The underside of planetary maps

Le dessous des

cartes planétaires

Benoît Seignovert

Unité d'Appui à la Recherche de l'OSUNA (UARO)

Séminaires LPG | Nantes | 2026-03-19

slid.es/seignovert/lpg-map

doi: pending…
NASA/NOAA/R. Stöckli
Creative Commons “Attribution-Share Alike” license icon

LPGMAP

wooclap.com

References:

- Twitter 4 PB / yr | 500 PB total (1, 2) [2024]

- Netflix 3 PB (1) [2013] - 20 GB x 15,000 movies -> 300 TB (2)

- Google Search 62 PB [2021] (1)

- Facebook 250 PB/yr (1)

- Youtube 250 PB/yr (1)

- Google photos 2e12 pictures in 2025 (7) - 4e12 pictures in 2020 (wiki)

- What app 7e9/day -> 2e12/yr pictures in 2023 (7)

- Number of smartphone in the world in 2025 : 7.4e9 (8) with 2,000 pictures on average (7) -> 15e12 pictures

Building maps

Sumerian map | -2500

Tabula Rogeriana | 1154

a multi-millennial story

70 sheets

Global vs. local

Mercator map

1569

18 sheets

1.2 m x 60 cm

Cassini map

1756 - 1815

(4 generations!)

181 sheets at 1:86 400

11 m x 11 m

1 km

Print

130 000 x 130 000 pixels

17 000 Mpix ~ 50 GB

Screen

32 000 x 32 000 pixels

1 000 Mpix ~ 3 GB

geoportail.gouv.fr/carte

IGN map

2011 - 2025

1 200 maps at 1:25 000

38 m x 38 m

300 m

Print

450 000 x 450 000 pixels

200 000 Mpix ~ 0.5 TB

Screen

110 000 x 110 000 pixels

12 000 Mpix ~ 33 GB

geoportail.gouv.fr/carte

Mercator projection

Global vs. local - as of today

1:1 000

40 km x 20 km

Print

450 000 000 x 225 000 000 pixels

100 000 000 000 Mpix ~ 300 PB

Screen

110 000 000 x 55 000 000 pixels

6 000 000 000 Mpix ~ 18 PB

4 000 km

Open Street Map
Level 0

Nantes

10 m

Open Street Map
Level 19

1:1 000

40 km x 20 km

Print

450 000 000 x 225 000 000 pixels

100 000 000 000 Mpix ~ 300 PB

Screen

110 000 000 x 55 000 000 pixels

6 000 000 000 Mpix ~ 18 PB

Global vs. local - as of today

How to store it?

 

🗺️ 🤔

and…

how to do this!

Wikipedia / S. Viinamäki

Building different maps at different scales

IGN / Géoportail

1 : 10,000

1 : 100,000

1 : 1,000,000

1 : 10,000,000

1 cm = 100 km

1 cm = 10 km

1 cm = 1 km

1 cm = 100 m

Slice the map

in tiles

Tile size:

Number of tiles:

Tiles indexing:

/Z/X/Y.[jpg|png]
N(Z) = 22 x Z
256 x 256 pixels
Z=
Z=
Z=
Z=
N=1
N=4
N=16
N=22n
200 ko

WMTS (Web Map Tile Service)

Client side

1 x 256 x 256 pix
200 ko

Server side

256 x 256 pix
200 ko
256 x 256 pix
200 ko
256 x 256 pix
200 ko
256 x 256 pix
200 ko
256 x 256 pix
200 ko
256 x 256 pix
200 ko
1.2 Mo
0.5%
4 x 256 x 256 pix
800 ko
16 x 256 x 256 pix
3.2 Mo
64 x 256 x 256 pix
13 Mo
256 x 256 x 256 pix
51 Mo
1024 x 256 x 256 pix
324 Mo
+26 %

256 Mo
(Z, X, Y)

(0, 0)

(0, 1)

(1, 0)

(1, 1)

(0, 0)

(0, 1)

(0, 2)

(0, 3)

(1, 0)

(1, 1)

(1, 2)

(1, 3)

(2, 0)

(2, 1)

(2, 2)

(2, 3)

(3, 0)

(3, 1)

(3, 2)

(3, 3)

<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>Map tiles demo</title>
     <link rel="stylesheet" href="https://unpkg.com/leaflet@1.9.4/dist/leaflet.css"/>
</head>
<body>
    <div id="map" style="height: 800px"></div>
    <script src="https://unpkg.com/leaflet@1.9.4/dist/leaflet.js"></script>
    <script>
        const map = L.map('map').setView([47.2386, -1.5554], 0);

        const tiles = L.tileLayer('https://tile.openstreetmap.org/{z}/{x}/{y}.png', {
            maxZoom: 19,
            attribution: '© <a href="http://www.openstreetmap.org/copyright">OpenStreetMap</a>'
        }).addTo(map);
    </script>
  </body>
</html>

How to use tiles on the web?

gitlab.univ-nantes.fr/osuna/demo/tiles

Looking into the network traffic

🔎

0 → 19

= 3 MB

≠ 18 PB

Mapping tiles providers

Apple Maps

Open Street Map

maps.obs-nantes.fr

<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>Map tiles demo</title>
     <link rel="stylesheet" href="https://unpkg.com/leaflet@1.9.4/dist/leaflet.css"/>
</head>
<body>
    <div id="map" style="height: 800px"></div>
    <script src="https://unpkg.com/leaflet@1.9.4/dist/leaflet.js"></script>
    <script>
        const map = L.map('map', {crs: L.CRS.EPSG4326}).setView([-4.8099, 137.3818], 0);

        const tiles = L.tileLayer('https://maps.obs-nantes.fr/wmts/mars/tianwen-1/{z}/{x}/{y}.jpg', {
            maxZoom: 10,
            attribution: '🌍 <a href="https://maps.obs-nantes.fr">Osuna</a> | 🇨🇳 <a href="https://clpds.bao.ac.cn/mall/MarsDATA">Tianwen-1</a>'
        }).addTo(map);
    </script>
  </body>
</html>

Changing the tiles provider

gitlab.univ-nantes.fr/osuna/demo/tiles

Use the tiles in a GIS software 🔨

<!DOCTYPE connections>
<qgsXYZTilesConnections version="1.0">
    <xyztiles
        name="Osuna | Mars | Tianwen 1"
        url="https://maps.obs-nantes.fr/wmts/mars/tianwen-1/{z}/{x}/{y}.jpg"
        zmin="0"
        zmax="10"
    />
</qgsXYZTilesConnections>

Render the tiles in 3D

Made by Simeon Schmauß in metashape

Interactive globe

Add temporal layers ⏳

/{t}/{z}/{x}/{y}.fmt

2025

2002

1943

Strasbourg
Google Earth

Add multi/hyper-spectral 🌈

Eumetsat
Meteosat FCI
/{w}/{z}/{x}/{y}.fmt

0.6 µm

10.5 µm

But…

there is a little catch 🫢

N=22z

💥

Tiling may not be the best strategy…

Level # tiles Pixel size ~ Scale Coverage # folder / files Size Serveur
0 1 160 km 1:500 000 000 Global 1 / 1 / 1 200 kB 😅
1
 4 80 km 1:250 000 000 2 / 2 / 2 0.8 MB 🌈
2 16 40 km 1:150 000 000 Continent 3 / 4 / 4 3.2 MB 🦄
3 64 20 km 1:70 000 000 4 / 8 / 8 12 MB ✨
4 256 10 km 1:35 000 000 4 / 16 / 16 50 MB 🤘
5 1 024 5 km 1:15 000 000 Country 5 / 32 / 32 200 MB 👋
6 4 096 2.5 km 1:8 000 000 6 / 64 / 64 0.8 GB 🤙
7 16 384 1.2 km 1:4 000 000 7 / 128 / 128 3.2 GB 👌
8 65 536 600 m 1:2 000 000 8 / 256 / 256 12 GB ✌️
9 262 144 300 m 1:1 000 000 9 / 512 / 512 50 GB 👊
10 1 048 576 150 m 1:500 000 Megapoles 10 / 1024 / 1024 200 GB 💪
11 4 194 304 75 m 1:250 000 City 11 / 2 048 / 2 048 0.8 TB 🏋️
... ... ... ... ... ... ...
15 1 073 741 824 5 m 1:15 000 Road 15 / 32 768 / 32 768 200 TB 💣
16 4 294 967 296 2.5 m 1:8 000 16 / 65 536 / 65 536 0.8 PB 💥
17 17 179 869 184 1.2 m 1:4 000 Address 17 / 131 072 / 131 072 3.2 PB 🔥
18 68 719 476 736 60 cm 1:2 000 18 / 262 144 / 262 144 12 PB 😵
19 274 877 906 944 30 cm 1:1 000 Ped. crossing 19 / 524 288 / 524 288 50 PB 💀 
256 x 256 pixels

