User:CitySketcher/Project Waimate

Background

This page, and Project Waimate, are a work in progress!

I'm staying in Waimate for a few weeks over new year 2021-2022. I noticed that there was very little infrastructure on the map, so I decided I'd spend some time mapping as much as possible. This is Project Waimate.

Previously missing features

• Almost all building outlines.
• Almost all POIs like shops, especially on Queen Street, though churches are already mapped well.
• Green area details.
• Pedestrian features like footpaths, benches, picnic tables, drinking fountains, playground equipment.
• All car parks.

Strategy

• Draw building outlines using aerial imagery and RapiD based on the LINZ building outlines dataset.
• If I cannot align existing address nodes to the outlines based on imagery, visit the houses.
• Survey green areas, pedestrian features, and POIs on foot, noting elements using GeoNotes and then entering data at the computer.
• Add road details through StreetComplete surveys.
• Record my trips using OsmAnd, then use the GPX data to add any footways and paths that aren't already plotted. The aerial imagery is aligned, but it may be obscured by buildings or trees, so being able to refer to a map trace is very helpful as an alternative.
• TODO: Upload the traces to help other editors?

Reports

Generating reports to see statistics about what I've done.

I will use the bounding box -44.7523,171.0214,-44.7217,171.0629 for all queries to keep things consistent. This area extends to include Knottlingly Park in the south, the whole of Mill Road in the west, Cashel Street in the north, and the residential area along Timaru Road in the east.

Most active users

This counts how many objects each user has created or edited. I don't know how to do this in Overpass only so I'm using fish shell to count changes and generate wikitable syntax.

Shortcomings of counting edits

Counting edits is useful for getting a general overview of which people have spent the most time on the map, but it is never going to be an accurate assessment of any kind of "value" of their edits.

• This counter only counts the last editor of each object. If somebody draws a road, and somebody else adds tags for whether the road has pavements, the whole road is worth 1 credit for the second person and 0 credits for the first person.
• Depending on the counting method, a 4-sided house might be worth 5 credits, 1 for each of the 4 nodes and 1 for the closed way. However, in this count, I have only considered elements that have tags, so each way is 1 credit no matter its size.
• Some edits may be graffiti. The counters don't know whether you've just created a major highway or a pictogram of a middle finger made of highways.
• Some edits may be more accurate than others. An on-the-ground survey of a building is more likely to discover the correct opening hours than if the opening hours were added by an armchair mapper referencing an out-of-date website. (I still love you, armchair mappers <3)
• Some objects are much more complex than others. For example, a amenity=place_of_worship building has complex outlines and may have many attribute tags compared to a highway=crossing which is a single node at the intersection of two existing ways with no other tags. Adding the church is worth 1 credit, and adding the crossing is worth 3 credits.
• Some objects are more practically useful than others. Adding an natural=tree has very little impact to almost everybody. Joining existing bus stops with a bus route relation will likely help many pedestrians travel a city quickly and easily. Adding an amenity=hospital or amenity=vetinary could save lives.

Query code

printf '{| class="wikitable"\n|-\n! Username !! Objects edited\n'; curl -Ss https://overpass-api.de/api/interpreter -d '[bbox:-44.7523,171.0214,-44.7217,171.0629][out:csv(::user; false)]; nwr(if:count_tags() > 0); out meta;' | string replace -ir '.*linz.*' '[linz data importers]' | sort | uniq -c | sort -nr | while read count user; printf '|-\n| %s || %s\n' $user $count; end; printf '|}\n'

Results

Username	Objects edited
cityofgod	4437
[linz data importers]	928
Robert Ancell	245
Rudy355	127
Aleksandar Matejevic	33
aleksaJov	7
SteveFish	6
Friendly_Ghost	6
lcmortensen	5
kylenz	5
TaraV	4
DoubleA	3
AndjelaS	2
Wallip	1
Shengnan Song	1
pizzaiolo	1
locke	1
gunnar1m	1
b-jazz-bot	1

Though the results aren't visible on a map, you can also open the query in overpass turbo. https://overpass-turbo.eu/s/1ePn

Types of features edited by me

Background

OpenStreetMap tagging is famously freeform, so first I need to figure out which tags I want to analyse as features. I did this by querying everything, and then gradually iterating on the query by removing features. I ended up with this:

[bbox:-44.7523,171.0214,-44.7217,171.0629];
nw(if:count_tags()>0)[!highway][!building][!"addr:housenumber"][!landuse][!amenity][!leisure][!natural][!barrier][!tourism][!playground][!disc_golf][!historic][!attraction][!man_made][!shop];
out tags center;

This means that I care about features with these keys: highway, building, address, landuse, amenity, leisure, natural, barrier, tourism, playground, disc_golf, historic, attraction, man_made, shop.

After separating those keys, the remaining few features were not important to me, so that is my final list of keys to analyse.

