City Walk/WissenZum StadtspaziergangViewer
ImpressumDatenschutz
Leitfaden
  • So funktioniert der Stadtspaziergang
  • Woher die Daten kommen
  • Quellen nach Stadt
  • Wie der Viewer zu einer Angabe kommt
  • Der Weg der Daten
  • Bedienung
  • Glossar
Guide (English)
  • How the city walker works
  • Where the data comes from
  • Sources by city
  • How the viewer comes by what it shows
  • From download to browser
  • Using the viewer
  • Glossary
Entwicklung (English)
  • docs/ — the project knowledge base
  • Rendering — how the data becomes pixels
  • Data pipeline — the bakes, the build step and the artifact contracts
  • Architecture decision records
  • Data → feature provenance
  • Transformation catalog
  • Portability — rendering any location, with whatever data it has
  • Writing a walk
  • Routines: the twin's scheduled agents
  • Implementation plans, backlog and audit history
  • What the data streams carry that the walker does not use yet
Contents
Leitfaden
  • So funktioniert der Stadtspaziergang
  • Woher die Daten kommen
  • Quellen nach Stadt
  • Wie der Viewer zu einer Angabe kommt
  • Der Weg der Daten
  • Bedienung
  • Glossar
Guide (English)
  • How the city walker works
  • Where the data comes from
  • Sources by city
  • How the viewer comes by what it shows
  • From download to browser
  • Using the viewer
  • Glossary
Entwicklung (English)
  • docs/ — the project knowledge base
  • Rendering — how the data becomes pixels
  • Data pipeline — the bakes, the build step and the artifact contracts
  • Architecture decision records
  • Data → feature provenance
  • Transformation catalog
  • Portability — rendering any location, with whatever data it has
  • Writing a walk
  • Routines: the twin's scheduled agents
  • Implementation plans, backlog and audit history
  • What the data streams carry that the walker does not use yet
Guide (English) · 2 / 7

Where the data comes from

Deutsch: Woher die Daten kommen

Everything in the viewer is derived from open data: datasets that a public authority or a volunteer community publishes for anyone to use. Nothing was surveyed or drawn by hand for this project. This page lists each dataset, where it was downloaded, what it is good at, where it falls short, and under which licence it may be used. The abbreviations are explained in the glossary.

The providers

GeoSN — the Landesamt für Geobasisinformation Sachsen, Saxony's state survey office. It publishes the terrain and surface models, the 3D building model, the land-use map and the aerial photos as free downloads on its open-geodata portal, geodaten.sachsen.de. All of them are cut into the same 2 km × 2 km tiles, which is why the viewer thinks in tiles too. Which tiles it shows is written down in one place, the site config sites/dresden.ts: the fifteen tiles (the first is the one you start on) and the viewpoints; the credits come from the provider.

Other places, other providers. Every German state runs its own survey office, and the viewer can draw on five of them: GeoSN for Dresden, Leipzig, Meißen and Grimma; Geobasis NRW for Unna; the Bavarian survey administration (LDBV) for Munich; Hamburg's LGV; and Berlin's Senate department for urban development. They publish the same kinds of datasets in different cuts — 1 km tiles, one archive for the whole city, aerial photos without the infrared band — and a small adapter per office cuts them into the viewer's 2 km tiles. Hamburg and Berlin do not publish the landscape model in the form the viewer reads, so their ground colours come from OpenStreetMap instead — and so do their bridges: OpenStreetMap's bridge ways stand in, while each deck's height is still measured in the surface model. (Every city's railway tracks come from OpenStreetMap, one by one; elsewhere the landscape model fills in only where it maps none.) Munich's aerial photos have no infrared band, so the colour of its trees and meadows is computed from the photo's visible colours instead (see NDVI below). And where a building's material is not mapped, it takes its neighbourhood's: in a street of mapped brick houses it is drawn in brick, elsewhere in plaster — so Hamburg's clinker quarters come out in brick without anyone having to say so for the whole city. One deployment shows every city whose data it was built with, each at its own address (/dresden, /leipzig, …), and the start page lists them; the list of offices with their licences is in sites/providers.ts. Which city draws what from which source — and where a stand-in steps in because a Land or a city does not publish something openly — is the table Sources by city.

OpenStreetMap (OSM) — the world map maintained by volunteers. It fills gaps the official datasets leave: street lamps, benches and other street furniture, station platforms, the railway tracks one by one, retaining walls with their heights, what kind of structure a bridge is, the shape of fountain basins, and the trees in courts and gardens the city's register does not list — in a city without a register, every tree it maps, as tall as the surface model measures it.

Wikidata — the free knowledge base behind Wikipedia. For named bridges it says what kind they are (arch, suspension, truss …) and often their main span and number of spans. It also names each city's landmarks: the buildings and structures with Wikipedia articles, ranked by how many languages write about them. The viewer marks the buildings that are one, lists the city's twelve best-known in the panel, and uses the material Wikidata gives (glass, brick, stone …) where OpenStreetMap names none. Asked, a landmark's name stands beside it in the city with what Wikidata knows of it: when it was built and by whom, when it was rebuilt, how tall it is. For monuments and sculptures it often says what they are made of — the Goldener Reiter is copper under gold leaf — and the viewer colours them in a muted tone of that material where OpenStreetMap names none. A building Wikidata records as demolished is left out, unless OpenStreetMap still tags a standing building with it — before, Dresden's Neues Rathaus was marked as the Alte Reformierte Kirche that stood there until 1894. For a mural OpenStreetMap maps without them, Wikidata may give its size and a photo. An excerpt of what Wikidata states about each landmark (architect, style, inception, height, heritage status, what it is named after …) is kept with the city's data — fetched once, not by your browser — as material for the walks and for Frag die Stadt, the questions in words.

Wikimedia Commons — the free media archive behind Wikipedia, where photographers share their pictures, each under a free licence of their choice. A mural OpenStreetMap maps with its size is drawn on its wall after a photo from Commons, where one shows the whole mural straight on: in Dresden the Fürstenzug and Der Weg der Roten Fahne on the Kulturpalast. Drawn, not photographed: pen lines in one dark ink over a light wash of the mural's own colours, in the pastel of the rest of the city. A photo taken at a slant, of a detail or of a mural of another size is not used, and the wall stays plain. Each photo's author and licence are listed in the sources in the panel's footer; the drawing is under its photo's licence.

Landeshauptstadt Dresden — the city itself. Its street-tree register (Stadtbaumkataster) lists about 124,000 municipal trees with species, height, crown width and trunk. The viewer plants those trees where they really stand, with their measured height, crown and trunk, a crown shape that follows the species, and the species' year: when it leafs out, which colour it turns in autumn and when it is bare. The same source provides two traffic datasets the viewer shows as switchable data layers (sidebar, Erkunden → Verkehrsdaten; off at start): the motor vehicles counted per road section, and the permanent bicycle counters, whose counts the browser reads live from the city.

DELFI and gtfs.de — DELFI e.V. collects the timetables of every transport association in Germany (in Dresden the VVO's, with the DVB's lines; in Leipzig the LVB's, in Munich the MVG's); gtfs.de publishes them in the common GTFS format. The trams of the third data layer run from it, as the timetable has them run, each in its operator's colours.

Copernicus — the European Union's Earth-observation programme. Its elevation model GLO-30, made by DLR and Airbus Defence and Space, covers the whole world at 30 m. The viewer reads it only for the land beyond each city's tiles: the hills on the far horizon, out to 10 km, which the baked sky light and the far shadows take in, and the rim of the valley the city lies in, which sets how deep the valley haze pools. Before it stands in for the ground, it is moved onto the city's own terrain model where both see open ground; inside the tiles the terrain model alone counts.

Deutscher Wetterdienst (DWD) — Germany's national weather service. Its volunteer observers note every year on which day the trees and shrubs around them leaf out, flower, turn and drop their leaves. The viewer's year of the trees is Dresden's; every other city's is shifted by the difference the observers within 30 km of it recorded for twelve trees and shrubs, 2010–2023 — Hamburg's trees leaf out three days later and turn five days later, Unna's leaf out six days earlier. The observations were read once, offline, into a table of shifts; the browser asks the weather service nothing.

