What is a GeoTIFF?

A GeoTIFF is a TIFF image file that carries georeferencing metadata in standardized tags, so each pixel maps to a coordinate on Earth. The tags record the coordinate reference system, the pixel size, and the position of the grid origin. A Cloud Optimized GeoTIFF (COG) arranges the same file so clients can read parts of it over HTTP.

What a GeoTIFF is

GeoTIFF is the dominant file format for raster geospatial data. It extends the Tagged Image File Format, an image standard from 1986, with tags that describe where the image sits on the Earth.

The format began in the 1990s as a community specification and became OGC GeoTIFF Standard 1.1 in 2019, which fixed the tag definitions and their meaning.

The defining property is that a GeoTIFF is still a valid TIFF. Any TIFF reader opens the pixels; a geospatial reader also reads the tags and places them.

Three tags do the work. ModelPixelScaleTag gives the ground size of a pixel, and ModelTiepointTag ties a pixel to a ground coordinate. GeoKeyDirectoryTag encodes the coordinate reference system, usually as an EPSG code.

A GeoTIFF can hold one band or many, in integer or floating point, with a NoData value, compression, and internal tiling. BigTIFF lifts the 4 GB size limit of classic TIFF.

What a GeoTIFF does

GeoTIFF carries almost every raster dataset in Earth observation from producer to analyst.

Satellite agencies deliver Landsat and Sentinel-2 scenes as GeoTIFF, one file per band. Elevation programs deliver DEM tiles as single-band GeoTIFF.

Weather and climate teams export model output as GeoTIFF for GIS users who cannot read GRIB or NetCDF.

Analysts write derived products such as NDVI, classified land cover, and flood depth back to GeoTIFF.

Web map servers read COGs directly from object storage and serve tiles from them.

Machine learning pipelines read training chips and write prediction rasters as GeoTIFF, since every raster library reads it.

The format’s job is interchange. A GeoTIFF written by one tool opens in every other, with its georeferencing intact.

How a GeoTIFF works

A GeoTIFF is a TIFF with three layers of structure: the image data, the TIFF tags, and the geo tags.

Image data is stored as strips or tiles. Strips are rows of pixels written in sequence; tiles are square blocks, typically 256 or 512 pixels, which allow a reader to fetch one region without reading the rows above it.

Compression applies per strip or tile. LZW and Deflate are lossless and universal, while ZSTD and LERC are faster or smaller but need newer readers.

Overviews are downsampled copies of the image stored inside the file at 2x, 4x, 8x and so on. A viewer at continental zoom reads the smallest overview instead of billions of full-resolution pixels.

Feature	Plain GeoTIFF	Cloud Optimized GeoTIFF
Layout	Any	Tiled, with overviews, header at the front, tiles in order
Read pattern	Download whole file	HTTP range requests for header, then needed tiles
Standard	OGC GeoTIFF 1.1	OGC COG Standard (2023), built on GeoTIFF 1.1
Validity	Valid TIFF	Also a valid GeoTIFF; any reader opens it

A Cloud Optimized GeoTIFF is a GeoTIFF whose internal layout follows rules. The header and tile index sit at the front, the image is internally tiled, and overviews are present, so a client reads the header with one request and then fetches only the tiles for its view.

A 50 GB COG in object storage serves a 512 by 512 window in a few range requests, with no download and no server-side process.

GeoTIFF in Wherobots

WherobotsDB reads GeoTIFF as a native raster type and writes it back, including as Cloud Optimized GeoTIFF.

RS_FromGeoTiff loads a file’s bytes into an in-db raster stored in the table. RS_FromPath creates an out-db raster, a reference to the GeoTIFF in S3, and reads tiles only when a function touches them. For a COG, that means range requests for the tiles a query needs and nothing else.

Havasu, the Apache Iceberg table format extended with a native raster type, stores millions of GeoTIFF references as rows with metadata columns, and pushes spatial filters into the scan so a query for one county reads only intersecting files.

RS_ functions then operate directly. RS_MapAlgebra, RS_Clip, RS_Resample, RS_Tile, and RS_ZonalStats each run in parallel across every raster in the table.

Writers close the loop. RS_AsGeoTiff serializes a raster to GeoTIFF bytes, and RS_AsCOG writes a Cloud Optimized GeoTIFF with tiling and overviews. Results land in the customer’s S3 bucket ready for any other tool.

The STAC reader registers every GeoTIFF asset in a catalog as an out-db raster in one step.

Because raster and vector data share one engine, a GeoTIFF joins to parcels or buildings in the query that reads it. WherobotsDB is built by the original creators of Apache Sedona, 100% code compatible across all spatial functions.

-- Register COGs in S3 as out-db rasters, clip to a boundary, write the result back as COG
WITH scenes AS (
  SELECT scene_id, RS_FromPath(path) AS rast FROM eo.scene_index
),
aoi AS (
  SELECT geometry FROM boundaries WHERE name = 'Study Area'
)
SELECT s.scene_id,
       RS_AsCOG(RS_Clip(s.rast, 1, aoi.geometry)) AS cog_bytes
FROM scenes s CROSS JOIN aoi
WHERE RS_Intersects(s.rast, aoi.geometry);

Related terms

• Raster vs vector data: why GeoTIFF is a raster container
• Digital elevation model: the most common single-band GeoTIFF
• Sentinel-2: distributed as one GeoTIFF per band
• Earth observation data: the category GeoTIFF carries most of

Read more from Wherobots

• Measuring the Strait of Hormuz shutdown with Sentinel-2, WherobotsDB and Overture Maps
• Out-DB rasters performance guide
• Raster loaders
• Raster writers

Query and write Cloud Optimized GeoTIFFs in SQL with WherobotsDB on the free tier at cloud.wherobots.com.