This blog post is a summary of my talk at the FOSSGIS & OpenStreetMap conference 2017 (german slides). I guess some of the content might be feasible for a research article, however, here we go:

Vandalism is (still) an omnipresent issue for any kind of open data project. Over the past few years the OpenStreetMap (OSM) project data has been implemented in a number of applications. In my opinion, this is one of the most important reasons why we have to bring our quality assurance to the next level. Do we really have a vandalism issue after all? Yes, we do. But first we should take a closer look at the different vandalism types.

It is important to distinguish between different vandalism types. Not each and every unusual map edit should be considered as vandalism. Based on the OSM wiki page, I created the following breakdown. Generally speaking, vandalism can occur intentionally and unintentionally. Therefore we should distinguish between vandalism and bad-map-editing-behavior. Oftentimes new contributors make mistakes which are not vandalism because they do not have the expert mapper knowledge. In my opinion, only intentional map edits such as mass-deletions or “graffiti” are real cases of vandalism.

To get an impression of the state of vandalism in the OSM project, I conducted a case study for a four week timeframe (between January 5th and February 12th, 2017). During my study I analyzed OSM edits, which mostly deleted objects from new contributors who created fictitious data or changesets for the Pokemon game. If you did not hear or read about OSM’s Pokemon phenomena, you can read more about it here. The OSM wiki page for quality assurance lists some tools that can be used for vandalism detection. However, for this study I applied my own developed OSM suspicious webpage and the quite useful augmented OSM change viewer (Achavi). Furthermore, a webpage that lists the newest OSM contributors may also be of interest to you.

So what can you do when you find a strange map edit that could be a vandalism case? The OSM help page contains an answer for that. First of all: Keep calm! Use changeset comments and try to ask in a friendly manner for the suspicious mapping reasons.

Results of the study: Overall I commented 283 Changesets in the aforementioned timeframe of four weeks. Unfortunately I did not count the number of analyzed changesets, but I assume that it should be around 1,200 (+- 200). The following chart shows the commented changesets per day. The weekends tend to have a larger number of commented/discussed changesets.

As mentioned in the introduction of the vandalism types, we should distinguish between different vandalism types. The following image shows the numbers for each category. In my prototype study, 45% of the commented changesets were vandalism related and 24% have already been reverted which was not documented in the discussion of the changeset. Sometimes I also found imported test- and fictitious data, which the initial contributor of the changeset didn’t revert. It should be clarified to everyone that the live-database should never be used for testing purposes. Interested developers can use the test API and a test database (see sandbox for editing).

Responses and spatial distribution: Overall I received 70 responses for the discussed changesets, sadly only 20 from the owner/contributor of the changeset. But, more or less every response was in a friendly manner. Most often the contributors wrote “thank you” or “I didn’t know that my changes are going to be saved in the live database”. Furthermore, if I received a response, it was within 24 hours.

The following map contains some clustered markers. Each one highlights areas where the discussed changesets are located. As you can see on the map, the commented changesets are spread almost all over the world. In some areas they tend to correlate with the number of active OSM’ers. However, here is some additional information about three selected areas: 1: USA – Several cases of Pokemon Go related and fictitious map edits. 2: Japan/China – Some mass deletions and 3: South Africa – Oftentimes new MissingMaps or HOT contributors tend to delete and redraw more or less the same objects such as buildings. I guess it was not explained well enough to these editors that this destroys the object history? However, the article about “Good practice” in the OSM wiki is quite useful in this case.

Conclusion: The study reveals that there is an ongoing issue with vandalism in OSM’s map data. I think we do need to simplify the tools for detecting vandalism. In particular we should omit work where several users review identical suspicious map edits. Maybe the best possible solution should be a tool which is integrated directly in the OSM.org infrastructure. However, my presentation also contained some statistics and charts about the OSM changeset discussions feature. This will be the content of a separate blog post in following weeks. Also, the prototype introduced at the end of my talk will (hopefully) be presented in the next few months.

Thanks to maɪˈæmɪ Dennis.

Over the past few years I have written several blog posts about the (non-) activity of newly registered OpenStreetMap (OSM) members (2015, 2014, 2013). Similarly to the previous posts, the following image shows the gap between the number of registered and the number of active OSM members. Although the project still shows millions of new registrations, “only” several hundred thousand of these registrants actually edited at least one object. Simon showed similar results in his yearly changeset studies.

