COG vs Zarr Explained — And Where MrSID Fits In

Anvita Shrivastava
Apr 21
3 min read

Updated: Apr 26

As the amount of geospatial data continues to increase significantly and cloud-based workflows have become commonplace, selecting the correct format for geospatial data has grown increasingly important. Two formats that are frequently discussed in relation to modern digital geospatial/scientific data systems are Cloud Optimized GeoTIFFs ('COG') and Zarr. While both formats are intended to promote effective access and usage of very large datasets, each is designed for different types of datasets and different types of infrastructure.

What are Cloud Optimized GeoTIFF's?

Cloud Optimized GeoTIFFs (also known as COG) are GeoTIFFs that have been structured in a manner to allow for effective HTTP range requests on files stored in cloud environments. This enables clients to access only the portion(s) of the file they require, without downloading the entire file.

Key characteristics:

Tiled internal structure to allow for partial read-access
Embedded overview (pyramid) data in a file, designed for multi-resolution access
HTTP range request compatible with cloud storage (e.g., AWS S3, Azure Blob)
Backward-compatible with existing GeoTIFF tools

Strengths:

Ideal for raster data types (e.g., satellite, Aerial imagery)
Works with existing GIS software (e.g., QGIS, GDAL, web mapping servers) without additional effort to integrate
Easily deployed and shared via cloud storage.

Limitations:

Optimized for 2D raster data only
Less versatile for working with multi-dimensional datasets (e.g., time series, climate datasets)

What is Zarr?

Zarr is a file format for storing large amounts of multidimensional data in the form of chunks, which can be compressed.

Zarr is designed to work with distributed systems, including cloud-based systems.

Key Characteristics:

Stores data as chunks across a variety of files and/or objects.
Stores N-dimensional arrays of values, where N represents the number of dimensions/axes.
Allows for parallel accessing and processing of data.
Compatible with popular data science and analysis libraries for the Python programming language, such as xarray and Dask.

Strengths:

Good for storing critical, time-based ecological data (for example, data generated from weather models or satellite imagery).
Extremely scalable for use in cloud-based and high-performance computing (HPC) environments.
Allows users to access and process multiple datasets at the same time.

Limitations:

Not designed for use in GIS applications by default.
Specialized tools are needed for visualizing and analyzing data.
Not as easy to use as traditional raster-based workflows.

https://www.youtube.com/watch?v=M1bnBrqsyuM

COG vs Zarr — And Where MrSID Fits In

COG vs Zarr: A Side-by-Side Comparison

Feature	COG	Zarr
Data Type	2D raster	N-dimensional arrays
Storage	Single file	Directory/object store
Access Pattern	HTTP range requests	Chunk-based parallel access
Cloud Optimization	Yes	Yes
GIS Compatibility	High	Limited
Ideal Use Case	Satellite/aerial imagery	Scientific & temporal datasets

Where Does MrSID Fit In?

MrSID (Multi-resolution Seamless Image Database) is a very efficient image compression format specifically designed for geospatial images. Even though MrSID has been around longer than cloud-native image formats (like COG), it is still an important option today, especially when the density of storage space is very important, and/or where high-performance visualization is essential to your business operations.

Advantages of Using MrSID

Superior Image Compression

The wavelet compression technology that MrSID uses can produce files that are significantly smaller than their original (uncompressed) sizes and keep the same visual quality as their originals, often exceeding the quality of compressed Geotiff files.

Multi-Resolution Access

Like COG, MrSID supports the use of pyramids to provide users with the ability to navigate around large images quickly using zoom and pan methods to display image data to the end user without needing to download the entire (full resolution) image to their workstation.

Fast Visualization of Images

MrSID images are optimized for fast rendering when viewing the final image as:

Web mapping applications
Desktop GIS (Geographical Information System) products
Military and intelligence products

Proven Long-Term Exposure

MrSID is used by many organizations throughout multiple industries; therefore, it is well embraced and supported by most of the prominent geospatial software applications and will continue to be utilized as the industry grows.

COG vs MrSID: Complementary, Not Competitive

While COG is gaining traction as a cloud-native standard, MrSID still offers distinct benefits:

Feature	COG	MrSID
Compression	Moderate	High
File Size	Larger	Smaller
Cloud Native	Yes	Not inherently
Visualization Speed	Good	Excellent

Geospatial data formats do not have just one answer or way to do things. COGs, ZARRs, and MrSID each serve different purposes, but they may also serve similar functions. Here are the three main geospatial format types: