Quick Insight
Raster tile data is a set of pre-rendered, fixed-size raster images, usually 256 x 256 or 512 x 512 pixels, that together cover a map extent at multiple zoom levels. Tiles are typically stored as PNG, JPEG or WebP (or GeoTIFF for analytic use) and organized as a pyramid. In the common XYZ scheme, zoom level z divides the world into 2^z by 2^z tiles, and each tile is addressed by its z/x/y index, usually in Web Mercator. Clients such as Leaflet or OpenLayers request tiles directly or through the OGC WMTS standard and stitch them together. Each tile holds only pixel values, with no feature geometry or queryable attributes, so tiles behave as static files that cache well on a CDN and ask little of the client’s hardware.
Why Raster Tile Data Matters
Tiling lets a map load only the pixels in view instead of one enormous image. That keeps pan and zoom fast on slow connections and modest devices, and it scales to millions of users because the server just hands out files. Styling is baked in, so every viewer sees the same finished look.
The honest limit: a raster tile is a picture, not data. Zoom past its native resolution and it pixelates. You can’t restyle it, click a feature or filter it without re-rendering the whole tile set, and storing every zoom level costs more storage and bandwidth than vector tiles.
How Raster Tiles Are Built
The source raster, such as satellite imagery, a rendered basemap or a heat map, is reprojected into the tiling scheme, then cut into tiles at each zoom level. Every step deeper quadruples the tile count while doubling the detail. When a user pans, the viewer works out which z/x/y tiles fall inside the window, fetches them and places them side by side. SafeGraph’s geospatial data types guide explains where raster sits next to other spatial formats.
Types of Raster Tiles
Pre-rendered (static) tiles are generated ahead of time and served as plain files. They are fast and cheap to host, but they only change when you rebuild them. Dynamic tiles are rendered by a tile server on request from the source data, which allows custom styling and fresher results at the cost of server compute.
Tiling Inside a Database
Raster databases apply the same idea internally, storing a large raster as many tile records, often as binary objects. Tile size is a trade-off: small tiles read quickly for narrow queries but multiply row counts and overhead, while large tiles cut row counts but pull in more data than a query needs. Some systems default to about 128 x 128 pixels. Compression such as LZ77-based schemes, JPEG or JPEG 2000 reduces storage, with lossy options trading some fidelity for size.
Common Uses of Raster Tile Data
- Web basemaps: street maps and satellite views in apps and dashboards.
- Imagery and environmental overlays: weather, elevation and vegetation layers. SafeGraph’s geospatial data ecosystem post shows where imagery and environment layers fit among other location datasets.
- Heat maps and density surfaces: rendered once, then served quickly. See SafeGraph’s guide to visualizing geospatial data for map types that suit this approach.
- Display layers beneath analysis: tiles supply the picture, while precise vector layers such as building footprints and polygon boundaries from SafeGraph Geometry supply geometry you can measure, join and query.
To compare raster and vector data, read SafeGraph’s Geospatial Data 101 and the broader geospatial data guide.
Frequently Asked Questions
1. Should I choose 256 or 512 pixel tiles?
256 is the long-standing default and works everywhere. 512 means fewer requests to fill a screen and sharper results on high-density displays, but each file is larger. If your users are mostly on phones or retina screens, 512 is usually worth testing.
2. Can I overlay my own data on top of a basemap as raster tiles?
Yes. Render your layer into tiles with a tool such as GDAL’s tile generator or a tile server, then stack it above the basemap. Use PNG or WebP rather than JPEG, because those formats support transparency, so the basemap shows through.
3. How much storage will a tile pyramid need?
It grows fast, because each added zoom level has roughly four times as many tiles as the one before. The deepest levels dominate the total. To keep it manageable, cap the maximum zoom and generate only the area you need rather than the whole world.
4. What happens when my underlying data changes?
Pre-rendered tiles are a snapshot, so you rebuild the affected tiles and refresh your CDN cache, ideally with versioned URLs. If the data changes often, dynamic tiles are a better fit.
5. Can raster tiles work offline?
Yes. Tiles can be packaged into a single file, commonly an MBTiles database (a SQLite container), and bundled into a mobile or field app so maps still load without a connection.