The following image shows, that the project still has some loyal contributors. More specifically, it shows the increase in monthly active members over the past few years and their consistent data contributions based on the first and latest changeset:

However, this time I would like to combine the current study with some additional research. I tried to identify three different OSM contributor groups, based on the hashtag in a contributor’s comment or the utilized editor, for the following analysis:

1. Contributors of the MissingMaps-Project: A contributors of the project usually use #missingmaps in their changeset.
2. Contributors that utilized the Maps.Me app: The ‘created_by’-tag contains ‘MAPS.ME’.
3. All other ‘regular’ contributors of the OSM project, who don’t have any #missingmaps in their changesets and neither used the maps.me editor.

In the past 12 months, almost 1.53 million members registered to the OSM project. So far, only 12% (181k) ever created at least one map edit: Almost 12,000 members created at least one changeset with the #missingmaps hashtag. Over 70,000 used the maps.me editor and 99,000 mapped without #missingmaps and the maps.me editor. The following diagram shows the number of new OSM contributors per month for the three aforementioned groups.

The release of the maps.me app (more specifically the OSM editor functionality) clearly has an impact on the monthly number of new mappers. Time for a more detailed analysis about the contributions and mapping times: The majority of the members of the groups don’t show more than two mapping days (What is a mapping day, you ask? Well, my definition would be: A mapping day is day, where a contributor created at least one changeset). Only around 6% of the newly active members are contributing for more than 7 days.

Some members of the #missingmaps group also contributed some changesets without the hashtag. But many of those members (70%) only contributed #missingmaps changesets. Furthermore, 95% of this adjusted group doesn’t map for more than two days. Anyway, despite identifying three different contributor groups, the results are looking somewhat similar. Let’s have a look at the number of map changes. The relative comparison shows that the smaller #missingmaps group produces a large number of edits. The maps.me group only generates small numbers of map changes to the project’s database.

Lastly, I conducted an analysis for three selected tag-keys: building, highway and name. The comparison shows that the #missingmaps group generates a larger number of building and highway features. In contrast “regular” OSM’ers and maps.me users contributed more primary keys such as the name- or amenity-tag.

I think the diagrams in this blog post are quite interesting because they show that the #missingmaps mapathons can activate members that contribute many map objects. But they also indicate that the majority of these elements are traced from satellite imagery without primary attributes. In contrast the maps.me editor functionality proofed to be successful with its in-app integration and its easy usability, which resulted in a huge number of new contributors. In summary, I think it would be good to motivate contributors not only to participate in humanitarian mapathons but also to map their neighborhood in an attempt to stick to the project. Also, I guess it would be great if the maps.me editor would work on the next steps in providing easy mapping functionality for its users (of course with some sort of validation to reduce questionable edits).

Thanks to maɪˈæmɪ Dennis.

Back in 2010 & 2011 I conducted several studies to detect underrepresented regions a.k.a. “unmapped” places in OpenStreetMap (OSM). More than five years later, some people asked if I could rerun the analysis. Based on the latest OSM planet dump file and Taginfo, almost 1 million places have been tagged as villages. Furthermore, around 59 million streets have a residential, unclassified or service highway value. My algorithm to find unmapped places, works as follows:

1. Use every place node of the OSM dataset which has a village-tag (place=village).
2. Search in a radius of ca. 700 m for a street with one of the following highway-values: residential, unclassified or service.
3. If no street can be found, mark the place as “unmapped”!

My results for the entire OSM planet can be found under the following webpage.

Overall we have more than 440,000 unmapped places in OSM. As you can see in the picture above, most of the places are around Central Africa, Saudi Arabia or China. However, I hope that this analysis helps to complete some of the missing areas or to revise some incorrect map data. Some remarks about “false=positives” or why your village is marked as unmapped? Some possible reasons: Is the used tag for your place correct? Compare the wiki page for further information. Sometimes “hamlet” could be the correct tag value. Are the nearby highways tagged correctly? (OSM wiki)

Amount of unmapped places for each continent:

• Africa 119,084
• Asia 241,833
• Australia 212
• Europe 44,819
• North America 16,464
• Oceania 837
• South America 15,576

Technical Stuff: The OSM data for the analysis is prepared by a custom OSM PBF reader. The webpage, which shows the results, is based on Leaflet 1.0.0-rc1 and the really fast PruneCluster plugin.

*Update*: You’ll find the date of the latest data update in the header -> “(Date: Apr. 9th, 2018)”

Thanks to maɪˈæmɪ Dennis.

Almost one year has passed again. This means it’s time for the fourth OpenStreetMap (OSM) member activity analysis. The previous editions are online here: 2014, 2013 and 2012. Simon Poole already posted some interesting stats about the past few years. You can find all his results on the OSM wiki page. However, similar to last year, I try to dig a little deeper in some aspects.