Hamburg, Leipzig and Berlin publish registers of the same kind, read the same way. Hamburg's Straßenbaumkataster (the environment authority, BUKEA) lists the street trees with species, crown width, trunk girth and planting year, but no height: the viewer derives each tree's height from its genus and trunk thickness, learned from Dresden's and Leipzig's registers, which record both. Hamburg's surface model was flown with the trees bare and reads them low, so it only raises a height it clearly exceeds and drops a tree it shows gone. Leipzig's Baumkataster (Amt für Stadtgrün und Gewässer) lists street and park trees with height, crown and trunk; felled trees are left out. Berlin's Baumbestand (Geoportal Berlin) lists street and park trees the same way; Berlin is configured but not built yet.

The datasets at a glance

DatasetPlain-English nameProviderWhat the viewer uses it for
DGM1Terrain model, 1 m gridGeoSNThe ground; seating every object on it; bridge abutment heights and the water level under the fairway; one level for each body of water, and the beaches and gravel bars inside it; the input for tree heights
DOM1Surface model, 1 m grid (terrain plus everything standing on it)GeoSNTree heights (surface minus terrain), and where the trees of a mapped tree row or orchard stand; a bridge's roadway height and width and what stands above it: trusses, pylons, arches, and the piers under a stone bridge; chimneys, towers and masts, buildings the 3D model does not carry yet, and platform roofs, carports and shelters — only where OpenStreetMap names them; the roof shape of a landmark the 3D model draws flat; the roofs of buildings the 3D model gets badly wrong; the dormers on pitched roofs and the house chimneys; the canopy over a shop front; the decks of the landing stages and how high the pontoons float; hedge heights where there is no laser scan
LoD23D building model with roof shapesGeoSNEvery building's footprint, height, roof shape and attributes (the viewer leaves out its bridge slabs)
Basis-DLMDigital landscape model (the land-use map)GeoSNGround colours (fields too), water outlines, hedges and tree rows, railway areas, tracks where OpenStreetMap maps none, bridge outlines, monuments and fountains (position and official name)
DOPDigital orthophoto, 20 cm, with a near-infrared channelGeoSNRoof colours; the colour of each body of water; vegetation greenness for tree crowns, meadows and crops; which way the vine rows run; where no laser scan is read, the trees in courtyards and gardens (green where the surface model is rough); in a mixed forest, where the photo was flown before the leaves came out, the crowns still green are conifers, stay green all year and stand as cones, as every crown in a needle forest the land-use map names does
LSCLaser-scan point cloudGeoSN (Dresden, Meißen, Grimma, Leipzig), Geobasis NRW (Unna), Bavarian survey administration (Munich)Hedge heights and hedges OpenStreetMap lacks (long, slim, green ones clear of trees); trees in courtyards and gardens; the garden houses, sheds and container buildings the 3D building model lacks
OSMOpenStreetMapVolunteersStreet lamps, hedges, street furniture (benches, bins, bicycle stands, bollards, post boxes, stop shelters and stop signs, advertising columns, traffic signals, hydrants, clocks, drinking fountains, flagpoles), playgrounds and their equipment, station platforms and their heights, railway tracks one by one with their electrification and overhead-line masts, walls, cliff edges, stairs, bridge structure types and navigation clearances, fountain basins, what streets, pavements and car parks are paved with, sports grounds, shops and cafés on the ground floor, listed buildings, fences, railings and gates, road markings (crossings, stop lines, cycle and centre lines), allotment gardens, orchards and vineyards, trees the city's register does not list (in a city without one, every mapped tree), tram tracks with their overhead-line masts, the landing stages, groynes and ferry routes on the Elbe, what buildings are made of and their wall and roof colours, chimneys, towers and masts, the buildings too new for the 3D model (at their mapped height where the surface model was flown before they were finished), platform roofs, carports and shelters, and murals with their size
Wikidatafree knowledge baseVolunteersThe kind, main span and number of spans of named bridges; the city's landmarks (the list in the panel), their facade material and the facts under an asked landmark's name (built, architect, rebuilt, height); what monuments are made of; a mural's size and photo where OpenStreetMap gives none
Wikimedia Commonsfree media archivePhotographers, each photo under its own licenceThe photos the murals are drawn after, where one shows the whole mural straight on (in Dresden the Fürstenzug and Der Weg der Roten Fahne), and the photos on the walks' cards
MapillaryStreet photos and the objects found in themContributors' photos, Mapillary's detectionIn Dresden, Grimma, Meißen and Unna, street lamps OpenStreetMap does not map (about 5 400 in Dresden, 950 in Grimma, 120 in Meißen and 130 in Unna). Mapillary places each object it recognises in several photos; the viewer keeps those with no OpenStreetMap lamp within 8 m, seen since 2020, seen from close by (a photo within 25 m) and by daylight at least once, not inside a building and not over a road tunnel, moved off the carriageway to the kerb. Their positions are a few metres off, and a lamp it placed twice from two drives stands only once (7 m apart at least). Its litter bins are not used: in street photos most of them are household wheelie bins put out for collection. Asked, a lamp says whether it came from Mapillary. The panoramas — in every city but Unna, where Mapillary has none — also say, per building (nearly 2 000 in Dresden and 1 600 in Leipzig, mostly along the inner city's streets), how busy and how dark a street front is and whether a shop sign hangs on it; the clay turns that into a fine relief, a slightly darker or lighter tone and a shop plinth (Fassadenbild). Two further measurements give the windows. The shop window: on every tile the panoramas are measured once more along each wall's ground floor, and where two photos agree on wide openings — or where a modernist shop row shows glass nearly throughout — the building gets shop windows there as soft niches: the glass a muted tone a little darker and cooler than the wall, slightly reflecting, set in a shallow reveal with rounded edges, no mullions, framed by piers, with a fascia where a shop sign was seen (about 1 000 walls on 780 buildings in Dresden, 1 130 on 843 in Leipzig, 3 079 on 2 229 in Munich, 152 on 112 in Hamburg, where few panoramas were taken, 79 in Grimma and 23 in Meißen). Where the surface model (DOM1) shows a flat canopy over such a shop front, as on the Hauptstraße's GDR shop pavilions, the canopy is drawn too, and the glass runs along under it. A shop OpenStreetMap maps where no photo shows one gets no window: a map point says a shop is there, not where its windows are. The window rhythm: the panoramas are measured once more over each wall above the ground floor, for what does not depend on exactly where the camera stood — how far apart the window axes are and how regular, the storey height, the windows' width and proportions, ornament. A trait is used only where two separate drives past the same walls agree on it (a rank correlation of at least 0.6); the window surrounds (Faschen), sill bands, pilaster strips and plinths did not. That gives a measured rhythm for 1 697 walls of 1 382 buildings in Munich, 577 of 513 in Dresden, 448 of 380 in Leipzig, 87 in Grimma, 18 in Meißen and 12 in Hamburg (as of 2026-11-02; Mapillary has no panoramas in Unna). Every other house borrows the rhythm of the nearest measured one within 150 m with the same roof form and about its eave height (a block's houses share their era), or else takes its type's — a town house of tall storeys, a walk-up, a small house, a flat-roofed block — whose numbers are the medians the photos measured on Dresden's houses of that type. The windows are drawn in the wall, not modelled: an opening the eye looks into, its reveals in the wall's own plaster, its back darker and cooler, a slim sill under it, its back mirroring a little sky at a grazing view — no panes, no mullions —, one row per storey. Where the photos also show where the windows stand, they are drawn there: the strips of one wall from several drives are matched by their window edges and their position errors averaged (about 0.1–0.2 m left), so the axes sit where the photos saw them, at the spacing the photos measured on that very wall (on 580 walls in Munich, 157 in Dresden, 79 in Leipzig, 32 in Grimma, 6 in Meißen and one in Hamburg); elsewhere the windows are centred on the wall, their height following the storey. None on a party wall, beside a door, over a shop front, on a church, palace, theatre, museum or hall, or on a glass or metal facade. Which facades are glass, where OpenStreetMap does not say, the photos tell too — any photo, not only the panoramas, since a glass front often stands back from the street behind a plaza that no car drives past: each photo that sees a wall well is laid onto it as for the window rhythm, and an open image model (CLIP) is asked whether the wall looks like a glass facade or a wall with windows; where most of at least two photos from separate drives or walks say glass, the wall is glass, and a building whose walls are glass for at least half the length read — and for a quarter of all its walls, or 15 m and a twentieth of them, so one glazed stair tower does not make a stone house glass — becomes a glass facade, with no windows drawn on it (thirteen buildings in Dresden, twelve of them not tagged in OpenStreetMap, among them the Congress Center on the Elbe and the adult education centre (Volkshochschule)). The model runs only when the data is prepared, never in the browser
StadtbaumkatasterThe city's street-tree registerLandeshauptstadt DresdenStreet and park trees at their surveyed positions, with height, crown width, trunk, a crown shape from the species and the species' autumn colour and leaf fall; asked, their species, location, tree number and age
Straßenbaumkataster / Baumkataster / BaumbestandThe street-tree registers of Hamburg, Leipzig and BerlinFreie und Hansestadt Hamburg (BUKEA); Stadt Leipzig (Amt für Stadtgrün und Gewässer); Geoportal BerlinThe same as Dresden's; Hamburg's register has no heights: they come from each tree's genus and trunk thickness, learned from Dresden's and Leipzig's registers
VerkehrsmengenMotor vehicles per day and road sectionLandeshauptstadt DresdenData layer Kfz-Verkehr: glass flows per direction, wider, taller and deeper in colour where more traffic runs
TagesgangEach hour's share of the day's trafficFreie und Hansestadt Hamburgthe time of day of the Kfz-Verkehr data layer
Other cities' traffic counts, road censusesMotor vehicles per day and section in Berlin and Hamburg; the road censuses of Saxony and NRW on their roadsBerlin (SenMVKU); Hamburg (BVM); Free State of Saxony (LASuV); Straßen.NRWthose cities' Kfz-Verkehr data layer, both directions counted together
Hamburg's cycle counting networkBicycles of the last hour, liveFreie und Hansestadt Hamburgthe Radverkehr (live) data layer in Hamburg
Rad-DauerzählstellenBicycles of the last hour, liveLandeshauptstadt DresdenData layer Radverkehr (live): two glass columns per counter, one per direction
GTFS timetableThe scheduled local-transport timetableDELFI e.V. via gtfs.deData layer Straßenbahnen (Fahrplan): every tram (DVB, LVB, MVG) at the scene's time on its track, in its operator's colours and length
GLO-30Copernicus DEM, a 30 m surface model of the whole worldDLR / Airbus Defence and Space, under the EU's Copernicus programmeThe land beyond the site: the hills on the far horizon (to 10 km) and the valley rim for the valley haze
DWD phenologyThe weather service's yearly plant observations: when trees and shrubs leaf out, flower, turn and drop their leavesDeutscher WetterdienstHow many days earlier or later than in Dresden each city's trees leaf out and turn

