osm_changeset c, (SELECT ST_SetSRID(ST_MakeEnvelope(-30.0,-48.0,55.0,34.0),4326) AS geom) s
WHERE ST_CoveredBy(c.geom, s.geom) and created_at > '2023-03-01 00:00:00' and created_at < '2023-04-01 00:00:00';
count
--------
119761
(1 row)
and worldwide:
WHERE created_at > '2023-03-01 00:00:00' and created_at < '2023-04-01 00:00:00'; count
---------
1214902
(1 row)
As a caveat I haven’t actually sanity checked these numbers against known numbers derived from elsewhere.
2 Likes
I looked at usage from both iD and JOSM and by far the countries with the most Maxar usage have been
- India,
- Indonesia,
- Turkey,
- Brazil,
- El Salvador,
- Guatemala, and
- Tanzania.
I also checked, and 35% of Maxar usage is from changesets that mention hotosm in the comment.
3 Likes
Just one example of how out of date Bing imagery is. This is an area near the town of Mwinilunga in Zambia that is fairly remote. The development that was originally mapped as part of the #hotosm-project-13416 #zambia-census in 2022 used Maxar imagery (my thoughts on the effectiveness of this project are for another discussion). A HOT validator with any sense would not remove this but it would not be so obvious where there were buildings mapped in a remote village where a number of new huts and houses were added based on the Maxar imagery that do not appear in Bing imagery. The basis of this project is to map remote human settlements in Zambia, I assume to assist with the census that took place last year.
Had to add comparison as single image as I can not embed more than one image and can only reply 3 times.
2 Likes
Building on top of @SomeoneElse and @pnorman’s great work, I have looked at relative percentages instead of absolute numbers over the last year. For example, 19613 changesets in Germany and 15404 in Afghanistan used Maxar, but in Afghanistan, that’s 88% of changesets; in Germany, about 1%.
If you’re mapping in Germany, you probably haven’t noticed that Maxar is gone. If you’re mapping in Afghanistan, well, you’re probably not mapping right now. Like @ftcat has pointed out, there’s a very good reason why 83% of changesets in Zambia over the past year have
used Maxar.
Many of the most affected countries are in the global south, where OpenStreetMap is often the most usable map, sometimes the only usable map. People rely on it for their jobs and daily lives.
The OpenStreetMap Foundation is very much aware of the issue, and is continuing to look for solutions. I have reached out to Maxar, and I will update this thread if/when we hear back.
| Country | Maxar count | Total count | Maxar percentage |
|---|---|---|---|
| Guatemala | 100412 | 107552 | 93,36% |
| El Salvador | 102280 | 114432 | 89,38% |
| Afghanistan | 15404 | 17485 | 88,10% |
| Somalia | 14745 | 17620 | 83,68% |
| Zambia | 60130 | 72095 | 83,40% |
| Timor-Leste | 10364 | 12913 | 80,26% |
| Congo (DRC/Kinshasa) | 45216 | 57122 | 79,16% |
| Qatar | 12379 | 15759 | 78,55% |
| Libya | 18980 | 24256 | 78,25% |
| United Arab Emirates | 40728 | 52429 | 77,68% |
| Syria | 49353 | 64668 | 76,32% |
| Mali | 13417 | 17898 | 74,96% |
| Togo | 10533 | 14058 | 74,93% |
| Madagascar | 24456 | 33235 | 73,59% |
| Ethiopia | 16406 | 22328 | 73,48% |
| Tanzania | 86890 | 118618 | 73,25% |
| Sudan | 14346 | 19585 | 73,25% |
| Malawi | 40710 | 56123 | 72,54% |
| Bangladesh | 60622 | 83580 | 72,53% |
| Pakistan | 17161 | 26172 | 65,57% |
| Turkey | 131282 | 204181 | 64,30% |
| Honduras | 32385 | 51313 | 63,11% |
| Egypt | 16848 | 28513 | 59,09% |
| India | 253789 | 431936 | 58,76% |
| Peru | 37037 | 68006 | 54,46% |
| Kenya | 30045 | 59749 | 50,29% |
| China | 58226 | 116383 | 50,03% |
| Nepal | 31692 | 63876 | 49,61% |
| Malaysia | 27224 | 57220 | 47,58% |
| Indonesia | 138214 | 293674 | 47,06% |
| Nigeria | 55355 | 118617 | 46,67% |
| Saudi Arabia | 17270 | 39509 | 43,71% |
| Viet Nam | 20203 | 49958 | 40,44% |
| Iraq | 15819 | 39679 | 39,87% |
| Philippines | 46432 | 117337 | 39,57% |
| Mexico | 33012 | 85477 | 38,62% |
| Kazakhstan | 18037 | 50624 | 35,63% |
| Uganda | 19534 | 55471 | 35,21% |
| Romania | 26477 | 76604 | 34,56% |
| Brazil | 121062 | 351765 | 34,42% |
| Thailand | 17879 | 53934 | 33,15% |
| Ecuador | 13919 | 43575 | 31,94% |
| Chile | 11902 | 38158 | 31,19% |
| Argentina | 20264 | 66341 | 30,55% |
| Belarus | 20925 | 70071 | 29,86% |
| Colombia | 23876 | 86638 | 27,56% |
| Ukraine | 33664 | 141756 | 23,75% |
| South Africa | 24262 | 104901 | 23,13% |
| Sweden | 33762 | 217931 | 15,49% |
| Russia | 71243 | 465210 | 15,31% |
| Hungary | 11284 | 91060 | 12,39% |
| Australia | 16134 | 183779 | 8,78% |
| Iran | 10899 | 127491 | 8,55% |
| Canada | 14223 | 227851 | 6,24% |
| United States of America | 69600 | 1704614 | 4,08% |
| United Kingdom | 19687 | 558880 | 3,52% |
| Italy | 15283 | 477605 | 3,20% |
| Germany | 19613 | 1941128 | 1,01% |
(Countries with less than 10k Maxar changesets aren’t shown. The centroid of a changeset isn’t always in the country where the changes were made. Changes made in more than one country are only counted once.)
26 Likes
As a mapper from South Korea I also feel the unavailability of Maxar imagery immensely. Bing imagery is hopelessly outdated (at least 4 years old), and while Esri works in the meantime (at least 1 year old), it still doesn’t show new developments and is slightly lower resolution than Maxar. In a country where the government still strictly controls export of geographic data, and most of the existing POIs referring to Yahoo (stopped service in 2012), up to date imagery is the most valuable thing we have.
12 Likes
It would be interesting to compare Profiles of Changesets vs Objects edited
to assure tjat the Changesets statistic dont give more weight to projects / countries with contributors having very small edits per changeset. The number of objects edited are available from the changesets.
“ChangesetMD” (which I suspect we’re all using) does have the “number of changes” that you can see from the XML for a changeset on the website. However, you’d need to think carefully about how to use that to get anything meaningful. If you can suggest some SQL or a calculation someone can probably do it.
In changsetMD, num_changes refers to objects editted ( nodes + ways + relations) but do reference how many tags edited. Even if this is not a perfect measure of the intensity of mapping, it is significantly better then the simple changeset count. If we want both statistics, we simply use :
select created_a, count(*) as changesets, sum(num_changes) as obj_edited, geom from osm_changeset ...
One option might be checking if local government has Orthophoto imagery and if licensing allows using in OSM, if not email them and work towards government opening data… All this is long and tiresome process… And even after that you need someone to host tiles.
In Slovenia all this data is open and refreshes 1/3 of country every year and local GIS company https://level2.si/ is willing host tiles for OSM.
In Timor-Leste the Bing (which had very good quality, maybe better than today, but could be near as 10 years old depending of the region) changed.