Overall the OSM project has officially more than 2.2 million registered members (Aug, 9th 2015). For several of my OSM related webpages I create a personal OSM contributor database, based on the official OSM API v0.6. Anyway, when using this API, the final table will show a list with more than 3 million individual OSM accounts (Aug, 9th 2015). I’m not sure what the cause for this gap of almost 1 million members between the official number and the member number extracted with the API could be. Maybe some of you have a possible explanation? However, I think many accounts are created by spammers or bots.

The following chart shows a trend similar to the one of previous years: The project attracts a large number of newly registered members, but the sum of contributors that actively work on the project is fairly small. As mentioned in earlier posts, this phenomenon is nothing special for an online community project and has been analyzed for previous years already.

Described in numbers (July 31st, 2015):

• Registered OSM Members (OSM API): 3,032,954
• Registered OSM Members (Official): 2,201,519
• Members who created 1 Changeset: 562,670
• Members who performed >= 10 Edits: 343,523
• Members who created >=10 Changesets: 137,591

Personally, I really like the following diagram: It shows the increase in monthly contributor numbers over the past few years and their consistencies in collecting OSM data based on the first and latest contributed changeset of an OSM member. It’s great to see that at least some experienced mappers are still contributing to the project after more than five years.

Some background information on how I created the stats: To retrieve the registration date of the members, I used the aforementioned OSM API. The other numbers are based on the OSM changeset dump, which is available for download here.

Next to the presented results above, you can find some daily updated statistics about the OSM project on OSMstats.

Thanks to maɪˈæmɪ Dennis.

OSMstats contains several statistics about the OpenStreetMap (OSM) project, such as daily-created objects, the amount of active contributors or detailed numbers for individual countries. One way to determine the sum of created or modified Node objects, is to use the minutely, hourly or daily OSM replication change files and counting the values for each country of the world. Sadly, this approach has some drawbacks. Firstly, the official files do not contain, for example, all Nodes of a modified way, which is required, when trying to find the country where the change took place. Furthermore, the determination of the country for a specific OSM object really depends on the border’s level of detail: More detailed country borders make the processing quite time-consuming. Some of you probably experienced this problem before when using Osmosis or a different OSM processing tool. Anyway, for calculating additional country statistics I tried a new approach:

1. Determine the country of a changset based on its center position
2. Use the changeset country information for all objects within this changeset.

Of course, the determined country of the changeset can “only” be generalized for the entire changeset content, but how does it compare with the current method utilized in OSMstats? I compared last week’s numbers of OSMstats for each country of the world with the newly introduced approach. In total, the number of active members per country differs for each weekday by around 3% (min. 1% and max. 5%). The average difference of created, modified and deleted Nodes per country is quite similar with 4% (min. 2% and max. 9%). The presented approach could produce partially incorrect results whenever a changeset contains border changes of two or more countries or if the center of the changeset is in the wrong country. But IMHO the assumption to use the changeset centers is sufficient to calculate results and determine changes per country. As you can see in the figure above, most OSM changesets happen in a manageable area within one country. Yes I know, exceptions prove the rule.

So, why am I doing this? The main idea behind this approach is to change the entire processing task for OSMstats within the coming weeks. The changes per country will then be based on the introduced approach. Another advantage will be, that this newly created information, gathered from the changesets, can be utilized to create additional contributor statistics.

Thanks to maɪˈæmɪ Dennis.

This blog post is a summary of Dennis’ and my State of the Map (SotM) United States presentation. Maybe some of you already know about our publication: “Comparison of Volunteered Geographic Information Data Contributions and Community Development for Selected World Regions”. From the abstract: “Our findings showed significantly different results in data collection efforts and local OSM community sizes. European cities provide quantitatively larger amounts of geodata and number of contributors in OSM …”. “Furthermore, the results showed significant data contributions by members whose main territory of interest lies more than one thousand kilometers from the tested areas.” Especially the last finding is quite interesting when considering “arm-chair-mapping” in OSM.

However, for our SotM US session we repeated some of the conducted analyses for 50 urban areas in the United States to see whether similar patterns could be determined. You can find the session abstract here; additionally the ppt slides and also a video are online. The following animation shows the number of contributor’s evolution in the US from 2007 to 2014.

Similar to our prior research results for the selected 12 world regions, the US urban areas showed different individual patterns. Some cities such as Fargo (ND) experienced several data imports in the past which resulted in strong data density values (Nodes and Ways), whereas other areas solely rely on a small community of volunteers and contributors.