Available from the same portal but not used yet: the cadastral parcels (ALKIS), and the topographic base map (DTK). The planned section of the transformation ledger records what they could add.

How the official datasets are produced

Most of what the viewer shows traces back to two kinds of flights over the city and to the survey office's own databases. The diagram shows how the products depend on each other; the fact sheets below give the details.

Survey-office databases, maintained continuously

Photo flight — Dresden: 19 Mar 2024

Laser-scan flight (LiDAR) — Dresden: 27–30 Nov 2024

used to rectify

Point cloud (LSC)
every laser echo, classified
ground / not ground

DGM1 — terrain model
ground points → 1 m grid

DOM1 — surface model
first echoes → 1 m grid

nDOM = DOM1 − DGM1
height of everything on the ground

DOP20 RGBI — orthophoto
red, green, blue, near-infrared · 20 cm

NDVI — greenness index
(NIR − red) / (NIR + red)

Basis-DLM — landscape model
roads, rails, water, land use, …

LoD2 — 3D buildings
footprints from the databases,
roof shapes fitted to the point cloud

Dataset by dataset

Each fact sheet answers the same questions: what the abbreviation means, how the data is collected, how often it is updated, how precise it is, what it is generally good for, what the viewer uses it for, and where it falls short. Figures marked "per GeoSN" come from the provider's product documentation.

DGM1 — the terrain model

Stands forDigitales Geländemodell 1 — digital terrain model, 1 m grid. "Terrain" means the bare ground: buildings, bridges and vegetation are removed.
How it is collectedAirborne laser scanning (LiDAR): an aircraft sweeps the ground with laser pulses and records each echo. The resulting point cloud is classified into ground and non-ground points; the ground points are interpolated into a regular grid. Gaps under buildings are filled with interpolated points.
Update cycleRegion by region, after each new laser flight. For Dresden the previous scan dated from 2016, the current one from 27–30 November 2024.
Resolution and accuracy1 m cells, height stored in the cell centre. Height accuracy up to ±0.15 m and position ±0.30 m at 95 % confidence, per GeoSN.
Generally suited forAny "how high is the ground here" question: terrain analysis, flood and drainage modelling, visibility studies, slope maps, rectifying aerial photos.
Used here forThe ground itself — a triangle mesh that keeps every point of the 1 m grid within 15 cm (25 cm on the outer tiles), dense where the ground bends and sparse where it is flat, so walls and embankments keep their edges; seating buildings, trees, lamps and the player on it; the abutment heights of bridges; the base term of the tree heights.
StrengthsThe true shape of the ground, including river banks, embankments and the terraces of the old town, to a few centimetres.
WeaknessesAnything vertical: a laser-scanned wall becomes a steep ramp one to two metres wide, never a vertical face — and once the viewer resamples the 1 m grid to its 2 m mesh, monumental walls such as the Brühlsche Terrasse "disappear" into a gentle bank about 3 m wide. Bridges are removed by definition, so a deck taken from this model would sink to the river bed. The viewer fixes both with other sources.
Format and downloadGeoTIFF, 2000 × 2000 pixels of 32-bit floats, 13.6 MB per 2 km tile, with a .tfw world file and an _akt.csv stating the survey date. Heights in metres above sea level (DHHN2016). Digitale Höhenmodelle.

Special role: this is the only bulk raw dataset committed to the repository, because the build step reads it directly. See From download to browser.