Data de captura:1 de jan. de 2012 - 12 de fev. de 2023
The photo is not even from the city area (which sometimes imagery tends to be more recent), but rural/remote area in Timor-Leste and is from 2023-02-12.
So, not sure about “how much outdated” Bing is expected to be, but even then, there’s ESRI, which often I found to be fairly recent (Maxar the iD does not reveal the imagery date, but I would not assume that just because Maxar, Mabox, etc don’t advertise date they are necessarily the most recent ones. Also, maybe coincidence, but in the places I do map, often Maxar had almost same imagery (but just don’t mention date) as ESRI or Bing).
A general statement about quality and actuality of different imagery providers simply is not possible.
All global imagery is a patchwork of countless small images stitched together from various raw images in order to avoid cloud cover, reduce license fees etc. etc.
Very often imagery from different providers is derived from the same raw source. On the other hand image capture may vary more than ten years.
This may vary drastically not only between countries but sometimes within kilometers. In some cases it may even vary within zoom levels.
In addition the “ranking” of imagery changes constantly as new imagery patches are made available by providers.
Maxar is no exception from it. There were regions in middle asia where the imagery was roughly ten years ahead of all others estmated from the amount of new roads built in the meantime. New built houses and replacement of wood by farmland are an additional hint for actuality.
Actuality of the road net on maps is a very important feature in difficult accessible regions. I just want to mention the Nepal earthquake.
In this sense the unavailability of Maxar imagery is a considerable loss in several regions whereas is does not matter at all in others.
8 Likes
Actually, Maxar -is- an exception. Essentially they are only constrained by their own economic considerations, this is very different than any other provider of an imagery mosaic that we have access to.
1 Like
On that tangent: The OpenStreetMap Foundation can and does happily host imagery too, if that’s needed. chef/cookbooks/imagery/recipes at master · openstreetmap/chef · GitHub is the current list of chef configs. Send me a message if you’d like to host something there.
5 Likes
I have summarized the whole story here:
-
March 15, 2017, Facebook’s AI Assisted Tracing was announced on the OpenStreetMap mailing list. It used Digital Globe’s satellite imagery of Thailand to trace roads with AI and manually verify them using human input.
-
Several members of the OSM community demanded that Facebook should release the Digital Globe satellite imagery to facilitate the re-verification of all submitted contributions.
-
Digital Globe agreed to release the imagery only to registered OSM editors, through funding from “a group of organizations”.
-
May 9, 2017, DG-Premium and DG-Standard layers were released on OSM. Their EULA specifically stated that the imagery could not be downloaded for any other purpose than OSM contribution.
-
June 28, 2019, DigitalGlobe was renamed to Maxar, and the DG-Standard layer was renamed to Maxar Standard, while the DG-Premium layer was renamed to Maxar Premium.
-
December 19, 2019, Maxar announced that it had seen a sharp increase in usage through automated requests coming from a few areas of concern. As a result, it would soon suspend its imagery service.
-
2023 : It has been stated that the reason for the current outage is that a small number of people are using the OpenStreetMap API key to steal imagery for non-OpenStreetMap usage.
See details here :
15 Likes
To present a complete yearly profile, I wrote my own analysis grouping this same Changesets data. I have a slighlty different perspective with 2022 changesets (the data I have on hand) by continent / Sub-continent (a least 70% of the changeset BBOX must be inside the Continent geometry). I also show other major imagery providers. I did analyse also with the variable num_changes (no. of objects edited). See PierZen's Diary | OpenStreetMap Imagery sources by Continent / Sub-Continent, 2022 | OpenStreetMap
2 Likes
Has the option for Maxar imagery now disappeared out of iD everywhere, or only in Australia?
In Greece too, I assume everywhere, since no new keys yet.
2 Likes
@tyr_asd removed the Maxar option from iD, globally, but it’ll be restored once the layer becomes available again.
6 Likes