We also conducted a simple statistical analysis to evaluate whether certain socio-economic factors have an impact on the development of OSM communities in the different cities. Variables such as population density, per capita income and education showed a moderate to strong correlation with contributor numbers, highlighting that all of the aforementioned factors can have an impact on the success of OSM in the selected urban areas in the US.

It was also quite useful to take a look at the local contributor numbers vs. external contributors. Certain cities such as Miami (FL) heavily rely on data contributions made by mappers whose home region is more than 1000 km away, whereas other cities such as Los Angeles(CA) show large values for both, local and external mappers. You can check out your own area here too. The corresponding blog post is online here: “The OpenStreetMap Contributors Map aka Who’s around me?”. A more detailed analysis that is currently being conducted will reveal if cities that prove to have large external mapper contributions show the same quality as areas with lots of local mappers or not.

Last year we wrote a journal paper in which we analyzed the OpenStreetMap (OSM) dataset of the United States which was published on May 28th, 2013 in the Transactions in GIS Journal. You can download a free pre-print version here. This paper has been published just on time to add to the discussion at the upcoming State of the Map United States conference which will take place in San Francisco and includes some presentations about data imports to OSM. Unfortunately, Dennis and I cannot attend the conference this year, so we decided to write a blog post with some additional and up-to-date numbers.

In January there was an announcement on the OSM mailing list that in the past few months many connectivity errors in the United States OSM dataset had been fixed. Probably a lot of these fixes can be attributed to Martijn’s Maproulette website or to Geofabrik’s OSM Inspector (OSMI) Routing View. However, a short discussion started on the mailing list about the total number of errors that are left and how long it would take to fix all those errors. Thus, we downloaded four OSM planet files dated Jan 4th 2012, June 13th 2012, Jan 2nd 2013 and Jun 2nd 2013 to get some new results. After cutting the United States dataset from the planet files, we used the same algorithm as utilized in OSMI’s Routing View, to receive some stats about the street network of the US datasets.

First of all the, the following image shows the number of errors for each dataset that we included in the analysis. The errors that were detected are separated into unconnected and duplicate ways. You can find some additional information about both error types here.

As you can see, the number of unconnected OSM ways has been rapidly reduced in the past 17 months from around 141,000 to 19,000. The number of “duplicate way” errors has been reduced from 17,500 to 11,500. You can find the exact numbers in the following table and an updated error layer on the mentioned OSMI website. In certain cases the duplicate way error created several errors for one and the same way. For these particular cases the number of unique OSM way IDs were counted.

Date – Unconnected Ways – Duplicate Ways

• Jan 4th, 2012 – 141,578 – unique 17,563 (overall errors: 535,923)
• June 13th, 2012 – 145,468 – unique 17,977 (overall errors: 518,536)
• Jan 2nd, 2013 – 15,911 – unique 12,287 (overall errors: 257,388)
• Jun 2nd, 2013 – 19,073 – unique 11,582 (overall errors: 220,451)

Overall the length of the US street network did not really change a lot. At the beginning of 2012 it was around 11.07 million km while in 2013 it is 11.1 million km, which means an increase of around 30,000 km. The following image shows the distribution of the US street network divided by different OSM road classes.

The length of the residential roads is still decreasing (-496,000 km), similar to what we saw during the analysis for our paper, while the length of the other road types (+276,000 km) and secondary/tertiary roads (+205,000 km) is increasing. This is the result of a massive retagging process of the imported TIGER/Line dataset in OSM. Dennis mentioned this already in his SotM US 2012 presentation. Motorways also experienced an increase of around +44,000 km in 2012. You will find some additional, quite interesting statistics, charts and of course maps in the aforementioned journal publication. In particular a few more thoughts and facts about the effect and impact of data imports on OSM can be found in our research study about the United States OSM dataset.

One month ago I wrote a blog post about a new website which allows you to see other OpenStreetMap contributors in your area. Overall the feedback was very positive, thank you very much for that! However, now it is time for a new extension to the “How did you contribute to OpenStreetMap?” (HDYC) webpage. As I mentioned in my last blog post, I used an algorithm (which is described in a paper that I wrote here) to compute and determine the activity area of a contributor based on her/his changeset centers. The following figure shows the new function that was added to the HDYC website visualizing the activity area of a contributor! Sorry Harry, as always you have to be our guinea pig, but you have a really awesome activity area 🙂

Next to the visualization of the overall activity area of a contributor, you can also click on a link at the bottom of the map to switch to the contributors’ activity area of the past six months. Furthermore, all maps on HDYC now use the great Leaflet map library instead of Openlayers. Also, your activity areas’ first and last Nodes have a direct link to the “Overview of OpenStreetMap Contributors aka Who’s around me?” webpage. This provides an easy way to locate other contributors in your area. I have to mention that not every contributor has an activity area for the past six months. It highly depends on the activity of the contributor within this time frame!