DOM1 — the surface model

Stands forDigitales Oberflächenmodell 1 — digital surface model, 1 m grid. "Surface" means the first thing the laser hit: roofs, tree tops, bridge decks, parked lorries.
How it is collectedThe same laser flight as the DGM1; the first echoes of the point cloud are interpolated into the grid instead of the ground points.
Update cycleTogether with the DGM1; identical dates for these tiles (27–30 November 2024).
Resolution and accuracy1 m cells; the same height accuracy as the point cloud (up to ±0.15 m), per GeoSN. Note that a cell on the edge of a crown or a roof can hold either height.
Generally suited forBuilding and tree heights (as DOM − DGM, the nDOM), canopy mapping, distinguishing deciduous from coniferous trees (leafless crowns let the ground show through), finding bridges and sunken entrances, solar-roof studies.
Used here forDOM1 − DGM1 gives the height of everything standing on the ground. Where the land-use map says forest, copse or park, the viewer places one tree per 7 m cell at the tallest point, with that height; the trees of a mapped tree row or orchard stand at the crown peaks it shows along the row or inside the orchard, as tall as measured. Bridges take from it the height and width of their roadway (a bridge mapped only as a line is as wide as the surface shows its deck — the Marienbrücke 20.5 m, not the 11 m of its kind) and everything standing above it: along a bridge's axis, the highest point over the deck traces the Blaues Wunder's truss between its two pylons and the Waldschlößchenbrücke's steel arch; lamps and cars are filtered out as too short. Under a stone arch bridge, the piers stand where the deck bulges out over them. It is also the only reliable record of what a monument or a fountain's sculpture looks like: its measured bulk (the Albertplatz sculpture groups are 3.7 m tall bodies about 4 × 5 m) becomes a soft clay form of that size and outline; a tall column or obelisk stands at its measured height (Munich's Mariensäule 13.2 m and Friedensengel 36.6 m, read on the laser scan there), and on a playground OpenStreetMap maps without equipment, what stands 1.2–4 m tall there and is no tree becomes a swing or a climbing frame. The small buildings the 3D model lacks are found in the same surface, but read from the point cloud itself at 0.5 m, where each echo also says whether the pulse split (see LSC). What stands higher than the 3D building model — the surface minus the higher of ground and LoD2 roof — is added only where OpenStreetMap says what it is: a mapped chimney, tower, mast, water tower or lighthouse becomes a round column of the measured height (Unna's 48 m Lindenbrauerei chimney), and a mapped building the 3D model does not carry yet a block of the measured height (under the roof's shape the surface shows, where it measured one) — or of OpenStreetMap's, where the surface model was flown before the building was finished (Hamburg's, flown 2022-11-21, measured The Lyte at 21.6 m; OpenStreetMap maps it at 51.2 m). A mapped platform roof, carport or shelter over which the surface shows a level roof becomes a thin roof on posts at that height. Unnamed spikes are left alone: in Leipzig's centre dozens of them, 50–95 m tall, were the construction cranes of the flight day. A tower the 3D model already draws — a church's, a castle's — is not added a second time. On a landmark, a roof the 3D model draws flat or cuts short but the surface shows shaped is added as the measured surface itself, metre by metre and lightly smoothed: the Elbphilharmonie's crests roll over its 96 m block, and a spire the model stops at 137 m (Unna's Stadtkirche) rises to its measured tip near 178 m. And where the 3D model's roof misses this surface badly — more than 2 m off over 40 % of the roof, or over a fifth of it where the surface is level and the model slopes (a low hipped roof drawn over a flat department store) — the viewer rebuilds that building's roof from it, inside the building's own footprint, in the shape it was measured: flat where the surface is flat, and where it slopes in the forms roofs are built in: as plane faces meeting at ridges and valleys and turned the ways one roof's flanks turn, or as a barrel running straight along its axis. The measured surface itself, lightly smoothed, only on a landmark, where Wikidata vouches that the building is that unusual — the Hauptbahnhof's ribbed arched sheds; what the surface shows in no such form stays flat, in steps. Measured and smoothed, ordinary pitched roofs had looked like a crumpled blanket (An der Frauenkirche 5). Where the 3D model has the right shape and only sits at the wrong height, it stays: the Frauenkirche's modelled dome is some 5 m lower than the surface model, but it is the right dome (see LoD2, weaknesses). The dormers on the pitched roofs come from it too: wherever the surface stands 0.7 m or more over a sloping roof of the 3D model — of any size and however close to the next, compact, lower than the ridge, not along a ridge or valley the model drew a little off, and with no tree near it — the viewer adds a dormer of that size and height, facing down the roof — flat-roofed, or as a shed dormer where the surface shows its roof falling with the main roof — in the house's own colours and without a window; a row of dormers the 1 m grid runs together is parted again. A bump of a cell or two at the ridge, one standing over the ridge, or one on a flat roof is a house chimney: a slim box up to its measured top, in the house's colour. And over a shop front the street photos found, a level surface 2.6–6.5 m up that runs 1.5–8 m out of the wall and then drops away is a canopy: the GDR shop pavilions on the Hauptstraße carry their flat roof some 4 m out over the pavement. The viewer draws that slab with its deep edge in the building's colour and runs the shop glass along under it; the columns that hold it up do not show in the surface model and are not drawn. The street lamps' heights come from it as well: where the scan shows a post clear of trees and facades, its head is read, and every lamp along a street takes the median of its street's posts (or, with too few, of its kind of street) — Leipzig's main roads' lamps stand about 10 m tall, its side streets' 5 m; where the scan shows too few posts, a lamp stays 5 m. It also gives the height of every freestanding wall OpenStreetMap gives none (its top over its lower side), and of every mapped hedge the laser scan does not measure: the median along the line, anything over 3 m left out as a tree; a hedge it reads under 0.4 m is not there and is not drawn. On the Elbe, a fixed landing stage's deck lies at the height the surface measures it, and a pontoon floats as high above the water as it was measured.
StrengthsMeasured heights for every tree and roof in the city at once, in a true top-down view without the sideways lean of a photograph.
WeaknessesIt cannot tell a tree from a building or a bus; tree placement is therefore restricted to vegetation classes and excluded from roads, rails and water. On road and rail land a smooth, bare surface — a gravel heap, a shed roof — is not taken for a tree. Vegetation under 3 m is ignored; 1 m is too coarse for individual shrubs. A November scan shows deciduous trees without leaves, so their crowns are thinner in the data than in summer.
Format and downloadGeoTIFF like the DGM1, same portal page. Used offline only; the raw file is not committed and never reaches the browser.

LSC — the laser-scan point cloud

Stands forLaserscandaten — the raw point cloud the two height models are computed from.
How it is collectedThe laser flight itself; every echo is a point with position, height and intensity, classified into ground and non-ground.
Update cycleThe same as the height models (27–30 November 2024 for these tiles).
Resolution and accuracyIrregular points, several per square metre; ±0.15 m in height, ±0.30 m in position, per GeoSN.
Generally suited forEverything the grids simplify away: individual tree crowns, roof edges, wall faces, power lines.
Used here forThe height of the hedges mapped in OpenStreetMap, and trees in courtyards and gardens that the land-use map does not mark as green (unless the city's tree register already has a tree there). Of the hedges found in the scan alone, only the long, slim ones are shown (since 2026-10-21), clear of tree crowns and green in the aerial photo: about a third of all of them turned out to be the rims of tree crowns. The shrubs it finds are not shown. And the small buildings the 3D building model leaves out — garden and allotment houses, sheds, carports, container buildings, pavilions, about 6 600 on the fifteen tiles: whatever stands 2–6.5 m above the ground outside every LoD2 building, returns one echo per pulse and has a flat or evenly sloping top becomes a simple box of that footprint and height, in the same clay as the other buildings, its roof in the colour the aerial photo shows where it lies in sun; in the photo's shadow only a clean, box-like top is taken (a shadow on a courtyard looked like a shed). The flight caught the Christmas markets, which had just opened (the Striezelmarkt on 27 November 2024): nothing is taken from pedestrian areas and squares, marketplaces, building sites and car parks as OpenStreetMap maps them, and nothing the size of a van unless OpenStreetMap maps a building there. The same is done for Meißen, Grimma and Leipzig with GeoSN's scan (Meißen: 475 small buildings and 6 449 trees the canopy leaves out) for Unna with North Rhine-Westphalia's (3D-Messdaten) and for Munich with Bavaria's laser points (flown in June 2022, with the leaves on); Hamburg and Berlin have no scan read: they show the OpenStreetMap hedges only, as tall as the surface model reads them, and their trees in courtyards and gardens are found in the aerial photo instead — where it shows green and the surface model is rough as a crown is, off every building (in Hamburg, whose photo was flown in summer, that more than doubled the trees found). The rules were tuned on Dresden's flight.
StrengthsSees below 3 m and between buildings, where the height grids and the land-use map see nothing. Each point knows how bright its echo was and whether the pulse split — tall trees split it almost always, roofs almost never.
WeaknessesIts classes do not separate vegetation from buildings, cars or fences. A clipped hedge rarely splits a pulse, so for low plants the viewer leans on the greenness of the (spring) aerial photo instead; a hedge under a tree crown stays invisible. Echo brightness is not comparable between two offices' scanners (NRW records it on a 16× larger scale): every scan is scaled so that its ground echoes are as bright, in the middle, as those of Dresden's scan, on which the hedge rule was tuned — an assumption (the same asphalt and lawn reflect alike for every scanner), not yet checked against Unna's or Munich's hedges. Bavaria sorts its points differently — buildings in a class of their own, vegetation together with other "object points" — so its classes are first translated into Saxony's.
Format and downloadSaxony: LAZ per 2 km tile, large (≈380 MB for 60 million points); same portal page as the DGM1. North Rhine-Westphalia: 3D-Messdaten on opengeodata.nrw.de (hm/3dm_l_las), LAZ per 1 km tile (≈100 MB each) with the same ground / non-ground classes; the four files of a 2 km tile are merged into one. Bavaria: classified laser points on geodaten.bayern.de, LAZ per 1 km tile (≈105 MB each), merged the same way. Downloaded only on request (bun run fetch <site> --lsc).

