COG (Cloud Optimized GeoTIFF)
Aug 27,2026

GISBox is a one-stop 3D GIS data editing, conversion and publishing platform that supports editing in multiple GIS formats such as OSGB/GEOTIFF/RVT, converting to 3DTiles/Terrain and publishing.

Introduction

COG (Cloud Optimized GeoTIFF) is an official OGC standard officially released in July 2023. It is optimized based on the traditional GeoTIFF format. Through a special file organization method featuring built-in tiles and multi-scale overviews, it enables clients to use HTTP range requests to on-demand retrieve the required local image content without downloading the entire large raster file. It is fully compatible with conventional GIS software while significantly improving the online streaming loading, visualization, and processing efficiency of geographic raster data such as remote sensing imagery and DEMs in cloud environments. It has now been widely adopted in geoscience data distribution scenarios by organizations such as NASA.

File Structure

COG is a cloud-optimized format that is fully compliant with the standard GeoTIFF specification. Its file structure is specially designed to accommodate HTTP range requests. The core components can be summarized as follows:

Front-loaded, centralized metadata area: All critical metadata (including the TIFF header of the full-resolution image, the Image File Directory (IFD), band information, georeferencing tags, tile offset tables, etc.) are centrally placed at the very beginning of the file, typically occupying only the first 16 KB to 50 KB. Clients only need to make one small HTTP request to retrieve all metadata, eliminating the need for multiple seek operations to locate scattered directory information as required in traditional GeoTIFFs.
Ordered multi-scale overview area: Within the file, a complete image pyramid (multi-scale overviews) is embedded in order from low to high resolution. The lowest-resolution thumbnail is placed at the very front, followed by progressively higher-resolution overview layers. All overviews are stored as independent small images within the same file, eliminating the need for additional auxiliary files such as .ovr. This directly supports fast preview loading at different zoom levels.
Tiled full-resolution data area: The full-resolution original image is uniformly divided into fixed-size regular tiles. All tiles are arranged contiguously and in order at the end of the file. The position of each tile can be directly located using the offset table in the front metadata, allowing clients to precisely issue HTTP range requests to download only the local tiles required for the current view. This avoids downloading entire multi-gigabyte image files and significantly reduces network transmission overhead.
Standardized and unified structural constraints: COG strictly adheres to the TIFF (ISO 12234-2) and GeoTIFF standards without any custom private format fields. All software that supports standard GeoTIFF reading can directly open it. By optimizing only the internal data arrangement, it achieves efficient access in cloud environments, balancing format universality with cloud-native performance.

Pros

Extremely high cloud access efficiency: Through the ordered arrangement of built-in tiles and multi-scale overviews, it perfectly adapts to the HTTP range request mechanism. Clients only pull the local data within the current viewport on demand. In practice, it reduces data transfer by 70% to 92% compared to traditional WMS services. Even for 4 GB remote sensing imagery, response times can be kept within 150 ms, providing smooth panning and zooming experiences.
Very low deployment barriers: No reliance on heavy-weight GIS middleware such as GeoServer or ArcGIS Server is required. Plain static file servers like Nginx and Apache, or even object storage services, can directly publish imagery services. A 2-core cloud instance can support concurrent access to province-scale large imagery, greatly reducing server hardware resource consumption.
Controllable storage overhead: Compared to traditional solutions where pre-generated tiles cause storage space to expand several times over, COG increases storage overhead by only about 15%. A 1.2 TB original image converted to COG rarely experiences a doubling in file size, significantly saving cloud storage costs.
Strong ecosystem compatibility: It remains essentially a standard GeoTIFF file, fully compatible with traditional desktop GIS software such as QGIS and ArcGIS, and also complies with official OGC standards. It is the recommended raster format for the STAC (SpatioTemporal Asset Catalog) and can be directly integrated into existing GIS workflows without additional adaptation.
Suitable for lightweight scenarios: The file is overall lightweight and does not require additional auxiliary tile packages or overview files. A single file carries complete geographic information and multi-scale data, making it highly suitable for lightweight scenarios with limited resources, such as mobile GIS applications and online map services.

Cons

Limited front-end rendering support: Currently, among mainstream web mapping APIs, only a few libraries such as OpenLayers natively support direct COG rendering. Most traditional WebGIS frameworks have not yet natively adapted it. Some Canvas-based layer post-processing functions may encounter compatibility issues when rendering COG with WebGL.
Lack of partial editing capability: The tiles and overviews in COG are fixed in order according to the specification. Unlike ordinary GeoTIFFs, local pixel content cannot be directly modified. Once local image data needs to be updated, the entire COG file must be regenerated, making it unsuitable for incremental editing scenarios.
Certain barriers in the generation process: Simply changing the file extension does not produce a compliant COG. Parameters such as tile size, compression method, and overview levels must be strictly configured according to the specification. Improper parameter settings may produce non-standard pseudo-COGs, leading to significant performance degradation or even read failures during remote access.
Additional projection adaptation requirements: If the coordinate reference system of the COG does not match the target display platform (e.g., Google Earth), rendering artifacts or display anomalies can easily occur. Smaller COG files are particularly sensitive to projection conversion, requiring additional reprojection and projection validation steps.
Strong dependence on network conditions: In high-latency or weak network environments, frequent small-range HTTP requests may lead to request stacking and loading stuttering. Compared to the mode of reading the entire file locally, network fluctuations have a more noticeable impact on the access experience.

Application Scenario

As a cloud-optimized GeoTIFF format, COG is core to online map imagery service publishing. It enables smooth loading of large-volume remote sensing imagery without heavy-weight GIS middleware. It also serves as an ideal exchange format for cross-team sharing and collaboration on geographic data, allowing different users to directly read cloud-based data online without downloading entire large files. It further supports lightweight loading requirements for mobile GIS applications and can be deeply integrated into cloud GIS platforms to support various remote sensing analysis tasks such as ecological monitoring and natural resource surveys. It is also a mainstream choice for large-scale public distribution of Earth observation datasets by geoscience institutions like NASA, significantly lowering the barriers for researchers and industry users in acquiring and processing massive raster data.