One more thing: The aforementioned “Who’s around me?” webpage has three new overlays. Two overlays show the contributors of the past six months with their first and last Nodes and one additional layer shows the activity areas also based on the past six months for each contributor. You can find all new layers in the upper right corner in the so-called “Layerswitcher”.

My HDYC database is updated more or less on a daily basis. The information about your changeset activities is updated once a week (based on the weekly changeset dumps from here). “The Created Nodes per Country”-section can only be updated when a new full history dump is available, but you can always find the latest date in the section-label. The “Who’s around me?” webpage uses almost the same database as HDYC, so the data up-to-dateness is similar.

Have fun with the new gadgets!

¡Muchas gracias maɪˈæmɪ Dennis

Some of you might already know that OpenStreetMap released a first bulk GPS point dataset last weekend. It contains almost 2.8 milliard (or for readers in the US 2.8 billion) points and is provided in its raw format, which means that only coordinate information is available for each point. Unfortunately it does not include any additional information or metadata. You can read more about it at the OSM Foundation Blog.

The first idea that came to my mind was a simple comparison analysis to answer the following questions: Where are all those points located and which country has the most GPS points? In a first try I conducted some results that showed that all points are distributed over 238 countries. For my analysis I used the OSM Mapnik world boundaries from the wiki. As you can see in the following pie chart, nearly 21% of the points are located in Russia (about 570 million points) and another 18% in Germany (about 500 million points). Does Russia have so many GPS points because of the country size or is the community just exceptionally active with GPS devices? However, the strange thing is that Germany is, with about 18%, “only” on the second place this time, weird isn’t it? 😉

I think overall these are some quite interesting numbers. We all hope to see some more metadata information in the OSM GPS point dataset soon.

thx @ maɪˈæmɪ Dennis & Good luck for next week!

*UPDATE* April 11th, 2012
The following map shows the OSM GPS points per 1000km²:

The second map shows the OSM GPS points per 1000 inhabitans:

The OpenStreetMap project will possibly finalize its license change on April 1st 2012. There are certain concerns in the community about possible data losses and to keep them as little as possible, several remapping activities have been started. A really nice overview of “Remapping principles” and “Tools to help you” can be found here.

Frederik’s OSMInspector (OSMI) and Simon’s CLEANMAP are two very handy remapping tools. Both display data that will likely be removed after April 1st due to the fact that this data was collected by contributors that did not accept the license change. In Germany you will find several areas that are affected by these changes and might even leave some new blank spots in the map. But what impact do these changes have on the total length in kilometers per street category in Germany?

You can find several files regarding the OSMI license change view at a Geofabrik server here. Based on the “ways” shape file that you can find there it is possible to calculate the total length of the ways, which will likely be removed with the license change. However, sadly the “ways” shape file does not include any “highway” attribute, but luckily it includes the OSM IDs. This means that to be able to do a Germany street network analysis you will have to download the Geofabrik Germany OSM *.pbf file. By applying a short script you can get all OSM way IDs in Germany with their highway=* key/value pair. Combining these with the “ways” shapefile allows us to calculate the total lengths of each highway type for “Germany” (based on the Geofabrik extract!).

The following image shows the results of a first attempt to visualize the values per street category. Overall this means that based on the current (January 15th 2012) license agreement/disagreement situation about 5.4% (94000 km) of the current street network in Germany will be removed after the license change in April. The relative difference for each highway type lies between 3-8%. Last week (Jan. 7th, 2012) the total amount in Germany was 5.9% and 103000 km.

The OSMI License Change view contains not only the data that will potentially be removed in the future but also some information on two additional feature-types: Features, which have been modified and features which have been modified in some minor way by a contributor that declines the license change. In the first case we have a total street network length of about 58000 km that is affected and in the second case about 17000 km. Remember, these numbers only reflect the situation in Germany! You can find more information about the different feature types here: “Understanding the Colour Scheme“.

Notice: This was a short hack done last night, but I think those numbers look realistic. Can anyone confirm this for Germany? I am very curious how and if these numbers will decrease in the next few weeks. What do you guys think?

thx @ maɪˈæmɪ Dennis