LoD2 — the 3D building model

Stands forLevel of Detail 2 of the Digitales 3D-Stadtmodell: every building as a simple solid with its real footprint, measured height and a standardised roof shape (flat, gable, hip, mansard and others). LoD1 would be flat boxes; LoD3 would add facade detail.
How it is collectedGenerated automatically by the survey office: the building footprints come from its cadastral and landscape databases (ALKIS / ATKIS), the roof shapes and heights are fitted to the laser point cloud. Since 2021 the product can also include bridges, walls, towers, wind turbines and masts.
Update cycleA tile is regenerated when its inputs change; the statewide model was last updated in August 2025. The model of a tile is dated by its inputs: for these tiles the roofs come from the 2016 laser scan, the footprints from the 2021 or 2022 Basis-DLM and the ground from the 2016 DGM; the model itself was produced in 2023 (western pair) and 2024 (eastern pair). The 2025 dates inside the files are export dates, not survey dates.
Resolution and accuracyFootprints at cadastral precision (decimetres); roof heights at laser precision; roof shapes are the nearest standard type, not the real roof.
Generally suited forCityscape visualisation, shadow and solar studies, noise and wind modelling, counting storeys and volumes.
Used here forEvery building's footprint, height, roof shape and attributes; the minimap footprints; the demolish tool; and what the card an asked building opens says about it — its identifier, official use, roof form and pitch, measured height and ground area, and the edition of the model. The small structures it lacks come from the laser scan (see LSC); chimneys, towers, masts and buildings newer than the model from the surface model where OpenStreetMap names them (see DOM1). Its canopies and carports, which it models as solid blocks (Hamburg, Munich and Unna have them), are opened up into a roof on posts.
StrengthsExact silhouettes: footprints and roof shapes are consistent citywide, heights are measured.
WeaknessesNothing below the eaves: no windows, doors, materials or colours (the viewer draws a door where OpenStreetMap maps an entrance, opens a passage through the building where it maps one — a gateway into a courtyard, the Zwinger's gates —, and windows in the rhythm street photos measured, a neighbour's or the house type's). The building-function attribute is "unspecified" for about 86 % of buildings and the storey count is filled in for only a few percent, so the viewer takes the storeys from OpenStreetMap where they are mapped and otherwise derives the storey bands from the measured height. Bridges are in the tiles as flat, 1 m thick slabs at one height (function 53001_1800, "structure in the traffic area") — the viewer leaves them out and builds bridges from the landscape and surface models. A building too complex for the catalogue of roof shapes gets a "free-form" roof, and on a large complex that is often a few huge sloping faces over the whole footprint: the Westin Bellevue hotel on the Elbe, a flat-roofed slab with lower wings and two courtyards, stood under one tent-like roof rising to a peak in the middle. Buildings finished after the roofs were measured stand as 3 m placeholders. The viewer measures every roof against the surface model (DOM1) and rebuilds the 1 080 that miss it badly — about 2 % — in the shape the surface model shows: flat parts at their measured heights, sloping ones as plane faces or a barrel; the measured surface itself only on a landmark.
Format and downloadCityGML per 2 km tile (also DXF and Shape); the project converts to CityJSON, a compact JSON form of the same standard, before committing (8–11 MB per tile). Digitale 3D-Stadtmodelle.

Basis-DLM — the landscape model

Stands forDigitales Basis-Landschaftsmodell: the most detailed level of ATKIS, the Amtliches Topographisch-Kartographisches Informationssystem, the nationwide system of official topographic data. A vector map of what covers the ground — roads, paths, railways, water, forest, farmland, settlements — plus line features such as hedges and tree rows, with attributes.
How it is collectedMaintained by the survey office from current orthophotos, the terrain model, local measurements and data from other authorities (road and rail operators, municipalities). Object types and attributes follow a nationwide catalogue (GeoInfoDok / AAA schema 7.1.2 since 2025), so they are identical in every German state.
Update cycleTwo rhythms, per GeoSN: a basic update in which every object is checked every 3–5 years, and a priority update in which important objects (roads, railways and the like) are checked every 3, 6 or 12 months. The download package is refreshed quarterly.
Resolution and accuracyPositional accuracy ±3 m for the essential line objects (roads, rails, rivers) and ±15 m for everything else, per GeoSN. Roads are centrelines with a width attribute, not surfaces.
Generally suited forA consistent, attributed base layer for GIS: joining specialist data to it, routing and navigation, land-use statistics, cartography at scales of about 1:10 000 to 1:25 000.
Used here forThe ground colours (ten land-cover classes, fields among them), the water outline, hedge and tree rows, the vegetation mask that gates tree placement, railway areas, the track lines where OpenStreetMap maps no track, and whether a line is electrified where OpenStreetMap does not say, bridge centrelines and, where present, deck outlines; the monuments, memorial stones, columns and named fountains (the Albertplatz fountains "Stilles Wasser" and "Stürmische Wogen", the Neptunbrunnen, …) with their official names.
StrengthsAuthoritative classification with useful attributes: road widths, number of tracks, electrification, bridge names.
WeaknessesAnything small or exact: roads must be widened by rule, railway lines and road axes come in short fragments that need merging and say nothing of running underground — the viewer drops a stretch that runs along a mapped tunnel for more than 15 m (Munich's U-Bahn and S-Bahn under the Marienplatz, Leipzig's City-Tunnel, the road through Meißen's Schottenbergtunnel), while a track or road that only crosses over one stays —, deck outlines exist mostly for major bridges, there is no street furniture and no platforms. Monuments are points with a name only: no size, no shape, and nothing says which of them is a fountain — the viewer takes a monument's form from the surface model where it can be measured (an abstract marker where not), and a fountain's basin from OpenStreetMap. At ±3 m a road edge can sit a lane off.
Format and downloadStatewide package in Shape (or NAS or GeoPackage), about 1.2 GB; the project clips each tile out of it. Basis-DLM.

DOP20 RGBI — the aerial photos

Stands forDigitales Orthophoto, 20 cm ground resolution, channels Red, Green, Blue and near-Infrared. An orthophoto is an aerial photograph rectified onto the terrain model so that every pixel sits at its true map position and distances can be measured as on a map.
How it is collectedA survey aircraft photographs strips of the state with a calibrated camera (80 % forward and 60 % side overlap since 2021); the images are oriented, rectified over the DGM and mosaicked into 2 km tiles.
Update cycleAbout half of the state every year, alternating spring and summer flights, i.e. every tile roughly every two years, per GeoSN. These tiles were flown on 19 March 2024 — a leaf-off spring flight.
Resolution and accuracy20 cm per pixel; a pixel's standard deviation from its true position ≤ 0.4 m; 8 bits per channel; 10 000 × 10 000 pixels per tile, per GeoSN.
Generally suited forThe base image for mapping and GIS capture; environment, agriculture and forestry monitoring (the infrared channel separates vegetation from everything else); transport and urban planning.
Used here forThe colour of each roof, read from the parts of it in sun (a robust median, avoiding the edge). Every city's aerial photos come with their own colour cast — Leipzig's are cyan, Munich's yellow — so each city's photo is first balanced on its own sunlit roads to Dresden's, and a grey roof is grey in every city. Also the greenness index NDVI from the red and infrared channels, which tints tree crowns from dry pale sage to lush deep green and does the same for meadows, and turns a field green where a crop grows. The rows of a vineyard run the way the photo shows them. It also says what a sports ground OpenStreetMap does not describe is laid with: grass where the infrared shows something growing, artificial turf where it is green and nothing grows, the red of tartan, the orange of clay or cinder, or a grey or blue hard court — in Dresden only one of the twelve untagged football pitches is grass.
StrengthsReal colours and real greenness where the photo looks straight down.
WeaknessesFacades are invisible (a nadir photo sees roofs only); tall buildings lean sideways in the image, so roof footprints are eroded inward before sampling; the raw colours read drab and hazy, which the viewer corrects with a hue-preserving saturation lift. Because the flight was leaf-off, deciduous trees are bare in the image: the greenness index is low almost everywhere and the crown colouring recentres it on the median rather than reading it as an absolute value.
Format and downloadGeoTIFF per 2 km tile, about 380 MB for the 4-channel variant. DOP download area. Only derivatives are committed (roof colours, NDVI raster).

