GeoTIFF vs Cloud-Optimized GeoTIFF
A Cloud Optimized GeoTIFF is a GeoTIFF organised for efficient partial access over HTTP; the difference is internal layout and serving behaviour, not a different raster data model.
Explain what makes a COG cloud-optimised, why range requests and overviews matter, and when an ordinary GeoTIFF is still sufficient.
GeoTIFF vs Cloud-Optimized GeoTIFF
A Cloud Optimized GeoTIFF (COG) is a GeoTIFF organised so software can efficiently read selected parts of the raster over HTTP without downloading the entire file.
The important relationship is:
every conformant COG is built on GeoTIFF/TIFF, but not every GeoTIFF is cloud-optimised.
The pixel values, CRS, and georeferencing can be conceptually identical; the difference lies in how the raster is internally arranged and served.
An ordinary GeoTIFF can require too much reading
Imagine a 20 GB satellite mosaic stored as one GeoTIFF. A user zooms into a small city that occupies less than 1% of the raster.
If the file layout is poorly suited to random access, the client may need to fetch large unrelated sections before it can read the required pixels. Remote analysis then becomes slow even though the file is a perfectly valid GeoTIFF.
A COG arranges data so the relevant spatial blocks and overview levels can be located and retrieved using HTTP byte-range requests.
Tiling provides spatially local blocks
TIFF can store imagery in internal tiles, so instead of organising the raster as very long strips, a tiled layout groups nearby pixels into rectangular blocks.
When a client needs one region, it can request the blocks intersecting that region rather than reading the whole image.
This is one of the building blocks of COG behaviour.
Overviews provide lower-resolution representations
A map zoomed out to continental scale does not need every native-resolution pixel of a 10-metre raster.
COGs include or use reduced-resolution subfiles—commonly called overviews—which allow software to read a coarser representation appropriate to the requested scale.
Without overviews, a client may need to access large amounts of high-resolution data only to downsample it for display.
HTTP range requests make partial access possible
The server must support byte-range access so the client can request selected parts of the file.
A COG-aware reader first reads enough metadata to understand where tiles and overviews are stored and then requests only the relevant byte ranges.
This allows one object in cloud storage to behave more like a remotely queryable raster source than a file that must be downloaded in full.
COG does not automatically make the raster smaller
Cloud optimisation is primarily about access pattern, not maximum compression.
A COG may even be slightly larger than a minimal GeoTIFF because internal tiling and overviews add structure and additional reduced-resolution imagery. The benefit is that users do not need all of those bytes for every request.
That distinction matters when evaluating storage costs versus read performance.
When an ordinary GeoTIFF is sufficient
You may not need COG when:
the raster is small;
it is always used locally;
workflows download the whole file intentionally;
software does not support COG-aware partial reading;
the file is archival and remote random access is not part of the requirement.
An ordinary well-formed GeoTIFF is still a standard geospatial raster.
When COG is a strong default
COG is valuable when large rasters are stored on HTTP-accessible object storage and users repeatedly read subsets, windows, or lower-resolution views.
Common cases include:
satellite imagery catalogues;
elevation models;
climate and environmental rasters;
web visualisation of large imagery;
cloud analysis that reads spatial windows rather than complete files.
It works especially well when paired with catalogues and software that know how to exploit range requests.
Validate the file and the server
A file can have a COG-like internal layout yet still be inefficient remotely if the server does not support byte ranges correctly. Conversely, HTTP range support cannot rescue a badly organised TIFF whose required blocks are scattered in a way that forces excessive requests.
The OGC COG standard therefore covers both file characteristics and HTTP range behaviour.
References
OGC Cloud Optimized GeoTIFF Standard. Describes COG's use of TIFF tiling, reduced-resolution subfiles, GeoTIFF keys, and HTTP range access.
OGC COG 1.0. Approved standard specifying file organisation and HTTP requirements.
OGC GeoTIFF Standard. Underlying georeferencing standard used by COG.
Related content
What Is GeoTIFF? — raster-format foundation
What Is PMTiles? — single-file random access for map tiles