200 ko

Impossible to store high resolution maps

in a file storage…

you need an object stockage!

Proact

 Object stockage you said? 🧐

Proact

What's the advantage of object storage  🪣

Very large individual file size (> 1TB)

Unlimited number of files (even very small ones)

No limits on folder / files

 

Web natif (gateway)

Static storage (you don't need an active server)

Object key : {z}/{x}/{y}.fmt

 

Requests parallelism: high performance

Use as a CDN: replication / sync / high availability / region specific

Very low cost: ~100€/TB/yr (⚠️ egress-cost 💸)

Object storage providers

Can we do better?

💪

Cloud Optimized GeoTiff ☁️

/{z}/{x}/{y}.tiff 👉 /cog.tiff

Cloud Native Geo

Cloud Optimized GeoTiff ☁️

/{z}/{x}/{y}.tiff 👉 /cog.tiff +
🙋 HTTP range requests
$ curl -r 128 https://example.org/cog.tiff
Start : 256
Length : 256
Nb tiles : 1
Start : 1024
Length : 256
Nb tiles : 16

Cloud Native Geo

Cloud Optimized GeoTiff ☁️

/{z}/{x}/{y}.tiff 👉 /cog.tiff +
🙋 HTTP range requests
$ curl -r 256-512 https://example.org/cog.tiff
Start : 256
Length : 256

Cloud Native Geo

Cloud Optimized GeoTiff ☁️

/{z}/{x}/{y}.tiff 👉 /cog.tiff +
🙋 HTTP range requests
$ curl -r 1024-5120 https://example.org/cog.tiff
Start : 1024
Length : 16 * 256 = 4096

Cloud Native Geo

Cloud Optimized GeoTiff ☁️

/{z}/{x}/{y}.tiff 👉 /cog.tiff +
🙋 HTTP range requests
$ curl -r 1280-1536 https://example.org/cog.tiff
Start : 1024
Length : 1 * 256 = 256
Offset : 1 * 256 = 256

Cloud Native Geo

Cloud native database 📍

Cloud Native Geo
Carto

Cloud native database 📍

Duck DB
ATTACH 's3://example.org/stations.duckdb' AS stations_db;

SELECT count(*) AS num_stations FROM stations_db.stations;
num_stations
578
ID_PARCEL geometry
123563 POLYGON ((3.33896 49.84122, 3.33948 49.8...
5527076 POLYGON ((-1.44483 49.61280, -1.44467 49...
11479241 POLYGON ((2.87821 46.53674, 2.87820 46.5...
... ... 
SELECT PARCEL_ID, geometry FROM stations_db.stations
	WHERE ST_Contains(ST_Point(0, 45), geometry) LIMIT 100;

Cloud native database 📍

Conclusions

  • Creating global map requires a lot of storage (PB scale)
  • Distribution them to the user is possible though efficient tiling
  • Storage the tiles need to be hosted on object storage
  • New cloud native format are required to scale

Backup

data.obs-nantes.fr/wmts

Carte du ciel ✨

Observatoire Vera Rubin (LSST) - 2025

8.4 m miroir

3 200 Mpix / image

200 000 images / an

20 To / jour

~ 6 Po / an

STAC catalogue 📚

Standard Open Geospatial Consortium (OGC) pour décrire les jeux de données

stacspec.org

Modèle numérique de Terrain (MNT) 🏔️

/{z}/{x}/{y}.fmt