NDVI — the greenness index (derived, not downloaded)

Stands forNormalized Difference Vegetation Index: (near-infrared − red) / (near-infrared + red), computed per pixel from the DOP.
Why it worksHealthy leaves reflect near-infrared strongly and absorb red, so the index is high on lush vegetation and near zero on roofs, roads and water.
Used here forTree crown colour and the meadow tint; baked into a 1024² raster per tile (about 2 m per pixel).
CaveatOnly as good as the photo's date: a March image separates evergreens and grass from everything else but says little about summer foliage. Where the aerial photo has no infrared band (Bavaria's, for Munich), a stand-in from the visible colours takes its place — the Green Leaf Index, (2·green − red − blue) / (2·green + red + blue), scaled to match the NDVI where both exist (on North Rhine-Westphalia's summer photos it agrees with the NDVI for 85 % of the pixels on whether a pixel is green). It tells green from grey well, but how vigorous a plant is less well. And every flight has its own brightness per channel — the same asphalt reads −0.30 in Dresden and −0.07 in Hamburg —, so each city's index is put onto one scale by two surfaces it shows everywhere: its carriageways and its mown lawns read as Dresden's do, so "green" means the same in every city.

OSM — OpenStreetMap

Stands forOpenStreetMap, the free world map built by volunteers since 2004.
How it is collectedContributors map from GPS traces, on-the-ground surveys and by tracing aerial imagery (including official orthophotos where their licence allows), and describe each object with free-form key=value tags such as highway=street_lamp or barrier=retaining_wall + height=9.
Update cycleContinuous: edits are live within minutes. Extracts for download (Geofabrik) are rebuilt daily; the project reads such an extract, not the live database.
Resolution and accuracyNo guarantee; in a well-mapped city typically metre-level positions. Completeness and tag consistency vary from street to street and mapper to mapper.
Generally suited forThings no official dataset has: street furniture, points of interest, names, informal paths, structure types; near-worldwide coverage; quick to fetch.
Used here forLamp positions (highway=street_lamp), street furniture — benches (amenity=bench, with backrest and direction where tagged), picnic tables, litter bins, bicycle stands (with capacity), bollards, post boxes and stop shelters (shelter=yes) — a bus stop without one gets its "H" sign —, advertising columns (advertising=column), traffic signals (highway=traffic_signals, moved from the stop line in the road to the kerb on the side of the traffic they face, traffic_signals:direction), fire hydrants (emergency=fire_hydrant: a pillar, or for an underground one its small sign plate), clocks (amenity=clock on a pole or on a wall), drinking fountains and flagpoles (man_made=flagpole, with height and the flag it flies where flag:name, subject, country or their Wikidata items name one the viewer draws — Germany, the EU, Saxony, Bavaria, Hamburg; country only on a national flag that names no item of its own, flag:type=european always Europe's —, else a pale cloth), playground outlines (leisure=playground) with the equipment mapped on them (playground=swing, slide, sandpit, …) — turned towards the nearest street or path where no direction is tagged, station platforms (railway=platform), hedges (barrier=hedge), retaining walls, city walls, embankments and cliff edges (barrier=*, man_made=embankment, natural=cliff) with their height tag and material (material, wall=*), flights of steps (highway=steps with width, step_count, surface and handrail, the width else from an area:highway=steps outline), whether a bridge is an arch bridge (bridge:structure, where Wikidata does not know the bridge), the Elbe bridges' navigation clearance from the inland-waterway marks (seamark:bridge:clearance_height: it gives the deck's structural depth, and no pier stands in the fairway), and fountains (amenity=fountain): the outline of each basin, whether it is a splash pad or a still pool, and the many small fountains the landscape model does not list. Where an official monument stands in an OSM basin, the fountain keeps the official name. Memorials and artworks (memorial=*, artwork_type=*): whether a monument is a statue on a pedestal, a bust, a free sculpture, a stele, a stone or an obelisk — the official list does not say — with its artist and material for the card and its colour (a muted tone of bronze, sandstone, granite …), and the park sculptures the official list leaves out. Also what a street, walkway or car park is paved with (surface=asphalt, paving_stones, sett, … on the ways, sidewalk:*:surface on the roads) and which way it runs, so slabs and cobbles lie along the street; and where cars park (parking:left/right/both with its orientation on the roads, amenity=parking car parks with their aisles, mapped bays), drawn as bays — with painted lines only where OpenStreetMap says they are marked (parking:<side>:markings), and in a car park unless it says they are not or the car park is unpaved; and where a kerb is lowered (highway=crossing, `kerb=lowered
StrengthsHuman-readable tags for exactly the details the survey office does not model; the Brühlsche Terrasse exists here and nowhere else.
WeaknessesNot every lamp or bench is mapped, and few benches say which way they face, heights are often missing (the viewer measures an untagged wall in the surface model, else uses a default per wall type), tags vary. Volunteer data must be credited (ODbL).
Download and licenceOne regional extract of the whole state, sachsen-latest.osm.pbf, downloaded from Geofabrik (about 250 MB) and read locally, which avoids rate limits and makes the result reproducible. The lamp, platform and bridge-structure files committed today are older: they were fetched through the Overpass API, a live query service, before the bakes switched to the extract, and move to the extract at their next re-bake. Licence: ODbL, credit "© OpenStreetMap contributors".

Verkehrsmengen — the counted motor traffic

Stands forThe road office's Verkehrsbelegung: motor vehicles per day (AADT, the annual average daily traffic) for every counted road section between two junctions, per direction, with the heavy goods vehicles.
How it is madeMostly counted by hand on one day and scaled to the average day; on some roads induction loops or infrared detectors giving yearly means; on a few an estimated "Hilfswert".
UpdatesSection by section, when counted again; in the site mostly 2023–2026, a few sections back to 2010.
Used here forThe Kfz-Verkehr data layer: one glass flow per counted direction on the carriageway, on the right of travel, as wide and as tall as the root of the vehicles per day, tinted in five steps from sage through peach, coral and rose to wine, with light running through it in the direction of travel; a high heavy-goods share turns it slate blue. The flows follow the time set: the day's count is spread over the hours by the daily curve (below) — slim and nearly dark at night, full and slow at the rush hours. The street behind shows through the glass, gently bent; from the air it thickens so the layer still reads as a map. A street without a count stays empty — nothing is estimated.
StrengthsMeasured, per direction; nearly every main road counted.
WeaknessesA daily mean, no curve of its own over the day (the daily curve supplies one, the same for every street); counts from different years side by side; minor streets often uncounted.
Download and licenceThe city's WFS (kommisdd.dresden.de, layer cls:L363 "Kfz/Tag"), read per tile. Licence dl-de/by-2-0, credit "Landeshauptstadt Dresden".

Tagesgang — how traffic spreads over the day

Stands forWhich share of a day's vehicles drives in which hour, for working days, Saturdays and Sundays.
How it is madeMeasured by Hamburg's infrared detectors on 38 inner-city main roads, hourly, in September 2026: each station's complete days normalised to one and averaged, then the stations. Dresden publishes daily values only; the shape of a German city's main roads stands in for its hours.
Used here forThe time of day of the Kfz-Verkehr data layer: a flow's colour, size and light follow its street's daily count times the share of the hour set. The sidebar names the hour and how busy it is.
StrengthsMeasured, urban, with working day, Saturday and Sunday.
WeaknessesOne curve for every street, from another city; public holidays count as working days; one month, no seasons.
Download and licenceHamburg's Urban Data Platform (SensorThings, iot.hamburg.de, HH_STA_Verkehrsdaten_Kfz_Infrarotdetektoren), evaluated once; the values live in lib/city/traffic-hours.ts. Licence dl-de/by-2-0, credit "Freie und Hansestadt Hamburg".

The other cities' traffic counts

Stands forMotor vehicles per day and road section, both directions together: in Berlin the Verkehrsmengen 2023 (working-day mean, main road network), in Hamburg the main roads' Verkehrsmengen 2019 (rounded to the thousand, with the heavy-goods share), for Grimma and Meißen Saxony's road census 2021, for Unna Straßen.NRW's Verkehrswerte.
How it is madeThe cities scale counts onto their network; the Länder's road censuses count every five years at fixed points of the federal, state and district roads and scale to the average day of the year.
Used here forThose cities' Kfz-Verkehr data layer, drawn as in Dresden; as only the sum of both directions is known, each direction gets half — the sidebar says so.
StrengthsOfficial and open; the same picture everywhere.
WeaknessesNo directions; one-way streets are not marked; the Länder's censuses do not reach the city centres (no counted road lies on Leipzig's or Munich's tiles, so the layer is not offered there).
Download and licenceBerlin: WFS gdi.berlin.de/services/wfs/verkehrsmengen_2023, dl-de/zero-2-0. Hamburg: WFS geodienste.hamburg.de/HH_WFS_Verkehrsmengen, dl-de/by-2-0, "Freie und Hansestadt Hamburg, Behörde für Verkehr und Mobilitätswende". Saxony: list.smwa.sachsen.de/gdi/download/DE-SN-SBV-SVZ2021.zip, dl-de/by-2-0, "Freistaat Sachsen, LASuV". NRW: WFS wfs.nrw.de/wfs/strassen_nrw (ms:Verkehrswerte), dl-de/by-2-0, "Straßen.NRW".

Hamburg's cycle counting network — bicycles, live

Stands forThe infrared counters of Hamburg's cycle counting network, per junction arm and direction — 21 on Hamburg's tiles.
UpdatesHourly.
Used here forThe Radverkehr (live) data layer in Hamburg, as in Dresden: the browser reads the last hour's values straight from the city.
Download and licenceThe Urban Data Platform (SensorThings, iot.hamburg.de, HH_STA_Verkehrsdaten_Rad_Infrarotdetektoren). Licence dl-de/by-2-0, credit "Freie und Hansestadt Hamburg". Leipzig (daily), Munich (monthly) and Berlin (yearly) do not publish their cycle counts live — the layer is not offered there.

Rad-Dauerzählstellen — bicycles, live

Stands forThe city's automatic bicycle counters — 35 in the site, on the Albert and Waldschlößchen bridges and along the Elbe cycle path among others.
How it is madeSensors in the cycle path count each bicycle and its direction and report every hour's sum.
UpdatesHourly, a few minutes after the hour.
Used here forThe Radverkehr (live) data layer: the browser reads the counts from the city when it is switched on and every five minutes after. Each counter is two glass columns either side of the street, one per direction (teal, lilac), as tall as the root of the last hour's bicycles, with rings of light rising faster the more bicycles passed; a counter silent for three hours turns grey. The sidebar lists the numbers; a click flies there.
StrengthsCurrent, measured, per direction.
WeaknessesFew points; says nothing about the streets in between. When the city's service is down the last counts stay.
Download and licenceThe same WFS, layer cls:L1781 ("aktuelle Zählwerte"; the whole hourly series since 2017 is cls:L1780). Licence dl-de/by-2-0, credit "Landeshauptstadt Dresden".

GTFS timetable — the trams

Stands forGeneral Transit Feed Specification, the common exchange format for timetables: lines, stops, trips and their times.
How it is madeThe operators plan their trips; DELFI e.V. collects the timetables of every German transport association (as NeTEx), and gtfs.de converts them to GTFS every week.
UpdatesWeekly, each about a month ahead.
Used here forThe Straßenbahnen (Fahrplan) data layer: every tram trip in the site (Dresden's DVB — about 2,600 on a working day —, Leipzig's LVB, Munich's MVG) laid onto the OpenStreetMap tracks (the timetable knows the stops, not the way between them). At the scene's time, which runs on in real time from the time set, each tram is a car in its operator's colours and of one of its fleet's lengths (Dresden's yellow, Leipzig's silver-grey with a yellow roof band, Munich's blue and white), trailing light along its track, that stands at its stops and crosses the bridges. The viewer tells working days, Saturdays and Sundays apart; a public holiday runs as its weekday.
StrengthsComplete: every planned trip of every line.
WeaknessesNot live — delays, diversions and cancellations are not shown. In a few places the way between two stops lies on the wrong one of two tracks.
Download and licencenv_free from gtfs.de (all of Germany, about 290 MB, not committed); only the cut for each site, data/<site>/transit/trams.json with the lines' operators from the feed's agency.txt, is. Licence CC BY 4.0, credit "DELFI e.V. via gtfs.de"; the tracks "© OpenStreetMap contributors".

Dataset editions in use

The exact edition matters when the picture disagrees with reality. GeoSN publishes, for every tile and product, a currency field ("Stand") through the download service behind its portal; the values below were read from it on 2026-09-22 and match the metadata files shipped in the tile ZIPs. The machine-readable version, with the download link of every file, is data/dresden/provenance.json. The editions below are Dresden's; another site's are read from the same kind of service when its data is fetched. "Committed" is the date a file entered the repository; the download happened on or shortly before it.

DatasetTilesEdition / survey date (provider's "Stand")How we knowCommitted
DGM1all fifteen2024-11-30; the northern row (5658) 2024-11-27 and 2024-11-30the _akt.csv in each tile ZIP; GeoSN download service2026-06-11; the eleven tiles south, east and west 2026-09-25
DOM1all fifteenthe same laser flight as the DGM1: 2024-11-30 / 2024-11-27 and 2024-11-30GeoSN download service (DOM1, DGM1 and the point cloud carry identical dates)derived canopy files 2026-06-12; the eleven new tiles 2026-09-25
LoD2all fifteenthe western column (33408_*) and 33410_5656/5658: model 2023, built from the 2016 laser scan, the 2021 Basis-DLM footprints and the 2016 DGM; 33412_5656/5658 and 33414_5656/5658: model 2024, from the 2016 laser scan, the 2022 Basis-DLM and the 2016 DGM; the southern row (5654) and 33416_5656: model 2023, footprints from the 2020–2024 Basis-DLM; 33416_5658 (forest, one building): model 2021. The objects were exported 2025-04-26 … 2025-07-07 (creationDate)GeoSN download service; a few older objects still carry Stand_* attributes with the same values2026-06-11; the eleven new tiles 2026-09-25
DOP (RGBI)all fifteenflown 2024-03-19 (leaf-off)GeoSN download servicederived roof colours and NDVI 2026-06-16/17; the eleven new tiles 2026-09-25
Basis-DLMstatewide packagethe package released 2026-11-04 (basisdlm_akt.csv and the package inside the ZIP are dated that day)the ZIP's metadata; data/dresden/provenance.jsonevery derived file re-baked 2026-11-05 (from the 2026-10-20 package before)
OSM via BBBikethe Dresden city extractthe extract of 2026-10-31 (Last-Modified 16:49 UTC), every OSM layer of all fifteen tiles; only the hedges and the scan's small buildings still come from the extracts of 2026-09-19/23, because their re-bake needs the laser scan GeoSN does not serve at the momentthe file's Last-Modified; data/dresden/provenance.json2026-11-01
Verkehrsmengenall fifteencounts from 2010 to 2026, mostly 2023–2026, read on 2026-10-01the city's WFS; data/dresden/provenance.json2026-10-01
Rad-Dauerzählstellen—live, when the layer is switched on and every five minutesthe city's WFSnot committed
GTFS timetablethe whole sitethe feed of 2026-09-26; the days 2026-10-01 (working day), 2026-10-10 (Saturday), 2026-10-04 (Sunday)the file's Last-Modified; data/dresden/transit/trams.json2026-10-01
Mapillary objects (street lamps)Dresden (fifteen tiles), Grimma (two), Meißen (one), Unna (two)detections last seen 2020 – 2026each object's last_seen_at, each photo's place and time; the cache under data/_raw/<provider>/mapillary/<tile>.v2.json, fetched again 2026-11-052026-11-05
Mapillary panoramas (facade readings, shop windows)Dresden (fifteen tiles), Leipzig, München, Hamburg (four each), Grimma (two), Meißen (one); Unna has nonetaken 2017 – 2026: Dresden mostly 2025, Leipzig 2023 – 2025, München 2021 – 2026, mostly 2026, Hamburg 2025, Grimma 2024 – 2026, Meißen 2026each image's captured_at; the measurements under data/_raw/<provider>/mapillary/facades/ (the photos are not kept)2026-10-08 (Dresden), 2026-10-09
Mapillary photos of every camera (glass facades)Dresden (fifteen tiles)taken 2016 – 2026, mostly 2019 and 2025 (about 38 000 photos read)each image's captured_at; the readings under data/_raw/<provider>/mapillary/facades/ (the photos are not kept)2026-11-10
Copernicus DEM GLO-30every site: its tiles and 10 km around themthe model as its public cloud copy serves it on the fetch day (© DLR e.V. 2010-2014, © Airbus Defence and Space GmbH 2014-2018)the window cached under data/_raw/glo30/<site>.tif, read 2026-10-21not committed (only the sky view and horizon baked from it)
DWD phenologythe observers within 30 km of each citythe years 2010–2023the viewer's season tableas part of the code

Note the mismatch of dates inside one picture: the ground and the tree heights are from late 2024, the building shapes from a 2016 laser scan with 2021/2022 footprints, the roof colours from March 2024, and the lamps and walls from late-October-2026 OpenStreetMap. A building finished in 2023 can have a 2024 roof colour and no 3D shape.