For some of the keys, like amenity, the value is also important since the value of the amenity key says what type of amenity it is. For other features, like addresses which are represented by addr:housenumber, the type of feature is already known and it would be silly to try to separate the values from each other.

I decided to render my results as a table, and to only include table rows where there were at least 2 objects edited.

Now that I know what I am looking for, I can write a query.

Note that some keys may be counted twice in the table, due to the aggregate count first and then the individual value counts afterwards. Some keys are only in the first section, some keys are only in the second section, and some are in both.

Query code

View full query

printf '{| class="wikitable sortable"\n|-\n! Key !! Value !! Count\n'; curl -Ss https://overpass-api.de/api/interpreter -d '
[bbox:-44.7523,171.0214,-44.7217,171.0629]
[out:csv(count, key, value; false)];

(
  make x key="addr:housenumber", type="node";
  make x key="attraction", type="node;way";
  make x key="barrier", type="node;way";
  make x key="building", type="way";
  make x key="disc_golf", type="node;way";
  make x key="highway", type="way";
  make x key="historic", type="node;way";
  make x key="landuse", type="way";
  make x key="playground", type="node;way";
  make x key="shop", type="node;way";
  make x key="tourism", type="node;way";
);
foreach->.feature_key {
  nwr(user:cityofgod)(if: t[feature_key.u(t["key"])] && lrs_in(type(), feature_key.u(t["type"])));
  make x key=feature_key.u(t["key"]), count=count(nwr);
  out;
};

(
  make x key="building", type="way";
  make x key="amenity", type="node;way";
  make x key="highway", type="way";
  make x key="leisure", type="node;way";
  make x key="man_made", type="node;way";
  make x key="natural", type="node;way";
  make x key="shop", type="node;way";
);
foreach->.feature_key {
  nwr(user:cityofgod)(if: t[feature_key.u(t["key"])] && lrs_in(type(), feature_key.u(t["type"])));
  for (t[feature_key.u(t["key"])]) (
    make x key=feature_key.u(t["key"]), value=_.val, count=count(nwr);
    out;
  );
};
' | while read count key value; test $count -ge 2; and printf '|-\n| %s || %s || %s\n' $key $value $count; end; printf '|}\n'

Results

View table of results

Key	Value	Count
addr:housenumber		1188
attraction		4
barrier		57
building		2550
disc_golf		27
highway		246
historic		2
landuse		80
playground		41
shop		32
tourism		8
building	church	3
building	commercial	2
building	detached	1080
building	garage	651
building	garages	2
building	house	8
building	kindergarten	3
building	residential	9
building	roof	14
building	school	11
building	semidetached_house	26
building	shed	7
building	terrace	8
building	yes	720
amenity	bench	115
amenity	bicycle_parking	12
amenity	cafe	2
amenity	clock	2
amenity	community_centre	2
amenity	drinking_water	15
amenity	fast_food	3
amenity	fuel	2
amenity	kindergarten	2
amenity	parking	35
amenity	place_of_worship	6
amenity	public_bookcase	3
amenity	school	3
amenity	shelter	8
amenity	toilets	4
amenity	waste_basket	42
highway	footway	71
highway	path	7
highway	primary	5
highway	residential	83
highway	service	50
highway	steps	7
highway	tertiary	11
highway	track	2
highway	unclassified	10
leisure	fitness_station	4
leisure	garden	8
leisure	park	4
leisure	picnic_table	20
leisure	pitch	36
leisure	playground	8
leisure	swimming_pool	4
leisure	track	3
man_made	drinking_fountain	14
man_made	mast	2
man_made	silo	4
man_made	tower	2
natural	tree	122
natural	tree_row	4
natural	water	2
shop	agrarian	2
shop	clothes	6
shop	collector	2
shop	hairdresser	2
shop	vacant	3

Conclusions

• Buildings and addresses were my most edited feature. Of buildings, detached and garage were the most common values. This makes sense since Waimate is primarily a low density residential town, and most of the town's area is taken up by detached houses.
• After this, the element counts drop significantly. The next largest is natural=tree with 122 uses. I added trees based on aerial imagery and scattered them around in parks where I felt like it. Trees are commonly planted physically close to each other and they are very quick to add based on imagery, making them my second most added feature.
• My third most added feature was benches with the tag amenity=bench. I was expecting benches to feature high on the list since they are one of the things I was paying the most attention to while journeying through the town. They are also quick to survey, even though I take the time to add the backrest, armrest, material, and sleepable status for each bench. I either add them by photographing them with GeoNotes, or writing a short text description of their features.

GPS tracks and distance travelled

TODO

Ideas for the future

I want software which takes spoken audio from me, notes my current location, processes my words into text, creates a point of interest based on my commands, and adds it to a GeoJson file with the correct tags which I can very quickly bulk review and import using JOSM later when I get home. This would also make it much easier to survey while riding in a vehicle, because buses and cars move so fast that I would miss features if I looked down at my phone to write a note, and using a phone while riding a bicycle would be dangerous unless I stopped and dismounted each time I had something to take a note of.