The other sites took the same re-bake on 2026-11-01: Leipzig and Grimma from BBBike's Leipzig extract, Meißen from its Dresden extract, Hamburg and München from their own (all 2026-10-31), Unna from openstreetmap.fr's Arnsberg extract (2026-11-01); the Saxon sites from the Basis-DLM of 2026-10-20, Unna from NRW's of 2026-10-20 and München from Bavaria's of 2026-07-03. On 2026-11-05 every site was re-baked once more, from the same OSM extracts (Unna from the Arnsberg extract of 2026-11-05) and the Saxon sites from the Basis-DLM of 2026-11-04; only the Saxon sites' hedges and small buildings from the laser scan and the facade readings stayed from the bake before. Each site's provenance.json records it.

Where exactly each file came from

GeoSN serves every tile ZIP from public folders on its cloud share; the portal's download app finds them through a map service that also carries the "Stand" of each tile. The folders are per product and format:

ProductPackagePortal page
DGM1 (GeoTIFF + .tfw + _akt.csv)…/JCcXyifaNdLDnxZ/dgm1_<tile>_tiff.zipDigitale Höhenmodelle
DOM1 (GeoTIFF)…/S6wwnFwX7882sZm/dom1_<tile>_tiff.zipsame page
Laser-scan point cloud (LAZ)…/EpkzyJHScGb5ndd/lsc_<tile>_laz.zipsame page
LoD2 (CityGML)…/GVzwbSyp7Yl7mBD/lod2_<tile>_citygml.zipDigitale 3D-Stadtmodelle
DOP20 RGBI (GeoTIFF)…/sX3GPcdBMGrfXT9/dop20rgbi_<tile>_tiff.zipDOP
Basis-DLM (Shape, statewide, 1.23 GB)…/DtPWngtLEJP8K3k/basisdlm_sn_shape.zipBasis-DLM

… stands for https://geocloud.landesvermessung.sachsen.de/public.php/dav/files/. The folder tokens can rotate; the download service described in data-pipeline.md lists the "Stand" for any tile, and GeoSN's batch-download page carries the current tokens. bun run fetch dresden reads those tokens from the batch page and fetches every product of each tile — terrain, surface model, building model (converted to CityJSON on the way), aerial photo — plus the statewide Basis-DLM package and the OpenStreetMap extract. Dresden's terrain and building models are committed as they were first downloaded. OpenStreetMap data now comes only from the Geofabrik Saxony extract (sachsen-latest.osm.pbf); the committed lamp, platform and bridge-structure files still date from earlier Overpass API queries.

Licences and credits

SourceLicenceRequired credit
GeoSN datasets (DGM1, DOM1, LoD2, Basis-DLM, DOP)Datenlizenz Deutschland – Namensnennung – Version 2.0 (dl-de/by-2-0), per GeoSN's terms of use (checked 2026-09-22)"Quelle: GeoSN, dl-de/by-2-0"
Geobasis NRW (Unna; including its laser scan)Datenlizenz Deutschland – Zero – Version 2.0 (dl-de/zero-2-0): no credit requiredgiven anyway: "Geobasis NRW, dl-de/zero-2-0"
Bavarian survey administration (Munich; including its laser points)CC BY 4.0"Bayerische Vermessungsverwaltung – www.geodaten.bayern.de, CC BY 4.0"
LGV Hamburgdl-de/by-2-0"Freie und Hansestadt Hamburg, Landesbetrieb Geoinformation und Vermessung (LGV), dl-de/by-2-0"
Geoportal Berlindl-de/zero-2-0: no credit requiredgiven anyway: "Geoportal Berlin, dl-de/zero-2-0"
OpenStreetMapOpen Database License (ODbL)"© OpenStreetMap contributors"
WikidataCC0 (public domain)none required
Wikimedia Commons (the photos the murals are drawn after)each photo's own free licence (CC BY, CC BY-SA …), read from Commons with its author; the drawing is under the sameper photo: "Wandbild „…“: gezeichnet nach Foto author, licence, Wikimedia Commons"
Wikimedia Commons (the walks' photos)each file's licence as its Commons page states (mostly CC BY-SA or public domain)author, licence and link on the stop's card
Copernicus DEM GLO-30 (the land beyond the tiles)Licence for the use of the Copernicus WorldDEM-30: free of charge for any use; a modified product carries the credit, and its legal notice the liability sentence"Gelände im Umland: produced using Copernicus WorldDEM-30 © DLR e.V. 2010-2014 and © Airbus Defence and Space GmbH 2014-2018 provided under COPERNICUS by the European Union and ESA; all rights reserved"; and: "The organisations in charge of the Copernicus programme by law or by delegation do not incur any liability for any use of the Copernicus WorldDEM-30."
DWD phenology (the annual reporters' observations of wild plants, 2010–2023)dl-de/by-2-0"Quelle: Deutscher Wetterdienst"
Stadtbaumkatasterdl-de/by-2-0"Landeshauptstadt Dresden"
Straßenbaumkataster Hamburgdl-de/by-2-0"Freie und Hansestadt Hamburg (BUKEA)"
Baumkataster Leipzigdl-de/by-2-0"Stadt Leipzig, Amt für Stadtgrün und Gewässer"
Baumbestand Berlindl-de/zero-2-0: no credit requiredgiven anyway: "Geoportal Berlin"
Verkehrsmengen, Rad-Dauerzählstellendl-de/by-2-0"Landeshauptstadt Dresden"
Daily curve (Hamburg's traffic counters)dl-de/by-2-0"Freie und Hansestadt Hamburg"
Hamburg's traffic counts and cycle countersdl-de/by-2-0"Freie und Hansestadt Hamburg"
Berlin's traffic countsdl-de/zero-2-0: no credit requiredgiven anyway: "SenMVKU Berlin"
Saxony's road censusdl-de/by-2-0"Freistaat Sachsen, LASuV"
NRW's Verkehrswertedl-de/by-2-0"Straßen.NRW"
GTFS timetableCC BY 4.0"DELFI e.V. via gtfs.de"
Mapillary (detected street lamps, facade readings, shop windows, window rhythm)CC BY-SA 4.0"Mapillary" — kept in its own file, apart from the OpenStreetMap data

The viewer shows the credit of the site's provider, the OSM credit, the Copernicus credit and, where there is one, the city's tree cadastre in the footer of its settings panel (and Mapillary's, and the data layers', where the city takes them, and each mural photo's); the baked sky view and horizon files carry the Copernicus credit inside as well. The weather service's observations reach the viewer only as a table of a few days' shift per city, so its credit stands here. The derived lamp and wall files carry the OSM credit inside the file as well; the monument file carries both credits, because it combines the two, and the street-tree file carries the city's and, for the trees it adds from OpenStreetMap, the OSM credit.

Previous
How the city walker works
Next
Sources by city
Source on GitHub: docs/guide/en/data-sources.mdDeutsche Fassung →
On this page
  • The providers
  • The datasets at a glance
  • How the official datasets are produced
  • Dataset by dataset
  • DGM1 — the terrain model
  • DOM1 — the surface model
  • LSC — the laser-scan point cloud
  • LoD2 — the 3D building model
  • Basis-DLM — the landscape model
  • DOP20 RGBI — the aerial photos
  • NDVI — the greenness index (derived, not downloaded)
  • OSM — OpenStreetMap
  • Verkehrsmengen — the counted motor traffic
  • Tagesgang — how traffic spreads over the day
  • The other cities' traffic counts
  • Hamburg's cycle counting network — bicycles, live
  • Rad-Dauerzählstellen — bicycles, live
  • GTFS timetable — the trams
  • Dataset editions in use
  • Where exactly each file came from
  • Licences and credits